RTO of NATO. Chazot O., Rini P. (Eds) - Experiment, Modeling and Simulation of Gas-Surface Interactions For Reactive Flows in Hypersonic Flights

NORTH ATLANTIC TREATY RESEARCH AND TECHNOLOGY
ORGANISATION ORGANISATION
AC/323(AVT-142)TP/111 www.rto.nato.int
RTO EDUCATIONAL NOTES EN-AVT-142
Experiment, Modeling and Simulation of

Gas-Surface Interactions for
Reactive Flows in Hypersonic Flights
(L’expérimentation, la modélisation et la simulation
d’interactions gaz-surface pour
écoulements réactifs en vol hypersonique)
Papers presented during the AVT-142 RTO AVT/VKI Lecture Series

held at the von Karman Institute, Rhode St. Genèse,
Belgium, 6-10 February 2006.
Published July 2007

The Research and Technology
Organisation (RTO) of NATO
RTO is the single focus in NATO for Defence Research and Technology activities. Its mission is to conduct and promote
co-operative research and information exchange. The objective is to support the development and effective use of
national defence research and technology and to meet the military needs of the Alliance, to maintain a technological
lead, and to provide advice to NATO and national decision makers. The RTO performs its mission with the support of an
extensive network of national experts. It also ensures effective co-ordination with other NATO bodies involved in R&T
activities.
RTO reports both to the Military Committee of NATO and to the Conference of National Armament Directors.
It comprises a Research and Technology Board (RTB) as the highest level of national representation and the Research
and Technology Agency (RTA), a dedicated staff with its headquarters in Neuilly, near Paris, France. In order to
facilitate contacts with the military users and other NATO activities, a small part of the RTA staff is located in NATO
Headquarters in Brussels. The Brussels staff also co-ordinates RTO’s co-operation with nations in Middle and Eastern
Europe, to which RTO attaches particular importance especially as working together in the field of research is one of the
more promising areas of co-operation.
The total spectrum of R&T activities is covered by the following 7 bodies:
• AVT Applied Vehicle Technology Panel
• HFM Human Factors and Medicine Panel
• IST Information Systems Technology Panel
• NMSG NATO Modelling and Simulation Group
• SAS System Analysis and Studies Panel
• SCI Systems Concepts and Integration Panel
• SET Sensors and Electronics Technology Panel
These bodies are made up of national representatives as well as generally recognised ‘world class’ scientists. They also
provide a communication link to military users and other NATO bodies. RTO’s scientific and technological work is
carried out by Technical Teams, created for specific activities and with a specific duration. Such Technical Teams can
organise workshops, symposia, field trials, lecture series and training courses. An important function of these Technical
Teams is to ensure the continuity of the expert networks.
RTO builds upon earlier co-operation in defence research and technology as set-up under the Advisory Group for
Aerospace Research and Development (AGARD) and the Defence Research Group (DRG). AGARD and the DRG share
common roots in that they were both established at the initiative of Dr Theodore von Kármán, a leading aerospace
scientist, who early on recognised the importance of scientific support for the Allied Armed Forces. RTO is capitalising
on these common roots in order to provide the Alliance and the NATO nations with a strong scientific and technological
basis that will guarantee a solid base for the future.
The content of this publication has been reproduced

directly from material supplied by RTO or the authors.
Published July 2007
Copyright © RTO/NATO 2007

All Rights Reserved
ISBN 978-92-837-0057-9
Single copies of this publication or of a part of it may be made for individual use only. The approval of the RTA
Information Management Systems Branch is required for more than one copy to be made or an extract included in
another publication.
RTO-EN-AVT-142
Experiment, Modeling and Simulation of
Gas-Surface Interactions for
Reactive Flows in Hypersonic Flights
(RTO-EN-AVT-142)
Executive Summary
The need for a closer interaction between materials and fluid sciences as a key point of the future progress
in space-entry research, rockets and high speed missiles engineering is well known. Indeed, under typical
flight conditions the high energetic exchanges cause sharp temperature increases followed by molecule
dissociation and particle ionizations governed by chemical reactions. Several empirical models are known
for the description of such chemical processes and their differences turn out to affect sensibly the flow
behavior mainly through the species composition. Another source of discrepancies arises when computing
the interaction of such reactive flow with particles or solid structures using specific models, including
catalytic properties, which affect the prediction of the total energy flux experienced by the surface.
In order to increase the efficiency of the technologies currently used to sustain these flow conditions it
requires on the R & D side a better understanding and control of the complex physico-chemical processes
occurring in the external flow fields, wakes and propulsion systems, of such configurations. Therefore,
a systematic review of the different aspects of this field appears essential in the development strategies of
new technologies, dealing with aerothermochemical phenomena, as well as their military applications and
for the more general space transportation domain.
The objectives of the collection of papers are to review the up-to-date experimental techniques,
the theoretical models, as well as the numerical simulation developments involved in the treatment of the
chemical characters of high temperatures gases. Aspects of both materials and high temperatures fluid
sciences will be discussed together, with the aim of enhancing the improvements in the understanding of
the processes of heat release on solid surfaces close to reactive gases taking place in re-usable hypersonic
flight vehicles. The course notes will then focus on the presentation of the experimental techniques used to
determine chemical non-equilibrium reaction rates, on the modeling and on the experimental investigation
of gas surface interactions; and, finally, on the application of the previous concepts to both numerical and
experimental investigation of flight and ground-test conditions.
We wish to thank the following for their contribution to the success of this RTO-VKI Lecture Series:
European Office of Aerospace Research and Development, Air Force Office of Scientific Research,
United States Air Force Research Laboratory EOARD.
RTO-EN-AVT-142 ES - 1
L’expérimentation, la modélisation et la simulation
d’interactions gaz-surface pour
écoulements réactifs en vol hypersonique
(RTO-EN-AVT-142)
Synthèse
Le besoin d’une interaction plus étroite entre les sciences des matériaux et des fluides comme point clé du
progrès futur dans l’ingénierie de recherche en matière d’entrée dans l’espace, de fusées et de missiles à
grande vitesse, est bien connu. En effet, dans des conditions de vol particulières, les échanges hautement
énergétiques provoquent des augmentations brutales de température suivies d’une dissociation moléculaire
et d’une ionisation des particules régies par des réactions chimiques. Plusieurs modèles empiriques sont
connus pour décrire ces processus chimiques et leurs différences s’avèrent affecter sensiblement le
comportement des flux, principalement par la composition des espèces. Une autre source de divergences
apparaît lors du calcul de l’interaction de ce type d’écoulement réactif avec des particules ou des structures
solides utilisant des modèles spécifiques, notamment des propriétés catalytiques, qui affectent la
prédiction du flux d’énergie total subi par la surface. Augmenter l’efficacité des technologies utilisées à
l’heure actuelle pour supporter ces conditions de flux nécessite que la R & D comprenne et contrôle mieux
les processus physico-chimiques survenant dans les champs de courants externes, les sillages et les
systèmes de propulsion, de ces configurations. Par conséquent, une revue systématique des différents
aspects de ce champ semble essentielle dans le cadre des stratégies de développement de nouvelles
technologies, traitant des phénomènes aérothermochimiques, et leurs applications militaires ainsi que pour
le domaine du transport spatial de manière plus générale.
L’objectif de ce recueil d’articles consiste à revoir les techniques expérimentales de pointe, les modèles
théoriques ainsi que les développements de simulations numériques impliqués dans le traitement des
caractéristiques chimiques des gaz à température élevée. Les aspects de ces deux sciences des matériaux et
des fluides à haute température feront l’objet de discussions communes dont l’objectif sera d’améliorer
davantage la compréhension des processus de dégagement de chaleur sur des surfaces solides à proximité
de gaz réactifs, se produisant dans des véhicules aériens hypersoniques réutilisables. Le cours portera
ensuite sur la présentation des techniques expérimentales utilisées pour déterminer des vitesses de réaction
chimiques hors d’équilibre, sur la modélisation et la recherche expérimentale d’interactions gaz-surface,
et enfin sur l’application des anciens concepts de recherche numérique et expérimentale des conditions
d’essai en vol et au sol.
Nous souhaitons remercier les organismes suivants pour leur contribution au succès de cette série de
conférences RTO-VKI : l’EOARD (le bureau européen de recherche et développement aérospatial de
l’armée de l’air américaine), l’AFOSR (le bureau de recherche scientifique de l’armée de l’air
américaine), l’AFRL (le laboratoire de recherche de l’armée de l’air américaine).
ES - 2 RTO-EN-AVT-142
List of Lecturers
Lecture Series Directors
Mr. O. Chazot and Mr. P. Rini

von Karman Institute
Chaussée de Waterloo 72
1640 Rhodes St. Genèse
Belgium
Email: vdb@vki.ac.be
Members
Belgium Italy
Mr. J.M. Muylaert Prof. C. Bruno
ESA/ESTEC Università degli Studi di Roma
Aerothermodynamic Section La Sapienza
Postbus 299 Dipartimento Meccanica Aeronautica
2200 AG Noordwijk Via Eudosssiana, 18
Email: Jean-Marie-Muylaert@esa.int 00148 Roma
Email: cbruno@dma.ing.uniroma1.it
France
Prof. S. Candel Russian Federation
EM2C Lab. Ecole Centrale Paris Dr. A.F. Kolesnikov
Grande Voie des Vignes Institute for Problems in Mechanics
92295 Chatenay Malabry IPM-Russian Academy of Sciences
Email: candel@em2c.ecp.fr Prospect Vernadskogo, 101-1
117526 Moscow
Germany
Email: koles@ipmnet.ru
Mr. G.S.R. Sarma
DLR Institute for Fluid Mechanics United States
SM-SK Mr. G. Candler
Bunsenstrasse 10 University of Minnesota
37073 Goettingen Aerospace Engineering and Mechanics
Email: sarma@ts.go.dlr.de 110 Union Street
Minneapolis, MN 55455
Email: candler@aem.unm.edu
Lecturers
France
Dr. M. Balat-Pichelin Mrs. A. Bourdon
Laboratoire Procédés Ecole Central de Paris
Materiaux et Energie Solaire Laboratoire EM2C UPR 288 CNRS
PROMES-SNRS, UPR 8521 Grande voie des vignes
Rue du four solaire 92295 Chatenay-Malabry Cedex
66120 Font-Romeu Odeillo Email: anne.bourdon@em2c.ecp.fr
Email: balat@promes.cnrs.fr
RTO-EN-AVT-142 PC - 1
France (cont’d) Russian Federation
Mr. A. Bultel Mr. A. Kolesnikov
Ecole Central de Paris Institute for Problems in Mechanics
Université de Rouen IPM – Russian Academy of Sciences
CORIA/CNRS UMR 6614 Prospect Vernadskogo, 101-1
Place Emile Blondel 117526 Moscow
76821 Mont Saint Aignan Cedex Email: koles@ipmnet.ru
Email: arnaud.bultel@coria.fr
Mr. V. Kovalev
Italy Moscow State University
Dr. P. Barbante Leninskije Gory
Pol. di Milano 119992 Moscow
Dept. of Mathematics Email: kovalev@mech.mth.msu.su
Piazza Leonardo da Vinci 32 Mrs. A. Sergievskaya
20133 Milano Moscow State University
Email: barbante@mate.polimi.it M.V. Lomonosov Moscow State University
Dr. M.A. Cacciatore 1 Michurinsky prospect.
Institute of Inorganic Methodologies and Moscow, 119192
Plasmas Email: sergievska@imec.msu.ru
CNR-IMIP South Korea
c/o Dipartimento di Chimica
Dr. C. Park
Università di Bari
Korea Advanced Institute of Science and
V. orabona, 4
Technology
70126 Bari
Dept. of Aerospace Engineering
Email: mario.cacciatore@ba.imip.cnr.it
373-1 Guseong-dong, Yuseong-gu
Prof. A. Lagana Daejeon, 305-701
University of Perugia
United States
Dipartimento di Chimica
06123 Perugia Mr. J. Marschall
Email: lagana05@gmail.com SRI International
333 Ravenswood Avenue
Germany Menlo Park, CA 94025
Prof. Dr. A. Gross Email: jochen.marschall@sri.com
Abteilung Theortetische Chemie
Dr. M.J. Wright
Universitat Ulm
NASA Ames Research Center
Albert-Einstein-Allee 11
MS 230-2 Moffet Field
D-89069 Ulm
CA 94035
Email: axel.gross@uni-ulm.de
Email: mjwright@mail.arc.nasa.gov
Dr. K. Reuter
Fritz-Haber-Institut der Max Planck-
Gesellschaft
Faradayweg 4-6
D-14195 Berlin
Email: reuter@fhi-berlin.mpg.de
PC - 2 RTO-EN-AVT-142
Contents
Cover Pages.
Executive Summary and Synthèse.
Programme Committee.
1. Not released for publishing.
2. Ab Initio Atomistic Thermodynamics for Surfaces: A Primer - (Rogal, J.; Reuter, K.).
3. Introduction to Theoretical Surface Science - (Groß, A.).
4. Simulation of Gas-Surface Dynamical Interactions - (Groß, A.).
5. Molecular Dynamics Simulations of Surface Processes: Oxygen Recombination on Silica
Surfaces at High Temperature - (Cacciatore, M.; Rutigliano, M.).
9. Detailed and Simplified Kinetic Schemes for High Enthalpy Air Flows and their Influence on
Catalycity Studies - (Bourdon, A.; Bultel, A.).
10. Theory and Computing of Gas Phase Chemical Reactions: From Exact Quantum to
Approximate Dynamical Treatments. Theoretical Estimates of Reaction Observables vis-à-vis
Modern Experiments - (Laganà, A.).
11. Laboratory Determination of Thermal Protection System Materials Surface Catalytic Properties
- (Marschall, J.).
12. Interaction of Reactive Gas Flows and Ceramics at High Temperature – Experimental Methods
for the Measurement of Species Recombination during Planetary Entry - (Balat-Pichelin, M.).
13. Experimental Studies on Hypersonic Stagnation Point Chemical Environment - (Chazot, O.).
14. Reacting Flows Simulation with Applications to Ground to Flight Extrapolation - (Barbante,
P.F.).
15. Determination of Effective Recombination Probability: Detailed Aspects of a Macroscopic
Methodology - (Rini, P.; Degrez, G.).
16. Numerical Implementation of Surface Catalysis, Reaction, and Sublimation - (Park, C.).
17. A Risk-Based Approach for Aerothermal/TPS Analysis and Testing - (Wright, M.J.; Grinstead,
J.H.; Bose, D.).
Ab Initio Atomistic Thermodynamics for Surfaces: A Primer
Jutta Rogal and Karsten Reuter

Fritz-Haber-Institut der Max-Planck-Gesellschaft
Faradayweg 4-6
D-14195 Berlin
Germany
reuter@fhi-berlin.mpg.de
1.0 INTRODUCTION
Rational design and advancement in materials science will ultimately rely on an atomic-scale
understanding of the targeted functionality. Corresponding modeling must then address the behavior of
electrons and the resulting interactions (often expressed in the terminology of chemical bonds) that govern
the elementary processes among the atoms and molecules in the system. Modern electronic structure
theory methods like density-functional theory (DFT) [1-5] have matured to a standard tool for this task,
allowing a description that is often already accurate enough to allow for a modeling with predictive
character. These techniques are referred to as first-principles (or in latin: ab initio) to indicate that they do
not rely on empirical or fitted parameters, which then makes them applicable for a wide range of realistic
conditions, e.g. realistic environmental situations of varying temperatures and pressures [6].
The latter type of application seems at first sight at variance with the frequent argument describing DFT as
a zero-temperature, zero-pressure technique. Such a confusion arises, when thinking that DFT provides
(apart from a wealth of information about the electronic structure) “only” the total energy of the system.
Instead, it is crucial to realize that this kind of energetic information can be obtained as a function of the
atomic configuration {RI}. This leads to the so-called potential energy surface (PES) E({RI}), which then
contains the relevant information needed to describe the effect of temperature on the atomic positions.
Obviously, a (meta)stable atomic configuration corresponds to a (local) minimum of the PES. The forces
acting on the given atomic configuration are just the local gradient of the PES, and the vibrational modes
of a (local) minimum are given by the local PES curvature around it.
One possibility to go from this to situations of finite temperature and finite pressure is to achieve a
matching with thermodynamics. This is the general idea behind ab initio atomistic thermodynamics,
namely to employ the information on the first-principles PES to calculate appropriate thermodynamic
potential functions like the Gibbs free energy [7-10]. Once such a quantity is known, one is immediately in
a position to evaluate macroscopic system properties using the standard methodology of thermodynamics.
Apart from bridging to any (T,p)-conditions, this methodology is particularly useful for larger systems,
which may readily be divided into smaller subsystems that are mutually in equilibrium with each other.
Each of the smaller and thus potentially simpler subsystems can then be first treated separately, and the
contact between the subsystems is thereafter established by relating their corresponding thermodynamic
potentials. Such a “divide and conquer” type of approach can be especially efficient, if infinite, but
homogeneous parts of the system like bulk or surrounding gas phase can be separated off, and are then
merely represented by corresponding reservoirs [11-16]. Although not further discussed here, another
aspect could be to consider situations of “constrained equilibria” [14,15], where not all, but only some of
the subsystems are in thermodynamic equilibrium.
Rogal, J.; Reuter, K. (2007) Ab Initio Atomistic Thermodynamics for Surfaces: A Primer. In Experiment, Modeling and Simulation of Gas-
Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 2-1 – 2-18). Educational Notes RTO-EN-AVT-142, Paper 2. Neuilly-sur-
Seine, France: RTO. Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 2-1
Figure 1: Schematic representation of the system discussed here, a single-crystal metal surface
in contact with a surrounding gas phase characterized by defined temperature T and pressure p.
The shaded area represents the finite part of the system that is affected by the presence of the
surface.
2.0 SURFACES IN REALISTIC ENVIRONMENTS
2.1 Surface free energy

Here we will illustrate how this quite general concept works and what it can contribute in practice by
using it to determine the equilibrium geometry and composition of a solid surface in contact with a given
environment at finite temperature and pressure. For the sake of simplicity we consider the case of a
monoatomic metal and an oxygen atmosphere described by an oxygen pressure p and a temperature T [17-
20], and refer to the literature for the generalizations to compounds like oxides [13] or alloys [21], and to
environments containing multiple gas phase species [14,15]. Under conditions of defined (T,p), the
appropriate thermodynamic potential to consider is the Gibbs free energy. For this quantity we introduce
the following notation: Capital G refers to an absolute Gibbs free energy, while lower case g is used to
denote a Gibbs free energy per formula unit or particle. In the case of an infinite, homogeneous system the
latter is then equivalent to the chemical potential µ, i.e. if the homogeneous system is viewed as a
reservoir, µ gives the cost at which this reservoir provides particles. For convenience we will use the
symbol µ instead of g, when we want to emphasize a system’s role as a reservoir. In the present context it
is particularly the oxygen environment which acts as such a reservoir, because it can give (or take) any
amount of oxygen to (or from) the sample without changing the temperature or pressure.
Figure 1 shows a schematic representation of the system discussed here, a solid phase in contact with a
surrounding gas phase. We can break down the Gibbs free energy of this entire system into contributions
coming from the bulk of the solid phase Gsolid, from the homogeneous gas phase Ggas and an additional
term introduced through the surface ∆Gsurf
(1) G = Gsolid + Ggas + ∆Gsurf .
2-2 RTO-EN-AVT-142
If the surface is homogeneous as in the case of an ideal single-crystal surface, ∆Gsurf will scale linearly
with the surface area A, and we can introduce the surface free energy per unit area, γ. Upon rearranging eq.
(1) we therefore have
(2) γ =
1
A
(
G − G solid − G gas ) .
We notice that γ is well defined by a finite part of the total (infinite) system. With increasing distance
from the surface, eventually both the solid and the gas phase part of the total system will no longer be
affected by the created surface. Although contained in G, these (infinite) parts of the total system are then
effectively canceled out in eq. (2) by the subtraction of the equivalent amounts of homogeneous systems
(Ggas and Gsolid). We can therefore concentrate on the finite part of the system that is affected by the
surface. If this part contains NM metal atoms and NO oxygen atoms per surface area, this allows us to
rewrite eq. (2) as
1
(3) γ(T , p ) = ( G (T , p, N O , N M ) − N M g M (T , p) − N O µO (T , p) ) ,
A
where we have introduced the Gibbs free energy per metal atom gM in the bulk, and the oxygen chemical
potential µO of the gas phase.
At this stage it is appropriate to spend a few words on the sign convention. In this text, a more negative
Gibbs free energy will indicate a more stable state of the system. In the interpretation of a chemical
potential this translates to µ approaching -∞ in the limit of an infinitely dilute gas, since adding a particle
will then yield an infinite gain in entropy. As a consequence, γ > 0 indicates the cost of creating the
surface between the solid bulk phase and the homogeneous gas phase. Alternatively, when discussing the
stability of phases that result from adsorbing species at the solid surface, it can be convenient to choose
another zero reference. Instead of γ as a measure of the cost of creating the surface at all, one can
introduce the surface free energy of the clean surface
1
(4) γclean (T , p ) = ( G (T , p, 0, N M ) − N M g M (T , p) )
A
and evaluate the Gibbs free energy of adsorption as a measure of the cost with respect to the clean surface
[17,19,20]
∆G ad (T , p ) = γclean (T , p,0, N ' M ) − γ (T ,p, N O , N M ) =

(5) ,
=
1
A
(
G (T , p, N O , N M ) − G (T , p, 0, N ' M ) − N O µO (T , p ) − ( N M − N ' M )g M (T,p ) )
where the last term accounts for a possible difference in the number of metals atoms between the reference
clean surface and the oxidized surface structural model. Obviously, the most stable surface structure and
composition at given (T,p) in the gas phase is the one that minimizes the surface free energy, or
equivalently the one that leads to the most positive Gibbs free energy of adsorption at the corresponding
oxygen chemical potential.
2.2 Calculating Gibbs free energies

As apparent from eqs. (3) and (5) the quantities determining γ or ∆Gad are the Gibbs free energy of the
solid surface and of the solid bulk, as well as the chemical potential of the oxygen environment. Since the
RTO-EN-AVT-142 2-3
contributions to free energies in the gas phase and in the solid phase are quite different, we will discuss the
evaluation of these quantities separately. First we concentrate on the oxygen chemical potential, and then
on the computation of the solid phase Gibbs free energies.
2.2.1 Gas phase chemical potential

The chemical potential of oxygen, which enters eqs. (3) and (5), is determined by the condition of
thermodynamic equilibrium with the surrounding gas phase reservoir. At the accuracy level relevant for
our purposes, this gas is well described by ideal gas laws. If we write the chemical potential at given
temperature T and pressure p as
(6) µO (T , p) = 1 2 µO2 (gas) (T , p) = 1

2 (− k T ln Q
B
tot
O 2 (gas) )
+ pV / N ,
we therefore have to evaluate the partition function QOtot2 (gas) of an ideal gas composed of N
indistinguishable O2 molecules
(7) QOtot2 (gas) =

1
( )
qO
N! 2
N
=
N!
(
1 trans rot vib electr nucl
q q q q q ) N
with qO2 the partition function of one O2 molecule, which can further be subdivided into different
partition functions with obvious nomenclature. In writing eq. (7) we assumed that the nuclear/electronic
(nucl, electr) degrees of freedom are decoupled from the vibrational/rotational (vib, rot) ones (Born-
Oppenheimer approximation), and further that also vibrational and rotational motions are decoupled as
they take place on different time scales.
Inserting eq. (7) into eq. (6) we then arrive at
1 ⎡ ⎛ 1 trans N ⎞ ⎤
(8) µO (T , p) = − ⎢k B T ln⎜ N! (q ) ⎟ − pV ⎥ + 1 2 µ rot + 1 2 µ vib + 1 2 µ electr + 1 2 µ nucl ,
2N ⎣ ⎝ ⎠ ⎦
and proceed here by briefly commenting on the translational, rotational, vibrational, electronic and nuclear
free energy terms. A more detailed derivation can be found in most textbooks on Statistical Mechanics,
e.g. the one by Mc Quarrie [22].
Translational free energy

In the classical limit the energy due to the center-of-mass motion of a particle in a box is
= 2k 2
(9) εk =
2m
, k=
π
L
( )
n x xˆ + n y yˆ + n z zˆ ,
where L = V1/3 characterizes the box size, m is the mass of one particle, xˆ / yˆ / zˆ are unit-vectors in the
three cartesian directions, and nx/y/z go from 1 to ∞. In the thermodynamic limit (L →∞), the one-particle
partition function becomes
3/ 2
trans ⎛ 2πmk BT ⎞
(10) q =V⎜ 2 ⎟ .
⎝ h ⎠
2-4 RTO-EN-AVT-142
With this we rewrite the term in brackets in eq. (8) to (employing the ideal gas law pV = NkBT and the
Stirling formula ln N! ≈ N lnN - N at some stage)
1 ⎡ ⎛ 1 trans N ⎞ ⎤ ⎡⎛ 2πm ⎞ 3 / 2 ( k BT ) 5 / 2 ⎤
(11) − ⎢k BT ln⎜ N ! ( q ) ⎟ - pV ⎥ = − 2 k BT ln ⎢⎜ 2 ⎟
1
⎥ .
2N ⎣ ⎝ ⎠ ⎦ ⎣⎢⎝ h ⎠ p ⎦⎥
Rotational free energy

In the rigid rotator approximation the rotational partition function is written as
∞
⎛ − J ( J + 1) Bo ⎞
(12) q rot = ∑ (2 J + 1) exp⎜⎜⎝
J =0 k BT
⎟⎟
⎠
,
where Bo = ħ2/(2I) is the rotational constant and I the moment of inertia depending primarily on mass and
equilibrium bond lengths. In case of homonuclear diatomic molecules like O2 (or other molecules that
have multiple indistinguishable orientations), qrot is a bit more tricky as it couples with the nuclear spin
degrees of freedom (the total wave function must be antisymmetric under exchange of the
indistinguishable particles). At the temperatures of interest to us, this can be approximately combined into
a classical symmetry number σsym indicating the number of indistinguishable orientations that the molecule
can have (e.g. =1 for heteronuclear diatomic molecules, =2 for homonuclear diatomic molecules). At such
temperatures, where the spacing of the rotational levels is small compared to kBT, the sum in qrot can be
converted into an integral with the Euler-Maclaurin series and one ends up with
⎛ k BT ⎞
(13) µ rot ≈ − k BT ln⎜⎜ sym ⎟
⎟ .
⎝ σ Bo ⎠
Notice that this holds only for linear molecules, where then only Bo enters. In more complex cases one
would need to diagonalize the inertial tensor and consider all three eigenvalues Ao, Bo and Co.
Vibrational free energy

The vibrational contribution is obtained within the harmonic approximation by writing the partition
function as a sum over the harmonic oscillators of all M fundamental modes ωi of the particle,
M ∞
⎛ − (n + 1 2 )=ωi ⎞
(14) q vib = ∑∑ exp⎜⎜⎝
i =1 n = 0 k BT
⎟⎟
⎠
.
Evaluation of the geometric series yields
M ⎡ =ωi ⎛ ⎛ =ω ⎞ ⎞⎤
(15) µ vib = E ZPE + ∆µ vib = ∑ ⎢⎢
i =1 ⎣ 2
+ k BT ln⎜⎜1 − exp⎜⎜ i ⎟⎟ ⎟⎟⎥ ,
⎝ k BT ⎠ ⎠⎥⎦
⎝
where the first term arises from the zero-point vibrations.
Electronic and nuclear free energy

For most molecules internal excitation energies are large compared to kBT, so that the only term
contributing significantly to the partition function is the ground state. Taking a possible spin degeneracy of
RTO-EN-AVT-142 2-5
this ground state into account, we end up with
(16) µ electr ≈ EOtotal

2
( )
− k BT ln I spin ,
i.e. with the total energy of the O2 molecule and a term depending on the electronic spin degeneracy of the
ground state, Ispin. The same form would also be obtained for the nuclear degrees of freedom, which are
even more confined to the ground state due to the much larger separation of nuclear energy levels.
Although the nuclear partition function may thus be different from unity (according to the degeneracy of
the nuclear spin ground state), we will omit it here, since the nuclear state is rarely altered in chemical
processes and therefore does not contribute to the thermodynamic changes discussed here.
Figure 2: Temperature dependence of the relative oxygen chemical potential ∆µO(T,1 atm) at p=1
atm. Compared are tabulated values from ref. [23] (crosses) with the result of eq. (18) using the
sym spin
material parameters: σ =2, I =3, Bo=0.18 meV, and ωO=196 meV [23]. Additionally shown are
the sums of the individual contributions: vibrational (dashed line, almost coinciding with the
zero axis), vibrational+nuclear (dotted line), vibrational+nuclear+rotational (dash-dotted line).
The remaining large difference to the full result (solid line) is due to the translational
contribution.
Bringing it all together

Equations (11), (13), (15), (16) allow the analytic evaluation of the terms entering eq. (8), based on the
total energy and vibrational modes of the gas phase molecule, as well as the molecular mass, the rotational
constant, the symmetry number and the spin degeneracy (which are available in thermochemical tables
[23]). As will become clear below, it is convenient to separate out the total and zero point energy terms
and write eq. (8) in the following form
total
(17) µO (T , p) = 1
2 EO2 + 1 2 EOZPE
2
+ ∆µO (T , p ) ,
where ∆µO(T,p) contains now all temperature and pressure dependent free energy contributions
2-6 RTO-EN-AVT-142
⎧⎪ ⎡⎛ 2πm ⎞ 3 / 2 ( k T ) 5 / 2 ⎤ ⎛ k BT ⎞ ⎡ ⎞⎤ ⎫
(18) ∆µO (T , p) = − 2 k BT ⎨ln ⎢⎜ 2 ⎟
1 B
p
⎥ + ln⎜⎜ sym ⎟
⎛ =ω
⎟ − ln ⎢1 − exp⎜ o
⎜k T (
⎟⎟⎥ + ln I spin )⎪⎬ .
⎪⎩ ⎢⎣⎝ h ⎠ ⎥⎦ ⎝ σ Bo ⎠ ⎢⎣ ⎝ B ⎠⎥⎦ ⎪⎭
From the structure of eq. (18) it is apparent that one can rewrite eq. (17) in the following form
⎛ p ⎞
(19) µO (T , p) = 1 total
2 EO2 + 1 2 EOZPE + ∆µO (T , p 0 ) + 1 2 k BT ln⎜⎜ 0 ⎟⎟ ,
⎝p ⎠
2
which allows to use tabulated enthalpy and entropy values at standard pressure p0=1atm [23] to determine
∆µO(T,p) instead of using eq. (18) [13]. For oxygen, both approaches yield virtually identical results in the
temperature range discussed here, as illustrated in Fig. 2 for ∆µO(T,1 atm). However, for other (more
complex) molecules one or the other approach may be more convenient, depending on the availability of
the gas phase data.
2.2.2 Gibbs free energy of solid bulk and surface

Similar to the procedure applied for the gas phase chemical potential we address the computation of the
solid phase Gibbs free energies by first decomposing them into several contributing terms [13,15]
(20) G = Etotal + Fvib + Fconf + pV ,
where Etotal is the total (internal) energy, Fvib the vibrational free energy, and Fconf the configurational free
energy. A crucial aspect that governs our analysis of all of these terms is that the quantities of interest to
us, namely surface free energies or Gibbs free energies of adsorption, do not depend on absolute Gibbs
free energies. What enters into the corresponding eqs. (3) or (5) are differences of Gibbs free energies
only. This can allow for quite some degree of error cancellation, in particular if the different terms
correspond to rather similar situations like the Gibbs free energies of solid bulk and solid surface.
The dominant term in eq. (20) is the total energy, which as discussed in the introduction is provided
through the DFT calculations. We note in passing that the thermodynamic formalism is, of course, general
and would be equally valid when using total energies coming from e.g. less accurate schemes. However,
when using say a semi-empirical potential to provide the total energies, the accuracy of the ensuing
thermodynamic reasoning would also only be at this level. It is precisely the idea of ab initio atomistic
thermodynamics to carefully evaluate the total energy contributions and thereby carry over the predictive
power of the first-principles technique to finite (T,p)-conditions.
Turning to the other terms in eq. (20), we find from a simple dimensional analysis that its contribution to
the surface free energy (normalized to unit surface area) will be [pV/A] = atm Å3/Å2 ~ 10-3 meV/Å2. Even
for p ~ 100 atm, the pV-contribution will therefore still be less than ~0.1 meV/ Å2. We will see below that
in the intended application to metal surfaces in contact with realistic environments this is a negligible
contribution compared to the other free energy terms, which are of the order of tens of meV/ Å2.
The contribution from the vibrational degrees of freedom can be handled with the same harmonic
approximation applied already for the gas phase chemical potential. However, instead of writing the
vibrational free energy as arising from a sum of discrete fundamental modes ωi, cf. eq. (15), we now
introduce the phonon density of states (DOS) σ(ω) and obtain
(21) F vib = dω F vib (T , ω) σ (ω)

∫ ,
RTO-EN-AVT-142 2-7
Figure 3: Estimated vibrational contribution ∆Fvib,ad to the Gibbs free energy of adsorption for
the p(2x2) O/Pd(100) structure. Used is the Einstein approximation to the phonon density of
states, where the characteristic frequency of O atoms is changed from 196 meV in the gas phase
to 40 meV (black line), 80 meV (blue line) or 120 meV (red line) at the surface. Changes of the
vibrational modes of surface Pd atoms upon O adsorption are neglected.
with
=ω ⎛ ⎛ =ω ⎞ ⎞
(22) F vib (T , ω) = + k BT ln⎜⎜1 − exp⎜⎜ ⎟⎟ ⎟
⎟ .
2 ⎝ ⎝ k BT ⎠ ⎠
A proper evaluation of the vibrational contribution to ∆Gad or γ amounts therefore to computing the
phonon DOS of the solid bulk and surface. This information is contained in the PES and can
correspondingly be calculated by DFT [5]. However, since surface phonon DOS calculations are
computationally still quite involved, getting away with simpler approximations would be particularly
worthwhile. For this, we come back to the afore mentioned cancellation in the differences of Gibbs free
energies. It is crucial to realize that the vibrational free energy of the solid bulk and surface is not small or
negligible, yet what matters for the determination of e.g. the Gibbs free energy of adsorption is only the
following difference
∆F vib, ad (T )
(23) =− [∫ dω(σ ( N
1⎧
⎨
A⎩
O , N M , ω) − σ clean (0, N ' M ]
, ω) − (N M − N ' M )σ M (ω) )F vib (T , ω) −
N O vib ⎫
2
µO 2 (T )⎬ ≈
⎭
≈ − ⎨ [∫ dω(σ ( N (N M − N 'M )σ M (ω) )F vib (T , ω)]− N O EOZPE ⎫⎬

1⎧
O , N M , ω) − σ clean (0, N M , ω) − ,
A⎩ 2 2
⎭
where σ is the phonon DOS of the surface considered, σclean the one of the corresponding clean surface, σM
the one of the metal bulk, and in the second line we have realized that the temperature dependent
vibrational contribution to µOvib2 (T ) is negligible, cf. Fig. 2. Equation (23) shows that only the changes in
the vibrational properties of the atoms at the surface compared to their counterparts in the clean surface,
the solid bulk or the gas phase are decisive. Before initiating a full-blown surface phonon DOS
2-8 RTO-EN-AVT-142
calculation, it can therefore be very valuable to use approximations to σ, σclean, and σM , and obtain an order
of magnitude estimate of ∆Fvib,ad first [13]. A very simple approximation would e.g. be an Einstein model,
which considers only one characteristic frequency for each atom type. Allowing this frequency to change
significantly for surface atoms compared to those in the solid bulk or in the gas phase provides then a first
idea of the magnitude of the vibrational contribution. This approach is exemplified in Fig. 3 for the below
discussed case of a Pd(100) surface in contact with an oxygen environment. Even for quite extensive
variations of the characteristic vibrational modes, the resulting ∆Fvib,ad always stays within ≈ ±5 meV/ Å2
for the entire temperature range up to 600 K that is of interest to our study. Comparing with the much
larger contribution from the total energy terms, we will use this information to justify neglecting the
vibrational ∆Fvib,ad below. However, we stress that this is not a general result. There might well be
applications where the inclusion of vibrational effects on the surface free energy or Gibbs free energy of
adsorption can be important as e.g. in the adsorption of larger molecules [24] or in hydrogen containing
environments [25]. In such cases, it may already be sufficient to only consider some prominent vibrational
modes, but eventually an explicit calculation of the surface and bulk phonon DOS may be required.
This leaves as remaining term the configurational free energy. A full evaluation of this contribution is
computationally most involved, since it requires a proper sampling of the huge configuration space
spanned by all possible surface structures. Although modern statistical mechanics methods like Monte
Carlo simulations [26,27] are particularly designed to efficiently fulfill this purpose, they still require a
prohibitively large number of free energy evaluations to be directly linked with electronic structure
theories [6]. A way to circumvent this problem is to map the real system somehow onto a simpler,
typically discretized model system, the Hamiltonian of which is sufficiently fast to evaluate. Obvious
uncertainties of this approach are how appropriate the model system represents the real system, and how
its parameters can be determined from the first-principles calculations. The advantage, on the other hand,
is that such a detour via an appropriate (“coarse-grained”) model system often provides deeper insight and
understanding of the ruling mechanisms. If the considered problem can be described by a lattice defining
e.g. the possible adsorption sites for the gas phase species in the system, a prominent example for such a
mapping approach comes under the names lattice-gas Hamiltonians (LGHs) or cluster expansions (CEs)
[6,28-30].
Here, we will instead concentrate on a much simpler alternative, which focuses on screening a number of
known (or possibly relevant) ordered surface structures by directly comparing which of them turns out to
be most stable under which (T,p)-conditions, i.e. which of them exhibits the lowest surface free energy or
Gibbs free energy of adsorption. For sufficiently low temperatures, the remaining configurational entropy
per surface area is then only due to a limited number of defects in these ordered structures and can be
estimated as
TS conf k BT ⎛ (N + n )! ⎞
(24) = ln⎜ ⎟ ,
NA NA ⎝ N !n! ⎠
where n is the small number of defects in a system with N surface sites (n << N). For N, n >> 1 we can
apply the Stirling formula which leads to
(25)
TS conf k BT ⎡ ⎛
NA
= ⎢
A ⎣ ⎝
n⎞ ⎛n⎞ ⎛
N⎠ ⎝N⎠ ⎝
N ⎞⎤
n ⎠⎦
k T
ln⎜1 + ⎟ + ⎜ ⎟ ln⎜1 + ⎟⎥ < 0.34 B
A
(∀n N < 10%) .
With a typical area per surface site of A ~ 10 Å2 for transition metal surfaces, we correspondingly deduce
that the configurational entropy contribution to the Gibbs free energy is less than 3 meV/ Å2 for any T <
1000 K [15]. We will see below that in the application to a Pd surface in contact with an oxygen
environment this is almost always negligible and will qualify to which changes its explicit consideration
will lead. While the effect of configurational entropy due to disorder is therefore not of much concern, the
RTO-EN-AVT-142 2-9
obvious limitation in this direct screening approach is that its reliability is restricted to the number of
considered configurations, or in other words that only the stability of those structures plugged in can be
compared. The predictive power extends therefore only to those structures that are directly considered, i.e.
the existence of unanticipated surface geometries or stoichiometries cannot be predicted. As such,
appropriate care should be in place when addressing systems where only limited information about the
surface structures is available. With this in mind, the direct screening approach to ab initio atomistic
thermodynamics can still be a particularly valuable tool, since it allows, for example, to rapidly compare
the stability of newly devised structural models against existing ones. In this way, it gives tutorial insight
into what structural motives may be particularly important, which may even yield ideas about other
structures that one should test as well. Still, the limited reliability to the set of actually considered
structural models must always be borne in mind and can really only be overcome by a proper sampling of
configurational space, which then leads also to a more general and systematic way of treating phase
coexistence and order-disorder transitions.
Figure 4: Generic free energy plot (a) and surface phase diagram (b) for a surface in equilibrium
with a surrounding oxygen gas phase. a) An adsorbate phase will become more stable than the
zero reference clean metal surface, if its ∆Gad >0 for some oxygen chemical potential (note the
inverted y-scale!). If there is more than one adsorbate phase, always the one with the largest
∆Gad will become most stable, as indicated here by the red line. Finally, for ∆µO > 1/y ∆Hf(T=0K),
the bulk oxide will always result as the most stable phase. b) Converting the obtained ∆µO
stability range for each phase (indicated schematically by the different shaded regions in the
free energy plot on the left) into (T,p)-conditions using eq. (18) allows to draw the resulting
surface phase diagram.
2 - 10 RTO-EN-AVT-142
2.3 Free energy plots and surface phase diagrams

We will now employ the general thermodynamic framework developed in sections 2.1 and 2.2 to
investigate the structure and composition of a solid surface in contact with a given environment at finite
temperature and pressure. More specifically, we aim at applying it to describe a single-crystal transition
metal surface in contact with an oxygen gas phase, but before we address the actual case of a Pd(100)
surface below let us first spend some time with more general considerations. In light of the discussion in
section 2.2.2 we will neglect the vibrational contribution ∆Fvib,ad to the Gibbs free energy of adsorption,
and within the spirit of the direct screening approach we will also neglect the configurational entropy term
in the solid phase Gibbs free energies for the time being. This transforms the general eq. (5) into the
working equation
1 ⎛ total N ⎞
(26) ∆G ad (T , p ) ≈ − ⎜ E ( N O , N M ) − E total (0, N M ) − (N M − N ' M )E Mtotal − O E Ototal − N O ∆µO (T , p ) ⎟ .
A ⎝ 2 2
⎠
Evaluating ∆Gad requires now primarily total energies which are directly amenable to electronic structure
theory calculations. In the direct screening approach to ab initio atomistic thermodynamics, these
quantities would be calculated for a number of ordered surface structural models. Equation (26) allows
then to directly plot the Gibbs free energy of adsorption for each model as a function of the oxygen
chemical potential ∆µO, as illustrated schematically in Fig. 4a. This yields a straight line for each model
considered, and at any given ∆µO the model with the lowest lying line (most stable ∆Gad) is identified as
the most stable one under environmental conditions corresponding to this particular oxygen chemical
potential. Using eq. (18) in section 2.2.1 allows to relate specific (T,p)-conditions to this chemical
potential, and this information can e.g. be included in graphs like Fig. 4a in form of additional x-axes,
which give the pressure dependence at some specific temperature. Alternatively, one can concentrate only
on the most stable structures, convert the range of chemical potential in which each is most stable into
corresponding (T,p) ranges, and plot these stability ranges in surface phase diagrams of the form of Fig.
4b. It is important to realize that both kinds of plots are based on exactly the same information. Surface
phase diagrams (Fig. 4b) provide a more direct insight to the experimentally accessible (T,p) conditions,
whereas free energy plots (Fig. 4a) summarize the two-dimensional dependence conveniently in the one-
dimensional, but less intuitive dependence on the chemical potential. Additionally, it is only in the latter
kind of plots that also information about the energetic difference to alternative, less stable surface
structural models is provided. These plots make also immediately apparent that the transition from one
stable phase to another occurs within the present framework always at a specific value of the oxygen
chemical potential, cf. Fig. 4a, which is the reason why the phase boundaries in (T,p) surface phase
diagrams of the type of Fig. 4b exhibit similar curvatures (lines of constant ∆µO) . We note in passing that
a third and equally equivalent way of plotting the results would be to plot the stability ranges in (1/T, p)-
figures, in which case these boundaries between the stable phases would then result as straight lines [31].
Realizing that the definition for the average binding energy at T = 0 K is
1 ⎛ total total N total ⎞

(27) Eb = − ⎜ E ( N O , N M ) − E (0, N M ) − O EO2 ⎟ ,
NO ⎝ 2 ⎠
with Eb > 0 for exothermicity, we arrive at the expression
NO N
(28) ∆G ad (T , p ) ≈ Eb + (N M − N 'M )E Mtotal + O ∆µO (T , p ) ,
A A
which has a rather intuitive structure: Forming the oxidized surface by accommodating NO oxygen atoms
yields an energy gain of NOEb (per surface area), that is opposed by the cost of taking these O atoms out of
RTO-EN-AVT-142 2 - 11
the gas phase reservoir, hence NO∆µO. The equivalent term (N M − N ' M )E Mtotal comes only into play for
oxidized surfaces where the total number of metal atoms is different to the one of the reference clean
metal surface, and represents then the cost of transferring the corresponding number of metal atoms to or
from the reservoir represented by the metal bulk.
On the basis of eq. (28) the general structure of a surface free energy plot for a metal surface in contact
with an oxygen environment can nicely be discussed. In the limit of an infinitely dilute gas (∆µO → –∞),
any surface structure containing oxygen (NO ≠ 0) will exhibit an infinitely negative ∆Gad, reflecting that it
is very unfavorable to maintain oxygen adsorbed at the surface under such conditions. As intuitively clear,
the clean surface will therefore always result as most stable in such environments. With increasing oxygen
content in the gas phase, ∆µO will become less negative and so will the ∆Gad of oxygen containing surface
structural models. Eventually, one of them will exhibit a ∆Gad > 0 and will then become more stable than
the clean surface. At corresponding oxygen pressures and temperatures, oxygen is getting stabilized at the
surface, cf. Fig. 4a, and the governing factors for this are immediately revealed by the structure of eq. (28):
The slope of each ∆Gad-line is determined by NO/A, i.e. the more oxygen is contained in the structure, the
faster this structure becomes more favorable with increasing oxygen chemical potential. For surfaces
preserving the number of metal atoms (NM =N’M), the x-axis intercept (i.e. the moment when the structure
becomes more stable than the clean surface) is reached at ∆µO = –Eb. A more stable binding of oxygen in
the surface structural model will correspondingly shift the ∆Gad-line in the free energy plot down and the
x-axis intercept to the left, and will thereby render the structure more stable than the clean surface at
already lower oxygen contents in the gas phase. A large ∆Gad at increasing oxygen chemical potential and
therewith the chance to become the most stable structure can therefore be reached by surface structural
models that either offer a strong binding of their oxygen species or contain a large number of oxygen
atoms per surface area. This way, a surface structure that strongly binds a few oxygen atoms could for
example become more favorable than the clean surface at low chemical potentials, while another surface
structure with weaker binding, but higher oxygen coverage will eventually become more stable at
somewhat higher chemical potentials due to its steeper slope, cf. Fig. 4a.
The highest number of oxygen atoms per surface area is ultimately reached by bulk oxide structures, i.e.
when the oxygen content in the environment is high enough to create an infinitely thick bulk oxide on top
of the metal substrate. Since then NO → ∞, the corresponding ∆Gad-line in the free energy plot is vertical,
cf. eq. (28). The intercept of this line with the x-axis is given by the condition that the bulk oxide becomes
thermodynamically more favorable than an equivalent amount of bulk metal and gas phase oxygen
(29) g M x Oy ≤ xg M + yµO ,
where g M x O y is the Gibbs free energy per formula unit of the oxide bulk. Using eq. (17), this yields
1⎛ y ⎞ 1
(30) > ⎜ g M x O y (T,p ) − xg M (T,p ) − EOtotal
∆µO ~ ⎟ = ∆H f (T = 0K ) ,
y⎝ 2 2 ⎠ y
where ∆Hf (T=0 K) is the heat of formation of the bulk oxide at T=0 K [13,17]. For any ∆µO higher than
this limit, the bulk oxide will always be the stable phase, cf. Fig. 4a.
After this more general discussion we proceed with the specific case of a Pd(100) surface in contact with
an oxygen atmosphere, to illustrate how the direct screening approach to ab initio atomistic
thermodynamics works and what it can contribute in practice. Typical for late transition metal surfaces,
the interest in this system comes from the widespread technological use of Pd, for example in the area of
oxidation catalysis [32]. Although this material is known for its propensity to form oxidic structures in
technologically-relevant high oxygen pressure environments, the possible formation of sub-nanometer thin
2 - 12 RTO-EN-AVT-142
oxidic films (so-called surface oxides) has only recently been addressed [17,18,33,35,36]. While
traditionally such films were conceived as closely related thin versions of the corresponding (known) bulk
oxides, recent atomic-scale characterizations of initial few-atomic-layer thick oxide overlayers especially
on Pd and Ag surfaces have revealed structures that have little resemblance to their bulk counterparts,
and/or are influenced to a large degree by a strong coupling to the underlying metal substrate
[17,19,20,33-37]. Due to this coupling and structures particularly suited for layered configurations, one
may expect the stability range for such surface oxides to exceed that of the hitherto discussed bulk oxides
[18].
Figure 5: Computed free energy plot (a) and surface phase diagram (b) for O/Pd(100), following
the general style of Fig. 4. Considered are two ordered adlayers with O in the fourfold hollow
sites (p(2x2), ¼ monolayer (ML) coverage, and c(2x2), ½ ML coverage) and the (√5 x √5)R27°–O
surface oxide (0.8 ML coverage). Note the extended stability range of the surface oxide
compared to the known PdO bulk oxide. The total energies (DFT-GGA, PBE) used to construct
this graph via eq. (26) are taken from refs. [36,40], the surface unit-cell area of Pd(100) is 7.8 Å2.
In the spirit of the direct screening approach we therefore consider here the three experimentally
characterized oxygen-containing surface structures to date, namely two ordered adlayers with O in the
fourfold hollow sites (p(2x2), ¼ monolayer (ML) coverage, and c(2x2), ½ ML coverage) and the (√5 x
√5)R27°–O surface oxide (0.8 ML coverage) [38,39]. The latter structure corresponds to a rumpled, but
commensurate PdO(101) film with a strong coupling to the underlying substrate [35]. Evaluating the
calculated DFT binding energies for these three surface structures leads to the results displayed in Fig. 5a
[35,36,40]. They nicely follow the more general structure discussed above: While the clean surface is the
most stable structure at the lowest oxygen chemical potentials, the p(2x2) structure exhibits a higher ∆Gad
RTO-EN-AVT-142 2 - 13
for ∆µO > –1.35 eV. Due to its steeper slope (higher coverage), the (√5 x √5)R27°–O surface oxide
becomes even more favorable for ∆µO > –1.15 eV, while ultimately in the most oxygen-rich environments
(∆µO > –0.87 eV) the PdO bulk oxide results as most stable phase. Converting this information about the
stability ranges of these different phases by means of eq. (18) into (T,p)-conditions leads to the plot also
shown in Fig. 5b.
Referring to the more detailed original literature [17,18,35,36], we restrict our discussion of these results
here to two noteworthy points: First, as motivated above, there is indeed a surprisingly large range of
(T,p)-conditions, where the surface oxide structure represents the thermodynamically most stable
structure. In corresponding oxygen environments, this sub-nanometer thin film will thus eventually form
on time scales set by possible kinetic limitations, but never grow thicker. Due to this finite thickness, the
coupling at the oxide-metal interface and an atomic structure that can be quite different to the one of the
known bulk oxides, one might suspect new properties that are distinct to those of surfaces of both bulk
metals and bulk oxides, and could thus be of potential interest for applications. Second, the c(2x2)
structure is never a most stable phase. This implies that the frequent observation of this structure in ultra-
high vacuum (UHV) experiments [38,39] is a mere outcome of the limited O supply offered, as well as of
kinetic barriers to the formation of the surface oxide, e.g. due to limitations in the O penetration at the low
temperatures employed (UHV experiments are typically performed by depositing a finite number of
adatoms, rather than by maintaining a given gas pressure [41]).
Instead of further dwelling on the physics of this system, let’s return now to the methodological
discussion. As already stated several times, the validity of these results is restricted by the limited number
of surface structural models considered in the direct screening approach. Apart from that, uncertainties are
introduced due to the neglected vibrational free energy contribution to ∆Gad, as well as due to inaccuracies
in the total energy difference entering eq. (26). Inspecting the y-axis scale of the free energy plot in Fig.
5a, we see for example that even small changes in ∆Gad of the order of ~ ±5 meV/ Å2 may still
considerably shift the (T,p) stability ranges for the various phases (due to the shifted crossing points of the
various lines with different, but similar slopes). On the other hand, the sequence of stable phases in
increasingly oxygen-rich environments (clean surface, p(2x2) adlayer, (√5 x √5)R27°–O surface oxide,
PdO bulk oxide) is not affected by such changes [40]. Both the numerical uncertainty in the total energy
difference due to the finite basis set employed in the DFT calculations and the neglected ∆Fvib,ad term are
of this order of magnitude and in the corresponding light the reported results have to be seen. An even
larger uncertainty in the total energy difference may result from the approximate exchange-correlation
(XC) functional in the DFT calculations. While in the present example using local-density or several
generalized gradient functionals still led to the same sequence of stable phases [40], extreme caution is
advisable in general. As always, the accuracy level is dictated by the questions one wants to get answered.
If required, systematic improvement on the numerical uncertainty in the total energy difference and on the
vibrational free energy contribution is in principle always possible (albeit in practice at possibly high or
prohibitive computational cost). Concerning the uncertainty due to the approximate XC functional, at least
one may compare the results obtained with differently constructed functionals. If doubts remain, a regional
XC correction or higher-level electronic structure calculations may be necessary.
As a final point, we briefly comment on the effect of the neglected configurational entropy contribution.
As discussed in section 2.2.2 at sufficiently low temperatures this term is quite small and can therefore
only have an effect when two competing ∆Gad-lines come very close to each other [15]. This is the case at
the transitions between stable phases, and in fact, the deliberately neglected configurational entropy term
is the reason why these boundaries are drawn abrupt in the surface phase diagram in Fig. 5 – even at the
highest temperatures shown. In reality, finite phase coexistence regions should occur at finite
temperatures, i.e. regions in which with changing pressure one phase gradually becomes populated and the
other one depopulated. With increasing temperature, the width of these coexistence regions around the
phase transitions increases, until eventually there are no pressures left in which one still finds the well-
ordered surface structures now displayed in Fig. 5. Only a proper evaluation of the configurational entropy
2 - 14 RTO-EN-AVT-142
term, e.g. through Monte Carlo simulations, can provide detailed insight into these order-disorder
transitions and/or the phase coexistence regions themselves, and corresponding care has to be taken in the
interpretation of results at elevated temperatures, when the configurational entropy term is neglected as in
the direct screening approach to ab initio atomistic thermodynamics discussed here [6].
3.0 SUMMARY
A predictive modeling of materials properties requires a consistent treatment in the wide hierarchy of
scales from the electronic level to macroscopic lengths and times. The central idea of ab initio atomistic
thermodynamics is to employ the information on the potential energy surface provided by modern
electronic structure theories, in order to calculate appropriate thermodynamic potential functions. With the
latter, macroscopic system properties at finite temperatures can immediately be discussed. At surfaces,
such a thermodynamic description can be particularly useful, since it provides the possibility to suitably
divide the total system into smaller subsystems that are mutually (or partly) in equilibrium with each
other. This way, infinite, but homogeneous parts of the system like bulk or surrounding gas phase can be
efficiently represented by corresponding reservoirs, which e.g. allows to address surfaces in contact with
realistic environments.
In this tutorial text we have focused on a very simple realization of this general scheme, namely the direct
screening approach, to determine the equilibrium geometry and composition of a solid surface in contact
with a given environment at finite temperature and pressure. For the sake of clarity we considered the case
of a monoatomic metal and an oxygen atmosphere, but the conceptual framework is readily generalized to
more complex systems, involving compounds like oxides or alloys, or environments containing multiple
gas phase species. In the direct screening approach one focuses on a number of known (or possibly
relevant) ordered surface structures, and directly compares which of them turns out to be most stable under
which (T,p)-conditions, i.e. which of them exhibits the lowest surface free energy or Gibbs free energy of
adsorption. This provides first valuable insight into the structure and composition of the surface in realistic
or technologically relevant environments at virtually no extra computational cost compared to the
underlying electronic structure theory calculations.
The major limitation of the direct screening approach is that its reliability is restricted to the number of
considered configurations, i.e. the existence of unanticipated surface geometries or stoichiometries cannot
be predicted. This can only be overcome by a proper sampling of configurational space, as e.g. provided
by modern statistical mechanics methods like Monte Carlo simulations, which then leads also to a more
general and systematic way of treating phase coexistence and order-disorder transitions. Last, not least,
one should always keep in mind that (regardless of whether direct screening or statistical sampling) ab
initio atomistic thermodynamics is – as reflected by the name – a thermodynamic theory and as such
describes systems that had infinitely long time to fully equilibrate. It provides no information on what time
scale (with which kinetic hindrance) this equilibration took place. For this, one necessarily needs to go
beyond a thermodynamic description and explicitly follow the kinetics of the system over time.
Acknowledgments
We gratefully acknowledge valuable discussions and contributions from our (former) colleagues at the
Theory Department of the Fritz-Haber-Institut, in particular Wei-Xue Li, Cathy Stampfl and Mira
Todorova. Particular thanks go to Matthias Scheffler for his continued support and the many insightful
discussions that led to the development and application of the here described ab initio atomistic
thermodynamics approach to metal oxidation and oxide formation. This work was partially supported by
the Deutsche Forschungsgemeinschaft (DFG) in the priority program SPP-1091, and by the EU under
contract no. NMP3-CT-2003-505670 (NANO2).
RTO-EN-AVT-142 2 - 15
4.0 REFERENCES
[1] P. Hohenberg and W. Kohn, Phys. Rev. B 136, 864 (1964).
[2] W. Kohn and L. Sham, Phys. Rev. A 140, 1133 (1965).
[3] R.G. Parr and W. Yang, Density Functional Theory of Atoms and Molecules, Oxford University
Press, New York (1989).
[4] R.M. Dreizler and E.K.U. Gross, Density Functional Theory, Springer, Berlin (1990).
[5] R.M. Martin, Electronic Structure. Basic Theory and Practical Methods, Cambridge University
Press, Cambridge (2004).
[6] K. Reuter, C. Stampfl, and M. Scheffler, “Ab Initio Thermodynamics and Statistical Mechanics of
Surface Properties and Functions”. In: S. Yip (ed.), Handbook of Materials Modeling, Part A.
Methods, Springer, Berlin (2005). ISBN-10 1-4020-3287-0.
[7] C.M. Weinert and M. Scheffler, In: H.J. von Bardeleben (ed.), Defects in Semiconductors, Mat. Sci.
Forum 10-12, 25 (1986).
[8] E. Kaxiras, Y. Bar-Yam, J.D. Joannopoulos, and K.C. Pandey, Phys. Rev. B 35, 9625 (1987).
[9] M. Scheffler, “Thermodynamic Aspects of Bulk and Surface Defects – First-Principles

Calculations”. In: J. Koukal (ed.), Physics of Solid Surfaces – 1987, Elsevier, Amsterdam (1988); M.
Scheffler and J. Dabrowski, Phil. Mag. A 58, 107 (1988).
[10] G.-X. Qian, R.M. Martin, and D.J. Chadi, Phys. Rev. B 38, 7649 (1988).
[11] X.-G. Wang, W. Weiss, Sh.K. Shaikhutdinov, M. Ritter, M. Petersen, F. Wagner, R. Schlögl, and M.
Scheffler, Phys. Rev. Lett. 81, 1038 (1998).
[12] X.-G.Wang, A. Chaka, and M. Scheffler, Phys. Rev. Lett. 84, 3650 (2000).
[13] K. Reuter and M. Scheffler, Phys. Rev. B 65, 035406 (2002).
[14] K. Reuter and M. Scheffler, Phys. Rev. Lett. 90, 046103 (2003).
[15] K. Reuter and M. Scheffler, Phys. Rev. B 68, 045407 (2003).
[16] Z. Lodzianan and J.K. Nørskov, J. Chem. Phys. 118, 11179 (2003).
[17] K. Reuter and M. Scheffler, Appl. Phys. A 78, 793 (2004).
[18] K. Reuter, “Nanometer and Sub-Nanometer Thin Oxide Films at Surfaces of Late Transition
Metals”. In: U. Heiz, H. Hakkinen, and U. Landman (eds.), Nanocatalysis: Principles, Methods,
Case Studies, Springer, Berlin (2006).
[19] W.-X. Li, C. Stampfl, and M. Scheffler, Phys. Rev. B 68, 16541 (2003).
2 - 16 RTO-EN-AVT-142
[20] W.-X. Li, C. Stampfl, and M. Scheffler, Phys. Rev. Lett. 90, 256102 (2003).
[21] J.R. Kitchin, K. Reuter, and M. Scheffler, Phys. Rev. B (in preparation).
[22] D.A. Mc Quarrie, Statistical Mechanics, Harper and Row, New York (1976).
[23] D.R. Stull and H. Prophet, JANAF Thermochemical Tables, 2nd edn., U.S. National Bureau of
Standards, Washington, D.C. (1971).
[24] D. Loffreda, Surf. Sci. (in press).
[25] Q. Sun, K. Reuter, and M. Scheffler, Phys. Rev. B 67, 205424 (2003).
[26] D. Frenkel and B. Smit, Understanding Molecular Simulation, 2nd edn., Academic Press, San Diego
(2002).
[27] D.P. Landau and K. Binder, A Guide to Monte Carlo Simulations in Statistical Physics, Cambridge
University Press, Cambridge (2002).
[28] D. de Fontaine, In: P.E.A. Turchi and A. Gonis (eds.), Statics and Dynamics of Alloy Phase
Transformations, NATO ASI Series, Plenum Press, New York (1994).
[29] J.M. Sanchez, F. Ducastelle, and D. Gratias, Physica A 128, 334 (1984).
[30] A. Zunger, “First Principles Statistical Mechanics of Semiconductor Alloys and Intermetallic
Compounds”. In: P.E.A. Turchi and A. Gonis (eds.), Statics and Dynamics of Alloy Phase
Transformations, NATO ASI Series, Plenum Press, New York (1994).
[31] C.T. Campbell, Phys. Rev. Lett. (in press).
[32] G. Ertl, H. Knözinger, and J. Weitkamp (eds.), Handbook of Heterogeneous Catalysis, Wiley, New
York (1997).
[33] E. Lundgren, G. Kresse, C. Klein, M. Borg, J.N. Andersen, M. De Santis, Y. Gauthier, C. Konvicka,
M. Schmid, and P. Varga, Phys. Rev. Lett. 88, 246103 (2002).
[34] A. Michaelides, M.L. Bocquet, P. Sautet, A. Alavi, and D.A. King, Chem. Phys. Lett. 367, 344
(2003).
[35] M. Todorova, E. Lundgren, V. Blum, A. Mikkelsen, S. Gray, J. Gustafson, M. Borg, J. Rogal, K.

Reuter, J.N. Andersen, and M. Scheffler, Surf. Sci. 541, 101 (2003).
[36] E. Lundgren, J. Gustafson, A. Mikkelsen, J.N. Andersen, A. Stierle, H. Dosch, M. Todorova, J.

Rogal, K. Reuter, and M. Scheffler, Phys. Rev. Lett. 92, 046101 (2004).
[37] A. Michaelides, K. Reuter, and M. Scheffler, J. Vac. Sci. Technol. A 23, 1487 (2005).
[38] D. Kolthoff, D. Jürgens, C. Schwennicke, and H. Pfnür, Surf. Sci. 365, 374 (1996).
[39] G. Zheng and E.I. Altman, Surf. Sci. 504, 253 (2002).
RTO-EN-AVT-142 2 - 17
[40] J. Rogal, K. Reuter, and M. Scheffler (in preparation).
[41] D.P. Woodruff and T.A. Delchar, Modern Techniques of Surface Science, 2nd edn., Cambridge
University Press, Cambridge (1994).
2 - 18 RTO-EN-AVT-142
Introduction to Theoretical Surface Science
Axel Groß
Abteilung Theoretische Chemie
Universität Ulm
D-89069 Ulm
GERMANY
email: axel.gross@uni-ulm.de
Abstract
Recent years have seen a tremendous progress in the microscopic theoretical treatment of
surfaces and processes on surfaces. While some decades ago a phenomenological thermody-
namic approach was dominant, a variety of surface properties can now be described from first
principles, i.e. without invoking any empirical parameters. Consequently, the field of theoretical
surface science is no longer limited to explanatory purposes only. It has reached such a level of
sophistication and accuracy that reliable predictions for certain surface science problems have
become possible. Hence both experiment and theory can contribute on an equal footing to the
scientific progress.
In this lecture, the theoretical concepts and computational tools necessary and relevant for
theoretical surface science will be introduced. A microscopic approach towards the theoretical
description of surface science will be presented. Based on the fundamental theoretical entity,
the Hamiltonian, a hierarchy of theoretical methods will be introduced in order to describe
surface structures and processes at different length and time scales. But even for the largest
time and length scales, all necessary parameters will be derived from microscopic properties.
1 Introduction
It is the aim of theoretical surface science to contribute significantly to the fundamental under-
standing of the underlying principles that govern the geometric and electronic structure of surfaces
and the processes occuring on these surfaces such as growth of surface layers, gas-surface scattering,
friction or reactions at surfaces [1]. Processes on surfaces are play a tremendous technological role
since every device interacts with the environment through its surface. Some processes at surfaces
are very beneficial. For example, chemical reactions can be enormously facilitated at the surface
of a catalyst. Some processes at surfaces, however, are harmful, such as corrosion or also friction.
On the experimental side, modern surface science is characterized by a broad variety of tools with
microscopic resolution, the most prominent being the scanning tunneling microscope (STM) [2]
which allows to image single atoms of a surface. Integrating techniques based on scattering such
Groß, A. (2007) Introduction to Theoretical Surface Science. In Experiment, Modeling and Simulation of Gas-Surface Interactions for
Reactive Flows in Hypersonic Flights (pp. 3-1 – 3-22). Educational Notes RTO-EN-AVT-142, Paper 3. Neuilly-sur-Seine, France: RTO.
Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 3-1
11
00
00
11 11
00
00
11
00
11
Figure 1: Simulation of surface structures and processes at surfaces on the computer.
as low energy electron diffraction (LEED) also give precise information about atomic positions
on surfaces. As far as theory is concerned, while some decades ago theory could more or less
only give qualitative explanations for surface science problems, the considerable improvement of
computer power and the development of efficient algorithms have made it possible to describe
many surface properties qualitatively and even quantitatively from first principles, i.e. without
invoking any empirical parameters. It is true that the solution of the theoretical problems requires
a computational approach. However, this approach is advanced enough that computational surface
science may act as a virtual chemistry and physics lab at surfaces (see Fig. 1). Hence there is now
a very fruitful give-and-take situation between experiment and theory with theory also being able
to make reliable predictions [3].
In this chapter, I will give an introduction into the current status of theoretical surface science
based on electronic structure theory. I will follow a hierarchical approach. First the basic theoretical
entity, the Hamiltonian, will be discussed, and then methods to solve the corresponding Schrödinger
equation will be introduced on microscopic properties. Using prototypical examples, I will give an
overview over the geometric and electronic structure of surfaces and the interaction of molecules
with surfaces.
2 The Hamiltonian
In solid state physics as well as in chemistry, the fundamental particles are nuclei and electrons
interacting with each other through electrostatic forces. Neglecting relativistic and magnetic effects,
the Hamiltonian describing a system of nuclei and electrons is given by
H = Tnucl + Tel + Vnucl−nucl + Vnucl−el + Vel−el , (1)
where Tnucl and Tel are the kinetic energy of the nuclei and the electrons, respectively, and the
other terms describe the electrostatic interaction between the positively charged nuclei and the
3-2 RTO-EN-AVT-142
electrons. Explicitly, the single terms are

L N
X P2I X p2i
Tnucl = , Tel = , (2)
2MI 2m
I=1 i=1
and
1 X ZI ZJ e2 X ZI e2 1X e2
Vnucl−nucl = , Vnucl−el = − , Vel−el = . (3)
2 |RI − RJ | |ri − RI | 2 |ri − rj |
I6=J i,I i6=j
Here, atomic positions are denoted by capital letters, and CGS-Gaussian units have been used. In
principle, we could stop here because all what is left to do is to solve the many-body Schrödinger
equation using the Hamiltonian (1)
HΦ(R, r) = EΦ(R, r) , (4)
taking the proper quantum statistics such as the Pauli principle for the electrons into account.
Unfortunately, the solution of the Schrödinger equation (4) in closed form is not possible, except
for textbook examples such as the harmonic oscillator or the hydrogen atom. In order to solve this
Schrödinger equation, a hierarchy of approximations is necessary. Usually the first approximation
is the Born–Oppenheimer or adiabatic approximation [4] which is based on the separation in the
time scale of processes involving electrons and atoms because of their large mass mismatch. Hence
one assumes that the electrons follow the motion of the nuclei almost instantaneously.
In practice, one splits up the full Hamiltonian and defines the electronic Hamiltonian Hel for fixed
nuclear coordinates {R} as follows
Hel ({R}) = Tel + Vnucl−nucl + Vnucl−el + Vel−el . (5)
The eigenvalues Eel ({R}) of the electronic Schrödinger for a given fixed configuration {R} of the
nuclei
Hel ({R})Ψ(r, {R}) = Eel ({R})Ψ(r, {R}) . (6)
then define the Born–Oppenheimer energy surface which acts as the potential for the nuclear
motion. It is important to realize that within the Born-Oppenheimer approximation electronic
excitations and electronically non-adiabatic effects are entirely suppressed. In order to describe
these processes, one has to go beyond the Born-Oppenheimer approximation.
The electronic Schrödinger equation (6) is still complex enough that its solution is not trivial. As
indicated in Fig. 1, in surface science one typically deals with systems where a finite object, a
molecule, is interacting with a semi-infinite substrate whose ideal three-dimensional periodicity is
broken in one direction. Historically, quantum chemists were the first to treat surfaces theoretically
describing the surface in the cluster approach as a big molecule with wave-function based meth-
ods. In recent years, periodic calculations using density functional theory (DFT) [5] have become
dominant in this field since their computational effort scales much more favorably with system size
than wave-function based methods.
Before addressing methods to solve the electronic Schrödinger equation in more detail, we will first
discuss the symmetry properties of surfaces. This is directly relevant for solving the Schrödinger
RTO-EN-AVT-142 3-3
Figure 2: Structure of surfaces. Left panel: fcc crystal with (100) faces and one (111) face, right
panel: fcc crystal with (100) faces and one (110) face. The surface unit cells of the corresponding
faces are also indicated in the figure.
equation since every symmetry present in the Hamiltonian reduces the required computational
effort significantly. Furthermore, thereby we will also introduce the notation used for the charac-
terization of surfaces. The periodicity of a crystal lattice leads to the existence of the reciprocal
space which reflects the translational symmetry of a lattice. Its lattice vectors are defined by
a2 × a3
b1 = 2π . (7)
|a1 · (a2 × a3 )|
where the ai are the basis vectors of the real space. The other two basis vectors of the reciprocal
lattice b2 and b3 are obtained by a cyclic permutation of the indices in (7). By construction, the
lattice vectors of the real space and the reciprocal lattice obey the relation
ai · bj = 2πδij , (8)
where δij is the Kronecker symbol. Further details can be found in any textbook about solid state
physics [6, 7].
A surface can be thought as being created by just cleaving an infinite crystal along one surface
plane. A bulk-terminated surface, i.e. a surface whose configuration has not changed after cleavage,
is called an ideal surface. Surface planes in a crystal are denoted by the so-called Miller indices
hkl which are defined through the shortest reciprocal lattice vector hb1 + kb2 + hb3 that is
perpendicular to the particular plane. In Fig. 2, a fcc crystal with (100) faces is shown which have
a square symmetry. Furthermore, a (111) face with a hexagonal symmetry and a (110) face with a
rectangular symmetry are created by additional cuts. Together with the centered rectangular and
the oblique lattices these symmetry types form the five possible two-dimensional Bravais lattices.
It is important to realize that there is no surface that remains ideal. This is due to the fact
that at the surface compared to the periodic bulk there are nearest-neighbor atoms missing so
that the bonding situation is changed. If the atoms at the surface just change there position
3-4 RTO-EN-AVT-142
vertically without altering the surface symmetry, the surface is called to be relaxed. If, however,
the rearrangement of the surface atoms changes the symmetry and periodicity of the surface, the
surface is called to be reconstructed.
The symmetry of the surface of course also enters the Hamiltonian. It is always advantageous to
take into account all symmetries of the Hamiltonian since only states that belong to a particular
irreducible representation of the symmetry group of the Hamiltonian couple to each other through
the Hamiltonian. One consequence of the existence of the two-dimensional periodicity along the
surface is that the Bloch theorem is applicable in two dimensions. This means that the electronic
single-particle wave functions form two-dimensional Bloch states
ψkk (r) = eikk ·r ukk (r) , (9)
where ukk (r) has the two-dimensional periodicity of the surface.
3 Electronic structure calculations
The first calculations of surface structures were done by quantum chemists in the cluster approach
in which the surface is regarded as a big molecule and not as an infinite periodic structure. This
ansatz is guided by the idea that bonding on surfaces is a local process. However, these wave-
function based calculations become prohibitively expensive for larger systems. Furthermore, this
cluster approach is not really reliable for the description of extended electronic state [8] as present
in, e.g., metals.
Nowadays, electronic structure calculations dealing with surfaces are predominantly performed in
the framework of density functional theory (DFT) [9, 10]. They offer a good compromise between
computational efficiency and sufficient accuracy for many systems. DFT can be regarded as an
extension of the Thomas-Fermi theory [11] to inhomogeneous situations [9]. It is based on the
Hohenberg-Kohn theorem which states that the exact ground-state density and energy can be
determined by the minimization of the energy functional E[n],
Etot = min E[n] = min(T [n] + Vext [n] + VH [n] + Exc [n]) . (10)
n(r) n(r)
From the Hohenberg-Kohn theorem also follows that there is a one-to-one correspondence be-
tween the electron ground-state density n(r) and the external potential vext (r). Vext [n] and VH [n]
are the functionals of the external potential and of the classical electrostatic interaction energy,
respectively, while T [n] is the kinetic energy functional for non-interacting electrons. All quan-
tum mechanical many-body effects are contained in the so-called exchange-correlation functional
E xc [n], which is an universal functional of the density, i.e., it does not depend on any particular
element, but unfortunately it is not known in general.
Since the kinetic energy functional T [n] is not precisely known for inhomogeneous situations, the
Hohenberg-Kohn theorem is not useful for a direct implementation of the DFT (however, there is
recently a renewed interest in so-called orbital-free DFT calculations [12]). Instead, one expresses
RTO-EN-AVT-142 3-5
the electron density n(r) as the sum over so-called single-particle Kohn-Sham states
N
X
n(r) = |ψi (r)|2 , (11)
i=1
which are obtained by a self-consistent solution of a set of coupled effective one-particle equations,
the Kohn-Sham equations [10]
~2 2

− ∇ + vext (r) + vH (r) + vxc (r) ψi (r) = εi ψi (r) , (12)
2m
where vext is the external potential. The Hartree potential vH is given by
e2
Z
vH (r) = d3 r0 n(r0 ) . (13)
|r − r0 |
Furthermore, the exchange-correlation potential vxc (r) is the functional derivative of the exchange-
correlation functional Exc [n]
δExc [n]
vxc (r) = . (14)
δn
The ground state energy can then be expressed as
N
X Z
E= εi + Exc [n] − vxc (r)n(r) d3 r − VH . (15)
i=1
The first term in the total-energy expression (15) is also called the band structure term Ebs since
it corresponds to the sum over the single-particle energies. The Kohn-Sham equations have to be
solved iteratively in a self-consistent scheme since the solutions of the Kohn-Sham equations also
enter the effective one-particle Hamiltonian. Initially the electron density is guessed, for example
as a superposition of atomic densities. The Kohn-Sham equations are then solved and the resulting
density is compared to the initial guess. If the difference is larger than some pre-specified value,
the new density enters the Kohn-Sham equations (often using some mixing scheme), and the cycle
is repeated so often until the iterations no longer modify the solutions, i.e. until self-consistency is
reached.
In principle, DFT provides an exact expression of the total energy. Unfortunately, this is of no great
practical advantage since the correct form of the non-local exchange-correlation functional and the
related exchange-correlation potential vxc is not known. Therefore, approximative expressions are
needed. For bulk situations, the local density approximation (LDA), in which at any position r the
exchange-correlation potential of the homogeneous electron gas with the corresponding electron
density is used, has been surprisingly successful [13]. However, for chemical reactions at surfaces the
LDA not sufficiently reliable [14]. For large scale DFT calculations of surfaces and nanostructures
the so-called generalized gradient approximation (GGA) [15] represents the state of the art. In
the GGA, the gradient of the density also into account in the exchange-correlation functional,
but in such a way that important electronic sum rules are obeyed. The GGA yields a satisfactory
accuracy for many applications, although there are still exceptions [16, 17]. There is certainly a
strong need for further improvements of the exchange-correlation functionals.
3-6 RTO-EN-AVT-142
Figure 3: Illustration of the supercell approach. A substrate of a fcc crystal with a (410) surface
termination and an adsorbed periodic atomic layer in a (2 × 1) geometry is represented by an
infinite array of slabs. The supercell and the surface unit cell are indicated in the figure.
It is numerically very efficient to use a plane-wave expansion of the Kohn-Sham single-particle

states, but such an approach usually requires a three-dimensional periodicity of the considered
system. Still, plane-wave codes can be used to model surface problems in the so-called supercell
approach: the surfaces are modeled by periodically repeated slabs. In Fig. 3, a typical supercell
describing the adsorption of atoms at the step sites of a nanostructured fcc(410) surface in a
(2 × 1) geometry is shown. The slabs have to be thick enough to reproduce the correct electronic
structure of the substrate, and they have to be separated enough in order to avoid any interaction
between the slabs. One advantage of the slab approach is that the substrates are infinitely extended
in lateral directions which yields a correct description of the delocalized nature of the electronic
states of metals, a feature that is not present when the substrate is modeled by finite clusters [8].
4 Electronic and geometric structure of surfaces
At a surface of a solid, the electronic structure is strongly modified with respect to the bulk
electronic structure. Although the periodicity parallel to the surface is conserved, along the surface
normal the periodicity is broken so that the wave number kz of the Bloch waves no longer is a
good quantum number. There is a very simple model which is able to illustrate some fundamental
properties of the electronic structure of surfaces, in particular simple metal surfaces, namely the
so-called jellium model. In this model, the positive ion charges are replaced by a uniform charge
RTO-EN-AVT-142 3-7
rs = 5
1.0 11111111111111111
00000000000000000
00000000000000000
11111111111111111
Electron density (1/n)

00000000000000000
11111111111111111
11111111111111111
00000000000000000
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
−
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111 Positive
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111 background
0.5 11111111111111111
00000000000000000
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111 rs = 2
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
00000000000000000
11111111111111111
0.0 11111111111111111
00000000000000000
−1.0 −0.5 0.0 0.5 1.0
Distance from the surface (1/k F)
Figure 4: Charge density as a function of the distance from the surface in Fermi wavelengths
determined within the jellium model for two different background densities denoted by the Wigner–
Seitz radius in atomic units, i.e., in multiples of the Bohr radius (after [18]).
background, (
n̄, z ≤ 0
n+ (r) = . (16)
0, z>0
Here z denotes, as usual, the direction perpendicular to the surface. The bulk charge density in
the jellium model is commonly specified by the Wigner–Seitz radius
3 1/3

rs = , (17)
4πn
which corresponds to the radius of the sphere whose volume V /N = 1/n equals the volume per
electron in the homogeneous electron gas.
Using DFT, the charge density within the jellium model has been determined for several different
different bulk charge densities [18]. Two examples corresponding to a high-density (rs = 2) and
a low-density metal (rs = 5) are shown in Fig. 4. The electron distribution does not follow the
sharp edge of the positive background but rather exhibits a damped oscillatory structure inside the
jellium which is stronger in the low-density case. These Friedel oscillations are a consequence of
the sharp edge of the background density in the jellium model [1]. Furthermore, it is obvious that
some electronic charge density spills out into vacuum, thereby creating a dipole layer. This dipole
layer contributes to the so-called work function of the metal Φ which is defined as the minimum
work that must be done to remove an electron from a solid at 0 K.
The jellium model has been used to evaluate the work function of simple and noble metals [19]
which is typically in the range of 2 to 5 eV. For simple sp-bonded metals such as Na, K or Cs
with rather delocalized electron orbitals, the jellium model is surprisingly successful, however, for
metals with d electrons which are much more localized the jellium model is no longer appropriate.
Furthermore, the jellium model can also not describe the occurrence of surface states which are
electronic states localized near the surface. These states are a consequence of a band gap in the
projected bulk band structure which allows for metal states that are delocalized parallel to the
surface but decay exponentially into the bulk.
3-8 RTO-EN-AVT-142
111111111111
000000000000
000000000000
111111111111
4
Layer 1 111111111111
000000000000
000000000000
111111111111
3 000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
2 000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
LDOS (states/eV)
1 000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
4
3 Layer 2111111111111
000000000000
000000000000
111111111111
111111111111
000000000000
000000000000
111111111111
2 000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
1 000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
4
Layer 3 111111111111111111111111
000000000000000000000000
3
111111111111111111111111
000000000000000000000000
000000000000000000000000
111111111111111111111111
2
1
000000000000000000000000
111111111111111111111111
000000000000000000000000
111111111111111111111111
0
000000000000000000000000
111111111111111111111111
000000000000000000000000
111111111111111111111111
−8 −6 −4 −2 0
Energy E − EF (eV)
Figure 5: Layer-resolved, local d-band density of states of Pd(210) determined by GGA-DFT

calculations. The Fermi level and the center of the d-band are indicated by vertical lines. The
third-layer PDOS is already very close to the bulk density of states of palladium (after [23]).
Furthermore, at a surface the local density of states (LDOS)

X
n(r, ε) = |φi (r)|2 δ(ε − εi ) . (18)
i
is also modified compared to the bulk. In a simple tight-binding picture, the width of an electronic
band is directly related to the coordination and the overlap of the orbitals. Surface atoms have
a lower coordination than bulk atoms, hence the local band width will become narrower. In fact,
the same is also true for pseudomorphic overlayers under tensile strain which reduces the overlap
between the electronic orbitals [20, 21, 22]). If the band is more than half-filled but not completely
filled and the band center is kept fixed, the number of states below the Fermi energy will increase.
This would lead to an higher occupation of the band. However, the number of electrons is conserved.
In order to obey charge conservation, the narrower band has to shift up so that the number of
occupied states remains unchanged.
This effect is illustrated in Fig. 5 where the results of GGA-DFT calculations concerning the layer-
resolved, local d-band density of states of the stepped Pd(210) surface are shown [23]. The LDOS
of the third layer is still rather similar to the Pd bulk density of states. This is due to the good
screening properties of metals [6] which lead to a rapid recovery of bulk properties in the vicinity
of imperfections such as surfaces. However, the width of the d-band of the second and first layer
are significantly reduced, and this reduction in band width is accompanied by an upshift of the
d-band centers indicated by the vertical dashed lines.
Such an upshift of the d-band has significant consequences for the reactivity which can be un-
derstood within the so-called d-band model [24, 25]. This scheme is closely related to the frontier
orbital concept developed for gas-phase reactions [26, 27]. In the d-band model, the whole d-band
RTO-EN-AVT-142 3-9
3
fcc(111) b) (111) e)
Surface energy (eV/atom)
a) 2
fcc(100)
fcc(110)
[001] direction (Å)

2 0
d12
c) (100)
−2
d23
1 −4
d34
d) (110) −6
−4 −2 0 2 4
0
Y Zr Nb Mo Tc Ru Rh Pd Ag [100] direction (Å)
Figure 6: First-principles surface energies for the 4d transition metals calculated using density
functional theory within the local density approximation [32]. The energies have been determined
for the (111), (100) and (110) surfaces in the fcc structure even for the hcp metals Y, Zr, Tc and
Ru and for the bcc metals Nb and Mo. (a) surface energy in eV/atom, (b-d) structure of the fcc
(111), (100) and (110) surface, respectively; (e) electron density of a Cu(001) surface along a (010)
plane calculated by DFT-GGA calculations [1] together with the notation of the layer spacing dij .
is replaced by an effective level located at the center of the d-band εd . This level then interacts
with the molecular orbitals of an adsorbate. If the interaction is sufficiently strong, the molecular
levels split because of the strong hybridization with the metal d-states into a bonding and an
anti-bonding contribution. The closer the d-band center is to the Fermi energy, the more reac-
tive the system usually is because the less any anti-bonding contribution is occupied. If relatively
similar systems are compared which only differ in the position of the d-band center, to first order
a linear relationship between the d-band center shift and the change in the chemisorption strength
∆Ed [28, 29] results,
V2
δEd = − δεd , (19)
|d − a |2
which means that an upshift of the d-band leads to a stronger interaction or larger energy gain.
This explains why low-coordinated sites such as for example step sites can exhibit a significantly
higher reactivity than flat terrace sites [30, 31].
Energetically, however, stepped metal surfaces are unfavorable. In fact, the more densely packed
a certain lattice plane and the higher coordinated the atoms in that plane, the less bonds have to
be broken upon cleavage. Hence the most densely packed surface should have the lowest surface
energy. This is indeed the case for almost all 3d, 4d and 5d transition metals. We have illustrated
this trend in Fig. 6 where calculated surface energies for the fcc(111), (100) and (110) termination of
the 4d transition metals are plotted in eV/atom [32] together with an illustration of their structure.
The densely packed (111) surfaces with their ninefold coordinated surface atoms have the lowest
surface energy per atom, i.e. they correspond to the most stable structure, while the more open
(100) surfaces (eightfold coordination) and the (110) (sixfold coordination) are less stable.
Figure 6 also indicates the chemical trend in the surface energies. There is a parabolic shape of the
3 - 10 RTO-EN-AVT-142
surface energies as a function of the d-band occupation. In order to make this dependence more
obvious, all surface energies have been calculated for the fcc structure, even for the hcp metals
Y, Zr, Tc and Ru and for the bcc metals Nb and Mo. The surface energies are maximal for a
half-filled d-band, while they are minimal for either an empty or a completely filled d-band. The
same trend is also observed for the 3d and the 5d transition metals [33] and is already well-known
for the cohesive energies [6]. The relationship between the cohesive and the surface energies can
be made quantitative within the so-called bond-cutting model which takes into account that the
bond strength varies with the coordination number. For a low-coordinated atom the single bonds
are stronger than for a high-coordinated atom.
In a simple tight-binding picture, the energy per bond can be assumed to scale with (Nc )1/2 . If we
denote the surface energy per atom by σ, we can estimate it by
q
Ncbulk − Ncsurf 0
p
σ = p Ecoh , (20)
Ncbulk
where Ncbulk and Ncsurf are the coordination number of the bulk and the surface, respectively,
0
and Ecoh is the cohesive energy related to a non-magnetic atom for a non-magnetic surface. For
0
a fcc(111) surface Eq. (20) yields a surface energy per atom of σ = 0.134 Ecoh which gives results
rather close to the ones plotted in Fig. 6a.
The electron density at a metal surface is rather smooth, as illustrated in Fig. 6e. This is due to
the fact that at the surface the electrons are free to lower their kinetic energy by becoming more
uniformly distributed which results in the so-called Smoluchowski smoothing [34]. In addition, in
Fig. 6e the notation of the layer spacing dij is introduced. For an ideal surface, the layer spacing
would be the same as the bulk layer spacing. Real surfaces, however, show relaxation effect: due to
the modified environment compared to the bulk, the layer spacing is changed. Still the changes are
rather small for most metal surfaces, as is illustrated in table 1 for the low-index copper surfaces.
At metal surfaces, the smoothening of the electron density usually leads to a contractive relaxation
of the first layer. For the densely packed (111) and (100) Cu surfaces, this contraction is negligible,
but for the more open surfaces (see Cu(110)) it can be already rather significant. Furthermore,
Table 1: Surface energies γ and relaxations of the uppermost layer of various Cu surfaces. The
relaxations are given in percent relative to the bulk layer spacing d0 (hkl).
Surface Method γ (J/m2 ) ∆d12 ∆d23 ∆d34 d0 (hkl) (Å)
Cu(111) Theorya 1.30 -0.9 -0.3 2.10

Cu(111) Exp. ∼1.79b -0.7c
Cu(100) Theorya 1.45 -2.6 1.5 1.821
Cu(100) Exp. ∼1.79b -2.1d 0.4d 0.1d 1.807
Cu(110) Theorye 1.53 -10.8 5.3 0.1 1.29
Cu(110) Exp. ∼1.79b -8.5f 2.3f
References: a) [35], b) [36], c) [37], d) [38], e) [39], f) [40],
RTO-EN-AVT-142 3 - 11
Figure 7: Structural models for the GaAs(100) surface. Dark, filled circles and bright, empty circles
represent As and Ga atoms, respectively. (a) Ball and stick model of the ideal (1×1) As-terminated
GaAs(100) surface. (b) Schematic top and side view of the β2(2 × 4) reconstruction. Larger circles
correspond to atoms closer to the surface.
many metal substrates respond to the contraction of the first interlayer spacing by an expansion
of the second interlayer, as for example the Cu(100) and (110) surfaces. However, this oscillatory
behavior does not necessarily occur.
In addition, in table 1 the surface energies are included. The experimental values are derived
from surface tension measurements which are made in the liquid phase and extrapolated to zero
temperature [36]. This does also mean that these surface energies are not related to any particular
crystal face, but it is obvious that the measured surface energy has the same order of magnitude as
the calculate ones. Furthermore, for the low-index surfaces of Cu the trend γ(111) < γ(100) < γ(110)
is confirmed.
While for the transition metals the modified bond-cutting model Eq. (20) gives a reliable estimate
of the surface energies, the situation for the divalent fcc and bcc sp-metals is not that simple.
For metals such as Ca, Sr or Ba the surface energies of the second most close packed surface are
consistently smaller than the ones of the most closed packed surface [33]. Even more complex is
the situation for semiconductor surfaces. Truly directional bonds between atoms will be broken
upon cleavage which creates an highly unstable state. The surface will try to minimize the number
of unsaturated bonds, the so-called dangling bonds. A prominent example of the resulting surface
reconstruction is provided by the GaAs(100) surface.
The evaluation of the most stable surface structure for compound materials is more complex than
for elemental materials, because the relevant entity is no longer the total energy but rather the
free energy which at zero temperature is given by
1 X
γ = (Esurf − µi Ni ) . (21)
A
i
The chemical potentials µ reflect the specific thermodynamic conditions of the preparation of the
surface, i.e., the reservoir with which the atoms are exchanged in a structural transition. For gas-
3 - 12 RTO-EN-AVT-142
8
a) b)
O3
Al
lO
3
Surface energy (J/m )
Al
2
lA
6
A
OA
lAl 2 lAl
O 1A
4
AlO3Al
2
0
-6 -5 -4 -3 -2 -1 0
O
Al rich Chemical potential µO - µO2 (eV) O rich
Figure 8: Al2 O3 (0001) surface. a) Side view of the ideally Al-terminated surface structure showing
the Al-O-Al trilayer structure. b) Surface energies of different Al2 O3 (0001) (1 × 1) structure in
J/m2 as a function of the difference of the oxygen chemical potential (after [43]). The uppermost
layers of the corresponding structures are indicated in the figure. The perpendicular lines indicate
the range of possible oxygen chemical potentials.
phase deposition, they can be directly related to the partial pressure at a certain temperature.
Esurf is the total energy of the surface per unit cell. As far as the structure of GaAs(100) is
concerned, the ideal surface structure (Fig. 7a) is highly unstable because of the large number of
broken bonds. The GaAs surface tries to minimize the number of dangling bonds by dimerization.
Still, at the dimer atoms of a (100) surface one dangling bond per atom remains. In general, polar
semiconductors exhibit surface reconstructions with the anion dangling bonds filled and the cation
dangling bonds empty. This so-called electron-counting principle leads to a semiconducting surface
since the occupation of Ga dangling bonds which are higher than As dangling bonds would result
in a metallic surface which is energetically unfavorable. The β2(2×4) surface structure (see Fig. 7b)
fulfilling these requirements is stable over a wide range of As chemical potentials [41, 42].
Finally the structure of ionic surfaces shall be addressed. Typically alkali halide crystals in equilib-
rium are almost ideally terminated by non-polar surfaces without any dipole moment perpendicular
to the surface. A typical example is the {100} surface of the sodium chloride structure. Since the
formation of a dipole layer is energetically rather costly, polar surfaces are usually highly unsta-
ble. Because of that reason salt grains have an almost perfect cubic shape. Surface structures are
more complicated for insulating oxide materials where the bonds still have a covalent character
although there is a significant charge transfer between the constituents. As an example, we consider
the (0001) surface of α-Al2 O3 (corundum or sapphire) which has been studied extensively by both
theory [43, 44, 45, 46] and experiment [47, 48].
α-Al2 O3 (sapphire) crystallizes in the corundum structure that can be described by a primitive
rhombohedral unit cell with two Al2 O3 formula units, but more convenient is the hexagonal unit
cell that contains 12 Al atoms and 18 O atoms. The hexagonal cell corresponds to a layered
structure with six oxygen planes associated with aluminum planes above and below it, forming
RTO-EN-AVT-142 3 - 13
stoichiometric triple Al-O-Al layers with three O atoms and just one Al atom in each layer per
unit cell, as can be seen in a side view of the ideally Al-terminated α-Al2 O3 (0001) surface shown in
Fig. 8a. Because of this layered structure, there is no non-polar termination of the α-Al2 O3 (0001)
surface.
Figure 8b shows the calculated surface energies as a function of the oxygen chemical potential for
different (1×1) terminations [43, 46]. These surface free energies have been determined using Eq. 21.
In fact, the total energy of every surface structure has to be calculated only once, and the slope
of the curves is given by the stoichiometry of the surface with respect to the bulk stoichiometry.
The two perpendicular lines indicate the range of possible oxygen chemical potentials. For larger
oxygen chemical potentials molecular oxygen is more stable, for smaller chemical potentials bulk
Al becomes the preferred phase. Within the range of possible chemical potentials the structure
with the lowest free energy is the one that is stable. According Figure 8b, over the entire range
of oxygen chemical potentials the stoichiometric AlO3 Al-termination is by far the energetically
most favorable one. This can be understood by simple electrostatic arguments because the triple
AlO3 Al layer does not have a dipole moment while all other (1 × 1) surface terminations have one.
However, it should be noted that there is a strong inward relaxation of the first Al-layer so that
it practically becomes coplanar with the oxygen layer [43, 49]. In contrast, the interlayer spacing
between the second and third layer is only slightly modified compared to the bulk spacing.
5 Interaction of atoms and molecules with surfaces
The interaction of atoms and molecules with surfaces is ubiquitous since all materials interact with
their environment via their surfaces. Certain chemical reactions can be significantly accelerated
at surfaces which is used in heterogeneous catalysis. However, for many devices and applications
any interaction with the environment should be rather avoided. One example is the high energetic
exchange between the atmosphere and rockets or high-speed missiles which causes significant tem-
perature rises that lead to chemical reactions and particle ionizations. Hence a deeper knowledge
about the principles underlying gas-surface interactions and adsorption is not only of fundamental
interest, but it is also technologically very relevant.
Basically one distinguishes two kinds of adsorption at surfaces. If true chemical bonds between
substrate and adsorbate are formed, it is called chemisorption, otherwise it is called physisorption
which is usually much weaker. We will start by discussing the nature of physisorption.
Any molecule in front of a surface experiences an attraction towards the surface caused by the van
der Waals forces, i.e., the attraction is due to the induced dipole moment interacting with its own
image charges in the polarizable solid. Although this interaction is usually rather weak (∼ 0.1 eV),
it is in fact crucial for the bonding in a wide range of matter. Although the van der Waals interaction
is a quantum phenomenon, some basics about this interaction can be understood within a purely
classical picture. In Fig. 9, a hydrogen atom in front of a metal surface is schematically drawn.
This hydrogen atom is interacting with its image charges of both the nucleus and the electron in
the conductor. If one adds up the electrostatic interaction of all charges and performs a Taylor
expansion in the small parameter |r|/|R|, where |R| = Z and |r| are the atom-surface and the
3 - 14 RTO-EN-AVT-142
e+ e−
r’
r
− −R R +
metal vacuum
Figure 9: Van der Waals interaction: schematic illustration of a hydrogen atom in front of a perfect
conductor interacting with its image charges.
proton-electron distance, respectively, one obtains
e2 x2 + y 2 3e2 h z 2
i
Vim = − 3 2
+z + (x + y 2
) + z 2
+ O(Z −5 ) . (22)
8Z 2 16Z 4 2
Here, x, y and z are the components of r. Introducing the atomic polarizability
e2
α= 2 . (23)
me ωvib
and the van der Waals constant Cv = ~ωvib α/8, the electrostatic interaction can be written as
Cv 3Cv Z0 Cv
Vim (Z) = − 3
− 4
+ O(Z −5 ) = − + O(Z −5 ) (24)
Z Z (Z − Z0 )3
where Z0 is the so-called dynamical image plane. This expression confirms the long-range nature
of the van der Waals interaction which is proportional to Z −3 . In this derivation, we have treated
the hydrogen atom as a dipole interacting with its image dipole at a distance 2Z. However, a
hydrogen atom in the ground state has no permanent dipole moment. Therefore a rigorous quantum
mechanical derivation of the long-range interaction between a neutral atom and a solid surface is
necessary. Still, such a quantum treatment results in an analogous expression as Eq. (24) with the
van der Waals constant also basically proportional to the polarizability of the atom [50].
The attractive van der Waals force is always present for any molecule interacting with surfaces,
but if close to the surface true chemical bonds are formed, then the additional, relatively weak
van der Waals interaction is negligible. Only for inert atoms and molecules with closed valence
shells such as the noble gas atoms, the van der Waals interaction is the main source of binding to
the surface. Close to the surface, the molecular wave functions start to overlap with the substrate
orbitals. Because of the Pauli principle the wave functions have to be orthogonal to each other
leading to the short-range Pauli repulsion. The balance between the short-range Pauli repulsion
and the long-range van der Waals attraction leads to the existence of a physisorption minimum.
Figure 10 shows the physisorption potential for He interaction with jellium surfaces with densities
corresponding to Ag, Cu and Au where the Pauli repulsion has been evaluated using Hartree–Fock
RTO-EN-AVT-142 3 - 15
10
8
6
Li
Energy (meV)
4
2
0
Ag
-2 He/Ag
He/Cu
-4
He/Au Cu
-6 He/Li
Au
-8
0 1 2 3 4 5 6
Distance from the surface z(Å)
Figure 10: Interaction potential of He with different jellium surfaces as a function of the distance
from the jellium edge. The jellium electronic densities correspond to the noble metals Ag, Cu and
Au and the simple metal Li, respectively (after [51]).
theory [51]. It is obvious that the physisorption minimum for He is rather far away from the surface
(>
∼ 3 Å) and shallow (below 10 meV).
In contrast to physisorption, chemisorption corresponds to the creation of a true chemical bond
between adsorbate and substrate. This means that the electronic structure of both the substrate
and adsorbate are strongly perturbed by the interaction. As far as the energy gain upon adsorption
is concerned, the adsorption or bonding energy, there can be a delicate balancing between different
opposing effects so that it is not always easy to develop a true understanding. We will discuss here
some basic mechanisms, namely ionic and covalent bonding. Again we will use DFT calculations
addressing adsorption on a jellium surface in order to illustrate the main trends.
In order to identify the energetic contributions to chemisorption within density functional theory,
we regroup the different energetic terms in the total energy expression (15) yielding
XN Z
Etot = εi + Exc [n] − vxc (r)n(r)d3 r − VH + Vnucl−nucl
i=1
XN Z
= εi + Exc [n] − veff (r)n(r)d3 r + Ees . (25)
i=1
Here we have included the interaction between the nuclei or ion cores Vnucl−nucl (see Eq. 3) which
acts as a renormalization constant in any electronic structure
P calculation so that Ees corresponds
to the total electrostatic energy of the system. The sum εi = Ebs is the band-structure term.
The adsorption energy is given by the energy difference between the energies of the separate
constituents and the interacting system
Eads = (Etot (substrate) + Etot (atom)) − Etot (adatom/substrate) . (26)
Here the adsorption energy is positive if the adsorption is stable. However, there is no consis-
tency in the literature as far as the sign of the adsorption energy is concerned. There are several
3 - 16 RTO-EN-AVT-142
Lithium Silicon Chlorine H2/H/Pd(210)

3
Charge density
-
2
Surface distance z @ÅD

Li Si Cl +
1
1Å -
+ +
0
-
Charge density
difference
-1
1
0 11
00
-2
-3
-2 -1 0 1 2
Surface distance
a) x @ÅD b)
Figure 11: Charge density and charge density difference plots of chemisorption systems calculated
with DFT methods. a) Contours of constant charge density calculated for Li, Si and Cl adsorbed
on a high-density jellium substrate. The solid vertical line indicates the jellium edge. Upper panel:
Total charge density of states; lower panel: charge density difference, broken lines correspond to
charge depletion (after [52]). b) Charge density difference plot of H2 adsorbed on the top site above
a Pd atom of a H(1×1)/Pd(210) surface. Regions of charge depletion are denoted by a minus sign
(after [23]).
terms contributing to the total energy and consequently to the energy difference. Most often, the
chemical interaction is analyzed in terms of the electronic band structure term and electrostatic
contributions.
The upper panel of Fig. 11a shows contours of constant charge density calculated for Li, Si and
Cl adsorbed on a high-density jellium substrate representative for Al[52]. Although there are some
differences, these plots still look rather similar for the three adsorbates. It is much more instructive
to look at the charge density difference between the interacting system and the superposition of
the bare atom and the substrate which corresponds to the adsorption induced charge density
redistribution and rehybridization. These plots exhibit regions of charge depletion and charge
accumulation indicating the charge transfer upon the adsorption.
The lower panel of Fig. 11a illustrates that there is charge transfer from the vacuum side of the
Li adatom towards the metal while for Cl there is a significant charge transfer from the substrate
to the adsorbate. This charge flow is caused by the difference in the electronegativity between the
metal and the adsorbates leading to positive ionic chemisorption in the case of Li and to negative
ionic chemisorption in the case of Cl. The adsorbed silicon atom, on the other hand, shows charge
transfer from the region close to the nucleus to both the vacuum and the bond region. Such a
charge accumulation in the bond region is typical for the formation of a covalent bond with the
maximum charge density in between the two bonding partners.
For realistic systems, charge density difference plots can reveal much more details of the interaction
between substrate and adsorbate. Figure Fig. 11b shows the adsorption induced charge redistri-
bution upon the adsorption of molecular hydrogen on top of a Pd atom of a hydrogen covered
RTO-EN-AVT-142 3 - 17
Change in state density (arb. units)

Cl Cl 3p Si 3p Li 2s
1.0 Si
Li
0.8
anti-bonding s
bonding s
0.6 Si 3s
0.4
0.2
0.0
-15 -10 -5 εF 0
Energy relative to vacuum (eV)
Figure 12: Change of the density of states upon the adsorption of Cl, Si and Li on jellium with an
electron density corresponding to Al (after [52]).
H(1×1)/Pd(210) surface [23, 53]. The main bonding occurs via the d3z 2 −r2 orbital of the Pd atom
which can already be deduced from the pattern of the charge redistribution. Furthermore, the
strong polarization of the adsorbed H2 molecule is visible leading to a significant reduction of the
work function [23, 53].
In addition to the spatial information about the charge redistribution upon adsorption, the change
in the density of states can give insight into the delicate energetic balance between band-structure
and electrostatic contributions to the chemical bonding. The change of the density of states upon
the adsorption of Li, Si and Cl is plotted in Fig. 12. There are several peaks that correspond
to adatom levels which have been shifted and broadened due to the interaction with the jellium
substrate.
The Cl 3p derived state is basically fully occupied since it is almost entirely below εF indicating the
negative ionic chemisorption, whereas the Li 2s derived state which is singly occupied in the free
atom lies primarily above the Fermi energy εF . This confirms the charge transfer from the Li atom
to the substrate and hence the positive ionic chemisorption. The density of states of Si adsorbed
on jellium shows two prominent peaks which can be associated with the Si 3s and 3p atomic levels.
The Si 3p derived state is only half-filled. Typically the lower part of such a resonance adds charge
to the bond region while the upper parts substract charge from this region [1]. The lower parts can
therefore be associated with a bonding contribution while the upper parts have an antibonding
character. Hence a half-filled resonance level corresponds to a covalent bonding interaction in weak
chemisorption cases. These considerations also explain why d-band metals with a half-filled d band
show the largest cohesive and surface energies (see Fig. 6). For Cl adsorption, both the bonding
and antibonding contributions are occupied. In this case it is the electrostatic attraction between
the Cl core and the transferred electron that stabilizes the adsorption.
In order to illustrate the possible complexity of the molecule-surface interaction, we will use the
dissociative adsorption of methanol on Cu(110) as an example [54]. The interaction of methanol
3 - 18 RTO-EN-AVT-142
Clean surface Cu d-structure

Surface Cu d-structure
Oxygen p-state (Adsorbate)
LDOS (arb. units)

Oxygen p-state (Gas)
CH3O/Cu(110)
5a1
1e
2e
-10 -8 -6 -4 -2 0
(a) (b) Energy ε - εF (eV)
Figure 13: Dissociative adsorption of methanol (CH3 OH) on a oxygen covered Cu(110) surface. a)
Geometry of neighboring methoxy and hydroxyl after the spontaneous methanol hydroxyl bond
decomposition on O/Cu(110). b) Projected local density of states (LDOS) of the oxygen atom of
methoxy on Cu(110). The methoxy orbitals are strongly perturbed due to the interaction with the
Cu substrate (after [54]).
with copper surfaces is of strong current interest in the context of the catalytic oxidation and
synthesis of methanol. On oxygen-covered Cu(110), methanol (CH3 OH) dissociates spontaneously
into methoxy (CH3 O) and hydroxyl (OH, see Fig. 13a). The methoxy radical is strongly interacting
with the copper surface. The 2e orbital which corresponds to a nonbonding π orbital on oxygen
is only partially filled in the gas-phase [55]. As Fig. 13b shows, this 2e orbital is significantly
broadened by the direct coupling to the Cu d-states which causes the high adsorption energy. The
oxygen px and py (1e) states are no longer degenerate, but split by the reduced symmetry at the
adsorption site. Furthermore, the pz (5a1 ) orbital is shifted down by about 2 eV between the two
1e-derived peaks.
Such a significant rearrangement of the electronic structure is always indicative of a strong inter-
action between adsorbate and substrate. This is also reflected in the relatively large adsorption
energy of almost 3 eV [54]. However, a strong interaction is not always associated with a large bind-
ing energy. For example, formaldehyde (CH2 O) adsorbed above the short-bridge site of Cu(110)
also exhibits a considerable rehybridization of its levels upon adsorption. Still, it is only bound by
0.6 eV to Cu(110). This is due to the fact that the adsorbed formaldehyde is significantly distorted
so that the energy gain upon adsorption is compensated to a large extent by the energetic cost of
the deformation of the molecule [54].
6 Conclusions
In this short introduction into theoretical surface science I have given a brief overview over the main
theoretical tools to address surface problems from a microscopic point of view. Some illustrative
examples have been used in order to discuss the main mechanisms underlying the structure of
RTO-EN-AVT-142 3 - 19
surfaces and the interaction of atoms and molecules with surfaces. Of course, surface science
is a much broader field that also covers, e.g., phase transitions at surfaces or the dynamics and
kinetics of gas-surface interactions and reactions at surfaces. A decription of these processes requires
further theoretical methods and tools. However, the combination of methods such as molecular
dynamics simulations or thermodynamical approaches with information from electronic structure
calculations permits to address complex stuctures and processes at surfaces from first principles,
even on rather long length and time scales. This makes theoretical surface science to a very active
and fruitful research field in close collaboration with the experiment allowing detailed insights into
the microscopic world at surfaces.
References
[1] A. Groß, Theoretical surface science – A microscopic perspective, Springer, Berlin, 2002.
[2] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel, Phys. Rev. Lett. 49, 57 (1982).
[3] A. Groß, Surf. Sci. 500, 347 (2002).
[4] M. Born and J. R. Oppenheimer, Ann. Phys. 84, 457 (1927).
[5] W. Kohn, Rev. Mod. Phys. 71, 1253 (1999).
[6] N. W. Ashcroft and N. D. Mermin, Solid State Physics, Saunders College, Philadelphia, 1976.
[7] C. Kittel, Introduction to Solid State Physics, John Wiley & Sons, New York, sixth edition,
1986.
[8] J. L. Whitten and H. Yang, Surf. Sci. Rep. 24, 55 (1996).
[9] P. Hohenberg and W. Kohn, Phys. Rev. 136, B864 (1964).
[10] W. Kohn and L. Sham, Phys. Rev. 140, A1133 (1965).
[11] E. H. Lieb, Rev. Mod. Phys. 53, 603 (1981).
[12] B. J. Zhou, V. L. Ligneres, and E. A. Carter, J. Chem. Phys. 122, 044103 (2005).
[13] M. C. Payne, M. P. Teter, D. C. Allan, T. A. Arias, and J. D. Joannopoulos, Rev. Mod. Phys.
64, 1045 (1992).
[14] B. Hammer, M. Scheffler, K. Jacobsen, and J. Nørskov, Phys. Rev. Lett. 73, 1400 (1994).
[15] J. P. Perdew et al., Phys. Rev. B 46, 6671 (1992).
[16] B. Hammer, L. B. Hansen, and J. K. Nørskov, Phys. Rev. B 59, 7413 (1999).
[17] P. J. Feibelman et al., J. Phys. Chem. B 105, 4018 (2001).
[18] N. D. Lang and W. Kohn, Phys. Rev. B 1, 4555 (1970).
[19] N. D. Lang and W. Kohn, Phys. Rev. B 3, 1215 (1971).
3 - 20 RTO-EN-AVT-142
[20] M. Mavrikakis, B. Hammer, and J. K. Nørskov, Phys. Rev. Lett. 81, 2819 (1998).
[21] A. Roudgar and A. Groß, Phys. Rev. B 67, 033409 (2003).
[22] A. Roudgar and A. Groß, J. Electroanal. Chem. 548, 121 (2003).
[23] M. Lischka and A. Groß, Phys. Rev. B 65, 075420 (2002).
[24] B. Hammer and J. K. Nørskov, Surf. Sci. 343, 211 (1995).
[25] B. Hammer and J. K. Nørskov, Nature 376, 238 (1995).
[26] K. Fukui, Science 218, 747 (1982).
[27] R. Hoffmann, Rev. Mod. Phys. 60, 601 (1988).
[28] B. Hammer, O. H. Nielsen, and J. K. Nørskov, Catal. Lett. 46, 31 (1997).
[29] V. Pallassana, M. Neurock, L. B. Hansen, B. Hammer, and J. K. Nørskov, Phys. Rev. B 60,
6146 (1999).
[30] S. Dahl et al., Phys. Rev. Lett. 83, 1814 (1999).
[31] S. Dahl, E. Törnqvist, and I. Chorkendorff, J. Catal. 192, 381 (2000).
[32] M. Methfessel, D. Hennig, and M. Scheffler, Phys. Rev. B 46, 4816 (1992).
[33] L. Vitos, A. V. Ruban, H. L. Skriver, and J. Kollár, Surf. Sci. 411, 186 (1998).
[34] R. Smoluchowski, Phys. Rev. 60, 661 (1941).
[35] S. Sakong and A. Groß, unpublished.
[36] W. R. Tyson and W. A. Miller, Surf. Sci. 62, 267 (1977).
[37] S. Lindgren, L. Walldén, J. Rundgren, and P. Westrin, Phys. Rev. B 29, 576 (1984).
[38] S. Walter, V. Blum, L. Hammer, K. Heinz, and M. Giesen, Surf. Sci. 458, 155 (2000).
[39] S. Liem, G. Kresse, and J. Clarke, Surf. Sci. 415, 194 (1998).
[40] D. L. Adams, H. B. Nielsen, and J. N. Andersen, Surf. Sci. 128, 294 (1983).
[41] J. Northrup and S. Froyen, Phys. Rev. B 50, 2015 (1994).
[42] N. Moll, A. Kley, E. Pehlke, and M. Scheffler, Phys. Rev. B 54, 8844 (1996).
[43] X. G. Wang, A. Chaka, and M. Scheffler, Phys. Rev. Lett. 84, 3650 (2000).
[44] C. Verdozzi, D. R. Jennison, P. A. Schultz, and M. P. Sears, Phys. Rev. Lett. 82, 799 (1999).
[45] R. Di Felice and J. E. Northrup, Phys. Rev. B 60, R16287 (1999).
[46] I. Batyrev, A. Alavi, and M. W. Finnis, Faraday Discuss. 114, 33 (1999).
RTO-EN-AVT-142 3 - 21
[47] P. Guénard, G. Renaud, A. Barbier, and M. Gautier-Soyer, Surf. Rev. Lett. 5, 321 (1998).
[48] J. Toofan and P. R. Watson, Surf. Sci. 401, 162 (1998).
[49] E. Soares, M. A. Van Hove, C. F. Walters, and K. F. McCarty, Phys. Rev. B 65, 195405
(2002).
[50] E. Zaremba and W. Kohn, Phys. Rev. B 13, 2270 (1976).
[51] E. Zaremba and W. Kohn, Phys. Rev. B 15, 1769 (1977).
[52] N. D. Lang and A. Williams, Phys. Rev. B 18, 616 (1978).
[53] P. K. Schmidt et al., Phys. Rev. Lett. 87, 096103 (2001).
[54] S. Sakong and A. Groß, J. Catal. 231, 420 (2005).
[55] C. F. Jackels, J. Chem. Phys. 82, 311 (1985).
3 - 22 RTO-EN-AVT-142
Simulation of Gas-Surface Dynamical Interactions
Axel Groß
Abteilung Theoretische Chemie
Universität Ulm
D-89069 Ulm
GERMANY
email: axel.gross@uni-ulm.de
Abstract
The interaction of atoms and molecule with surfaces is of great technological relevance. Both
advantageous and harmful processes can occur at surfaces. If an atom or molecule impinges
on a surface, it can either scatter back into the gas phase or become adsorbed on the surface.
Molecules can furthermore undergo chemical reactions at the surface. All these processes are
accompanied by energy transfer between the impinging projectile and the substrate.
The simulation of the dynamics of the gas-surface interaction still represents a considerable
challenge since the coupling of a low-dimensional object, the impinging atom or molecule, to the
substrate with in principle infinitely many degrees of freedom has to be modeled. Furthermore,
depending on the mass of the atom or molecule, quantum effects both in the molecular motion
as well as in the excitation of the substrate have to be taken into account. In this lecture, the
quantum and classical methods required for the simulation of gas-surface dynamical interactions
will be reviewed. Furthermore, the main processes occuring in the interaction of atoms and
molecules with substrates will be illustrated using quantum calculations and classical molecular
dynamics simulations.
1 Introduction
Understanding the interaction of atoms and molecules with surfaces plays an important role in a
wide range of technologically relevant applications [1]. Among those are the heterogenous catalysis
– the majority of reactions in the chemical industry employ catalysts; crystal growth, which
determines, e.g., the quality of semiconductor devices; corrosion and lubrication, which influences
the durability of mechanical systems; or friction, which determines the energy transfer to surfaces,
as for example in the interaction between the atmosphere and rockets or high-speed missiles which
causes significant temperature rises that lead to chemical reactions and particle ionizations.
In order to study the interaction of molecules with surfaces, it is mandatory to have a reliable de-
scription of the interaction energetics. The basic quantity reflecting this interaction is the potential
energy surface (PES) that corresponds to a hyperplane in the multidimensional configuration space.
Groß, A. (2007) Simulation of Gas-Surface Dynamical Interactions. In Experiment, Modeling and Simulation of Gas-Surface Interactions
for Reactive Flows in Hypersonic Flights (pp. 4-1 – 4-26). Educational Notes RTO-EN-AVT-142, Paper 4. Neuilly-sur-Seine, France:
RTO. Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 4-1
Lateral coordinate
Adsorbed reactants Activation barrier
Distance from the surface

on the surface
Gas phase
Reactands in barrier Products in
the gas phase the gas phase
Figure 1: Schematic illustration of the role of a catalyst using a two-dimensional representation

of the potential energy surface. A catalyst provides a detour in the multi-dimensional PES with
lower activation barriers
Nowadays potential energy surfaces can be mapped out in great detail by first-principles electronic
structure calculations, typically based on density functional theory. From the PES, quantities such
as adsorption or binding energies, transition state configurations and energies or minimum energy
paths can be derived.
Potential energy surface can also be used to illustrate processes at surfaces. An example is given in
Fig. 1 explaining the way a catalyst works. The presence of a catalysts lowers the activation barrier
for a particular chemical reaction, however, this involves a detour in the multi-dimensional PES on
the path from the reactants to the products. In heterogeneous catalysis, this usually means that
the reactants have to be adsorbed on the catalyst surface, where the activation barrier is much
smaller than for example in the gas phase. Hence the reaction rate is enormously enhanced in the
presence of a catalyst since the rate depends exponentially on the barrier height.
For a true understanding of processes on and at surfaces, however, the static information from
total-energy calculations is often not sufficient. Furthermore, experiments do usually not yield
direct information about the potential energy surface, but rather determine reaction probabilities
or transfer rates. In order to calculate these quantities and allow for a genuine comparison between
theory and experiment, dynamical or kinetic simulations are required. Unfortunately, processes
such as chemical reactions at surfaces often consist of many elementary steps which are too complex
to be studied as a whole. Therefore in surface science one tries to understand reaction mechanisms
by breaking them up into simpler steps which are then studied under well-defined conditions [2].
In this chapter, I will briefly review the theoretical methods necessary to determine the dynamics
of processes at surfaces. After presenting classical and quantum methods, scattering at surfaces
will be addressed. A substantial fraction of the chapter will then be devoted to the discussion of
atomic, molecular and dissociative adsorption. Finally, the system O2 /Pt(111) will be discussed as
an example, where scattering, molecular and dissociative adsorption can all occur.
4-2 RTO-EN-AVT-142
2 Molecular Dynamics
In order to follow the temporal evolution of a dynamical system, the corresponding quantum or
classical equations of motion have to be solved. They can be derived from the basic equations
describing the interaction of atoms and molecules with each other, namely the non-relativistic
Schrödinger equation on which most of chemistry and solid-state physics is based. It is of the
general form
HΨ({R, r})i = E Ψ({R, r}) . (1)
where the R are the ionic coordinates and r the electronic coordinates. It is well-known that
a complete analytical solution of the Schrödinger equation taking into account both ionic and
electronic degrees of freedom is not possible except for simple cases. One common approach is
to assume that – because of the large mass difference between electrons and the nuclei – the
electrons follow the motion of the nuclei adiabatically. This is the so-called Born-Oppenheimer
approximation [3].
In practice, in the Born-Oppenheimer approximation the nuclear coordinates are fixed so that they
enter the Schrödinger equation no longer as variable but just as parameters. Thus one obtains an
electronic Schrödinger equation
Hel ({R}) ψ({r}) = Eel ({R}) |ψ({r}). (2)
The many-electron ground state energy Eel ({R}) then defines the potential energy surface for the
motion of the nuclei. For extended systems the most efficient approach to determine the many-
electron ground state energy Eel ({R}) from first principles is density functional theory (DFT) [4]
in combination with the supercell concept. Once the electronic ground state energy is obtained, it
can be plugged into the Schrödinger equation for the nuclei,
!
X −~2
∇2 + Eel ({R}) Φ({Rm }) = Enucl Φ({R}), (3)
2Mi Ri
i
where Enucl is now the energy relevant for the dynamics of the nuclei. Alternatively, the potential
energy surface can be used to solve the classical equations of motion,
∂2 ∂
Mi 2
Ri = − Eel ({Rm }) . (4)
∂t ∂Ri
Instead of Newton’s equation of motion, also Hamilton’s equation of motion
∂H ∂H
q̇ = ṗ = − . (5)
∂p ∂q
can be solved.
The solution of the equations of motion can be obtained by standard numerical integration schemes
like Runge–Kutta, Bulirsch–Stoer or predictor-corrector methods (see, e.g., [5]). Very often the
rather simple Verlet algorithm [6, 7] is used which is easily derived from a Taylor expansion of the
RTO-EN-AVT-142 4-3
trajectory.
h2 d2 ri h3 d3 ri

dri
ri (t + h) = ri (t) + h + + + ...
dt h=0 2 dt2 h=0 6 dt3 h=0
h2 Fi (t) h3 d3 ri

= ri (t) + h vi (t) + + + ... (6)
2 m 6 dt3 h=0
Here we have introduced the velocity vi = ṙi . Furthermore, we have used Newton’s equation of
motion to include the force Fi acting on the i-th particle. Analogously we can derive
h2 Fi (t) h3 d3 ri

ri (t − h) = ri (t) − h vi (t) + − + ... (7)
2 m 6 dt3 h=0
Adding (6) and (7) yields the Verlet algorithm [6]
Fi (t)
ri (t + h) = 2ri (t) − ri (t − h) + h2 + O(h4 ) . (8)
m
The accuracy of the numerical integration of the equation of motion can be checked by testing the
energy conservation. In order to evaluate the kinetic energies, the velocities at time t are needed.
Note that they do not explicitly appear in Eq. (8). They can be estimated by
ri (t + h) − ri (t − h)
vi (t) = . (9)
2h
However, the kinetic energy evaluated with Eq. (9) belongs to the time step prior to the one
used for the positions (8) which enter the evaluation of the potential energy. This problem can be
avoided in the so-called velocity Verlet algorithm [7]
h2 Fi (t)
ri (t + h) = ri (t) + h vi (t) +
2 m
Fi (t + h) + Fi (t)
vi (t + h) = vi (t) + h , (10)
2m
which is mathematically equivalent to the Verlet algorithm.
In Fig. 2, various processes that can occur when atoms or molecules are impinging on a surface
are illustrated. For all atoms that are heavier than hydrogen or helium, the quantum effects in the
dynamics are often negligible [8]. Hence molecular dynamics simulations involving the solution of
the classical equations of motion are an appropriate tool to determine the time evolution of most
chemical systems. Even if hydrogen is contained in the system as is generally the case for organic
molecules, still the results of classical dynamical calculations might be meaningful qualitatively
or semi-quantitatively. However, for a real quantitative description, a quantum treatment is nec-
essary. Furthermore, there are certain phenomena in the gas-surface interaction that can only be
understood within a quantum framework, namely elastic scattering and diffraction.
There are two ways to determine quantum mechanical reaction probabilities: by solving the time-
dependent or the time-independent Schrödinger equation. Both approaches are equivalent [9] and
should give the same results. The answer to the question which method is more appropriate depends
4-4 RTO-EN-AVT-142
Diffraction Imn
In
co
m
Specular I oo
in
g
be
am
Inelastic
(Dissociative) adsorption Selective adsorption
Figure 2: Illustration of the different processes that can occur when atoms or molecules are imping-
ing on a surface. The substrate with lattice constant a is represented in a simple ball and spring
picture.
on the particular problem. Time-independent implementations are usually more restrictive as far
as the form of the potential is concerned, but often the choice of the method is a matter of training
and personal taste.
In the time-dependent or wave-packet formulation, the solution of the time-dependent Schrödinger
equation
∂
i~ Ψ(R, t) = H Ψ(R, t) (11)
∂t
can formally be written as
Ψ(R, t) = e−iHt/~ Ψ(R, t = 0) , (12)
if the potential is time-independent. The most common methods to represent the time-evolution
operator exp(−iHt/~) in the gas-surface dynamics community are the split-operator [10, 11] and
the Chebychev [12] methods. In the split-operator method, the time-evolution operator for small
time steps ∆t is written as
e−iH∆t/~ = e−iK∆t/2~ e−iV ∆t/~ e−iK∆t/2~ + O(∆t3 ) , (13)
where K is the kinetic energy operator and V the potential term. In an alternative approach, the
Chebyshev method, the time-evolution operator is expanded as
jX
max
−iH∆t/~
e = aj (∆t) Tj (H̄) , (14)
j=1
where the Tj are Chebyshev polynomials and H̄ is the Hamiltonian rescaled to have eigenvalues
in the range (−1, 1). Both propagation schemes use the fact that the kinetic energy operator
is diagonal in k-space and the potential is diagonal in real-space. The wave function and the
potential are represented on a numerical grid, and the switching between the k-space and real-
space representations is efficiently done by Fast Fourier Transformations (FFT) [5].
RTO-EN-AVT-142 4-5
In the time-independent formulation, on the other hand, the wave function is usually expanded
in a suitable basis set. This often requires to introduce the concept of reaction path coordinates.
Starting from the time-independent Schrödinger equation
(H − E) Ψ = 0 , (15)
one specific reaction path coordinate s is chosen. Then the kinetic energy operator in this coordinate
is separated yielding
−~2 2
( ∂ + H̃E) Ψ = 0 . (16)
2µ s
Here H̃ is the original Hamiltonian except for the kinetic energy operator in the reaction path
coordinate. Usually the use of curve-linear reaction path coordinates results in a more complicated
expression for the kinetic energy operator involving cross terms, but for the sake of clarity this has
been neglected in Eq. (16). In the coordinates perpendicular to the reaction path coordinate, the
wave function is expanded in some suitable set of basis functions,
X
Ψ = Ψ(s, . . .) = ψn (s) |ni . (17)
n
Here n is a multi-index, and the expansion coefficients ψn (s) are assumed to be a function of the
reaction path coordinate. This expansion of Ψ is inserted in the Schrödinger equation (16), and this
equation is multiplied by hm| which corresponds to performing a multi-dimensional integral. Since
the basis functions |ni are assumed to be independent of s, this yields the so-called coupled-channel
equations
X −~2
( ∂s2 − E) δm,n + hm|H̃|ni ψn (s) = 0 . (18)
n
2µ
Instead of a high-dimensional partial differential equation – the original time-independent Schrödin-
ger equation (15) – there is now a set of coupled ordinary differential equation. Still a straight-
forward numerical integration of the coupled-channel equations leads to instabilities, except for
in simple cases, due to exponentially increasing so-called closed channels. These problems can be
avoided, for example by making the potential stepwise constant so that the wave function can be
analytically propagated [13, 14]
3 Scattering at surfaces
In Fig. 2, possible collision processes in the scattering of atoms and molecules at surfaces are
summarized. A monoenergetic beam of atoms or molecules characterized by the wave vector Ki =
Pi /~ where Pi is the initial momentum of the particles is impinging on a periodic surface with
lattice constant a. Classically there will always be a certain energy transfer from the molecules
to the substrate when the incoming particles hit the surface. Quantum mechanically, however,
there will be a non-vanishing probability for elastic scattering, i.e. with no energy transfer to the
substrate. This probability is given by the so-called Debye-Waller factor.
Furthermore, if the de Broglie wavelength λ = 1/|Ki | of the incident beam is of the order of
the lattice spacing a, quantum effects in the momentum transfer parallel to the surface become
4-6 RTO-EN-AVT-142
important. The periodicity of the substrate leads to the conservation of the quasi-momentum
parallel to the surface which means that the component of the wave vector parallel to the surface
can only be changed by a reciprocal lattice vector of the periodic surface:
k k
Kf = Ki + Gmn , (19)
k k
where Gmn is a reciprocal lattice vector of the periodic surface and Ki and Kf are the initial
and final wave vectors parallel to the surface. As a consequence, diffraction results, i.e., there is
only a discrete number of allowed scattering angles. The intensity of the elastic diffraction peak
k k
mn according to Eq. (19) is denoted by Imn . The scattering peak I00 with Kf = Ki is called the
specular peak. It is important to note that the allowed scattering angles are a consequence of the
surface geometry which means that from the diffraction pattern the periodicity and lattice constant
of the surface can be derived, whereas the intensity of the different diffraction peaks depends on
the particular molecule-surface system. Therefore the coherent scattering of atoms or molecules
from surfaces can be used as a tool for probing surface structures which has first been realized in
1930 [15]. In particular helium atom scattering (HAS) experiments have been carried out to study
the surface crystallography (see, e.g., [16, 17] and references therein).
The probability for elastic scattering, i.e. the Debye-Waller factor, vanishes rapidly with increasing
mass of the impinging molecules. For heavier molecules, predominantly inelastic scattering occurs.
The main source for the energy transfer between the impinging molecules and the substrate is the
excitation and deexcitation of substrate phonons but also electron-hole pairs of the substrate may
be involved. Since phonons also carry momentum, the conservation of quasi-momentum parallel
to the surface is modified to
k k
X
Kf = Ki + Gmn + ±Q , (20)
exch.phon.
where Q is a two-dimensional phonon-momentum vector parallel to the surface. The plus-signs in

the sum correspond to the excitation or emission of a phonon while the minus-signs represent the
deexcitation or absorption of a phonon. The energy balance in phonon-inelastic scattering can be
expressed as
~2 K2f ~2 K2i X
= + ±~ωQ,j , (21)
2M 2M
exch.phon.
where ~ωQ,j corresponds to the energy of the phonon with momentum Q and mode index j.
From the change of the momentum and the energy in single-phonon scattering, the surface phonon
spectrum can be derived which has been extensively done using helium atom scattering as a
probe [16, 18]. The excitation of phonons usually leads to a reduced normal component of the
kinetic energy of the back-scattered atoms or molecules. Thus the reflected beam is shifted in
general to larger angles with respect to the surface normal compared to the angle of incidence.
The resulting supraspecular scattering is indicated in Fig. 2 as the inelastic reflection event.
In addition to diffraction, often resonances in the intensity of the specular peak as a function of
the angle of incidence are observed in the case of the scattering of weakly interacting particles at
smooth surfaces, [19]. These so-called selective adsorption resonances are also indicated in Fig. 2.
They occur when the scattered particle can make a transition into one of the discrete bound state
RTO-EN-AVT-142 4-7
of the adsorption potential [20]. This is only possible if temporarily the motion of the particle
is entirely parallel to the surface. The interference of different possible paths along the surface
causes the resonance effects. Energy and momentum conservation yields the selective adsorption
condition
k
~2 K2i ~2 (Ki + Gmn )2
= − |El | , (22)
2M 2M
where El is a bound level of the adsorption potential. From the scattering resonances, bound
state energies can be obtained using Eq. (22) without any detailed knowledge about the scattering
process.
Since molecules, in contrast to atoms, have also internal degrees of freedoms, namely rotations
and vibrations, additional peaks may appear in the diffraction pattern of molecules. They are due
to the fact that these rotations and vibrations can be excited and de-excited during the collision
process. The total energy balance in the molecular scattering is therefore more complex:
~2 K2f ~2 K2i X
= + ∆Erot + ∆Evib + ±~ωQ,j . (23)
2M 2M
exch.phon.
Since the time-scale of the molecular vibrations is usually much shorter than the scattering time
or the rotational period and therefore the associated excitation energies are much larger, the
excitation of molecular vibrations in molecule-surface scattering is usually negligible, in contrast
to the phonon excitation. Molecular rotations, on the other hand, can be excited rather efficiently
in the scattering at highly corrugated and anisotropic surfaces. This leads to additional peaks in
the diffraction spectrum, the rotationally inelastic diffraction peaks.
Experimentally, rotationally inelastic diffraction of hydrogen molecules has been first observed in
the scattering at inert ionic solids such as MgO [21] or NaF [22]. At metal surfaces with a high
barrier for dissociative adsorption, the molecules are scattered at the tails of the metal electron
density which are usually rather smooth. Hence relatively weak diffraction and hardly any rota-
tionally inelastic diffraction has been observed for, e.g., the scattering of H2 from Cu(001) [23, 24].
This is different for the case of HD scattering, where the displacement of the center of mass from
the center of the charge distribution leads to a strong rotational anisotropy [25].
At reactive surfaces where non-activated adsorption is possible, the scattering occurs rather close
to the surface where the potential energy surface is already strongly corrugated and anisotropic.
For such systems, intensive rotationally inelastic peaks in the diffraction pattern have been pre-
dicted theoretically in six-dimensional quantum dynamical calculations [26] for the scattering of
H2 /Pd(100) where non-activated together with activated pathways to dissociative adsorption ex-
ist. One typical calculated angular distribution of H2 molecules scattered at Pd(100) is shown
in Fig. 3 [26]. The total initial kinetic energy is Ei = 76 meV. The incident parallel momentum
equals 2~G along the h01̄1i direction which corresponds to an incident angle of θi = 32◦ . The
molecules are initially in the rotational ground state ji = 0. Figure 3a shows the so-called in-plane
scattering distribution, i.e. the diffraction peaks in the plane spanned by the wave vector of the
incident beam and the surface normal. The label (m, n) denotes the parallel momentum transfer
∆Gk = (mG, nG). The specular peak is the most pronounced one, but the first order diffraction
peak (10) is only a factor of four smaller. Note that in a typical helium atom scattering experiment
the off-specular peaks are about two orders smaller than the specular peak [16]. This is due to
4-8 RTO-EN-AVT-142
0.3 90
(00)
a) 60
Scattering intensity
Final angle θy ( )
o
0.2 30
0.1 − (00) −30

(20)
∆j=2 (10) ∆j=2

−
(10)
− ∆j=2 −60

− (30)
(40) − −
∆j=2 (50) − ∆j=2 (3
− (10)

(40) 0) b)
0.0 −90
−90 −60 −30

0 30

60

90

−90 −60 −30 0 30 60 90

o
Scattering angle θf ( )
o
Final angle θx ( )
Figure 3: Diffraction spectrum of H2 scattered at Pd(100) for a kinetic energy of 76 meV at

an incidence angle of 32◦ along the [10] direction of the square surface lattice, obtained by six-
dimensional quantum coupled channel calculations. a) In-plane diffraction spectrum where all peaks
have been labeled according to the transition. b) In-plane and out-of-plane diffraction peaks. The
open and filled circles correspond to the rotationally elastic and rotationally inelastic scattering,
respectively, with he radius of the circles being proportional to the logarithm of the scattering
intensity (after [26]).
the fact that the chemically inert helium atoms are scattered at the smooth tails of the surface
electron distribution.
The peaks labeled with ∆j = 2 in Fig. 3a correspond to rotationally inelastic diffraction involving
the rotational excitation j = 0 → 2 summed up over all final azimuthal quantum numbers mj .
Except for the specular peak, the intensities of the rotationally inelastic diffraction peaks are
even larger than the corresponding rotationally elastic diffraction peaks with the same momentum
transfer (m, n). Note that because of the particular conditions with the incident parallel momentum
corresponding to the reciprocal lattice vector Gk = (2G, 0), the rotationally elastic and inelastic
(2̄0) diffraction peaks fall upon each other.
The full diffraction pattern including also the so-called out-of-plane scattering peaks is shown
in Fig. 3b. The open circles represent the rotationally elastic, the filled circles the rotationally
inelastic diffraction peaks. The radii of the circles are proportional to the logarithm of the scattering
intensity. It is obvious that there is a significant fraction of out-of-plane scattering with the sum
of all out-of-plane scattering intensities approximately equal to the sum of all in-plane scattering
intensities. Interestingly, some diffraction peaks with a large parallel momentum transfer still show
substantial intensities. This phenomenon is well known from helium atom scattering and has been
discussed within the concept of so-called rainbow scattering [27].
Experimentally, it is not so easy to measure diffraction patterns at reactive surfaces since during the
experiment, a significant fraction of the impinging molecules remains on the surface which destroys
the ideal periodicity of the surface and therefore also suppress the occurrence of diffraction patterns.
RTO-EN-AVT-142 4-9
Thus the surface has to be kept clean which is achieved by relatively high surface temperatures so
that adsorbates quickly desorb again. High surface temperatures, on the other hand, also smear
out the diffraction pattern. Still rotationally inelastic peaks in addition to rotationally elastic peaks
have been clearly identified in the diffraction pattern of D2 /Ni(110) [28] and D2 /Rh(110) [29].
At reactive surfaces, the particles can off course also adsorb. As it is indicated in Fig. 2, molecules
can adsorb both molecularly which means intact or dissociatively. In the case of the atomic or
molecular adsorption, the particles can only remain trapped at the surface if their initial kinetic is
transfered to the surface and dissipated. For light projectiles, the quantum nature of the substrate
phonons becomes important in the energy transfer process. These topics will be discussed in the
next section.
4 Atomic and molecular adsorption
The sticking or adsorption probability is defined as the fraction of atoms or molecules impinging on
a surface that are not scattered back, i.e. that remain on the surface. It is important to note that in
the case of atomic or molecular adsorption when the molecule stays intact, the particles can only
remain on the surface if they transfer their energy to the substrate. This is similar to gas-phase
reactions where a bond between two reactants can only be formed in a three-body collision where
a third reaction partner has to carry away the energy gained by the reaction unless there are other
dissipation channels such as radiation.
At a surface, there are two main channels for energy dissipation namely phonon and electron-hole
pair excitations. Here we focus on the energy transfer to phonons since they usually represent
the main channel for dissipation [30]. For the explicit evaluation of sticking probabilities, PE () is
defined as the probability that an incoming particle with kinetic energy E transfers the energy
to the surface. Only if the particle transfers more than its initial energy to substrate excitations,
it can remain at the surface. Hence the sticking probability can be expressed as
Z ∞
S(E) = PE () d. (24)
E
In order to discuss the essentials of atomic and molecular adsorption due to the energy transfer
to phonons, we will use the rather simple hard-cube model (HCM) [31, 32]. In this model that is
schematically illustrated in Fig. 4a, the impact of the atom on the surface is treated as a binary
elastic collision between a gas phase atom (mass m) and a substrate atom (mass Mc ) which is
moving freely with a velocity distribution Pc (vc ). Because of the adsorption well of depth Ead , the
particle is accelerated and impinges on the hard cube with a velocity
r
2Ead
vwell = − vg2 + . (25)
m
Due to the simplicity of the hard cube model, it can be solved analytically [32]. Assuming a
weighted Maxwellian velocity distribution for vc , the trapping probability becomes
exp −α2 vlim
2

1 1
Strap (vg ) = + erf(αvlim ) + √ , (26)
2 2 2 παvwell
4 - 10 RTO-EN-AVT-142
vc
vg

∆ vc

Mc m
v’well E ad
Figure 4: Hard cube model. a) Schematic illustration of the model. An atom or molecule with
mass m is impinging in an attractive potential with well depth Ead on a surface modeled by a
cube of effective mass Mc . The surface cube is moving with a velocity vc given by a Maxwellian
distribution. b) Trapping probability as a function of the kinetic energy evaluated according to the
hard cube model Eq. (26) for different adsorption energies Ead , mass ratios µ = m/Mc and surface
temperatures Ts .
p
where α = Mc /2kB Ts , vlim is given by
r
µ+1 2Ead µ − 1
vlim = − vwell , (27)
2 m 2
and µ is the mass ratio µ = m/M .
Typical sticking curves obtained with the hard cube model are shown in Fig. 4b. All curves show
the same behavior, namely a monotonic decrease of the sticking probability as a function of the
kinetic energy. This dependence is a consequence of the fact that the energy transfer to the surface
becomes less efficient at higher kinetic energies. Although more energy is transfered to the surface
at higher kinetic energies, the fraction of particles that loose more energy than their initial kinetic
energy becomes smaller.
The examples shown in Fig. 4b have been chosen in order to illustrate further general trends in
atomic adsorption. If the mass ratio between the impinging molecules and the substrate atoms
increases, i.e. if heavier atoms are hitting the surface, more energy is transfered to the surface
so that the trapping probability increases. If the adsorption well becomes deeper, the impinging
molecules become faster and loose more energy upon impact which also increases the trapping
probability. For an adsorption well of 0.6 eV, the effect of changing the surface temperature is
shown. A higher surface temperature Ts leads to a broader velocity distribution which results in
an averaging over a wider range of kinetic energies. This causes a decrease for negative curvature
of the sticking curve, i.e. at high sticking probabilities, and an increase for positive curvature, i.e.
at low sticking probabilities.
Adsorption well depths have often been estimated by fitting measured trapping probabilities to
curves derived from the hard cube model, also for molecular adsorption (see, e.g., [32, 33]). In the
simple hard cube model the surface is flat and structureless which means that in any scattering
and adsorption process the incident parallel momentum is conserved. For the sticking probability
RTO-EN-AVT-142 4 - 11
1.0
∗∗ ∗
0.8
∗ ∗∗Xe
Trapping probabilty
0.6
∗∗
∗
∗ Kr ∗ Experiment
0.4 ∗
∗∗ ∗
∗ Ar ∗∗ Theory
∗ ∗
0.2 ∗ ∗∗
∗
∗∗ Ne ∗ ∗ ∗∗ ∗
∗ ∗ ∗ ∗20 ∗ ∗ ∗ ∗
60 ∗
0.0
0 40 80
Kinetic energy (meV)
Figure 5: Sticking probability of rare gas atoms on Ru(0001) at a surface temperature of Ts = 6.5 K.
Stars (*): experiment; lines: theoretical results obtained with the forced oscillator model (after [34],
not all measured data points are included)
this leads to the normal energy scaling, i.e., the sticking probability is a function of the normal
component Ei cos2 θi of the incident energy alone, where θi is the angle of incidence. Real surfaces,
however, are not structureless as far as the interaction of atoms and molecules is concerned since
adsorption corresponds to the making of a chemical bond which strongly depends on the local
environment. This leads to corrugation in the potential energy surface, i.e., the potential depends
on the lateral position of the interacting particle on the surface. Trapping probabilities often scale
as Ei cosn θi with n < 2. An exponent of n = 0 corresponds to total energy scalingwhich is usually
associated with a highly corrugated potential energy surface.
A further prediction of the hard cube model is that the sticking probability in the limit E → 0 and
Ts → 0 always becomes unity no matter how small the adsorption well, no matter how small the
mass ratio between the impinging atom and the substrate oscillator, if there is no barrier before
the adsorption well. This is due to the energy transfer to the surface which in the limit of zero
energy does not allow the particle to return into the gas-phase.
Experiments indeed show that for many adsorption systems all impinging particles become trapped
at low kinetic energies, for example in the case of Xe impinging on Ru(0001) [35, 34] which is shown
in Fig. 5. However, for the lighter rare gases Kr, Ar and in particular Ne the trapping probability
extrapolated to E → 0 is clearly below one. This behavior can not be reproduced using classical
mechanics [35], instead it can only be understood if the quantum nature of the substrate phonon
system is taken into account leading to a non-zero probability for elastic scattering at the surface.
A classical treatment of the solid is only appropriate if the energy transfer to the surface is large
compared to the Debye energy of the solid [36].
The essential physics can be captured by just considering an atomic projectile interacting via
linear coupling with a single quantum surface oscillator in the so-called trajectory approximation.
In this model, the classical motion of the incoming particle is assumed not to be perturbed by the
coupling to the surface oscillator. The classical trajectory then introduces a time-dependent force
4 - 12 RTO-EN-AVT-142
in the Hamiltonian of the oscillator

1
Hosc = ~ω (a+ a + ) + λ(t) (a+ + a) , (28)
2
In this forced oscillator model [37, 38, 39], the mean number of excited phonon n̄ is given basically
by the square of the Fourier transform of the coupling λ(t),
2
Z∞

1 iωt0 0 0

n̄ = e λ(t )dt . (29)
~
−∞
The number n̄ also enters the expression of probability Pji for a transition from an initial oscillator
state i to the final state j [37]:
i! −n̄ j−i j−i
Pji = e n̄ [Lj (n̄)]2 , j≥i. (30)
j!
Here Lj−i
j is an associated Laguerre polynomial [40]. For an excitation Pj0 from the ground state
one just obtains a Poisson distribution. The probabilities (30) yield the energy distribution of
excited phonons and hence also the energy transfer from the impinging particle to the surface
which that they correspond to the probability PE () entering the expression (24) for the sticking
probability.
This forced oscillator model has been applied to evaluate the sticking probabilities of rare gas
atoms on a Ru(001) surface at a temperature of Ts = 6.5 K [34]. Using the potential well depth,
the potential range, the mass of the surface oscillator and the surface Debye temperature as
parameters, this simple model was able reproduce the measured data quite well (see Fig. 5). In
particular for the a light rare gas such as Ne impinging on a metal surface, the quantum effects in
the surface recoil are quite substantial.
At even lower kinetic energies than reached in the experiments [34] shown in Fig. 5, the quantum
nature of the impinging particles cannot be neglected any longer. Hence the trajectory approxi-
mation cannot be applied any more. In fact, in the limit E → 0 the de Broglie wavelength of the
incoming particle tends to infinity. In the case of a short-range attractive potential this means that
the amplitude of the particle’s wave function vanishes in the attractive region [36, 41]. Thus there is
no coupling and consequently no energy transfer between the particle and the substrate vibrations.
Therefore the quantum mechanical sticking probability also vanishes for E → 0. However, in order
to see this effect extremely small kinetic energies corresponding to a temperature below 0.1 K are
required [36]. Nevertheless, this quantum phenomenon in the sticking at surfaces has been verified
experimentally for the adsorption of atomic hydrogen on thick liquid 4 He films [42].
5 Dissociative adsorption
In the case of dissociative adsorption there is another channel for energy transfer, namely the
conversion of the kinetic and internal energy of the molecule into translational energy of the atomic
fragments on the surface relative to each other. This bond-breaking process which is illustrated in
RTO-EN-AVT-142 4 - 13
3.0 2.5
Center of mass distance Z (Å)

1
0
0
1 a) H2 /Pd(100)
0
1
0
1
0.4
0
1 b) H2 /Cu(100)
2
2.0 0.0
11
00
00
11
1.5 00
11
1.6
11
00
00
11
1.0 00
11
00
11
−0
00
11
.3
00
11 0.6
1
−1.0
0.2111
000
000
111
000
111
0.0
0.0
0.5
0.5 1 1.5 2 2.5
1.0 2.0 3.0
H−H distance d(Å) H−H distance d(Å)
Figure 6: Contour plots of the potential energy surface along two-dimensional cuts through the
six-dimensional coordinate space of H2 in front of (100) metal surfaces determined by DFT-GGA
calculations for H2 molecules above the bridge site. The contour spacing is 0.1 eV per H2 molecule.
(a) H2 /Pd(100) (after [43]), (b) H2 /Cu(100) [44]. The inset is a) illustrates the process of disso-
ciative adsorption while in b) two typical trajectories showing the effect of initial vibrations are
included.
the inset of Fig. 6 represents the fundamental difference to atomic or molecular adsorption. As far
as the complete adsorption process is concerned, it is true that eventually the atomic fragments
will also dissipate their kinetic energy and come to rest at the surface. However, in the case of light
molecules, in particular molecular hydrogen, dissociating on metal surfaces the energy transfer
to the substrate is very small because of the large mass mismatch. Hence the probability for
dissociative adsorption is almost entirely given by the initial dissociation probability upon the
impact on the surface since the fragments do in general not directly recombine and desorb again.
Therefore the dissociative adsorption process can be described within low-dimensional potential
energy surfaces neglecting the surface degrees of freedom. This also requires that there is no
substantial surface rearrangement upon adsorption, but this is usually the case in the dissociative
adsorption on close-packed metal surfaces.
The dynamics of the interaction of hydrogen with metal surfaces has been well-studied, both ex-
perimentally [45] and theoretically [2, 46, 47, 48]. In Fig. 6, two cuts through the six-dimensional
potential energy surface derived from DFT calculations for two benchmark systems are plotted,
for H2 /Pd(100) where non-activated dissociation is possible [43]) and for the activated system
H2 /Cu(100) [44]. The cuts including the respective minimum energy paths to dissociative adsorp-
tion correspond to so-called elbow plots which show the potential energy as a function of the H2
center of mass distance from the surface and the H-H distance.
In order to quantitatively evaluate the adsorption probability of hydrogen, the full six-dimensional
PES has to be known. Furthermore, the dynamics of hydrogen require a quantum treatment be-
cause of its light mass. Because of the high computational effort associated with quantum methods,
4 - 14 RTO-EN-AVT-142
1.0 1.0
Experiment th
0.8
6D QD, ji = 0 (Eichler et al.) a) pa b)
0.8 0.6 on
i at
e
act din surface
re or
Sticking probability
0.4
0.2 co
0.6
0.0
0.0 0.1 0.2 0.3 0.4 0.5
0.4
0.2
low medium high
Experiment energy energy energy
6D QD, ji = 0 (Gross et al.)
6D QD, beam simulation gas phase
0.0
0.0 0.1 0.2 0.3 0.4 0.5 surface coor

dinate
Kinetic energy Ei (eV)
Figure 7: Dissociative adsorption of H2 /Pd(100). a) Sticking probability of H2 /Pd(100) as a func-

tion of the initial kinetic energy. Circles: experiment [51], dashed and solid line: theory according
to H2 initially in the ground state and with a thermal distribution appropriate for a molecular
beam [49]. The inset shows the theoretical results using an improved ab initio potential energy
surface [52]. b) Illustration of the steering effect.
for a long time the theoretical treatment was limited to studies within a reduced dimensionality.
Only recently the first quantum studies were performed in which the full dimensionality of the
hydrogen molecule was taken into account [49, 50].
The corresponding dissociative adsorption probability of H2 /Pd(100) is shown in Fig. 7a. Experi-
ment [51] and theory [49, 52] agree well, as far as the qualitative trend of the adsorption probability
as a function of the kinetic energy is concerned. First there is an initial decrease, and after a min-
imum the sticking probability rises again. The initial decrease of the sticking probability is typical
for H2 adsorption at transition metal surfaces [51, 53, 45, 54, 55]. Originally, such an initial de-
crease was associated with the existence of a molecular adsorption precursor state through which
the dissociation was assumed to proceed. However, nowadays it is well established that this be-
havior in the dissociation of H2 on reactive metal surfaces is caused by dynamical effects, namely
steering [56, 49, 57] and dynamical trapping [58, 59, 60].
The steering effect is illustrated in Fig. 7b where typical trajectories are plotted as a function of
a lateral coordinate and a reaction path coordinate which connects the molecule in the gas phase
with the dissociated molecule on the surface. It is important to note that in these systems the
PES shows purely attractive paths towards dissociative adsorption, but the majority of reaction
paths for different molecular orientations and impact points exhibits energetic barriers hindering
the dissociation. Still, at low kinetic energies most impinging molecules are steered towards the
attractive dissociation channel leading to a high adsorption probability. In addition, the corrugation
and anisotropy leads to a conversion of the kinetic energy of the impinging molecule into internal
molecular degrees of freedom such as rotation and vibration and into lateral motion along the
surface. This energy is then not available for a direct escape from the adsorption well so that
the molecules become dynamically trapped. Both mechanisms that are only operative a low kinetic
energies result large adsorption probabilities, and their suppression for higher kinetic energies leads
to the decrease of the sticking probability. At even higher energies, the molecules can directly cross
RTO-EN-AVT-142 4 - 15
0.6
First excited 6D, v = 1
0.5 vibrational state
Adsopriton probability
0.4
Experiment, v = 1
0.3
∆Ev 6D, v = 0
0.2 Vibrational
groundstate
0.1 Experiment, v = 0
0.0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Incident kinetic energy (eV)
Figure 8: Dissociative adsorption probability of H2 on Cu(100) as a function of the incident kinetic

energy determined by six-dimensional quantum wave-packet calculations for molecules initially in
the vibrational ground state and first excited state, respectively [61]. The calculations are compared
to experimental results derived from an analysis of adsorption and desorption experiments [62].
the dissociation barriers causing again an increase in the adsorption probability.

As far as the activated adsorption is concerned, the interaction of hydrogen with copper surfaces
has served as a model system [46, 47, 63, 64, 50, 65, 66]. It was also the first system for which
high-dimensional potential energy surface were mapped out by DFT methods [67, 68, 69]. Figure 8
shows the calculated sticking probabilities for molecules either in the vibrational ground state or
in the first excited state, respectively, based on an six-dimensional wave-packet calculations on an
ab initio PES [50, 61]. The theoretical results are rather close to the experimental data which were
derived from an analysis of both adsorption and desorption experiments [62] by using the following
analytical form of the vibrationally resolved sticking probability as a function of the kinetic energy:

A E − E0 (v)
Sv (E) = 1 + tanh (31)
2 W (v)
The agreement between theory and experiment in Fig. 8 is very satisfactory except for the high-
energy behavior. However, the experimental data of H2 in Fig. 8 for kinetic energies above 0.5 eV
were derived from thermal desorption experiments. As far the saturation value of the experimental
sticking probability is concerned, there is a large uncertainty since the high energy contributions
in desorption are exponentially suppressed through the Boltzmann factor.
The sticking probability of H2 /Cu(100) shows a typical activated behavior with an onset at ap-
proximately 0.5 eV for H2 for molecules initially in the vibrational ground state. This onset is given
by the minimum energy barrier including zero-point effects which arise from the quantization of
the molecular levels due to the localization of the wave function at the minimum barrier position.
The slope of the sticking probability is directly related to the distribution of the barrier heights
for dissociative adsorption in the multidimensional potential energy surface [64]. Thus sticking can
be understood in terms of the region of the surface that classically is available to dissociation.
This so-called hole model [70] is valid at high kinetic energies when the incoming particles are not
4 - 16 RTO-EN-AVT-142
significantly redirected by the shape of the potential energy surface.

As Fig. 8 demonstrates, for initially vibrationally excited molecules the sticking probability is
significantly enhanced in the system H2 /Cu. Using the so-called vibrational efficacy,
∆Ev
χ= , (32)
~ωvib
where ∆Ev is the energetic shift between the sticking curves for molecules in the vibrationally
ground and first-excited state, the vibrationally enhanced dissociation can be quantified. The
shift ∆Ev is approximately 0.3 eV, as indicated in Fig. 8. Consequently, with the vibrational fre-
quency of H2 , ~ωvib = 0.516 eV, the vibrational efficacy becomes χ ≈ 0.6 which is often interpreted
to mean that 60% of the vibrational energy is used to overcome the barrier for dissociative adsorp-
tion.
Vibrationally enhanced dissociation has been known for years in gas phase dynamics [71]. The
basic mechanism can be discussed within a two-dimensional elbow plot as shown in Fig. 6b. The
system H2 /Cu(100) exhibits a barrier which is located after the curved region of the minimum
energy path which is called a late barrier. For such a topology of the PES, initial vibrations can
be very helpful, as is illustrated with two typical trajectories in Fig. 6b. Initially non-vibrating
molecules with a kinetic energy less than the barrier height are reflected at the adsorption barrier
(dashed line). However, if the molecule is already initially vibrating, i.e., if it is oscillating back
and forth in the d-direction, then the vibrational energy can be very efficiently used “to make it
around the curve” and enter the dissociation channel. In addition, there are vibrationally adiabatic
effects associated with the lowering of the vibrational frequency perpendicular to the reaction path
which also contribute to the vibrationally enhanced dissociation [72].
In addition to the vibrational state, also the rotational state can have a significant influence on the
dissociative adsorption probability. Usually additional rotational motion suppresses the adsorption
probability because the molecule will rotate out of a favorable orientation for dissociation during
the adsorption process [8]. Invoking the principle of microscopic reversibility of detailed balance,
this leads to the so-called rotational cooling in desorption, i.e., the mean rotational energy is below
the value expected for thermal equilibrium. Conversely, the vibrationally enhanced dissociation
corresponds to vibrational heating in desorption.
6 The full concert: molecular and dissociative adsorption and

scattering
In the examples discussed so far, we focused either on the energy transfer in atomic and molecular
adsorption or on the bond-breaking process in dissociative adsorption. However, there are molecule-
surface systems in which all these processes can occur at the same time. One example is the
interaction of O2 /Pt(111). This is a well-studied system motivated by the fact that the adsorption
of oxygen on platinum represents one of the fundamental microscopic reaction steps occuring in the
car-exhaust catalyst. On Pt(111), there exist both physisorbed and chemisorbed molecular oxygen
species [73, 74] as well as dissociatively adsorbed oxygen. The chemisorbed species have also been
RTO-EN-AVT-142 4 - 17
identified in total-energy calculations [75, 76] using density functional theory (DFT) within the
generalized gradient approximation (GGA) [77].
According to molecular beam experiments [78, 79], the sticking probability of O2 /Pt(111) first
exhibits a strong decrease, and then after passing a minimum at approximately 0.15 eV the sticking
probability levels off at a value of about 0.3 [78, 79]. Surprisingly, in molecular beam experiments it
was also found that oxygen molecules do not dissociate at cold Pt surfaces below 100 K [33, 80, 79],
even at kinetic energies above 1 eV which are much greater than the dissociation barrier.
The theoretical description of the adsorption dynamics of O2 /Pt(111) represents a significant
challenge. On the one hand, a realistic PES is needed that reliably describes both the molecular
as well as the dissociative adsorption channels. On the other hand, molecular trapping processes
can only be reproduced if the energy dissipation to the platinum substrate is properly taken into
account, as discussed in sect. 4. Direct ab initio molecular dynamics simulations represent a scheme
that meets these requirements, but it is computationally still very expensive [2, 81]. Using empirical
classical potentials, almost arbitrarily many trajectories can be computed, however, there are no
reliable interaction potentials available treating reactions on the surface and the surface recoil upon
impact on an equal footing.
As an intermediate method, tight-binding molecular dynamics simulations have been performed [82,
83] with the parameters of the tight-binding Hamiltonian derived from ab initio calculations [84, 75,
76]. In tight-binding, the exact many-body Hamiltonian is replaced by parameterized Hamiltonian
matrix elements of the effective one-particle Hamiltonian in an atomic-like basis set. The atomic-
like basis functions are usually not considered explicitly, but the matrix elements are assumed to
have the same symmetry properties as matrix elements between atomic states. The evaluation of
the tight-binding Hamiltonian still requires the diagonalization of a matrix, but it is about three
orders of magnitude faster than corresponding ab initio calculations.
A comparison between the calculated [83] and the measured [78, 79] sticking probabilities is shown
in Fig. 9a. It is obvious that the experimental data are qualitatively and even semi-quantitatively
reproduced by the ab initio based tight-binding molecular dynamics calculations. It is important
to note that also the experimental finding that O2 does not directly dissociate upon adsorption is
confirmed by the calculations.
There is a simple explanation for this result in terms of the topology of the elbow plots (see Fig. 9b).
Dissociation corresponds to an event in which the molecules enter the exit channel towards the
lower right corner of the figures. To enter this channel directly from the gas phase through the
molecular adsorption state requires a sharp turn of the trajectories. In Fig. 9b, a trajectory of an
O2 molecule directly aimed at the molecular precursor state is included. Its kinetic energy of 0.6 eV
is much higher than the dissociation barrier (∼ 0.2 eV [76]); still it does not dissociate. It becomes
accelerated by the attractive potential, hits the repulsive wall of the potential and is scattered
back. This shows that direct dissociation of O2 /Pt(111) is not impossible, but it is very unlikely.
Thus it follows that because of this steric hindrance, dissociation of O2 on Pt(111) is a two-step
process. First the molecule becomes trapped and accommodated in the molecular chemisorption
state, and only subsequently it dissociates at sufficiently high surface temperatures due to thermal
fluctuations which will make the O2 molecules enter the dissociation channel.
4 - 18 RTO-EN-AVT-142
1.0 4.0 4.0

Pt

Exp. Luntz et al., Ts = 200 K
Exp. Luntz et al., Ts = 90 K 3.6 3.6
0.8
O
Trapping probability
Exp. Nolan et al., Ts = 77 K

o 3.2 3.2
TBMD, Ts = 0 K, Θi = 0
0.6 TBMD, Ts = 300 K, Erot = 0.1 eV
2.8 2.8
(b)
(a) (b)
0.4
2.4 (c) 2.4
2.0 2.0
0.4
0.0
0.2 1.0
(a) 1.6 1.6
0.0 1.2 1.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.0 1.5 2.0 2.5 1.0 1.5 2.0 2.5 3.0
2
Normal kinetic energy Ekin cos Θi (eV) O−O distance d(Å)
Figure 9: O2 adsorption on Pt(111). a) Trapping probability of O2 /Pt(111) as a function of the

normal component of the kinetic energy. Results of molecular beam experiments for surface tem-
peratures of 90 K and 200 K (Luntz et al. [78]) and 77 K (Nolan et al. [79]) are compared to
tight-binding molecular dynamics simulations for the surface initially at rest (Ts = 0 K). b and
c) Elbow plots of the PES of the dissociation of O2 /Pt(111) determined by the ab initio derived
tight-binding Hamiltonian [76, 83]. The configurations of the remaining O2 degrees of freedom are
illustrated in the insets. The contour spacing is 0.2 eV per O2 molecule. In (b) a trajectory of an
O2 molecule with an initial kinetic energy of 0.6 eV scattered at Pt(111) is also plotted.
The molecular dynamics simulations also showed that contrary to common belief [33, 79] the
strong initial decrease of the sticking probability is not caused by the trapping into a shallow
physisorption state. Instead, the high sticking probability at low kinetic energies is again caused
by the steering effect which becomes quickly suppressed for higher kinetic energies. Thus it is
not the energy transfer per se that determines the sticking probability at low kinetic energies but
rather the probability to enter the molecular chemisorption state. All molecules that find their way
to the molecular chemisorption state at low kinetic energies do in fact remain trapped.
The magnitude of the steering effect is a consequence of the strong corrugation and anisotropy of
the PES of O2 /Pt(111). The significant corrugation is illustrated in Figs. 9b and c. The lateral
position of the O2 center of mass is only shifted by about 1 Å between the two cuts. Nevertheless,
there is no longer any chemisorption well present but rather a large barrier of about 1 eV towards
dissociative adsorption which becomes even larger for the molecule directly at the on-top site. In
fact, the majority of adsorption pathways are hindered by barriers; direct non-activated access to
the adsorption states is possible for only a small fraction of initial conditions.
The strong corrugation and anisotropy of the PES has further consequences. First of all it leads
to a high probability of dynamical trapping due to the conversion of the initial kinetic energy
into internal molecular degrees of freedom which for this particular system is almost independent
from the kinetic energy. This causes the leveling off of the trapping probability at higher kinetic
energies. Furthermore, in Fig 9a the results of calculations for non-normal incidence are shown
for two different total kinetic energies, Ei = 0.3 eV and 1.1 eV. Additional parallel momentum
strongly suppresses the sticking probability, in agreement with the experiment [78]. It is obvious
RTO-EN-AVT-142 4 - 19
180
Theory, Ei = 1.1 eV, Erot = 0.1 eV
Scattering intensity (arb. units)

Azimuthal scattering angle ( )
o
Theory, Ei = 1.1 eV, Erot = 0.3 eV
1
Experiment, Ei = 1.27 eV
90
0.8
o
0.6 b) Θi = 60
0
0.4
-90
o 0.2
a) Ei = 1.1 eV, Θi = 60
-180 0
0 20 40 60 80 0 20 40 60 80
o o
Polar scattering angle ( ) Final scattering angle ( )
Figure 10: Angular distribution of O2 scattered from Pt(111) with an angle of incidence of 60◦ . a)
Calculated angular distribution (surface temperature Ts = 300 K) for an initial rotational energy of
0.1 eV (filled circles) and 0.3 eV (open circles). b) Comparison of the measured and the calculated
angular distribution in in-plane scattering for an angle of incidence of 60◦ . The initial kinetic energy
in the experiment was 1.27 eV [85] while the TBMD simulations have been performed for an initial
energy of 1.1 eV. Theoretical results are shown for both an initial rotational energy of 0.1 eV and
0.3 eV.
that the trapping of O2 /Pt(111) does not obey normal energy scaling, i.e., it is not a function of
the normal kinetic energy alone, but total energy scaling is also not obeyed since the results for
non-normal incidence are smaller than those for normal incidence.
For non-normal incidence, also the angular distribution of the scattered O2 molecules that is shown
in Fig. 10a has been analyzed in detail. The distribution shows that there is predominantly in-
plane scattering, i.e. the molecules do not significantly change their azimuthal angle. In Fig. 10a,
the results for the inplane scattering are compared to the experiment [85]. There is a rather good
agreement between theory and experiment.
However, it is important to note that the angular distribution of Ar atoms scattered from Pt(111)
is very similar to that of O2 scattered at the same surface [86, 85]. The interaction potential of
noble gas atoms with low-index metal surfaces is usually relatively structureless and only weakly
corrugated. The similarity between Ar/Pt(111) and O2 /Pt(111) scattering indicates that also
the O2 scattering corresponds to the reflection from a rather flat surface. This has already been
indicated by the weak out-of-plane scattering in the system O2 /Pt(111). The width of the in-
plane distribution can in fact be explained by the energy transfer to a vibrating flat surface [85]
which leads to a certain width in the distribution of the normal component of the kinetic energy.
Together with the conservation of the parallel momentum of the scattered particles this causes the
broadened angular distribution around the specular direction in the scattering.
Thus there is the seemingly paradox conclusion that adsorption experiments suggest that the
O2 /Pt(111) interaction potential should be strongly corrugated while scattering experiments in-
dicate a rather small corrugation. This contradiction is caused by the fact that the scattered
molecules are reflected directly from the repulsive tails of the potential which is less strongly cor-
rugated while in the trapping the corrugation and anisotropy of the potential energy surface closer
to the surface become important.
4 - 20 RTO-EN-AVT-142
Figure 11: Illustration of the different apparent corrugation of a surface as a function of the angle
of incidence.
This phenomenon is illustrated in Fig. 11. In the system O2 /Pt(111) scattering corresponds to the
majority channel at higher kinetic energies while adsorption is the minority channel. Adsorption is
strongly dependent on the corrugation of the surface, however, for the scattered molecules at higher
angles of incidence, due to shadowing effects the surface looks rather smooth. Since adsorption is
the minority channel, the additional scattering flux which results from the suppression of adsorption
at higher angles of incidence does not crucially influence the scattering distribution.
7 Conclusions
In this chapter, I have briefly reviewed the fundamentals of the dynamics of gas-surface interactions
and their theoretical treatment. There are some phenomena in this interaction such as elastic
scattering and diffraction that can only be understood in a quantum treatment. Most of the
processes, however, can be understood qualitatively and even quantitatively based on classical
dynamics simulations.
While atomic and molecular adsorption require a appropriate treatment of the excitation spectrum
of the semi-infinite solid in order to correctly describe the energy transfer to the substrate, in the
dissociative adsorption of light molecules, in particular H2 , this transfer can be neglected since
the crucial process is the bond-breaking upon adsorption. However, a reliable high-dimensional
potential energy surface is needed in order to reproduce the dependence of the adsorption prob-
ability on the internal degrees of freedom of the impinging molecules. Ab initio based molecular
dynamics simulations are capable of giving a reliable and rather complete picture of the interaction
of molecules with surfaces.
Electronically non-adiabatic processes have not been discussed in this chapter. Their theoretical
treatment is still problematic. Such processes can either lead to further energy losses due to the
excitation of electron-hole pairs in the substrate, or they modify the reaction dynamics due to
transitions between electronically excited states. However, for most thermal processes at surfaces,
electronically non-adiabatic processes only seem to play a minor role.
RTO-EN-AVT-142 4 - 21
References
[1] A. Groß, Dynamics of molecule-surface interactions from first principles, in The Chemical
Physics of Solid Surfaces, edited by D. P. Woodruff, volume 11, chapter 1, Elsevier, Amster-
dam, 2003.
[2] A. Groß, Surf. Sci. Rep. 32, 291 (1998).
[4] W. Kohn, Rev. Mod. Phys. 71, 1253 (1999).
[5] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in

Fortran 90. The Art of Parallel Scientific Computing, Cambridge University Press, second
edition, 1996.
[6] L. Verlet, Phys. Rev. 159, 98 (1967).
[7] D. Heermann, Computer Simulation Methods in Theoretical Physics, Springer, Berlin, second
edition, 1990.
[8] A. Groß, Theoretical surface science – A microscopic perspective, Springer, Berlin, 2002.
[9] R. Newton, Scattering Theory of Waves and Particles, Springer, New York, second edition,
1982.
[10] J. A. Fleck, J. R. Morris, and M. D. Feit, Appl. Phys. 10, 129 (1976).
[11] M. D. Feit, J. A. Fleck, and A. Steiger, J. Comput. Phys. 47, 412 (1982).
[12] H. Tal-Ezer and R. Kosloff, J. Chem. Phys. 81, 3967 (1984).
[13] W. Brenig, T. Brunner, A. Groß, and R. Russ, Z. Phys. B 93, 91 (1993).
[14] W. Brenig, , A. Groß, and R. Russ, Z. Phys. B 97, 311 (1995).
[15] I. Estermann and O. Stern, Z. Phys. 61, 95 (1930).
[16] E. Hulpke, editor, Helium Atom Scattering from Surfaces, volume 27 of Springer Series in
Surface Sciences, Springer, Berlin, 1992.
[17] D. Farı́as and K.-H. Rieder, Rep. Prog. Phys. 61, 1575 (1998).
[18] B. Gumhalter, Phys. Rep. 351, 1 (2001).
[19] R. Frisch and O. Stern, Z. Phys. 84, 430 (1933).
[20] M. Patting, D. Farı́as, and K.-H. Rieder, Surf. Sci. 429, L503 (1999).
[21] R. G. Rowe and G. Ehrlich, J. Chem. Phys. , 4648 (1975).
[22] G. Brusdeylins and J. P. Toennies, Surf. Sci. 126, 647 (1983).
4 - 22 RTO-EN-AVT-142
[23] J. Lapujoulade, Y. Lecruer, M. Lefort, Y. Lejay, and E. Maurel, Surf. Sci. 103, L85 (1981).
[24] M. F. Bertino and D. Farı́as, J. Phys.: Condens. Matter 14, 6037 (2002).
[25] K. B. Whaley, C.-F. Yu, C. S. Hogg, J. C. Light, and S. Sibener, J. Chem. Phys. 83, 4235
(1985).
[26] A. Groß and M. Scheffler, Chem. Phys. Lett. 263, 567 (1996).
[27] U. Garibaldi, A. C. Levi, R. Spadacini, and G. E. Tommei, Surf. Sci. 48, 649 (1995).
[28] M. F. Bertino, F. Hofmann, and J. P. Toennies, J. Chem. Phys. 106, 4327 (1997).
[29] D. Cvetko, A. Morgante, A. Santaniello, and F. Tommasini, J. Chem. Phys. 104, 7778 (1996).
[30] J. T. Kindt, J. C. Tully, M. Head-Gordon, and M. A. Gomez, J. Chem. Phys. 109, 3629
(1998).
[31] E. K. Grimmelmann, J. C. Tully, and M. J. Cardillo, J. Chem. Phys. 72, 1039 (1980).
[32] E. W. Kuipers, M. G. Tenner, M. E. M. Spruit, and A. W. Kleyn, Surf. Sci. 205, 241 (1988).
[33] C. T. Rettner and C. B. Mullins, J. Chem. Phys. 94, 1626 (1991).
[34] H. Schlichting, D. Menzel, T. Brunner, and W. Brenig, J. Chem. Phys. 97, 4453 (1992).
[35] H. Schlichting, D. Menzel, T. Brunner, W. Brenig, and J. C. Tully, Phys. Rev. Lett. 60, 2515
(1988).
[36] R. Sedlmeir and W. Brenig, Z. Phys. B 36, 245 (1980).
[37] R. W. Fuller, S. M. Harris, and E. L. Slaggie, Am. J. Phys. 31, 431 (1963).
[38] W. Brenig, Z. Phys. B 36, 81 (1979).
[39] J. Böheim and W. Brenig, Z. Phys. B 41, 243 (1981).
[40] G. B. Arfken and H. J. Weber, Mathematical Methods for Physicists, Academic Press, San
Diego, 4th edition, 1995.
[41] W. Brenig, Z. Phys. B 36, 227 (1980).
[42] I. A. Yu et al., Phys. Rev. Lett. 71, 1589 (1993).
[43] S. Wilke and M. Scheffler, Phys. Rev. B 53, 4926 (1996).
[44] S. Sakong and A. Groß, Surf. Sci., accepted for publication.
[45] K. D. Rendulic and A. Winkler, Surf. Sci. 299/300, 261 (1994).
[46] G. R. Darling and S. Holloway, Rep. Prog. Phys. 58, 1595 (1995).
[47] G.-J. Kroes, Prog. Surf. Sci. 60, 1 (1999).
RTO-EN-AVT-142 4 - 23
[48] W. Diño, H. Kasai, and A. Okiji, Prog. Surf. Sci. 63, 63 (2000).
[49] A. Groß, S. Wilke, and M. Scheffler, Phys. Rev. Lett. 75, 2718 (1995).
[50] G.-J. Kroes, E. J. Baerends, and R. C. Mowrey, Phys. Rev. Lett. 78, 3583 (1997).
[51] K. D. Rendulic, G. Anger, and A. Winkler, Surf. Sci. 208, 404 (1989).
[52] A. Eichler, J. Hafner, A. Groß, and M. Scheffler, Phys. Rev. B 59, 13297 (1999).
[53] C. T. Rettner and D. J. Auerbach, Chem. Phys. Lett. 253, 236 (1996).
[54] M. Beutl, M. Riedler, and K. D. Rendulic, Chem. Phys. Lett. 256, 33 (1996).
[55] M. Gostein and G. O. Sitz, J. Chem. Phys. 106, 7378 (1997).
[56] D. A. King, CRC Crit. Rev. Solid State Mater. Sci. 7, 167 (1978).
[57] M. Kay, G. R. Darling, S. Holloway, J. A. White, and D. M. Bird, Chem. Phys. Lett. 245,
311 (1995).
[58] A. Groß and M. Scheffler, J. Vac. Sci. Technol. A 15, 1624 (1997).
[59] H. F. Busnengo, W. Dong, and A. Salin, Chem. Phys. Lett. 320, 328 (2000).
[60] C. Crespos, H. F. Busnengo, W. Dong, and A. Salin, J. Chem. Phys. 114, 10954 (2001).
[61] D. A. McCormack et al., Chem. Phys. Lett. 328, 317 (2000).
[62] H. A. Michelsen and D. J. Auerbach, J. Chem. Phys. 94, 7502 (1991).
[63] C. T. Rettner, D. J. Auerbach, and H. A. Michelsen, Phys. Rev. Lett. 68, 1164 (1992).
[64] A. Groß, B. Hammer, M. Scheffler, and W. Brenig, Phys. Rev. Lett. 73, 3121 (1994).
[65] D. A. McCormack and G.-J. Kroes, Phys. Chem. Chem. Phys. 1, 1359 (1999).
[66] Y. Miura, H. Kasai, and W. Diño, J. Phys.: Condens. Matter 14, L479 (2002).
[67] B. Hammer, M. Scheffler, K. Jacobsen, and J. Nørskov, Phys. Rev. Lett. 73, 1400 (1994).
[68] J. A. White, D. M. Bird, M. C. Payne, and I. Stich, Phys. Rev. Lett. 73, 1404 (1994).
[69] G. Wiesenekker, G.-J. Kroes, and E. J. Baerends, J. Chem. Phys. 104, 7344 (1996).
[70] M. Karikorpi, S. Holloway, N. Henriksen, and J. K. Nørskov, Surf. Sci. 179, L41 (1987).
[71] J. C. Polanyi and W. H. Wong, J. Chem. Phys. 51, 1439 (1969).
[72] A. Groß and M. Scheffler, Chem. Phys. Lett. 256, 417 (1996).
[73] W. Wurth et al., Phys. Rev. Lett. 65, 2426 (1990).
[74] C. Puglia et al., Surf. Sci. 342, 119 (1995).
4 - 24 RTO-EN-AVT-142
[75] A. Eichler and J. Hafner, Phys. Rev. Lett. 79, 4481 (1997).
[76] A. Eichler, F. Mittendorfer, and J. Hafner, Phys. Rev. B 62, 4744 (2000).
[77] J. P. Perdew et al., Phys. Rev. B 46, 6671 (1992).
[78] A. C. Luntz, M. D. Williams, and D. S. Bethune, J. Chem. Phys. 89, 4381 (1988).
[79] P. D. Nolan, B. R. Lutz, P. L. Tanaka, J. E. Davis, and C. B. Mullins, J. Chem. Phys. 111,
3696 (1999).
[80] P. D. Nolan, B. R. Lutz, P. L. Tanaka, J. E. Davis, and C. B. Mullins, Phys. Rev. Lett. 81,
3179 (1998).
[81] A. Groß, M. Bockstedte, and M. Scheffler, Phys. Rev. Lett. 79, 701 (1997).
[82] A. Groß, M. Scheffler, M. J. Mehl, and D. A. Papaconstantopoulos, Phys. Rev. Lett. 82, 1209
(1999).
[83] A. Groß, A. Eichler, J. Hafner, M. J. Mehl, and D. A. Papaconstantopoulos, Surf. Sci. 539,
L542 (2003).
[84] M. J. Mehl and D. A. Papaconstantopoulos, Phys. Rev. B 54, 4519 (1996).
[85] A. E. Wiskerke, F. H. Geuzebroek, A. W. Kleyn, and B. E. Hayden, Surf. Sci. 272, 256
(1992).
[86] J. E. Hurst, L. Wharton, K. C. Janda, and D. J. Auerbach, J. Chem. Phys. 78, 1559 (1983).
RTO-EN-AVT-142 4 - 25
4 - 26 RTO-EN-AVT-142
Molecular Dynamics Simulations of Surface Processes: Oxygen
Recombination on Silica Surfaces at High Temperature
M. Cacciatore and M. Rutigliano

Institute of Inorganic Methodologies and Plasmas CNR-IMIP
c/o Dipartimento di Chimica, Università di Bari v.Orabona,4
70126 Bari
Italy
mario.cacciatore@ba.imip.cnr.it
A) GENERAL CONSIDERATIONS
The interaction between a molecule and a solid surface can lead to a great variety of elementary processes
such as elastic, inelastic and reactive. Of particular importance is the dissociative chemisorption of a
diatomic molecule
O2 (v, j ) + wall − − > [O2 (v′, j′)]ad ∗ wall − − > O + O + wall (1)
where a molecule chemisorbed at the surface in a specific roto-vibrational state (v,j) dissociates with the
two atoms adsorbed or scattered into the gas-phase. Of great importance is also the atom recombination on
surfaces:
O + O + wall → O2 (v, j ) + wall (2)
here two atoms recombine thus forming a diatomic molecule that can be either chemisorbed or reflected in
the gas-phase in a given internal energy state. Reactions (1)-(2) are very often the rate determining step of
complex heterogeneous systems of interest in different branches of industrial and technological
applications, as for example in the ammonia synthesis, hydrocarbon production, chemical vapour
deposition, etching and thin solid film deposition via plasma, nuclear rector technologies [1].
Both processes are of central importance in aerothermodynamics and the chemistry of interstellar media.
Thus, the recombination of atomic O and N on silica and UHTC materials plays a central role for the
thermal protection system of the space shuttles entering into the terrestrial atmosphere, whereas the
recombination of hydrogen atoms on ice grains covered by carbon is, very likely, the main source of
molecular hydrogen observed in the interstellar media.
The interaction of chemical species with surfaces can lead to other non-reactive chemico-physical
processes such as the inelastic processes and adsorption. The adsorption processes occur when the particle
is trapped in a chemisorption site and its available energy is not enough to escape from the chemisorption
potential well. The inelastic processes can be of two types: direct and indirect. In the first case the
molecule hits the surface and it is scattered in the gas-phase after few bounces. Due to the interaction, the
internal energy state of the molecule leaving the surface is different with respect to its state before the
collision.
In the indirect scattering, or adsorption/desorption collisions, the interaction occurs on a longer time scale
and the collision is more involved: in fact, the molecule hits the surface several times, it is temporarily
adsorbed eventually forming a surface activated complex before desorbing in the gas-phase. The energy
distribution of the scattered molecules is different according to the different mechanism followed in the
interaction.
Cacciatore, M.; Rutigliano, M. (2007) Molecular Dynamics Simulations of Surface Processes: Oxygen Recombination on Silica Surfaces
at High Temperature. In Experiment, Modeling and Simulation of Gas-Surface Interactions for Reactive Flows in Hypersonic Flights
(pp. 5-1 – 5-22). Educational Notes RTO-EN-AVT-142, Paper 5. Neuilly-sur-Seine, France: RTO. Available from:
http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 5-1
Molecular Dynamics Simulations of Surface Processes:
Oxygen Recombination on Silica Surfaces at High Temperature
A further class of surface processes are: surface oxidation, chemical and physical sputtering, ion
implantation, etc. In these cases atomic and molecular radicals react with the atoms of the substrate, thus
forming volatile compounds. These processes, not discussed in this lecture, are effective in the energy
regime of several eV, or KeV, that is higher compared to the collisional energy regime of the processes
(1)-(2).
The dynamical quantity typically associated to the process (1) is the state-to-state sticking coefficient S0
which corresponds to the dissociation probability. S0 is a function of the several variables: S0= S0
(Ekin|{θi, ϕi}|v,j,TS), where Ekin is the impact kinetic energy, v and j the vibrational and rotational quantum
number of the incident molecule respectively, TS is the surface temperature. {θi, ϕi} are the angles that
define the relative orientation of the molecule with respect to the surface plane ( the collisional system,
sketched in Fig.1, is described in a cartesian frame of reference having the Z axis orthogonal to the
surface plane and the (X,Y) axis laying on the surface top layer).
r
Y
R
ϕ ϑ
Z
X
Fig. 1: The collisional system is sketched. The dynamics is described with respect
to a cartesian frame of reference having the Z axis orthogonal to the surface
plane and the (X,Y) axis laying on the surface top layer.
Different averaged quantities can be obtained from the state-to-state sticking coefficient. Thus the state-
average sticking coefficient <S0(Ekin|{θi, ϕi}|TS)>, usually measured in molecular beam experiments, is
obtained by averaging S0 over a known, not necessarily Boltzmann population distribution for the v and j
states. The statistical rate constant K(TS) used in kinetic modelling can be obtained by further averaging S0
over the incident angles and the Maxwell energy distribution function of the gas-phase molecule.Then we
get K(TS)=Kcoll *< S0 (TS)>.
In addition to S0, other collisional data needed to characterize the surface reactions and to include them in
kinetic modelling are: the energy accommodation coefficient β(Ts), the roto-vibrational distribution of the
scattered particles N(v,j|Ts), the angular distribution in intensity and energy I(ϑ,ϕ|Ts), the lifetimes of the
adsorbed species.
Collisional data can be determined by using a large variety of experimental techniques that have been
developed to probe directly or indirectly surface processes. Nevertheless, the observation of surface
5-2 RTO-EN-AVT-142
processes is not an easy task and in fact it is a matter of fact that collisional data for heterogeneous
processes are very sparse (if not absent) despite the large amount of information emerged from different
experimental works. Rate coefficients are still poorly known and somehow dubious, particularly for wall
processes occurring under extreme thermal conditions as those met in aerothermodynamics.
In Fig. 2 some of the experimental results reported in the literature for the recombination coefficient γ for O
atoms recombination on silica-based materials are reported in a large range of the surface temperature [2-4].
1000 500 330K

100
10-1
Recombination Coefficient
10-2
10-3
Scott (silica-based materials)
Greaves & Linnett (silica)
10-4 Berkut (silica-based materials)
10-5
0,0 0,5 1,0 1,5 2,0 2,5
1/T(Kx1000)
Fig. 2: The recombination coefficient γ for O atom recombination on silica-based
materials is reported as a function of the surface temperature.
In the high temperature regime the results are somehow contrasting, also due to the large experimental
uncertainties. From the theoretical point of view various kinetics schemes have been developed to describe
the kinetics of this reaction [5,6,7].
In these studies some physical quantities, typically the activation energy, surface coverage, the number of
active sites and the so-called 'steric factor', are treated as unknown parameters and varied to fit the
experimental data.
Due to the lack of accurate collisional data, a complete understanding of elementary surface processes
involving O, N, C, CO on silica-based materials has not yet completely achieved and a number of
fundamental aspects are open to question. Among the others, the most critical aspects discussed in
molecule-surface interaction studies concern, on the one hand the internal state selectivity of the
collisional coefficient and, on the other hand the energy transfer mechanisms that control the surface
processes. Further elements of complexity are introduced by the surface structure and surface coverage
effects on the catalytic activity of the substrate.
Possible answers to these issues can be given by performing Molecular Dynamics (MD) calculations
[8,9,10]. The MD approach has some important features in that predicts quantities that cannot be easily
measured in experiments: quite often, the state-selected sticking coefficient S0, or the energy
accommodation coefficient β can be established only on a theoretical ground.
RTO-EN-AVT-142 5-3
The MD simulation of surface chemical processes is quite complex both from the formal and the
computational point of view. The multidimensional nature of the collisional system, and thus the large
number of degrees of freedom introduced by the presence of the solid, increases considerably the
complexity of the dynamics. As a consequence, in the last few years a variety of collisional methods [8,9]
have been developed to simulate reactive surface processes, ranging from ab initio quantum mechanics
calculations of reduced dimension, semiclassical and mixed semiclassical-wave packet propagation
techniques, statistical simulations, classical and quasi-classical methods. MD simulations imply the
solution of three main problems: i.) the first, probably the most critical one that would prevent an accurate
description of the surface processes concerns the determination of the potential energy surface (PES)
where the dynamics takes place, that is the determination of the forces exerted between the atoms in the
gas-phase and the atoms of the surface. ii.) The second concerns the building up of a ‘model’ sample of
the solid substrate. iii.) Finally the third aspect to solve concerns the development of a suitable collisional
model to follow the nuclear motions of the particles propagating in the gas phase and interacting with the
solid surface.
In the following sections some interesting results obtained from MD simulations of atom recombination
processes over model silica surfaces that could be of interest in aerothermodynamics are briefly reviewed
and discussed.
B) INTERACTION POTENTIAL DETERMINATION FOR O,N/SILICA
B.1) Ab initio methods

Determining the PES where the reaction takes place is a prerequisite for any MD calculations. In principle,
this issue is addressed using computational methods developed in quantum chemistry [10]. Among the
proposed approximate schemes, the most promising approach is the density functional theory (DFT)
[9,11-13], which is accurate near the chemisorption well, but less accurate for long-range/weakly bound
interactions. The accurate determination of the interaction potential for heterogeneous systems poses
different problems, both from the theoretical and the computational point of view. Assuming that the
relaxation of the lattice atoms can be neglected, the PES is, in principle, a multidimensional function
which depends for diatom-surface interaction, upon six coordinates, that is the surface is too much
complex and the number of interactions is too large for ab initio electronic structure calculations, either
cluster or slab calculations, involving a large number of atoms and electrons. As a consequence, also due
to the large computational time demanded to assure the convergence of calculated energies, complete fully
dimension electronic structure calculations have been performed for few elementary systems, notably
those involving H2, CO, O2 and others diatomics on noble and transition metals. The 'true' interaction
potential remains rather unknown for a large class of heterogeneous systems, in particular for the catalytic
systems of interest in aerothermodynamics. In fact uncertainties exist on: the adsorption energies and their
dependence on the chemisorption site, the size and the nature of the activation energies, the energy
barriers to surface diffusion.
Due to the lack of ab initio calculations one has, quite often, to rely on low-dimension ab initio ‘model’
potentials where the dependence of the interaction potential upon the intramolecular distance of the diatom
and the distance from the surface is considered. PES of reduced dimension can be useful for understanding
the qualitative aspects of dissociation/recombination processes, but for a quantitative description a PES
that includes the dependence on the two (X,Y) surface plane coordinates is necessary.
An accurate two-dimensional adiabatic PES for N,N2 interacting with a silica surface has been recently
calculated [14] using the size-scalable cluster approach with SixOy clusters of increasing size cleaved from
the β-cristobalite unit cell. In this study the hybrid Hartree-Fock Self Consistent Field (HF-SCF) and the
DFT method was applied and the B3LYP functions [15] used in order to take into account the exchange
and the electron correlation contribution in the interaction. The interaction potential is calculated by
keeping the SiO2 geometry fixed at the experimental values of the β-cristobalite cell and extending the
5-4 RTO-EN-AVT-142
N/N2-SixOyHz cluster model in the intervals (x=1,3,7; y=2,4,6,14; z=2,6,14). All the calculations were
carried out with the Gaussian 03 package [16]. Complete scans of the RN-Si distance in the whole range of
energy interaction (RN-Si = 0.90 Å, RN-Si =4.00 Å) for just the perpendicular configuration (that is, with the
N-surface angles fixed to 90º) were performed. The N-N distance used in all N2-surface PES calculations
was kept fixed to 1.10 Å, close to the spectroscopic value [17]. As far as the N-SixOyHz interaction is
concerned, both doublet and quartet total electronic spin states were considered along the PES scan, since
the electronic spin state induced by chemisorption is not known. The calculated binding energies are
reported in Fig. 3. As expected, N is chemisorbed with a binding energy of about 2.79eV at a minimum N-
Si distance around 1.65 Å. N2 is slightly physisorbed.
Fig. 3: Full line: the interaction potential for N interacting with Si3O4H6 cleaved from the
β-cristobalite surface is reported as a function of the distance of the N atom from the
Si active atom. The dashed line shows the interaction potential of N2( req)
approaching the silica cluster in the perpendicular configuration on top
of the Si active atom. Results obtained at the DFT-B3LYP/6-311+G* level.
From the calculated quantum mechanical interaction energies a LEPS-type interaction potential can be
obtained [18,19]. In order to account for the motion of the lattice atoms explicitly considered in the
collision dynamics, the LEPS potential is, in this case, obtained as a sum of pair-wise N/silica atom
interaction and it is given by:
N
V= ∑ U (r) + U
1 2 (Ra,i ) + U 3 (Rb,i ) − A12 (r) + (A2 ( Ra,i ) + A3 (Rb,i ))2 − A1 (r)( A2 ( Ra,i ) + A3 (Rb,i )) (3)
i =1
where the sum is over the lattice atoms, N being the total number of atoms in the silica surface, r is the N-
N interatomic distance, Ra,i and Rb,i are the distances of nitrogen atom a and nitrogen atom b from the i-th
atom in the lattice, respectively. Uk (k=1,3) and Ak (k=1,3) are related to the two-body Coulomb and
RTO-EN-AVT-142 5-5
exchange integrals, respectively, in the energy expression for the interaction of the solid surface with the
molecule [18,19] and are given as a combination of modified Morse functions with the two additional Sato
parameters. The corresponding potential parameters are obtained by fitting the full potential given by Eq.
(3) to the ab initio DFT-B3LYP energies (Fig.3). Since the determination of the activation barrier for the
N2 formation would imply massive multidimensional DFT calculations, planned for the near future, the
Sato parameters are taken as a free parameters and changed in order to get PES’s with different barriers in
the recombination reaction channel. This allows to study the sensitivity of the calculated collisional
reaction coefficients upon the reaction barrier height.
In Fig. 4 the potential contour map of the PES for N2 interacting in the perpendicular geometry with the
active Si surface atom is shown. The interaction potential is plotted as a function of the distance from the
N atom closest to the surface, and the N-N bond distance. The PES shown is obtained assuming a value of
the Sato parameters that gives zero activation energy.
N2 distance from the surface (Å)
-7.6
5
-3.2
-5.1 -7.1
4
-8.5
-9.0
-9.5
-2.7
-1.2
-4.6
-8.1
-6.6
-2.2
-4.2
-6.1
3
-1.7
-3.7
-5.6-0.8
-0.3
0.2
0.7
2 1.7 1.2
2.6 2.2
4.63.1 4.1 3.6
5.1
6.1 6.5 5.6
7.0
7.5
8.0
1
1 2 3 4 5
N-N interatomic distance (Å)
Fig. 4: PES for N2 interacting in the perpendicular geometry with the active Si surface atom.
The interaction potential (in eV) is plotted as a function of the distance from
the N atom closest to the surface and the N-N bond distance.
The DFT cluster approach has been recently extended to the O,O2/ZrB2 system and preliminary results for
O chemisorbed on top of a Zr atom have been obtained by us.
B.2) Approximate semiempirical methods

For molecule-surface systems potential surfaces can be inferred following a semiempirical approach by
using the spectroscopic data available for the chemisorbed species, i.e. equilibrium geometries, adsorption
energies, inter and intra-molecular vibrational frequencies. The idea is to start with a parametric
interaction potential and to calculate the potential parameters such that the potential is able to reproduce
all the observed spectroscopic constants. This approach has been followed to derive a realistic interaction
potential for CO/Pt(111) [20], a system considered as a prototype in heterogeneous catalysis.
5-6 RTO-EN-AVT-142
Unfortunately, due to the lack of spectroscopic data for the O,O2/silica system, this approach cannot be
followed. However a tentative interaction potential can be constructed semiempirically on the ground of
reliable physical considerations. The method, developed to study the interaction of O with two silica
polymorphs [21,22], β-cristobalite and β-quartz, although quite crude has been successful in predicting the
recombination coefficient γ for the O+O/silica system at high temperature. As usual, the O-silica potential
is expressed as a sum of pair-wise interactions, assuming that the Ogas-Si interaction is the same as the O-
Si interaction in the lattice, while the interaction between the O atom in the gas-phase and the O of the
quartz surface is assumed as that known for the O-O interaction in the gas-phase. The validity of this latter
assumption is discussed later in the paper .Thus we have:
 NSi NO 
VG -S = ∑ ∑ BKS αj ∑
V (R ) + VO-O (R αj' ) + VC (4)
α =1,2 
 j=1 j' =1 
where the first term is the BKS potential [23] for the interaction of the gaseous O atoms and the Si lattice
atoms. Moreover, since the BKS potential at O-Si distances shorter than 1.5 Å is attractive, in order to get
the expected correct behaviour, the following potential is taken for the Ogas-Si interaction:
CO- Si   RO − Si − R0 
VO- Si = AO − Si e -bO - Si RO - Si - 1 + tanh  (5)
RO6 - Si   RO − Si 
where the repulsive and the attractive part of the BKS have been retained (and the Coulomb charge-charge
interaction dropped out). An appropriate switching function has been introduced in order to smooth the
repulsive potential to the C6 attractive interaction at the larger distances. R0 was set to 2.5 Å. The second
term in Eq. (4) is the interaction potential between the gaseous O atoms and the O lattice atoms and it is
given as:
VO-O = AO −O e -bO -O RO-O (6)

where the parameters AO-O and bO-O are those known from the O-O interaction in the gas-phase [21].
Finally, the last term in Eq. (4) represents the Coulomb interaction between the charge on the Si and O
lattice atoms and the charges of the molecular oxygen due to the quadrupole charge separation in O2.
Further details about these two last terms of the potential can be found in Ref. [21].
In Fig.5, the two-dimensional interaction potential is shown for O2 interacting in the perpendicular
orientation on top of a Si atom. The potential is plotted as a function of the O-O internuclear distance and
the distance of the closest O to the surface. In this configuration, the PES shows that the O2 formation
occurs without activation energy so that this configuration is very likely the most effective for the
recombination process.
RTO-EN-AVT-142 5-7
Fig. 5: Two-dimensional interaction potential for O2 interacting in the perpendicular orientation

on top of a Si atom of a β-quartz surface. The potential (in eV) is plotted as a function of
the O-O internuclear distance and the distance of the closest O atom to the surface.
C) RECOMBINATION DYNAMICS OF O ON ‘MODEL’ SILICA SURFACES
C.1) The semiclassical collisional method

In this section two aspects relevant to the MD approach to catalysis are discussed: firstly we briefly
discuss the semiclassical collisional model developed to describe the dynamics of the O atom
recombination and the O2 dissociation on silica, secondly the static properties of a silica surface are
examined. Compared to a fully classical MC simulation, a much detailed and physically coherent
description of the surface phenomena can be reached using the semiclassical time-dependent collisional
method [24]. This method is able to describe the most fundamental features of the molecule/surface
interaction: desorption (inelastic scattering) and sticking, diffusion, dissociation and atom recombination.
The following aspects can be incorporated in the model: lattice geometry, force constants, electronic
structure (band gap, Fermi level, etc).
The method is semiclassical in that, while the nuclear motions of the atoms/molecules approaching the
surface are treated classically, the vibrational motions of the lattice atoms and the electrons are treated
quantum-mechanically. Thus, the classical Hamilton’s equations of motion for the gas-phase species and
the time-dependent Schrödinger equations of motion for the phonons and the electrons of the substrate are
solved self-consistently. The time evolution of the phonons/electrons is dynamically coupled to the time
evolution of the chemical species through the definition and calculation of the effective Hamiltonian
Heff( t):
H eff = Ψ VI Ψ = H effph + H effe − h (7)
H effph is contribution due to the excitation/deexcitation processes of the surface phonons and it is defined
as the expectation value of the interaction potential over the total wave function of the phonon state at a
5-8 RTO-EN-AVT-142
given temperature. Similarly, the effective Hamiltonian due to the excitation/de-excitation of the electron
e−h
holes in the substrate, H eff , is given by averaging the coulomb potential over the wave function of the
electrons:
H effph = Ψ ph (t ) Vint (r , R ) Ψ ph (t ) H effe − h = Ψe − h (t ) VC (r , R ) Ψe − h (t ) (8)
where t is the interaction time, Vint(r, R) is the interaction potential between the gas-phase species and the
surface atoms determined as described in the previous section, Ψ ph is the total wave function of the
phonon state at a given surface temperature TS, VC(r,R) is the coulomb potential due to the interaction
between the electron charge in the solid and the charge temporally transferred to the molecule, Ψe− h is
the total wave function of the electrons perturbed by the external coulomb force.
The total hamiltonian for the molecular motion of a diatomic molecule is obtained by adding to the usual
hamiltonian for the free molecule the effective hamiltonian arising from the molecule-surface interactions:
1 Piα2
H cl =
2 ∑i ,α mi
+ V ( r ) + H effph + H effe − h + ∆E ph + ∆E e − h (9)
where P iα is the α-th cartesian component of the momentum of atom i and mi the corresponding mass.
∆E ph and ∆E e − h are the energy exchanged with the phonons and the electrons respectively, V(r) is the
intra-molecular potential of the gas-phase molecule.
The effective Hamiltonian H effph due to the phonon interaction can be obtained within the well known
linearly and quadratically forced harmonic oscillator model [24]. This leads to an analytical solution for
the time evolution of the phonon wave function and, consequently, for the excitation/de-excitation
ph
amplitudes of the phonons in each normal modes. Finally, an analytical expression of H eff is obtained in
terms of the ‘phonon excitation strengths’ ηk(t):
H effph = ∑V k
k
(1)
( R( t ))η k ( t ) (10)
where Vk(1) is the linear external force exerted on the k-th normal mode, R(t) is the position vector of the
incoming particle at t time. R(t) is the classical trajectory that is solution of the Hamilton's equations of
motion. ηk(t) are given in terms of the Fourier components of the forces exerted between the gaseous
species and the surface.
1 d ( ∆E ph )
η k ( t ) = − ∫ dt ′
=ωk dρ k
[
I c , k cos( ω k t ) + I s , k sin( ω k t ) ] (11)
+∞
I c ,k = ∫
−∞
dtVk( 1 ) ( R( t )) cos( ω k t ) (12)
ωk are the frequencies of the phonon normal modes.

The method outlined here has two important features: firstly, since the quantum phonon dynamics is
solved exactly both the single and the multi-phonon processes are accounted for in the collisional model,
and secondly the effective Hamiltonian depends explicitly upon the temperature of the solid. This later
effect is due to the energy symmetrization and to the use of a thermal phonon distribution instead of a pure
phonon eigenstate for the initial phonon state.
RTO-EN-AVT-142 5-9
A further important feature is the energy exchanged between the solid substrate (phonons and/or electrons)
and the chemical particles. In the case of surface processes assisted by the phonon excitation mechanism,
the exchanged energy is given as:
∆E ph = ∑∑∑ p
k nk nk0
nk0
( E nk − E n0 )Pn 0 →n =
k k k
∑ ( ∆E
k
+
k + ∆E k− ) (13)
where ( E nk − E n 0 ) is the energy exchanged in the transition nk ← nk0 between the quantum
k
state nk0 and nk of the k-th phonon mode. p n 0 is the Boltzmann distribution of the phonon energies,
k
Pn 0 →n is the transition probability.

k k
∆E k± is the energy loss (+) or gained (-) from the molecule due to the excitation/de-excitation phonon
processes in the substrate.
The method is formally rather involved, but nevertheless it is computationally affordable still providing a
realistic description of the gas-surface interactions in their full dimension. The method has been applied to
several heterogeneous systems, including the CO and CO2 formation on a platinum surface [25,26] and the
inelastic scattering of Xe atoms from a corrugated GaSe(001) surface [27].
C.2) Model silica surfaces and phonon dynamics
In order to perform MD calculations a sample of the solid substrate where the surface reactions occur must
be built up and incorporated in the collisional model. Model surfaces can be constructed starting from the
known arrangement of the Si and O atoms in the unitary cell of a specific silica polymorph [28], then the
unitary cell is repeated several times in the 3D space in order to have a sufficiently large sample. The
obtained surface can be considered as an ideal, perfect surface. In fact, while the bulk structure of real
surface is known, the atom arrangement on the top most layers is not due to the covalent character of
silica. Indeed, MD simulations of the static properties of silica have revealed a very complex structure of
the surface layers which appear reconstructed with structural and bond defects(see Fig. 6) [29].
Reconstructed surface
O
Si
Bulk
structure
Two-membered
rings
β-cristobalite:
Si-O bonding angle = 109.48°
Si-O bonding length = 1.54 Å
O-O distance = 2.5 Å
Fig. 6: A reconstructed surface is depicted.
5 - 10 RTO-EN-AVT-142
In addition to that, the chemical and structural properties of silica targets used in cell experiments, or in
real systems, can be strongly connected to the chemistry and the physics of the plasma environment. As a
consequence, the surface reactivity can be different under different operative plasma conditions. Since the
real structure of silica surfaces cannot be easily modelled, also due to the lack of structural data
determined from in situ observation, practically all the MD simulations are carried out on flat and perfect
surfaces. This poses some problems when the collisional data obtained in MD studies are compared with
those determined in cell experiments. On the other hand, MD data can be directly compared ( and
validated ) with data measured in molecular beam experiments, where a molecule/atom in a well defined
internal state hits a well defined single crystal surface (as in the MD simulations).
In Fig.7 the top view of the β-quartz and β-cristobalite surfaces used in recent MD calculations are shown.
Fig. 7: Top view of the β-cristobalite (left) and β-quartz (right)

surfaces recently used in MD calculations.
In order to describe the phonons dynamics of the silica surface a reliable interatomic potential is
necessary. Several interatomic potentials for silica polymorphs are available in the literature each of them
being able to reproduce on a quite good level different structural properties of the materials. The semi-
empirical BKS potential [23] is among the prime potentials used for silica polymorphs. The BKS potential
is a pair-wise potential given by:
-bij Rij C ij qi q j
V BKS (Rij ) = Aij e − 6
+ 14.4 (14)
R ij Rij
where the first term describes the purely repulsive interaction between two lattice atoms, i and j, Rij being
the distance between the two atoms. The second term is the C6 attractive dispersion term, while the last
term is the coulomb interaction between the charged Si and O atoms. Being the BKS potential a two-body
interaction potential with few parameters, the computational time needed to describe the dynamics of the
lattice phonons is lower than that required by other interaction potentials and its results in reproducing
various structural material properties are often better.
The phonon frequency spectrum,{ωk}, for the assumed silica surface can be deduced from the normal
mode analysis of the 3D lattice atom vibrations. The dynamical matrix H of the potential force constants is
constructed by calculating the second derivatives of the pair-wise BKS potential applied to the first nearest
atoms of the lattice
RTO-EN-AVT-142 5 - 11
 K iα , jβ   1 ∂ 2VBKS (Rij )
  =  (15)
 mij  i < j < N  mi m j ∂Riα ∂Riβ 
where Kiα,jβ stands for the inter-atomic force constants between the two nearest lattice atoms i and j, α
and β are the components of their respective cartesian coordinates. N is the total number of atoms in the
lattice. Then the secular equation is solved:
H − Eω = 0 (16)
and the eigenvalues {ωk} corresponding to the frequencies for the localized vibrational normal modes of
the lattice are obtained ( E is the unitary matrix).
In Fig. 8, the frequency distribution calculated for a model β-quartz surface is shown for a 3D crystal
consisting of 198 atoms displayed over 8 layers [22].
Density of phonon states
0
0 200 400 600
-1
Frequency (cm )
Fig. 8: Frequency distribution calculated for a 3D model crystal
of β-quartz consisting of 198 atoms displayed over 8 layers.
C.3) E-R and L-H Oxygen recombination dynamics on silica

The oxygen atom recombination on surfaces is generally described according to two extreme collisional
mechanisms, the so-called E-R and L-H mechanism. The E-R reaction is a two steps process, firstly the
oxygen is adsorbed at the surface in a specific surface site:
O + silica -------> Oad + silica (17)
Then the adsorbed oxygen reacts with an oxygen atom approaching the surface from the gas-phase.
O + Oad + silica -------> O2(v,j) + silica (18)
The first step is a thermal equilibrium process, the second reaction is the rate-determining step.
5 - 12 RTO-EN-AVT-142
According to the L-H mechanism the reaction occurs between two adsorbed atoms which propagate on the
surface from site to site and then, eventually, recombine.
Oad + Oad + silica -------> O2(v,j) + silica (19)
MD simulations have been performed for both reactions catalysed by β-cristobalite in the temperature
range from Ts=600K to Ts=1600K, using a semiempirical interaction potential determined according to
the basic considerations discussed in a previous section. Details can be found in Ref.s [21,30]. In Ref.[21]
the effect of first layer composition on the recombination reaction was also explored. In Fig. 9.a the E-R
recombination probability at Ts=1000K is reported for an oxygen covered and a clean β-cristobalite
surface, full dot and open dot respectively.
0.8 0.8
Recombination Probability
Reaction probability
0.6 0.6
0.4 0.4
0.2 0.2
0.0
0.0
4 56 0.01 2 3 4 56 0.10 2 3 4 56 1.00 2
0.001 2 3 4 5 0.01 2 3 4 5 0.1 2 3 4 5 7 1 2 3 45
Kinetic Energy (eV)
Kinetic Energy(eV)
9a 9b
Fig. 9: (a) The E-R oxygen recombination probability at Ts=1000K is reported for a clean (open dots)
and oxygen covered (full dots) model surfaces of β-cristobalite. (b) The E-R recombination
probability of O on β-cristobalite (full dot) and β-quartz (open dot) is reported as a function
of the impact energy of the gas-phase oxygen atom. The surface temperature is Ts=1000K.
In Fig.9b the E-R recombination probability of O on β-cristobalite (full dot) and on β-quartz (open dot) is
shown as a function of the impact energy of the gas-phase oxygen. The surface temperature is Ts=1000K.
In the trajectory simulations the oxygen atom sitting on the surface is in thermal equilibrium with the
surface and randomly adsorbed within the unit cell area, the initial Oad-surface distance being 2.Å. The
incoming gas-phase O atom approaches the silica surface in the perpendicular orientation, the initial (X,Y)
position coordinates are randomized within the unit silica cell and Z far in the asymptotic region ( Z=10 Å
at t=0 ).
From the calculated recombination probabilities the recombination coefficient γ(Ts) is calculated
assuming that a flux of oxygen atoms hits the surface with a maxwellian kinetic energy distribution
corresponding to the surface temperature.
The γ(Ts) coefficient is obtained by averaging the calculated recombination probabilities over a flux of O
atoms having a Maxwellian kinetic energy distribution at a given temperature (usually the temperature of
the surface):
∞ E kin
2 E kin − K BTS
γ E −R =
K B TS π
∫ PE−R (E kin ) K B TS
e dE kin (20)
0
where KB is the Boltzmann constant.
RTO-EN-AVT-142 5 - 13
The L-H reaction mechanism was also simulated and the recombination probabilities calculated for
different adsorption geometries of the two ad-atoms (see Ref. [21] for details). The L-H simulation was
performed at three different temperatures of the substrate: Ts=600K,800K,1000K.
The semiclassical recombination coefficient γ for the E-R and the L-H recombination is reported in Fig.
10, together with the high temperature experimental data determined recently by M.Balat et coll. on β-
cristobalite using the MESOX set up [31,32].
1000 500 330K

100
Recombination Coefficient γO
10-1
10-2
10-3
10-4
10-5
0.0 0.5 1.0 1.5 2.0 2.5
1/T(Kx1000)
Fig. 10 : The semiclassical recombination coefficient γ for the E-R (X ) and the L-H (+)
recombination reaction [21] is reported together with the experimental data
( ■Scott (silica-based materials)Ref. [3]; ● Greaves et al. (silica) Ref. [4] ; Berkut et al.
(silica-based materials) Ref. [2] ; ◊ Balat et al. (β-cristobalite) Ref. [31] ).
A cross-comparison between the experimental and the semiclassical results shows that a good agreement
is reached between the E-R γ values and the experimental values, whereas at the lower temperature the L-
H recombination probabilities match the experimental measurements. Note that a large surface
temperature effect is found for the L-H recombination process, whereas TS has only a weak impact on the
E-R reaction.
The theoretical γ values should not be directly compared with the cell measurements (see the previous
discussion on this point), nevertheless the agreement between theory and experiments is very encouraging.
The reliability of the semiclassical approach followed in the MD simulation has been demonstrated in a
combined experimental and theoretical investigation [22] recently carried out on the catalytic activity of
the β-quartz. In the table below the theoretical and experimental recombination coefficient γ for oxygen
recombination on β-quartz are reported at TS=1000K and compared with the corresponding data for β-
cristobalite.
5 - 14 RTO-EN-AVT-142
SURFACE (TS=1000K) γth γexp

β-quartz 0.008 0.008
β-cristobalite 0.03 0.027
Such agreement gives support to the E-R recombination mechanism as the mechanism followed in the
recombination reaction process monitored in the experiments and gives further evidence of the large effect
the surface crystallographic structure has on the catalytic activity of silica-based surfaces.
It is worth noticing that the semiclassical recombination coefficients reported in the table above were
obtained assuming that in the initial conditions the O ad-atom is randomly adsorbed within the unit cell.
The same trend is observed when the O ad-atom is chemisorbed in a specific silica site, on top of a Si
surface atom. In this case in fact we get γ= 0.04 and γ=0.3 for O recombination on β-quartz and β-
cristobalite at Ts=1000K, respectively. That is, although the recombination coefficients are higher than
that measured in the experiments, the ratio γ(cristobalite)/γ(quartz) is about a factor two higher compared
to the ratio observed in the experiments.
The semiclassical MD calculations have revealed some aspects that cannot be observed directly in the
experiments. The dynamics is, in fact, quite complex with several collisional processes that are in
competition with the O2 formation at the surface.
O + Oads *silica -------> O2(v, j)+silica recombination (21)
Ogas + Ogas +silica inelastic
Oads *silica+ Ogas adsorption/desorption 

Ogas + Oads *silica
[O2]ads adsorption
Oads + Oads + silica
In particular, the adsorption and adsorption/desorption processes of the oxygen atoms are quite effective in
a large collisional energy range. This is shown in Fig.11 where the probabilities for the different surface
processes active on β-cristobalite are reported at TS=1000K.
RTO-EN-AVT-142 5 - 15
Inelastic
Process Probability 0.8
Ads/Des
0.6
0.4
0.2 Ads
0.0
4 56 0.01 2 3 4 567 0.1 2 3 4 567 1
Kinetic Energy(eV)
Fig. 11: Probability for the different surface processes of Eq. (21) on β-cristobalite. TS=1000K.
In the full range of the explored energies the adsorption and adsorption/desorption processes are quite
effective. This is an interesting result that could be of some relevance for the surface damage problem. The
residence time for these processes is large so that chemical reactions between the adsorbed atoms and the
surface Si/O atoms can be eventually be promoted (with the formation of SiOx volatile compounds).
In Fig. 12 the probability for the different reaction channels of Eq. (21) is reported but for the interaction
of N atoms with β-cristobalite. In the initial conditions the N ad-atom is chemisorbed in a specific
chemisorption site, that is on top of a Si surface atom, in thermal equilibrium with the silica surface at
Ts=1000K. The gas-phase N hits the surface in the normal configuration, with the (X,Y) coordinates
randomly varied within the unit silica cell of the β-cristobalite. The potential energy surface assumed in
the dynamics is the LEPS surface of Fig. 4 obtained from DFT calculations. A further aspect of great
relevance for the surface heating problem concerns the energetics of the recombination process. Reactions
(18), (19) are both strongly exothermic so that part of the exothermic energy can be gained by the formed
O2 molecules while the remaining part can be transferred to the substrate as heat flux. The energy flux
pathways are shown in Fig 13. Here the energy partitioning among the translational (Etr), rotational (Erot)
and vibrational motions (Evib) of the O2 molecules and the surface phonons (∆Eph) is shown as a function
of the impact kinetic energy for oxygen formation on β-cristobalite at Ts=1000K (a) and Ts=1600K,
respectively.
5 - 16 RTO-EN-AVT-142
1.0 Inel
Ads/Des
0.8
Probability
0.6
0.4
0.2
Ads
0.0
0.01 2 3 4 5 6 7 80.1 2 3 4 5 6 78 1 2
Kinetic Energy (eV)

Fig. 12: The same as for Fig.11 but for N + Nad on β-cristobalite. Nad is initially adsorbed
on top of a Si surface atom. The PES of Fig. 4 was assumed in the dynamics.
0.75 0.75
0.60 Evib 0.60 Evib
0.45 0.45
Energy Fraction
Energy Fraction
Erot
Erot
0.30
0.30 ∆Eph
Etr
0.15 0.15
∆Eph
0.00 0.00
4 5 6 0.01 2 3 4 5 67 0.1 2 3 4 5 67 2 3 45 4 56 0.01 2 3 4 5 67 0.1 2 3 4 5 6 78 1

1
Kinetic Energy(eV) Kinetic Energy(eV)
13a 13b
Fig.: 13: Energy partitioning among the translational (Etr), rotational (Erot), vibrational (Evib)
motions of the O2 molecules and the surface phonons (∆Eph) as a function of the impact kinetic
energy in the reaction: O + Oad *silica ---> O2(v,j) + silica . (a) Ts=1000K and (b) Ts=1600K.
It is evident that the largest fraction is primarily shared among the vibrational and rotational motion of O2,
while only a small fraction of the total energy is transferred to the silica surface as phonon excitation
energy. However, at the higher energies and Ts=1000K almost 30% of the exothermic energy appears as
RTO-EN-AVT-142 5 - 17
phonon energy. Interesting different behaviours of the reaction energetics can be pointed out by comparing
the results reported at the two different surface temperatures. As a consequence of the energy sharing
mechanism, the O2 molecules are formed in highly vibrationally excited state.
This is shown in Fig.14 where a typical vibrational distribution is reported for the recombination reaction
over β-cristobalite at Ekin= 0.003eV, Ts=1000K.
0 .1 5
0 .1 2
Probability
0 .0 9
0 .0 6
0 .0 3
0 .0 0
0 4 8 12 16 20 24 28 32 36
V ib ra tio n a l Q u a n tu m N u m b er
Fig.14: A typical vibrational distribution of the O2 molecules formed after
atom recombination over β-cristobalite at Ekin= 0.003eV and TS=1000K.
The distributions are characterized by a peak in v=0, but virtually all the vibrational levels are populated
and involved in the surface process. The distributions are in non-thermal equilibrium.
From the calculated vibrational distribution, the vibrational state-selected recombination coefficient
γ(v,Ts) can be easily obtained. The spectral distribution of γ for O recombining on β-cristobalite at
Ts=1000K is shown in the Fig.15. The γ(v) coefficients are an important ingredient for the kinetic
modelling of the chemistry of the boundary layer [33-34] where, very likely, a non-thermal equilibrium
condition is established due to the fluodynamics conditions.
5 - 18 RTO-EN-AVT-142
γ(v,TS=1000K) 0.06
0.04
0.02
0.00
0 10 20 30
Vibrational Quantum Number
Fig.15: The vibrational spectral distribution of the recombination coefficient γ for O2(v)
formed in the reaction: O + Oad *silica ---> O2(v,j) + silica is reported as
a function of the vibrational quantum number v. TS=1000K.
The energy transferred to the surface in the recombination processes is generally claimed to be among the
primary sources of the surface damage. Molecular dynamics simulations for O ( and N) recombination on
silica surfaces have shown that this energy is only a small fraction of the total energy delivered in the
reaction, so that the energy accommodation coefficient β should be less than one ( contrary to what is
generally assumed in kinetic modelling calculations, i.e β=1 ). Indeed, the overall energy transferred to the
substrate is the result of a complex mechanism and it is due to the concurrence of various surface
processes. The adsorption processes can play an important role since the energy exchanged with the
surface can be quite large. In Fig. 16 a typical adsorption collision:
O + Oad + silica -------> [O2]ad + silica-------> O + Oad + silica/β-cristobalite Ts=1000K
is reported together with the corresponding energy transferred to the silica substrate.
RTO-EN-AVT-142 5 - 19
Atomic distance from the surface (Å)
2.0
O gas
6
1.6 O + O ad
1.2
4
∆Eph (eV)
0.8
O ad
2 (O 2 ) ad
0.4
0 0.0
10 30 50 10 30 50
Interaction T im e ( 10
-14
s) Interaction tim e (10 -14 s)
Fig. 16: A typical collision for the adsorption process: O + Oad + silica --->[O2]ad + silica--->
O + Oad + silica/β-cristobalite is reported together with the corresponding energy
transferred to the silica substrate. a) the z coordinate of the two oxygen atoms
is reported as a function of the collisional time. b) the corresponding
energy transferred to the silica surface is shown. TS=1000K.
FINAL REMARKS
The development and applications of the semiclassical approach to the molecule-surface dynamics have
made possible MD simulations of elementary surface processes involving oxygen and nitrogen atoms over
silica model surfaces relevant to the TPS system. In the simulations various important behaviours of the
surface substrate, such as the phonon structure, the surface polymorphic modifications, the site and the
intrinsic surface temperature effect can be incorporate and their impact on the surface catalytic activity
studied at a very elementary level. Collisional data, such as the global and the vibrational state selected
recombination coefficient γ, the energy distributions in the product states, the energy flow pathways and
the energy released to the surface in the recombination reaction as well as in the adsorption/desorption
processes can be determined. The accuracy of the calculated coefficients mostly relies on the accuracy of
the interaction potential assumed in the dynamics, rather than to the approximations made in the
collisional model. As a matter of fact, the determination of a reasonable PES for these complex
heterogeneous systems is the most critical problem to solve. MD simulations have some limitations,
among the others: i) the surface processes can be followed in the short time scale (few picoseconds). ii)the
surface inserted in the simulations is, in general, a flat perfect crystal surface, iii) the dynamics occurs on
an adiabatic potential surface so that surface processes involving electronically excited states are not
explored. Though the inclusion of these effects add more complexity and computational time
requirements, the importance of MD calculations for the catalytic systems candidate for the future TPS
systems will become more evident in the next future.
5 - 20 RTO-EN-AVT-142
REFERENCES
[1] M.C.Lin and G.J.Ertl Ann.Rev.Phys.Chem. 37 (1986) 587; G.J.Ertl Catal.Rev.-Sci.Eng. 21 (1980)
201
[2] V.D. Berkut, V.M. Doroshenko, V.V. Kovtun, N.N. Koudryavtsev, S.S. Novikov, N.V. Smirnov,
A.I. Sharotovov, Soviet J. Chem. Phys. 9 (1992) 2222
[3] C.D. Scott, AIAA Paper 83-1485 (1983).
[4] J.C. Greaves, J.W. Linnett, Trans. Faraday Soc., 55 (1959) 1355
[5] E. J. Jumper, W.A. Seward, J. Thermophys. & Heat Transfer 8 (1994 ) 460
[6] F. Nasuti, M. Barbato, C. Bruno, J. Thermophysics & Heat Transfer 10 (1996) 131
[7] V. L. Kovalev, “Phenomenological treatment of heterogeneous recombination”, these VKI Lecture

Notes
[8] A.E.Depristo and A.Kara Adv.Chem.Phys. 77 (1990) 163
[9] G.P.Brivio and T.B.Grimley Surf.Sci.Rep.17(1993) 1
[10] A.Groβ : these VKI Lecture Notes
[11] R. Dovesi, et al CRYSTAL 95 User's Manual’, ( University of Torino, Torino, 1996).
[12] E.g., see, K. G. Nakamura, Chem. Phys. Lett. 285 (1998) 21.
[13] W. Koch, W.G. Holthousen, ‘A chemist’s guide to density functional theory’, (Wiley-VCH,
Weinheim, 2000).
[14] 14) M. Rutigliano, A. Pieretti, M. Cacciatore, N. Sanna and V. Barone, “N atoms recombination on
a silica surface: a global theoretical approach”, Surf. Sci. 2006
[15] C. Lee, W. Yag and R. G. Parr, Phys. Rev. 37 (1988) 785.
[16] Gaussian 03 (Revision C.02), M. J. Frisch, et al, Gaussian, Inc., Wallingford CT, 2004.
[17] “Spectra of Diatomic Molecules”, Van Nostrand, Princeton, NJ, 1950; K.P. Huber and G. Herzberg,
“ Molecular Spectra and Molecular Structure, IV. Constants of Diatomic Molecules”, Van Nostrand,
New York, 1979.
[18] J. McCreery and G. Walken, J. Chem. Phys. 63 (1975) 2340
[19] B. Jackson and M. Persson, J. Chem. Phys. 96 (1992) 2378
[20] M.Cacciatore, E.Christoffersen, M.Rutigliano J. Phys. Chem. A, 108 (2004) 8810
[21] M.Cacciatore,M.Rutigliano and G.D.Billing , J. Therm,&Heat Transfer. 13 (1999)195
[22] L. Bedra, M. Rutigliano, M. Balat-Pichelin, M. Cacciatore, “Atomic Oxygen Recombination on

Quartz at High Temperature: Experiments and MD Simulation”, submitted
RTO-EN-AVT-142 5 - 21
[23] Van Beest B. W. H., Kramer G. J., Van Santen R. A., Phys. Rev.Lett., 64 (1990)1955.
[24] G.D.Billing "Dynamics of Molecule Surface Interactions", (Wiley & Sons, N. Y., 2000).
[25] G. D. Billing and M. Cacciatore, Chem. Phys. Lett. 113 (1985) 23
[26] G.D.Billing, M.Cacciatore Chem.Phys.103 (1986) 137
[27] S.Iannotta, C.Gravili,A.Boschetti,A.Cagol and M.Cacciatore Chem.Phys. 194 (1995)133
[28] R.W.G.Wickoff American Journal Science 448 (1925) 5
[29] B.P. Feuston and S. H. Garofalini, J. Chem. Phys. 91 (1989) 564
[30] M. Balat-Pichelin, J. M. Badie, M.Cacciatore, M.Rutigliano, ESA SP-521 (2003) 403
[31] M. Balat-Pichelin, J.M. Badie, R. Berjoan, P. Boubert, Chem. Phys. 291 (2003) 181
[32] M.Balat-Pichelin these VKI Lecture Notes
[33] I. Armenise, M. Capitelli, C. Gorse, M. Cacciatore and M. Rutigliano, JSR 37 (2000) 318
[34] D. Bruno, M. Cacciatore, S. Longo, M. Rutigliano, Chem. Phys. Lett. 320 (2000) 245
5 - 22 RTO-EN-AVT-142
Detailed and simplified kinetic schemes for high
enthalpy air flows and their influence on catalycity
studies
Anne Bourdon
Laboratoire EM2C UPR 288 CNRS
Ecole Centrale Paris, Grande voie des vignes
92295 Châtenay-Malabry Cedex, France
email : Anne.Bourdon@em2c.ecp.fr
Arnaud Bultel
UMR CNRS 6614 CORIA
Université de Rouen - Site universitaire du Madrillet - BP12
76801 Saint-Etienne du Rouvray cedex, France
email : Arnaud.Bultel@coria.fr
February 8, 2006
Contents
1 Abstract 3
2 Introduction 3
3 Collisional-radiative model for O atoms 4

3.1 Atomic model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 Elementary rate coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2.1 Electron-impact excitation and deexcitation rate coefficients . . . . 7
3.2.2 Electron-impact ionization and three-body recombination rate co-
efficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2.3 Spontaneous emission . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2.4 Radiative and dielectronic recombinations . . . . . . . . . . . . . . 8
3.3 Master equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4.1 General behaviour of time-dependent population densities . . . . . 9
3.4.2 General method to determine effective ionization and recombination
rate coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Bourdon, A.; Bultel, A. (2007) Detailed and Simplified Kinetic Schemes for High Enthalpy Air Flows and Their Influence on Catalycity
Studies. In Experiment, Modeling and Simulation of Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 9-1 – 9-60).
Educational Notes RTO-EN-AVT-142, Paper 9. Neuilly-sur-Seine, France: RTO. Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 9-1
Detailed and Simplified Kinetic Schemes for High
Enthalpy Air Flows and Their Influence on Catalycity Studies
3.4.3 Determination of effective three-body recombination rate coefficients 12

3.4.4 Determination of the effective ionization rate coefficient . . . . . . . 15
3.4.5 Study of the quasi-steady-state relaxation time . . . . . . . . . . . 17
3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4 Collisional-radiative model for air 19

4.1 Atomic and molecular model . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Elementary processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.1 Radiative processes . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.2 Collisional processes . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.3 Dissociative recombination of N2+ , O2+ and N O+ . . . . . . . . . . . 23
4.2.4 Vibrational processes . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3.1 Ionization situation . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3.2 Recombination situation . . . . . . . . . . . . . . . . . . . . . . . . 34
4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5 Stagnation point boundary layer 38

5.1 Stagnation point boundary layer code . . . . . . . . . . . . . . . . . . . . . 38
5.1.1 Catalycity and boundary conditions . . . . . . . . . . . . . . . . . . 40
5.1.2 Thermochemical model . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.1.3 Numerical method . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2 Catalycity models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2.1 Simplified models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2.2 Phenomenological models . . . . . . . . . . . . . . . . . . . . . . . 42
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.3.1 High pressure test case : P = 10000P a . . . . . . . . . . . . . . . . 47
5.3.2 Low pressure test-case : P = 2000P a . . . . . . . . . . . . . . . . . 50
6 Conclusions 52
9-2 RTO-EN-AVT-142
1 Abstract
The thorough understanding of the formation and the relaxation of the plasma produced
in the shock layer developed during the re-entry of a spacecraft in the upper layers of the
earth’s atmosphere is crucial in order to prevent damaging of its outer surface. Among
the different points to be studied, the chemical aspects are particularly important: the
mechanical characteristic time scale of the flow being short, the flow is indeed in chemical
nonequilibrium. In addition, the inner storage of energy of the different species of the
flow leads to other mechanical behaviors than those observed in classical low temperature
flows. These nonequilibrium effects have to be taken into account accurately in codes
devoted to the study of the interaction between the plasma and the surface.
In this paper, we propose to focus our attention on the detailed kinetics of an air plasma
under nonequilibrium conditions. First, we present a time-dependent collisional-radiative
(CR) model for atomic oxygen to study in detail ionization and three-body recombination
rates of oxygen in high temperature air plasma flows. Second we present a time-dependent
CR model for air taking into account 13 species and numerous excited states and working
over a wide range of pressure and temperature. For typical conditions encountered in re-
entry flows, we compare this detailed CR scheme for air with simplified schemes (proposed
by Park, Dunn & Kang and Gupta et al.) usually implemented in high enthalpy air flow
codes. Finally, a 1D code simulating the stagnation point boundary layer near the wall of
the spacecraft is presented and the influence of the gas chemistry on the wall catalycity
is discussed.
2 Introduction
During the re-entry of a spacecraft into the upper layers of the atmosphere of a planet, the
flow is hypersonic and a shock layer is formed in front of the vehicle (Anderson (1989)).
In crossing the shock layer, the flow is hardly slowed down and the gas is strongly heated
(up to 15000 K (Carlson and Hassan (1992))) and therefore becomes a plasma. Near the
surface, a boundary layer is formed due to the difference between the temperature inside
the shock layer and the wall temperature and this induces high energy flux densities to the
vehicle surface. These fluxes have to be reduced to prevent damaging of the spacecraft.
To choose correctly the material of the thermal shield it is necessary to study the thermal
interaction between the gas medium and the vehicle (Hankey (1994)) and therefore to
study thoroughly the plasma formed in the shock layer (i.e. different species concentra-
tions close to the wall, their internal degree of excitation, etc). In the present context,
the flow is far from equilibrium; after an ionization phase, the plasma is recombining as it
approaches the thermal shield and the velocity is high. As a result, the chemistry is time-
dependent in the frame of the moving plasma. Its calculation needs an accurate kinetic
scheme. Moreover, as the pressure varies over a wide range during the re-entry phase,
the kinetic scheme has to take into account a sufficient number of binary and ternary
collisions. Currently, despite the large number of experimental and theoretical studies on
high temperature air chemistry, uncertainties remain on the accuracy of numerous reac-
tion rates.
In particular, the effective ionization and recombination rates of atoms are difficult to
RTO-EN-AVT-142 9-3
determine experimentally and, then most available results have been derived from theo-
retical studies. However, a reaction such as O+ + 2e− ↔ O + e− results from numerous
different elementary processes between the atomic levels of oxygen. Therefore, the ac-
curacy of theoretical effective ionization and recombination rate coefficients depends on
the choice of the atomic model and on the choice of the rate coefficients for the elemen-
tary processes. In section 3, we present a time-dependent collisional radiative model for
atomic oxygen in order to determine effective ionization and recombination rate coeffi-
cients for 1018 m−3 ≤ ne ≤ 1021 m−3 and 8000K ≤ Te ≤ 20 000K. We will show that a
time-dependent approach allows to follow the relaxation of atomic level populations and
to study the validity of the quasi-steady-state hypothesis.
As far as we know, no time-dependent collisional-radiative model dedicated to re-entry
applications exists for air, taking into account N2 , O2 , N O, N , O, N2+ , O2+ , N O+ , N + ,
O+ , O2− , O− on their ground state and on numerous electronic excited states. Although
some collisional-radiative models have been earlier elaborated by Teulet et al. (2001) and
Sarrette et al. (1995), no model works between 1 kP a and the atmospheric pressure. In
addition, considerable improvements have been made about data in term of excitation by
electron impact, dissociative recombination of N O+ , O2+ and N2+ and their branching frac-
tions as well as vibrational processes. In section 4, we propose to focus on two important
points in the case of the plasma formed during the re-entry into the earth’s atmosphere:
(1) the time-dependent chemistry of an ionizing or recombining air plasma and (2) the
comparison with existing simplified kinetic schemes used up to now in Navier-Stokes codes.
This study is based on the elaboration of a time-dependent collisional-radiative model for
heavy particles temperature TA and electron temperature Te less than 10000 K. The re-
sults of the elaborated CR model are finally discussed for different re-entry conditions.
Finally, in section 5, we present the influence of the choice of the kinetic scheme for air on
the wall heat flux on a TPS material in the stagnation point boundary layer for different
conditions.
3 Collisional-radiative model for O atoms

In this section, we propose to set up a collisional radiative model for atomic oxygen in
order to determine effective ionization and recombination rate coefficients for 1018 m−3 ≤
ne ≤ 1021 m−3 and 8000K ≤ Te ≤ 20 000K. In an earlier study, Taylor and Ali (1986) de-
termined these rate coefficients for slightly different conditions: 1022 m−3 ≤ ne ≤ 1025 m−3
and 1eV ≤ Te ≤ 3eV . Other previous collisional radiative models for oxygen have been
devoted to the study of the population densities at the stationary state (Gomès et al.
(1990); Soon and Kunc (1990)) and relaxation times (Cacciatore and Capitelli (1976)).
Since the study of Taylor and Ali, numerous experiments and computations have been
carried out to improve the accuracy of atomic cross sections and Einstein coefficients in
oxygen. In this work, we have tried to determine the best possible coefficients available in
the literature. The data critically compiled by Soon and Kunc (1990) have been used with
a few updates on e−O inelastic collisions (Doering (1992)) and on Einstein coefficients
(Biémont and Zeippen (1992); Bhatia and Kastner (1995)). In this study, we consider
a spatially uniform and electrically neutral plasma. We assume that the distributions
of energies of particles are Maxwellian and that the atom-atom and ion-atom inelastic
9-4 RTO-EN-AVT-142
collisions are negligible. The treatment of radiation is limited to either optically thin or
optically thick cases.
Unfortunately, no experimental data are available on the effective ionization and recombi-
nation rates of atomic oxygen to compare with our results. In high temperature air kinetic
schemes used in plasma flow codes, different analytical expressions, which accuracy is un-
known, are proposed for these rate coefficients (Park (1990); Gupta et al. (1990); Losev
et al. (1994)). Therefore, this work is the opportunity to study the validity of these simple
analytical expressions for the effective ionization and recombination rates of atomic oxy-
gen. Furthermore, the time-dependent approach used in this paper allows us to follow the
relaxation of atomic level populations and to study the validity of the quasi-steady-state
hypothesis.
3.1 Atomic model

In this work, forty coalesced energy levels of the neutral oxygen atom have been considered
(Moore (1993)). Their characteristics (for each level index: level configuration, energy
and statistical weight) are given in Table 1. As in the nitrogen atom, there is a large
energy gap in the oxygen atom between the two first metastable levels (2s2 2p4 1 D and
2s2 2p4 1 S) and the other excited levels which lie in a narrow region of energy. However,
it is interesting to note that in the oxygen atom, the resonant level is only the fifth
level, located at 9.52 eV from the ground state. In their atomic model for oxygen, Soon
and Kunc (1990) considered only the nine first levels of our model. In this case, the
last level lies 1.69 eV under the ionization limit. With our model, this energy gap is
only of 0.124 eV. The influence of the atomic model on the results will be discussed in
the following sections. In the electron temperature range of our study, we consider only
the ground ionic term of O+ , 2s2 2p3 4 S (13.62 eV) which is taken as the ”core” for all
atoms. Processes involving O− ions could a priori also influence the determination of the
ionization and recombination rate coefficients. However, Soon and Kunc (1991) showed
that for Te ≥ 8000K, the contribution of O− ions to production of electrons, positive ions
and excited atoms is negligible. Therefore, in this study, the influence of O− ions has
been neglected.
3.2 Elementary rate coefficients

In a uniform plasma, the evolution of the population density of an excited atomic level i
is due to different elementary processes
Oi + e−
Oj + e− for i < j [Cij , Cji ] (1)
where Cij (m3 s−1 ) and Cji (m3 s−1 ) are, respectively, the electron-impact excitation rate
coefficient for the transition from the level i to the level j and its inverse deexcitation rate
coefficient
Oi + e−
O+ + 2e− [βic , βci ] (2)
where βic (m3 s−1 ) and βci (m6 s−1 ) are, respectively, the electron impact ionization rate
coefficient from the level i and the three-body recombination rate coefficient on the same
level
Oj −→Oi + hν for i < j [Aji ] (3)
RTO-EN-AVT-142 9-5
i 1s2 + Ei (eV ) gi
1 2s2 2p4 (3 P ) 0.000 9
2 2s2 2p4 (1 D) 1.970 5
3 2s2 2p4 (1 S) 4.190 1
4 2s2 2p3 3s(5 S o ) 9.150 5
5 2s2 2p3 3s(3 S o ) 9.520 3
6 2s2 2p3 3p(5 P ) 10.740 15
7 2s2 2p3 3p(3 P ) 10.990 9
8 2s2 2p3 4s(5 S o ) 11.840 5
9 2s2 2p3 4s(3 S o ) 11.930 3
10 2s2 2p3 3d(5 Do ) 12.090 25
11 2s2 2p3 3d(3 Do ) 12.100 15
12 2s2 2p3 4p(5 P ) 12.300 15
13 2s2 2p3 4p(3 P ) 12.370 9
14 2s2 2p3 3s0 (3 Do ) 12.550 15
15 2s2 2p3 5s(5 S o ) 12.670 5
16 2s2 2p3 5s(3 S o ) 12.710 3
17 2s2 2p3 3s0 (1 Do ) 12.740 5
18 2s2 2p3 4d(5 Do ) 12.760 25
19 2s2 2p3 4d(3 Do ) 12.770 15
20 2s2 2p3 4f (5 F, 3 F ) 12.780 56
21 2s2 2p3 5p(5 P ) 12.860 15
22 2s2 2p3 (3 P ) 12.890 9
23 2s2 2p3 6s(5 S o ) 13.030 5
24 2s2 2p3 6s(3 S o ) 13.050 3
25 2s 2p3 5d(5 Do , 3 Do )
2
13.080 40
26 2s2 2p3 5f (5 F, 3 F ) 13.087 56
27 2s2 2p3 6p(5 P ) 13.130 15
28 2s2 2p3 6p(3 P ) 13.140 9
29 2s2 2p3 7s(5 S o ) 13.220 5
30 2s2 2p3 7s(3 S o ) 13.230 3
31 2s 2p3 [6d(5 Do , 3 Do )
2
+6f (5 F, 3 F )
+6g(5 Go , 3 Go )] 13.250 168
32 2s2 2p3 8s(5 S o ) 13.330 5
33 2s2 2p3 8s(3 S o ) 13.340 3
34 2s 2p3 [7d(3 Do , 5 Do )
2
+7d(5 F, 3 F )] 13.353 96
35 2s2 2p3 9s(3 S o ) 13.412 3
36 2s2 2p3 8d(5 Do , 3 Do ) 13.418 40
37 2s2 2p3 10s(3 S o ) 13.459 3
38 2s2 2p3 9d(3 Do ) 13.464 15
39 2s2 2p3 11s(3 S o ) 13.493 3
40 2s2 2p3 10d(3 Do ) 13.496 15
41 (O+ ion) 2s2 2p3 (4 S o ) 13.620 4
Table 1: Atomic level model.
9-6 RTO-EN-AVT-142
where Aji (s−1 ) is the transition probability (Einstein coefficient) from the level j to the
level i; and
O+ + e− −→Oi + hν
RR RD
αi , αi (4)
where αiRR (m3 s−1 ) and αiRD (m3 s−1 ) are, respectively, the radiative and dielectronic
recombination rate coefficients on the level i. Dielectronic recombination is the result of
a two-stage process
O+ + e− −→ O∗ −→ Oi + hν (5)
where O∗ represents an autoionizing state of the O atom.
3.2.1 Electron-impact excitation and deexcitation rate coefficients
Currently, no complete set of precalculated cross sections exists for oxygen. Therefore,
for low atomic levels, measurements of cross sections have been used, when available.
For the 1 → j ≤ 7 transitions, the cross sections measured by Doering et al. ( Doering
and Gulcicek (1989b), Doering and Gulcicek (1989a), Gulcicek et al. (1988) and Gulcicek
and Doering (1988)) have been used in agreement with the discussion of Soon and Kunc
(1990). For the 1→2 transition, it is interesting to note that the shape and magnitude of
the cross section have been confirmed recently by Doering (1992). For the 1→ 9≤ j ≤
19 transitions, the theoretical and experimental cross sections reviewed by Laher and
Gilmore (1990) have been used. For all the other transitions, excitation cross sections
have to be determined theoretically. In this work, we have used for low atomic levels the
Vainshtein formalism which is certainly more adapted than hydrogenic models retained in
earlier studies. It would be interesting to use cross sections obtained with more accurate
quantum mechanical methods such as close-coupling and R-matrix methods. But these
methods are computer intensive, and have been completed, up to now, only for a few
systems (Pradhan (1988)). For ”difficult” transitions (i.e., involving two atomic electrons),
a binary-encounter approximation is used. All details are given in the paper of Soon and
Kunc (1990). For atomic levels close to the continuum, all atoms are almost hydrogenic.
Therefore, the formalism initially developed for hydrogen by Gryzinski (1965) is used for
these levels, and is no doubt more adapted than the Vainshtein formalism.
The use of different theoretical models may result in abrupt changes in the elementary
rate coefficient values. In order to eliminate this anomaly, already mentioned in other
plasma studies (Park (1971)) we introduce, in the same way, a correction factor such
that elementary rate coefficient values vary smoothly from one model to the other. As
an example Fig. 1 shows the rate coefficient for the 1 → j transition. Reliable rate
coefficients based on the review of Laher and Gilmore (1990) are available up to j = 19.
Gryzinski rate coefficients up to j = 19 have higher values but we note that in both cases,
the rate coefficient decreases rapidly up to j = 10 and decreases more slowly for higher
values. Therefore, in our model, rate coefficients for 1 → j > 19 are derived using the
Gryzinski rate coefficients scaled using the rate coefficient of the 1 → 19 transition. We
have checked that discrepancies on the effective ionization and recombination rate values
due to this correction remain within a factor of 2.
RTO-EN-AVT-142 9-7
−11
10
our model
−13
10 Gryzinski model
rate coefficients available
Rate coefficient [m s ]
3 −1
−15
10
−17
10
−19
10
−21
10
−23
10
0 10 20 30
j level
Figure 1: Rate coefficients of the 1 → j transitions. Open circles:

theoretical and experimental rate coefficients reviewed by Laher and
Gilmore (1990). Dashed line: Gryzinski (1965) rate coefficients.
Solid line: our model.
3.2.2 Electron-impact ionization and three-body recombination rate coeffi-

cients
For the ground state and the two lowest excited levels, the models retained by Soon and
Kunc (1990) are used. For the other atomic levels, we have used the classical model
approximation derived by Gryzinski and Kunc (1986).
3.2.3 Spontaneous emission

For excited levels the energy of which is less than or equal to 11.93 eV, the transitions and
the Einstein coefficients reviewed by Soon and Kunc are used. However, for the 5→1 and
9→1 transitions, the Einstein coefficients determined recently by, respectively, Biémont
and Zeippen (1992), and Bhatia and Kastner (1995) have been used. For transitions
issuing from levels lying higher than 11.93 eV, the best possible Einstein coefficients
available in the literature have been used (Bhatia and Kastner (1995); Fuhr and Wiese
(1990); Tayal and Henry (1989); Doering et al. (1985); Verner et al. (1996)).
3.2.4 Radiative and dielectronic recombinations

The analytical expressions proposed by Nussbaumer and Storey (1983) are used to cal-
culate the dielectronic recombination coefficients for the effective (direct plus cascade)
transitions to the terminating levels i =1 and 4. The total dielectronic recombination
rate is X
αRD = αiRD (6)
i=1,4
As in nitrogen, radiative recombination coefficients are significant only for the effective
transitions to the three first atomic levels. Using the same approach as Soon and Kunc
9-8 RTO-EN-AVT-142
(1990), we have calculated and fitted the associated total radiative recombination rate,
and obtained
X3
RR
α = αiRR = 2.12 × 10−18 Te−0.29 m3 s−1 (7)
i=1
3.3 Master equations

Taking into account the different elementary processes mentioned above, the rate equa-
tions for the production of an excited atom on the level i can be written as
dNi X X
Nj Aji κji + ne N + αiRR κRR RD RD

= Ṅi = ne Nj Cji + i + αi κi + n e βci
dt j6=i j>i
" #
X X
−Ni ne Cij + Aik κik + ne βic (8)
j6=i k<i
for 1≤ i ≤ np, where np is the total number of atomic levels, ne (m−3 ) the electron
number density, Ni (m−3 ) the population density of the atomic level i, and N + (m−3 ) the
number density of O+ . For the escape factors denoted κij , κRRi and κRD
i , respectively,
for bound-bound, free-bound, and dielectronic radiation, we consider only the limit cases
where they are equal either to 0 (optically thick plasma) or to 1 (optically thin plasma).
The rate equation for the production of electrons is
dne X X X
n2e N + βci − ne N + αiRR κRR + αiRD κRD

= n˙e = ne Ni βic − i i (9)
dt i i i
Finally, the electrical neutrality of the plasma imposes ne = N + . We denote Na = Ni ,

P
the total atomic density and Nt = Na + N + , the total particle density. The system of
differential equations [Eqs. (8) and (9)] has been integrated numerically in using LSODE
Hindmarsh (1980). In each set of calculations, Te and Nt are constant.
3.4 Results
3.4.1 General behaviour of time-dependent population densities
Figure 2 shows the temporal evolution of the level populations for a typical recombining
situation at t = 0 s: ne = 1022 m−3 , Te = 12 000K, and bi = Ni /Ni∗ = 10−4 , where Ni∗
is the Saha density of level i. The plasma is considered to be optically thin and all the
elementary processes described above are considered. This example clearly puts forward
that different time scales characterize the global relaxation process. For t < 10−10 s, in
this example, N1 , N2 , N3 and ne vary only slightly in comparison to other excited state
population densities. For 5 × 10−9 s ≤ t ≤ 10−6 s, the latter reach a plateau whereas N1 ,
N2 , N3 and ne vary significantly. Finally, all the system tends toward a stationary state
(reached at t ' 10−3 s in this example). It is interesting to note that the populations of
excited states at the plateau are very different from those at the stationary state. It is
important to mention that the general characteristics of the results are independent of
the initial ionizing or recombining condition.
RTO-EN-AVT-142 9-9
18
22
10
10
ne 4
21 17
10 10
10
20 20
10
Density [m ]
Density [m ]
–3
–3
16
10 30
10
19 Na
18 15
10 10 40
1
17
10 2 14
16
10
10 3 (a) (b)
15 13
10 –13 –11 –9 –7 –5 –3 10 –13 –11 –9 –7 –5 –3
10 10 10 10 10 10 10 10 10 10 10 10
t[s] t[s]
Figure 2: Temporal evolution of densities for an initial recombining situation: at t = 0 s,

Te = 12 000 K, ne = 1020 m−3 , and bi = Ni /Ni∗ = 10−4 , where Ni∗ is the Saha density of level
i. All elementary processes are taken into account, and the plasma is considered to be optically
thin. Numbers denote atomic level numbers. Atomic level population densities are divided by
their statistical weights.
13 2.0
12
Variation rates [10 m s ]
Variation rates [10 m s ]
1.5
6 −1
11
−3 −1
10
−39
26
9 1.0
7 0.5
5 −10 −9 −8 −7 −6 −5
0.0 −10 −9 −8 −7 −6 −5 −4
10 10 10 10 10 10 10 10 10 10 10 10 10
t[s] t[s]
Figure 3: For the same initial recombining Figure 4: For the initial recombining con-
condition as in Figure 2. Solid line: |n˙e |. dition of Fig. 2. Solid line: |n˙e |/n3e . Solid
Solid line with open circles: Ṅ1 . Dashed line with open circles: Ṅ1 /n3e . Dashed line:
line: Ṅ1 + Ṅ2 . Dot-dashed line: Ṅ1 + Ṅ2 + (Ṅ1 + Ṅ2 )/n3e . Dot-dashed line: (Ṅ1 + Ṅ2 +
Ṅ3 . Ṅ3 )/n3e .
9 - 10 RTO-EN-AVT-142
Figure 3 shows the temporal evolutions of |n˙e | and Ṅi for 1≤ i ≤3. For t ≥ τmin (τmin '
5×10−9 s in this example), we have
Ṅi ' 0 for i > 3 (10)
and then a quasi-steady-state condition of the form

3
X
|n˙e | = Ṅi (11)
i=1
can be written. This condition has been already established by Cacciatore and Capitelli
(1976) in oxygen. For hydrogen as shown by Bates et al. (1962), the corresponding quasi-
steady-state condition takes into account only the ground state of the atom. In oxygen
and also in nitrogen as shown by Bourdon and Vervisch (1996), it is necessary to take into
account the contribution of the two first low-lying levels. In oxygen, the fourth level is
metastable too and then one could have expected to have for oxygen a quasi-steady-state
(QSS) condition of the form
4
X
|n˙e | = Ṅi (12)
i=1
In the electron temperature and density range of our study, we have checked that Eq.
(11) is the quasi-steady-state condition for oxygen.
3.4.2 General method to determine effective ionization and recombination

rate coefficients
For the global reaction O+ + 2e− ↔ O + e− , we define kr , the effective three-body
recombination rate coefficient, and ki , the effective ionization rate coefficient, such that
dne
= n˙e = ki ne Na − kr n3e (13)
dt
Both rate coefficients have to be time independent, but may depend on electron temper-
ature and densities. To determine kr , the method is to consider at t =0 s a recombining
situation. In this case, in early times the first term on the right hand side of Eq. (13)
is far much smaller than the second one. Then, if kr exists, there is a plateau in the
temporal evolution of the ratio |n˙e |/n3e and kr is the value of this ratio at the plateau.
Conversely, to determine ki , one has to consider at t =0 s an ionizing condition and to
study the temporal evolution of the ratio n˙e /(ne Na ). ki is the value of this ratio at the
plateau. For the initial recombining situation of the last section, Fig. 4 shows that a value
of kr can be determined for τmin ≤ t ≤ τmax , with τmax ' 10−5 s. The minimal time τmin
corresponds to the necessary time to reach a quasi-steady-state (Figs. 2 and 3) and is
denoted τQSS in the following. For t > τmax , ionization processes become significant and
then the first term on the right hand side of Eq. (13) is no longer negligible in comparison
to the second one. Finally, Fig. 3 clearly shows that a quasi-steady-state condition has
to be reached to determine kr . In the case of an initial ionizing situation, the general
RTO-EN-AVT-142 9 - 11
characteristics of the results are the same.

Finally, to determine kr , respectively ki , we start from an initial recombining, respectively
ionizing situation, and, stop the calculations when |n˙e | and |Ṅ1 + Ṅ2 + Ṅ3 | agree to within
1%. Then we determine τQSS and the value of kr , respectively ki , with the instantaneous
values of n˙e , and n3e , respectively ne Na . The interest of this approach is the possibility to
follow the temporal evolutions of populations and to determine also τQSS .
It is interesting to point out that ionization and recombination rate coefficients are not
independent. Indeed, at the stationary state of the reaction O+ + 2e− ↔ O + e− , the two
terms on the right hand side of Eq. (13) have to balance to give n˙e = 0. Then, ki can be
derived from kr by
[nstat ]2
ki = kr estat = kr K stat (14)
Na
We have checked that both methods gave the same results.
3.4.3 Determination of effective three-body recombination rate coefficients

Optically thick cases
In order to represent an optically thick medium, in this section, all escape factors in
Eqs. (8) and (9) are set equal to zero. Then, as only elementary collisional processes are
considered, the results of this section are denoted with the superscript C. It is interesting to
note that these results may be applied to optically thin media where the electron number
density is high enough to neglect radiative processes in comparison to collisions. After
having considered different initial conditions, we have pointed out that in an optically
thick medium, the three-body recombination rate coefficient for oxygen depends only on
the electron temperature (Fig. 5). We have noted no influence of either densities or initial
distribution on atomic levels. The simplest best-fit curve for our computed result is
−4.66
−39 Te
C
kr = 1.32 × 10 m6 s−1 (15)
104
for 8000 K≤ Te ≤ 20 000 K. Park (1990) assumed that the three-body recombination rate
of oxygen is the same as the one of nitrogen, derived for 4000 K≤ Te ≤ 20 000 K in using
for all levels the elementary rates derived by Gryzinski (1965) for hydrogen
−5.27
−38 Te
krP ark = 1.15 × 10 m6 s−1 (16)
104
Figure 5 shows that, over the whole temperature range of this study, the value obtained
with our model is on the average five times smaller than the one proposed by Park.
With our method, we have also calculated the effective three-body recombination rate
coefficient (denoted krH ) obtained when one uses for all levels the elementary rates derived
by Gryzinski (1965) for hydrogen. In this case, Fig. 5 shows that the value of krH is
close to the one proposed by Park. Cacciatore and Capitelli (1976) proposed to modify
the elementary rate coefficients derived by Gryzinski to take into account that low-lying
levels of atomic oxygen are nonhydrogenic. That is, Gryzinski values are multiplied by
a factor 4 for the ground state. It is interesting to mention that this correction has
no influence on the calculated value of krH for atomic oxygen. Finally, on Fig. 5, the
9 - 12 RTO-EN-AVT-142
fitting law of the three-body recombination rate of nitrogen obtained in a similar work
on a collisional-radiative model for atomic nitrogen is represented. This law is given for
4000K ≤ Te ≤ 12 000K and, therefore should be handle with care for Te > 12 000K.
However, it is interesting to note that three-body recombination rates for oxygen and
nitrogen remain close to each other for 8000K ≤ Te ≤ 20 000K.
−38
10
−38
Recombination rate [m s ]
10
6 −1
6 −1
−39 −39
10 10
−40
10
−40
10
8000 12000 16000 20000
Te [K] 8000 12000 16000 20000
Te [K]
Figure 5: Recombination rate in optically
thick cases as a function of Te . Solid Figure 6: Recombination rate in optically
line: our model. Dot-dashed line: Park thick cases as a function of Te . Thick solid
(1990). Dashed line: our atomic model line: our 40 level model. Thin solid line:
with Gryzinski (1965) cross sections. Solid 9 levels. Solid line with open circles: 15
line with open circles: Recombination rate levels. Dashed line: 20 levels. Dot-dashed
of nitrogen (Bourdon and Vervisch (1996)). line: 30 levels.
Influence of the number of levels

Figure 6 shows the influence of the number of atomic levels on the calculated effective
three-body recombination rate coefficient value in optically thick cases. As the number
of atomic levels increases, the calculated recombination rate converges towards an upper
limit and the lower the electron temperature is, the more numerous atomic levels are re-
quired to determine krC . This figure puts forward clearly that the 9-level model proposed
by Soon and Kunc (1990) is unadapted to our temperature range and that it was nec-
essary to consider a more sophisticated atomic model. Our forty level model appears to
be a good compromise between the inaccuracy of atomic models for highly excited levels
and the reliability of the effective rate coefficients derived in this study. It is important
to mention that the determination of krC at low electron temperatures Te ≤ 2 000 K is
generally difficult since it depends strongly on highly excited levels. We have noted in
this case, that the more numerous highly excited atomic levels are taken into account, the
higher the derived recombination rate is.
Optically thin cases

In optically thin cases, the results obtained for different initial recombining conditions in-
dicate that the three-body recombination rate depends only on the electron temperature
RTO-EN-AVT-142 9 - 13
and the electron number density. First, neglecting radiative and dielectronic recombina-
tions, Fig. 7 shows that spontaneous emission increases the value of the recombination
rate (denoted, in this case, krCE ). As expected, at high electron number densities, the
recombination rate converges to the value obtained in optically thick cases.
−36
10
−37 18
10 −37
10
6 −1
18
6 −1
19
−38
19
−38
10 10
20 20
−39 −39
10 21 10 21
−40
22
−40 22 10
10
8000 12000 16000 20000

8000 12000 16000 20000 Te [K]
Te [K]
Figure 8: Recombination rate derived in
Figure 7: Influence of spontaneous emis-
taking into account all elementary pro-
sion on the recombination rate in an opti-
cesses, in an optically thin plasma. Same
cally thin plasma. Numbers at right denote
notations as Fig. 7. The thin solid line
x with ne = 10x m−3 . The thick solid line
corresponds to the expression proposed by
corresponds to optically thick cases.
Park (1990).
Conversely, for low electron number densities, krCE tends towards an upper limit. If all
the elementary processes of section 3.2 are considered, Fig. 8 shows that the influence
of radiative and dielectronic recombinations increases as ne decreases. In this case, the
recombination rate has no upper limit at low electron number densities. These results are
in qualitative agreement with those obtained earlier by Taylor and Ali (1986). Figure 8
indicates that Park’s expression [Eq. (16)], initially proposed for an optically thick case,
corresponds to an optically thin plasma where 1020 m−3 ≤ ne ≤ 1021 m−3 and 8000K ≤
Te ≤ 20 000K.
In order to implement the results of this work in numerical studies on air plasma flows,
simple analytical laws are required. In the general Eq. (9), radiative and dielectronic
terms can be isolated, then Eq. (13) can be written
dne
= n˙e = ki ne Na − krCE n3e − αR n2e (17)
dt
where krCE takes into account all the elementary collisional processes and spontaneous
emission (Figure 7 ), and
αR = αRR + αRD (18)
The total effective recombination rate kr is therefore krCE + αR /ne . This formulation is
of great interest since αRR [Eq. (7)] and αRD [Nussbaumer and Storey (1983)] can be
9 - 14 RTO-EN-AVT-142
expressed analytically. Then, we fitted krCE for different electron number densities to the
following expression
krCE = ATeb , m6 s−1 (19)
for 8000K ≤ Te ≤ 20 000K. The coefficients A and b are given in Table 2. The discrepancy
between the total recombination rate value calculated with the analytical expressions [Eqs.
(19) and (18)] and the one determined directly and represented on Fig. 8 remains within
a factor of 2.
ne (m−3 ) A (m6 s−1 K−b ) b

1018 7.27 × 10−31 -1.73
1019
4.29 × 10−30 -1.95
1020 2.58 × 10−28 -2.51
1021 3.19 × 10−25 -3.43
Table 2: A and b parameters for the effective three-body recombination rate coefficient
krCE = ATeb , m6 s−1 for different electron number densities
Comments on the definition of kr

Following the method described above to determine αR , one could propose to derive the
effective three-body recombination rate coefficient directly from Eq. 9
X
kr∗ = βci (Te ) = f (Te ) (20)
i
By definition, this coefficient depends only on Te and takes into account only direct
recombination processes on the atomic levels, and neglects the other processes which link
the atomic level populations together. Figure 9 shows that this coefficient kr∗ is slightly
CE
higher than kr,max which is the maximal value of krCE . This coefficient has no physical
interest to describe the O+ + 2e− → O + e− global recombination process. This result
already mentioned by Park (1969) is clearly illustrated here.
3.4.4 Determination of the effective ionization rate coefficient

In optically thick cases, the results obtained for different initial ionizing conditions indicate
that the effective ionization rate coefficient (denoted kiC ) depends only on the electron
temperature. This can be simply derived from Eq. (14), since in this case, krC depends
only on Te and, the stationary state of the reaction O+ + e− ↔ O + 2e− corresponds to a
Saha equilibrium, therefore K stat = K Saha (Te ). We have fitted the equilibrium constant
to the following expression
X5
stat
ln(K ) = Ai Z i , (21)
i=0
4
where Z = ln(10 /Te ), for 8000K ≤ Te ≤ 20 000K. For a Saha equilibrium, the coeffi-
cients Ai are given in Table 3. Figure 10 compares the value of kiC obtained in our study
with three other values often used in high temperature air kinetic schemes, but which
accuracy is unknown:
RTO-EN-AVT-142 9 - 15
• The effective ionization rate coefficient proposed by Losev et al. (1994) which is
derived from the ionization cross section from the ground state measured by Smith
et al. (1962).

−18 0.68 −157980
kiLosev = 8.64 × 10 Te exp m3 s−1 (22)
Te
• The value proposed by Gupta et al. (1990)

−158120
kiGupta = 59.8 Te−2.91 exp m3 s−1 (23)
Te
• The value proposed by Park (1990)

−158620
kiP ark = 6478.4 Te−3.78 exp m3 s−1 (24)
Te
−14
10
−37
10 10
−15
6 −1
−16
Ionization rate [m s ]
10
3 −1
−38
10 kr* −17
krCEmax
10
krCEmin
−18
10
−39
10
−19
10
−20
−40 10
10
−21
10
8000 12000 16000 20000 8000 12000 16000 20000
Te [K] Te [K]
Figure 9: Different definitions of the three- Figure 10: Ionization rate coefficient in op-
body recombination rate coefficient. kr∗ tically thick cases. Solid line: our model.
CE
is the one-way coefficient and kr,min and Dotted line: Park (1990) Dot-dashed line:
CE
kr,max are respectively the minimal and Gupta et al. (1990) Dashed line: Losev
maximal values of krCE . et al. (1994).
First of all, we note that there are strong discrepancies between the different models. As
expected, the value obtained with our model is greater than the one proposed by Losev et
al. since the latter takes into account only the ionization process from the ground state.
Figure 10 shows that the value proposed by Gupta et al. is overestimated by more than
one order of magnitude. In fact, the value of kiC obtained with our model turns out to
be in close agreement with the one proposed by Park. This result seems to be in dis-
agreement with the discrepancy observed on Fig. 5 between our value of the three-body
recombination rate in optically thick cases (krC ) and the one proposed by Park (krP ark ).
In fact, we have noted that the ionization and recombination rate coefficients proposed
9 - 16 RTO-EN-AVT-142
Nt (m−3 ) A0 A1 A2 A3 A4 A5
1017 30.5304 -34.0190 -23.6944 6.18581 -0.916160 -17.5735
1018 32.8715 -33.1659 -22.8399 7.45801 5.02374 -11.8446
1019 35.0493 -31.4172 -24.8850 -0.795804 17.3478 11.0059
1020 36.6971 -27.7121 -6.36738 -49.7549 -187.362 -149.453
1021 38.6205 -35.9083 6.47223 62.7877 -35.6484 -97.7333
1022 41.9019 -45.7554 -37.4442 166.309 443.193 298.333
1023 45.2287 -31.9500 -57.7671 -69.9729 -17.6998 19.8520
K Saha 47.0661 -17.1846 -7.99356 -2.64241 -0.579718 0.182226
Table 3: Parameters in K Saha and in the equilibrium constant K stat for different total
particle number densities.
by Park in optically thick cases are not related by the Saha equilibrium constant. That
is, krP ark K Saha is on the average 4.7 times greater than kiP ark ; which is also the average
discrepancy observed on Fig. 5 between krC and krP ark .

In optically thin cases, radiative processes enhance recombination and then the ionization
degree at the stationary state is lower than the one corresponding to the Saha equilibrium.
Consequently, K stat < K Saha (Te ), where K stat depends on Te but also on densities. As
explained by Kunc and Soon (1989) the most significative density of the nonequilibrium
properties of the stationary state is the total particle density Nt (note that the definition of
Nt proposed by these authors is the double of ours) and not the electron number density.
As expected, as the density increases, collisional processes become more efficient than
radiative ones and, then the equilibrium constant converges toward the Saha equilibrium
constant, all the faster since the electron temperature is high. First, neglecting radiative
and dielectronic recombination processes, the equilibrium constant tends toward a lower
limit for low values of Nt . When radiative and dielectronic recombination processes are
taken into account, this lower limit disappears. As Soon and Kunc (1990) considered
more accurately the reabsorption of the emitted radiation, a direct comparison with their
results at the stationary state is impossible. However, a qualitative agreement is observed.
Finally, to implement our results in flow codes, we have fitted K stat for different values
of Nt to the expression of Eq. (21) for 8000K ≤ Te ≤ 20 000K. The coefficients Ai are
given in Table 3. The discrepancy between the ionization rate coefficient value calculated
with the analytical expression [Eq. (21)] and the one determined directly remains within
a factor of 2 (Bourdon et al. (1998)).
3.4.5 Study of the quasi-steady-state relaxation time

In the previous sections, we have put forward that effective three-body recombination and
ionization rate coefficients exist only when the system is in a quasi-steady-state. There-
fore, before implementing these rate coefficients in flow codes, one has to check the validity
of the QSS hypothesis [Eq. (11)]. The time-dependent approach used in this work allows
us to determine the relaxation time necessary to reach a quasi-steady-state, starting at
RTO-EN-AVT-142 9 - 17
t = 0s from a given initial condition.

For optically thick cases, the results obtained for different initial conditions (either ioniz-
C
ing or recombining) indicate that the quasi-steady-state relaxation time (denoted τQSS )
depends only on the electron temperature and electron number density but not on the
C
initial distribution on the atomic levels. Figure 11 shows that τQSS depends only weakly
19 −3
on Te but increases as ne decreases. For ne = 10 m , the relaxation time derived in
using Gryzinski’s cross sections for all levels is also represented. The latter is one order
magnitude shorter than the one obtained with our model. It is interesting to note that re-
laxation times obtained in oxygen are very close to those calculated in nitrogen (Bourdon
and Vervisch (1996)). Finally, Fig. 11 shows that the effective ionization and recom-
bination rate coefficients determined in this work are adapted to study optically thick
plasma flows, where the characteristic flow time is about 10−6 s, only if ne ≥ 1020 m−3 .
For lower electron number densities or shorter flow times, it is necessary to consider one
conservation equation for each atomic level i of the oxygen atom.
−4
10
Relaxation time τQSS[s]
−6
Relaxation time τQSS[s]
−5 10
10
−6
10
−7
10
−7
10
8000 12000 16000 20000 −8

Te[K] 10
8000 12000 16000 20000
Te[K]
Figure 11: Relaxation time in optically
thick cases. Thin solid line with open cir- Figure 12: Relaxation time in optically
cles: ne = 1018 m−3 . Thin dashed line: thin cases with at t =0s, bi = Ni /Ni∗ =1,
ne = 1019 m−3 . Thin dot-dashed line: ne = for a given electron number density. Solid
1020 m−3 . Thin solid line: ne = 1021 m−3 . line with open circles: ne = 1018 m−3 .
The thick dashed line represents the results Dashed line: ne = 1019 m−3 . Dot-dashed
obtained for ne = 1019 m−3 in using the line: ne = 1020 m−3 . Solid line: ne =
cross sections derived by Gryzinski (1965). 1021 m−3 .

In an optically thin plasma, the relaxation time depends on densities and Te but also on
the initial distribution on atomic levels. Therefore, it is more difficult to derive general
results. However, we have noted that for initial recombining conditions with at t =0 s
bi = Ni /Ni∗ ≤ 1 (where Ni∗ is the Saha density of level i) for a given electron number
9 - 18 RTO-EN-AVT-142
density, the QSS relaxation time depends only on ne and Te . In these cases, Fig. 12
shows that the QSS relaxation time depends weakly on Te and increases as ne decreases.
In comparison to an optically thick case, we note that relaxation times are on the average
one order of magnitude shorter.
3.5 Conclusions
In this work, we have shown how to use a detailed collisional radiative model to determine
the effective ionization and three-body recombination rate coefficients of oxygen in a
plasma where 1018 m−3 ≤ ne ≤ 1021 m−3 and 8000K ≤ Te ≤ 20 000K. The time-dependent
approach used in this work has clearly put forward the necessity that the system be in a
quasi-steady state to derive meaningful rate coefficients. This has led us to study more
precisely the relaxation time τQSS necessary to reach a quasi-steady-state.
In this work, the treatment of radiation has been reduced to optically thin or thick cases.
It would be interesting to take into account a more accurate modeling of radiation escape
factors in order to study more precisely the influence of the reabsorption of radiation.
However, this effect will not change the general characteristics of the results obtained in
this paper.
4 Collisional-radiative model for air

The previous section has shown that a temporal description of the chemistry of a plasma
is the most relevant approach to understand its behavior by a fundamental point of view:
this approach is the only one allowing to check the conditions for using a ionization or
recombination rate coefficient. As a result, we have chosen to reproduce such a study
but for a more complicated medium where molecules play a role in order to see if a quasi
steady-state still exists. Moreover, this study is also motivated by our main objective:
the more realistic calculation of the interaction between a plasma formed during an at-
mospheric re-entry of a spacecraft and its thermal shield. In that case, the flow is indeed
very rapid: as illustrated in the following, the chemistry is therefore time-dependent in
Lagrangian description.
4.1 Atomic and molecular model

In this work, air considered is a mixture of heated nitrogen and oxygen and their products:
argon as other gas traces are neglected. Due to the wide range of temperatures and pres-
sure studied, we take into account N2 , O2 , N , O, N O, N2+ , O2+ , N + , O+ , N O+ , O2− , O−
and electrons. For each species, we consider electronic excited states except for N + , O+ ,
O− and O2− as shown in Table 4. The energetic data are due to Herzberg (1950), Capitelli
et al. (2000a) and Park (1990). 46 levels for N and the previous 40 levels for O are taken
into account in order to calculate correctly the recombination of N + and O+ respectively
by three-body processes and in order to estimate thoroughly the net population density
of the metastable states 2 D0 and 2 P 0 for nitrogen and 1 D and 1 S for oxygen resulting
from the electron and heavy particles induced processes. In addition, that provides the
calculation of the excited population densities which are the source of the radiative signa-
ture of the medium. The levels considered for N are those already mentioned by Bourdon
RTO-EN-AVT-142 9 - 19
Type Species State
N2 X 1 Σ+ 3 + 3 3
g , A Σ u , B Πg , C Πu
Molecules O2 −
X Σg , a ∆g , b Σg , A Σu , B 3 Σ−
3 1 1 + 3 +
u
NO X 2 Π, A2 Σ+ , B 2 Π, C 2 Π, B 02 ∆
N2+ X 2 Σ+ 2 2 + 2 +
g , A Πu , B Σ u , C Σ u
Molecular O2+ 2 4 2 4 −
X Πg , a Π u , A Πu , b Σ g
ions N O+ X 1 Σ+ 3 + 3 03 −
(v=0→14) , a Σ , b Π, b Σ , A Π
1
O2− X 2 Πg
4
Atoms N S 0 ,... (46 levels)
3
O P ,... (40 levels)
Atomic N+ 3
P
ions O+ 4
S0
O− 2
P
Table 4: Species taken into account in the present CR model.
and Vervisch (1996).

Moreover, since the rate coefficient for dissociative recombination of N O+ has been calcu-
lated for each excited vibrational level between v = 0 and v = 14 (Motapon et al. (2006)),
the vibrational excitation of this ion is explicitly considered contrary to the case of the
other molecular species for which the assumption of equilibrium at Te is done.
All in all, this atomic and molecular model represents 131 different levels and their cou-
pling in the frame of the present physical situation is ensured by the following elementary
processes.
4.2 Elementary processes

4.2.1 Radiative processes
We take into account 45 spontaneous emission lines for N and 24 for O for which Einstein’s
probability is higher than 5×103 s−1 for N and 3×105 s−1 pour O. In addition, we account
for the radiative and dielectronic recombinations. The calculations and data for all these
radiative processes are similar to those of the previous section.
The possible re-absorption of the radiations is estimated by the following escape factor:
1
e−k0 l x dx
Z
T (l) = √ (25)
0 −π ln x
derived from the work of Holstein (1947) with a typical plasma thickness equal to that of
the boundary layer (typically of the order of l = 1 cm) and assuming a Doppler broadening
for which k0 is the absorption coefficient by unit length at the centerline. Any pressure
broadening is assumed negligible, the maximum pressure level being atmospheric. In the
9 - 20 RTO-EN-AVT-142
same way, the role of charged particles is not considered since the electron density ne is
sufficiently weak.
Some of the states of the species mentioned in Table 4 radiate importantly. The β (B 2 Π →
X 2 Π) and γ (A2 Σ+ → X 2 Π) systems of N O as well as the first positive (B 3 Πg → A3 Σ+ u)
and the second positive (C 3 Πu → B 3 Πg ) systems of N2 and the first negative system
+
(B 2 Σ+ 2 +
u → X Σg ) of N2 have been considered. Since a vibrational equilibrium is assumed
for these species at the electron temperature Te , the equivalent transition probability for
the latter systems is calculated by:
G(v 0 )
P P −
v0 ( v” Av0 v” )e kB Te
A= P − kG(vT0 ) (26)
v0 e
B e
where G(v 0 ) is the vibrational energy, Av0 v” the vibrational transition probability and kB
the Boltzmann constant. Using the data calculated by Laux and Kruger (1992), we have
fitted each value of A by:
X 3
A= ai Tei (27)
i=0
whose coefficients ai are calculated for 2000 K ≤ Te ≤ 12000 K and may be obtained on
demand.
4.2.2 Collisional processes

Electron induced processes
When available, the cross sections are chosen in accordance with the more recent experi-
mental data. Since the Maxwell equilibrium assumption is done for most of the species,
the cross sections have been integrated and the rate coefficient obtained finally fitted
under the form:
ki→j (Te ) = A Te−α e−θ/Te (28)
In Table 5 are summed up the parameters A, α and θ hence calculated for 2000 K ≤
Te ≤ 10000 K. The rate coefficient for the excitation of molecular species under electron
impact has been calculated recently by Teulet et al. (1999) under the form of Eq.(28) and
are used in the present work.
There are few data concerning the ionization of N and O, especially for the highly excited
states. For the ground state, we have used for both atoms the cross sections reviewed
by Tawara and Kato (1999). Conversely, we have preferred the cross sections obtained
by Kim and Desclaux (2002) for metastable nitrogen and assumed that the cross section
for metastable oxygen has the same form when the shift due to the different threshold is
taken into account. The derived rate coefficients expressed by Eq.(28) use the parameters
mentioned in Table 5. For higher excited levels, the classical cross sections of Drawin are
used considering that each atom has only one optical electron.
The data related to the ionization of molecules are taken from the work of Teulet et al.
(1999) as those concerning the dissociation.
Heavy particles induced processes

The excitation of atoms due to atom collisions has been discussed by Capitelli et al.
RTO-EN-AVT-142 9 - 21
i j A α θ Ref.
N (4 S 0 ) N (2 D) 1.720 × 10−15 −0.280 28250 Berrington et al. (1975)

N (4 S 0 ) N (2 P ) 4.004 × 10−16 −0.324 42860 Berrington et al. (1975)
O(3 P ) O(1 D) 1.374 × 10−17 −0.566 24600 Itikawa et al. (1990)
O(3 P ) O(1 S) 9.352 × 10−19 −0.620 50110 Itikawa et al. (1990)
N (4 S 0 ) N + (3 P ) 8.583 × 10−16 −0.276 166870 Tawara and Kato (1999)
N (2 D) N + (3 P ) 7.732 × 10−18 −0.711 139900 Kim and Desclaux (2002)
N (2 P ) N + (3 P ) 1.026 × 10−17 −0.677 126310 Kim and Desclaux (2002)
O(3 P ) O+ (4 S 0 ) 5.393 × 10−18 −0.750 158330 Tawara and Kato (1999)
O(1 D) O+ (4 S 0 ) 8.571 × 10−18 −0.695 134190 Kim and Desclaux (2002)
O(1 S) O+ (4 S 0 ) 1.397 × 10−17 −0.647 108800 Kim and Desclaux (2002)
Table 5: Parameters A, α and θ for the calculation of the rate coefficient (in m3 s−1 ) given
by Eq.(28) for the transition i → j induced by electron collision. θ is expressed in K.
(2000a) between the ground and metastable levels. We have consequently considered all
these data in our CR model. On the other hand, we have adopted a linear form for the
cross section σijA with the relative kinetic energy of the colliding particles for excitation
from the ground or the metastable states towards highly excited levels:
σijA−A () = βijA−A ( − Eji ) (29)
where Eji is the threshold. Since the parameter βijA−A weakly depends on the mass
of the colliding particles (Drawin and Emard (1973)), we have assumed that the cross
section can be derived from the case of argon (Bultel et al. (2002)) taking the mean
value βijA−A = 3.4 × 10−25 m2 eV −1 . For excitation between higher excited levels, we have
considered, as in the case of argon, that Eq.(29) is applicable with:
−2.26
βijA−A = 8.69 × 10−22 Eji m2 eV −1 (30)
The cross section for excitation of atoms under molecule impact behaves approximately
as (Lotz (1968)):
ln(/Eji )
σijA−M () = σ0 (31)
/Eji
where σ0 ' 10−20 m2 . This form has been adopted.
Equation (31) is also assumed to condition the excitation of any molecular species under
atom or molecule impact except in the case when experimental data exist. We have then
preferred the rate coefficients compiled by Teulet et al. (2001), Capitelli et al. (2000a)
and Kossyi et al. (1992).
From these sets of rate coefficients, we have also extracted and used those related to
charge transfer (with possible reassociation of atoms to form a molecule), reassociation,
excitation transfer, ionization, dissociation, neutralization of O− and O2− and attachment
and detachment. Due to the pressure levels involved in our CR model, we have also taken
into account the previous processes occurring when a third particle interacts.
9 - 22 RTO-EN-AVT-142
Two reactions are known to highly influence the distribution of nitrogen and oxygen
between atomic and molecular systems. These reactions called Zeldovich’s reactions are:
(1)
O2 + N → N O + O (32a)
(2)
N2 + O → N O + N (32b)
Recently, Bose and Candler have up-dated their rate coefficient using a quasiclassical
trajectory method performed starting from an ab initio potential surfaces calculation.
They have obtained:
k1 (TA ) = 4.14 × 10−21 TA1.179 e−4000.5/TA m3 s−1 (33)
for the process (32a) (Bose and Candler (1997)) and:
k2 (TA ) = 9.45 × 10−18 TA0.420 e−42940/TA m3 s−1 (34)
for the process (32b) (Bose and Candler (1996)). The rate coefficients (33) and (34) are
used in our work.
The processes (32a) and (32b) can contribute to the destruction of O2 and N2 and forma-
tion of N O. Of course, the direct dissociation and the reverse process by heavy particles
impact have to be taken into account. The rate coefficients used are due to Losev and
Shatalov (1990) who have assumed the vibrational mode in equilibrium with the electrons
at Te . We refer to their paper for more details.
4.2.3 Dissociative recombination of N2+ , O2+ and N O+

The dissociative recombination (DR) of the molecular ions, which is the subject of a very
intensive work (Guberman (2003)), is well-known to play a very important role in the
case of recombining plasmas even at low pressure (Bultel et al. (2002)). In addition, the
inverse process, the associative ionization, allows in much cases the formation of the first
electrons in shock tube (Drawin and Emard (1973)) as well as for the re-entries problems
and consequently explain the ionizing situations. In our case, since N2+ , O2+ and N O+
are present in the plasmas described here, dissociative recombination has to be considered.
Dissociative recombination of N2+

The dissociative recombination of N2+ has been investigated in the CRYRING storage ring
by Peterson et al. (1998) over a wide range for electron energy (10 meV < < 30 eV ).
They have shown that its rate coefficient is weakly dependent on the vibrational state v
of the molecular ion and can be set equal to:
−0.30
−13 Te
αN2+ = 1.75 × 10 m3 s−1 (35)
300
In addition, they have obtained the branching fractions γi (N2+ ) for v = 0 at 0 eV collision
energy for the following products formation:
γ1 (N + )
N2+ + e− −→2 N (4 S 0 ) + N (2 D) (36a)
RTO-EN-AVT-142 9 - 23
γ2 (N + )
N2+ + e− −→2 N (4 S 0 ) + N (2 P ) (36b)
γ3 (N + )
N2+ + e− −→2 N (2 D) + N (2 D) (36c)
with γ1 (N2+ ) = 0.37, γ2 (N2+ ) = 0.11 and γ3 (N2+ ) = 0.52. The collision energy range of
their experiment includes our own (derived from the electron temperature Te levels con-
sidered in this work). Moreover, although the rate coefficient found weakly depends on
the vibrational excitation of the molecular ion, the vibrational temperature of N2+ in the
experiment of Peterson et al. corresponds well to the case of the plasmas considered in
this paper (Tv = Te ' 6000 K on average). For all these reasons, their results have been
integrated in our calculations.
Dissociative recombination of O2+

The dissociative recombination of O2+ has been investigated in the same heavy ion storage
ring by Peverall et al. (2001) but only on the collision energy range 1 meV < < 3 eV
due to the collapse of the cross section for higher energies. The rate coefficient derived is:
−0.70
−13 Te
αO2+ = 2.4 × 10 m3 s−1 (37)
300
where the molecular ion is in the vibrational fundamental state. The branching fractions
ascribed to the processes:
γ1 (O+ )
O2+ + e− −→2 O(3 P ) + O(3 P ) (38a)
γ2 (O+ )
O2+ + e− −→2 O(3 P ) + O(1 D) (38b)
γ3 (O+ )
O2+ + e− −→2 O(1 D) + O(1 D) (38c)
are respectively equal to 0.20, 0.40 and 0.40. Although the branching fraction for the
process:
O2+ + e− → O(1 D) + O(1 S) (39)
is not equal to 0, its order of magnitude and the fact that it becomes almost 0 for colli-
sion energies between 10 and 40 meV by increasing up to 0.06 for = 300 meV led us to
consider finally that this channel is closed and that the dissociative recombination for O2+
occurs by the processes (38a), (38b) and (38c) only. Founding no more information in the
literature about a possible influence of the vibrational excitation of O2+ on the dissociative
recombination rate coefficient, we have assumed no influence and consequently adopted
Eq.(37) and the previous branching fractions.
Dissociative recombination of N O+
The case of N O+ has to be particularly discussed: as we will see in the subsequent part
of this paper, this molecular ion plays a very important role. As in the previous case
of N2+ and O2+ , three factors condition the discussion: the vibrational excitation of the
molecular ion, the collision energy range and the branching fractions. Early works were
devoted to the experimental determination of the rate coefficient in room temperature
conditions for kinetics in post-discharge purposes (Gunton and Shaw (1965)) (Weller and
9 - 24 RTO-EN-AVT-142
Biondi (1968)): N O+ is then on its vibrational ground state and the collision energy
range is very weak. Other determinations of the rate coefficient were also performed for
larger collision energy ranges (Huang et al. (1975)). With the development of the hy-
personic flights, experiments were done also in shock-tubes where the rate coefficient was
determined (Dunn and Lordi (1969)) or calculated (Hansen (1968)): the vibrational exci-
tation of the molecular ion was higher consequently as well as the collision energy range
(Te < 6000 K in order of magnitude). Several theoretical treatments have been applied
to explain the experimental values. We mention the work of Bardsley (1983) and those
of Sun and Nakamura (1990) whose results, obtained with different methods, reveal that
the cross section for v > 0 can be higher by one order of magnitude than for v = 0 when
> 1 eV . Since a calculation of the Maxwell rate coefficient needs the value of the cross
section up to = 10 eV for Te = 1 eV , the vibrational excitation of N O+ may have an
influence on the global dissociative recombination. In order to use an accurate value for
the cross section on the previous collision energy range and in particular in order to access
to the role of the vibrational excitation, we have used the results of the extensive calcu-
lations using the Multichannel Quantum Defect Theory (MQDT) developed by Motapon
et al. (2006) resulting, among other data, in the cross section σDR (, v) of the dissociative
recombination:
σDR (,v)
N O+ (X 1 Σ+ , v) + e− () −→ N ∗ + O∗ (40)
This kind of reaction can occur in fact according to two channels.
−13
10
−14
10
−15
10
σDR,v(cm )
2
−16 v=0
10 v=1
v=2
v=3
10
−17 v=4
v=5
Figure 13: Schematic view of the two chan- 10

−18
−2 −1 0 1
10 10 10 10
nels allowing the dissociative recombina- ε(eV)
tion of a molecular ion AB with electron.

Left: the direct process, right: the indirect
process. The electron energy is . AB ∗∗ is a Figure 14: Cross section σDR (, v) for dis-
dissociative potential curve of the molecule sociative recombination of N O+ with elec-
AB. tron for different vibrational excitation of
the ion (0 ≤ v ≤ 5).
First, the DR can be due to a direct process: it leads to the formation of a doubly excited
state of N O, denoted N O∗∗ , which dissociate spontaneously afterwards by producing
separated N and O atoms possibly electronically excited (see Fig.13). The estimation of
the cross section needs consequently an accurate knowledge of the potential curves for
these states. The higher the energy for the electron, the more numerous the potential
RTO-EN-AVT-142 9 - 25
curves. Second, the DR can also occur through an indirect process: this process leads to
the formation of a Rydberg level of N O associated with a vibrationnally excited state of
N O+ which is predissociated by the N O∗∗ state. In this case, the Rydberg states being
numerous by definition, the calculation of the contribution of the indirect channel to the
cross section needs a great amount of molecular data.
Both channels need finally numerous data and one of the main issues to be addressed is the
thorough determination of the potential curves related to the states involved. After such
a determination, several methods can be applied in order to calculate the cross section.
One of the most efficient one at the present time is based on the Multi-Channel Quantum
Defect Theory. Its approach is based on a description of molecular states in which only
part of the electronic hamiltonian is diagonalized, within subspaces of electronic states
with similar nature. We refer to the paper of Motapon et al. (2006) for more details.
We can observe on Fig.14 the behavior of the cross section resulting from the energy of
the colliding electron and the vibrational excitation of the molecular ion calculated by
the MQDT approach. The influence of the great number of states is obvious: the cross
section presents much oscillations which correspond to resonances when the energy of
the electron is accorded with the difference between two particular states involved. In
addition, the global behavior does not follow a −1 law contrary to that usually assumed.
In particular, when the energy of electron exceeds 7 eV , the cross section collapse.
The Maxwellian rate coefficient of the DR is derived from the calculation of the cross
section. Each rate coefficient is finally fitted under the general modified Arrhenius law
(28). We can find in Table 6 the parameters A, α and θ for each value of the vibrational
quantum number v between 0 and 5. The usual form adopted for the cross section leads
v A α θ
0 8.486 × 10−11 0.793 859.99

1 3.958 × 10−14 0.076 −661.83
2 1.664 × 10−11 0.704 −39.13
3 6.335 × 10−13 0.383 −282.98
4 2.970 × 10−13 0.311 −632.21
5 2.666 × 10−13 0.297 −472.45
Table 6: Parameters A, α and θ for the calculation of the rate coefficient (in m3 s−1 ) given
by Eq.(28) for the dissociative recombination of N O+ (X 1 Σ, v) with electrons used in this
work Motapon et al. (2006). θ is expressed in K.
to a rate coefficient following a T −0.5 law. The characteristic temperature θ being rather
small, the temperature dependence is mainly due to α. Table 6 shows that the rate
coefficient is relatively different from a value of the vibrational quantum number v to
another with a form far from a T −0.5 law: as a result, the usual temperature dependence
is consequently not realistic and too simplistic. Assuming an equilibrium between the
vibration of N O+ and electrons, these values lead to that of the global rate coefficient for
the dissociative recombination:
kDR (T ) = 1.658 × 10−10 T −0.933 e−377.68/T m3 s−1 (41)
9 - 26 RTO-EN-AVT-142
where the main temperature dependence is due to the T −0.933 term. This form reveals
that the common law adopted for the DR is erroneous. This conclusion can be made
also in the case of N2+ and O2+ whose rate coefficient presents a temperature dependence
according to a T −0.3 and T −0.7 law respectively as experimentally determined.
The more realistic view of the process (40) needs to know the excitation of nitrogen and
oxygen atoms after the recombination. As far as we know, only three works were devoted
to this difficult issue. The earliest one, due to Kley et al. (1977) and performed by
photoionization of buffered nitric oxide by a flash lamp, reported that only two channels
are open:
γ1
N O+ (X 1 Σ+ ) + e− −→ N (4 S 0 ) + O(3 P ) (42)
and
γ2
N O+ (X 1 Σ+ ) + e− −→ N (2 D) + O(3 P ) (43)
with γ1 = 0.24 ± 0.06 and γ2 = 0.76 ± 0.06. It is important to note that these results
have been obtained with a Maxwellian distribution for the kinetic energy of colliding
particles. Controlling , Vejby-Christensen et al. (1998) have more recently determined
the branching fractions in the ASTRID ion storage ring for a collision energy = 0, 0.75
and 1.35 eV . They confirm in part the results of Kley et al. but with γ1 rather equal
to 0.15 ± 0.06 for = 0 eV and not well identified for = 0.75 and 1.35 eV . For these
two collision energies, they concluded only that the process (43) represents at least 65%
of the dissociative recombination events. Finally, Hellberg et al. (2003) have investigated
the dissociative recombination of the X 1 Σ+ and a3 Σ+ states in the CRYRING storage
ring and succeeded for the ground electronic state to measure the branching fractions
for = 0, 1.25 and 5.6 eV . Increasing the collision energy leads to a multiplication of
the branching. Table 7 illustrates their results compared to those of Kley et al. and
Vejby-Christensen et al. Since our CR model is elaborated to work over a wide range of
temperature, each one corresponding to a Maxwell distribution of the collision energy, the
temperature dependent branching fraction γi (Te ) for each i channel has to be determined.
The values finally retained in our work result from the following considerations and are
based on the data of Hellberg et al. due to their best precision. When the collision energy
is insufficient, the branching fraction is put equal 0. This is the case at different energies
of all the channels involved except those producing N (4 S 0 ) + O(3 P ) and N (2 D) + O(3 P )
for which the processes are exothermic. Since the branching fraction for the production
of N (4 S 0 ) + O(1 D) is always 0 considering the uncertainty range, we have assumed that
this channel is systematically closed.
Moreover, the branching fraction is assumed independent of the collision energy when
the collision energy is greater than the threshold (which is equal to 0.81, 1.59, 2.78, 3.81
and 5.0 eV for the processes 3 up to 7 respectively). For the channels (42) and (43), the
values adopted are chosen over the uncertainty range in order that the sum of branching
fractions be equal to unity. As a result, the branching fractions obtained according to the
energy ranges are summed up in Table 8. The temperature dependent branching fraction
γi (Te ) for the i-process is finally obtained by:
Z +∞
γi () x e−x σDR (x) dx
γi (Te ) = 0 Z +∞ (44)
−x
x e σDR (x) dx
0
RTO-EN-AVT-142 9 - 27
Ref. Process Process Branching

i fraction γi
K. N O+ (v̄ = 1.5) + e− (¯ = 0.34 eV ) −→ N (4 S 0 ) + O(3 P ) 1 0.24 ± 0.06

K. N O+ (v̄ = 1.5) + e− (¯ = 0.34 eV ) −→ N (2 D) + O(3 P ) 2 0.76 ± 0.06
V.-C. N O+ (0) + e− (0 eV ) −→ N (4 S 0 ) + O(3 P ) 1 0.15 ± 0.06
V.-C. N O+ (0) + e− (0 eV ) −→ N (2 D) + O(3 P ) 2 0.85 ± 0.06
V.-C. N O+ (0) + e− (0.75, 1.35 eV ) −→ N (2 D) + O(3 P ) 2 > 0.65
H. N O+ (0) + e− (0 eV ) −→ N (4 S 0 ) + O(3 P ) 1 0.05 ± 0.02
H. N O+ (0) + e− (0 eV ) −→ N (2 D) + O(3 P ) 2 0.95 ± 0.02
H. N O+ (0) + e− (1.25 eV ) −→ N (4 S 0 ) + O(3 P ) 1 0.10 ± 0.06
H. N O+ (0) + e− (1.25 eV ) −→ N (2 D) + O(3 P ) 2 0.70 ± 0.20
H. N O+ (0) + e− (1.25 eV ) −→ N (4 S 0 ) + O(1 D) - 0.10 ± 0.10
H. N O+ (0) + e− (1.25 eV ) −→ N (2 P ) + O(3 P ) 3 0.10 ± 0.07
H. N O+ (0) + e− (5.6 eV ) −→ N (4 S 0 ) + O(3 P ) 1 0.03 ± 0.02
H. N O+ (0) + e− (5.6 eV ) −→ N (2 D) + O(3 P ) 2 0.15 ± 0.05
H. N O+ (0) + e− (5.6 eV ) −→ N (2 P ) + O(3 P ) 3 0.11 ± 0.05
H. N O+ (0) + e− (5.6 eV ) −→ N (2 D) + O(1 D) 4 0.31 ± 0.05
H. N O+ (0) + e− (5.6 eV ) −→ N (2 P ) + O(1 D) 5 0.21 ± 0.05
H. N O+ (0) + e− (5.6 eV ) −→ N (2 D) + O(1 S) 6 0.09 ± 0.05
H. N O+ (0) + e− (5.6 eV ) −→ N (2 P ) + O(1 S) 7 0.10 ± 0.05
Table 7: Comparison between experimental information available on the branching frac-

tions of the dissociative recombination of N O+ (X 1 Σ+ , v). The value v̄ = 1.5 means that
N O+ is weakly vibrationally excited. The value ¯ = 0.34 eV means that the collision
energy is distributed according to a Maxwell distribution. K., V. − C. and H. mean Kley
et al. (1977), Vejby-Christensen et al. (1998) and Hellberg et al. (2003) respectively.
Energy γ1 γ2 γ3 γ4 γ5 γ6 γ7
range (eV )
[0; 0.81[ 0.07 0.93 0.00 0.00 0.00 0.00 0.00

[0.81; 1.59[ 0.15 0.75 0.10 0.00 0.00 0.00 0.00
[1.59; 2.78[ 0.03 0.55 0.11 0.31 0.00 0.00 0.00
[2.78; 3.81[ 0.03 0.34 0.11 0.31 0.21 0.00 0.00
[3.81; 5[ 0.03 0.25 0.11 0.31 0.21 0.09 0.00
[5; 10[ 0.03 0.15 0.11 0.31 0.21 0.09 0.10
Table 8: Branching fractions γi () for i-process (see Table 7) adopted for the calculation
of the temperature dependent branching fractions of the dissociative recombination of
N O+ (X 1 Σ+ , 0) with electrons.
9 - 28 RTO-EN-AVT-142
where x = /(kB Te ) and σDR the dissociative recombination cross section. Figure 15
shows the branching fractions calculated with the help of the cross section determined
by the MQDT approach. In the following, we have assumed that the branching fractions
behave for v > 0 like for v = 0.
v
v
8
10
0 v∆
i=2
vδ y T T
0δ ∆ ∆
10
−1 i=1 p Tδ
Te δ
i=3 p
∆ Te
10
−2
y
0δ ∆
γi(Te)
i=4
y
10
−3
i=5 0δ ∆ 0 δ ∆ y
Stagnation
i=7 streamline
10
−4 i=6 R
v
8
−5
10
2000 3000 4000 5000 6000 7000 8000 9000 10000
Te(K)
Blunt body
ry layer
Bounda
Figure 15: Temperature dependent Shock layer
branching fractions obtained with our cal- t
k fron
culated cross section for the dissociative re- Shoc
combination of N O+ (X 1 Σ+ , 0) for the dif-
ferent channels investigated by Hellberg et Figure 16: Considered situations of re-
al. (see Table 7 and 8). entry displaying the pressure, velocity and
temperatures profiles along the stagnation
streamline.
The fact that the rate coefficient depends strongly upon v has led us to develop a state
to state kinetics for N O+ .
4.2.4 Vibrational processes

Due to the lack of experimental or theoretical investigations allowing the determination
of the cross section of vibration-vibration (VV) and vibration-translation (VT) processes
involving N O+ , we assume that this ion behave like N2 for which Armenise et al. (1996)
and Billing and Fisher (1979) have produced very useful data. Conversely, the case of
vibration-electron (VE) processes like:
N O+ (X 1 Σ+ , vi ) + e− → N O+ (X 1 Σ+ , vf ) + e− (45)
corresponding to inelastic collisions when vf > vi and to super-elastic collisions in the

reverse case can be treated by the MQDT approach. The cross sections have been de-
termined and the Maxwellian rate coefficients derived. These rate coefficients have been
fitted under the form (28).
RTO-EN-AVT-142 9 - 29
4.3 Results
4.3.1 Ionization situation
Although the CR model elaborated with the help of the previous data can work up to
the atmospheric pressure, we will discuss in the following some typical re-entry situations
only displayed for instance by Fig.16 where the pressure is clearly lower.
We have chosen to develop our study from a temporal point of view as already developed
in the section devoted to the elaboration of CR model for atoms. This choice will be
justified later.
The general balance equation for a species X on one of its excited levels i can be written
indeed by displaying the hydrodynamic derivative D/Dt relevant under a Lagrangian
description of the flow:

D[X(i)] ~ ~v = −∇~ J~X + ∂[X(i)]
+ [X(i)] ∇ i
(46)
Dt ∂t CR
This form is particularly well adapted to estimate the importance of each contribution
in the balance: comparing their characteristic time scale is enough. In order to esti-
mate correctly the one related to the chemical source [∂[X(i)]/∂t]CR , we assume that the
characteristic time scale for the fluid expansion of contraction taken into account in the
balance equation by the term [X(i)] ∇ ~ ~v and that of the diffusion term ∇~ J~X tend to
i
infinity. The calculations are performed assuming that the Maxwell-Boltzmann equilib-
rium is reached for the kinetic energy of atoms and molecules at TA and electrons at Te .
These assumptions are verified a posteriori by comparing the characteristic time scale for
reaching a Maxwellian equilibrium with that of the chemistry. Since the electrons are
produced owing to the collisions resulting from the high temperature for heavy species,
we can have Te ≤ TA .
We can observe on Fig.17 and Fig.18 the time evolution of the number density of the
species of Table 4 when a fluid particle with T ' 200 K crosses the shock front and
is suddenly in high temperature (7000 K for Fig.17 and 10000 K for Fig.18) and high
pressure (6 kP a for Fig.17 and 15 kP a for Fig.18) conditions. We can observe that the
evolution of population densities starts immediately from t = 0 and stops at τc = 3×10−5 s
for Te = TA = 7000 K and at τc = 4 × 10−4 s for Te = 5000 K in the case of Fig.17. The
fact that the evolution starts immediately from t = 0 is due to the very strong initial
non equilibrium. The initial concentrations correspond as already mentioned to those
before the shock front at high altitude typically where the temperature is approximately
T ' 200 K: there is no populated excited states in these conditions. As soon as the
thermodynamic conditions change due to the crossing of the shock front, the density of
species begin to evolve. When electrons and heavy species are in the same thermal con-
ditions downstream the shock front, (Te = TA ), τc is shorter and the state following this
moment is quite close to chemical equilibrium, the pressure being sufficiently high. When
Te < TA , this state is also quite close to chemical equilibrium and is almost the same than
the latter. One of the important difference between both cases we can observe concerns
τc . Although the electrons have a relatively weak influence on the final state, their den-
sity remaining low, decreasing their temperature Te from 7000 K to 5000 K modifies τc
by one order of magnitude. Figure 17 shows clearly that the negative charged particles
are not electrons during the first part of the evolution (t < 10−6 s to 10−5 s according to
9 - 30 RTO-EN-AVT-142
the level of Te ). Changing Te cannot therefore be ascribed to the action of the electrons.
This behavior is in fact the result of the dissociation of N2 , O2 and N O under their own
impact which leads to higher concentrations observed before 10−7 s for Te = TA . Indeed,
the rate coefficients that we use are derived from the paper of Losev and Shatalov (1990)
and depends on both TA and Te assuming an equilibrium between the electrons and the
vibration of the molecules.
N2 N
23 23
10 10
O2 O2
22 22 + +
10 10 N2 e− NO O
21
NO 21
O
10 10
20 O 20
10 e− 10
NO
N
19 19
[X(i)](m )
[X(i)](m )
10 10
−3
−3
−
O −
+ O
10
18 NO 10
18
+
O2 + − +
17 N2 O O2 17
10 − 10
O2
16 16
10 − + 10 e− + + +
O O − O2 N2 N
+
O
10
15 e− N 10
15 −
O2
−
O2
14 14
10 −10 −9 −8 −7 −6 −5 −4 −3 −2 10 −10 −9 −8 −7 −6 −5 −4 −3 −2
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
t(s) t(s)
Figure 17: Time evolution in a typical re- Figure 18: Time evolution in a typical re-
entry situation: ionization (thermal equi- entry situation: ionization (thermal equi-
librium: Te = TA = 7000 K in thick lines librium: Te = TA = 10000 K in thick lines
and 5000 K = Te < TA = 7000 K in thin and 8000 K = Te < TA = 10000 K in thin
lines for a case of thermal disequilibrium, lines for a case of thermal disequilibrium,
negative ions in dot-dashed lines). negative ions in dot-dashed lines).
This hypothesis is highly questionable in the present case, the number density of electrons
N O+ (X)
being very weak before 10−7 s: this is confirmed by the evolution of Tv displayed
on Fig.19. This temperature is defined by:
+ (X) 1
TvN O =− h i (47)
d ln([N O+ (X,v)]/Z R (v))
kB dE(X,v)
lsl
where ZR (v) is the rotational partition function of the vibrational level v considered whose
energy is E(X, v) and lsl means that the derivative is in fact the slope of the least squares
line.
Before 10−7 s, the population density of the vibrational states are increasing and the
efficiency of the elastic collisions with electrons or heavy particles is insufficient to ensure
rapidly an equilibrium: the vibrational distribution does not follow a Boltzmann equilib-
rium neither with Te nor TA and depends on chemical processes. This is confirmed by
Table 9 listing the main processes responsible for the creation and destruction for N , O,
N O, O2− , O− , electrons and N O+ for t ' 10−7 s. This list shows clearly that populating
the state v = 0 for N O+ (X 1 Σ+ ) is dominated by associative ionization between O and
N (2 D) atoms and the inverse process of neutralization between N O+ and O2− in a reduced
RTO-EN-AVT-142 9 - 31
8000
7500
7000
6500
6000
5500
(K)
5000
NO+(X)
Tv 4500
4000
3500
3000
2500
2000 −11 −10 −9 −8 −7 −6 −5 −4 −3 −2
10 10 10 10 10 10 10 10 10 10
t(s)
Figure 19: Time evolution of the vibrational temperature of N O+ (X 1 Σ+ ) in the case of Fig.17,
the gas having crossed the shock front and whose temperatures are such that Te = TA = 7000 K
(thick line) or 5000 K = Te < TA = 7000 K (thin line). The thin dot-dashed line refers to the
typical recombination inside the boundary layer with p = 2 kP a from Te = TA = 7000 K to
Te = TA = 3000 K (see Fig.20 and 21) and the thick dot-dashed line refers to the same situation
but with p = 10 kP a (see Fig.22 and 23).
Processes (α) (β)
→N N2 + O → N O + N N 2 + N2 → N + N + N2
N→ N + O2 → N O + O(1 D) N + O2 → N O + O
→O O 2 + N2 → O + O + N2 O2 + O2 → O + O + O2
O→ O + N2 → O(1 D) + N2 O + N2 → N O + N
→ NO N2 + O → N O + N O2 + N → N O + O(1 D)
NO → N O + N2 → N + O + N2 NO + O → O + O + N
→ O2− O2 (a) + O− → O2− + O O2 + O2 → O2− + O2+
O2− → O2− + O2 → O2 + O2 + e− O2− + O → O− + O2 (a)
→ O− O2 + e− → O− + O O2− + O → O− + O2 (a)
O− → O− + O2 (a) → O2− + O O− + O → O2 + e−
→ e− O2 + O2− → e− + O2 + O2 O− + O → e− + O2
e− → e− + O2 → O− + O e− + O2+ → O + O(1 D)
→ N O+ O + N (2 D) → N O+ + e− N O + O2 → N O+ + O2−
N O+ → N O+ + N2 → N O+ (v = 1) + N2 N O+ + O2 → N O+ (v = 1) + O2
Table 9: Main processes at time t = 10−7 s during the formation of the plasma illustrated
by Fig.17 when Te = TA . The arrow indicates that the species X is produced (→ X) or
destroyed (X →). The second column (α) corresponds to the more efficient process and
the third one (β) to that just follows. The species are on their ground state if there is no
indication and N O+ means N O+ (X 1 Σ+ , v = 0) except in the case v = 1.
9 - 32 RTO-EN-AVT-142
way while populating the higher excited vibrational states are resulting from collisions
with N2 and O2 molecules. Afterwards, elastic collisions with electrons remain not very
efficient, ne remaining relatively low: this is illustrated by Fig.17 where changing the
N O+ (X)
electron temperature from 7000 K to 5000 K does not change Tv by a same factor.
Anyway, this behavior illustrates the high sensitivity of the kinetics towards the hypoth-
esis adopted under high degree of non equilibrium and the limitations of our CR model.
Table 9 is also interesting because it provides understanding of the global mechanism of
the plasma formation by coupling with Fig.17. Although the number density of O2− and
O− are completely negligible when the steady state is reached, these species play a key
role in the ionization phase. It is often assumed that the ionization in shock waves is
driven by associative ionization. When the gas contains molecules, they are dissociated
under heavy particles impact and the atoms produced associate to form electrons. This
is particularly true in the case of noble gases where atoms already exist: the associative
ionization can start immediately (Drawin and Emard (1973)). We see that this is not the
case for molecules being electronegative or having the possibility to form such species.
The most important negative charged particles are O2− and O− whose formation is de-
rived from the charge transfer process O2 (a) + O− → O2− + O and the inverse process
of neutralization O2 + O2 → O2− + O2+ for O2− and O2 + e− → O− + O and the charge
transfer process O2− + O → O− + O2 (a) for O− . These processes delay the formation
of free electrons and form a reservoir which will be a subsequent source for these parti-
cles: in order of decreasing importance, the electrons are produced first by the process
O + O− → O2 + e− and second by the associative ionization O + N (2 D) → N O+ + e− at
time t = 2 × 10−6 s for Te = TA . The numerical study of air shock waves do not usually
involve these processes (Abe et al. (2002)). Nevertheless, it could be interesting to adopt
them and observe their influence.
Apart from the problem of the vibrational distribution not necessarily in equilibrium dur-
ing the first part of the evolution, the concentrations calculated for t < 5 × 10−8 s are
not significant since it corresponds to too short time scales in regard to that for reaching
Maxwellian equilibrium for electrons and heavy species. Moreover, assuming a linear de-
crease of the velocity through the shock front between v∞ and v∆ (see Fig.16) allows to
estimate the duration of crossing the shock front (having a thickness equal to the mean
free path λ under upstream conditions) of a fluid particle such that:
λ v∞
τsc = ln (48)
v∞ − v∆ v∆
For a mean free path λ ' 10−3 m in order of magnitude, we have τsc ' 10−6 s. Before this
value, the fluid particle is not yet in the conditions downstream the shock front and the
calculation has to be considered prudently. Nevertheless, this calculation is interesting
because it provides the chemical scheme leading to the formation of the plasma starting
from a cold gas when the conditions change suddenly. Anyway, the state following τc is
quite close to equilibrium: it depends very weakly on the initial conditions. Assuming
once again a linear decrease of the speed from v∆ to 0 between y = ∆ and the wall, and
assuming also that the cold gas is suddenly in the conditions just behind the shock front
at t = 0, the y-location of the fluid particle at the time t is:
y(t) = ∆ e−v∆ t/∆ (49)
RTO-EN-AVT-142 9 - 33
When the near-equilibrium is reached (either at the time τc = 3 × 10−5 s with Te = TA =

7000 K or at the time τc = 4 × 10−4 s for the other case related to Fig.17), the fluid
particle is located either in y ' 6.3 cm or in y ' 0.8 cm respectively. In both cases, one
can conclude that the edge of the boundary layer is therefore close to chemical equilibrium
because y ≥ δ.
The global behavior when TA = 10000 K is the same as we can see on Fig.18. Due to
strong non equilibrium, the time evolution starts immediately but stops approximately
at the same time: when t = τc = 9 × 10−7 s for Te = TA = 10000 K and when t =
τc = 1.5 × 10−6 s for 8000 K = Te < TA = 10000 K. This is due to the same processes
of dissociation discussed previously whose rate coefficient varies more slowly between
8000 K and 10000 K than between 6000 K and 8000 K. As in that case, Fig.18 shows
that the electrons are not the most significant negative charged particles. The first part
of the evolution is also questionable regarding the characteristic time scales for reaching
Maxwell-Boltzmann equilibrium for electrons and heavy species as well as regarding τsc ,
Eq.(48) leading to the estimation τsc = 2 × 10−7 s. The same conclusion can be derived
from Eq.(49): whatever the conditions, the fluid particle is not still in the boundary layer
when the near-equilibrium is reached: its location is y ' 4.8 cm for thermal equilibrium
(Te = TA = 10000 K) and is y ' 4.7 cm when Te = 8000 K. The chemical relaxation
is therefore extremely rapid and the boundary layer’s edge is therefore close to chemical
equilibrium. Finally, the most important processes leading to the plasma formation are
globally the same as those listed in Table 9.
4.3.2 Recombination situation
In the literature, different simplified kinetic schemes have been implemented in Navier-
Stokes codes to study high temperature air flows and particularly boundary layers. Most
of these schemes were developed during the 70’s and 80’s in the frame of the shuttle’s
building. We have chosen to discuss three widely used simplified kinetic schemes due to
Dunn and Kang (1973), Gupta et al. (1990) and Park (1993), the latter being an up-dated
version of an older one Park (1985). Except O2− and O− , they consider the same species
as those we have taken into account. In addition, apart from the scheme of Park, they
assume a thermal equilibrium between electrons and heavy particles.
Conversely to the case of the kinetic scheme of Park, those of Dunn & Kang and Gupta
contain the rate coefficient of each inverse process: as a result, we have calculated each in-
verse rate coefficient for the kinetic scheme of Park starting from the equilibrium constant.
Its value is derived from his own work (Park (1990)) concerning the partition function of
the involved species. In addition, all radiative processes have been removed in our CR
model in order to test the kinetic scheme only.
The situations adopted for illustrating the difference between these schemes and our CR
model are related to two pressure levels of 2 kP a and 10 kP a framing the typical range
over which a re-entry occurs. In both cases, the plasma is initially at equilibrium with
Te = TA = 7000 K and is recombining at Te = TA = 3000 K. We can see on Fig.20 the
time evolution of the neutral species and that of the charged species except electrons on
Fig.21 calculated with our CR model and with the three other kinetic schemes. Figures
22 and 23 present the calculated evolutions in the same conditions but for p = 10 kP a.
Figure24 illustrates finally the time evolution of the electron density in the same thermal
9 - 34 RTO-EN-AVT-142
conditions as previously considered with p = 2 kP a (α) and with p = 10 kP a (β).
15
22 10
10 N2 +
21 10
10 N2
10 5
21
1016
10 10
10
19 O2 10
14
+
17
10
12 O2
10 10
10
15 10
18
[X(i)](m )
20
[X(i](m )
10
−3
10
−3
NO 16 +
18 10 NO
10 14
10
16 12
10 10
22 16
10 10
11 +
20 N 10 N
10 6
10
18 1
10 10
17
O 10
+
12 O
10
22 10
7
−10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1 2 10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
t(s) t(s)
Figure 20: Time evolution of N2 , O2 , N O, Figure 21: Time evolution of N2+ , O2+ ,
N and O in a typical re-entry situation: re- N O+ , N + and O+ in a typical re-entry sit-
combination at p = 2 kP a from Te = TA = uation: recombination at p = 2 kP a from
7000 K to Te = TA = 3000 K (thick solid Te = TA = 7000 K to Te = TA = 3000 K
line: our model, thin solid line: Dunn & (thick solid line: our model, thin solid line:
Kang, thin dotted line: Gupta, thick dot- Dunn & Kang, thin dotted line: Gupta,
dashed line: Park). thick dot-dashed line: Park).
Since the plasma considered in this section is optically thick in order to make a compar-
ison between chemical processes involved by the kinetic schemes only, the final state is
an equilibrium state. The agreement is rather good for this final equilibrium state con-
cerning N2 , O2 , N O, N and O except for the simplified kinetic scheme of Gupta which
leads to a final overpopulation of N and to a final underpopulation of N2 and N O. A
calculation under a higher pressure condition as the atmospheric one show the same trend
with in addition an overpopulation of O2 . This behavior is the direct result of the influ-
ence of the equilibrium constants. In particular, those of the Zeldovich’s reaction (32a) is
lower leading to a lower value for the rate coefficient in the case of the kinetic scheme of
Gupta: the population density of O2 and N are then higher while those of N2 and N O
are lower. We can observe for the ions on Figures 21 and 23 that their population density
is of the same order of magnitude except for O2+ and N O+ . At atmospheric pressure,
O2+ is overpopulated in the final equilibrium state in the case of the kinetic scheme of
Gupta. This is also the direct result of the different equilibrium constants involved by
each kinetic scheme. Moreover, we can observe that N O+ is the more significant ion in
the final equilibrium state: the temperature being relatively low, the number density of
N O+ is the highest one as a result of the high concentration of N O and its relatively low
ionization characteristic temperature (' 107500 K).
The final electron density is in strong disagreement from a kinetic scheme to another.
RTO-EN-AVT-142 9 - 35
24
10
15
23 10
10 N2 +
22 10
10 N2
10 5
22
1016
10 10
10
20 O2 10
14
+
18
10
12 O2
10 10
10
16 10
18
[X(i)](m )
21
[X(i](m )
10
−3
10
−3
NO 16 +
19 10 NO
10 14
10
17 12
10 10
22 16
10 10
11 +
20 N 10 N
10 6
10
18 1
10 10
23 17
10 O 10
+
12 O
10
7
−10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1 2 10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
t(s) t(s)
Figure 22: Time evolution of N2 , O2 , N O, Figure 23: Time evolution of N2+ , O2+ ,
N and O in a typical re-entry situation: re- N O+ , N + and O+ in a typical re-entry sit-
combination at p = 10 kP a from Te = TA = uation: recombination at p = 10 kP a from
7000 K to Te = TA = 3000 K (thick solid Te = TA = 7000 K to Te = TA = 3000 K
line: our model, thin solid line: Dunn & (thick solid line: our model, thin solid line:
Kang, thin dotted line: Gupta, thick dot- Dunn & Kang, thin dotted line: Gupta,
dashed line: Park). thick dot-dashed line: Park).
20
10
19
10
18
10
17
ne(m )
10
−3
16
10
15
10 (α)
14
10
13
10
20
10
19
10
18
10
17
ne(m )
10
−3
16
10 (β)
15
10
14
10
13
10
−10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10 10
t(s)
Figure 24: Time evolution of the electron density for a recombining plasma from Te = TA =
7000 K to Te = TA = 3000 K with p = 2 kP a (α) and with p = 10 kP a (β) (thick solid line: our
model, thin solid line: Dunn & Kang, thin dotted line: Gupta, thick dot-dashed line: Park).
9 - 36 RTO-EN-AVT-142
The reason can be ascribed once again to the equilibrium constants, but also to the elec-
troneutrality: since N O+ has the highest number density among the ions and since our
kinetic scheme provides the highest population density for this molecular ion, the value
of ne in the final equilibrium state is then the highest one.
Although the final state is related with too long time scales with respect to the conditions
typically involved in the re-entry conditions, nevertheless its study is very important: this
informs about the quality of the equilibrium constants which are crucial because they
provide the calculation of the inverse rate coefficients and consequently they influence
the time evolution of the species number densities. The comparison between the good
correlation of the final state for the neutral species and the bad one for that related to the
charged species lead us to conclude that the equilibrium constants for the neutral species
are better than for the charged ones.
Except the influence of the rate coefficients, the previous conclusion explain in part the
fact that the different population densities evolve strongly in a different way during the
transient step. We have already mentioned the influence of the rate coefficient for the
Zeldovich’s reaction (32a) which is very different from a kinetic scheme to another. The
equilibrium constant being also different from a kinetic scheme to another, it is clear that
the time evolution has no reason to be the same one. Nevertheless, the discrepancy is
sometimes in a ratio higher than 50. For relatively short time scales as those related to
typical re-entry (' 10−4 s), the concentrations of O2 and N O are very different.
In addition, this evolution depends strongly on the adopted initial conditions: equilibrium
at T = 7000 K induces here higher concentrations for N + and O+ than for N O+ : as a
result, the latter ion plays a relatively minor role. For a lower temperature as considered
in the next section, one have [N O+ ] = 4.9 × 1018 m−3 at Te = TA = 5400 K while [N + ] =
1.9 × 1017 m−3 and [O+ ] = 2.4 × 1017 m−3 when p = 2 kP a and [N O+ ] = 1.9 × 1019 m−3 ,
[N + ] = 1.4 × 1017 m−3 and [O+ ] = 3.5 × 1017 m−3 when p = 10 kP a. In this case, N O+
is the more important ion and the recombination of electrons is driven essentially by its
DR.
Figure 24 illustrates the evolution of ne : the different kinetic schemes are globally in
strong disagreement, even if the schemes of Dunn & Kang and Gupta are relatively well
correlated for relatively short time scales. They underestimate clearly the electron density
in relation to that calculated by the other schemes for t < 10−4 s which is of interest in
the case of a typical re-entry. In the case electrons play a role the temperature being
sufficiently high, the kinetic scheme chosen to solve the boundary layer’s equations will
lead to very different results. We have made a special effort to describe correctly the
recombination and as a result the ionization via the equilibrium constants: we can affirm
that the recombination is too strong with the three common simplified kinetic schemes
here studied.
Finally, Figure 19 shows that the relaxation of the vibrational distribution for N O+ (X 1 Σ+ )
is almost complete when the transient regime begins for the other species: the assump-
tion of a Maxwell equilibrium between the vibration of molecules with the other species
is validated.
RTO-EN-AVT-142 9 - 37
4.4 Conclusions
Three main conclusions can be derived from the study of the collisional radiative model
for air. First, ionization situations have shown that O2− and O− , due to their unsteady
behavior, may have an influence on the characteristic time scale necessary to reach the
equilibrium state. The common approach is to ignore these ions in shock layer computa-
tions. It could be interesting to account for them in a shock calculation.
The second important conclusion is that a CR model has to be time-dependent: it provides
chemical characteristic time scales which can be compared with the other characteristic
time scales of the flow. This temporal approach is also particularly powerful to check the
quality of the equilibrium constants (by the equilibrium state finally reached and also by
the transient regime).
Third, the comparison between the three common simplified kinetic schemes used for
re-entry studies and the detailed CR model has shown that the results are very different
from a scheme to another. This is the reason why it is clear that the better solution to
validate a kinetic scheme is in fact to study the transient regime. We have shown that
the final state allows to test the quality of the equilibrium constants only. The next step
of this work is to derive a new simplified kinetic scheme to be implemented in a Navier-
Stokes code devoted to re-entry studies. A time-dependent approach will allow to verify
the relevance of this new reduced kinetic scheme over a wide temporal range.
5 Stagnation point boundary layer

A method has been proposed by researchers (Kolesnikov (1999)) at IPM (Moscow) to
determine the catalytic properties of thermal protection systems (TPS). This method is
indirect and is based on comparisons of the wall heat flux measured on a TPS sample
set in a high enthalpy flow generated by a ICP torch with the flux calculated using a
stagnation point boundary layer code. This method is popular and is used in different
places in Europe, for example, at CORIA (France), VKI (Belgium) and in EADS-ST
(France). Studies have been carried out for different materials. Most results available
in the literature are for pure silica or for material containing mostly silica (e.g. RCG
contains 92% silica). In this section, we present the stagnation point boundary layer code
for air with different models for catalycity effects. Then for two typical flow conditions
encountered in catalycity studies, the influence of the catalycity model and of the kinetic
scheme is disccused.
5.1 Stagnation point boundary layer code

Classical boundary layer equations are written assuming that the flow is axisymmetric
and laminar, dissociated but non ionized, in chemical nonequilibrium and in thermal
equilibrium. Using the classical Lees-Dorodnitsyn (x, y) → (ξ, η) transformation, the
boundary layer equations at the stagnation point reduce to a set of ordinary differential
equations (Barbante (2001))
• continuity
∂V 0
+f =0 (50)
∂η
9 - 38 RTO-EN-AVT-142
• momentum 0 0
∂f 1 ρδ 0 ∂ ∂f
V = −f 2 + l0 (51)
∂η 2 ρ ∂η ∂η
• species i continuity
∂yi ∂Jiη
V + = Wi (52)
∂η ∂η
• energy
N
! Ns
!
s
∂g ∂ l0 ∂g ∂ l0 X ∂yi hi ∂ X hi
V = − − Jiη (53)
∂η ∂η P r ∂η ∂η P r i=1 ∂η hδ ∂η i=1
hδ
with
0 u
f = (54)
uδ
and
h
g= (55)
hδ
where the subscript δ indicates a quantity evaluated at the boundary layer outer edge. yi
is species i mass fraction, u is the velocity component parallel to the wall and h is the
enthalpy of the mixture. l0 is defined as
ρµ
l0 = (56)
ρδ µ δ
The local Prandtl number is
µcp
Pr = (57)
λ
The dimensionless diffusion flux is
Jiy
Jiη = q (58)
2ρδ µδ ∂u
∂x
δ
and the dimensionless mass production term is

wi
Wi = ∂uδ (59)
2ρ ∂x
where wi is the mass source term of species i in the (x,y) framework.
In order to carry out a stagnation point boundary layer computation, it is necessary to
know the wall temperature Tw and three parameters at the boundary layer edge: the
velocity gradient β = ∂u
∂x
δ
, Tδ and Pδ .
It is important to mention that to do comparisons with catalycity studies in subsonic flows
generated by an ICP torch, the finite thickness of the boundary layer at low Reynolds
numbers has to be taken into account (Barbante (2001)). Then, the set of equations
for the stagnation point boundary layer has to be slightly modified. These corrections
require to know precisely the boundary layer thickness and different velocity gradients.
In this work, we propose to discuss the influence of different models on the results of the
stagnation point boundary layer simulation, and therefore these corrections have been
neglected.
RTO-EN-AVT-142 9 - 39
5.1.1 Catalycity and boundary conditions

Catalycity effects are generally taken into account using two types of dimensionless co-
efficients: a recombination probability γ (0 ≤ γ ≤ 1) and an energy accommodation
coefficient at the wall β. The γi coefficient for a species i is defined as
Fi,recomb = γi Fi (60)
where Fi is the flux of species i impinging on the wall, and Fi,recomb is the recombining
flux of species i on the wall. Assuming a Boltzmann distribution of velocities close to the
wall r
kB Tw
Fi = ni (61)
2πmi
where mi is the mass of species i. Then, the production rate of species i at the wall wi,cat
can be written as r
kB Tw
wi,cat = γi mi ni (62)
2πmi
The β coefficient is defined as
q̇
β= (63)
DFrecomb /2
where q̇ is the effective energy flux given to the wall, D is the dissociation energy of the
molecule formed. Then the heat flux transferred to the wall depends on the product
0 0
γ = γβ. It is interesting to note that γ and γ are equal when the time of stay of the
formed molecules on the surface is long enough. In this case, the whole recombination
energy is given to the wall (β = 1). In fact, there has been only a few studies on the β
coefficient determination and its modeling (Daiss et al. (1997); Scott (1990); Suslov and
Tirskyi (1994); Barbato et al. (1998)), therefore generally this coefficient is assumed to
be equal to 1 and only the γ coefficient is used. In this work, we have also used this
approximation. Then, boundary conditions are written as
At the wall (η = 0)
u=v=0 (64)
h = h(Tw ) (65)
y
Ji,w = wi,cat (66)
y
where Ji,w is the diffusion flux of species i in the direction normal to the wall. At the
boundary layer edge (η → ∞)
u = uδ (67)
h = hδ = h(Tδ , yi,δ ) (68)
yi = yi,δ (69)
The boundary layer flow is assumed to be in chemical equilibrium at the boundary layer
edge, then yi,δ = yi (Tδ , Pδ ). Finally, the wall heat flux is given by

Ns
y ∂T
X y

qw = −λw + hi Ji (70)
∂y w i=1

w
9 - 40 RTO-EN-AVT-142
5.1.2 Thermochemical model

In the stagnation point boundary layer code, air is considered as a 5 species mixture (N2 ,
O2 , N O, N and O). To determine the viscosity and the thermal conductivity of the
mixture, the mixing laws given by Gupta et al. (1990) have been used with the collision
integrals calculated recently by Capitelli et al. (2000b). For boundary layer flows, it is
particularly important to calculate accurately diffusion fluxes. The classical Fick’s law
gives a simple expression of diffusion fluxes. For example, the y component of the diffusion
flux of species i in the mixture is
∂yi
Jiy = −ρDim (71)
∂y
where Dim is the diffusion coefficient of species i in the
PNmixture. However, this simple
s y
expression does not satisfy the mass conservation: i=1 Ji = 0 (unless all the Dim
coefficients are equal) and therefore should not be used when diffusion processes are
important. The rigorous expression of diffusion fluxes is given by Stefan - Maxwell and
Sutton and Gnoffo (1998) have proposed an efficient iterative technique to determine
them. In this case, the calculated fluxes satisfy the constraint of mass conservation.
In this work, the kinetic schemes proposed by Dunn and Kang (1973), Park (1993) and
Gupta et al. (1990) adapted to a five species air mixture have been used.
5.1.3 Numerical method

The boundary layer equations (except the continuity equation) can be written in the
generic form
∂2f ∂f
a 2 +b + cf = d (72)
∂η ∂η
with boundary conditions
∂f
η = 0 : al + bl f = c l (73)
∂η
η = ηδ : f = fδ (74)
and are solved with a 4th order finite difference method (Barbante (2001)). 500 points
are taken in the η direction.
5.2 Catalycity models

5.2.1 Simplified models
The simplest model is to consider that only two recombination reactions occur at the
wall (i.e. N + N → N2 and O + O → O2 ) and that these two reactions have the
same γ coefficient. This model is widely used to determine an average γ coefficient for a
material. From different measurements in pure nitrogen and oxygen flows, Scott (1990)
and Kolodziej and Stewart (1987) have determined recombination rates of N in N2 and
O in O2 on different surfaces as a function of the wall temperature Tw . Scott’s fits for
recombination coefficients are
γN N = 0.0714e−2219/Tw , 950 < Tw < 1670K (75)
RTO-EN-AVT-142 9 - 41
γOO = 16.e−10271/Tw , 1400 < Tw < 1650K (76)

Kolodziej and Stewart (1987) give
γN N = 0.061e−2480/Tw , 1410 < Tw < 1640K (77)
γOO = 40.e−11440/Tw , 1435 < Tw < 1580K (78)

These rates are usually implemented in air flow codes assuming that no N O is formed on
the surface. They are very simple to use, but are valid in a small temperature range.
5.2.2 Phenomenological models

Different authors (Kovalev and Suslov (1996), Jumper and Seward (1994), Nasuti et al.
(1996), Deutschmann et al. (1995) and Daiss et al. (1997)) have developed phenomeno-
logical models in which the gas-surface interaction is represented by different elementary
reactions
• atom adsorption/desorption (Fig. 25)
• recombination of an atom of the gas with an atom adsorbed on the wall: Eley-Rideal
(E-R) type process (Fig. 26)
• recombination of two adsorbed atoms at the wall: Langmuir-Hinshelwood (L-H)

type process (Fig. 27)
For the temperatures and pressures studied, molecule adsorption at the wall is neglected.
0
A rate coefficient is associated to each elementary process. In this case, the global γ
coefficient depends on temperature but also on the gas composition close to the wall.
Figure 26: Eley-Rideal (E-R) type recom-

Figure 25: atom adsorption/desorption bination process
Figure 27: Langmuir-Hinshelwood (L-H)

type recombination process
9 - 42 RTO-EN-AVT-142
In the model derived by Deutschmann et al., simplified expressions for elementary rate
coefficients are used. In works of Jumper et al., Daiss et al. and Nasuti et al., a better
description of each process is given. In the model of Jumper et al., recombinations of O
in O2 and N in N2 are assumed to be Eley-Rideal type processes. In the models of Daiss
et al. and Nasuti et al., recombination reactions are assumed to be of both types. It is
interesting to note that the model of Nasuti et al. takes into account the N O formation
at the wall. In the model of Nasuti et al, the following processes are taken into account
• N and O atom adsorptions and desorptions
O + s ↔ O − s, N + s ↔ N − s (79)
• Eley-Rideal type processes
O − s + O ↔ O2 + s, N − s + N ↔ N2 + s (80)
O − s + N ↔ N O + s, N − s + O ↔ N O + s (81)
• Langmuir-Hinshelwood type processes
O − s + O − s ↔ O2 + 2s, N − s + N − s ↔ N2 + 2s (82)
O − s + N − s ↔ N O + 2s (83)
where s is a free surface site. With this approach, it is necessary to define four recombi-
nation probabilities at the wall : γOO , γN N , γN O , γON .
Expressions of γ coefficients
Due to a competition between O and N recombination processes at the wall, the four
recombination probabilities are not independent of each other. Therefore γN N and γOO
coefficients obtained in pure nitrogen and oxygen flows certainly overestimate these coef-
ficients in an air flow where both atoms are present. Furthermore, the number of surface
sites for N and O adsorption may be limited due to the chemisorption on the surface of
other species such as the H2 O molecule which forms Si-OH bounds on a silica surface. To
determine, the surface coverage rate by OH, Nasuti et al. proposes
θOH = A(1 − exp(−EOH /kB T )) (84)
The surface coverage rates θO et θN are related to the free surface sites θf by
θf = 1 − θO − θN − θOH (85)
with 0 ≤ θf ≤ 1. At the stationary state, the number of adsorbed atoms on the surface is
constant and therefore, the flux of adsorbed atoms i on the surface (Fai ) is equal to the
flux of atoms recombining following a L-H (Flhii ) or a E-R (Ferii ) process and to the flux
of atoms thermally desorbed (Ftdii ). Then, for O and N we have
FaO = FerN O + FerOO + FlhN O + FlhOO + FtdO (86)
RTO-EN-AVT-142 9 - 43
FaN = FerON + FerN N + FlhN O + FlhN N + FtdN (87)

As we will see in the following, these different fluxes are functions of the two variables
θO and θN . The system of equations (86) and (87) is therefore a quadratic system of
equations for θO and θN variables which can be easily solved by an iterative method
(e.g. Newton-Raphson method). When θO et θN are known, the different recombination
probabilities can be expressed as
(FlhOO + 2FerOO )
γOO = (88)
FO
(FlhN N + 2FerN N )
γN N = (89)
FN
(FlhN O + FerN O + FerON )
γON = (90)
FO
(FlhN O + FerN O + FerON )
γN O = (91)
FN
These four coefficients depend on pO , pN , θN , θO and T .
Wall boundary conditions for the different species

The production rate wi,cat at the wall of species i = N2 , O2 , N O, N and O can be written
as
wO = mO (γOO + γON )FO (92)
wN = mN (γN N + γN O )FN (93)
wO2 = −mO γOO FO (94)
wN O = −(mO γON FO + mN γN O FN ) (95)
wN2 = −mN γN N FN (96)
where Fi is the flux of species i impinging on the wall and is given by Eq. (61). The flux
of atoms i adsorbed at the wall Fai is proportional to Fi and to the number of free surface
sites θf
Fai = si θf Fi (97)
where si is a sticking coefficient (0 ≤ si ≤ 1). The flux Ferij of atoms i in the gas,
recombining with adsorbed atoms j to form an ij molecule following an E-R process is
proportional to Fi and θj
Ferij = γij∗ θj Fi (98)
where 0 ≤ γij∗ ≤ 1 with:
γij∗ = Perij exp(−Qerij /kB T ) (99)
In a L-H recombination process, both atoms are adsorbed at the wall. The flux of ad-
sorbed atoms which recombine following a L-H type process to form O2 , N2 and N O are
respectively
2
ZlhOO = 2νO nθO exp[−(QlhOO /kB T )] (100)
2
ZlhN N = 2νN nθN exp[−(QlhN N /kB T )] (101)
9 - 44 RTO-EN-AVT-142
ZlhN O = (νN + νO )nθN θO exp[−(QlhN O /kB T )] (102)

where n is the number of surface sites per unit surface and νi is a characteristic frequency
given by p
νi = (cA /∆) πkB T /2mi (103)
where ∆ is the mean distance between surface sites and cA is a constant. Activation
energies are given by
QlhOO = M AX(2QaO − DOO , EmO ) (104)
where Qai is the surface-atom bond energy for atom i and Emi is the energy necessary for
species i to move on the surface and Dij is the bond energy between i and j atoms.
QlhN N = M AX(2QaN − DN N , EmN ) (105)
QlhN O = QaO + QaN − DN O (106)

The flux of thermal desorption Ftdi is
Ftdi = nθi (kB T /h)exp[−Qtdi /kB T )] (107)
where Qtdi is an activation energy.
For N
sN 0.015
PerN N 0.1
Wall
QerN N 20kJ/mol
n 4.51018 m−2
QtdN QaN
∆ 510−10 m
For O
For N
sO 0.05
EmN 82.2kJ/mol
PerOO 0.1
QaN 530.8kJ/mol
QerOO 20kJ/mol
DN N 941.3kJ/mol
QtdO QaO /2.
For O cA 3.5
EmO 159.1kJ/mol
For OH
QaO 499.8kJ/mol
A 1.
DOO 498kJ/mol
EOH 2.5kJ/mol
Table 10: Data for the model derived by
Table 11: Parameters derived to have
Nasuti et al. (1996).
the best agreement with experimental
results obtained in pure nitrogen and
oxygen flows (Nasuti et al. (1996)).
Determination of the parameters of the model

Among the different parameters of the model, there are:
• parameters which depend on the model chosen for the surface : Qai , ∆, n and Emi
RTO-EN-AVT-142 9 - 45
• fundamental data available in the literature as bond energies Dij
• parameters which can be derived from semi-empirical models : Qerij , Qai and Qtdi ,
• parameters for which no theoretical model is available: Perij , si and cA .
The three first types of parameters are given in table 10. Figure 28 shows numerous
measurements reviewed by Jumper and Seward (1994) and Wiley (1993) of γN N et γOO
for pure silica or RCG in pure nitrogen and oxygen flows. These results have been used
to determine the parameters of the model given in table 11. For the thermal desorption
threshold, we have assumed that QtdN = QaN and QtdO is taken equal to QaO /2 to have
a maximum of γOO at T ' 1600K. On Fig. 28, the results obtained by the model
proposed by Jumper are also shown. In this model, recombination processes are of the
E-R type and Perij and si are assumed to depend on the temperature. Experimental
results seem to indicate that γOO has a maximum at T ' 1600K. This evolution is
well reproduced by both models and is due to thermal desorption which becomes very
efficient as temperature increases. With the model of Nasuti, γOO is independent of
pressure for T < 1000K and depends on the pressure for higher temperatures. For
O in O2 recombination, we have checked that in the studied wall temperature range
300K − 2000K, the L-H type process is negligible, and then the recombination process is
an E-R type recombination process. For γN N , experimental data are scarce and therefore
the existence of a maximum at T ' 1600K is not obvious. This maximum exists in the
model of Jumper but not in the model of Nasuti. With the model of Nasuti, γN N depends
on pressure only for 600K < Tw < 1000K. For T < 600K, the recombination process is
an E-R type process and it is a L-H process for T > 1000K. For air, no measurements
of the different γ coefficients are available. For the recombination of O in O2 and of N in
N2 , the parameters derived hereabove are used. The other parameters are given in table
12 following the assumptions of Nassuti et al.
−1 −1
10 10
Nasuti − P=1000Pa
experiments
Nasuti − P=1000Pa Jumper −100Pa
−2 Nasuti − P=100Pa −2
10 experiments 10 Jumper −1000Pa
Jumper −100Pa Nasuti −100Pa
Jumper −1000Pa
−3 −3
10 10
γΝΝ
γΟΟ
−4 −4
10 10
−5 −5
10 10
0.5 1.0 1.5 2.0 2.5 3.0 0.5 1.0 1.5 2.0 2.5 3.0
1000/T [K] 1000/T [K]
Figure 28: Recombination probabilities of N in N2 and of O in O2 on pure silica and RCG.

Symbols correspond to different experimental results reviewed in Jumper and Seward (1994)
and Wiley (1993).
9 - 46 RTO-EN-AVT-142
For ON
QerON QerN N
PerON PerN N
For NO
QerN O QerOO
PerN O PerOO
DN O 625.1kJ/mol
Table 12: Parameters for ON and NO (Nasuti et al. (1996)).
5.3 Results
In this section, we present results for two typical conditions of catalycity studies carried
out in high enthalpy flows generated by a ICP torch. For both test-cases, we propose to
discuss the influence of the model used for catalycity and for the gas kinetic scheme on
the calculated results.
5.3.1 High pressure test case : P = 10000P a

We consider typical boundary edge conditions at Pδ = 10000P a: Tδ = 4500K and
β = 11600s−1 (Barbante (2001)). The simplest model for catalycity presented in Sec-
tion 5.2.1 is to consider that only two recombination reactions occur at the wall (i.e.
N + N → N2 and O + O → O2 ) and that these two reactions have the same γ coefficient.
Figure 29 presents the wall heat flux as a function of the wall temperature for different
values of γ. In the classical method derived by IPM to determine the catalycity of a TPS,
measured heat fluxes are set in this figure to determine the γ coefficient of the studied
wall.
As expected the calculated wall heat flux increases as γ increases and as the wall temper-
ature decreases. For a given wall temperature, the heat flux in the fully catalytic case is
about twice the flux in the noncatalytic case. The heat flux for a fully frozen flow ( i.e.
no reactions in the gas and at the wall) is also shown. This flux is the lower heat flux
limit. In the noncatalytic case, the difference between the flux calculated in the frozen
case and the curve denoted γ = 0, shows the influence of the gas chemistry on the heat
flux. Figure 29 shows that this influence decreases as the wall temperature increases,
as the recombination of atoms in the gas is more efficient at low temperature. Figure
30 presents the wall heat flux calculated using the phenomenological model derived by
Nasuti et al and presented in Section 5.2.2 . At low temperature, this model shows that
the wall is noncatalytic and that its catalycity increases up to Tw ' 1400K. For higher
temperatures, thermal desorption becomes very efficient and the calculated wall heat flux
decreases.
RTO-EN-AVT-142 9 - 47
1.4e+06
γ=0
γ=1e−3 γ=0
1.2e+06 γ=1e−2 1.3e+06
γ=1e−3
γ=1e−1 γ=1e−2
γ=1. γ=1e−1
frozen flow Catalycity model − Dunn & Kang
1.1e+06
wall heat flux [W.m ]
1.0e+06
−2

−2
8.0e+05 9.0e+05
6.0e+05 7.0e+05
4.0e+05
5.0e+05
2.0e+05
500 1000 1500 2000 3.0e+05
Tw [K] 500 1000 1500 2000
Tw [K]
Figure 29: Wall heat flux as a function

Figure 30: Same as Fig. 29 with the heat
of the wall temperature for different values
flux calculated using the phenomenological
of the average γ coefficient. The kinetic
model derived by Nasuti et al.
scheme of Dunn and Kang is used.
Figure 31 shows that the γ coefficients of the different species calculated with the phe-
nomenological model have very different values from each other and different evolutions
with the wall temperature. At low temperature Tw ≤ 1000K, γOO and γN O are close to
each other and much higher than the two other coefficients. The γN N coefficient increases
as the wall temperature increases and becomes the highest recombination coefficient for
Tw ≥ 1800K. We note that the γON remains very small in the whole temperature range.
0
10
γNN 1.0
−1 γOO
10 γNO
θO
γON 0.8 θN
−2 θOH
10
surface coverage
γ coefficients
−3
0.6
10
−4
10 0.4
−5
10 0.2
−6
10
500 1000 1500 2000 0.0
Tw [K] 500 1000 1500 2000
Tw [K]
Figure 31: γ coefficients of the phe-

Figure 32: Different values of θ coefficients
nomenological model derived by Nasuti et
for the same condition as Fig. 31.
al. for the conditions of Fig. 30.
Figure 32 shows the surface coverage for O, N atoms and the OH molecule for the
conditions of Fig. 31. The surface coverage by the OH molecule decreases as the wall
temperature increases. The surface coverage by O atoms is on the average of 0.5 for
9 - 48 RTO-EN-AVT-142
300K ≤ Tw ≤ 1500K and then decreases rapidly for higher temperature due to thermal
desorption. The N atom surface coverage remains very small in the whole temperature
range.
Figure 33 shows the species concentration for Tw = 1500K and the phenomenological
model proposed by Nasuti et al. Strong variations are observed close to the wall for N ,
O2 and N O. Figure 33 compares the wall heat flux calculated with the phenomenological
model proposed by Nasuti et al. using different kinetic schemes (Dunn and Kang , Park
and Gupta) and for a frozen flow. In the scheme of Park, backward rate coefficients are
derived from forward rates using equilibrium constants fitted as a function of temperature
for T ≥ 500K. To avoid numerical problems at low temperatures with these fits, we
present results for the kinetic scheme of Park for T ≥ 1000K. The heat fluxes calculated
using the kinetic scheme of Dunn and Kang are very close to the results obtained for
the scheme of Gupta for T > 700K. The most significant discrepancies are observed for
lower temperatures. The heat fluxes calculated using Park’s kinetic scheme are slighlty
higher than those obtained with Dunn and Kang scheme but the discrepancy remains
small. The heat flux calculated assuming a frozen flow is the lower heat flux limit.
We note that the discrepancies with the results obtained assuming a reacting boundary
layer are rather small and decrease as the wall temperature increases. Then, for wall
temperatures Tw ' 1500K usually encountered in catalycity studies, the flow is nearly
frozen at P = 10000P a. Figure 34 also shows the heat flux calculated using the kinetic
scheme of Dunn and Kang and Fick’s law for diffusion fluxes. In this case, the influence
of the modeling of the diffusion flux appears to be rather small and the use of Fick’s law
slightly overestimates the calculated wall heat flux.
0
10
7.5e+05
N2
O2 7.0e+05
NO
species concentrations [mol.m ]
−1
−3
10 N
0
6.5e+05
−2
−2 6.0e+05
10
5.5e+05
Dunn and Kang
Gupta
−3
10 5.0e+05 Park
Dunn et Kang − Fick’s law
frozen flow
4.5e+05
−4
10
0 1 2 3 4 5 6 7 8 4.0e+05
500 1000 1500 2000
η Tw [K]
Figure 33: Species concentrations in the Figure 34: Wall heat flux calculated for
boundary layer for Tw = 1500K using different gas kinetic schemes and with the
the phenomenological model. The kinetic kinetic scheme and Dunn and Kang and the
scheme of Dunn and Kang is used. simple Fick’s law for diffusion fluxes.
RTO-EN-AVT-142 9 - 49
5.3.2 Low pressure test-case : P = 2000P a

We consider typical boundary edge conditions at Pδ = 2000P a : Tδ = 5400K and β =
14000s−1 (Bourdon et al. (2003)).
γ=0
1.2e+06 γ=1e−3
γ=1e−2 1.2e+06 γ=0
γ=1e−1 γ=1e−3
γ=1. γ=1e−2
1.0e+06 frozen flow γ=1e−1
1.0e+06 Catalycity model
−2

−2
8.0e+05
8.0e+05
6.0e+05
6.0e+05
4.0e+05
4.0e+05
2.0e+05
500 1000 1500 2000 2.0e+05
Tw [K] 500 1000 1500 2000
Tw [K]
Figure 35: Wall heat flux as a function

Figure 36: same as Fig. 35 with the heat
of the wall temperature for different values
flux calculated using the phenomenological
of the average γ coefficient. The kinetic
model derived by Nasuti et al.
scheme of Dunn and Kang is used.
0
10
γNN 1.0
−1 γOO
10 γNO
θO
γON 0.8 θN
−2 θOH
10
surface coverage
γ coefficients
−3
0.6
10
−4
10 0.4
−5
10 0.2
−6
10
500 1000 1500 2000 0.0
Tw [K] 500 1000 1500 2000
Tw [K]
Figure 37: γ coefficients of the phe-

Figure 38: Different values of θ coefficients
nomenological model derived by Nasuti et
for the same condition as Fig. 37.
al. for the conditions of Fig. 36.
Figure 35 presents the wall heat flux as a function of the wall temperature for different
values of γ and Fig. 36 shows the flux obtained using the phenomenological model derived
by Nasuti et al. As for the high pressure case, we note a strong nonlinear dependence of
9 - 50 RTO-EN-AVT-142
the heat flux with the recombination probability γ. The heat flux for a fully frozen flow
(i.e. no reactions in the gas and at the wall) is also shown. The difference of this flux with
the one denoted γ = 0, puts forward (as already observed for the high pressure case) that
for Pδ = 2000P a in the noncatalytic case, the influence of the chemistry on the heat flux
decreases as the wall temperature increases. Figure 37 shows that γ coefficients of the
different species calculated with the phenomenological model. The coefficients γOO , γN O
and γON have a similar evolution with temperature. They increase up to Tw ' 1500K
and decrease for higher temperatures. For 300K ≤ Tw ≤ 2000K, γN N increases with the
temperature and for Tw ≥ 1700K, this is the highest recombination coefficient. We note
that the γON coefficient is much higher in this condtion at Pδ = 2000P a than for the
previous case at Pδ = 10000P a. Figure 38 shows the surface coverage for O, N atoms
and the OH molecule for the conditions of Fig. 37. The surface coverage by the OH
molecule decreases as the wall temperature increases. The surface coverage by O atoms
is rather high for 300K ≤ Tw ≤ 1500K and then decreases rapidly for higher temperature
due to thermal desorption. The N atom surface coverage remains small in the whole
temperature range but is much higher for Tw ≤ 1000K in this case than in the previous
case at Pδ = 10000P a .
5.0e+05
4.5e+05
−2
4.0e+05
3.5e+05 Dunn and Kang

Gupta
Park
Dunn et Kang − Fick’s law
3.0e+05 frozen flow
500 1000 1500 2000

Tw [K]
Figure 39: Wall heat flux calculated for different gas kinetic
schemes and with the kinetic scheme and Dunn and Kang and the
simple Fick’s law for diffusion fluxes.
Figure 39 compares the wall heat flux calculated with the phenomenological model pro-
posed by Nasuti et al. using different kinetic schemes (Dunn and Kang , Park and Gupta)
and for a frozen flow. The heat fluxes calculated using the kinetic scheme of Dunn and
Kang are very close to the results obtained for the scheme of Gupta for the whole tem-
perature range. The heat fluxes calculated using Park’s kinetic scheme for Tw ≥ 1000K
are slightly higher than those obtained with Dunn and Kang scheme but the discrepancy
remains small. The heat flux calculated assuming a frozen flow is the lower heat flux
limit. We note that the discrepancy with the results obtained assuming a reacting bound-
ary layer is of about 20% for Tw = 300K, 15% for Tw = 1300K and decreases for higher
RTO-EN-AVT-142 9 - 51
temperatures. We note that the influence of the gas chemistry on the calculated heat flux
appears to be more significant for the boundary layer conditions at Pδ = 2000P a than
at Pδ = 10000P a. Therefore, for wall temperatures Tw ' 1500K usually encountered in
catalycity studies, the chemistry of the gas has a non negligible influence at Pδ = 2000P a
on the catalycity coefficients derived for the material. Figure 39 also shows the heat flux
calculated using the kinetic scheme of Dunn and Kang and Fick’s law for diffusion fluxes.
In this case, the influence of the modeling of the diffusion flux appears to be rather small
and the use of Fick’s law slightly overestimates the calculated wall heat flux.
Figure 40 shows the species concentrations for Tw = 1500K and the phenomenological
model proposed by Nasuti et al. We note that the evolution of the different species con-
centrations in the boundary layer is very different than the one obtained for 10000P a.
Figure 41 shows the species concentrations for the same boundary conditions, but a frozen
chemistry. We note that a 15% difference on the wall heat flux (Figure 39) corresponds to
significant discrepancies on the different species evolutions in the boundary layer. It would
be very interesting to compare these results with species concentrations measurements in
the boundary layer.
−1 −1
10 10
N2 N2
−2 −2
10 O2 10 O2
NO NO

−3
−3
N N
−3
0 −3
0
10 10
−4 −4
10 10
−5 −5
10 10
−6 −6
10 10
−7 −7
10 10
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
η η
Figure 40: Species concentrations in the Figure 41: Species concentrations in the
boundary layer for Tw = 1500K using boundary layer for Tw = 1500K using the
the phenomenological model. The kinetic phenomenological model. The boundary
scheme of Dunn and Kang is used. layer is assumed to be frozen.
6 Conclusions
Two main conclusions can be derived from the work developed in the previous sections
on (1) the gas phase chemistry and on (2) the interaction between a gas and a surface
during the re-entry of a spacecraft into the upper layers of the earth’s atmosphere.
First, the elaboration of a relevant and detailed kinetic scheme for a given gas phase is a
difficult task which has to be done carefully, since numerous species and elementary pro-
cesses are involved. For each process a rate coefficient has to be found and it is important
to try to find the most accurate theoretical or experimental data. The chemistry being
9 - 52 RTO-EN-AVT-142
time-dependent by nature, it is important to take into account species that may play a
role in the time evolution even if their concentration at the stationary state is negligible.
Furthermore, it is important to be able to reduce the kinetic scheme after having carefully
compared the results of the simplified scheme with the detailed scheme.
We have used this procedure to derive the three-body recombination and effective ioniza-
tion rate coefficients of atomic oxygen from a detailed collisional radiative model. The
time-dependent approach used in this work has clearly put forward the necessity that
the system be in a quasi-steady state to derive meaningful rate coefficients. In addition,
we have also shown that the value of these rate coefficients depends on the number of
excited levels considered. Then, to elaborate a CR model for air, we have shown that
it is necessary to take into account a sufficient number of excited levels in order to cal-
culate correctly the ionization or recombination as well as all processes occurring in this
type of medium. The results obtained put forward significant discrepancies with common
simplified kinetic schemes used for re-entry studies. The CR scheme developed in this
work can be used to derive up-dated and new simplified kinetic schemes for air for various
applications. For the validation of a simplified kinetic scheme, we recommend careful
comparisons of the time-dependent results obtained with the simplified scheme and with
the CR model.
Second, the stagnation point boundary layer study has shown that the problem of the
determination of the catalycity of the surface interacting with a re-entry plasma is far
from being solved. The surface chemistry plays a crucial role and the net heat flux is
highly dependent on the surface temperature. We have shown that the chemistry of the
gas phase plays a non negligible role in particular for P ≤ 10000P a. For the conditions
studied in section 5 the discrepancies between different kinetic schemes remain small, as
the temperature in the boundary layer remains less than 5400K. However, the time-
dependent CR model has put forward that these discrepancies may be significant when
the temperature in the boundary is high enough to have a non negligible ionization degree.
In this work, we have used a phenomenological model for wall catalycity. We have shown
that this type of model gives very interesting information on the evolution of the catalyc-
ity with the wall temperature. It would be interesting to further improve the modeling
of the gas surface interaction in flow and boundary layer codes. For example it could be
very interesting to take into account the dynamics of the lattice atoms according to the
semi-classical collisional method, which has been developed and successfully applied for
the description of oxygen recombination on silica surfaces (Cacciatore et al. (1999)).
Finally, to validate the results obtained, a comparison with experiments is necessary.
However, it is difficult to measure the concentrations of a large number of species in a
flow simple enough to be a good test-case to validate accurately a nonequilibrium chem-
istry. For catalycity studies, up to now, only a limited number of parameters have been
measured in the stagnation point boundary layer and at the wall. It would be interesting
to carry out further measurements of the evolutions of different species concentrations in
the boundary layer to better validate the results.
RTO-EN-AVT-142 9 - 53
References
Abe, K., Kihara, H., Uchida, T., and Nishida, M. (2002). Experimental and numerical
studies of rotational relaxation behind a strong shock wave in air. Shock Waves, 11:413.
Anderson, J. (1989). Hypersonic and High Temperature Gas Dynamics. McGraw-Hill

Book Company, New York.
Armenise, I., Capitelli, M., Colonna, G., and Gorse, C. (1996). Nonequilibrium vibrational
kinetics in the boundary layer of re-entering bodies. Journal of Thermophysics and Heat
Transfer, 10(3):397.
Barbante, P. (2001). Accurate and efficient modelling of high temperature nonequilibrium

air flows. PhD thesis, VKI, Université libre de Bruxelles, Belgique.
Barbato, M., Belluci, V., and Bruno, C. (1998). Effects of catalytic boundary conditions
accounting for incomplete chemical energy accomodation. AIAA Paper 98-2846.
Bardsley, J. (1983). Dissociative recombination of electrons with no+ ions. Planet and
Space Science, 31(6):667.
Bates, D., Kingston, A., and McWhirter, R. (1962). Recombination beween electrons
and atomic ions i. optically thin cases. Proceedings of the Royal Society of London,
267:297–312.
Berrington, A.K. and Burke, P.G. and Robb, W.D. (1975). The scattering of electrons by
atomic nitrogen. Journal of Physics B: Atomic and Molecular Physics, 8(15):2500–2511.
Bhatia, A. and Kastner, S. (1995). The neutral oxygen spectrum 1: Collisionally excited
level populations and line intensities under optically thin conditions. Astrophysical
Journal Supplement Series, 96(1):325–341.
Biémont, E. and Zeippen, C. (1992). Electric dipole transitions in atomic oxygen and
lifetimes of the 2p3 (4s) 3s 5s and 3s states. Astronomy and Astrophysics, 265:850–856.
Billing, G. and Fisher, E. (1979). Vv and vt rate coefficients in n2 by a quantum-classical

model. Chemical Physics, 43:395.
Bose, D. and Candler, G. (1996). Thermal rate constants of the n2 + o → no + o reaction

using ab initio 3 a” and 3 a0 potential-energy surfaces. Journal of Chemical Physics,
104(8):2825.
Bose, D. and Candler, G. (1997). Thermal rate constants of the o2 + n → no + o reac-

tion based on the 2 a0 and 4 a0 potential-energy surfaces. Journal of Chemical Physics,
107(16):6136.
Bourdon, A., Bultel, A., Desportes, A., and Vervisch, P. (2003). Cnes-nasda-coria collab-
oration on catalycity : γ coefficient determination of c/c samples at coria. Technical
report.
9 - 54 RTO-EN-AVT-142
Bourdon, A., Térésiak, Y., and Vervisch, P. (1998). Ionization and recombination rates
of atomic oxygen in high temperature air plasma flows. Physical Review E, 57(4):4684–
4692.
Bourdon, A. and Vervisch, P. (1996). Three body recombination rate of atomic nitrogen
in low-pressure plasma flows. Physical Review E, 54(2):1888–1898.
Bultel, A., van Ootegem, B., Bourdon, A., and Vervisch, P. (2002). Influence of ar2+ in
an argon collisional-radiative model. Physical Review E, 65:046406.
Cacciatore, M. and Capitelli, M. (1976). The temporal evolution of population densities

of excited states in atomic oxygen thin plasmas. Journal of Quantitative Spectroscopy
and Radiative Transfer, 16(4):325–334.
Cacciatore, M., Rutigliano, M., Billing, G.D. (1999). Elley-Rideal and Langmuir-
Hinshewood recombination coefficients for oxygen on silica surfaces. Journal of Ther-
mophysics and Heat Transfer, 13(2):195–203.
Capitelli, M., Ferreira, C., Gordiets, B., and Osipov, A. (2000a). Plasma Kinetics in
Atmospheric Gases. Springer, Berlin.
Capitelli, M., Gorse, C., and Longo, S. (2000b). Collision integrals of high temperature
air species. Journal of thermophysics and heat transfer, 14(2):259–268.
Carlson, A. and Hassan, H. (1992). Direct simulation of re-entry flows with ionization.
Journal of Thermophysics and Heat Transfer, 6(3):400.
Daiss, A., Fruhauf, H., and Messerschmid, E. (1997). Modeling of catalytic reactions on
silica surfaces with consideration of slip effects. Journal of Thermophysics and Heat
Deutschmann, O., Riedel, U., and Warnatz, J. (1995). Modeling of nitrogen and oxygen
recombination on partial catalytic surfaces. Journal of Heat Transfer, 117:495.
Doering, J. (1992). The atomic oxygen (p-3)-(d-1) electron-excitation cross section near
threshold. Geophysical Research Letters, 19:449.
Doering, J. and Gulcicek, E. (1989a). Absolute differential and integral electron excitation
cross section for atomic oxygen 7. the and transitions from 4.0 to 30 ev. Journal of
Geophysical Research, 94:1541–1546.
Doering, J. and Gulcicek, E. (1989b). Absolute differential and integral electron excitation
cross sections for atomic oxygen. viii - the 3p - 5s0 transition (1356 a) from 13.9 to 30
ev. Journal of Geophysical Research, 94:2733–2736.
Doering, J., Gulcicek, E., and Vaughan, S. (1985). Electron impact measurement of oscil-
lator strengths for dipole-allowed transitions of atomic oxygen. Journal of Geophysical
Research, 90:5279–5284.
Drawin, H. and Emard, F. (1973). Atom-atom excitation and ionization in shock waves
of the noble gases. Physics Letters, 43A(4):333.
RTO-EN-AVT-142 9 - 55
Dunn, M. and Kang, S.-W. (1973). Theoretical and experimental studies of reentry
plasmas. Technical report, NASA CR-2232.
Dunn, M. and Lordi, J. (1969). Measurement of electron temperature and number density
in shock-tunnel flows. part ii: no+ + e− dissociative recombination rate in air. AIAA
Journal, 7(11):2099.
Fuhr, J. and Wiese, W. (1990). CRC Handbook of Chemistry and Physics, pages 128–179.
CRC Press, Boca Raton, FL.
Gomès, A., Essoltani, A., and Bacri, J. (1990). Collisional-radiative modelling of a non-
equilibrium stationary oxygen plasma at atmospheric pressure, te= 3000-18,000 k. Jour-
nal of Quantitative Spectroscopy and Radiative Transfer, 43:471–498.
Gryzinski, M. (1965). Two particle collisions ii. coulomb collisions in the laboratory
system of coordinates. Physical Review A, 138:322–335.
Gryzinski, M. and Kunc, J. (1986). Collisional ionization and the atomic model. Journal
of Physics B: Atomic, molecular and Optical Physcis, 19:2479–2504.
Guberman, S. (2003). Dissociative recombination of molecular ions with electrons. Kluwer,

Dordrecht.
Gulcicek, E. and Doering, J. (1988). Absolute differential and integral electron excitation
cross sections for atomic oxygen. v - revised values for the 3p - 3s0 (1304 a) and 3p -
3d0 (989 a) transitions below 30 ev. Journal of Geophysical Research., 93:5879–5884.
Gulcicek, E., Doering, J., and Vaughan, S. (1988). Absolute differential and integral
electron excitation cross sections for atomic oxygen. vi - the 3p - 3p and 3p - 5p
transitions from 13.87 to 100 ev. Journal of Geophysical Research, 93:5885–5889.
Gunton, R. and Shaw, T. (1965). Electron-ion recombination in nitric oxide in the tem-
perature range 196 to 358 k. Physical Review, 140(3A):A756.
Gupta, R., Yos, J., Thompson, R., and Lee, K. (1990). Calculations and curve fits of
thermodynamic and transport properties for equilibrium air to 30000 k. Technical
report, NASA RP1232.
Hankey, W. (1994). Re-entry aerodynamics. AIAA Education Series, Washington.
Hansen, C. (1968). Temperature dependence of the no+ + e− dissociative recombination

rate coefficient. Physics of Fluids, page 904.
Hellberg, F., Rosen, S., Thomas, R., Neau, A., Larsson, M., Petrignani, A., and van der
Zande, W. (2003). Dissociative recombination of no+ : Dynamics of the x1 σ + and a3 σ +
electronic states. Journal of Chemical Physics, 118(14):6250.
Herzberg, G. (1950). Spectra of Diatomic Molecules. John Wiley & Sons, New York.
Hindmarsh, A. (1980). Lsode and lsodi, two new initial value ordinary differential equation
solvers. ACM SIGNUM Newsletter, 15:10.
9 - 56 RTO-EN-AVT-142
Holstein, T. (1947). Imprisonment of resonance radiation in gases. Physical Review,

72(12):1212.
Huang, C.-M., Biondi, M., and Johnsen, R. (1975). Variation of electron-no+ recombina-
tion coefficient with electron temperature. Physical Review A, 11(3):901.
Itikawa, Y. and Ichimura, A. (1990). Cross sections for collisions of electrons and photons
with atomic oxygen. Journal of Physical and Chemical Reference Data, 19:637.
Jumper, E. and Seward, W. (1994). Model for oxygen recombination on reaction-cured

glass. Journal of Thermophysics and Heat Transfer, 8(3):460.
Kim, Y.-K. and Desclaux, J.-P. (2002). Ionization of carbon, nitrogen and oxygen by
electron impact. Physical Review A, 66:012708.
Kley, D., Lawrence, G., and Stone, E. (1977). The yield of n(2 d) atoms in the dissociative
recombination of no+ . Journal of Chemical Physics, 66(9):4157.
Kolesnikov, A. (1999). Combined measurements and computations of high enthalpy and

plasma flows for determination of tpm surface catalycity. In Measurement techniques for
high enthalpy plasma flows, Belgique. von Karman Institute for fluid dynamics, RTO
EN 8, NATO.
Kolodziej, P. and Stewart, D. (1987). Nitrogen recombination on high-temperature

reusable surface insulation and the analysis of its effects on surface catalysis. AIAA
Paper 87-1637.
Kossyi, I., Kostinsky, A., Matveyev, A., and Silakov, V. (1992). Kinetic scheme of the
non-equilibrium discharge in nitrogen-oxygen mixtures. Plasma Sources Science and
Technology, 1:207.
Kovalev, V. and Suslov, O. (1996). Simulation of the interaction between partially ionized
air and the catalytic surface of high temperature reusable thermal insulation. Fluid
Dynamics, 31(5):775.
Kunc, J. and Soon, W. (1989). Collisional radiative nonequilibrium in partially ionzed

atomic nitrogen. Physical Review A, 40(10):5822–5842.
Laher, R. and Gilmore, F. (1990). Updated excitation and ionization cross sections for
electron impact on atomic oxygen. J. Phys. Chem. Ref. Data, 19(1):277–305.
Laux, C. and Kruger, C. (1992). Arrays of radiative transition probabilities for the n2
first and second positive, no beta and gamma, n+
2 first negative and o2 schumann-runge
band systems. Journal of Quantitative Spectroscopy and Radiative Transfer, 48(1):9.
Losev, S., Makarov, N., Pogosbekyan, M., and Shatalov, O. (1994). Aiaa 94-1990. In Pro-
ceedings of the American Institute of Aeronautics and Astronautics, 6th AIAA/ASME
Joint Thermophysics and Heat Transfer Conference, Washington,DC. AIAA 94-1990,
American Institute of Aeronautics and Astronautics.
RTO-EN-AVT-142 9 - 57
Losev, S. and Shatalov, O. (1990). Elaboration of computer data bank for physicochemical
gasdynamics. Soviet Journal of Chemical Physics, 6(12):3299.
Lotz, W. (1968). Electron-impact ionization cross sections and ionization rate coefficients
for atoms and ions from hydrogen to calcium. Zeitschrift fur Physik, 216(3):241.
Moore, C. (1993). Tables of Spectra of Hydrogen, Carbon, Nitrogen, and Oxygen Atoms
and Ions, pages 203–209. CRC Press, Boca Raton, FL.
Motapon, O., Fifirig, M., Florescu, A., Waffeu-Tamo, F.-O., Crumeyrolle, O., Varin-
Bréant, G., Bultel, A., Vervisch, P., Tennyson, J., and Schneider, I. (2006). Reactive
collisions between electrons and no+ ions: rate coefficient computations and relevance
for the air plasma kinetics. Plasma Sources Science and Technology, 23(15):23–32.
Nasuti, F., Barbato, M., and Bruno, C. (1996). Material-dependent catalytic recombi-
nation modeling for hypersonic flows. Journal of Thermophysics and Heat Transfer,
10(1):131–136.
Nussbaumer, H. and Storey, P. (1983). Dielectronic recombination at low temperatures.

Astronomy and Astrophysics, 126:75–79.
Park, C. (1969). Collisional ionization and recombination rates of atomic nitrogen. AIAA
Journal, 7(8):1653–1654.
Park, C. (1971). Electron-impact excitation rate coefficients for hydrogen, helium and
alkali atoms. Journal of Quantitative Spectroscopy and Radiative Transfer, 11:7–36.
Park, C. (1985). Problems of rate chemistry in the flight regime of aeroassisted orbital
transfer vehicles, page 511. AIAA, New York.
Park, C. (1990). Non-Equilibrium Hypersonic Aerothermodynamics. Wiley, New York.
Park, C. (1993). Review of chemical-kinetic problems of future nasa missions, i: Earth

entries. Journal of Thermophysics and Heat Transfer, 7(3):385.
Peterson, J., Le Padellec, A., Danared, H., Dunn, G., Larsson, M. Larson, A., Peverall,
R., Strömholm, C., Rosen, S., af Ugglas M., and van der Zande, W. (1998). Dissocia-
tive recombination and excitation of n+
2 : Cross sections and product branching ratios.
Journal of Chemical Physics, 108(5):1978.
Peverall, R., Rosen, S., Peterson, J., Larsson, M., Al-Khalili, A., Vikor, L., Semaniakand,
J., Bobbenkamp, R., Le Padellec, A., Maurellis, A., and van der Zande, W. (2001).
Dissociative recombination and excitation of o+ 2 : Cross sections, product yields and im-
plications for studies of ionospheric aifglows. Journal of Chemical Physics, 114(15):6679.
Pradhan, A. (1988). Atomic Processes in Plasmas. Academic, New York.
Sarrette, J.-P., Gomes, A.-M., Bacri, J., Laux, C., and Kruger, C. (1995). Collisional-
radiative modelling of quasi-thermal air plasmas with electronic temperatures between
2000 and 13000 k -i. θe > 4000 k. Journal of Quantitative Spectroscopy and Radiative
9 - 58 RTO-EN-AVT-142
Scott, C. (1990). Catalytic recombination of nitrogen and oxygen on high temperature

reusable surface insulation. In Aerothermodynamics and planetary entry, New York.
AIAA, Progress in Astronautics and Aeronautics.
Smith, A., Caplinger, E., Neynaber, R., Rothe, E., and Trujillo, S. (1962). Electron
impact ionization of atomic nitrogen. Physical Review, 127:1647–1649.
Soon, W. and Kunc, J. (1990). Thermal nonequilibrium in partially ionized atomic oxygen.
Physical Review A, 41(2):825–842.
Soon, W. and Kunc, J. (1991). Kinetics and continuum emission of negative atomic ions
in partially ionized plasmas. Physical Review A, 43(2):723–734.
Sun, H. and Nakamura, H. (1990). Theoretical study of the dissociative recombination of

no+ with slow electrons. Journal of Chemical Physics, 93(9):6491.
Suslov, O. and Tirskyi, G. (1994). the kinetics of the recombination of nitrogen atoms
on high temperature reusable surface insulation in hypersonic thermochemical non-
equilibrium flows. In 2nd Europ. symp. on Aerothermodynamics for Space Vehicles,
Noordwijk, The Netherlands. ESTEC, ESA.
Sutton, K. and Gnoffo, P. (1998). Multi-component diffusion with application to com-

putational aerothermodynamics. In 7th AIAA/ASME Joint Thermophyscis and Heat
transfer conference. AIAA Paper 98-2575.
Tawara, H. and Kato, M. (1999). Electron impact data for atoms and ions: up-dated in
1998. NIFS Data Serie S, 51:256P.
Tayal, S. and Henry, R. (1989). Oscillator strengths and electron collisional excitation
cross sections for atomic oxygen. Physical Review A, 39(9):4531–4536.
Taylor, R. and Ali, A. (1986). Saha decrements and collisional-radiative recombination

and ionization coefficients for a nonequilibrium oxygen plasma. Journal of Quantitative
Spectroscopy and Radiative Transfer, 36:373–383.
Teulet, P., Sarrette, J.-P., and Gomes, A.-M. (1999). Calculation of electron impact inelas-
tic cross sections and rate coefficients for diatomic molecules. Journal of Quantitative
Spectroscopy and Radiative Transfer, 62:549.
Teulet, P., Sarrette, J.-P., and Gomes, A.-M. (2001). Collisional-radiative modelling of
one- and two- temperature air and air-sodium plasmas at atmospheric pressure with
temperatures of 2000 − 12000 k. Journal of Quantitative Spectroscopy and Radiative
Transfer, 70:159.
Vejby-Christensen, L., Kella, D., Pedersen, H., and Andersen, L. (1998). Dissociative
recombination of no+ . Physical Review A, 57(5):3627.
Verner, D., Verner, E., and Ferland, G. (1996). Atomic data for permitted resonance lines
of atoms and ions from h to si, and s, ar, ca and fe. Atomic Data Nucl. Data Tables,
64:1.
RTO-EN-AVT-142 9 - 59
Weller, C. and Biondi, M. (1968). Recombination, attachment and ambipolar diffusion of

electrons in photo-ionized no afterglows. Physical Review, 172:198.
Wiley, R. (1993). Comparison of kinetic models for atom recombination on high tem-
perature reusable surface insulation. Journal of Thermophysics and Heat Transfer,
7(1):55.
9 - 60 RTO-EN-AVT-142
THEORY AND COMPUTING OF GAS PHASE
CHEMICAL REACTIONS: FROM EXACT
QUANTUM TO APPROXIMATE DYNAMICAL
TREATMENTS
A. Laganà
∗
Dipartimento di Chimica, Università di Perugia,

06123 Perugia, Italy
8 February 2006
Contents
1 Abstract 3
2 Introduction 3
3 The a priori dynamical approach 4

3.1 Electronically adiabatic potential energy surfaces . . . . . . . . . . . . . . 5
3.2 Quantum formalism for few body reactions . . . . . . . . . . . . . . . . . . 7
3.2.1 Time independent approaches . . . . . . . . . . . . . . . . . . . . . 7
3.2.2 Time dependent approaches . . . . . . . . . . . . . . . . . . . . . . 9
3.2.3 Direct calculations of Reactions rate coefficients . . . . . . . . . . . 10
4 Approximate Methods 12
4.1 Reduced dimensionality quantum methods . . . . . . . . . . . . . . . . . . 12
4.2 Mixed quantum and classical mechanics approaches . . . . . . . . . . . . . 13
4.3 The pure classical mechanics computational machinery . . . . . . . . . . . 13
∗
Interesting discussions with the members of the Quantum Reactive Scattering (QRS) group of interest
is acknowledged. The work illustrated here has been financially supported by MIUR, ASI, CNR and
COST Chemistry (Action D23). Specific mention needs to be made to Italian Space Agency (Project
ASI PQE2000) and the Italian MIUR FIRB Grid.it project (RBNE01KNFP) on High performance Grid
Platforms and Tools, and to the MIUR CNR Strategic Project L 499/97-2000 on High performance
Distributed Enabling Platforms.
Laganà, A. (2007) Theory and Computing of Gas Phase Chemical Reactions: From Exact Quantum to Approximate Dynamical Treatments.
In Experiment, Modeling and Simulation of Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 10A-1 – 10A-24).
Educational Notes RTO-EN-AVT-142, Paper 10A. Neuilly-sur-Seine, France: RTO. Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 10A - 1
Theory and Computing of Gas Phase Chemical Reactions:
From Exact Quantum to Approximate Dynamical Treatments
5 Advanced Computing 14
5.1 Parallel computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.2 Virtual organizations and Grid enabled applications . . . . . . . . . . . . . 16
5.3 The Grid implementation of a molecular science simulator . . . . . . . . . 17
10A - 2 RTO-EN-AVT-142
1 Abstract
A realistic simulator of complex gas phase systems needs to handle accurately processes
spanning different scales of space and time. The paper outlines the key theoretical and
computational features of modern a priori treatments of the dynamics of elementary
processes in these systems with particular emphasis on reactive processes.
2 Introduction
Realistic a priori simulations of rarefied gas flows are conceptually articulated into various
blocks of operations characterized by different time and space scales (1). These blocks
are concerned with the calculation of electronic structures, molecular collisions, collec-
tive (fluid-, electro-, etc.) dynamics and the averaging over randomly sampled variables
to work out a priori estimates of observed and measurable properties of real situations
(like the formation of shock waves in rarefied gas flows). As a matter of fact, these are
the pillars on which the COST Chemistry Metalaboratory SIMBEX has developed the
homonymous simulator of molecular beam experiments (1; 2) and the COMPCHEM (3)
virtual organization (VO) 1 has assembled the Grid Enabled Molecular Simulator (GEMS)
proposed for European funding as a specific targeted research project within the Activity
IST-2005-2.5.6 Research Network Testbed of the 6th Framework Program (4).
The mentioned blocks of computations are carried out using separate suites of codes
thanks to the fact that time and space scales associated with them are profoundly dif-
ferent. This means that the electronic energy of the molecular systems and the related
wavefunctions can be calculated to a high level of accuracy from first principles using ab
initio techniques within the Born-Oppenheimer (BO) (5) approximation (provided that,
when necessary, at certain times of the process or at certain arrangements of the nuclei in
the corresponding stationary scheme, the coupling between nuclear and electronic degrees
of freedom is regained). In the BO approximation the electronic motion is assumed to
depend only parametrically on the nuclear coordinates and the potential energy (elec-
tronic energy plus nuclear repulsions) of the nuclear motion is calculated (at several fixed
geometries of the nuclei) using well consolidated quantum chemistry suites of programs
(see for example refs. (6; 7; 8; 9; 10)) which will not be discussed here. The ab initio
calculation of the electronic energy can be performed either time by time at the actual
geometry of the moving system (on-the-fly) or, once for ever, at the first step of the com-
putational procedure. In the latter case, ab initio calculations are performed for a large
matrix of nuclear geometries (including those of the initial and final states of the consid-
ered process) and calculated values are then best-fitted by optimizing the parameters of
a suitable potential energy functional.
The integration of the motion equations of the nuclei on the adopted potential energy
surface (PES) allows to estimate the efficiency parameters of the elementary chemical
processes being considered. These detailed efficiency parameters (with their temperature,
energy, spatial, angular momentum, etc. dependencies) are the key set of input data
needed to carry out realistic simulations of the kinetic regime of real gaseous systems.
In this paper we shall discus on:
1
COMPCHEM is registered at grid-it.cnaf.infn.it.
1. the development of few atom rigorous dynamical methods (section 2);

2. the extension of dynamical methods to larger systems (section 3);
3. the exploitation of the potentialities of advanced computing for Molecular Sciences
calculations (section 4).
3 The a priori dynamical approach

From a theoretical point of view elementary chemical processes are many body problems
concerning the encounter of two or more aggregates of electrons and nuclei. This allows to
treat the molecules as deformable objects which can collide and break into parts and/or
recombine. Simultaneous encounters of three or more objects are quite unlikely to oc-
cur (especially in the low pressure gas phase processes mentioned before). Accordingly,
most of the theoretical and computational studies have focused on two body encounters
(bi-molecular). After all, the treatment of multi-molecular encounters are usually ratio-
nalized by chaining relevant bi-molecular collisions (though, recently, they are increasingly
dealt directly using classical molecular dynamics). For uni-molecular processes too the
theoretical approach can be easily reconducted to that of half a bi-molecular encounter.
Typical efficiency parameters of elementary bi-molecular processes are either (thermal,
state-specific and state-to-state) rate coefficients or integral (state-specific and state-to-
state) cross sections. The thermal rate coefficient k(T ) is formulated as (11)
X X wi exp [−ǫi /kB T ]
k(T ) = ki,f (T ) (1)
i f
Qint (T )
where i and f are respectively the initial and final internal states, wi is the multi-
plicity
P of the initial internal state i including the nuclear spin symmetry, Qint (T ) =
i wi exp [−ǫi /kB T ] is the partition function associated with the initial internal states,
kB is the Boltzmann constant, ǫi is the energy of state i, ki,f (T ) is the state-to-state rate
coefficient and T is the temperature of the system. The evaluation of the state-to-state
rate coefficient can be reconducted to the calculation of the state-to-state cross section
σi,f (Etr ) using the following equation
1/2 Z ∞
8
ki,f (T ) = 3 3
Etr σi,f (Etr )e−Etr /kB T dEtr (2)
πµkB T 0
if the energy distribution is of the Boltzmann type with µ being the reduced mass of the
system in its reactant arrangement and Etr the translational energy. The state-to-state
cross section (σi,f (Etr )) can be calculated from the state-to-state cumulative reaction
probability (Pi,f (Etr )) as follows
∞ ∞ J
π π X J π X X
JΛ
σi,f (Etr ) = 2 Pi,f (Etr ) = 2 (2J + 1)Pi,f (Etr ) = 2 (2J + 1) Pi,f (Etr ) (3)
ki ki J=0 ki J=0 Λ=−J
JΛ
where ki is the system wavenumber and the individual terms Pi,f (Etr ) of the right hand
side member of expression 3 can be derived from the square modulus of the S matrix
elements calculated at a given pair of total angular momentum quantum number J and
helicity quantum number Λ.
3.1 Electronically adiabatic potential energy surfaces

As already mentioned, a priori calculations of state-specific and state-to-state probabil-
ities of elementary chemical processes start from the consideration that the fixed nuclei
electronic wavefunctions Φe ({w}; {W}) are a suitable basis set for the expansion of the
system wavefunction Z ({w}, {W}, t) (in our notation {w} and {W} are the sets of elec-
tronic and nuclear position vectors, respectively, and t is time). The fixed nuclei electronic
wavefunctions Φe ({w}; {W}) are the eigensolutions of the fixed nuclei electronic problem
Ĥe ({w}; {W}) Φe ({w}; {W}) = E({W})Φe ({w}; {W}) (4)
where Ĥe is the time independent electronic Hamiltonian. The solution of eq. 4 provides
us with the desired set of Ith adiabatic electronic eigenvalues EI ({W}) for the molecular
geometry and electronic wavefunction associated with it. The ensemble of the EI ({W})
values, once summed to the corresponding nuclear repulsion, represent a pointwise de-
scription of the PES (VI ({W}) on which the motion of the Nnucl nuclei of the system
takes place.
This gives us a means for solving the general time dependent Schrödinger equation of
the system (12; 13)
∂
i~ Z ({w}, {W}, t) = Ĥ({w}, {W})Z ({w}, {W}, t) (5)
∂t
(in eq. 5 Ĥ = Ĥe + ĤN is the total many body Hamiltonian of the system whith ĤN being
its nuclear component) by expanding Z, as already mentioned, in terms of the electronic
eigenfunctions which parametrically depend on the position vectors of the nuclei as follows
X
Z ({w}, {W}, t) = ΨI ({W}, t) ΦeI ({w}; {W}) . (6)
I
After averaging over the electronic coordinates, the resulting differential equations
for the ΨI ({W}, t) coefficients of the expansion contain some terms coupling nuclear
and electronic degrees of freedom. As already mentioned, the BO decoupling scheme is
usually applied at this point by assuming these terms to be negligible. As a result, for
each electronic state I (hereafter, the index I will be dropped from the formalism because
we shall confine our attention to the single electronically adiabiatic PES BO regime) the
calculations reduce to the following electronically adiabatic time dependent Schrödinger
equation (12; 13; 14)
∂ h i
i~ Ψ ({W}, t) = ĤN ({W})Ψ ({W}, t) = T̂N ({W}) + V ({W}) Ψ ({W}, t) (7)
∂t
where T̂N ({W}) is the nuclear kinetic operator.
In regions where the BO approximation breaks down, as is the case of closely spaced
electronic eigenvalues, ad hoc treatments can be adopted which deal at the same time
with different terms of the electronic functions manifolds (15). In these approaches the
elements coupling electronic and nuclear degrees of freedom are evaluated consistently us-
ing information coming from electronic structure calculations. Then resulting differential
equations coupling different electronic states and wavefunctions are integrated using the
same standard numerical algorithms available to integrate equations 7.
The crucial starting phase of the calculation of cross sections and rate coefficients is
therefore the assemblage of a suitable PES. This task is most often highly demanding in
terms of computing time and chemical ingenuity. For this reason the calculation of the
potential energy values is increasingly tackled at an on-the-fly ab initio level. Yet, up to
date, the most popular approaches are those based on a two step procedure. The first
step consists of the collection of all the available (both ab initio and empirical) local or
global data on the system interaction while the second step consists of their fitting using
an appropriate functional form.
For small systems (Nnucl < 10), as is the case of the majority of the elementary reac-
tions considered up to date for theoretical dynamical studies (in some cases this is true
also for larger systems if the complexity of the computational procedure is reduced by
imposing suitable dynamical constraints), the above mentioned two step procedure is,
indeed, the preferred one. The reason for this preference is the fact that the quality of
present electronic structure calculations is often insufficient to guarantee an accurate re-
production of the interaction of the system over the full range of internuclear distances
unless calculated ab initio values are adjusted using empirical considerations before un-
dertaking the fitting. The functional representation to be used for fitting the PES of
reactive systems is more difficult to formulate and most of the computations have been
confined to systems made of three and four atoms (16; 17; 18; 19; 20).
The most popular functional forms used for this purpose are polynomials either in
physical coordinates (17) (like the internuclear distances defined as rij = |Wi − Wj |) or
in bond order variables (21) (defined as exponentials of the displacement from equilibrium
of the related internuclear distances). When using physical coordinates the polynomials
need to be damped to avoid divergence at long range. Polynomial functionals are usually
adopted within a Many Body Expansion approach to formulate the individual compo-
nents of the expansion. Other simple functional forms are either derived from drastically
simplified formulations of ab initio methods (22; 23) or from intuitive models (such as
diatomic rotating potentials (24; 25)). Alternative approaches make use of local inter-
polation methods in which for each interval low order polynomials are employed and the
value of related parameters are determined by imposing pointwise or switching continuity
conditions. Similar approaches are also used for multi-surface treatments by fitting each
surface using a functional form (except for methods directly providing multiple solutions).
When moving to complex systems it becomes more convenient to compose the PES
by summing simple few body functions (stretches, bends, torsions, van der Waals, non-
bonded interactions, etc) containing empirically determined parameters (force fields) (26).
Parameters used by these approaches are transferable within the same family of systems.
Usually these surfaces are scarcely suited to describe the making or breaking of bonds
while they are better suited for conformational studies.
3.2 Quantum formalism for few body reactions

Quantum techniques based on the integration of eq. (7) are widely used to evaluate the
observable properties of reactive elementary processes. This task is, nowadays, a largely
routine work when dealing with atom diatom systems. From the integration of eq. (7)
one can evaluate the elements of the scattering matrix S whose square moduli are the
elements of the probability matrix P used in eq. 3.
For a few body isolated system (as is the case of laser free crossed molecular beam
experiments) the motion of the center of mass of the system can be factored out to reduce
the dimensionality of the problem (to six spatial coordinates in the atom diatom case
that is the simplest prototype of elementary reactions) without introducing additional
approximations.
Accordingly, for a three atom (A, B, C) system, the time dependent Schrödinger
equation for the nuclei reads as
2
∂ ~ 2 2
i~ Ψ(Rτ , rτ , t) = − (∇Rτ + ∇rτ ) + V (Rτ , rτ , Θτ ) Ψ(Rτ , rτ , t) (8)
∂t 2µ
where Rτ and rτ are the mass scaled atom-diatom Jacobi vectors (of modulus Rτ and
rτ ), Θτ is the angle formed by Rτ and rτ and Ψ(Rτ , rτ , t) is the time dependent nuclear
wavefunction. As usual, Jacobi coordinates are labeled after the arrangement τ (τ = 1, 2
and 3 means A + BC, B + CA and C + BA respectively) to which they refer.
The dimensionality of the problem is further reduced if the Laplacian (the kinetic
component of the Hamiltonian) is written in terms of angular momentum operators
~2 2 ~2 1 ∂2 1 ∂2 (Jˆ − ĵτ )2 ĵτ2

2
− (∇Rτ + ∇rτ ) = − Rτ + rτ + + (9)
2µ 2µ Rτ ∂Rτ2 rτ ∂rτ2 2µRτ2 2µrτ2
with Jˆ being the total angular momentum operator given by the sum of ĵτ and ˆlτ (the
rotational and the orbital angular momentum operators of the system, respectively). This
makes it convenient to express the Ψ(Rτ , rτ , t) wavefunction in terms of products of the
ΨJM p (Rτ , rτ , Θτ , t) partial waves (which are eigenfunctions of the eigenvalue J(J +1) of the
total angular momentum operator Jˆ2 , of its projection M on a space fixed (SF) reference
axis and of the parity p) in the internal coordinates Rτ , rτ and Θτ , and the appropriate
spherical harmonics. For computational convenience to integrate the scattering equations
from the reactant to the product arrangement one can also adopt a body fixed (BF)
representation (in which the reference frame is allowed to rotate in order to have the z
axis always aligned with the Rτ vector and the xz plane having a fixed orientation with
respect to the molecular plane) where Λ is the projection of the total angular momentum
Jˆ on the BF z axis.
3.2.1 Time independent approaches

To further reduce the dimensionality of the problem (this was the approach usually
adopted in the past) the time variable can be factored out from the system wavefunc-
tion and a time independent formulation of the Schrödinger equation can be obtained
without introducing additional approximations (this means that the system can be de-
scribed by a stationary wave) (27). To integrate the stationary Schrödinger equation one
needs to define a particular coordinate (usually called reaction coordinate) by properly

combining the internuclear distances or Jacobi coordinates to ensure a smooth switch
from reactant oriented to product oriented arrangement coordinates.
The most popular reaction coordinate adopted in recent studies has been the hyper-
radius ρ defined as ρ2 = Rτ2 + rτ2 (this relationship holds for all arrangements because ρ
is invariant under kinematic rotations). The hyperradius together with two hyperangles
constitutes the set of hyperspherical coordinates. These coordinates can be defined in
various ways depending on what arrangement is to be preferred. Here, for illustrative
purposes, we make use of their APH (28) (democratic) version in which the hyperangles
are θ and χ (the value of χ depends on the choice of the reference geometry though
for simplicity the related label is dropped here from the notation) and the partial wave
equations take the form
h i
T̂ρ + T̂h + T̂r + T̂c + V (ρ, θ, χ) ΨJM p (ρ, θ, χ) = EΨJM p (ρ, θ, χ) (10)
where subscripts “h”, “r” and “c” stand for “hypersphere”, “rotational,” and “Coriolis”,
respectively and the operators T̂ρ , T̂h , T̂r and T̂c are formulated as:
~2 ∂ 5 ∂
T̂ρ = − ρ ,
2µρ5 ∂ρ ∂ρ
~2 1 ∂2

4 ∂ ∂
T̂h = − sin 2θ + ,
2µρ2 sin 2θ ∂θ ∂θ sin2 θ ∂χ2
T̂r = A(ρ, θ)Jx2 + B(ρ, θ)Jy2 + C(ρ, θ)Jz2 ,
and
i~ cos θ ∂
T̂c = − 2 Jy ,
µρ sin θ ∂χ
2
with A(ρ, θ), B(ρ, θ) and C(ρ, θ) being defined as A−1 (ρ, θ) = µρ2 (1 + sin θ), B −1 (ρ, θ) =
2µρ2 sin2 θ, C −1 (ρ, θ) = µρ2 (1 − sin θ).
Eq. (10) is integrated by segmenting the hyperradius in several sectors and expanding
Ψ (within each sector i) in terms of the ΦJp tΛ surface functions which are eigensolutions of
the following equation
15~2

T̂h + + (C − D) ~ Λ + D~ J(J + 1) + V (ρi , θ, χ) − εtΛ (ρi ) ΦJp
2 2 2 Jp
tΛ (θ, χ; ρi ) = 0
8µρ2i
(11)
with D = (A + B)/2 (though the Φ functions could be chosen also to be independent of
J by setting J = 0 in eq. 11 and regaining the J dependence during the integration over
ρ). Once the expansion is performed one gets the following set of equations to integrate
over the hyperradius ρ
2
∂ 2µE Jpn 2µ X
+ ψtΛ (ρi ) = < ΦJp Jp Jp Jp Jpn
tΛ (θ, χ; ρi )D̂ΛM |Ĥi |Φt′ Λ′ (θ, χ; ρi )D̂Λ′ M ′ > ψtΛ (ρi )
∂ρ2 ~2 ~2 ′ ′
tΛ
(12)
where the internal Hamiltonian Ĥi is defined as
15~2
Ĥi = T̂h + T̂r + T̂c + + V (ρ, θ, χ). (13)
8µρ2
Accordingly, the computational procedure can be partitioned into three parts. Part
one is devoted to the calculation of the Φ functions over suitable fixed ρ grid points of θ and
χ values. This part is dominated by the evaluation of (two dimensional for three atom, five
dimensional for four atom systems) integral quadratures and eigenvalues finding of large
real dense square matrices. The second part of the calculation consists in propagating
Jpn
the ψtΛ (ρi ) solutions. This part of the calculation is dominated by the inversion of large
matrices having the same dimension as the number of channels. The third part is less
computationally demanding and is devoted to the mapping of the asymptotic solution
into the proper arrangement space and to the imposition of boundary conditions to work
out the S matrix.
Moving from three to four atom systems (say diatom-diatom) the number of internal
coordinates doubles. A possible choice is the set of coordinates R, r1 , r2 , Θ1 , Θ2 and φ
with R being the distance between the centers of mass of the two diatoms, r1 and r2 being
the two diatomic internuclear distances, Θ1 and Θ2 being the two planar angles formed
by r1 and r2 with R and φ being the dihedral angle formed by the (R, r1 ) and (R, r2 )
planes. Using these coordinates the Hamiltonian of the AB + CD system takes the form
(29)
~2 ∂ 2 (Jˆ − ĵ12 )2 ĵ12 ĵ22

ĤN = − + ĥ (r
1 1 ) + ĥ (r
2 2 ) + + + + ∆V (14)
2µ4 ∂R2 2µ4R2 2µ1 r12 2µ2 r22
where µ4 is the reduced mass of the AB and CD reduced masses (µ1 and µ2 , respectively),
Jˆ is, as already mentioned, the total angular momentum operator, ĵ12 is the sum of ĵ1
and ĵ2 which are the rotational angular momentum operators of AB and CD, respectively,
ĥ1 (r1 ) and ĥ2 (r2 ) are the vibrational Hamiltonians of AB and CD, respectively, while ∆
V is the difference between the total interaction potential V (R, r1 , r2 , Θ1 , Θ2 , φ) and the
V (r1 ) and V (r2 ) vibrational potentials of ĥ1 (r1 ) and ĥ2 (r2 ).
The increased complexity of the four atom Hamiltonian makes the definition of the
reaction coordinate, the calculation of the sector basis functins and the switch from one
arrangement to another (and therefore the solution of the time independent Schrödinger
equation) very difficult. As a matter of fact, only recently significant advances have been
made in describing the reaction coordinate of four atom systems by using row-orthonormal
hyperspherical democratic coordinates made of a hyperradius and five hyperangles (30).
3.2.2 Time dependent approaches

The difficulty of handling the problem associated with the definition of a suitable smoothly
evolving spatial continuity variable brings the discussion back to the decision of reduc-
ing the dimensionality of the calculation by factoring out the time dependence of the
wavefunction. In this respect the simplicity obtained in formulating the computational
machinery when using the Jacobi coordinate time dependent formalism is increasingly
considered a suitable reward for keeping the extra variable of time in the formalism. As a
matter of fact, in time dependent approaches one has the advantage of straightforwardly
shaping the initial wavepacket. This is, in fact, chosen to correspond to a given reac-
tant state or to a mixture of them, its component along the atom diatom coordinate R
is formulated as a product of a gaussian wave (exp [−α(R − Ro )2 ]) times a phase shift
factor (exp [−iko (R − Ro )]) associated with its inward traveling nature having an average
momentum ko ~. Then the wavepacket can be mapped into any other set of arrangement
coordinates ofh interestiand let propagate in time by repeatedly applying the evolution
operator exp −iĤτ /~ . Eventually, after the wavepacket has spread all over the whole
accessible configuration space one can carry out its analysis at the product asymptotic
line by expanding the cut of the wavefunction into the related diatomic wavefunctions
(for atom-diatom systems). The time dependent coefficients of the expansion Cv′ j ′ Λ′ (t)
read (31)
Z Z
Cv′ j ′ Λ′ (t) = sin ΘPjΛ (Θ)φv′ j ′ (r)ΨJΛ (R = R∞ , r, Θ, t)dΘdr (15)
r Θ
when using the reactant Jacobi coordinates R, r and Θ (primed variables are for products,
unprimed for reactants). In eq. (15) PjΛ (Θ) is the normalized associated Legendre function
of the angular part of the wavefunction. By performing a half Fourier transform of the
Cv′ j ′ Λ′ (t) coefficients one gets the time independent (energy dependent) A matrix whose
elements read Z ∞
1
Av′ j ′Λ′ (E) = exp (iEt/~) · Cv′ j ′ Λ′ (t)dt. (16)
2π t=0
From them one can easily determine the S matrix elements whose square modulus (the
probabilities) allow to calculate the atom-diatom cross section (31)
π X X X
S J ′ ′ ′ 2 .

σvj,v′ j ′ (Etr ) = 2
(2J + 1) vjΛ,v j Λ (17)
kvj (2j + 1) J p=±1 ′ Λ,Λ
3.2.3 Direct calculations of Reactions rate coefficients

The simplicity of the time dependent method has also facilitated its extension to larger
systems (29; 32). However, when one is interested in evaluating the (less detailed) thermal
rate coefficient there is no need to carry out in the computation all the details embodied
in the S-matrix. After all, the value of the rate coefficient is rather insensitive to the
detailed structure of the whole PES while it is strongly dependent on the shape of the
reactant side of the saddle and in particular on the height and the width of the reaction
barrier (33).
This has motivated the formulation of the rate coefficient directly in terms of the
cumulative reaction probability N(E) defined as
XX
N(E) = Pi,f (E). (18)
i f
and that does not refer to any asymptotic state and depends only on the dynamics of the
system in the vicinity of the reaction barrier. In terms of N(E) the rate coefficient is then
formulated as Z ∞
1
k(T ) = N(E)e−E/kB T dE (19)
hQtr (T )Qint (T ) 0
where Qtr (T ) is equal to (2πµkB T )3/2 /h3 . This means that the efficiency of the reactive
process is expressed in terms of the fraction of the system wavefunction left over after
10A - 10 RTO-EN-AVT-142
projecting out all its components which do not have outgoing character in the product
asymptotic region. The projection operator can be given the form
Pprod = lim eiĤt θ(R − R0 ) e−iĤt . (20)
t→∞
Accordingly, the wavefunction can be propagated forward (out of an arbitrarily located

dividing surface) infinitely in time. Then components not located on the product side of
the dividing surface are projected out and the remaining wavefunction is propagated
backward in time. In this approach, the rate constant k(T ) can be formulated as:
1
k(T ) = lim tr(F̂ e−Ĥ/kB T eiĤt ĥ e−iĤt ) (21)
Qtr (T )Qint (T ) t→∞
when such a limit exists. In equation 21, ĥ can be chosen to be any operator discriminating
between reactants and products and F̂ any operator measuring the flux from reactants
to products (in Eq. (21) the following correlation function
Ĥ Ĥ
Cf p (t) = tr(e− 2kT F̂ e− 2kT eiĤt ĥ e−iĤt ) (22)
of the flux-position type is used). As an alternative use can be made of the following
correlation function
Ĥ Ĥ
Cf f (t) = tr(e− 2kT F̂ e− 2kT eiĤt F̂ e−iĤt ) (23)
of the flux-flux type with Cf f (t) being the time derivative of Cf p (t) and F̂ the time
derivative of ĥ in the Heisenberg picture. Accordingly, the thermal rate coefficient takes
the form
Z∞
1 1
k(T ) = lim Cf p (t) = Cf f (t)dt (24)
Qtr (T )Qint (T ) t→∞ Qtr (T )Qint (T )
0
and the cumulative reaction probability becomes (34; 35):

N(E) = 2π 2 tr F̂ δ(Ĥ − E) F̂ δ(Ĥ − E)
Z 2
1 2E/kB T X X
′ iEt

−iĤt ′
= e fT fT e < fT e fT > dt (25)
2 f ′
T fT
where the evaluation of the trace (whose detailed calculation would imply the propagation
of the whole set of basis functions) has been obtained in terms of the eigenstates fT of
the thermal flux operator F̂T defined as
F̂T = e−Ĥ/2kB T F̂ e−Ĥ/2kB T (26)
which need only a small number of applications of F̂T on a sample wavefunction.
An interesting analogy between the classical and the quantum description of the rate
coefficient can be obtained if one factorizes the correlation factor into a static and a
dynamic component. The dynamic component provides, in fact, in the t → ∞ limit,
a description of the amount of F̂T eigenstates ending up on the product side of the
dividing surface. Different ways of exploiting these ideas to the end of carrying out the
actual calculations of the value of the thermal rate coefficients and of cumulative reaction
probabilities, are given in refs. (36; 37; 38; 39).
RTO-EN-AVT-142 10A - 11
4 Approximate Methods
The practical impossibility of carrying out exact calculations for more complex systems
has prompted the development of computational procedures based on various decoupling
schemes effective in reducing the dimensionality of the problem.
4.1 Reduced dimensionality quantum methods

The goal of working out approximate formulations of the cross section and of the rate
coefficient was first accomplished for atom diatom systems by introducing decoupling
schemes of the kind energy sudden (40), centrifugal sudden (41) and infinite order sudden
(42). In these decoupling schemes either diatomic rotations or atom diatom orbiting (or
both) are treated in an approximate way. For systems made of four or more atoms the
coupling of the various degrees of freedom, while significantly increasing the complexity
of the dynamical treatment, most often plays a negligible role in determining the reaction
outcome.
This has allowed a split of the dynamical treatment of strongly coupled degrees of
freedom from weakly coupled ones. Strongly interacting degrees of freedom are treated
rigorously while weakly interacting ones are treated approximately. The most popular
approximations are based either on adiabatic assumptions or on the parametrization of
some variables. As an example, in the rotating bond approximation (RBA) of diatom-
diatom non linear collisions, two radial coordinates and one bending angle are explicitly
treated while the other three degrees of freedom are kept frozen (43). The overall result
is then worked out by averaging quantities calculated at the different values of the frozen
variables. In the adiabatic bend approximation (ABA) (44; 45), instead, the three radial
coordinates are treated explicitly while the three bending angles are treated adiabatically.
This means that the overall wavefunction is factored out and eigenvalues associated with
the effect of the terms of the Hamiltonian in the related coordinates are used to obtain
the effective Hamiltonian for the coordinates to be treated exactly.
As already mentioned the calculation of dynamical properties can be performed by
determining the cumulative probability which can be estimated in an approximate way
using a transition state (TS) schematization of the reactive process. In this view the role
played by the overall rotation of the system is that of shifting in energy (of a quantity
Eshif t related to the rotational energy of the system at the TS geometry) the reaction
J
probability Pi,f (Etr ). This approximation, usually called J-shifting (45), links the value of
the state-to-state probability calculated at a given value of the total angular momentum
quantum number J to that calculated at a reference value (Jref ). The relationship used
is
J Jref
Pvj,v ′ j ′ (Etr ) = Pvj,v ′ j ′ (Etr − Eshif t ) (27)
where Eshif t is the difference in energy between the overall rotational eigenstates J and
Jref at the transition state and Jref is usually taken equal to zero. A more elaborated
J
method linearly interpolates Pi,f (Etr ) between the probabilities calculated at the two Jref
values most closely sandwiching the actual J value of the calculation.
10A - 12 RTO-EN-AVT-142
4.2 Mixed quantum and classical mechanics approaches

Approaches alternative to quantum calculations are those based on classical mechanics.
Several of these approaches use classical mechanics formulations to describe slower mo-
tions and quantum mechanics formulations to describe the remaining (faster) degrees
of freedom. Related schemes have been applied to divide electronic from nuclear mo-
tion and to formulate widely general scaling procedures (47; 48). Some of these mixed
quantum-classical approximations have been derived in a rigorous way in refs. (49; 50) by
introducing a specific basis-set of the orthogonal polynomial type centered at a ”classical”
trajectory and expanded around it as the dynamics evolves in time. In this case the forces
appearing in the classical equations of motion are not the usual Newtonian ones but more
general forces usually called ”quantum forces”.
The chosen basis can be used in combination with the usual time independent basis
functions or collocated representations. This allows to work out mixed quantum classical
approaches and to monitor the quantum classical correlation and measure the accuracy
of classical path treatments.
When systems become very large, the computational procedures need to be simplified
further also at the level of the calculation of the electronic energies (in this case, in fact, the
accuracy of the calculated electronic wave function becomes intrinsically poor (51)). As a
result, the choice of election is to combine the use of classical mechanics for treating the
dynamics of the nuclei with the adoption of approximate schemes to calculate electronic
energy (like in density functional approaches (51)) and greatly simplify the on-the-fly (52)
computational machinery.
These Ab-initio Molecular Dynamics simulations are presently (53; 54) applied to
the study of physico-chemical properties, such as reactivity and dynamic relaxation, of
several systems (55). Hybrid procedures linking Quantum-Mechanical parameterized de-
scriptions of the ‘active sites’ of the molecule with a Molecular Mechanics description of its
inactive framework are highly popular. These Quantum Mechanics Molecular Mechanics
(QMMM) treatments (56) show advantages and limitations. They still lead, in fact, to
dynamical results which cannot be inferred by a mere static analysis of the features of
the potential energy surface though they do not treat most of them accurately.
At various stages of a trajectory calculation one has the possibility of treating semiclas-
sically those degrees of freedom for which a classical approach is inadequate by associating
to them a semiclassical wave depending on the classical action accumulated along the clas-
sical path. This allows to regain concepts like flux, interference, resonance and tunneling
within a trajectory framework and reproduce quite closely some quantum features of the
results.
4.3 The pure classical mechanics computational machinery

Purely classical mechanics approaches found the dynamical treatment on the integration
of classical trajectories (CT). For this reason they are called quasiclassical or QCT when
initial and final states are in some way discretized. CT methods assume that the nuclei
involved in a chemical reaction obey classical mechanics and roll as point mass particles on
the potential energy surface of the system. Accordingly, HNCT (R), the classical analogue
RTO-EN-AVT-142 10A - 13
of the quantum nuclear Hamiltonian HN (R) of eq. (7), reads

N
X nucl p2k,xR + p2k,yR + p2k,zR
HNCT ({R}) = k k k
+ V ({R}) (28)
k
2mk
and the evolution of the system is followed by integrating the equations of classical me-
chanics starting from different sets of initial conditions (for atom diatom systems initial
conditions are given by the vibrational number ν, the rotational number j, the transla-
tional energy Etr , the impact parameter b, the elongation of the diatom, the phase of
the rotating diatom, the angles formed by the rotational angular momentum and the
velocity with the molecular plane). Various formulations of the classical particles motion
equations can be given. In the widely used Hamilton’s version they read as
dsRk ∂HN
=
dt ∂psRk
dpsRk ∂HN
=−
dt ∂sRk
for each cartesian projection sRk (of the position vector Rk ) and psRk (of the momentum
vector pk ) of each atom k of mass mk of the molecular system. The equations are inte-
grated starting from one set of allowed initial conditions of the reactants in state i and
are stopped either when the maximum number of interactions steps has been reached or
when an asymptotic geometry of the products has been reached. As already mentioned,
discrete features of quantum results are then enforced in QCT ones by arbitrarily dis-
cretizing the energy of classical bound motions. Usually for atom diatom systems this
means that the counter Nνjν ′ j ′ (associated with trajectories starting from νj and ending
with a classical vibrotational energy closer to that of the ν ′ j ′ state than to any other one)
is incremented by a unit. When all planned Ntj trajectories are integrated Pνj,ν ′j ′ is set
equal to Nνj,ν ′ j ′ /Ntj and σνj,ν ′ j ′ is set equal to πbmax Pνj,ν ′j ′ (with bmax being the maximum
value used for the impact parameter). CT and QCT approaches can often provide esti-
mates of rate coefficients, cross sections, angular distributions and reaction probabilities
of reasonable accuracy. Moreover, they allow a pictorial view of the mechanisms governing
chemical reactions. Obviously, the CT method is an approximation to the nuclear motion
and it becomes more accurate when quantum effects are negligible (as is the case of heavy
nuclei, large collision energies and highly averaged reactive properties).
5 Advanced Computing
As it has been already mentioned, the demand for computer resources prompted by chem-
ical reactivity calculations and related realistic a priori simulations is as high as that of
other grand challenges of modern computational science. To guarantee the computational
feasibility of these applications one has to resort to the exploitation of the innovative
features of parallel and distributed computers by decomposing the problem in simpler
decoupled subproblems and distributing the resulting independent blocks of the codes
for execution on a large quantity of processing elements. The application of some de-
compositions based on physical considerations (like separating the electronic structure
10A - 14 RTO-EN-AVT-142
calculation, factorizing the time variable, disentangling the center of mass motion and
performing a partial wave expansion) have been already discussed before. However, when
dealing with large systems and complex applications one has to effectively exploit the in-
novative features of parallel and distributed computing by carrying out a decomposition
of the problem at an algorithmic level.
5.1 Parallel computing

To effectively tackle the problem of parallel restructuring a computational application the
following aspects need to be specifically considered (57; 58; 59)
1. the key algorithms need to be optimized for parallel execution,
2. existing software modules need to be integrated efficiently,
3. different programming models and languages need to be used at various levels,
4. retrieving and storing of data structures need to be reorganized and streamed,
5. performances of the adopted articulation need to be measured under different con-

ditions for its improvement.
This particularly difficult and time consuming job is usually carried out by reorganizing
the relevant suites of codes and by inserting the appropriate directives and commands
chosen among those of the most popular parallelization libraries (60). However, the need
for ensuring both reliability and standardization on one side and the difficulty of keeping
the pace of the continuous evolution of architectures and simulation techniques on the
other side have made it necessary to produce tools guaranteeing the automatic or semi-
automatic portability of applications (also in the sense of performance portability) across
computing platforms.
Significant progress along this direction has been made using structured environments.
A typical structured environment useful for a semi-automatic parallelization of the appli-
cations is SKIE (61). SKIE is an integrated environment providing a new application ori-
ented set of instruments allowing the rapid development and prototyping of applications.
Such an environment is based on some optimized and ready-to-use parallel structures,
called skeletons. The skeletons can embody sections of codes allowing so far an extended
reuse of the existing sequential (written in the most popular high level sequential lan-
guages) or parallel programs by encapsulating them in modules. Examples of skeletons
are processor farms (a pool of worker processes computes a pool of independent tasks);
pipelines, (different processes carry out in a sequence the various phases of the computa-
tion); map data-parallel computations (all the elements of a data structure are updated
or computed at the same time). These structures are handled using a coordination lan-
guage (CL) called SKIE-CL and can be utilized to coordinate and connect any sequential
or parallel module encapsulated using a SKIE-CL wrapper. The wrapper ensures that
parameter passing and data representation are consistent among the modules composing
a parallel application. SKIE-CL makes use of instruments like control, stream parallel
and data parallel though it accepts also in the input and output parameter lists all the
usual basic types of variables (integers, real, etc), records, and multi-dimensional arrays.
RTO-EN-AVT-142 10A - 15
In addition, SKIE-CL provides a stream data type which allows programmers to generate
or consume sequences of items of indefinite length. The peculiar feature of these patterns
is the fact that they can be freely composed ”a la carte” to build complex structures.
SKIE automatically generates also an optimized implementation of the skeleton com-
position. This means that when using a SKIE skeleton the support not only generates
automatically the code needed for parallel interaction, but it also optimizes the resources
allocated to each skeleton, decides the best granularity of computation and locates ineffi-
ciencies in the global structure.
Going to a more abstract level skeletons can be generalized practically to any form
(or combination of forms) and size. This is indeed the key feature of the coordination
language ASSIST (62; 63) that is an evolution of SKIE. ASSIST is made of a graph
whose nodes are parallel (parmod) and sequential (seqmod) modules. The arches of the
graph represent streams of data. Interaction among parmods can occur also via shared
objects (data, memory regions, functions, etc.) for which ASSIST provides a primitive
access mechanism by implementing or emulating a shared memory access. Non primitive
accesses are instead provided for external objects (like DSM libraries, CORBA servers,
etc.) for which the user has to take care of the access, synchronyzation and consistency
via directives to be inserted in the code. Activities in different parmods can be parallel
or concurrent (like, for example, in a pipeline). Parallel or concurrent activities can take
place also within the same parmod. They can be either farm-like or data-parallel-like (or
mixed in a non-deterministic way depending on the structure of the data and the status of
the computation). This means that ASSIST allows the following two hierarchical levels:
among various parmods and within the same parmod. The first level describes a graph
of the data flow type while the second can describe computations both of a data parallel
domain decomposition type and of a functional replication farm type. ASSIST is also
scarcely invasive since the computation is described using a set of procedures wrapping
the sequential user code and organized in the parallel fashion specified by the used parmod
type.
ASSIST has been already used to parallelize time dependent reactive scattering ap-
plications. Applications implemented using ASSIST and its libraries (see ref. (64)) have
shown to clearly outperform their versions implemented using MPI. Similar studies have
been carried out for time independent applications separately for the surface functions
calculations and the propagation along the reaction coordinates. Tests performed using
MPI on an eigenvalue finding routine are illustrated in ref. (65)
5.2 Virtual organizations and Grid enabled applications

Moving to realistic simulations of complex chemical systems, however, parallelism on a
single (no matter how powerful) machine is still insufficient. As a matter of fact, the
emerging computing paradigm is the computing Grid (66; 67). The Grid offers the pos-
sibility of solving complex problems using a (preferably very large) set of distributed
computers as a single unified computing resource. To this end the Grids enable the shar-
ing, selection, and aggregation of a wide variety of geographically dispersed resources
ranging from PCs to supercomputers, storage systems, data sources and specialized de-
vices. These may be owned by different organizations and work for completely different
purposes within the virtual organization. Accordingly a Grid can be viewed as a seamless
10A - 16 RTO-EN-AVT-142
integrated computational platform.

Moreover, the Grid is by definition at the same time both a collaborative environment
and a market in which the users interact to find a solution for their problems and/or to offer
their solutions to other users. Just to mention a few services which can be ”traded” on the
Grid we can list: Computational services, Data services, Application services, Information
services and Knowledge services. This makes Grid applications intrinsically large scale
and multidisciplinary. Obviously, in order to make the Grid computing effective, a large
number of tools concerning security, resource allocation, costs management, information
flux, software development, process execution, resource aggregation and scheduling need
to be implemented. As a matter of fact the Grid is able to use in a synergistic way
resources of all kinds including those which cannot be physically replicated in a single
site. This means that one can easily scale up computing cycles as well as competences
regardless of their location to work in a coordinated fashion. In summary, the Grid is a
virtual place for composition of multiple administrative domains and autonomies to work
concertedly on a large variety of heteregenous machines and problems which naturally
scales up from a few integrated resources to very many of them a smooth, dynamic,
adaptable and interoperable way.
The European Grid of election is the already mentioned EGEE (68). Within EGEE
a memorandum of understanding has been signed last March with the COMPCHEM vir-
tual organization to foster the development of molecular science complex simulations. The
mapping of a complex application on the computing grid is, indeed, a difficult task. Such
a process is not simple because there is not a unique correspondence between the variation
of physical and mathematical parameters of a complex computational application and the
distribution on the grid of the computing blocks. This is not only due to the fact that
a variation of the computational parameters alters the relative importance of the various
computational blocks and of the related demand of computational resources but also to
the fact that the support of the grid infrastructure at run time is not deterministic (69).
For this reason it is vital to work out data graphs and build workflow managers allowing
a proper independent handling of the computing blocks at various levels of distribution.
The first step of this process is, therefore, the breaking of the computational procedure
into independent or loosely coupled computing blocks. In fact, the singling out of inde-
pendent (or almost independent) computational tasks is propedeutic to the design of any
concurrent organization of the relevant computer programs.
5.3 The Grid implementation of a molecular science simulator

The particular Grid enabled application considered as a study case by COMPCHEM is,
as it has been already mentioned, GEMS the Grid Enabled Molecular Simulator. A demo
version of GEMS (GEMS.0) has been already implemented on the production Grid of
EGEE and presented at the first EGEE review workshop (70). GEMS.0 is derived from
the SIMBEX simulator (see ref. (1)) developed by the homonimous working group of the
COST Chemistry Action D23 (Metalaboratories for complex computational applications
in Chemistry) (71). It takes care of evaluating the cross sections and the product distri-
butions (plus some non observable quantities) of a crossed molecular beam atom-diatom
experiment using a quasiclassical approach. In a quasiclassical approach, the observable
properties of a scattering experiment are determined by performing a multidimensional
RTO-EN-AVT-142 10A - 17
integral over the unselected variables of the experiment. The kernel of the integral is
given by a boolean function whose value depends on the result of the integration of the
Hamilton equations of the molecular system.
This means that one has to integrate large batches (say Ntj ) of classical trajectories
for which some of the initial conditions are randomly selected. The calculation of each
individual trajectory implies the integration of the equations of motion of the Nnucl atoms
composing the system. This means that the simulation has to generate, possibly in a
deterministic way, a large number of subsets of pseudorandom numbers (with each subset
determining the initial conditions of a given trajectory). The overall workflow of GEMS.0
is, therefore, articulated into:
1. a first part defining the parameters of the calculation, computing quantities of gen-
eral use and assembling the information needed for the calculation of the potential,
2. a final part performing the averaging of the calculated quantities and carrying out
the graphical elaboration of the properties to be rendered possibly in real time on
the virtual monitors,
3. a central (the key) part iterating over the distribution to the workers of the trajec-
tories to be integrated and to the recollection of the related results to update the
reactive probability.
The central part of GEMS.0 is the kernel of the calculation that can be efficiently dis-
tributed on the Grid using a task-farm scheme. The distribution can take place by as-
signing the integration of individual (or blocks of) trajectories to the worker nodes after
generating initialseed (the first seed of the subset needed by each trajectory or subset
of trajectories) at Master process level. In GEMS.0, in order to keep the generation of
initialseed as much deterministic as possible, it has been chosen to perform the iterat on
the individual trajectories. Accordingly, to integrate Ntj trajectories on a Grid made of
Nnode nodes (Nnode is assumed to be smaller than Ntj ) the Master process (see Fig. 1)
generates and SENDS out, for the first Nnode iterations, the initial seed of each trajec-
tory without waiting for the result of the integration to be returned. For the subsequent
iterations the initialseed of each trajectory is sent out only after one of the worker nodes
(anynode ) has sent back its result. After the sending to the worker nodes of initialseed for
the Ntj trajectories is completed the Master process still needs to iterate over the Nnode
nodes to collect the results of the remaining trajectories and send a conventional signal
(we chose to adopt for this purpose a negative value of the initial seed) to stop the work
at worker process level.
As sketched in Fig. 2 the worker process devoted to the integration of each individual
trajectory receives at the beginning (once for ever) the general information. Then it
iterates on receiving initialseed of the subset necessary to generate the trajectory initial
conditions. After receiving the trajectory initial seed the worker process, if the seed is
not negative, generates the remaining pseudorandom numbers of the needed subset to
work out either at random (fully or partially according to the chosen distribution) or the
specific initial value of the parameters of the trajectory like Etr , J, the velocity orientation,
the diatomic internuclear distance, orientation and position as well as its vibrotational
(vj) quantum state. The iteration on time then starts to integrate the trajectory from
10A - 18 RTO-EN-AVT-142
Read input data: v, j, Etr , ∆to , error, maxstep, Ntj ...

Perform preliminary calculations
BROADCAST to the worker nodes the needed general information
Generate initialseed for the first trajectory
inode =0
Do itj = 1, Ntj
inode =inode +1
Generate next initialseed of the of the pseudorandom series
IF inode > Nnode THEN
RECEIVE from anynode the result and update Pvj,v′ j ′
SEND the initialseed to anynode
ELSE SEND the initialseed to inode
ENDIF
EndDo itj
Do inode = 1, Nnode
RECEIVE from anynode the result and updated Pvj,v′ j ′
SEND to anynode a negative seed to stop activities
EndDo inode
Figure 1: Pseudocode of the trajectory Master program.
an atom-diatom distance large enough to consider the system in its asymptotic reactant
arrangement. The integration is eventually terminated either when one of the atom-
atom distances has reached a value large enough to consider again the arrangement as
asymptotic or when the maximum number of integration steps (maxstep) has been reached.
J
If at the integration end point the system has reached the product asymptote the Pvj,v ′j′
′ ′
probability of the product diatom vibrotational (v j ) quantum state closest in energy to
the computed classical value is updated.
In GEMS.0 the PES is assumed to be of the LEPS type and the value of its parameters
to be available from a library (they may have been already calculated using two other
computational procedures, SUPSIM and FITTING, implemented in our Laboratory (72;
73)). The integration of a trajectory can be disposed to any computing node of the grid
while the integration outcome is accumulated by the master process sketched in Fig. 1 by
updating the value of the related quasiclassical probability Pvj,v′ j ′ which can be displayed
to the user on a virtual monitor.
Test runs of GEMS.0 (70) performed as EGEE (68) demonstrations have shown that
the simulator is highly suitable for a distribution on the heterogenous networked envi-
ronment of the Grid. In these runs from several thousands to millions or even billions of
trajectories (to this end particular attention has to be paid to the generation of the pseu-
dorandom sequence) can be run depending on how much the considered event is likely
to occur. The main feature of GEMS.0 is that of being cpu bound (this is in general
true even for systems larger than atom diatom ones). The implementation of GEMS.0
has impacted EGEE in two ways. The first of them is related to the specific requests of
GEMS.0 in terms of infrastructure, middleware and services. The second is related to the
characteristics that molecular simulations need to possess in order to be suitable for Grid
RTO-EN-AVT-142 10A - 19
RECEIVE from the master node the needed general information

[∗] RECEIVE the trajectory initialseed
IF seed < 0 STOP
Generate the needed subset of pseudorandom numbers
Calculate initial values of the integration variables
t=0
tstep = ∆to
Do it = 1, maxstep
t = t + tstep
Perform the time step integration
IF Energy and total angular momentum are conserved THEN
IF an asymptotic arrangement has been reached THEN
perform the asymptotic analysis
SEND results to Master
GOTO [∗]
ENDIF
ELSE t = t − tstep and reduce tstep
ENDIF
ENDDO it
GOTO [∗]
Figure 2: Pseudocode of the trajectory worker program.
implementation.
As to the first item it has become apparent that molecular simulations force the Grid
to better exploit various levels of parallelization and distribution using an appropriate
workflow computational procedures of different nature and origin. In other words, GEMS
requests EGEE to focus more on an effective coordination of brainware and knowledge.
As to the second item it has also become apparent that EGEE prefers direct calcula-
tions (with respect to data transfers) and coarse grain granularity schemes.
References
[1] Gervasi, O., Laganà, A., 2004, Future Generation Computer Systems, 20(5),
703.
[2] O. Gervasi, D. Cicoria, A. Laganà and R. Baraglia, Pixel, 10, 19 (1944).
[3] NA4: Memorandum of understanding for the EGEE applications, March

2005.
[4] Laganà, A., GEMS-testbed: Grid Enabled Molecular Simulator, an EGEE

testbed devoted to on demand interoperable cooperative simulations for
Molecular Science Applications, proposal n. IST035244 GEMS-testbed of
the 6th Framework Program.
10A - 20 RTO-EN-AVT-142
[6] http://www.molpro.net/
[7] http://www.msg.ameslab.gov/GAMESS/GAMESS.html
[8] http://www.gaussian.conv/g03.htm
[9] http://www.emsl.pnl.gov/docs/nwchem
[10] http://www.itc.univie.ac.at/ hans/Columbus/columbus main.html
[11] P.L. Houston, Chemical kinetics and reaction dynamics, McGraw Hill,
Boston 2001, ISBN 007-243537-2
[12] M. Baer, The theory of chemical reaction dynamics CRC Press, Boca Raton,
FL, 1985
[13] D.C. Clary, The theory of chemical reaction dynamics, D. Reidel, Boston,
1986.
[14] W.H. Lester Jr, The N coupled channel problem in Dynamics of Molecular
Collisions, Plenum Press, New York, 1976.
[15] J.C. Tully, Non Adiabatic Processes in Molecular Collisions in Dynamics of

Molecular Collisions, Plenum Press, New York, 1976.
[16] G. C. Schatz, Rev. Mod. Phys. 61, 669 (1989); G. C. Schatz, Lect. Not.
Chem. 75, 15 (2000).
[17] J. N. Murrell, S. Carter, S. C. Farantos, P. Huxley, and A. J. C. Varandas,

Molecular Potential Energy Functions (Wiley, Chichester, 1984).
[18] A. J. C. Varandas, Adv. Chem. Phys. 74, 255 (1988).
[19] A. J. C. Varandas, J. Molec. Str. (Theochem) 166, 59 (1988).
[20] D. G. Truhlar, R. Steckler, and M. S. Gordon, Chem. Rev. 87, 217 (1987).
[21] E. Garcia and A. Laganà, Mol. Phys 56, 629 (1985).
[22] F. O. Ellison, J. Am. Chem. Soc. 85, 3540 (1963).
[23] P. J. Kuntz, Ber. Bunsenges. Phys. Chem. 86, 367 (1982).
[24] A. Laganà, G. Ochoa de Aspuru, and E. Garcia, J. Chem. Phys. 108, 3886
(1998).
[25] G. Ochoa de Aspuru and D. C. Clary, J. Phys. Chem. A 102, 9631 (1998).
[26] U. Burkert and N.L. Allinger, Molecular Mechanics (ACS Monograph Series,
Washington DC, 1982).
RTO-EN-AVT-142 10A - 21
[27] R. G. Newton Scattering theory of waves and particles (Springer-Verlag, New

York, 1982).
[28] R. T Pack and G. A. Parker, J. Chem. Phys. 87, 3888 (1987).
[29] D.H. Zhang, M. Yang, S. Lee, M. Collins, First principle quantum dynamical
study of four atom reactions in Modern trends in chemical reaction dynamics
Experiment and theory (Part II) Adv. Series in Phys. Chem, X. Yang and
K. Liu Ed., World Scientific, Singapore 2004.
[30] A. Kuppermann, J. Phys. Chem. 100, 2621 (1996); J. Phys. Chem.

1997,101,6368; J. Phys. Chem. 108, 8894 (2004)
[31] G. G. Balint-Kurti, in International review of Science, Series II, Vol. 1. Eds.

A.D. Buckingham and C.A. Coulson (Butterworths), page 286 (1975).
[32] D.M. Medvedev, E.M. Goldfield, S.K. Gray, Comp. Phys. Comm 166, 94
(2005).
[33] U. Manthe, Lecture Note in Chemistry 75, 130 (2000).
[34] W. H. Miller, S. D. Schwartz, and J. W. Tromp, J. Chem. Phys. 79, 4889

(1983).
[35] F. Matzkies and U. Manthe, J. Chem. Phys. 110, 88 (1999).
[36] D. H. Zhang and J. C. Light, J. Chem. Phys. 104, 6184 (1996).
[37] S. M. Miller and T. Carrington, Chem. Phys. Lett. 267, 417 (1997).
[38] W. H. Thompson and W. H. Miller, J. Chem. Phys. 106, 142 (1997).
[39] F. Matzkies and U. Manthe, J. Chem. Phys. 108, 4828 (1998).
[40] D.S. Chase, Phys. Rev. 104,838 (1956).
[41] R.T Pack J. Chem. Phys. 60, 633 (1974); P. McGuire and D.J. Kouri, JCP
60, 248 (1974).
[42] M. Baer, E. Garcia, A. Laganà, O. Gervasi, Chem. Phys. Letters 158, 362
(1989)
[43] D.C. Clary J. Chem. Phys. 95, 7298 (1991)
[44] D. Wang and J. Bowman J. Chem. Phys. 96, 8906 (1992)
[45] J. M. Bowman, J. Chem. Phys. 95, 4960 (1991); J. M. Bowman, Lecture

Notes in Chemistry, 75, 101 (2000).
[46] U. Manthe and W. H. Miller, J. Chem. Phys. 99, 3411 (1993).
[47] D. C. Clary, J. Phys. Chem. 98, 10678 (1994).
10A - 22 RTO-EN-AVT-142
[48] M. Baer and D. J. Kouri, in The Theory of Chemical Reaction Dynamics,

ed. D.C. Clary (Reidel, Dordrecht, 1986), pp. 167-192.
[49] G. D. Billing, J. Chem. Phys. 107, 4286 (1997).
[50] G. D. Billing, J. Chem. Phys. 110, 5526 (1999); Chem. Phys. 242, 341(1999).
[51] R. M. Dreizler and E. K. U. Gross, Density Functional Theory (Springer

Verlag, Berlin, 1990).
[52] V. Bakken, J. M. Millam, and H. B. Schlegel, J. Chem. Phys. 111, 8773

(1999).
[53] R. Car, M. Parrinello, Phys. Rev. Lett. 55, 2471 (1985).
[54] M. E. Tuckerman, P. J. Ungar, T. von Rosenvinge, and M. L. Klein, J. Phys.

Chem. 100, 12878 (1996).
[55] For a review of ab initio classical trajectory methods see K. Balton, W.

L. Hase, and G. H. Peshlherbe, in Modern Methods for Multidimensional
Dynamics Computation in Chemistry, edited by D. L. Thompson (World
Scientific, Singapore, 1998), p. 143.
[56] F. Bernardi, M. Olivucci, M. A. Robb, J. Am. Chem. Soc. 114, 1606 (1992).
[57] M. Vanneschi, Lecture Notes in Chemistry 75, 168 (2000)
[58] A. Hey, C.J. Scott, M. Surridge, C. Upstill, Integrating computation and

information resources, an MPP perspective, in: Proceedings of 3rd Inter-
national Working Conference on Massively Parallel Programming Models,
(IEEE Computer Society Press, Los Alamitos, 1998), 174–184.
[59] M. Vanneschi, PQE2000: HPC tools for industrial applications, IEEE Con-
currency 6 (1998) 68–73.
[60] http://www3.niu.edu/mpi/
[61] B. Bacci, B. Cantalupo, P. Pesciullesi, R. Ravazzolo, A. Riaudo, and L. Van-

neschi, Skie user guide (version 2.0), Technical report, QSW Ltd., Roma,
Italia, December 1998.
[62] M. Vanneschi, The programming model of Assist, an environment of parallel

and distributed portable applications, Parallel computing 28, 12 (2002)
[63] P. Ciullo, M. Danelutto, S. Magini, L. Potiti, ASSIST-CL User Manual,

Technical report, ASI-PQE2000, 2002.
[64] S. Gregori, S. Tasso, A. Laganà, Lecture Notes in Computer Science, 3044

437-444 (2004).
[65] A. Laganà, ASI-PQE2000 WP4 Technical report 003-005, March 2002.
RTO-EN-AVT-142 10A - 23
[66] Foster I., Kesselman C. Eds., The Grid Blueprint for a Future Computing
Infrastructure, Morgan Kaufmann: San Francisco CA, 1999.
[67] Foster I., Kesselman C. Tuecke S. The anatomy of the grid: Enabling scalable
virtual configurations, Int. J. Supercomp. Appl. 15 (2001).
[68] Enabling Grids for E-Science in Europe (EGEE) http://www.eu-egee.org
[69] Laganà, A., Crocchianti, S., Bolloni, A., Piermarini, V., Baraglia, R., Ferrini,
R., Laforenza, D.: Computational granularity and parallel models to scale up
reactive scattering calculations. Computer Physics Communications 128/1-
2 295 (2000)
[70] First Egee Application Migration Report, EGEE-

DNA4.3.1.523422-v4-7-13.doc (2005); Egee review workshop,
(http://agenda.cern.ch./fullAgenda.php?ida=a043803) Geneva, 2005.
[71] Preparatory wokshop ”METACHEM”, Brussels, November, 1999

and Action D23 of COST in Chemistry Memorandum of under-
standing (see http//costchemistry.epfl.ch/docs/D23/d23-main.htm,
http://www.unil.ch/cost/chem).
[72] Storchi, L., Tarantelli, F., Laganà, A.: Computing Molecular Energy Sur-
faces on a Grid, ICCSA 2006, Glasgow (2006).
[73] Arteconi, L., Pacifici, L., Laganà, A.: Fitting potential eneergy surfaces for
grid enabled molecular simulators, ICCSA 2006, Glasgow (2006).
10A - 24 RTO-EN-AVT-142
THEORETICAL ESTIMATES OF REACTION
OBSERVABLES VIS-A-VIS MODERN
EXPERIMENTS
A. Laganà
∗
Dipartimento di Chimica, Università di Perugia,

06123 Perugia, Italy
8 February 2006
Contents
1 Abstract 3
2 Introduction 3
3 The universal CMB 4

3.1 The reactant beam generation . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1.1 The beam specificity . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1.2 Continuous and pulsed supersonic beams of transient species . . . . 6
3.1.3 Reagent state-selection . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2 The product detection and characterization . . . . . . . . . . . . . . . . . . 7
3.2.1 Rotatable mass spectrometry . . . . . . . . . . . . . . . . . . . . . 7
3.2.2 Photoionization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2.3 Laser spectroscopic detection . . . . . . . . . . . . . . . . . . . . . 8
3.2.4 Ion-imaging and Doppler methods . . . . . . . . . . . . . . . . . . . 9
4 Some case studies 9

4.1 The atom-diatom reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.1.1 The H + H2 reaction . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.1.2 The Li + HF reaction . . . . . . . . . . . . . . . . . . . . . . . . . 13
4.2 Polyatomic reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2.1 The OH + H2 reaction . . . . . . . . . . . . . . . . . . . . . . . . . 15
∗
Interesting discussions with several scientists involved in CMB experimental studies and, in particular,
with P. Casavecchia (a world expert in experimental molecular dynamics who has provided us with some
of his material) is acknowledged. Financial support from ASI, MEC and MIUR and computer time
support from CINECA (I), EPCC (UK) and SGI/IZO-SGIker (E) is also acknowledged.
Laganà, A. (2007) Theoretical Estimates of Reaction Observables vis-à-vis Modern Experiments. In Experiment, Modeling and Simulation
of Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 10B-1 – 10B-26). Educational Notes RTO-EN-AVT-142,
Paper 10B. Neuilly-sur-Seine, France: RTO. Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 10B - 1
Theoretical Estimates of Reaction
Observables vis-à-vis Modern Experiments
4.2.2 The OH + CO reaction . . . . . . . . . . . . . . . . . . . . . . . . 16
10B - 2 RTO-EN-AVT-142
1 Abstract
Recent advances in experimental chemical reaction dynamics are examined. The paper
focuses on crossed beam technologies because the detailed single collision information
they provide is a stringent test of the adopted molecular interaction. In this respect some
prototypal atom-diatom and diatom-diatom elementary reactions are analysed.
2 Introduction
Modern understanding of chemical reactive processes is based on the assemblage of a
suitable potential energy surface (PES) out of the first principles (ab initio) calculations
of electronic energies and on the integration (on the adopted PES) of the equation of
motions of the nuclei. In this respect theoretical and computational machineries have
made impressive progress and it is now possible to compute rigorously using a priori
techniques several averaged as well as highly detailed properties of gas phase reactions
(see the preceeding lecture (1) and references therein).
This is indeed a clear example of synergistic development of theory and experiment.
As a matter of fact, the traditional quantity to measure for chemical reactions used
to be the concentration (of either the reactants or the products) as a function of the
time elapsed from the starting of the process while keeping under control temperature
and pressure. Concentrations can be easily measured for processes occurring on the
time-scale of minutes or longer (this is the case of the popular thermostated vessels,
discharge flow, flash photolysis, shock wave, etc. measurements). For shorter time-scales
this technique is impractical unless the determination of the concentration (or even the
starting of the process) is reconducted to the determination of fast to measure physical
properties of the system (like some spectroscopic ones). Moreover, from bulb experiments
no information can be obtained on the intervening individual state-to-state elementary
uni and bi-molecular processes.
Such an information can be obtained, instead, from Crossed Molecular Beams (CMB)
experiments. As a matter of fact, the measurement of the efficiency of fast bi-molecular
reactions has become possible in the mid fifties (of the 20th century) when Datz and
Taylor (2) carried out the first CMB experiment of the K + HBr → KBr + H reaction.
The distinct innovative feature of CMB experiments that marked the birth of the experi-
mental molecular dynamics, is the possibility of working in a single-collision regime. The
conceptual foundations of molecular dynamics date back to the advent (in the twenties) of
quantum mechanics. Quantum mechanics studies prompted, in fact, the request for inves-
tigating the collision processes in beams and their outcomes. Theoretical studies of atomic
and molecular collisions are usually articulated in a first step devoted to the calculation
of electronic energies (for which separate accurate ab initio computational procedures are
available) and a second step integrating the equations of motion of the nuclei on the PES
associated with the electronic energies calculated in the first step considered (for which
software is less standard). As a matter of fact, this has led to the construction of the
first semiempirical PES (3) for the H + H2 reaction, to the extension of the theoretical
investigation to other systems of chemical interest (4) and to the conduction, as well, of
the first classical trajectory (QCT) computational study of reaction dynamics (5).
The real breakthrough of molecular reaction dynamics, however, has occurred in the
sixties. As to theory, in fact, significant improvements were made by the early 1960s
in calculating both ab initio and semiempirical PESs as well as in integrating classical
trajectories for the full three-dimensional treatment of atom-diatom reactions (6). As
to the experiment, in the sixties the first experiments measuring product angular and
translational energy distributions for a large variety of alkali atom reactions (7; 8) were
performed. During the same period, the infrared-chemiluminescence technique (9) was
developed which allowed the determination of nascent product vibrational-rotational dis-
tributions for many halogen and hydrogen atom reactions. Finally, at the end of the
’60s, the first universal CMB apparatus was built. The universal CMB machines are
characterized by the fact of having supersonic beam sources for the reagents and an elec-
tron impact mass spectrometer detector (10) and a pseudo-random time-of-flight (TOF)
velocity analyser for the products. This allows the universal CMB machine to measure
not only the integral (over the angular range) cross section (ICS) but also the differen-
tial (DCS) one. This has opened the field to other experimental innovations (see for an
introduction to the field the books (12; 13) as a general review and for focus on some
specific aspects see the papers (8; 14; 15; 16)). More recent reviews are given in refs.
(17; 18; 19; 20; 21; 22; 23; 24; 25; 26; 27; 28; 29; 30; 31; 32; 33; 34; 35; 36; 37; 38; 39; 40; 41)
with the latter being the main source of experimental information for the present paper.
The paper is articulated into the following sections: section 2 illustrates the modern
evolution of crossed molecular beams as a universal tool for investigating bi-molecular
processes; section 3 discusses the case of few atom systems to analyses the interaction
between experimental and theoretical studies.
3 The universal CMB

As already mentioned the advent of the universal CMB machine has profoundly innovated
the field of reaction dynamics (as testified also by the award of the 1986 Nobel Prize to
three top researchers of the field). The universal CMB has, in fact, made it possible
to investigate a large variety of elementary reactions. As an example, during the 1970s
and early 1980s several reactions involving hydrogen, oxygen and halogen atoms, were
investigated in detail using universal CMB apparatuses (15; 42). The distinctive feature
of CMB techniques, due to the use of ultra-high-vacuum (UHV) technologies, is the
possibility of dealing specifically with all the parameters of an elementary reactive process.
Experiments performed on universal CMB machines, in fact, can measure the lab-
oratory (lab) fixed angle (Ilab (θ′ , u′)) flux intensity of products (by convention primed
quantities refer to products) as a function of θ′ (the angle of deviation of the recoil direc-
tion from the original one) and u′ (the speed of the recoiled flux) starting from reagents
in well defined states for virtually any reaction (11). However, theoretical treatments
work on the center-of-mass (cm) reference frame. This means that, for example, the sim-
plest bi-molecular elementary reactive process i.e. the atom-diatom reaction that can be
schematized as:
′ ′ ′
A(vA , EA ) + BC(vBC , EBC , νBC , jBC ) → AB(vAB , EAB , νAB , j′AB ) + C(vC′ , EC′ ) (1)
in which reactants and products are characterized by the speed v, the angle Θ, and a set of
internal (electronic E, vibrational ν, and rotational j) quantum states (primed quantities
are for products, unprimed for reactants). For CMB experiments several techniques have
been developed to generate the reactants in specific quantum states and with a narrow
velocity distribution as well as to detect the velocity, angular and in internal quantum
states of the products.
3.1 The reactant beam generation

The first key feature of a CMB experiment is the production of sufficiently intense contin-
uous or pulsed beams of reactants with a sufficiently narrow velocity distribution. At first
the problem was tackled by mechanically selecting effusive beams. More recently the prob-
lem has been addressed by generating supersonic beams. Supersonic beams have a higher
intensity, a narrower velocity spread and are generated by expanding a high-pressure gas
through a nozzle into a high-vacum chamber (43). This is particularly difficult for unsta-
ble reactants like atoms and radicals. However, the progress made in dealing with pulsed
lasers and seeded beam techniques has allowed the generation of supersonic beams of
virtually any atomic species and of a large variety of radicals (e.g. continuous and pulsed
beams of N, C and of several diatomic radicals like OH and CN). It has also allowed
to generate pulsed beams of polyatomic free radicals allowing the study of reactions of
combustion-, atmospheric-, and astro-chemistry interest. The main methods used to gen-
erate beams of transient reactants (atoms or molecular radicals) are similar to those used
in bulb studies, i.e., pyrolysis, photolysis, or discharge of a suitable precursor (eventually
to be followed by a chemical reaction inside the nozzle).
3.1.1 The beam specificity

Intensity, purity, and cooling are the three main factors which come into play when one
desires to generate a beam of highly specific characteristics. High beam intensity is, of
course, a prerequisite for reactive scattering, to guarantee a detectable quantity of final
products. Purity is also a key issue to uniquely identify the detailed outcome of elementary
processes. Mixtures of species, however, and in particular mixtures of different electronic
states of the same species have been in various occasions exploited to study the dependence
of the cross section on the internal energy especially when one can play with the percentual
composition. Efficient cooling, or state selection, is of concern mainly in state-to-state
studies. Radical beams, when generated in a supersonic expansion, usually attain a
high degree of state selection, albeit restricted to the lowest rotational levels. However,
if the radical is initially created in highly excited states, for instance by photolysis or
high temperature pyrolysis, the cooling to the lowest ro-vibronic levels may not be very
efficient.
Further specificity of the reactants is obtained through the control of the orienta-
tion and alignment of reagents using exapole state-selection, brute force (by magnetic
or electric fields) alignment or orientation, and polarised laser radiation methods. These
methods have enabled the investigation of stereodynamical aspects of chemical reactions
at a fundamental level. Advances were also made in photo-initiated reactions, real-time
studies (femtochemistry) and transition state spectroscopy experiments. In more recent
years CMB methods have been extended to the generation and study of a large variety of
atoms, molecules and radicals of increasing complexity including biological systems with
a consequent shift of the research focus from fundamental to more applicative scopes.
3.1.2 Continuous and pulsed supersonic beams of transient species

Supersonic beams of metal atoms seeded in a carrier gas are usually generated continuously
in a high temperature oven (alkali, alkali earth) (26; 44) or by pulsed laser ablation (Mg,
Al, C, Si, and transition metals such as Ti, V, Cr, Cu, Y, Zr, Mo) (45; 46). This source has
also been successfully adapted to a universal CMB apparatus to measure reactive DCSs
for a large series of reactions with unsaturated hydrocarbons, of particular relevance to
astrochemistry and combustion. Other minor products of these generation methods have
also been used as reactant beams. Supersonic continuous beams of non-metal atoms
were produced traditionally by pyrolysis and/or electrical discharge (micro-wave or radio-
frequency), starting from dilute mixtures of the stable molecules in a rare gas. Continuous
N-atom and C-atom (47) beams both in the ground and the first electronically excited
state, sufficiently intense to allow measurements of DCS, have been produced using the
same radio frequency discarge built for O atom beams. Intense beams have been produced
also for D, O, S, Cl and F radicals either by pulsed laser photolysis or by pulsed discharge.
As to molecular radicals, important advances have been made in the generation of
continuous beams of OH and pulsed beams of CN. Beams of C3 H3 have been obtained by
laser photolysis of C3 H3 Br (48) and CN (49) by laser ablating carbon atom from graphite
in the presence of N2 . Beams of C6 H5 and C2 D have also been produced for use in CMB
experiments. A pulsed fully state-selected beam of OH(X2 Π3/2 , j = 3/2, Λ) radicals has
been generated using a pulsed electrical discharge with hexapole selection. More intense
beams of jet-cooled molecular radicals (OH, CH3 ) have been obtained by combining slit
supersonic expansion and pulsed electrical discharges (50).
3.1.3 Reagent state-selection

Most of the beam sources allow a selection of the reactants collision energy and have been
used to determine the translational energy dependence of DCSs and ICSs (the dependence
of ICS from the collision energy is often called also ”excitation function”). Translational
energy can be varied using rotating velocity selectors, changing the nozzle temperatures
and exploiting the seeding effect in a (rare gas) carrier gas. TOF selection of one reagent
has also been used in beam-cell experiments (16). Accurate excitation functions have
been obtained by using crossed pulsed beam apparatuses with independently rotating
beam sources, which make it possible to vary the translational energy down to very low
values by simply varying the crossing angle of the two beams (17; 43; 51). This has
allowed, for example, accurate measurements of the excitation function for the reactions
O(1 D), S(1 D), Cl(2 P), F(2 P), and CN + H2 (D2 , HD) (52; 53; 54; 55; 56; 57) and for the
reactions C(3 P)+O2 and C(3 P)+NO, C2 H4 (51).
An efficient way of preparing the reagent molecules in the lowest rotational states
is through the cooling which occurs during the expansion of the supersonic beam. The
preparation of H2 and D2 beams in the ground j = 0 rotational level has been obtained via
a catalytic conversion to p-H2 or o-D2. The production of rotationally excited HCl(ν=2,
j=1, 6, 9) with O(3 P) and O(1 D) (54; 58) has allowed the study of the effect of rotational
excitation on the product energy disposal.
To study the effect of vibrational energy on the DCS one can use laser radiation to
selectively excite vibrational modes (usually of stable molecules). An example of CMB
measurements of DCSs for a rovibrationally excited reagent is represented by the studies
of Li, K, Sr + HF(v=1, j=1, m=0) reactions (26). Highly detailed studies of selective
vibrational excitation have been carried out using the pump-probe and photo-initiation
methods (33; 59; 60).
Electronic and spin-orbit excitation have been mainly studied for atoms. The effect of
electronic excitation on chemical reactivity and dynamics is quite distinct because through
electronic excitation strongly endoergic (inefficient) reactions may become very efficient.
The effects of electronic excitation on ICS (especially of metal atoms) and on DCS have
been investigated by generating short-lived electronically excited alkali (Li, Na, K, Cs)
(61; 62), alkali earth (Ca, Sr, Ba) (44) and transition metals (V, Mo) (63). Electronic
excitation of Ca atoms has also been generated by DC discharge (31) while that of the O,
S, N and C non metal atoms has been obtained by RF discharge (O1 D), N(2 D), C(1 D))
(23; 64; 65) and by laser photolysis of a suitable precursor molecule. These species have
long lifetimes and can be generated well before the collision region. Spin-orbit effects and
differences in branching ratios for spin-orbit states of product atoms are also observed in
chemical reactions by measuring both quenching rates and chemiluminescent yields (see
(66; 67; 68)).
3.2 The product detection and characterization

In addition to the possibility of measuring the speed and the direction of reactants, the
other key feature of a CMB experiment is the possibility of measuring the speed and the
angular distributions of the products (or, when is the case, the angular distributions of
the state-selected products and their orientation or alignment). To this end several tech-
nologies have been developed to deal with the low density of scattered products. In the
’70s the Laser-Induced-Fluorescence (LIF) technique was adopted (69). Advances were
also made in the detection of products using photo-ionization by tunable synchrotron
or VUV laser radiation, CARS (Coherent-AntiStokes-Raman-Spectroscopy), infrared ab-
sorption, and REMPI (Resonance-Enhanced-Multiphoton-Ionization). Alternative and
complementary techniques were implemented for determining reactive integral and differ-
ential cross sections like ion-imaging, Doppler methods and the H-Rydberg atom TOF
spectroscopy. In addition, advances in the probing of the orientation of the products, by
exapole state-selection, by brute force, and by polarised laser radiation methods have en-
abled the investigation of stereodynamical aspects of chemical reactions at a fundamental
level.
3.2.1 Rotatable mass spectrometry

The main detection device able to measure Ilab (θ′ , u′) at different values of θ′ is the univer-
sal rotatable electron impact mass spectrometer. Rotatable detectors are usually made of
an (electron-impact or photo) ionizer, a quadrupole or magnetic sector mass filter, and an
ion counting device, such as a Daly detector or a secondary electron multiplier mounted
on a UHV chamber that can rotate around the collision region (10). In this way one can
plot the speed of the atomic or molecular products as a function of the scattering angle.
The contour maps of the speed flux of the reaction products can be regarded as an image
of the reaction in angle-velocity space. The kinetic energy distribution of one of the prod-
ucts of a bimolecular reaction measured by TOF spectroscopy provides also information
about the undetected product through conservation of energy and linear momentum (70).
UHV conditions are mandatory in the ionization and detection zones to measure fluxes as
low as 104 molecules cm−2 s−1 . This requires a special care in choosing materials, drawing
mechanical design, adopting UHV pumping (including differential pumping) techniques.
A detector of this type must have a reduced size and weight to be housed in a UHV
chamber rotated both in and out the plane. Optical detection can provide an alternative
(in favorable cases) to mass/TOF spectroscopy detection.
3.2.2 Photoionization
Electron-impact ionization has, however, some disadvantages like dissociative ionization
of the neutral reaction product (this is the most serious problem because fragmentation
in the ionizer can complicate the identification of reaction products especially when sev-
eral reaction channels are open at the same time), the high background partial pressure
produced by the hot filament used in the electron-impact source especially for H, H2 ,
CH4 , H2 O, CO, CO2 , energy and time limitations to the resolution induced by space-
charge effects. To the end of improving the signal-to-noise ratio magnetically suspended
turbomolecular pumps and efficient cryopumping (down to 10 K) of the electron-impact
ionization region have been introduced (71). To overcome the disadvantages associated
with the electron impact ionization method photo-ionisation with tunable radiation tech-
niques have been adopted. Dissociative ionization can be avoided by tuning the radiation
below the dissociative ionization potential of a molecule or radical, and so detection of a
species of mass m at the parent m/e mass ratio is possible. The residual background at
critical masses can be kept to a minimum because of the negligible amount of heat gener-
ated. The tunability of the radiation adds selectivity in the internal energy of the detected
species, in addition to universality. Further, no space charge effects accompany the pro-
cess, so by spatially focusing the photons it is possible to attain high time-resolution for
TOF measurements. The low ionization efficiency problem of this technique has been
recently solved using synchrotron radiation sources (72), although applications to reac-
tive scattering have been very limited, so far, for sensitivity reasons. Photoionization
detection using commercially available laser sources has also been reported
3.2.3 Laser spectroscopic detection

In CMB experiments two main spectroscopic schemes have been applied for product
detection: LIF and REMPI (recently, also, direct IR absorption has been used). The great
advantage of LIF and REMPI techniques is that they are both state-specific and sensitive.
Probing the reaction products at the collision region, has permitted the derivation of ICSs
for many elementary reactions. The main limitation of these two techniques is their lack
of generality. The availability of the proper light source is less of a problem nowadays
than the detailed knowledge of the spectroscopy of the molecule or radical to be probed
(detection of atomic products is usually straightforward). Predissociation in LIF, and the
availability of suitable schemes in REMPI, in addition to the properties of the available
tunable light sources, in practice restrict LIF and REMPI detection to small radicals and
molecules (35). Furthermore, probing the entire ro-vibrational distribution of a reaction

product is often difficult, if several vibrational levels are populated. Nevertheless, the
amount of detail obtained from these state-specific detection schemes is impressive, and
they have played an important role in the development of reaction dynamics during the
’80s and ’90s. Laser probing at the collision region, exploiting ion-imaging techniques
or Doppler profiles, represent an alternative as well as a complementary approach to the
CMB method with rotating detector and TOF analysis for measuring reactive DCS.
3.2.4 Ion-imaging and Doppler methods

Imaging techniques are an alternative modern approach to the simultaneous, in principle
state specific, measurement of angular and velocity distributions of reaction products
(73). Full three-dimensional velocity distributions can be recovered from the analysis
of the ion images. Using this technique some vibrational resolution was achieved (74).
Recently, the resolution of the ion-imaging method has been dramatically improved with
the introduction of the velocity mapping scheme (75; 76).
The technique of measuring state-resolved DCS by Doppler spectroscopy following op-
tical detection at the collision region by LIF of the scattered product was introduced by
Kinsey (77). An interesting feature of this technique is that LIF probe with sub-Doppler
resolution leaves encoded in the absorption spectral lines the angle-resolved velocity distri-
butions of the state resolved product. Unfortunately, resolution and sensitivity problems
due to low velocities and intrinsically lower signal levels do not allow a widespread appli-
cation of Doppler techniques to reactive scattering though significant advances have been
made (28; 62; 78).
A significant step ahead in the development of methods for measuring double DCSs is
represented by the Doppler-selected TOF technique implemented by Liu and coworkers,
and used to map out the three-dimensional velocity distribution of a reaction product,
directly in the center-of-mass system. This technique combines the conventional Doppler-
shift and ion TOF in an innovative manner, such that the net performance becomes more
than additively enhanced.
4 Some case studies

In order to compare experiments with calculations and carry out their physical interpre-
tation, one needs to express the measured Ilab (θ′ , u′) in the cm coordinates Θ′ and v ′ via
the relationship
X u′2
Ilab (θ′ , u′ ) = f (Etr )fi (ur1 , ur2, γ) Icm (Θ′ , vi′ ). (2)
i
v ′2
In eq. 2 f (Etr ) is the function expressing the dependence of the cross section from
the collision energy Etr and fi (ur1, ur2 , γ) is the weight of the ith Newton diagram (the
composition diagram of the velocities of the bi-molecular collision partners under the
conservation laws of closed systems (79)) depending on the velocities of reagents r1 and
r2 and the related collision angle γ. Then for any given set of reactant initial parameters
(including when is the case orientation and alignment) which for sake of simplicity have
been dropped from the notation, the value of the center of mass flux Icm (Θ′ , v ′ ) the product
flux is formulated in terms of its energy dependent analogous as follows
R∞
′ ′ dE ′ ′ ′ o
Icm (Θ′ , E ′ )dE ′
Icm (Θ , v ) = Icm (Θ , E ) = R ∞ ′ I (Θ′ , E ′ )
′ , E ′ )dv ′ cm
(3)
du′ o
v Icm (Θ
Then Icm (Θ′ , v ′ ) is linked to the differential cross section σ(Θ′ , Etr
′
) (that can be directly
obtained by a proper combination of the elements of the scattering S matrix of the theory
′
see ref. (1) and form which the integral cross section σ(Etr ) can be derived by integrating
′
over Θ ) using the relationship
Icm (Θ′ , E ′ ) = nr1 nr2 utr σ(Θ′ , E ′ )∆V J(Θ′ )F ∆Ω′ (4)
where nr1 and nr2 are the densities of the first and second reagent, respectively, utr is
the relative collision velocity of the reactant species, σ(Θ′ , E ′ ) is the differential cm cross
section for the given initial state (or distribution of initial states) whose labels have been,
however, dropped, ∆V is the collision volume, J(Θ) is the Jacobian of the cm → lab
transformation, F is the detection efficiency, ∆Ω′ is the detection solid angle.
However, eq. 4 cannot be inverted and different sets of cross section values may satisfy
equation 4 and ultimately the measured scattering data. For this reason the problem is
usually tackled using forward convolution techniques. This means that cm angular and
velocity distributions are given a functional representation whose parameters are varied
to best fit the measured laboratory distributions (obviously, in doing so one has to take
into account the velocity spread of the beams and the resolution of the TOF (that is the
ratio between the length of the ionization zone and that of the flight length)).
4.1 The atom-diatom reactions

Atom-diatoms are the systems which have attracted most of the work comparing the
results of theoretical and experimental investigations from the very beginning of CMB
studies. In particular for atom-diatom systems some examples of comparison of state
averaged, state resolved and state-to-state experimental ICSs and DCSs with both QM
and QCT estimates are given in the literature. As a matter of fact, as already mentioned,
the comparison of dynamical calculations with CMB data is, indeed, the most important
probe of the accuracy of a proposed PES. The rigorous QM approach, however, becomes
progressively heavier as the mass of the system and the structure of the PES increases.
In fact, one needs to achieve convergence of the calculated cross sections with the value of
the total angular momentum quantum number and with the size of the basis set (or grid
points). For QCT calculations there are fewer difficulties in achieving convergence because
it depends only on the number of trajectories integrated. Moreover, while trajectories are
independent computational tasks which can be efficiently distributed for calculation on
the grid (80), this is not so for QM calculations.
4.1.1 The H + H2 reaction

The hydrogen exchange reaction H + H2 → H2 + H is the lightest and simplest neutral
particles reaction. This reaction has been extensively studied, both theoretically and
10B - 10 RTO-EN-AVT-142
experimentally, since the 1930s (see refs. (81; 82) for a review of older work). As a
matter of fact the first accurate calculation of an ab initio PES (LSTH) (83; 84) and the
first quantum (QM) calculation of reactive scattering properties (85) were carried out for
H+H2 . More recently, it has become possible to make an accurate comparison between
experiment and theory.
On the experimental side product-state and angular distributions have been measured
(82; 86; 87) and DCS have been determined using the photoloc technique for H + D2
leading to HD(ν=1, j=1,5,8) + D at Etr =1.7 eV and to HD(ν=2, j=0,3,5) at Etr =1.5
eV (88; 89). Extensive and detailed top CMB experiments were carried out by Welge
and co-workers (74; 90; 91; 92; 93; 94). In these experiments state-to-state (rotationally
resolved) DCS for the reaction H + D2 (ν=0,j=0) → HD(ν ′ , j ′ ) + D were measured for a
wide range of collision energies (from 0.53 eV up to 2.67 eV) using the ultra-high-resolution
H-Rydberg atom TOF spectroscopy method. These results are the most comprehensive
and accurate experimental data ever measured for a chemical reaction and have enabled
a very detailed assessment of the quantum mechanical treatment of chemical elementary
reactions.
The above listed properties of the H + H2 reaction have been the subject of detailed
comparisons with both QCT and QM scattering calculations (74; 95). In particular,
state-to-state DCS at collision energies of 0.52, 0.531 and 0.54 eV have been compared
with converged QM scattering calculations on the different ab initio PESs available from
the literature for this system. This has allowed a direct assessment of their quality (96).
Differences found in the QM calculations were traced back to slight differences in the
topology (barrier height and bending potential) of the different PESs used in the calcu-
lations. The comparison showed that BKMP2 (97) is at present the PES most suitable
to describe the reactive scattering process at low collision energy. At the higher value of
Etr =1.29 eV, the older and more popular LSTH PES still provides a good description of
the experimental state-to-state DCS. A comparison of QM and QCT (though performed
only on the BKMP2 PES) calculations showed that quasiclassical methods do a good job
in reproducing the experimental results, except for the low collision energy rotationally
state-to-state DCSs (98).
Table 1 H + H2 LSTH REACTIVE T = 300 K

′
v v 1 3 5 7 9 11
0 .329(-13) .248(-11) .592(-11) .587(-11) .453(-11) .156(-11)
1 .118(-12) .289(-11) .814(-11) .762(-11) .471(-11) .339(-11)
2 .638(-11) .110(-10) .957(-11) .708(-11) .409(-11)
3 .560(-11) .177(-10) .140(-10) .107(-10) .579(-11)
4 .327(-10) .244(-10) .160(-10) .858(-11)
5 .289(-10) .396(-10) .228(-10) .131(-10)
6 .685(-10) .397(-10) .214(-10)
7 .592(-10) .648(-10) .292(-10)
8 .285(-13) .112(-09) .551(-10)
9 .997(-10) .964(-10)
10 .887(-13) .184(-09)
11 .168(-09)
12 .958(-12)
RTO-EN-AVT-142 10B - 11
Table 2 H + H2 LSTH REACTIVE T = 500 K

v′ v 1 3 5 7 9 11
0 .684(-12) .568(-11) .993(-11) .109(-10) .726(-11) .359(-11)
1 .992(-12) .775(-11) .119(-10) .115(-10) .710(-11) .527(-11)
2 .148(-10) .181(-10) .174(-10) .109(-10) .586(-11)
3 .150(-10) .305(-10) .234(-10) .150(-10) .966(-11)
4 .670(-14) .529(-10) .418(-10) .243(-10) .129(-10)
5 .473(-10) .581(-10) .364(-10) .211(-10)
6 .984(-13) .100(-09) .575(-10) .329(-10)
7 .847(-10) .941(-10) .500(-10)
8 .535(-12) .162(-09) .865(-10)
9 .126(-09) .144(-09)
10 .208(-11) .240(-09)
11 .199(-09)
12 .111(-10)
13 .360(-12)
Table 3 H + H2 LSTH REACTIVE T =1000 K

′
v v 1 3 5 7 9 11
0 .472(-11) .152(-10) .190(-10) .152(-10) .130(-10) .551(-11)
1 .852(-11) .186(-10) .226(-10) .219(-10) .131(-10) .762(-11)
2 .248(-12) .400(-10) .312(-10) .270(-10) .190(-10) .129(-10)
3 .408(-10) .567(-10) .411(-10) .291(-10) .164(-10)
4 .123(-11) .970(-10) .642(-10) .466(-10) .233(-10)
5 .846(-10) .102(-09) .629(-10) .418(-10)
6 .342(-11) .153(-09) .103(-09) .623(-10)
7 .887(-13) .131(-09) .155(-09) .846(-10)
8 .760(-11) .198(-09) .144(-09)
9 .295(-12) .177(-09) .221(-09)
10 .204(-10) .290(-09)
11 .197(-11) .250(-09)
12 .792(-13) .514(-10)
13 .930(-14) .752(-11)
14 .371(-12)
These conclusions encouraged us to carry out an extensive campaign of rate coefficients

calculations for the H + H2 reaction by running trajectory calculations on the LSTH PES.
Calculated rate coefficients for reactive transitions are reported in the tables 1 - 5
for increasing values of the temperature T . In the tables values of the rate coefficients
are given in units of 10x cm3 molec−1 s−1 with x being given in the round brackets. Initial
vibrational states ν are given in the first row while final states ν ′ are reported in the first
column.
10B - 12 RTO-EN-AVT-142

v′ v 1 3 5 7 9 11
0 .237(-10) .279(-10) .334(-10) .286(-10) .211(-10) .973(-11)
1 .403(-10) .465(-10) .436(-10) .339(-10) .272(-10) .126(-10)
2 .433(-11) .853(-10) .633(-10) .510(-10) .374(-10) .217(-10)
3 .265(-12) .912(-10) .102(-09) .759(-10) .572(-10) .332(-10)
4 .888(-14) .136(-10) .162(-09) .114(-09) .775(-10) .446(-10)
5 .131(-11) .132(-09) .163(-09) .117(-09) .645(-10)
6 .112(-12) .252(-10) .209(-09) .159(-09) .963(-10)
7 .191(-13) .320(-11) .166(-09) .225(-09) .142(-09)
8 .273(-12) .492(-10) .271(-09) .199(-09)
9 .197(-12) .922(-11) .196(-09) .257(-09)
10 .150(-11) .728(-10) .299(-09)
11 .242(-12) .202(-10) .259(-09)
12 .554(-13) .424(-11) .900(-10)
13 .167(-13) .935(-12) .242(-10)
14 .729(-13) .275(-11)

v′ v 1 3 5 7 9 11
0 .717(-10) .687(-10) .509(-10) .436(-10) .317(-10) .203(-10)
1 .115(-09) .934(-10) .800(-10) .582(-10) .492(-10) .285(-10)
2 .116(-09) .135(-09) .116(-09) .912(-10) .703(-10) .438(-10)
3 .462(-10) .183(-09) .152(-09) .122(-09) .948(-10) .558(-10)
4 .130(-10) .151(-09) .198(-09) .156(-09) .122(-09) .699(-10)
5 .396(-11) .655(-10) .215(-09) .208(-09) .168(-09) .925(-10)
6 .105(-11) .238(-10) .186(-09) .232(-09) .214(-09) .122(-09)
7 .547(-12) .915(-11) .835(-10) .248(-09) .243(-09) .161(-09)
8 .612(-13) .246(-11) .324(-10) .209(-09) .261(-09) .196(-09)
9 .291(-13) .100(-11) .125(-10) .106(-09) .209(-09) .217(-09)
10 .280(-13) .125(-12) .507(-11) .447(-10) .102(-09) .230(-09)
11 .000(-00) .174(-12) .159(-11) .206(-10) .489(-10) .210(-09)
12 .299(-13) .888(-13) .693(-12) .493(-11) .151(-10) .927(-10)
13 .316(-13) .200(-12) .121(-11) .381(-11) .299(-10)
14 .202(-12) .106(-11) .462(-11)
4.1.2 The Li + HF reaction
In the realm of three-atom systems, the prototypal role of truly A + BC asymmetric

reactions is played by Li + HF → LiF + H. After the early CMB investigation of Lee and
coworkers (99) in more recent years further experiments have been carried out involving
also HF vibrational excitation and orientation (100; 101). For this system extensive ab
initio calculations have been reported in ref. (102) and further calculations have been
discussed in ref. (103). On these ab initio values a PES was fit whose parameters,
functional formulations and results obtained from dynamical calculations are discussed in
RTO-EN-AVT-142 10B - 13
refs. (104; 105; 106). Here, we plot in figure 1 the QM reactive probability calculated
at ν = 0 and j = 0 at different values of the total angular momentum quantum number
J. Similar calculations performed on a different PES are given in refs. (107; 108; 109).
The situation for this system, as is for many other atom-diatom and, a fortiori, for the
diatom-diatom systems we discuss in the followings, is still that of assessing the accuracy
of the proposed PESs.
0.3
0.25 J=10 J=35
0.2
0.15
0.1
0.05
0
0.25 J=5 J=30
0.2
0.15
0.1
0.05
0
0.25 J=2 J=25
0.2
P00(E)
0.15
J
0.1
0.05
0
0.25 J=1 J=20
0.2
0.15
0.1
0.05
0
0.25 J=0 J=15
0.2
0.15
0.1
0.05
0
0.2 0.3 0.4 0.5 0.6 0.2 0.3 0.4 0.5 0.6 0.7
E / eV
Figure 1: QM reaction probabilities calculated for the Li + FH reaction at ν = 0 and j

= 0 for different values of the total angular momentum quantum number J plotted as a
function of the total energy E.
4.2 Polyatomic reactions

Reactions involving more than three-atoms can be divided into two classes: atom poly-
atomic molecules processes and free radicals (diatomic or polyatomic) molecules (diatomic
or polyatomic) ones. While there has been a number of CMB studies of atom-polyatomic
molecule reactions much fewer have been those concerned with free radical molecule re-
actions.
In the past, reactive DCS for radical-molecule reactions have been measured using
effusive beams of CH3 and OH radicals (110; 111; 112) and using mildly supersonic CH3
beams generated by pyrolysis (113; 114). More recently, DCS radical-molecule reactions
have been measured for reactions of OH (with H2 and CO) and CN (with H2 ) by mass
10B - 14 RTO-EN-AVT-142
spectrometric/TOF and Doppler selected/TOF detection, respectively. ICS resolved over

the final products states were also determined for CH (17) NH, NH2 (115) and CN (116)
using laser spectroscopy probing the collision region. CMB studies were also carried out
for CN reactions with unsaturated hydrocarbons and polyatomic radical (C2 D, C3 H3 ,
C6 H5 ).
4.2.1 The OH + H2 reaction

The development of a continuous, intense supersonic beam of OH radicals has been the
key to the first successful measurement of product angular and velocity distributions for
the OH + H2 → H2 O + H reaction
DCS of the 18 OH + D2 reaction were measured by Casavecchia and collaborators (118;
119) using a crossed beam apparatus to detect the HOD product. The product cm angular
and translational energy distributions obtained from the experiment were measured at a
collision energy of 6.3 kcal/mol. The experiment indicates that the angular distribution
of the HOD product is strongly backward (with respect to the OH incoming direction).
The authors attributed this to the rebound nature of the reactive mechanism. They also
found that the average fraction of energy available as translation of the products is only
0.32 and that, accordingly, the percentage of internal energy is high. More recently, Daris
and coworkes have obtained vibrationally resolved DCS for the same reaction by using
the H-atom Rydberg tagging techniche citecasa-science. Experimental kinetic studies of
this reaction have also been performed (121; 122; 123; 124; 125; 126).
1.0
OC-QCT
0.8 SE-QCT
0.6
P(E’tr)
0.4
0.2
0.0
0 5 10 15 20
-1
E’tr /kcal mol
Figure 2: Product translational distribution for the 18 OH + D2 → H18 OD + D reaction

evaluated at Etr =6.3 kcal/mol. Experimental values from ref. (119) are given as a shaded
area, QCT values calculated on the OC PES are given as a dashed line and QCT values
calculated on the SE PES are given as a dashed-dotted line.
Here, results of highly converged QCT calculations are compared with the experimen-
tal data (118; 119). The PES mostly widely used for dynamical calculations of the reactive
RTO-EN-AVT-142 10B - 15
properties of the OH + H2 system is the one proposed by Schatz and Elgersma (SE) (127)
which is based on the ab initio calculations of Walch and Dunning (128). The SE PES
adopts an MBE (Many-Body Expansion) formulation of the interaction and expresses the
two body terms as Morse functions, the three body terms as LEPS and quadratic Sorbie
Murrell functionals, and the four body term as a sum of two exponentials of polynomials
of the related internuclear distances.
A more recent fit of the PES (129) based to the ab initio energies of Werner and
coworkers was also performed using an extension to four atoms of the LAGROBO (Largest
Angle Generalization of ROtating Bond Order) many process functional representation
of the interaction (130; 131; 132) by Ochoa and Clary (OC).
CMB results of the 18 OH + D2 → H18 OD + D reaction allow a comparison of calculated
cm product translational energy (PTD) distributions obtained by plotting the reactive
′ ′
probability P (Etr ) as a function of Etr with measured data. To this end half million
trajectories were integrated on both the SE and the OC PESs at Etr = 6.3 kcal/mol,
ν = 0, jOH = 0 and jD2 = 2 to mimic experimental conditions (119). The different
properties of the two PESs show up clearly in the PTD structures plotted in Fig. 2.
′
The PTD calculated on the OC PES has, in fact, its maximum at Etr = 3 kcal/mol
′
(dashed line) while the one calculated on the SE PES has its maximum at Etr = 11
kcal/mol (dashed dotted line). As apparent from the figure the measured distribution
(given as the dashed area of the experimental error bars) has an intermediate position
′
with the maximum located at Etr = 5.5 kcal/mol (2.5 kcal/mol above that of OC results
and 5.5 kcal/mol below that of the SE ones). Other differences between calculated and
experimental properties are the lower value of the theoretical cross sections (with respect
to the measured ones). This means that also the OC PES is still far from having the
accuracy necessary for use in rigorous dynamical calculations. As a matter of fact new ab
initio calculations have been performed and their fit based on a Shepard intepolation has
been proposed (133). Though the new surface does not seems yet to behave properly in
the long range region it marks a definite improvement over the previous ones and therefore
deserves to be tested by running massive trajectory calculations.
4.2.2 The OH + CO reaction

Extensive experimental investigations (134; 135; 136; 137; 138; 139; 140; 141; 142; 143;
144; 23; 22; 145; 146; 147; 148) have been carried out for the OH + CO → H + CO2
reaction. As to theory, also several calculations (149; 150; 151; 152; 153; 154; 155; 156;
157; 158; 159; 160; 161; 162; 163; 164; 165) have been carried out. Calculations and
experiments have also been carried out for the reverse H + CO2 → OH + CO reaction
(166; 167; 168; 169; 170; 171; 172; 173; 174) as well as for related isotopic variants.
Previous work is reviewed in ref. (117) and a great deal of research is still reported on
these systems at the top international conferences in the field (see for recent examples
refs. (175; 176)).
The first full-dimensional PES for the OH + CO system based on the many-body
expansion approach (177) was proposed by Schatz and coworkers (166). The two and
three body terms consist of polynomials in the related internuclear distances multiplied
by damping functions to make the potentials vanish at large distances. The four body
term consists of products of polynomials (of the same kind used for the two and three
10B - 16 RTO-EN-AVT-142
body terms) and Gaussians enforcing the reproduction of the ab initio stationary points.
Subsequently the PES was further improved (149; 167) in order to better reproduce the
ab initio calculations of ref. (166) and to smooth out singularities occurring during the
numerical evaluation of the potential energy and its derivatives.
Recently, using the potential energy values obtained from a new set of ab initio calcu-
lations, a more accurate evaluation of the stationary points of the surface was carried out
and a new fit (YMS) was performed (178; 179; 180). More recently, the PES of Schatz and
coworkers was also further modified (159) and the new PES (LTSH) is in better agreement
with the most recent ab initio values and avoids some of the singularities occurring during
the calculation of the derivatives. To this end the parameters of the four body term and
the coefficients of the long range part of the OH + CO entrance channel were optimized
to reproduce the ab initio values of refs. (178; 179; 180). To study the influence on the
dynamics of the wells located in the reactant channel also a modified version of the LTSH
PES (called mod-LTSH) was produced (165) by suppressing the long range tail of the
surface in the reactant channel.
To work out a theoretical estimate of the PTD at the conditions of the experiment, we
performed QCT calculations on both the YMS and LTSH PESs by setting the collision
energy at 8.6 and 14.1 kcal mol−1 and the vibrational energy of both the OH and CO
molecules at that of their ground rovibrational state. Batches of more than one million
trajectories were run using a maximum impact parameter of 2.6 and 2.4 Å for the YMS
and the LTSH PESs respectively and accepting only trajectories conserving total energy
to better than 4 × 10−2 kcal/mol.
The PTDs calculated on the YMS PES have a location of the maximum depending on
the reactant collision energy (that is at Etr′ = 19 and 21 kcal mol−1 for reactant collision
energies of 8.6 and 14.1 kcal mol−1 , respectively). On the contrary the PTDs calculated
on the LTSH PES at the same collision energies have both maxima at Etr′ = 23 kcal
mol−1 . The PTDs are basically symmetric (the average product translational energies is
coincident with the value at the maxima) with the distributions calculated at Etr = 14.1
kcal mol−1 being slightly broader than the ones calculated at Etr = 8.6 kcal mol−1 . An
even more dramatic difference between the two sets of PTDs is that those calculated
on the LTSH PES are systematically closer to the experimental distribution than those
calculated on the YMS one. However, none of the calculated PTDs coincides with the
experimental one.
These results and the need for rejecting a large fraction of trajectories not conserving
total energy within the imposed boundaries are a clear indication that the proposed
PESs are still too far from the quality standards required for carrying out dynamical
calculations.
RTO-EN-AVT-142 10B - 17
References
[1] Laganà A Theory and computing of gas phase chemical reactions: from
exact quantum to approximate dynamical treatments, preceeding paper in
this book.
[2] Taylor E H and Datz S 1955 J. Chem. Phys. 23 1711-8
[3] London F 1929 Z. Elektrochem. 35 552
[4] Eyring H and Polanyi M 1931 Z. Physik. Chem. B12 279
[5] Hirschfelder J O, Eyring H and Topley B 1936 J. Chem. Phys. 4 170
[6] Porter R A R, Brown G R and Grosser A E 1979 Chem. Phys. Lett. 61 313-4
[7] Ross J ed 1966 Molecular Beams. Adv. Chem. Phys. (New York: Wiley) vol
10
[8] Herschbach D R 1966 Adv. Chem. Phys. 10 319-93
[9] Polanyi J C 1972 Acc. Chem. Res. 5 161-8
[10] Lee Y T, McDonald J D, LeBreton P R and Herschbach D R 1969 Rev. Sci.

Instrum. 40 1402
[11] Lee Y T 1987 Science 236 793-8
[12] Levine R D and Bernstein R B 1987 Molecular Reaction Dynamics (New

York: Oxford University Press)
[13] Levine R D 2005 Molecular Reaction Dynamics and Chemical Reactivity

(Cambridge University Press)
[14] Farrar J M and Lee Y T 1974 Annu. Rev. Phys. Chem. 25 357
[15] Levy M R 1979 Prog. React. Kinet. 10 1-252
[16] Gonzales Urena A 1987 Adv. Chem. Phys. 66 213-335
[17] Liu K, Macdonald R G and Wagner A F 1990 Int. Rev. Phys. Chem. 9
187-225
[18] Neumark D M 1992 Annu. Rev. Phys. Chem. 43 153-76
[19] Orr-Ewing A J and Zare R N 1994 Annu. Rev. Phys. Chem. 45 315-66
[20] Orr-Ewing A J and Zare R N 1995 The Chemical Dynamics and Kinetics of
Small Radicals ed K Liu and A F Wagner (Singapore: World Scientific) pp
936-1063
[21] Casavecchia P, Balucani N, Volpi G G 1993 Research in Chemical Kinetics

ed RG Compton and G Hancock Vol 1 (Amsterdam: Elsevier) pp 1-63
10B - 18 RTO-EN-AVT-142
[22] Casavecchia P, Balucani N and Volpi G G 1995 The Chemical Dynamics

and Kinetics of Small Radicals ed K Liu and A F Wagner (Singapore: World
Scientific) pp 365-437
[23] Alagia M, Balucani N, Casavecchia P, Stranges D and Volpi G G 1995 J.

Chem. Soc. Faraday Trans. 91 575-596
[24] Grice R 1995 Int. Rev. Phys. Chem. 14 315-26
[25] Heck A J R and Chandler D W 1995 Annu. Rev. Phys. Chem. 46 335-72
[26] Loesch H J 1995 Annu. Rev. Phys. Chem. 46 555-94
[27] Brouard M and Simons J P 1995 The Chemical Dynamics and Kinetics of
Small Radicals ed K Liu and A F Wagner (Singapore: World Scientific) pp
795-841
[28] Mestdagh J M, Visticot J P and Suits A G 1995 The Chemical Dynamics

and Kinetics of Small Radicals ed K Liu and A F Wagner (Singapore: World
Scientific) pp 668-729
[29] Polanyi J C and Zewail A H 1995 Acc. Chem. Res. 28 119-32
[30] Gonzales-Urena A and Vetter R 1995 J. Chem. Soc. Faraday Trans. 91 389
- 98
[31] Gonzales-Urena A and Vetter R 1996 Int. Rev. Phys. Chem. 15 375-427
[32] Houston P L 1996 J. Phys. Chem. A 100 12757-70
[33] Crim F F 1996 J. Phys. Chem. 100 12725-34
[34] Moore C B and Smith I W M 1996 J. Phys. Chem. A 100 12848-65
[35] Whitehead J C 1996 Rep. Progr. Phys. 59 993-1040
[36] Loomis R A and Lester M I 1997 Annu. Rev. Phys. Chem. 48 643-73
[37] Alexander M H, Werner H J and Manolopoulos D E 1998 J. Chem. Phys.

109 5710-3
[38] Zewail A H 1997 Adv. Chem. Phys. 101 3-43
[39] Zhong D and Zewail A H 1998 J. Phys. Chem. A 102 4031-58
[40] Zare R N 1998 Science 279 1875-9
[41] Casavecchia P. 2000 Rep. Progr. Phys. 63 355-414.
[42] Whitehead J C 1983 Comprehensive Chemical Kinetics Vol 24: Modern

Methods in Kinetics ed C H Bamford and C F H Tipper (Amsterdam: El-
sevier) pp 357-506
RTO-EN-AVT-142 10B - 19
[43] Gentry R 1987 Atomic and Molecular Beam Methods vol 1 ed G Scoles (New
York: Oxford University Press) pp 54-82
[44] Davis H F, Suits A G and Lee Y T 1992 Gas-Phase Metal Reactions ed A

Fontijin (Amsterdam: Elsevier) pp 319-47
[45] Costes M, Naulin C, Dorthe G, Daleau G, Joussot-Dubien J, Lalaude

C,Vinckert M, Destor A, Vaucamps C and Nouchi G 1989 J. Phys. E 22
1017-23
[46] Willis P A, Stauffer H U, Hinrichs R Z and Davis H F 1999 Rev. Sci. Instrum.
70 2606-14
[47] Casavecchia P, Balucani N, Cartechini L, Capozza G and Bergeat A 2001

Faraday Discuss. 119, 27. on Free Radicals (Flagstaff, USA) p 42
[48] Kaiser R I, Sun W, Suits A G and Lee Y T 1997 J. Chem. Phys. 107 8713-6
[49] Kaiser et al 1999 J. Chem. Phys. 110 10330-43
[50] Davis H F, Anderson D T, Duxbury G and Nesbitt D J 1997 J. Chem. Phys.

107 5661-75
[51] Costes M and Naulin C 1998 C.R. Acad. Sci. Paris, Serie IIc 771-5
[52] Lee S H and Liu K 1998 J. Phys. Chem. A 102 8637-40
[53] Lee S H and Liu K 1998 Chem. Phys. Lett. 290 323-8
[54] Lee S H and Liu K 1999 J. Chem. Phys. 111 4351-2
[55] Lee S H and Liu K 1999 J. Chem. Phys. 111 6253-9
[56] Hsu Y T, Wang J H and Liu K 1997 J. Chem. Phys. 107 2351-6
[57] Lee S H, Lai L H, Liu K and Chang H 1999 J. Chem. Phys. 110 8229-32
[58] Zhang R, van der Zande W J, Bronikowski M J and Zare R N 1991 J. Chem.
Phys. 94 2704-12
[59] Sinha A, Hsiao M C and Crim F F 1990 J. Chem. Phys. 92 6333-5
[60] Bronikowski M J, Simpson W R and Zare R N 1993 J. Phys. Chem. 97

2194-203; 2204-8
[61] Mestdagh J M, Balko B A, Covinsky M H, Weiss P S, Vernon M F, Schmidt

and Lee Y T 1987 Faraday Discuss. 84 145
[62] LHermite J M, Ramat G and Vetter R 1990 J. Chem. Phys. 93 434-44
[63] Willis P A, Stauffer H U, Hinrichs R Z and Davis H F 1998 J. Chem. Phys.

108 2665-8
10B - 20 RTO-EN-AVT-142
[64] Alagia M, Aquilanti V, Ascenzi D, Balucani N, Cappelletti D, Cartechini L,

Casavecchia P, Pirani F, Sanchini G and Volpi G G 1997 Israel J. Chem. 37
329-42
[65] Bergeat A, Cartechini L, Casavecchia P and Volpi G G 1999 Proc. XVI Int.
Conf. on Molecular Energy Transfer (Assisi, Italy) p 51
[66] Casavecchia P et al 1999 Faraday Discuss. 113 206-8
[67] Nizkorodov S A, Harper W W and Nesbitt D J 1999 Faraday Discuss. 113

107-17
[68] Naulin C and Costes M 1999 Chem. Phys. Lett. 310 231-9
[69] Zare R N and Dagdigian P J 1974 Science 185 739-47
[70] Lee Y T 1987 Atomic and Molecular Beam Methods vol 1 ed G Scoles (New
York: Oxford University Press) pp 553-68
[71] Lin J J, Hwang D W, Harich S, Lee Y T and Yang X 1998 Rev. Sci. Instrum.
69 642-46
[72] Yang X, Lin J, Lee Y T, Blank D A, Suits A G and Wodtke A M 1997 Rev.
Sci. Instrum. 68 3317-26
[73] Chandler D W and Houston P L 1987 J. Chem. Phys. 87 1445-
[74] Wrede E and Schnieder L 1997 J. Chem. Phys. 107 786-90
[75] Parker D H and Eppink A T J B 1997 Rev. Sci. Instrum. 68 3477
[76] Chang B Y, Hoetzlein R C, Mueller J A, Geiser J D and Houston P L 1998

Rev. Sci. Instrum. 69 1665-70
[77] Kinsey J L 1977 Annu. Rev. Phys. Chem. 28 349
[78] Girard B, Billy N, Gonedard G and Vigu J 1991 J. Chem. Phys. 95 4056-69
[79] Bernstein R B 1982 Chemical Dynamics via Molecular Beam and Laser Tech-
niques (New York: Oxford University Press)
[80] First Egee Application Migration Report, EGEE-

DNA4.3.1.523422-v4-7-13.doc (2005); Egee review workshop,
(http://agenda.cern.ch./fullAgenda.php?ida=a043803) Geneva, 2005.
[81] Valentini J J and Phillips D L 1989 Advances in Gas Phase Photochem-

istry ed M N R Ashfold and J E Baggot Vol 2 (London: Royal Society of
Chemistry) pp 1-51
[82] Buchenau H, Toennies J P, Arnold J and Wolfrum J 1990 Ber. Bunsenges.

Phys. Chem. 94 1231-48
RTO-EN-AVT-142 10B - 21
[83] Siegbahn P and Liu B 1978 J. Chem. Phys. 68 2457-65
[84] Truhlar D G and Horowitz C J 1978 J. Chem. Phys. 68 2466-76; 1979 ibid
71, 1514E
[85] Schatz G C and Kuppermann A 1976 J. Chem. Phys. 65 4642-92
[86] Buntin S A, Giese C F and Gentry W R 1987 J. Chem Phys. 87 1443-5
[87] Continetti R E, Balko B and Lee Y T 1990 J. Chem Phys. 93 5719-40
[88] Fernandez-Alonso F, Bean B D and Zare R N 1999 J. Chem. Phys. 111

1035-42
[89] Fernandez-Alonso F, Bean B D and Zare R N 1999 J. Chem. Phys. 111

2490-8
[90] Schnieder L, Seekamp-Rahn K, Borkowski J, Wrede E, Welge K H, Aoiz F

J, Banares L, DMello M J, Herrero V J, Saez Rabanos V and Wyatt R E
1995 Science 269 207-10
[91] Schnieder L, Seekamp-Rahn K, Wrede E and Welge K H 1997 J. Chem. Phys.

107 6175-95
[92] Wrede E, Schnieder L, Welge K H, Aoiz F J, Banares L and Herrero V J

1997 Chem. Phys. Lett. 265 129-36
[93] Wrede E, Schnieder L, Welge K H, Aoiz F J, Banares L, Herrero V J,

Martnez-Haya B and Saez Rabanos V 1997 J. Chem. Phys. 106 7862-4
[94] Wrede E, Schnieder L, Welge K H, Aoiz F J, Banares L, Castillo J F,

Martnez-Haya B and Herrero V J 1999 J. Chem. Phys. 110 9971-81
[95] Wu Y S M and Kuppermann A 1995 Chem. Phys. Lett. 235 105-10
[96] Banares L, Aoiz F J, Herrero V J, DMello M J, Niederjohann B, Seekamp-

Rahn K, Wrede E and Schnieder L, 1998 J. Chem. Phys. 108 6160-9
[97] Boothroyd A I, Keogh W J, Martin P G and Peterson M R 1996 J. Chem.

Phys. 104
[98] Aoiz F J, Banares L, Herrero V J, Saez-Rabanos V, Stark K and Werner H

J 1994 J. Phys. Chem. 98 10665-70 7139-52
[99] Becker C H, Casavecchia P, Tiedemann P W, Valentini J J and Lee Y T

1980 J. Chem. Phys. 73 2833-50
[100] Loesch H J and Stienkemeier F 1993 J. Chem. Phys. 98 9570-84
[101] Hobel O., Menendez M., Loesch H J 2001 J Phys. Chem. Chem. Phys. 3(17)
3633-3637
10B - 22 RTO-EN-AVT-142
[102] Chen M M L and Schaefer H F III 1980 J. Chem. Phys. 72 4376

[103] Palmieri P and Laganà A, 1989 J. Chem. Phys. 91 7302
[104] Parker G A, Laganà A, Crocchianti S and Pack R T 1995 J. Chem. Phys.
102 1238-50
[105] Laganà A, Bolloni A, Crocchianti S and Parker G A, 2000 Chem. Phys.
Letters 324 466-474
[106] Piermarini V., Crocchianti S, Laganà A, 2002 J. Comp. Meth. in Science and
Eng. 2 361-367
[107] Aguado A, Paniagua M, Lara M and Roncero O 1997 J. Chem . Phys. 107
10085-95
[108] Lara M, Aguado A, Roncero O, Paniagua M 1998 J. Chem . Phys. 109 9391-
400
[109] Aoiz F J, Martinez M.T., Saez-Rabanos V, Verdasco E 1999 Chem. Phys.
Letters 299 25-34
[110] Carter C F, Levy M R and Grice R 1972 Chem. Phys. Lett. 17 414-5
[111] McFadden D L, McCullough E A, Kalos F and Ross J 1973 J. Chem . Phys.
57 1351-2
[112] Logan J A, Mims C A, Stewart G W and Ross J 1976 J. Chem. Phys. 64
1804-15
[113] Hoffmann M A, Smith D J, Bradshaw N and Grice R 1986 Mol. Phys. 57
1219-31
[114] Robinson G N, Nathanson G M, Continetti R E and Lee Y T 1988 J. Chem.
Phys. 89 6744-52
[115] Patel-Misra D and Dagdigian PJ 1992 J. Phys. Chem. 96 3232-6
[116] Sauder D G, Patel-Misra D and Dagdigian P J 1991 J. Chem. Phys. 95
1696-707
[117] J. M. Bowman, G. C. Schatz, Annu. Rev. Phys. Chem. 46, 169 (1995).
[118] M. Alagia, N. Balucani, P. Casavecchia, D. Stranges, G. G. Volpi, J. Chem.
Phys. 98 (1993) 2459.
[119] M. Alagia, N. Balucani, P. Casavecchia, D. Stranges, G. G. Volpi, D. C.
Clary, A. Kliesh, H.-J. Werner, Chem. Phys. 207 (1996) 389.
[120] B. R. Strazisar, C. Lin, and H. F. Davis, Science 290, (2000) 958.
[121] see for example, J. Warnatz, Sandia National Laboratories Report No.
SAND83-8606, Livermore, 1983.
RTO-EN-AVT-142 10B - 23
[122] F. P. Tully, A. R. Ravishankara, J. Phys. Chem. 84 (1980) 3126.
[123] A. R. Ravishankara, J. M. Nicovich, R. L. Thompson, F. P. Tully, J. Phys.

Chem. 85, (1981) 2498.
[124] R. K. Talukdar, T. Gierezak, L. Goldfarb, Y. Rudich, B. S. Madhava Rao,

A. R. Ravishankara, J. Phys. Chem. 100 (1996) 3037.
[125] R. Zellner, W. Steinert, Chem. Phys. Lett. 81 (1981) 568.
[126] G. P. Glass, B. K. Chaturvedi, J. Chem. Phys. 75 (1981) 2749.
[127] G. C. Schatz, H. Elgersma, Chem. Phys. Lett. 73 (1980) 21.
[128] S. P. Walch, T. H. Dunning, Jr., J. Chem. Phys. 72 (1980) 1303.
[129] G. Ochoa de Aspuru, D. C. Clary, J. Phys. Chem. A 102 (1998) 9631.
[130] A. Laganà, J. Chem. Phys. 95 (1991) 2216.
[131] A. Laganà, G. Ferraro, E. Garcı́a, O. Gervasi, A. Ottavi, Chem. Phys. 168

(1992) 341.
[132] A. Laganà, G. Ochoa de Aspuru, E. Garcı́a, J. Chem. Phys. 108 (1998) 3886.
[133] Yang, M., Zhang, D. H., Collins, M. A., Lee, S.-Y., 2001, J. Chem. Phys.,
115, 174.
[134] S. Gordon, W. A. Mulac, Int. J. Chem. Kin. 7, 289 (1975).
[135] R. Atkinson, R. A. Perry, J. N. Pitts, Chem. Phys. Lett. 44, 204 (1976). R.
A. Perry, R. Atkinson, J. N. Pitts, J. Chem. Phys. 67,5577 (1977).
[136] J. E. Spencer, H. Endo, G. P. Glass, 16th Symposium on Combustion, The

Combustion Institute, 829 (1977).
[137] C. M. Stevens, L. Kaplan, R. Gorse, S. Durkee, M. Compton, S. Cohen, K.

Bielling, Int. J. Chem. Kin. 12, 935 (1980).
[138] T. Dreier, J. Wolfrum, 18th Symposium on Combustion, The Combustion

Institute, 801 (1981).
[139] G. Paraskevopoulos, R. S. Irwin, J. Chem. Phys. 80, 259 (1984).
[140] J. Wolfrum, Faraday Discuss. Chem. Soc. 84, 191 (1987).
[141] J. Brunning, D. W. Derbyshire, I. W. M. Smith, M. D. Willians, J. Chem.

Soc., Faraday Trans. 2 84, 105 (1988).
[142] M. J. Frost, P. Sharkey, I. W. M. Smith, Faraday Discuss. Chem. Soc. 91,

305 (1991).
10B - 24 RTO-EN-AVT-142
[143] M. Alagia, N. Balucani, P. Casavecchia, D. Stranges, G. G. Volpi, J. Chem.

Phys. 98, 8341 (1993).
[144] M. S. Wooldridge, R. K. Hanson, C. T. Bowman, 25th Symposium on Com-

bustion, The Combustion Institute, 741 (1994).
[145] D. Fulle, H. F. Hamann, H. Hippler, J. Troe, J. Chem. Phys. 105, 983 (1996).
[146] D. M. Golden, G. P. Smith, A. B. McEwen, C.-L. Yu, B. Eiteneer, M. Fren-

klach, G.L. Vaghjiani, A. R. Ravishankara, F. P. Tully, J. Phys. Chem. A
102, 8598 (1998).
[147] T. Rockmann, C. A. M. Brenninkmeijer, G. Saueressig, P. Bergamaschi, J.

N. Crowley, H. Fisher, P. J. Crutzen, Science 281, 544 (1998).
[148] K. L. Feilberg, S. R. Sllevag, C. J. Nielsen, D. W. T. Griffith, M. S. Johnson,

Phys. Chem. Chem. Phys. 4, 4687 (2002).
[149] K. Kudla, G. C. Schatz, in The Chemical Dynamics and Kinetics of Small

Radicals, K. Liu and A. Wagner Ed., World Scientific, Singapore (1995), p.
438.
[150] D. C. Clary, G. C. Schatz, J. Chem. Phys. 99, 4578 (1993).
[151] M. I. Hernández, D. C. Clary, J. Chem. Phys. 101, 2779 (1994).
[152] E. M. Goldfield, S. K. Gray, G. C. Schatz, J. Chem. Phys. 102, 8807 (1995).
[153] D. H. Zhang, J. Z. H. Zhang, J. Chem. Phys. 103, 6512 (1995).
[154] N. Balakrishnan, G. D. Billing, J. Chem. Phys. 104, 4005 (1996).
[155] F. N. Dzegilenko, J. M. Bowman, J. Chem. Phys. 108, 511 (1998).
[156] R. Valero, G. J. Kroes, J. Chem. Phys. 117, 8736 (2002).
[157] D. A. McCormack, G. J. Kroes, Chem. Phys. Lett. 352, 281 (2003). Erratum,
ibid. 373 , 648 (2003)
[158] G. D. Billing, J. T. Muckerman, H. G. Yu, J. Chem. Phys. 117, 4755 (2002).
[159] M. J. Lakin, D. Troya, G. C. Schatz, L. B. Harding, J. Chem. Phys. 119,

5848 (2003).
[160] D. M. Medvedev, S. K. Gray, E. M. Goldfield, M. J. Lakin, D. Troya, G. C.

Schatz, J. Chem. Phys. 120, 1231 (2004).
[161] R. Valero, D. A. McCormack, G. J. Kroes, J. Chem. Phys. 120, 4263 (2004).
[162] R. Valero, M. C. van Hemert, G. J. Kroes, Chem. Phys. Lett. 393, 236 (2004).
[163] R. Valero, G. J. Kroes, Phys. Rev. A 70, 040701 (2004).
RTO-EN-AVT-142 10B - 25
[164] R. Valero, G. J. Kroes, J. Phys. Chem. A 108, 8672 (2004).
[165] Y. He, E. M. Goldfield, S. K. Gray, J. Chem. Phys. 121, 823 (2004).
[166] G. C. Schatz, M. S. Fitzcharles, Faraday Discuss. Chem. Soc. 84, 359 (1987).
[167] K. Bradley, G. C. Schatz, J. Chem. Phys. 106, 8464 (1997).
[168] D. Troya, M. J. Lakin, G. C. Schatz, L. B. Harding, M. González, J. Phys.

Chem. B 106, 8148 (2002).
[169] K. Kleinermanns, J. Wolfrum, Chem. Phys. Lett. 104, 157 (1984). K. Klein-
ermanns, E. Linnebach, J. Wolfrum, J. Phys. Chem. 89, 2525 (1985).
[170] A. Jacobs, M. Wahl, J. Wolfrum, Chem. Phys. Lett. 158, 161 (1989).
[171] N. F. Scherer, C. Sipes, R. B. Bernstein, A. H. Zewail, J. Chem. Phys. 92,

5239 (1990).
[172] S. K. Shin, Y. Chen, S. Nickolaisen, S. W. Sharpe, R. A. Beaudet, C. Wittig,

in Adevances in Photochemistry, Volume 16, Eds. D. Volman, G. Hammond,
D. Neckers, Wiley (1991). p. 115.
[173] M. Brouard, D. W. Hughes, K. S. Kalogerakis, J. P. Simons, J. Phys. Chem.

A 102, 9559 (1998).
[174] M. Brouard, D. W. Hughes, K. S. Kalogerakis, J. P. Simons, J. Chem. Phys.

112, 4557 (2000).
[175] XX Conference on the Dynamics of Molecular Collisions, Asilomar, CA, July

10-15, 2005.
[176] VIII workshop on Quantum Reactive Scattering, Santa Cruz, July 15-19,
2005
[177] J. N. Murrell, S. Carter, S. C. Farantos, P. Huxley, A. J. C. Varandas,

Molecular Potential Energy Functions, Wiley, Chichester, 1984.
[178] H.-G. Yu, J. T. Muckerman, T. J. Sears, Chem. Phys. Lett. 349, 547 (2001).
[179] T. V. Duncan, C. E. Miller, J. Chem. Phys. 113, 5138 (2000).
[180] R. S. Zhu, E. G. W. Diau, M. C. Lin, A. M. Mebel, J. Phys. Chem. A 105,

11249 (2001).
10B - 26 RTO-EN-AVT-142
Laboratory Determination of Thermal Protection System
Materials Surface Catalytic Properties
Jochen Marschall
Molecular Physics Laboratory, SRI International
333 Ravenswood Avenue
Menlo Park, CA 94025
USA
jochen.marschall@sri.com
ABSTRACT
Different experimental approaches to measuring catalytic reaction efficiencies on thermal protection system
materials are reviewed. Special emphasis is given to the theory and application of the diffusion tube side-arm
reactor technique. In this technique, reactants diffuse into a dead-end tube and are progressively removed
from the gas phase by surface reactions on the walls, establishing unique steady-state concentration profiles
along the length of the tube. Reactant loss probabilities are determined by matching experimentally measured
species profiles to calculated solutions of a reaction-diffusion model. The advantages of laser-based methods
for species concentration measurements are summarized and different approaches to reactor modeling and
the extraction of reaction efficiencies from measured data are presented. The advantages and limitations of
the diffusion-tube side-arm technique, the associated uncertainties in derived loss probabilities, and the
prospects for further laboratory development, are presented.
1.0 INTRODUCTION
During hypersonic flight through a planetary atmosphere, shock waves form as gases are rapidly compressed
ahead of leading surfaces. High energy intermolecular collisions within the shock layer heat the gas and cause
molecular species to dissociate. The extent of molecular dissociation depends on many factors, including the
atmospheric composition and the vehicle shape, velocity, and altitude. Dissociated gas species can diffuse
through the boundary layer to the vehicle surface, where they have the potential to undergo exothermic
surface-catalyzed recombination reactions. Some fraction of the energy liberated by recombination may be
transferred directly to the vehicle surface as heat. The greater the extent of dissociation and the more efficient
the surface is at catalyzing recombination reactions, the more likely it is that recombination contributes
significantly to the total aerothermodynamic heat load.
Oxygen and nitrogen molecules are the relevant species in the Earth’s atmosphere. Oxygen is more easily
dissociated than nitrogen, since it has a lower dissociation energy (~5.1 eV compared to ~9.8 eV.) The
importance of surface catalyzed reactions to Earth re-entry heating was demonstrated in a series of flight
experiments on the Space Shuttle Orbiter.1-3 Adjacent thermal protection system (TPS) tiles in the base heat
shield of the orbiter attained significantly different surface temperatures depending on the catalytic efficiency
of their surface coating.
Marschall, J. (2007) Laboratory Determination of Thermal Protection System Materials Surface Catalytic Properties. In Experiment,
Modeling and Simulation of Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 11-1 – 11-32). Educational Notes
RTO-EN-AVT-142, Paper 11. Neuilly-sur-Seine, France: RTO. Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 11 - 1
Laboratory Determination of Thermal
Protection System Materials Surface Catalytic Properties
During high speed entry into extraterrestrial atmospheres, the dissociation and recombination of other species
can be important. For example, the dissociation of CO2 during Martian entries enables the surface reactions
CO + O and O + O, and the dissociation of methane during Titan entries creates atomic hydrogen for
heterogeneous H + H recombination.
The quantity of heat transferred to a catalytic surface by a particular recombination reaction pathway depends
first on the efficiency of surface in catalyzing the reaction, and second, on the efficiency of the energy transfer
process between the energetic product molecules and the surface. Since surface reaction must precede energy
transfer, reaction efficiency is the more globally important parameter. It is also more experimentally
accessible, although by no means straightforward to determine.
Neither reaction nor energy transfer efficiencies can be measured directly. Experiments must probe the
effects of surface reactions on the gas phase concentrations of reactants or products near catalytic surfaces or
the flow of heat to catalytic bodies. Measurements of the latter phenomenon are generally insufficient to
separate reaction and energy transfer efficiencies without additional assumptions or information, whereas
species concentration data can be related much more directly to reaction efficiencies through appropriate
chemical and transport models.
Experiments to determine catalytic efficiencies for TPS materials can be divided into two categories:
measurements in high enthalpy facilities such as arc-jets, induction plasmatrons, and shock tubes, and
measurements in low enthalpy laboratory systems such as diffusion tubes or flow tubes. The former more
closely reproduce aspects of the flow environments seen during hypersonic flight, but are also much more
complicated to operate and generally generate temperature or heat flux data. In contrast, laboratory systems
are more amenable to controlled and well-characterized experiments, although under conditions far from those
in flight.
In this lecture, laboratory approaches for determining catalytic efficiencies are discussed, with a specific focus
on the diffusion-tube side-arm reactor technique in combination with laser-based species detection
diagnostics.
2.0 SURFACE CATALYSIS: DEFINITIONS AND MODELS
2.1 Basic Quantities
We consider heterogeneous chemical reactions mediated by adsorption of one or more reactants on the surface
of a TPS material. Discussion is limited to chemical reactions in which the TPS material is not consumed or
converted. The TPS material participates by providing an environment for the reaction of impinging gas
phase species, and acts as a catalytic surface when it enhances the net rate of a thermodynamically-favoured
chemical reaction. The catalytic characteristics of a surface for a particular reaction can range between two
limits. When the reaction rate is negligible the surface is said to be non-catalytic; when the reaction proceeds
at the maximum possible rate the surface is termed fully-catalytic.
The net production of species i by surface-catalyzed reactions must be balanced by its net diffusive flux Ji,w at
the wall. For a one-dimensional body-fitted coordinate system
11 - 2 RTO-EN-AVT-142
∂n
J i , w = − Di = w i , (1)
∂z w
where z is the normal coordinate to the wall (positive outward), ni is the number density, Di is the diffusion
coefficient, and w i is the net species production rate (positive for formation, negative for consumption).
For reactant species it is often convenient to express the net flux to the surface in terms of a loss probability,
γ i , times the surface impingement flux, Γi . Then the production rate of the reactant at the surface is
w i = −γ i Γi . (2)
The impingement flux is typically approximated by the classic gas kinetic expression
vi
Γi = n i , w , (3)
4
with ni , w the species concentration at the wall and vi the average thermal speed given by
8ℜT
vi = , (4)
πM i
where ℜ is the universal gas constant; T is the temperature, and Mi is the molar mass of the species.
The loss probability is defined here as the fraction of impinging reactant flux removed permanently from the
gas phase. With this definition, the loss probability is equivalent to the “uptake coefficient” found in the
atmospheric heterogeneous chemistry literature. For atom recombination via a single reaction pathway, the
loss probability is often called the “recombination coefficient.”
It is important to remember that γ i is not uniquely associated with a single chemical process in the same way
that a reaction rate coefficient is associated with a specific gas phase reaction. Rather, γ i reflects the total
efficiency of all operating surface reaction pathways that remove species i on a particular surface under a
particular combination of temperature, pressure, and gas composition. Similar to other surface properties like
emittance and reflectance, the loss probability depends both on the chemical nature and the microscopic
geometric characteristics of the surface. The absolute lower and upper bounds on γ i are 0 and 1, but this range
may be limited further by various other factors such as the availability of partner reactants or free reaction
sites on the surface.
2.2 Element Conservation
Element conservation is the most basic physical constraint on numerical values of loss probabilities. Element
conservation also ensures mass conservation. For a single reactant participating in a single recombination
reaction, element conservation is consistent with the 0 through 1 range of loss probability. However, when a
reactant participates in more than one surface reaction or when surface reactions involve two different reactant
species, element conservation can constrain the range further.
RTO-EN-AVT-142 11 - 3
Consider a partially dissociated mixture of oxygen and nitrogen interacting with a catalytic surface via three
net steady-state heterogeneous reaction pathways: O + O → O2, N + N → N2, and O + N → NO. Element
conservation dictates
w NO + 2w O + w O = 0
2
(5)
w NO + 2 w N + w N = 0 .
2
(6)
The production rate of nitric oxide can be eliminated from this equation set, and with some rearrangement the
relation
γ O ΓO (1 − f O ) = γ N ΓN (1 − f N
2 2
) (7)
is obtained, where f O is defined as −2w O w O , the fraction of the total O-atom loss that forms O2, and f N is
2 2 2
defined as −2w N w N , the fraction of the total N-atom loss that forms N2. By definition the four parameters
2
γ O , γ N , f O , and f N can be no smaller than zero and no larger than one, however all four values cannot be
2 2
specified independently because they must also be consistent with Eq. 7.
For any allowed set of parameters, the individual production rates are then
w O = −γ O ΓO (8a)
w N = −γ N ΓN (8b)
γO
w O = f O ΓO (8c)
2 2
2
γN
w N = f N ΓN (8d)
2 2
2
w NO = γ O ΓO (1 − f O ) = γ N ΓN (1 − f N
2 2
) (8e)
2.3 Surface Reaction Models
Element conservation is a necessary condition for a physically consistent description of surface catalytic
effects, but it does not directly relate loss probabilities to chemical mechanisms on the surface. A surface
chemistry model is required to link net species production rates to a specific set of finite-rate surface reactions
and their dependencies on gas composition, temperature, and active site density (or reactive surface area).
The “true” chemical processes that take place at a catalytic surface are extremely complex, even for simple
reactions on pristine surfaces.4 A phenomenological mechanistic description of the temperature and
compositional dependencies of the dominant kinetic processes suffices in many chemical engineering
applications; this approach has been applied in various forms to describe surface catalysis for aerothermal
computations, primarily in the European aerospace community.5-8
11 - 4 RTO-EN-AVT-142
Table 1: Surface Reaction Model for O-O2-N-N2-NO system For the partially-dissociated oxygen-
Site Filling Flux nitrogen system, a simple finite-rate
Reaction (moles m-2 s-1) surface reaction model can be
Adsorption constructed based on two reactants (O
R1 O + [s] → Os S Oθ ΓO and N atoms) and the four kinetic
R2 N + [s] → Ns S N θ ΓN processes of adsorption, thermal
Thermal Desorption desorption, Eley-Rideal (ER)
- ν D ,Oθ O Φ s recombination, and Langmuir-
R3 Os → O + [s] Hinschelwood (LH) recombination. All
R4 Ns → N + [s] - ν D , Nθ N Φ s processes for both reactants are
Eley-Rideal Recombination envisioned to take place at a finite
R5 O + Os → Ο2 + [s] - γ ER ,OOθ O ΓO number of identical active sites, [s], on
- γ ER , NN θ N ΓN the surface.
R6 N + Ns → Ν2 + [s]
R7 O + Ns → ΝΟ + [s] - γ ER ,ON θ N ΓO ER recombination occurs between an
R8 N + Os → ΝΟ + [s] - γ ER , NOθ O ΓN adsorbed atom and an impinging gas
Langmuir-Hinschelwood Recombination phase atom; LH recombination occurs
-2 ν LH ,OOθ O2 Φ s between two adsorbed atoms, one of
R9 Os,m + Os → Ο2 + 2[s] which is mobile. In both types of
R10 Ns,m + Ns → Ν2 + 2[s] -2 ν LH , NN θ N2 Φ s recombination, the product molecule is
R11 Os,m + Ns → ΝΟ + 2[s] -2 ν LH ,ON θ Oθ N Φ s assumed to desorb instantaneously. The
-2 ν LH , NOθ N θ O Φ s reverse processes of dissociative O2, N2,
R12 Ns,m + Os → ΝΟ + 2[s] and NO adsorption are not included.
Variables
Φs Active site density, (moles m-2) Table 1 lists the individual reactions and
θ ,θi Fraction of active sites, free or occupied by species i corresponding expressions for the active
Si Sticking coefficient for species i site filling fluxes, as well as the variable
ν D ,i definitions.
Thermal desorption frequency for species i, (s-1)
γ ER ,ij Eley-Rideal reaction efficiency between gas phase At steady-state, the number of active
species i and adsorbed species j sites occupied by each atomic species is
ν LH ,ij Langmuir-Hinshelwood reaction frequency between constant and the site filling fluxes must
mobile species i and immobile species j, (s-1) sum to zero:
dθ O
Φs = 0 = S Oθ ΓO −ν D ,Oθ O Φ s − γ ER , OOθ O ΓO − γ ER , NOθ O ΓN − 2ν LH , OOθ O2 Φ s −ν LH , NO θ N θ O Φ s −ν LH ,ON θ Oθ N Φ s (9)
dt
dθ N
Φs = 0 = S N θ ΓN −ν D , N θ N Φ s − γ ER , NN θ N ΓN − γ ER ,ON θ N ΓO − 2ν LH , NN θ N2 Φ s −ν LH , ON θ Oθ N Φ s −ν LH , NO θ N θ O Φ s (10)
dt
From these relations and the identity θ = 1 − θ O − θ N , the fractional surface coverage of each atomic species can
be obtained from the coupled quadratic equations:
a Oθ O2 + bOθ O + c O = 0 (11a)
a N θ N2 + b N θ N + c N = 0 , (11b)
RTO-EN-AVT-142 11 - 5
with
a O = 2ν LH , OO Φ s , bO = ν D ,O Φ s + γ ER ,OO ΓO + γ ER , NO ΓN + (ν LH , NO + ν LH , ON )θ N Φ s + S O ΓO , c O = S O ΓO (θ N − 1) (12a)
a N = 2ν LH , NN Φ s , b N = ν D , N Φ s + γ ER , NN ΓN + γ ER ,ON ΓO + (ν LH , ON + ν LH , NO )θ O Φ s + S N ΓN , c N = S N ΓN (θ O − 1) (12b)
Expressions for the species production rates are
w O = − S Oθ ΓO + ν D ,O θ O Φ s − γ ER ,OOθ O ΓO − γ ER , ON θ N ΓO (13a)
w N = − S N θ ΓN + ν D , N θ N Φ s − γ ER , NN θ N ΓN − γ ER , NO θ O ΓN (13b)
w O = γ ER , OOθ O ΓO + ν LH , OOθ O2 Φ s
2
(13c)
w N 2 = γ ER , NN θ N ΓN + ν LH , NN θ N2 Φ s (13d)
w NO = γ ER ,ON θ N ΓO + γ ER , NO θ O ΓN + ν LH , ON θ Oθ N Φ s + ν LH , NO θ N θ O Φ s (13e)
and expressions for the atom loss probabilities follow directly from Eq. 2:
ν D ,Oθ O Φ s
γ O = S Oθ + γ ER ,OOθ O + γ ER ,ON θ N − (14a)
ΓO
ν D, N θ N Φ s
γ N = S N θ + γ ER , NN θ N + γ ER , NOθ O − . (14b)
ΓN
This simple surface reaction model illustrates the complexity in systems of coupled heterogeneous reactions.
The loss probability of each individual atomic species is dependent on the availability of both species and
their individual chemical interactions with the surface.
The complexity of the model can be methodically reduced by eliminating certain interactions. For example,
the assumptions that NO is not produced heterogeneously and that O + O and N + N recombination proceed
on independent sets of active surface sites, Φ s ,O and Φ s , N , decouples the oxygen and nitrogen chemistry. So
for oxygen,
a O = 2ν LH , OO Φ s , O , bO = ν D ,O Φ s , O + γ ER , OO ΓO + S O ΓO , c O = − S O ΓO (15)
w O = − S Oθ ΓO + ν D ,Oθ O Φ s ,O − γ ER ,OOθ O ΓO (16)
w O = γ ER , OOθ O ΓO + ν LH , OOθ O2 Φ s , O
2
(17)
ν D ,Oθ O Φ s ,O
γ O = S Oθ + γ ER ,OOθ O − (18)
ΓO
with analogous expressions for nitrogen.
11 - 6 RTO-EN-AVT-142
Even further simplification is obtained by assuming that thermal desorption is negligible and that only Eley-
Rideal processes take place. Then for oxygen we have the very simple relationships,
a O = 0, bO = γ ER ,OO ΓO + S O ΓO , c O = − S O ΓO (19)
w O = − S Oθ ΓO − γ ER ,OOθ O ΓO (20)
w O = γ ER , OOθ O ΓO
2
(21)
2 S O γ ER ,OO
γO = , (22)
S O + γ ER ,OO
again with symmetric expressions for nitrogen atoms. By their formulation in terms of steady-state species
fluxes, all surface reaction models above automatically satisfy element conservation.
2.4 Application to Simulation and Data Analysis
Equation set 8 together with relation 7 can be used to specify species production rates in diffusive flux
boundary conditions (Eq. 1), to ensure that numerical simulations of reactive flow interactions with a catalytic
surface are at least constrained to a parameter space consistent with element (and mass) conservation. The
extremes within this parameter space can be used to bound the effects of different surface reaction pathways
and their efficiencies on gas composition, catalytic heating rates, etc.
Laboratory experiments investigating catalytic surfaces can involve measurements of species concentrations
and concentration gradients near surfaces, since these quantities reflect the species production rates at the
surface. Reactive flow simulations can be tuned to match measured data by adjusting loss probabilities in the
boundary conditions to obtain the best agreement possible. The resulting values are not fundamental chemical
quantities, but rather emperical parameters that relate experimental species production rates to reactant
impingement fluxes under specific gas composition, temperature, and pressure conditions.
Analytic fits to experimental loss probabilities are often used in reactive-flow simulations attempting to
predict the catalytic effects on heating and gas composition in different flight and test environments.
Typically, these fits contain only the temperature dependence of the loss probability, with no information on
pressure or gas composition dependencies.9-11 By using an analytic fit of this sort, the same assumptions
required to obtain Eq. 22 are implicitly assumed; i.e., a single atom recombination reaction on an independent
sets of active sites, negligible atom desorption, and only ER recombination.
A preferred approach is to use a general surface chemistry model to extrapolate and interpolate loss
probabilities. The variables describing the efficiencies or frequencies of the various heterogeneous processes -
S i , ν D,i , γ ER, ij , and ν LH ,ij - are not constants in general. Analytic formulas for their temperature dependences
can be obtained from kinetic and transition state theory, and typically involve Arrhenius-type expressions with
temperature-dependent prefactors.7,8,12 Thus at least nine quantities – four activation energies, four prefactors,
and the active site density, Φ s , must be assigned numerical values to complete the model in Table 1.
RTO-EN-AVT-142 11 - 7
A surface chemistry model can be fit to experimental data by adjusting these model parameters to reproduce
experimental results. Experimental data must be available over a sufficiently wide range of pressures,
temperatures, and gas compositions, so that model parameters are both robustly determined and retain
physically plausible numerical values. Such wide-ranging data sets are rarely available, and in practice, some
arbitrary choices must be made, guided by ancillary information in the chemistry and physics literature.
Nevertheless, a surface reaction model is a fitting function with a mathematical formulation that can be
justified, at least to some extent, by underlying physical and kinetic mechanisms.
3.0 OVERVIEW OF LABORATORY METHODS

Laboratory methods characterizing atom recombination reactions at an “engineering” level have primarily
used flow or diffusion tube reactors, in conjunction with different methods of generating reactants, measuring
species concentrations, and extracting numerical values for loss probabilities. Experiments in these reactors
involve spatially resolved reactant concentration measurements in the vicinity of a catalytic specimen and/or
energy transfer measurements to a catalytic sample or probe. Smith13 first introduced the diffusion-tube side-
arm reactor design in 1943 to study surface reactions on materials of low catalytic efficiency. One of the
primary advantages of the diffusion tube approach is that species transport is greatly simplified, and
complications associated with measuring and modelling the interactions of different convective and diffusive
flow regimes are avoided.
A great deal of work was performed in both types of reactor systems from the late 1950’s through the early
1970’s, including frequently cited works by Melin and Madix14,15, Linnett et al.16-24, and Wise, Wood and
coworkers.25-40 Atoms were generated using a variety of electrical discharge sources. Traditional methods of
atom detection have included chemical titration,41 electron spin resonance,28,42-44 Wrede-Harteck gauges,16-18
isothermal calorimeters,39,40 and thermocouple probes.13,19,20,22-24,45 Some of this early work was summarized
in reviews by Kaufman,41 Brennan,46 and Wise and Wood.36
In more recent times, a significant amount of surface catalytic efficiency data have been generated in the
MESOX test facility in the PROMES-CNRS laboratory on a variety of ceramic materials.47-50 The MESOX
facility uses a flow tube configuration with the test sample facing the oncoming flow. A microwave discharge
is used to dissociate air and a unique solar radiation concentrator is used to heat specimens to temperatures as
high as 2300 K. O-atom concentration gradients in the vicinity of the sample are inferred using actinometry, a
technique based on measuring the relative emission intensities of electronically-excited oxygen and argon
atoms at selected wavelengths.48,51,52 O-atom recombination coefficients are extracted using an analytic
expression for the O-atom gradient approaching the catalytic surface.47 Simultaneous measurements of
sample temperature enable the estimation of the energy transfer efficiency using an energy balance
calculation.47
Amouroux and coworkers53-55 used a similar flow tube configuration to expose a wide range of metallic, oxide
and ceramic materials to O-atoms. A radio frequency53,54 or microwave discharge55 was used to dissociate
oxygen molecules and a cartridge heater was used to heat samples to different test temperatures.
Recombination coefficients were obtained by fitting an analytic expression for the O-atom gradients near the
surface to actinometry data,53,54 or calculated from titration data with or without the sample in place.55 A flow
tube configuration was also used by Mozetič, and coworkers56-58 to study oxygen and hydrogen recombination
on steel and niobium surfaces. O-atoms were generated using a radio frequency discharge and recombination
coefficients were calculated from measured sample heating curves.
11 - 8 RTO-EN-AVT-142
During the 1990’s, NASA Ames Research Center supported laboratory measurements of atom recombination
coefficients using the diffusion-tube side arm technique, including the study of Kim and Boudart on O, N, and
H atom recombination on silica.45 In this study, a microwave discharge source was used to generate atomic
species, a thermocouple probe was used to measure atomic concentration gradients along the side-arm axis,
and a furnace was used to control tube temperatures.
In subsequent work, Pallix and Copeland59 introduced the use of laser-induced fluorescence (LIF) as an atom
detection method for diffusion tube side-arm reactor experiments. The LIF technique is species specific and
minimally intrusive. This avoids many problems associated with the traditional atom detection techniques
listed above; e.g., the introduction of contaminates, the disturbance of the gas flow or concentration gradient,
the lack of spatial sensitivity, or the lack of species discrimination.
Since 1995, the diffusion tube side-arm reactor in conjunction with laser induced species detection diagnostics
has served as the main method by which NASA Ames Research Center has measured O and N atom
recombination coefficients in the room temperature to 1000 K range.
4.0 DIFFUSION-TUBE SIDE-ARM REACTOR
4.1 Measurement Theory
The diffusion tube side-arm reactor consists of a dead-end side-arm tube connected at right angles to a main
flow tube; see Fig. 1. Free-radical species are produced in the main flow tube upstream of the tube
intersection, typically by dissociating molecular gasses using some type of low-pressure electrical discharge.
The dissociated species flow past the opening of the side-arm tube. As reactants diffuse into the side-arm
tube, they are progressively removed from the gas phase by surface reactions on the walls, establishing a
unique steady-state species concentration profile down the length of the tube. This reactor design simplifies
subsequent data analysis because gas transport in
the side-arm is restricted to species diffusion.
Vacuum Pumps
4.2 Diffusion-Reaction Models
Diffusion Tube Side-Arm
The complexity of the diffusion-reaction model
required to predict species concentrations in the
Discharge side-arm reactor depends on many things: the
Gas Inlet number of gas species present, the number of
potential surface reactions and their reaction orders,
1
the number of potential gas-phase reactions, and
U = n/n0 spatial variations in tube material, geometry, and
temperature.
0
0 Distance Along Side-Arm L The simplest systems are binary atom-molecule
mixtures in a uniform-diameter isothermal tube of a
Fig. 1 Schematic diagram of a diffusion-tube side- single wall material. Operation at sufficiently low
arm reactor and a characteristic concentration pressures ensures that gas-phase reactions are
profile for a reactant depleted by surface reactions. negligible in comparison to heterogeneous atom
losses at the wall, and working with low atom
RTO-EN-AVT-142 11 - 9
concentrations ensures that the linear diffusion equation is sufficiently accurate. Further simplifying
assumptions are first-order irreversible surface reaction kinetics, a purely catalytic wall, and a single
heterogeneous loss pathway leading to atom recombination. Described below are one and two dimensional
reaction-diffusion models that can be applied to evaluate surface recombination in binary systems such as O-
O2 and N-N2.
4.2.1 One-Dimensional Model: Single Wall Material
The most basic solution for the steady-state atom concentration along a finite tube of length Lz is obtained by
solving the linear one dimensional diffusion equation with surface loss included as a first-order sink term:
∂2n
D − Rsink = 0 (23)
∂z 2
where R sink ≡ (loss probability)(atom-surface impingement rate)(surface area per volume), or
 v n  2πRdz   γv 
R sink = γ wall   2  =  n . (24)
 4  πR dz   2 R 
Here n is the atom number density, z is the axial tube coordinate, R is the tube radius, D is the binary diffusion
coefficient, and v is the average thermal speed of the atoms. For a binary atom-molecule mixture where the
net loss process is heterogeneous atom recombination, the loss probability γ wall is commonly referred to as the
recombination coefficient. Appropriate boundary conditions are constant atom concentration at the tube
entrance and heterogeneous atom recombination at the tube endplate with loss probability γ end which may
differ from that of the tube wall:
∂n v n( L z )
n(0 ) = n 0 −D = γ end . (25a,b)
∂z Lz 4
It is useful to recast the equations in non-dimensional form by normalizing the axial coordinate and tube
length by the tube radius and the atom concentration by the concentration at the tube entrance; viz., x ≡ z / R ,
L ≡ L z / R and U (x ) ≡ n(x ) n 0 . This leads to the equation set
∂ 2U
− λ2U = 0 , (26)
∂x 2
∂U
U (0 ) = 1 = −αU (L ) . (27a,b)
∂x L
λ = v Rγ wall 2 D α = v Rγ end 4 D (28a,b)
A general solution to Eq. 26 is
U (x ) = A cosh (λx ) + B sinh (λx ) . (29)
11 - 10 RTO-EN-AVT-142
Evaluation of the constants A and B with the boundary conditions 27a and b gives the particular solution
 α cosh (λL ) + λ sinh (λL ) 

U (x ) = cosh (λx ) − sinh (λx )  . (30)
 λ cosh (λL ) + α sinh (λL ) 
In the limit of an infinitely long tube, the final bracketed term goes to unity and Eq. 30 reduces to the decaying
exponential solution
U (x ) = exp(− λx ) . (31)
Solutions to Eq. 30, representative of O-atom diffusion into quartz tube at room temperature and low pressure
(298 K, 67 Pa, R = 1.1 cm, γ wall= 10 −4 ) are plotted in Fig. 2 for three different tube lengths and four different
endplate reaction efficiencies. The infinite tube solution of Eq. 31 is also included.
Figure 2 illustrates several general features of this diffusion-reaction system. At locations far from the
endplate, the atom concentration profile along the tube is well approximated as an exponential decay. In this
region, a plot of ln(U (x )) vs x is linear with a slope of –λ that can be directly related to the wall catalytic
efficiency through Eq. (28a).
As the endplate is approached, the atom concentration profile deviates from a pure exponential dependence on
x and this deviation can be either positive or negative depending on the catalytic efficiency of the endplate
material. In this region of the tube, the atom
concentration profile depends on both γ wall and
0 Infinite Tube γ end .
10 L=50R -2
γend = 0 γend = 10
-3
γend = 10 γend = 1 One attractive feature of this binary diffusion-
reaction system is that only relative values of
10
-1 L=25R atom concentrations along the tube are required
to relate concentration profiles to the catalytic
efficiencies of the wall and endplate materials.
Experimentally, relative atom concentrations
U(x)
-2 L=10R
10 are much easier to quantify than absolute atom
concentrations.
-3 Because the solution for U(x) can be anchored

10
(started) anywhere along the tube, modelling
the entire length of the diffusion tube is often
unnecessary. For example, if measurements of
10
-4 relative atom concentrations are made over a
0 10 20 30 40 50 central region of the tube far from either the
x opening or the endplate, then the exponential
solution (Eq. 31), with x = 0 at the location of
Fig. 2 O-atom concentration profiles calculated for a
finite quartz tube with three different tube lengths and the first concentration measurement (nearest the
four different endplate recombination coefficients; tube opening), is sufficient to relate measured
R = 1.1 cm, T = 298 K, P = 67 Pa, and γwall = 10 .
-4
concentration profiles to the wall catalytic
efficiency.
RTO-EN-AVT-142 11 - 11
4.2.2 One-Dimensional Model: Three Tube Sections
Several solutions of the form of Eq. 29 can be joined to model a diffusion tube of multiple sections having
walls of different reaction efficiency. A three-section solution is particularly useful for simulating atom
concentration profiles along a relatively inert tube containing a more catalytic specimen of finite length. The
atom concentration in each tube section of length Li can be written as:
U 1 (x ) = A1 cosh (λ1 x ) + B1 sinh (λ1 x )

U 2 (x ) = A2 cosh (λ 2 (x − L1 )) + B2 sinh (λ 2 (x − L1 )) (32a,b,c)
U 3 (x ) = A3 cosh (λ3 (x − L1 − L2 )) + B3 sinh (λ3 (x − L1 − L2 ))
To join the different solutions, additional boundary conditions are specified enforcing continuity of atom
concentration and concentration gradient at the interfaces between different tube sections. The required
constants Ai and Bi are obtained by solving the linear equation set
 1 0 0 0 0 0 
 cosh (λ L ) sinh (λ1 L1 ) −1   A1  1
 1 1 0 0 0   B1  0
λ1 sinh (λ1 L1 ) λ1 cosh (λ1 L1 ) − λ2 0 0 0    
   A2  = 0
 0 0 cosh (λ 2 L 2 ) sinh (λ 2 L 2 ) −1 0   B  0 
 0 0 λ 2 sinh (λ 2 L2 ) λ 2 cosh (λ 2 L 2 ) − λ3 0   2  
  A  0 
 λ 3 sinh (λ 3 L3 ) + λ 3 cosh (λ 3 L3 ) +   3   
0 0 0 0
α sinh (λ 3 L3 )  3   
B 0
 α cosh (λ 3 L3 )
(33)
If the tube is infinite, the general solution for the final (third) tube section is replaced by
U 3 (x ) = A3 exp(− λ3 (x − L1 − L2 )) and equation set 33 is reduced by one order to
 1 0 0 0 0   A1  1
 cosh (λ L ) sinh (λ L ) − 1 0 0  B   
 1 1 1 1   1  0 
λ1 sinh (λ1 L1 ) λ1 cosh (λ1 L1 ) − λ2 0 0   A2  = 0 . (34)
    
 0 0 cosh (λ 2 L 2 ) sinh (λ 2 L 2 ) − 1  B 2  0

 0 0 λ 2 sinh (λ 2 L 2 ) λ 2 cosh (λ 2 L 2 ) λ 3   A3  0
Figure 3 shows the O-atom concentration profiles calculated for a quartz tube γ 1 = γ 3 = γ wall = 10 −4 containing
samples of three different catalytic efficiencies γ 2 = γ s = 10 −3 ,10 −2 ,10 −1 . Again, with sufficient separation
between the sample location and the end of the diffusion tube, the influence of the endplate catalytic
efficiency on the atom concentration profile near the catalytic sample is negligible.
The presence of the catalytic sample distorts the exponential atom concentration profile along the quartz tube.
The decay of the atom concentration with distance is accelerated in the region preceding the sample, is most
rapid within the sample region, and then recovers the characteristic exponential decay along the quartz tube
beyond the sample, though at lower atom concentrations. The larger the difference between the quartz and the
sample catalytic efficiencies, the more pronounced these effects. Analogous trends are observed when the
11 - 12 RTO-EN-AVT-142
length of a catalytic sample is increased, as shown

10
0 Infinite Tube in Fig. 4.
γend = 0
γend = 1
γs = 10
-3
In an experiment, one would try to measure
changes in atom concentration before and after a
10
-1
target sample, and relate these measurements to
the catalytic efficiencies of the wall and sample.
-2
From an experimental perspective, it is
γs = 10 advantageous if concentration profile with the
U(x)
-2
10 sample in place shows a large deviation from the
baseline exponential decay observed for the
uniform quartz tube. However, if the sample
10
-3
γs = 10
-1 removes too many atoms from the gas phase, the
atom concentration beyond the sample may be too
Sample small to detect accurately. The optimal length for
Location
a particular test specimen is determined by
10
-4
balancing these two factors, though this can only
0 10 20 30 40 50 be estimated before the experiment since the
x exact catalytic efficiency of the sample is
Fig. 3 O-atom concentration profiles calculated for a presumably unknown.
three-section tube with three different sample and two
different endplate recombination coefficients; R = 1.1 cm,
3.2.3 Two-Dimensional Model: Arbitrary
T = 298 K, P = 67 Pa, and γwall = 10-4. The lengths of the
Tube Sections
tube sections are L1 = 14R, L2 = 1R, and L3 = 35R.
The next step in modelling complexity is a two-

10
0 dimensional diffusion-reaction model that can
capture radial concentration gradients. Radial
concentration gradients may be important when
surface catalytic efficiencies are high. Two-
dimensional effects may also become significant
near material discontinuities along the wall or
-1
10 near the tube endplate.
For materials with high catalytic efficiencies

U(x)
(γ → 1), the atom velocity distribution function in

the vicinity of the surface may become
-2 significantly non-Maxwellian, affecting the atom
10
Ls = 0.5R impingement flux. Motz and Wise27 have derived
Ls = 1.0R a more accurate expression for the boundary
Ls = 2.0R
Ls = 4.0R
condition at a catalytic surface in a binary mixture
with low atom concentrations:
0 5 10 15 20 25
x ∂n γ n wall v
−D = . (35)
∂x wall (1 − γ 2) 4
Fig. 4 O-atom concentration profiles calculated for an
infinite three-section tube with samples of four different
lengths. Same conditions as in Fig. 3; γs = 10 .
-2
RTO-EN-AVT-142 11 - 13
Note that the 1 (1 − γ 2 ) correction term only changes the flux by more than 10% for catalytic efficiencies
greater than about 0.1.
As long as the atom concentration remains moderate – ratios of atom to molecule concentrations below ~1/5 –
the linear diffusion equation remains satisfactory; in tube-radius normalized cylindrical coordinates
∂ 2U 1 ∂U ∂ 2U
+ + 2 =0 . (36)
∂r 2 r ∂r ∂x
Wise and co-workers have presented a number of analytic solutions to this equation for different tube and
endplate configurations.25,27,31,36 Their approach has been extended to include both finite and infinite tubes
with N tube sections.60 The methodology is summarized below.
Using separation of variables, the solution satisfying Eq. 36 for each tube section has the form
U n (r , x) = Fn (r )G n ( x) = [c1 J 0 (kr ) + c 2Y0 (kr )][c 3 cosh(kx n ) + c 4 sinh(kx n )] , (37)
where x n = x − Ln −1 and L0 = 0 . The radial boundary conditions for each tube section are i) a symmetric
concentration profile about the centreline (zero concentration gradient) and ii) first order atom loss to the tube
walls.
∂U n ∂U n U (1, x n )
=0 =− (38a,b)
∂r 0 , xn
∂r 1, x n
δn
with δ n = 4 D (1 − γ n 2) γ n v R .
Application of these boundary conditions leads to
∑ [A cosh(k n ,i x n ) + Bn ,i sinh (k n ,i x n )]J 0 (k n ,i r ) ,

∞
U n (r , x n ) = n ,i (39)
i =1
where the kn,i are the positive roots of J 0 (k n ,i ) = δ n k n ,i J 1 (k n ,i ) .
To solve for the remaining constants, two axial boundary conditions are required for each tube section. For
the first tube section, uniform concentration at the entrance gives
U 1 (r ,0 ) = 1 . (40)
For the final tube section, two choices are available for finite and infinite tubes, respectively:
∂U N U (r , ∆L N )
U N (r , ∞ ) = 0 =− . (41a,b)
∂x r , ∆L N
δ end
Additional continuity boundary conditions are imposed at the interfaces between neighbouring tube sections:
11 - 14 RTO-EN-AVT-142
∂U n ∂U n +1
U n (r , ∆Ln ) = U n +1 (r ,0 ) = . (42a,b)
∂x r , ∆Ln
∂x r ,0
Application of these boundary conditions results in the following relations:
2 (43)
A1,i =
[
1 + δ k k1,i J 1 (k1,i )
1
2 2
1, i
],
δ ′k N ,i sinh( k N ,i ∆L N ) + cosh(k N ,i ∆L N )
B N ,i = − AN ,i , or B N ,i = − AN ,i . (44a,b)
δ ′k N ,i cosh(k N ,i ∆L N ) +sinh(k N ,i ∆L N )
(45)
∞
sinh(k n ,i ∆Ln ) ∞
An ,i = cosh(k n ,i ∆Ln )∑ Rn (i, j )An +1, j − ∑k n +1, j

Rn (i, j )Bn +1, j
j =1 k n ,i j =1
(46)
∞
cosh(k n ,i ∆Ln ) ∞
Bn ,i = − sinh(k n ,i ∆Ln )∑ Rn (i, j )An +1, j + ∑k n +1, j Rn (i, j )Bn +1, j

j =1 k n ,i j =1
with
2k n ,i k n +1, j [δ n − δ n +1 ]J 1 (k n +1, j ) . (47)

R n (i, j ) =
[k 2
n +1, j
−k 2
n ,i
][1 + δ 2
n
k 2
n,i
]J 1
(k n ,i )
These relations can be represented as a set of matrix equations that can be evaluated numerically to a specified
accuracy (i.e., to the Ith approximation of the infinite series).60
Figure 5 shows examples of the 2-dimensional O-atom concentration profile computed for a room-
temperature quartz tube containing a catalytic sample of γs = 1, 0.1, or 0.01. The concentration along the
centreline (r = 0) varies smoothly with axial location, even in the vicinity of the highly catalytic sample.
Strong radial gradients are observed in close proximity to the highly catalytic wall.
0 0 0
10 10 10
-1 -1 -1
10 10 10
U(r,x) U(r,x) U(r,x)

-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
γs = 1 0.0
γs = 0.1 0.0 γs = 0.01 0.0
0.2 0.2 0.2
0 0.4 0 0.4 0 0.4
5 0.6 5 0.6 5 0.6
10 10 10
15 0.8 r 15 0.8 r 15 0.8 r
x 20 x 20 x 20
25 1.0 25 1.0 25 1.0
Fig. 5 O-atom concentrations calculated with a 2-dimensional three-section tube model; R = 1.1 cm,
T = 298 K, P = 67 Pa, and γwall = 10 . The lengths of the tube sections are L1 = 14R, L2 = 1R, and L3 = ∞.
-4
RTO-EN-AVT-142 11 - 15
The radial variation in atom concentration at the sample

1.0 midpoint location (x = 14.5) is plotted in Fig. 6. Note that
the atom concentration near the centreline does not
U(r,14.5)/U(0,14.5)
0.8 change rapidly in the radial direction, even for the fully
catalytic sample. The radial distortion of the atom
concentration is rapidly dissipated before and after the
0.6 sample along the axial direction.
0.4 -3
γs = 10 4.3 Species Detection
-2
γs = 10
0.2 -1
γs = 10 In LIF detection, a tunable laser is used to generate light
γs = 1 at a wavelength resonant with a particular electronic
0.0 transition in the atom or molecule of interest. Absorption
0.0 0.2 0.4 0.6 0.8 1.0 of this light excites the target species to a higher-energy
r electronic state. The excited species subsequently emits
Fig. 6 Radial variations of the O-atom photons as it drops back down to a lower energy state.
concentration at the axial location of the The intensity of this fluorescence is directly proportional
sample midpoint (x = 14.5); same parameters to the initial number of ground state species in the gas
as in Fig. 5. volume excited by the laser, provided that competitive
losses from the upper state, such as ionization, amplified
spontaneous emission, and collisional quenching, are minimized.
For the detection of ground-state atomic oxygen and atomic nitrogen, Pallix and Copeland59 applied two-
photon LIF schemes that were originally developed at SRI International61,62 and have seen wide application in
combustion diagnostics. More recently, an alternate two-photon scheme for N-atoms has been used.63 The
relevant energy level diagrams for oxygen and nitrogen are shown in Figs. 7 a,b,c.64
The oxygen atom ground state has three fine structure components (for J = 2, 1, or 0) that are populated at
thermal equilibrium according to the Boltzmann distribution. The J = 2 component has the highest relative
population over the temperature range of the experiment, ranging from about 74% at 300 K to 60% at 1500 K.
Ultraviolet radiation near 226 nm is used to excite the J = 2 component of ground state oxygen via the
3 p 3P1, 2 , 0 ← 2 p 3P2 two-photon transition. The excited state has a radiative lifetime of 33.3 ± 1.7 ns and an O2
quenching constant of (7.8 ± 0.8) × 10-10 cm3 s-1.65 Fluorescence occurs via the 3 p P1, 2 , 0 → 3s S 1 transitions.
3 3 0
The three fine structure components of the 3 p P state are within 1 cm-1 of each other. Thus, the fluorescence
3
wavelength is about 845 nm for radiative transitions originating from any of the three upper states.
Ground-state nitrogen atoms can be detected using two different two-photon schemes. The energy level
diagram for the traditional scheme is shown in Fig. 7b.64 The 3 p 4 D70 / 2 ← 2 p 3 4 S 30/ 2 two-photon transition is
excited using ultraviolet radiation at 211 nm. The radiative lifetime of the upper state is about 43 ± 3 ns and
the N2 quenching constant is (4.6 ± 0.6) ×10-10 cm3 s-1.66 Fluorescence near 868 nm is obtained from the
3 p 4 D70 / 2 → 3s 4 P5 / 2 radiative transition. The 3 p 4 D 0 state has four fine structure components (J = 1/2, 3/2, 5/2,
and 7/2) and the 3s 4 P state has three fine structure components (J = 1/2, 3/2, and 5/2). Therefore, a manifold
4 0
of allowed transitions (∆ J = 0, ±1) exist between these two states. The advantage of preparing the 3 p D7 / 2
state is that the entire fluorescence signal is concentrated at one wavelength because only the ∆J = -1
11 - 16 RTO-EN-AVT-142
transition is allowed. Collisional transfers among the

109837.02 cm-1 O+ multiplets of the excited nitrogen state are possible,67
however at pressures below ~0.5 Torr (67 Pa) this transfer
88631.32 cm-1 0
88631.16 cm-1 2 2p33p 3P is probably not significant. Additionally, the most
88630.60 cm-1 1 probable intramultiplet transfer is from the 7/2 to the 5/2
λ ≈ 845 nm state, which fluoresces predominantly near 868 nm and
76794.98 cm-1 1 still contributes to the measured signal.
2p33s 3So
The energy level diagram for a second two-photon LIF
λ ≈ 226 nm scheme for N-atom detection is given in Fig. 7c.64 The
226.99 cm-1 0
158.29 cm-1 2p4 3P 3 p 4 S 30/ 2 ← 3 p 3 4 S 30/ 2 two-photon transition is excited by
1
0.00 cm-1 2 ultraviolet radiation near 207 nm. The radiative lifetime of
(a)
the 3 p 3 4S 30/ 2 upper state is about 26 ± 2 ns with an N2
quenching rate constant of (6.7 ±0.9) × 10-11 cm3 s-1.63
Fluorescence occurs near 745 nm via the
117225.4 cm-1 N+ 3 p 4 S 30/ 2 → 3s 4 P50/ 2 , 3 / 2 , 1 / 2 radiative transitions. This scheme
94881.79 cm-1 7/2 has the advantages that the collisional quenching rate
94830.86 cm-1 5/2 2p23p 4Do constant is significantly smaller and that the fluorescence
94793.46 cm-1 3/2 is at shorter wavelengths than the traditional scheme.
94770.85 cm-1 1/2
λ ≈ 868 nm
83364.62 cm-1 5/2 4.4 System Configuration
83317.83 cm-1 3/2 2p23s 4P
83284.07 cm-1 1/2 Figure 8 shows a diagram of the diffusion tube side-arm
λ ≈ 211 nm reactor facility at NASA Ames Research Center. Similar
0.00 cm-1
facilities have been set up at SRI International.
3/2
2p3 4So
(b)
The reactor consists of a glass main arm and a quartz side
arm joined at a quartz cross. Approximately 47 cm (18
inches) of the side arm tube is enclosed in a three-section
N+ Lindberg electric clam-shell furnace capable of reaching
117225.4 cm-1 ~1500 K. Test samples in the form of tubes are placed
2p23p 4So inside the side-arm tube near the center of the furnace.
96750.81 cm-1 3/2
Gases are introduced into the main arm upstream of the
λ ≈ 745 nm
cross. By a simple adjustment of valves (see Fig. 8) the
83364.62 cm-1 5/2
83317.83 cm-1 3/2
side arm can be turned into either a dead-end diffusion
2p23s 4P tube or a flow tube. In both cases gases are evacuated by a
83284.07 cm-1 1/2
Varian V-90 turbomolecular pump backed with a
λ ≈ 207 nm mechanical pump. A fluorinated pump oil (Fomblin) is
0.00 cm-1 3/2 used in the mechanical pump to avoid explosion hazards
2p3 4So when pumping oxygen.
(c)
Fig. 7 Energy level diagrams of relevant two- Gas flow is measured using calibrated Tylan flow meters
photon LIF schemes: (a) O-atom excitation at and gas pressures are measured using a 1 Torr Baratron
226 nm, (b) N-atom excitation at 211 nm; (c) capacitance manometer gauge. High purity oxygen and
N-atom excitation at 207 nm. nitrogen gases (99.998%) are used directly from their
RTO-EN-AVT-142 11 - 17
cylinders. Typical gas flow rates are 10-20 standard cubic centimeters per minute (sccm) and typical reactor
pressures are ~40 Pa (0.3 Torr). A microwave discharge is positioned on the main arm upstream of the cross.
The discharge operates at 2450 MHz and is powered by an Opthos power supply. Molecular oxygen or
nitrogen is introduced into the reactor upstream of the discharge. The discharge is lit using a Tesla coil.
During operation, the discharge is operated at powers between 40 W and 100 W, with 90 W being typical, and
is convectively cooled using compressed air. Chemical titration experiments indicate that oxygen dissociation
fractions are in the 1-3% range, with nitrogen dissociation about one order of magnitude lower.
The O + O2 + Z → O3 + Z gas phase three-body reaction becomes significant at pressures above about 65 Pa
(0.5 Torr). To eliminate the influence of this gas phase reaction on species concentration profiles, the total
reactor pressure is always kept below 53 Pa (0.4 Torr). The O + O + Z → O2 + Z reaction, although ~5 times
faster than the ozone producing reaction at room temperature, is less important because of the low O-atom
concentrations present in the system.
678 nm,
678 nm, 339 nm, 678 nm,
339 nm 226 nm THG 339 nm SHG 678 nm dye 532 nm Nd:YAG
crystal crystal laser laser
dump
to vacuum
pumps
Cu mesh trap
226 nm
valve clam-shell tube furnace valve
dielectric
energy
mirror
meter
Galilean
telescope sample pressure tap
microwave
furnace inserts discharge
filters and lenses PMT4 PMT3 PMT2 PMT1
mass flow
computer meters
preamplifiers O2 in
boxcar delay
computer
integrators generator
interface
Fig. 8 Schematic diagram of side-arm reactor facility at NASA Ames Research Center. The indicated
wavelengths are for O-atom detection.
11 - 18 RTO-EN-AVT-142
Tunable red light is provided by a Continuum ND60 dye laser pumped by the second (532 nm) harmonic of a
Continuum NY81 Nd:YAG laser. For O-atom detection, red light is generated near 678 nm using LDS 698
laser dye. For N-atom detection red light is generated near 621 nm using Rhodamine 640 laser dye, or near
633 nm using DCM laser dye. All three dyes are dissolved in HPLC grade methanol. Ultraviolet radiation for
atom excitation is obtained by frequency tripling the red light using two beta-barium borate (BBO) crystals for
second and third harmonic generation. The same BBO crystals can be used for all three excitation schemes.
Residual red light in the beam is separated from the desired UV radiation using several dichroic mirrors. The
beam is directed through a down-collimating Galilean telescope, through a quartz window oriented at
Brewster’s angle and down the centerline of the side arm tube. The laser pulse energy is monitored with a
Molectron J3-09 pyroelectric energy meter that intercepts the beam as it exits the reactor through a similar
Brewster’s angle window located on the cross.
Fluorescence is detected at normal incidence by four red-sensitive Hamamatsu R636 photomultiplier tubes
(PMTs) with gated C1392 sockets. Gated operation reduces saturation from background light. The PMT
anode bias voltages of -1300 V are provided by four high voltage DC power supplies. A single custom built
+250 V DC power supply is used to supply the PMT gate voltages. A periodic external trigger signal removes
this gate voltage, allowing the PMT anode biases to be applied.
A wedge shaped refractory insert allows optical access while the furnace is closed and operating. This
refractory insert has 1 cm diameter optical ports lined with quartz tubes to view the centerline of the side-arm
tube every 8.9 cm (3.5 inches) along its axis. The pseudo-collimated light emerging from the ports is
collected by the PMTs, which are fitted with narrow-band interference filters (~3 nm bandwidth) centered at
845 for O-atom detection and 745 nm for N-atom detection. The PMTs are labeled 1 through 4 in Fig. 8, with
PMT1 the closest to the side-arm opening. Test samples are positioned between the regions viewed by PMT2
and PMT3.
A Stanford Research Systems data acquisition system is used along with a personal computer to collect,
process, and store output signals from the energy meter and the four PMTs. The data acquisition system
consists of an SR240 preamplifier, five SR250 boxcar integrators, and an SR245 computer interface all
housed in an SR280 mainframe. The Nd:YAG laser, the gated PMTs and the data acquisition system are
triggered at 10 Hz and timed with respect to one another using a Stanford Research System DG535 digital
delay and pulse generator.
4.5 Experimental Procedures
4.5.1 PMT Normalization
The fluorescence detection sensitivity at the four PMTs differs for a number of reasons: PMT sensitivity, filter
transmission characteristics, and alignment of the detection optics and laser beam alignment. These
differences must be accounted for when using the relative PMT signals at the different positions as proxies for
the relative atom concentrations. Relative PMT sensitivities can be trivially determined when detecting stable
species, by simply filling the tube with a uniform stagnant gas. However, this is not possible for reactive
species such as O or N atoms since they are lost on the walls and hence not uniformly distributed.
An initial approach to this problem was to measure the relative PMT response to the thermal background of
the heated furnace; however, this proved unsatisfactory for a number of reasons. One problem is that thermal
RTO-EN-AVT-142 11 - 19
gradients exist in the furnace walls since the embedded heating elements are hotter than the surrounding
surfaces. It is not assured that all four PMTs view similar backgrounds. Additionally, the LIF signal is
collected from a line source (the excited atoms in the laser beam path) while thermal radiation appears as a
disk source, consequently focusing differently on the PMT cathodes. Also, the convolution of the filter
transmission band with a uniform thermal radiation background does not capture differences in LIF sensitivity
due to the wavelength alignment of the filter band pass with the atomic line emissions.
An alternate approach, currently in use, is to use LIF signals collected with the reactor configured as a flow
tube as normalizing factors. This approach removes uncertainties introduced by using an emission source of
different geometry and wavelength characteristics than the atom fluorescence to be detected. The side-arm
can be turned into a flow tube by simple adjustment of two valves, as shown in Fig. 8.
Fast flow minimizes differences in atom concentrations between the different PMT locations because the
influence of surface reactions is greatly diminished. However, surface losses cannot be completely eliminated
and the pressure gradients that drive the flow also cause axial variations in number density. In order to use
LIF measurements made in flow tube mode as normalizing factors for fluorescence detection sensitivity, the
collected signals must first be corrected for both wall losses and pressure effects. A correction procedure is
described below
The relationship between mass flow m and the pressure (P1 , P2 ) at two axial locations (x1 , x 2 ) in a uniform
tube of radius R can be described using the Hagen-Poiseulli relation including a correction for slip at the wall:
 πR 4 M  P1 − P2  P1 + P2 µ 8πℜT 
m =    +  ; (48)
 8µℜT  x1 − x 2  2 R M 
ℜ is the universal gas constant, M molar mass, and µ is the viscosity. The sufficiency of this relation in
describing the flow can be checked by measuring pressures on either side of the furnace, calculating the mass
flow rate, and comparing the value to that measured with the mass flow meter. Generally, discrepancies
between calculated and experimental mass flow rates are less than 10%. Combining mass conservation and
the perfect gas law allows calculation of the bulk flow velocity at each PMT position:
m ℜT
v= . (49)
πR 2 PM
The ratio of the actual atom number densities viewed by two PMTs at different axial locations can be
approximated by a simple one-dimensional flow analysis leading to
n(x 2 ) P2  γ v ( x 2 − x1 ) 
= exp −  , (50)
n(x1 ) P1  R(v 2 + v1 ) 
where the wall loss probability must initially be estimated.
The pressure gradient has the additional effect of changing the fluorescence yield and the fraction of
fluorescence collected at each position. The ratio of measured fluorescence at two different PMT locations
viewing the same atom concentration is approximately
11 - 20 RTO-EN-AVT-142
 τg 
1 − exp − 
f 2 τ eff , 2  τ 
 eff , 2 
= , (51)
f 1 τ eff ,1  τg 
1 − exp − 
 τ 
 eff ,1 
where τg, is the gate length of the boxcar integrator, and the effective lifetime of the excited state is given in
terms of the natural radiative lifetime τ rad and the quenching rates k q ,i of the collision partners by
1 1
τ eff
=
τ rad
+ ∑n k
i
i q ,i . (52)
In practice, the fluorescence collection corrections amount to only 1 or 2 %, while the corrections for wall loss
and pressure differences can approach 25% between PMT1 and PMT4. Because of their sensitivity to beam
and PMT alignment, the fluorescence detection sensitivities are re-measured periodically during every testing
program. A consistent value of loss probability can be obtained by iterating between flow tube and diffusion
tube measurements several times.
4.5.2 Measurement Sequence
The data collection for a typical test consists of acquiring approximately 200 bins of data on five channels
(i.e., the energy meter and four PMTs), with each bin containing the mean signal from 10 laser shots. The
appropriate baseline offset is subtracted from each bin, and the average signal and associated standard
deviation are calculated for each channel. The average PMT signals and the standard deviations are then
corrected for the relative detection sensitivities at the different PMT locations. The final corrected PMT
signals are directly proportional to the species concentration at each detection location and are fit by analytic
solutions to extract recombination coefficients.
In order to extract recombination coefficients for samples of interest, the atom loss on the quartz reactor walls
must either be measured, assumed known, or estimated during the data fitting procedure. Typically, atom
recombination coefficients on the quartz side-arm tube are measured as a function of temperature several
times before and after testing other materials, to capture any systematic changes that have been caused by
temperature cycling or contamination from out-gassing samples.
For the geometry and dimensions of the side-arm reactor facility at NASA Ames Research Center, the simple
one-dimensional exponential solution for an infinite tube (Eq. 31) provides a sufficiently accurate description
of the atom decay along the tube centreline. Fits of ln (U ( x) ) vs x to the measured data at the four PMT
locations yields a slope that is directly related to the recombination coefficient (via Eq. 28a.)
With quartz recombination coefficients in hand, data measured with a sample inside the side-arm can be fit
using solutions to a three section tube model, with the recombination coefficient for the first and third sections
fixed at the experimental value. Again, the infinite tube solutions are sufficient for the final side-arm section.
In practice, centreline solutions to the 2-dimensional diffusion model are used to match experimental data
(Eq. 39), though the 1-dimensional three-section tube model provides similar results.
RTO-EN-AVT-142 11 - 21
4.6 Examples of Experimental Results
The majority of the side-arm reactor data generated for a variety of TPS materials during the late 1990’s is
summarized in publicly available NASA reports and AIAA conference papers.9-11,60 Additional measurements
are contained in proprietary documents.
Figures 9a and b show experimental data and curve fits used to extract O-atom loss probabilities on quartz and
platinum surfaces, during a recent side-arm reactor study to support shock tube experiments conducted in
dissociated CO2.68 Table 2 lists the number of independent determinations, the average room temperature
values, and the standard deviations of the O-atom loss probabilities obtained for quartz, stainless steel 304,
Constantan, Chromel, and platinum samples. The standard deviations reflect the reproducibility of the
experimental measurements, which were made over the course of several days for each material.
The measured ~1 × 10-4 value of O-atom loss probability on room temperature quartz is well within the wide
range of experimental values that have been reported in the literature. For example, Berkowitz69 cited 15
different determinations that produced values from (0.17 - 7.1) × 10-4 for measurements on silica, quartz,
Pyrex and Vycor samples, and Marschall60 cited 8 different measurements made on quartz surfaces that span
the range (0.27 - 2.5) × 10-4.
Table 2: Room-Temperature O-atom Loss Probabilities

Material Experiments γO, × 10-3
Quartz 11 0.098 ± 0.013
SS304 (Cr 18%, Ni 10%, Fe balance) 11 16 ± 7
Constantan (Ni 45%, Cu balance) 10 46 ± 17
Chromel (Cr 10%, Ni Balance) 9 6.8 ± 1.3
Platinum (99.95% Pure) 10 5.3 ± 1.1
0.0 Quartz Platinum

0.0
-0.5
-0.5
ln( U(x) )
ln( U(x) )
-1.0
-1.0
-1.5
-1.5
(a) -2.0
(b) Sample
0 5 10 15 20 25 0 5 10 15 20 25
x x
Fig. 9 Measured relative O-atom concentrations and the analytic curve fits used to extract room-
temperature O-atom recombination coefficients: a) quartz data fit to Eq. (31) giving γ O , quartz = 1.1 × 10-4, and
b) platinum data fit with centerline solution of Eq. (39) with γ O , quartz fixed at 1.0 × 10-4 giving γ O , Pt = 5.7 × 10-3.
From reference 68.
11 - 22 RTO-EN-AVT-142
The value of (16 ± 7) × 10-3 measured for stainless steel 304 can be compared to (70 ± 9) × 10-3 reported for
the same material by Mozetič and Zalar,57 (7.1 ± 1.4) × 10-3 reported for “steel” by Cauquot et al.55, and
(10 ± 4) × 10-3 reported for Fe by Melin and Madix.15 It is highly likely that most metallic surfaces are
covered by thin layers of oxide scales. Dickens and Sutcliffe70 give a value of 8.5 × 10-3 for Fe2O3 and Guyon
et al.54 give (15 ± 3) × 10-3 for Fe3O4. The value of (5.3 ± 1.1) × 10-3 measured for platinum can be compared
to (4 ± 3) × 10-3 given by Hartunian et al.,71 and (14 ± 4) × 10-3 reported by Melin and Madix.15
No O-atom recombination coefficients have been reported for Constantan or Chromel samples. However, for
copper and nickel samples, Melin and Madix15 measured (15 ± 5) × 10-3 and (17 ± 5) × 10-3, Myerson72
measured 31 × 10-3 and 8.5 × 10-3, and Hartunian et al.71 reported (150 ± 50) × 10-3 and (40 ± 20) × 10-3. The
more recent measurement of Cauquot et al.55 for copper is (25 ± 5) × 10-3. Data cited by Melin and Madix,15
as well as the measurements of Dickens and Sutcliffe,70 suggest that copper oxides are about one order of
magnitude more catalytic than nickel oxides. Thus it is perhaps reasonable that the Cu containing Constantan
alloy is more catalytic than the Ni-rich Chromel alloy.
A major source of discrepancies among atom recombination coefficients reported in the literature for
nominally identical materials can safely be attributed to differences in the surface morphology (e.g.,
microscopic roughness) and surface composition (e.g., oxidation). In most studies the condition of the surface
before and after testing is not characterized.
During recent measurements of surface catalytic efficiency on HfB2/SiC and ZrB2/SiC monolithic composites,
experimental results clearly indicated that the catalytic properties were changing during testing.73 Each
sample was tested on three different days, at several discreet temperatures, with multiple measurements at
each temperature. During the first test run measurements were made at 295 K, 473 K, and 673 K, during the
second test run these measurements were repeated and extended to 923 K, and during the third test run
A-8, Sample 2
-1
10
0.016
ation Coefficient
A-8 Tm B O = 723 K
Oxygen Recombination Coefficient
2 3
0.014
-2
10 0.012
0.010
Oxygen Recombin
0.008
-3
10
0.006
0.004
Quartz
-4 0.002
10 1000
First Run 0.000 0 900
Second Run 20 B2O3 Tmelt: 723 K 800
Third Run 40 700
60 600
-5 80 500 ,K
10 Tim 100 re
300 400 500 600 700 800 900 1000 e, m 120 400 ratu
in. e
140 300 mp
Temperature, K
160 200 Te
a) b)
Fig. 10 O-atom recombination coefficient as a function of temperature for ZrB2-20 vol. % SiC: a) sequential
experiments on three different days; b) measured values as a function of time during the second test
sequence. From reference 73.
RTO-EN-AVT-142 11 - 23
measurements at all four temperatures were repeated once more. The reactor was not opened between runs
and the samples remained continuously under vacuum.
Figures 10a and b show the results. During the first test, room temperature recombination coefficients were
~1-2 × 10-3. Upon heating to 473 K, the recombination coefficients rise by an order of magnitude to
~1-2 × 10-2. With further heating to 673 K they increase slightly to the ~2-3 × 10-2 range. During the second
test sequence, the measured recombination coefficients at each of these three temperatures are basically the
same within some experimental scatter. When the sample is heated to 923 K, a steady decrease in catalytic
efficiency with time is observed as indicated by the downward arrow in Fig. 10a. The trend is more clearly
shown in Fig. 10b, which plots the measured recombination coefficient as a function of time and temperature.
During the third test, the sample’s catalytic efficiency for O-atom recombination is drastically reduced, by one
to two orders of magnitude, to the point that the recombination coefficients of the test specimen cannot be
distinguished from those of the quartz tube at the two lowest test temperatures.
This change in catalytic efficiency is an indication that the surface of the test specimen is changing. The most
likely process is surface oxidation, an interpretation supported by energy dispersive X-ray analysis (EDX) and
X-ray photoelectron spectroscopy (XPS) conducted on identical samples exposed to similar oxidizing
environments and heating cycles.73 SiC oxidation under the oxygen and temperature conditions of the side-
arm reactor during catalysis testing should be negligible; however, oxidation of the diboride to form zirconia
and B2O3 is possible and the large drop in surface catalytic efficiency correlates well with the first temperature
excursion above the B2O3 melting point of ~723 K during the second test sequence.
The side-arm reactor has also been used to investigate catalytic phenomena in a qualitative way. In a study by
Copeland et al.,74 a nitrogen discharge was titrated by adding NO downstream to produce O-atoms by the fast
gas phase reaction N + NO → N2 + O (k = ~3 × 10-11 cm3 s-1). As the NO flow is increased, the concentration
of N-atoms falls and the concentration of O-atoms rises. At the titration endpoint all N-atoms produced by the
discharge are consumed. Further additions of NO to the reactor result in the appearance of NO in the gas-
phase. Laser induced fluorescence was used to simultaneously monitor N-atom and NO concentrations at
locations downstream of the titration inlet.
Downstream of the NO inlet only surface reactions can change the concentrations of N and O atoms
significantly, because all other gas-phase reactions are too slow under the low pressure conditions of the
experiment. In a flow tube environment surface reactions are minimized and the decay of N-atom
concentration was observed to be linear with NO addition. However, in a diffusion tube environment, a
greater than linear dependence on NO addition was consistently observed during the initial portion of the
titration curve, and N-atom losses were greater than expected when O-atoms were present in comparable
concentrations. The likely explanation for this observation is that O + N recombines to NO on the surface
and that the desorbing NO molecule quickly consumes another N atom via the titration reaction. This cycle
effectively enhances N-atom loss from the gas-phase. Kurotaki has shown how this mechanism is important
for simulating measured heat flux data on the OREX flight experiment using a mechanistic surface chemistry
model similar to that in Table 1.75
Experiments have also been performed to look for qualitative indications of the relative importance of the
O + O and O + CO surface reactions, as might be important for Mars atmospheric entries.68,76 In these studies,
LIF was used to look for changes in either the O-atom or CO concentration with the addition of the other
species to the reactor. Experiments were performed in O2/O/CO/He gas mixtures, on quartz, and a variety of
metal samples, including Inconel 617 and all the metals listed in Table 2.
11 - 24 RTO-EN-AVT-142
Figure 11 shows the results of these types of experiments on quartz.76 No significant or systematic changes
are observed in the concentration profiles of either gas species with the introduction of the other. If the
O + CO reaction were operating efficiently, one would expect to see the exponential O-atom profile along the
side-arm decay more rapidly in the presence of CO and the uniform CO concentration change to a decaying
profile in the presence of O-atoms. However neither of these changes is seen in Figs. 11a or b.
Similar measurements with Inconel 617 at room temperature,76 and SS304, Constantan, and Chromel at room
temperature and 250 °C,68 also suggest that the O + CO reaction is not competitive with the O + CO reaction
on these materials under the pressure, temperature, and gas composition of these experiments. For platinum at
room temperature the same result is found. At 250 °C the well known surface catalyzed O2 + 2CO oxidation
reaction was clearly observed, and no additional effect of O-atoms could be discerned.
1.2 1.5
Detecting O-Atoms De tecting CO
Relative O-Atom Concentration
1.0
Relative CO Concentration
0.8 295 K 1.0
0.6
623 K
0.4 0.5
With CO With O
0.2
Without CO Without O
0.0 0.0
0 4 8 12 16 0 4 8 12 16
Distance/Tube Radius Distance/Tube Radius
a) b)
Fig. 11 a) Relative O-atom concentrations with and without CO present; b) relative CO concentrations with
and without O-atoms present. From reference 76.
4.7 Accuracy, Reproducibility, and System Improvements
The “absolute” accuracy of measured loss coefficients is problematic to define. There are no reference
standards that can be used with any confidence. Examples can be found in the literature where certain
materials – typically copper or silver surfaces – have been used as reference materials with “known” values
for recombination coefficients. However, any cursory survey of the literature quickly uncovers numerous
measured values for each material, many of which do not agree within their stated uncertainties. This should
not be surprising. First, loss coefficients obtained are “effective values” in the sense that both the intrinsic
material chemistry and the microscopic reactive surface area contribute to the measured value. For aerospace
applications, this highlights the importance of conducting measurements on test samples that reproduce as
closely as possible the material source and fabrication techniques used in the actual TPS components.
Second, loss coefficients extracted from the experimental data are closely coupled to the surface chemistry
and gas flow assumptions used during data analysis. Because these vary with different techniques, only
measurements made using the same experimental and data analysis approaches can really be expected to agree
RTO-EN-AVT-142 11 - 25
closely. A broad (and admittedly subjective) review of published measurements suggests that any individual
experimental result is probably uncertain by a factor of at least 3 when taken to represent a particular type of
material such as “steel” or “silica”.
For a chemically stable low catalytic efficiency material such as quartz, measurements made in rapid
succession in the diffusion-tube side-arm reactor generally produce recombination coefficients that differ by
less than 10%. Exceptions occur at high temperatures where the decreasing atom number density and the
increasing thermal background from the furnace degrades the signal-to-noise ratio, or when initial heating or
discharge operation produces transients associated with the removal of surface contaminates such as water.
For materials with higher catalytic efficiency the scatter in successive measurements is typically higher at a
given temperature because more atoms are removed from the gas phase, lowering the LIF signal. However,
this scatter still rarely exceeds 25%. Measurements made on different days, especially after temperature
cycling, can yield values that differ by 50% or more, as reflected in the data listed in Table 2 for the metals.
These larger changes are typically due to the evolution of the sample in response to the test environment, but
may also contain contributions from various subtle changes in the experiment; e.g., optical alignments,
excitation beam shape, etc.
Many improvements can be made in both the experimental and data analysis techniques to extend the
experimental range of the apparatus and to make the resulting experimental data more useful to the
aerothermal and computational fluid dynamics (CFD) modelling communities.
A clear improvement in the present data analysis can be achieved if the experimental data is fit using the same
computational tools that are used to simulate high-enthalpy test and flight environments. Such a procedure
would ensure consistency in transport property approximations (i.e., diffusion coefficients) and allow for
experiments in multispecies gas mixtures at pressures where gas-phase reactions operate in parallel with
surface reactions. It would also enable more accurate modelling of the side-arm and sample geometries, and
account for wall temperature distributions. The first steps along this path were taken recently, when the CFD
code DPLR was applied to simulate concentration gradients in O2/O/CO/He gas mixtures inside the side-arm
reactor.68 DPLR77 is a parallel multiblock finite-volume code that solves the Navier-Stokes equations
including finite-rate chemistry and the effects of thermal nonequilibrium. DPLR, along with the CFD code
LAURA,78 are the main aerothermal analysis tools used by NASA for planetary entry vehicles.
On the experimental side, the two main system improvements would be a quicker and more accurate method
of measuring the relative fluorescence detection sensitivities at the four PMTs, and extension of the
experimentally accessible high temperature range. For the former task, the calibration scheme of Döbele and
coworkers appears promising.79-81 They use two-photon excitation of rare gas atoms at wavelengths very
close to the excitation wavelengths for O and N atoms to generate a fluorescence volume with the similar
geometric and emissive characteristics. Combined with knowledge of the filter transmission curves and PMT
spectral sensitivity at the pertinent fluorescence wavelengths, the relative fluorescence detection sensitivities
at the four PMTs could then be determined simply by filling the reactor with the appropriate gas (Xe for O-
atom detection and Kr for N-atom detection).
Currently, the upper temperature limit for usable experimental measurements in the side-arm reactor is about
850 °C. This limit is determined by the signal-to-noise ratio in the LIF signal. LIF signals decrease with
temperature for a number of reasons, including lower gas densities, greater surface loss rates and in the case of
O-atoms, depletion of the J = 2 fine structure level population in the electronic ground state. However, in
general, signal strength is not the limiting factor; rather it is the fluctuating thermal background emanating
from the furnace at high temperatures that overwhelms the atom signal. This background cannot be filtered
11 - 26 RTO-EN-AVT-142
out since it arises in the near infrared region at the same wavelengths transmitted by the narrow band filters on
the PMTs.
One solution to this problem is to seek alternate excitation schemes that fluoresce in the visible range.
Preliminary experiments were conducted exciting the 4 p P2,1, 0 ← 2 p P2 two-photon transition at 200.6 nm.
3 3
Fluorescence from the excited state via the 4 p P2 ,1, 0 → 3s S 1 transition generates light near 437 nm. At this
3 3 0
wavelength, gas temperatures above 1200 °C can be reached before background thermal radiation levels reach
comparable levels. The two-photon cross-section is a factor of four smaller and the excited state lifetime a
factor of ~6 longer than in the traditional O-atom LIF scheme. Further, only ~10% of the total fluorescence is
emitted at 437 nm.82 Nevertheless, the signal to noise ratio observed in experiments at elevated temperatures
was significantly improved. Unfortunately, additional complications arise from the unintended photo-
dissociation of vibrationally excited O2 molecules in the Schumman-Runge bands above ~700 °C, making the
ultimate utility of this scheme doubtful.
A more promising approach is to combine the present LIF schemes with on-axis fluorescence detection.
Exploratory measurements demonstrate that this detection orientation eliminates almost all of the thermal
background interference, since the background field of view seen by the PMT is essentially the cold exit
window on the cross (Fig. 8). The trade off for this arrangement is the difficulty in achieving spatial
resolution along the tube centreline, which must be done either by spatially discriminating collection optics or
by bringing the excitation laser in through the side of the furnace at normal incidence to the side-arm tube.
5.0 CONCLUDING REMARKS

The experimental characterization and numerical modelling of surface reactions on TPS materials remain
challenging problems. Uncertainties in the catalytic component of aeroconvective heating contribute
significantly to the overall uncertainty in aerothermal heating predictions for many flight and test
environments. One ramification of this situation is the default assumption of fully-catalytic surfaces in TPS
sizing calculations. Another consequence is the difficulty in separating catalytic and turbulent contributions to
measured heating rates in turbulent hypersonic test environments.
Improvements in our capability to incorporate robust and validated models for high temperature surface
chemistry on TPS in hypersonic environments will require continued experimentation both in high-enthalpy
and low-enthalpy facilities. With continuing development of experimental techniques and data analysis
procedures, the diffusion tube side-arm reactor approach can be used to make useful contributes to this effort.
6.0 ACKNOWLEDGEMENTS
Many people have contributed to the work described in this lecture. The guidance and collaboration of Joan
Pallix, Doug Fletcher, and Doug Bamford are gratefully acknowledged. Special thanks are due the David
Stewart of NASA Ames Research Center who introduced me to the subject of surface catalysis, and to
Richard Copeland of SRI International who taught me about laser diagnostics. My work in this area has been
supported through NASA and AFOSR funding.
RTO-EN-AVT-142 11 - 27
7.0 REFERENCES
1
Rakich, J.V., Stewart, D.A., and Lanfranco, M.J., “Results of a Flight Experiment on the Catalytic
Efficiency of the Space Shuttle Heat Shield,” AIAA Paper 82-0944, 1982.
2
Stewart, D.A., Rakich, J.V., and Lanfranco, M.J., “Catalytic Surface Effects Experiment on the Space
Shuttle,” AIAA Paper 81-1143, 1981.
3
Stewart, D.A., Rakich, J.V., and Lanfranco, M.J., “Catalytic Surface Effects on Space Shuttle
Thermal Protection System during Earth Entry of Flights STS-2 through STS-5,” NASA CP-2283,
Hampton, VA, March 1983.
4
Masel, R.I., Adsorption and Reaction on Solid Surfaces, John Wiley & Sons, Inc., New York, 1996.
5
Deutschmann, O., Riedel, U., and Warnatz, J., "Modeling of Nitrogen and Oxygen Recombination on
Partial Catalytic Surfaces," Journal of Heat Transfer, Vol. 117, 1995, pp. 495-501.
6
Jumper, E.J. and Seward, W.A., "Model of Oxygen Recombination on Reaction-Cured Glass,"
Journal of Thermophysics and Heat Transfer, Vol. 8, 1994, pp. 460-465.
7
Daiß, A., Frühauf, H.-H., and Messerschmid, E.W., "Modeling Catalytic Reactions on Silica Surfaces
with Consideration of Slip Effects," Journal of Thermophysics and Heat Transfer, Vol. 11, 1997, pp.
346-352.
8
Natsui, F., Barbato, M., and Bruno, C., "Material-Dependent Catalytic Recombination Modeling for
Hypersonic Flows," Journal of Thermophysics and Heat Transfer, Vol. 10, 1996, pp. 131-136.
9
Stewart, D.A., “Determination of Surface Catalytic Efficiency for Thermal Protection Materials -
Room Temperature to Their Upper Use Limit,” AIAA Paper 96-1863, June 1996.
10
Stewart, D.A., “Surface Catalysis and Characterization of Proposed Candidate TPS for Access-to-
Space Vehicles,” NASA TM-112206, July 1997.
11
Stewart, D.A., “Surface Catalytic Efficiency of Advanced Carbon Carbon Candidate Thermal
Protection Materials for SSTO Vehicles,” NASA TM-110383, February 1996.
12
Laidler, K.J., Chemical Kinetics, McGraw-Hill Book Company, New York, 1965.
13
Smith, W.V., "The Surface Recombination of H Atoms and OH Radicals," Journal of Chemical
Physics, Vol. 11, 1943, pp. 110-125.
14
Melin, G.A. and Madix, R.J., "Energy Accomodation During Hydrogen Atom Recombination on
Metal Surfaces," Transactions of the Faraday Society, Vol. 67, 1971, pp. 2711-2719.
15
Melin, G.A. and Madix, R.J., "Energy Accomodation During Oxygen Atom Recombination on Metal
Surfaces," Transactions of the Faraday Society, Vol. 67, 1971, pp. 198-211.
16
Greaves, J.C. and Linnett, J.W., "Recombination of Atoms at Surfaces, Part 4 - Theory of Method and
Measurement of Atom Concentrations," Transactions of the Faraday Society, Vol. 55, 1959, pp.
1338-1345.
17
Greaves, J.C. and Linnett, J.W., "Recombination of Atoms at Surfaces, Part 5 - Oxygen Atoms at
Oxide Surfaces," Transactions of the Faraday Society, Vol. 55, 1959, pp. 1346-1354.
18
Greaves, J.C. and Linnett, J.W., "Recombination of Atoms at Surfaces, Part 6 - Recombination of
Oxygen Atoms on Silica from 20 °C to 600 °C," Transactions of the Faraday Society, Vol. 55, 1959,
pp. 1355-1361.
19
Greaves, J.C. and Linnett, J.W., "The Recombination of Oxygen Atoms at Surfaces," Transactions of
the Faraday Society, Vol. 54, 1958, pp. 1323-1330.
20
Linnett, J.W. and Marsden, D.G.H., "The Kinetics of the Recombination of Oxygen Atoms at a Glass
Surface," Proceedings of the Royal Society of London, A, Vol. 234, 1956, pp. 489-504.
21
Linnett, J.W. and Rahman, M.L., "Recombination of Atoms at Surfaces, part 13 - Oxygen Atoms on
ZnxFe3-xO4 and NixFe3-xO4," Transactions of the Faraday Society, Vol. 67, 1971, pp. 191-197.
22
Rahman, M.L. and Linnett, J.W., "Recombination of Atoms at Surfaces, part 10 - Nitrogen Atoms at
Pyrex Surfaces," Transactions of the Faraday Society, Vol. 67, 1971, pp. 170-178.
11 - 28 RTO-EN-AVT-142
23
Some Acid, Base, and Salt Surfaces," Transactions of the Faraday Society, Vol. 67, 1971, pp. 179-
182.
24
Some Metal and Alloy Surfaces," Transactions of the Faraday Society, Vol. 67, 1971, pp. 183-190.
25
Ablow, C.M., Motz, H., and Wise, H., "Diffusion and Heterogeneous Reaction. VII. Effect of
Different Catalyst Geometries," Journal of Chemical Physics, Vol. 43, 1965, pp. 10-14.
26
Ablow, C.M. and Wise, H., "Diffusion and Heterogeneous Reaction. VIII. Kinetic Considerations of
Surface Reactions," Journal of Chemical Physics, Vol. 46, 1967, pp. 3424-3428.
27
Motz, H. and Wise, H., "Diffusion and Heterogeneous Reaction. III. Atomic Recombination at a
Catalytic Boundary," Journal of Chemical Physics, Vol. 32, 1960, pp. 1893-1894.
28
Sancier, K.M., "Luminescence of Solids Excited by Surface Recombination of Atoms. V.
Quantitative Dependence of Luminescence Response on Oxygen- and Nitrogen-Atom Densities,"
Journal of Chemical Physics, Vol. 42, 1965, pp. 1240-1243.
29
Sancier, K.M., Fredericks, W.J., Hatchett, J.L., and Wise, H., "Luminescence of Solids Excited by
Surface Recombination of Atoms. II. Recombination Coefficients," Journal of Chemical Physics,
Vol. 37, 1962, pp. 860-864.
30
Sancier, K.M. and Wise, H., "Diffusion and Heterogeneous Reaction. XI. Diffusion Coefficient
Measurements for Gas Mixture of Atomic and Molecular Hydrogen," Journal of Chemical Physics,
Vol. 51, 1969, pp. 1434-1438.
31
Wise, H. and Ablow, C.M., "Diffusion and Heterogeneous Reaction. I. The Dynamics of Radical
Reactions," Journal of Chemical Physics, Vol. 29, 1958, pp. 634-639.
32
Wise, H. and Ablow, C.M., "Diffusion and Heterogeneous Reaction. IV. Effects of Gas-Phase
Reaction and Convective Flow," Journal of Chemical Physics, Vol. 35, 1961, pp. 10-18.
33
Wise, H., Ablow, C.M., and Sancier, K.M., "Diffusion and Heterogeneous Reaction. VI. Surface
Recombination in the Presence of Distributed Atom Sources," Journal of Chemical Physics, Vol. 41,
1964, pp. 3569-3573.
34
Wise, H., Ablow, C.M., and Schott, D.J., "Diffusion and Heterogeneous Reaction. V. Transition from
a Surface- to a Diffusion-Controlled Process during Atom Recombination," Journal of Chemical
Physics, Vol. 39, 1963, pp. 2063-2067.
35
Wise, H. and Wood, B.J., "Energy Transfer during Atom Recombination on Solid Surfaces," in
Rarefied Gas Dynamics; Vol. 1, edited by J. A. Laurmann, (Academic Press, New York, 1963), p.
470-477.
36
Wise, H. and Wood, B.J., "Reactive Collisions Between Gas and Surface Atoms," in Advances in
Atomic and Molecular Physics; Vol. 3, edited by D. R. Bates and I. Estermann, (Academic Press, New
York, 1967), p. 291-353.
37
Wood, B.J. and Wise, H., "Diffusion and Heterogeneous Reaction. II. Catalytic Activity of Solids for
Hydrogen-Atom Recombination," Journal of Chemical Physics, Vol. 29, 1958, pp. 1416-1417.
38
Wood, B.J. and Wise, H., "The Interaction of Atoms with Solid Surfaces," in Rarefied Gas Dynamics,
edited by L. Talbot, (Academic Press, New York, 1961), p. 51-59.
39
Wood, B.J. and Wise, H., "The Kinetics of Hydrogen Atom Recombination on Pyrex Glass and Fused
Quartz," Journal of Physical Chemistry, Vol. 66, 1962, pp. 1049-1053.
40
Wood, B.J. and Wise, H., "Kinetics of Hydrogen Atom Recombination on Surfaces," Journal of
Physical Chemistry, Vol. 65, 1961, pp. 1976-1983.
41
Kaufman, F., "Reactions of Oxygen Atoms," in Progress in Reaction Kinetics; Vol. 1, edited by C.
Porter, (Pergamon Press, 1961), p. 3-39.
42
Marshall, T., "Surface Recombination of Nitrogen Atoms Upon Quartz," Journal of Chemical
Physics, Vol. 37, 1962, pp. 2501-2502.
RTO-EN-AVT-142 11 - 29
43
Marshall, T.C., "Studies of Atomic Recombination of Nitrogen, Hydrogen, and Oxygen by
Paramagnetic Resonance," Physics of Fluids, Vol. 5, 1962, pp. 743-753.
44
Evenson, K.M. and Burch, D.S., "Atomic-Nitrogen Recombination," Journal of Chemical Physics,
Vol. 45, 1966, pp. 2450-2460.
45
Kim, Y.C. and Boudart, M., "Recombination of O, N, and H Atoms on Silica: Kinetics and
Mechanism," Langmuir, Vol. 7, 1991, pp. 2999-3005.
46
Brennan, D., "Heterogeneous Atomisation and Recombination," in Comprehensive Chemical Kinetics.
Volume 21, Reactions of Solids with Gases, edited by C. H. Bamford, C. E. H. Tripper, and R. G.
Compton, (Elsevier, New York, 1984), p. 151-231.
47
Balat, M., Czerniak, M., and Badie, J.M., "Thermal and Chemical Approaches for Oxygen Catalytic
Recombination Evaluation on Ceramic Materials at High Temperature," Applied Surface Science,
Vol. 120, 1997, pp. 225-238.
48
Balat-Pichelin, M., Badie, J.M., Berjoan, R., and Boubert, P., "Recombination Coefficient of Atomic
Oxygen on Ceramic Materials Under Earth Re-Entry Conditions by Optical Emission Spectroscopy,"
Chemical Physics, Vol. 291, 2003, pp. 181-194.
49
Balat-Pichelin, M. and Duqueroie, F., "Heat Transfer Modeling at High Temperature for the
Evaluation of Atomic Oxygen Recombination Energy on Ceramic Materials," International Journal
of Thermal Sciences, Vol. 40, 2001, pp. 279-287.
50
Bedra, L. and Balat-Pichelin, M., "Comparative Modeling Study and Experimental Results of Atomic
Oxygen Recombination on Silica-Based Surfaces at High-Temperature," Aerospace Science and
Technology, Vol. 9, 2005, pp. 318-328.
51
Booth, J.P., Joubert, O., Pelletier, J., and Sadeghi, N., "Oxygen Atom Actinometry Reinvestigated:
Comparisons with Absolute Measurements by Resonance Absorption at 130 nm," Journal of Applied
Physics, Vol. 69, 1991, pp. 618-626.
52
Katsch, H.M., Tewes, A., Quandt, E., Goehlich, A., Kawetzki, T., and Döbele, H.F., "Detection of
Atomic Oxygen: Improvement of Actinometry and Comparison with Laser Spectroscopy," Journal of
Applied Physics, Vol. 88, 2000, pp. 6232-6238.
53
Guyon, C., Cavadias, S., and Amouroux, J., "Heat and Mass Transfer Phenomenon from an Oxygen
Plasma to a Semiconductor Surface," Surface and Coating Technology, Vol. 142-144, 2001, pp. 959-
963.
54
Guyon, C., Cavadias, S., Mabille, I., Moscosa-Santillan, M., and Amouroux, J., "Recombination of
Oxygen Atomic Excited States Produced by Non-Equilibrium RF Plasma on Different Semiconductor
Materials: Catalytic Phenomena and Modelling," Catalysis Today, Vol. 89, 2004, pp. 159-167.
55
Cauquot, P., Cavadias, S., and Amouroux, J., "Thermal Energy Accommodation from Oxygen Atoms
Recombination on Metallic Surfaces," Journal of Thermophysics and Heat Transfer, Vol. 12, 1998,
pp. 206-213.
56
Mozetič, M., Drobnič, M., and Zalar, A., "Recombination of Neutral Hydrogen Atoms on AISI 304
Stainless Steel Surface," Applied Surface Science, Vol. 144-145, 1999, pp. 399-403.
57
Mozetič, M. and Zalar, A., "Recombination of Neutral Oxygen Atoms on Stainless Steel Surface,"
Applied Surface Science, Vol. 158, 2000, pp. 263-267.
58
Mozetič, M., Zalar, A., Cvelbar, U., and Poberaj, I., "Heterogeneous Recombination of Neutral
Oxygen Atoms on Niobium Surface," Applied Surface Science, Vol. 211, 2003, pp. 96-101.
59
Pallix, J.B. and Copeland, R.A., "Measurement of Catalytic Recombination Coefficients on Quartz
Using Laser-Induced Fluorescence," Journal of Thermophysics and Heat Transfer, Vol. 10, 1996, pp.
224-233.
60
Marschall, J., “Experimental Determination of Oxygen and Nitrogen Recombination Coefficients at
Elevated Temperature Using Laser-Induced Fluorescence,” AIAA Paper 97-3879, August 1997.
11 - 30 RTO-EN-AVT-142
61
Bischel, W.K., Perry, B.E., and Crosley, D.R., "Two-Photon Laser-Induced Fluorescence in Oxygen
and Nitrogen Atoms," Chemical Physics Letters, Vol. 82, 1981, pp. 85-88.
62
Bischel, W.K., Perry, B.E., and Crosley, D.R., "Detection of Fluorescence from O and N Atoms
Induced by Two-Photon Absorption," Applied Optics, Vol. 21, 1982, pp. 1419-1428.
63
Adams, S.F. and Miller, T.A., "Two-Photon Absorption Laser Induced Fluorescence of Atomic
Nitrogen by an Alternative Excitation Scheme," Chemical Physics Letters, Vol. 295, 1998, pp. 305-
311.
64
Bashkin, S. and Stoner, J.O., Jr., Atomic Energy Levels and Grotrian Diagrams, Vol. 1 Elsevier
Publishing Company, Inc., New York, 1975.
65
Dagdigian, P.J., Forch, B.E., and Miziolek, A.W., "Collisional Transfer Between and Quenching of
the 3p 3P and 5P States of the Oxygen Atom," Chemical Physics Letters, Vol. 148, 1988, pp. 299-308.
66
Copeland, R.A., Jeffries, J.B., Hickman, A.P., and Crosley, D.R., "Radiative Lifetime and Quenching
of the 3p 4Do State of Atomic Nitrogen," Journal of Chemical Physics, Vol. 86, 1987, pp. 4876-4884.
67
Jeffries, J.B., Copeland, R.A., and Crosley, D.R., "Intermultiplet Energy Transfer in the Collisions of
3p 4Do Nitrogen Atoms with Nitrogen Molecules," Journal of Chemical Physics, Vol. 91, 1989, pp.
2200-2205.
68
Marschall, J., Copeland, R.A., Hwang, H.H., and Wright, M.J., “Surface Catalysis Experiments on
Metal Surfaces in Oxygen and Carbon Monoxide Mixtures,” AIAA Paper 2006-181, January 2006.
69
Berkowitz, J., "Catalytic Oxygen Atom Recombination on Solid Surfaces," in The Structure and
Chemistry of Solid Surfaces, edited by G. A. Somorjai, (John Wiley & Sons, Inc., New York, 1968),
p. 80.81-80.16.
70
Dickens, P.G. and Sutcliffe, M.B., "Recombination of Oxygen Atoms on Oxide Surfaces,"
Transactions of the Faraday Society, Vol. 60, 1964, pp. 1272-1285.
71
Hartunian, R.A., Thompson, W.P., and Safron, S., "Measurement of Catalytic Efficiency of Silver for
Oxygen Atoms and the O-O2 Diffusion Coefficient," Journal of Chemical Physics, Vol. 43, 1965, pp.
4003-4006.
72
Myerson, A.L., "Exposure-Dependent Surface Recombination Efficiencies of Atomic Oxygen,"
Journal of Chemical Physics, Vol. 50, 1969, pp. 1228-1234.
73
Marschall, J., Chamberlain, A., Crunkleton, D., and Rogers, B., "Catalytic Atom Recombination on
ZrB2/SiC and HfB2/SiC Ultra-High Temperature Ceramic Composites," Journal of Spacecraft and
Rockets, Vol. 41, 2004, pp. 576-581.
74
Copeland, R.A., Pallix, J.B., and Stewart, D.A., "Surface Catalyzed Production of NO From
Recombination of N and O Atoms," Journal of Thermophysics and Heat Transfer, Vol. 12, 1998, pp.
496-499.
75
Kurotaki, T., "Catalytic Model on SiO2-Based Surface and Application to Real Trajectory," Journal of
Spacecraft and Rockets, Vol. 38, 2001, pp. 798-800.
76
Sepka, S., Chen, Y.-K., Marschall, J., and Copeland, R.A., "Experimental Investigation of Surface
Reactions in Carbon Monoxide and Oxygen Mixtures," Journal of Thermophysics and Heat Transfer,
Vol. 14, 2000, pp. 45-52.
77
Wright, M.J., Candler, G.V., and Bose, D., "Data-Parallel Line Relaxation Method for the Navier-
Stokes Equations," AIAA Journal, Vol. 36, 1998, pp. 1603-1609.
78
Cheatwood, F.M. and Gnoffo, P.A., “User's Manual for the Langley Aerothermodynamic Upwind
Relaxation Algorithm (LAURA),” NASA TM-4764, April 1996.
79
Niemi, K., Gathen, V.S.-v.d., and Döbele, H.F., "Absolute Atomic Oxygen Density Measurements by
Two-Photon Absorption Laser-Induced Fluorescence Spectroscopy in an RF-Excited Atmospheric
Pressure Plasma Jet," Plasma Sources Science and Technology, Vol. 14, 2005, pp. 375-386.
RTO-EN-AVT-142 11 - 31
80
Niemi, K., Gathen, V.S.-v.d., and Döbele, H.F., "Absolute Calibration of Atomic Density
Measurements by Laser-Induced Fluorescence Spectroscopy with Two-Photon Excitation," Journal of
Physics D: Applied Physics, Vol. 34, 2001, pp. 2330-2335.
81
Goehlich, A., Kawetzki, T., and Döbele, H.F., "On Absolute Calibration with Xenon of Laser
Diagnostic Methods Based on Two-Photon Absorption," Journal of Chemical Physics, Vol. 108,
1998, pp. 9362-9370.
82
Bamford, D.J., Saxon, R.P., Jusinski, L.E., Buck, J.D., and Bischel, W.K., "Two-Photon Excitation of
Atomic Oxygen at 200.6, 192.5, and 194.2 nm: Absolute Cross Sections and Collisional Ionization
Rate Constants," Physical Review A, Vol. 37, 1988, pp. 3259-3269.
11 - 32 RTO-EN-AVT-142
Interaction of Reactive Gas Flows and Ceramics at High Temperature –
Experimental Methods for the Measurement of Species
Recombination during Planetary Entry
Marianne BALAT-PICHELIN
Laboratoire Procédés, Matériaux et Energie Solaire, PROMES-CNRS, UPR 8521
rue du four solaire
66120 Font-Romeu Odeillo
France
Tél : +33 468 307 768
Fax : +33 468 302 940
balat@promes.cnrs.fr
1. INTRODUCTION
During the atmospheric re-entry phase of aerospace vehicles, several physico-chemical phenomena taking
place on the hot parts (nose cap and wing leading edges) can lead to an important excess of heating and a
possible damage of the protective materials.
PROMES-CNRS laboratory has developed since several years experimental methods to study oxidation
and catalytic recombination of atomic oxygen under conditions of atmospheric re-entry on Earth and on
Mars. The most important conditions for the ground simulation of re-entry (high temperature, low pressure
plasma…) have been realized in the MESOX set-up associating a reactor placed at the focus of a solar
radiation concentrator and a microwave generator. Concerning the dynamic contribution, only low
enthalpy flow can be reproduced on this set-up.
A multi-scale experimental and theoretical approach has been developed to evaluate the recombination
parameters.
On one hand, the study of the atomic oxygen recombination on partially catalytic based -silicon or -
aluminum ceramic materials, at high temperature (850-2000 K) has been done at different pressures (200-
2000 Pa) by a thermal approach, at a mesoscopic scale and leads to the determination of the thermal flux
of recombination transferred to the material and to the chemical energy accommodation coefficient β.
Influences of total pressure, surface temperature and material microstructure are studied.
On the other hand, a chemical approach, at a microscopic scale is developed for the evaluation of the
recombination coefficient γ using optical emission spectroscopy and actinometry technique on the same
device. The values obtained by this method are complementary of the ones of the thermal approach.
At the atomic scale, a simulation by Molecular Dynamics is performed in collaboration with Cacciatore
from IMIP-CNR, Bari in order to predict atomic oxygen recombination process over silica-based ceramics
at high temperatures.
These last decades, the development of thermal protection materials for aerospace vehicles needs surface
flux calculations during atmospheric re-entry to predict heat rates. Most of the experiments were realized
Balat-Pichelin, M. (2007) Interaction of Reactive Gas Flows and Ceramics at High Temperature – Experimental Methods for the
Measurement of Species Recombination during Planetary Entry. In Experiment, Modeling and Simulation of Gas-Surface
Interactions for Reactive Flows in Hypersonic Flights (pp. 12-1 – 12-26). Educational Notes RTO-EN-AVT-142, Paper 12.
Neuilly-sur-Seine, France: RTO. Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 12 - 1
Interaction of Reactive Gas Flows and Ceramics
at High Temperature – Experimental Methods for the
Measurement of Species Recombination during Planetary Entry
on arc-jet plasma facilities where the recombination is obtained from stagnation point heat flux
measurement in dissociated flow. The relative heating rates of a known catalytic sample are compared to
the surface being studied [1-10].
Some evaluation of catalytic recombination is also obtained by measurement of atom concentration by

several techniques (LIF, spectroscopy...) to reach the recombination coefficient γ [11-13]. Other authors
have proposed models for recombination and compared to experimental values [2, 4, 14, 15].
Another coefficient β called chemical energy accommodation coefficient is rarely evaluated and often its
value is taken equal to 1 (total accommodation). Some authors have tried to measure β and the result is
often far from 1 [16-20]. This coefficient is very important because it allows to take into account the real
energy transferred to the surface.
In many times, both measurements of flux and γ coefficient to study recombination process are not
realized on the same experiment. More often, the flux is measured and the recombination coefficient γ is
obtained by calculation. The experimental set-up MESOX that we have developed in our laboratory allows
to measure simultaneously the thermal and chemical contributions of the atomic oxygen recombination on
surfaces at high temperatures, on the same set-up so the accuracy could be better.
2. SURFACE REACTIONS
2.1 Various aspects of dynamics of surface reactions

Catalysis concerns the rate of a chemical reaction, it is a dynamic phenomenon. A heterogeneously
catalyzed reaction consists of a sequence of elementary steps such as adsorption, surface diffusion,
chemical transformation of adsorbed species and desorption, needing the identification and
characterization of the reaction mechanisms and the transport processes.
The efficiency of a catalyst can be determined by the macroscopic kinetics of the overall reaction,
including chemical transformations and physical processes of energy and mass transport. The resulting
yield depends on external parameters such as temperature, flow rates, partial pressures of the species
participating in the reaction. Modelling of the macroscopic kinetics is often achieved by fitting empirical
equations, for example using exponential dependent temperature laws (Arrhenius) leading to apparent
activation energies as parameters.
The development of sophisticated surface physical methods enabled detailed insights into the atomic
processes on surface as well as identification and characterization of the properties of the surface species.
Description of the progress of a catalytic reaction in an approach called microkinetics is achieved by
modelling the macroscopic kinetics through correlating the atomic processes with the macroscopic
parameters in the framework of a suitable continuum model. Even if the assumptions underlying a specific
continuum model are not completely fulfilled, the model may nevertheless provide a good approximation
over a limited range of external parameters. The formulation of the rate laws can lead to a set of nonlinear
coupled differential equations for the concentrations of the various species involved at the surface. The
resulting temporal behaviour under constant continuous-flow conditions will not be necessarily stationary.
Either, the spatial distributions may not be uniform, and the existence of local variations in surface
coverage causes coupling of the reaction with adsorbate diffusion or heat transfer processes. So, the
formation of spatio-temporal concentration patterns on a mesoscopic scale may occur.
The occurrence of interactions between the adsorbed species and their occupation of non equivalent
adsorption sites complicates an appropriate description, apart from the fact that the surface is non-uniform
and it may undergo structural transformations under the influence of the adsorbates (oxidation,
12 - 2 RTO-EN-AVT-142
degradation…). Detailed investigation of these effects on an atomic scale represents the classical domain
of surface science.
The basic idea of the transition state theory consists of the assumption that at all stages along the reaction
coordinate thermal equilibrium is established, leading to the temperature as the only essential parameter.
This assumption requires that energy exchange between the degrees of freedom of the particles interacting
with the surface and the heat bath of solid occurs much faster than the elementary step initiating nuclear
motion. Processes of energy transfer between the various degrees of freedom on the quantum level form
the ultimate basis for chemical reactions.
2.2 Definition of the recombination parameters

The fraction of the chemical energy transferred to the surface by atom recombination depends on the rate
at which atoms are recombining and releasing their dissociation energy on the surface.
Four parameters can describe the global recombination reaction:
γ : the recombination coefficient, defined as the ratio of the number of recombined atoms to the total
number of atoms impinging the surface γ = [O]rec/[O]tot
β : the chemical energy accommodation coefficient, defined as the ratio of the chemical energy transferred
to the surface by recombination to the total energy due to the recombination of atoms.
These two coefficients are comprised between 0 (no recombination, no energy transferred to the surface)
and 1 (total recombination or total energy transferred to the surface). The product (γ.β) is often called the
effective recombination coefficient γ.
qrec,acc : the thermal flux due to the recombination of species at the surface, linked to the γ coefficient
kw : the catalytic recombination rate linked to γ by the relation
k B .T
kw = γ
2πM (1)
2.3 Mechanisms of recombination: L-H and E-R
The catalytic recombination process can be divided into main three steps:
• Adsorption : 2 types of adsorption are distinguished, physical adsorption (physisorption) where the
atom is held to the surface by Van der Waals forces and chemical adsorption, where the bond between
atom and surface is of chemical nature (Lennard-Jones potential). In atmospheric re-entry conditions,
the temperature of the surface is too high for significant physisorption to take place, which leaves only
chemisorption as the mechanism through which catalytic recombination can occur ;
• Recombination : Here too, 2 types are considered. In the first, two adhered atoms, located at adjacent
sites combine to form a molecule (Langmuir-Hinshelwood mechanism) and in the second,
recombination occurs when a gas-phase atom collides with a surface-adhered atom (Eley-Rideal
mechanism).
• Desorption : The final step in catalytic recombination is the desorption of the recombined molecule
from the surface, leaving the site it occupied for new atom adsorption.
RTO-EN-AVT-142 12 - 3
At low temperatures, the L-H mechanism seems favourable to the E-R mechanism, because of the high
surface coverage. For high temperatures, the thermal agitation of adsorbed atoms becomes more
important, and the fraction of surface sites occupied, is expected to drop. Therefore, unless the mobility of
atoms adsorbed at the surface is high, the L-H mechanism will not play an important role in the catalytic
recombination process. Highly mobile adhered surface atoms are an unlikely phenomenon, because of the
nature of the adsorption. Furthermore, in most experiments, catalytic recombination is found to be a first
order reaction, indicating that the E-R mechanism is dominant in the reaction process.
3. EXPERIMENTAL SET-UP MESOX

The MESOX set-up (Moyen d’Essai Solaire d’OXydation) that associates a quartz tube reactor placed at
the focus of the 6 kW solar radiation concentrator for sample heating (up to 2500 K under pressures
comprised between 102 and 105 Pa) and a microwave generator (2450 MHz, 1200 W max) is presented on
Figure 1. Originally developed for oxidation studies, it has been adapted to allow atomic recombination
measurements. Atmospheric entry conditions can be partially simulated, pressure and temperature can be
reproduced independently with a high accuracy.
The experimental device is placed at the focus of a solar furnace equipped with a variable opening shutter.
It can be moved away from the focus to be replaced by a calorimeter to measure the incident concentrated
solar flux that can reach 5 MW.m-2. The temperature measurements on front and back faces of the sample
are realized using a single optical pyrometer (5 µm) with a system of one rotating mirror and two
stationary mirrors. The experimental reactor consists of a quartz tube, 50 cm length and 5 cm diameter
with CaF2 viewports. This reactor crossing the refrigerated waveguide contains the sample (25 mm
diameter and 3 mm height) placed in stagnation point position at the center of the discharge. A regulator, a
gauge and a vacuum pump are used to control precisely the total pressure during the experiment.
The spectroscopic bench is composed of an optical sampling system including a lens and a mirror, and a
monochromator (spectrometer Triax 550 Jobin-Yvon) equipped with an Optical Multichannel Analyzer
(OMA). The microwave discharge is imaged by the silica lens (magnification 0.1) on the slit entrance of
the monochromator. The 55 cm focal length monochromator working with a 1200 grooves/mm grating
(also available 1800 and 2400 grooves/mm gratings) and a 100 µm width slit allows a spectral resolution
of 0.2 nm. The dispersed light is analyzed by means of the CCD matrix (1024 x 128) of the OMA
detector. Each of the 128 lines of the matrix gives information on the relative atomic oxygen concentration
at different distances from the surface of the sample with a spatial resolution around 270 µm. A spectral
analysis over the 128 lines is performed very quickly after the solar radiation breaking. The total duration
of a scan is 200 ms. Therefore, all the spectral and spatial needed informations are taken simultaneously
allowing a good accuracy.
12 - 4 RTO-EN-AVT-142
Figure 1: Experimental set-up MESOX with the irradiated shutter (top, right) placed at the focus of
the 6 kW solar furnace (top, left). Images of plasma environment: CO2 (down, left)
and air (down, right).
RTO-EN-AVT-142 12 - 5
4. EXPERIMENTAL DETERMINATION OF THE RECOMBINATION OF

ATOMIC OXYGEN AT HIGH TEMPERATURE
In this chapter, the presentation of the experimental methods developed for the evaluation of
recombination parameters is carried out according to a multi-scale approach.
4.1 Mesoscopic approach: thermal flux of recombination

A mesoscopic approach for the evaluation of the atomic oxygen recombination has been developed using a
thermal balance done on a reference cylinder volume in the sample. The surface of this cylinder of 6 mm
diameter and 3 mm height in the sample, considered for the thermal balance (Fig. 2), represents the
measurement area by pyrometry. Convective phenomena are neglected (rarefied gas flows) compared to
the radiative fluxes to establish the equations for steady state heat transfer under different environments.
The thermal balance is established under each atmosphere: standard air, air plasma, standard argon and
argon plasma, the determination of the recombination flux being done by difference between the
experiments under air plasma (reactive) and argon plasma (inert).
q sol
q rad,f
q rec q mw
front face
q lat
Reference cylinder
6 mm diameter, 3 mm heigth
back face
q rad,b
Figure 2: Reference cylinder considered for the evaluation of the thermal flux of recombination
under air plasma conditions
Under standard air: the sample is heated only by solar radiation

α.qsolair = qrad,fair + qrad,bair + qlatair (2)
Under air plasma: the sample is heated by solar radiation, by microwave-material interaction and by
atomic oxygen recombination
α.qsolair,* + qmwair,* + qrec,accair,* = qrad,fair,* + qrad,bair,* + qlatair,* (3)
Under standard argon: the sample is heated only by solar radiation, like in standard air
α.qsolarg = qrad,farg + qrad,barg + qlatarg (4)
Under argon plasma: the sample is heated by solar radiation and by microwave-material interaction
α.qsolarg,* + qmwarg,* = qrad,farg,* + qrad,barg,* + qlatarg,* (5)
12 - 6 RTO-EN-AVT-142
The thermal flux of recombination is obtained by difference between Eq. (3) and (5), and we can neglect
the difference between the radial losses by conduction out of the reference cylinder because the radial
losses are of same order in air and argon plasmas.
Under the following assumptions (some are experimentally verified) a like-one dimensional equation is
obtained:
- without plasma, the same temperature level is reached for both air and argon atmospheres, that is Tfair =
Tfarg and Tbair = Tbarg ; this is due to the constant solar flux, experiments being done around the period of
solar zenith ;
- under plasma, the microwave-material interaction for both the atmospheres is supposed equal so qmwair,*
= qmwarg,* ; this will be verified when the dielectric permittivity of the samples tested may be measured at
high temperature and so the absorbed power may be known ;
- the absorbed solar energy is constant under the different atmospheres for a given shutter opening because
during all the experiment the incident solar flux is constant ;
- the radial losses by conduction out of the reference cylinder are neglected. Infrared imaging on the front
face of the sample has allowed to determine a weak radial temperature gradient as function of distance
from the centre, even when the sample is heated under air or argon plasma.
So, the thermal balance is reduced to a 1D approach. Finally, it comes
qrec,accair,* = qrad,fair,* - qrad,farg,* + qrad,bair,* - qrad,barg,* (6)
equivalent to :
qrec,accair,* = ε σ [(Tfair,*)4 - (Tfarg,*)4 + (Tbair,*)4 - (Tbarg,*)4] (7)
The thermal flux of recombination transferred to the surface can be calculated from the following
parameters: ε the total hemispherical emissivity and Tf*and Tb* the front and back face temperatures
under air and argon plasmas.
The uncertainties ∆qrec,acc/qrec,acc have been calculated taking into account the errors on temperature
measurements due to the accuracy of the optical pyrometer (0.5%), on the spectral emissivity (at 5 µm, 1
to 2% depending on the materials) and on the total hemispherical emissivity (1%). The emissivity
measurements are done in our laboratory by a direct method using a two-color (1.3 and 1.55 µm)
pyrometer for the temperature measurement and a spectroradiometer for the radiance determination [21].
The accuracy is mentioned on Table 2 but not on Figure 3 for a better visualization of the experimental
points.
4.1.1 Experimental results

The pressure and the flow rate (10-6 m3.s-1) are fixed at the beginning of each experiment. Three total gas
pressures have been applied: 200, 1000 and 2000 Pa, for five temperature levels (1000, 1200, 1400, 1600
and 1800 K). The surface temperature depends on the incident solar flux controlled by the opening of the
shutter.
The study has been realized on sintered materials (SiC, Si3N4, AlN, Al2O3) or on samples obtained by
oxidation at 1300 K under atmospheric pressure during 24 hours (SiC + SiO2 , AlN + Al2O3).
RTO-EN-AVT-142 12 - 7
Table 2 shows the temperature increase due to the atomic oxygen recombination on the surface of
different materials deduced from the difference under air and argon plasmas at 2000 and 200 Pa
according to the relation:
∆Tf = Tfair,* - Tfair - (Tfarg,* - Tfarg ) (8)

The temperature increase due to the catalytic recombination for all the tested materials decreases rapidly
with temperature. The silicon-based ceramics have weaker ∆Tf than the aluminum ones and particularly
than the sintered alumina (∆Tf = 350 K at 1000 K).
A graphical representation of the results for the thermal fluxes of recombination obtained from Eq. (7) is
shown on Figure 3 for some materials tested at 2000 and 200 Pa, at five temperature levels.
The determination of the thermal flux of recombination confirms the tendencies obtained with the
temperature increase. Silicon-based ceramic materials have similar and low catalytic activities. The
sintered SiC has the weakest recombination flux (at 200 Pa) that is nearly constant between 1000 and 1600
K (around 30 kW.m-2), except under active oxidation [22], at 1800 K (13 kW.m-2). This confirms its very
low catalytic activity, the higher values obtained between 1000 and 1600 K being due to the presence of a
thin silica layer formed during experiment, like in the case at 2000 Pa (see Table 2). Aluminum-based
ceramic materials have more pronounced catalytic activities and sintered alumina is by far the more
catalytic material with an increasing recombination flux that reaches 151 kW.m-2 at 1400 K for 200 Pa and
184 kW.m-2 at 2000 Pa, before decreasing at 1800 K (respectively 65 kW.m-2 at 200 Pa and 93 kW.m-2 at
2000 Pa). Here, on Table 2, we can see the influence of the total pressure: higher total pressures lead to
higher thermal flux of recombination.
Sintered SiC 2000 Pa Sintered AlN + Al2O3 Sintered Al2O3

Tf ∆Tf qrec,acc 1D/ 2D Tf ∆Tf qrec,acc 1D Tf ∆Tf qrec,acc 1D/ 2D
(kW/m2) (kW/m2)
(K) (kW/m2) (K) (K)
984±3 74 37±3 / 43 994±6 125 58±8 982±6 354 136±22 / 154
1196±5 45 35±3 / 40 1174±10 94 59±9 1173±9 301 149±27 / 168
1386±7 30 33±3 / 38 1409±14 52 48±9 1364±11 304 184±37 / 202
1598±9 21 34±3 / 36 1572±17 40 48±9 1597±16 154 113±25 / 126
1775±11 19 36±4 / 40 1769±21 28 42±9 1767±21 107 93±23 / 106
Sintered SiC 200 Pa Sintered AlN + Al2O3 Sintered Al2O3

Tf ∆Tf qrec,acc 1D/ 2D Tf ∆Tf qrec,acc 1D Tf ∆Tf qrec,acc 1D/ 2D
(kW/m2)
(K) (kW/m2) (K) (K) (kW/m2)
984±3 64 30±2 / 34 972±6 113 48±7 975±6 281 97±16 / 109
1191±5 38 29±2 / 31 1179±10 88 56±9 1168±9 287 139±24 / 153
1377±7 27 30±3 / 32 1380±14 55 51±9 1399±11 251 151±29 / 166
1591±9 16 25±2 / 26 1577±17 36 38±7 1577±16 153 108±23 / 120
1780±11 7 13±2 / 15 1771±21 26 34±7 1710±20 81 65±15 / -
Table 2 : Experimental data for atomic oxygen recombination on ceramic materials at 2000 (top)
and 200 (down) Pa where Tf is the front face temperature, ∆Tf the front face temperature increase
due to the recombination and qrec,acc the thermal fux of recombination.
12 - 8 RTO-EN-AVT-142
SiC SiC+SiO2 AlN Si3N4 AlN+Al2O3 Al2O3

300
200
100
0
900 1100 1300 1500 1700 1900
Tf (K)
Al2O3 SiC SiC+SiO2 Si3N4 AlN AlN+Al2O3

200
P = 200 Pa
150
100
50
0
900 1100 1300 1500 1700 1900
Tf (K)
Figure 3: Increasing in temperature due to recombination (top) and thermal flux of recombination
transferred to the surface (down) for some materials under 200 Pa air plasma.
4.1.2 Comparison of modelling and experimental results

A bi-dimensional model has been developed in order to estimate the radial thermal losses by conduction
that have been neglected in the experimental part. This 2-D model is developed in order to estimate the
radial thermal losses that have been neglected in equation (7), on the whole sample, and to calculate by
parameter identification the thermal flux of recombination. This model is based on the simulation of axial
and radial heat transfers in the material at high temperature. The 2-direction heat equation with added
source terms (one due to the microwave-material interaction, since the sample is in the electromagnetic
field and another one due to the recombination) is solved with a finite difference scheme that is applied to
a cylindrical shape. This code is written with Matlab™ and gives a temperature mapping of the sample
front face (where recombination occurs) that is compared to camera-obtained infra-red images. The code
gives also the sample back face temperature mapping and that of its thickness [23]. A comparison is done
between:
• the experimental values of the thermal flux of recombination and the values obtained by modelling ;
• the temperature mapping of the samples obtained by modelling and those experimentally acquired
using an infra-red camera.
RTO-EN-AVT-142 12 - 9
The problem is solved by using the codes called MAT2D and REC2D. With the material characteristics
and the boundary conditions, the codes compute the temperature distribution in the sample (MAT2D) and
determine the thermal flux of recombination qrec,acc2D (REC2D) which is compared to the experimental
value qrec,acc1D.
For the determination of the thermal flux of recombination using the code REC2D, several stages are
necessary. At each stage, one parameter is identified according to the experimental values of temperature.
This parameter is validated when the computed temperatures are equal to the experimental ones. The
different stages are the following:
• First, an experiment is done under standard air. The microwave flux absorbed by the sample and the
thermal flux of recombination are equal to zero. In this case, the value of the incident solar flux qsol is
measured using a calorimeter. The only unknown value is the apparent solar absorptivity α. Its value
is assumed to be correct when T2Dair = Texpair, that is to say when the temperature obtained by the
modelling and the experiment are equal ;
• Second, an experiment is performed under argon plasma. The thermal flux of recombination is equal
to zero. The apparent solar absorptivity is known by the first stage, so the only unknown is, in this
case, the microwave flux absorbed by the sample. As previously, its value is supposed to be correct
when T2Darg*= Texparg* under plasma conditions (*) ;
• The last stage is an experiment under air plasma (reactive gas). The apparent solar absorptivity α and
the microwave flux absorbed by the sample qmw are known using the preceding stages, so the only
unknown is now the thermal flux of recombination transferred to the surface sample qrec,acc. Its value is
determined when T2Dair* = Texpair* under air plasma conditions (air*).
The results of the calculations are given in Table 2 for two sintered materials (SiC and Al2O3). Most of the
results obtained by the model REC2D are in good agreement with the experimental values taking into
account the accuracy of the experimental results. The results obtained for sintered SiC (less catalytic
ceramic) and Al2O3 (highest catalytic ceramic) at 200 and 2000 Pa air, under plasma conditions, are given
in Table 2. The values of the thermal flux of recombination obtained by experiment (qrec,acc1D) are very
near the calculated ones (qrec,acc2D) thus proving that the hypotheses chosen for the experimental evaluation
are valid for the ceramic materials tested. The main hypothesis (difference between both radial losses
negligible) appears to be less important at high temperature as shown in Table 2, so the experimental
results present a better accuracy at high temperature (1600-1800 K).
4.2 Microscopic approach: recombination coefficient

The actinometry technique is used to follow the relative atomic oxygen concentration profile along the
discharge. A low known quantity of argon is introduced in the flow and the evolution of the intensities
ratio IO/IAr of an oxygen line to an argon line is measured along the discharge zone.
Some assumptions are necessary using actinometry:
• the actinometer must be introduced in low quantity so as not to disturb the plasma ;
• the excited species must be solely produced by electronic impact from the ground state ;
• the desexcitation of the species must be essentially radiative ;
• the energy dependencies of the cross sections of electronic excitation of O and Ar must be identical in
theory and, at the least, the energy thresholds of the transition must be similar.
To determine the spatial variation of the relative concentration of atomic oxygen, we use its most reliable
transition at 844.6 nm. For the actinometer line, we choose the argon transition at 842.4 nm that presents a
similar energy threshold (13.1 eV) than for the atomic oxygen transition (11 eV). Thus, it is reasonable to
12 - 10 RTO-EN-AVT-142
assume that the ratio of the intensities of the two lines is proportional to the oxygen atom concentration.
Moreover, the lines 842.4 and 844.6 nm can be recorded simultaneously that increasing the accuracy of
the intensities ratio measurements [24].
We have chosen to work at a constant microwave power of 300 W, a total air pressure of 200 Pa and a
total flow of 10-6 m3 s-1 with 5 % argon. In this type of microwave plasma, as shown by several authors
[25-28], the oxygen atom concentration is strongly increased when the amount of nitrogen is increased.
Recently, we have used fiber-optics catalytic probe to measure the absolute density of neutral oxygen
atoms in our reactor in collaboration with Mozetic and Vesel from the Jozef Stefan Institute (Ljubljana,
Slovenia). The degree of the dissociation of oxygen molecules in our 2450 MHz air plasma is around 80%
for the above conditions of flow rate (to be published).
A cylindrical volume corresponding to the discharge zone is considered, every point being represented by
the coordinates (r, x). As the mean free path of the atoms (0.043 cm at 200 Pa) is less than the diameter of
the reactor (5 cm), the atom diffusion is given by the diffusion equation written in cylindrical coordinates
that describes the variation of the concentration CO of an oxygen atom O versus time for a fixed point in
the cylinder (r, x):
∂CO
+ divCO.Ux + divCO.Ur + ω = 0
∂t
(9)
with ω the variation of the concentration due to the recombination in the gaseous phase and on the reactor
walls.
In steady state conditions, equation (9) is reduced to:
 ∂2C 2 
D. O + ∂ CO + 1 ∂CO + ω = 0

 ∂x
2
∂r2 r ∂r 
(10)
We suppose that the convective transfer is negligible. The radial gradient in the reactor is negligible
compared to the axial one, so the concentration is only function of x. Moreover, the stability of the ratio
IO/IAr in the reactor allows neglecting the recombination in volume and on the reactor wall.
Thus, equation (10) can be simplified in:
∂2CO
D. =0
∂x2
(11)
This equation has two limit conditions:
• the ratio IO/IAr is constant along the discharge, thus, far from the sample, the concentration has a
known fixed value : CO(x = L) = constant
• at the surface sample (x = 0), the mass balance in oxygen atoms is established by the equality between
the oxygen arriving at the surface by diffusion and the atomic oxygen recombined at the surface :
∂C O γ .V
−DO, air. − CO(x = 0). =0
∂x x = 0 4
(12)
with V the mean square velocity of oxygen atoms.
RTO-EN-AVT-142 12 - 11
The evolution of the atomic oxygen concentration is given by the solution of equation (11).
Finally, the intensities ratio obtained by actinometry leads to the determination of the recombination
coefficient γ by the following equation:
 
 IO 
I 
TS  4.DO,air
γ =  Ar x= L
−1 .
 TL  V.L
 IO 
I 
 Ar x= 0  (13)
with IO/IAr the ratio of the intensities respectively at the entrance of the reactor (x = L) and at the surface
sample (x = 0), DO,air the binary diffusion coefficient of atomic oxygen in air, V the mean square atomic
velocity and L the thickness of the recombination boundary layer.
4.2.1 Experimental results

The uncertainties ∆γ/γ have been calculated taking into account the errors on IO/IAr and L but also on the
flow parameters: the binary diffusion coefficient DO,air determined using the Chapman-Enskog theory and
the mean square atomic velocity V determined using the gas kinetic theory (rarefied gas). The accuracy on
these two last values is due essentially to that of the gas temperature, measured by emission spectroscopy
(N2 rotational temperature), this leading to a total accuracy on γΟ of ± 30 %.
For nearly all the materials, after the acquisition of the plasma image, we observe a constant level for the
concentration profile of atomic oxygen until 6 mm from the surface leading by a progressive decrease and
then, at 1.5 mm from the surface, there is a change in the slope that we consider representative of the
recombination phenomena occurring on the material surface.
We have applied the actinometry method on sintered SiC oxidized during the γ measurement in SiO2 β-
cristobalite and on SiO2 quartz, on two sintered alumina and on an alumina layer obtained by oxidation at
1300 K at atmospheric pressure during 24 h on sintered AlN. The values of the recombination coefficient
have been calculated for each material with the concentration profile and are given on Figures 4 to 7.
Figure 4 presents the results for the two polymorphs of silica, Figure 5 the AFM images of both the silica
samples and Figures 6 and 7 the γ results for the three alumina samples different by their impurity content.
4.2.1.1 Silica polymorphs samples

The silica polymorphs samples studied were (i) the silica layer (β-cristobalite) obtained by oxidation on
sintered SiC, α-6H containing less than 1% mass boron and (ii) silica, quartz.
The results presented on Figure 4 show the very weak catalytic activity of quartz. On the contrary, the
recombination coefficient of SiO2 β-cristobalite is roughly four times higher [24]. According to Figure 4,
the apparent activation energy of recombination Ea using an Arrhenius law for the recombination
coefficient gives:
12 - 12 RTO-EN-AVT-142
• for SiO2 β-cristobalite :

γ = 0.6382 exp(-3374/T) and Ea = 28.0 ± 0.6 kJ/mol between 800 and 1830 K
• for SiO2 quartz :
γ = 0.0622 exp(-2158/T) and Ea = 17.9 ± 0.6 kJ/mol between 850 and 1430 K
The accuracy on the activation energy values is calculated from several measurements.
For βcristobalite and quartz, the recombination coefficient increases with temperature, so the mechanism
of recombination seems to be the same along the temperature range and probably with a preponderance of
the Eley-Rideal mechanism. To determine the main recombination mechanism between the Eley-Rideal
and Langmuir-Hinshelwood processes for the recombination of atomic oxygen at high temperatures on
these ceramic materials, a study was carried out using Molecular Dynamics Simulation in collaboration
with Cacciatore from IMIP-CNR (Bari, Italy) [29-31, also see in this lecture series].
It can also be noticed the importance of the microstructure of the material as far as the crystalline structure
and the surface morphology are concerned. Around 1000 K, the recombination coefficients determined
experimentally are 27.10-3 for β-cristobalite and 8.10-3 for quartz. Such a trend is confirmed at higher
temperatures (at 1400 K, 60.10 for β-cristobalite and 14.10-3 for quartz). These results confirm the
-3
importance of the crystalline structure because, in spite of a greater specific area for quartz than for β-
cristobalite as revealed by the micrographs obtained by SEM and AFM (Fig. 5), the catalytic activity is
weaker for quartz.
Many models developed to describe the evolution of the recombination coefficient γ with temperature
predict a very important decrease of γat high temperature [2-6, 14, 15]. Figure 4 shows the decreasing of
the recombination coefficient around 1800 K. This effect can be attributed:
• to the change in the reaction order : before the “critical temperature” corresponding to the fall-off of
the recombination coefficient, the reaction is first order and above this temperature (around 1800 K),
the reaction becomes second order with a square dependence on pressure as proposed by Jumper [2] ,
• or to the thermal desorption of atoms that becomes important at high temperature, further reducing the
numbers of adsorbed atoms as proposed by several authors [3, 4, 14, 15, 20].
In order to confirm these experimental results and to eventually prove experimentally the decrease of γ
after a critical temperature, we have then perform measurements at higher surface temperatures up to 2300
K on refractory oxides such as alumina, because at these temperature levels the alumina surface remains
unchanged. Also, further investigations based on AFM characterization to reveal the roughness at a
nanometer scale and the possible active sites have to be done for a better understanding of these
experimental results on alumina.
RTO-EN-AVT-142 12 - 13
γΟ s SiO2 β-cristobalite
m SiO2 quartz
0,1
0,01
0,001
0,4 0,5 0,6 0,7 0,8 0,9 1,0 1,1 1,2 1,3
1000/T (1/K)
Figure 4: Atomic oxygen recombination coefficient γ for SiO2 β-cristobalite

and quartz versus reciprocal temperature
Figure 5: AFM images (10 x 10 µm) for β-cristobalite (left) and quartz (right)
4.2.1.2 Alumina samples
Two sintered alumina called A and B (A for AF 997 Saint-Gobain Céramiques, and B for AluSiK 99 ZA
from Anderman Céramiques) and a coating of alumina obtained by oxidation of an AlN sintered material
called C (AlN from Saint-Gobain Céramiques) were studied. The two sintered alumina A and B have
respectively a purity of 99,7% and 99,8%, a density of 3,9 and 3,8 g/cm3 and an open porosity equal to
zero for both. The AlN material contains around 3 to 5% Y2O3 as a sintering aid and some yttrium is
present in the oxide layer as revealed by XRD and XPS analyses.
The choice of these α-alumina samples to study the recombination was made according to the fact that:
12 - 14 RTO-EN-AVT-142
• we have previously shown that the crystal lattice is an important parameter to take into account for
the catalycity, β-cristobalite being more catalytic than quartz
• so, we chose an α-Al2O3 and modify the microstructure of the material by the presence of
impurities in low or relative high content to see if it is an influent parameter
• alumina is a high stable refractory oxide, that allows to perform measurement of recombination
parameters until very high temperatures with no modification of the material, and also around the
melting point (2323 K) to eventually measure the sharp decrease in the curve Log γ = f(1/T).
The recombination coefficients γ have been calculated for alumina with the measured concentration
profiles, obtained in steady-state conditions, and are plotted on Figures 6 and 7. These results show the
strong catalytic activity of alumina if we compare with the previous results obtained on silica (Fig. 4).
Hence, in all the temperature range, Arrhenius fits of the experimental dots is still possible. We have to
notice that, for both the sintered alumina, around 1400 K, there is a change in the slope of the curves. This
was already observed for the measurement of the thermal flux of recombination on the same alumina at
the same temperature and was not explained (Fig. 3). Now, we think that probably, there is a change in the
recombination mechanism passing from a Langmuir-Hinshelwood reaction (below 1400 K) to an Eley-
Rideal one (above 1400 K). This has to be verified using Molecular Dynamics simulation, such as
demonstrated by Cacciatore on silica [32]. This is also under study with Guerra using modelling with
asymptotic analytic solutions [33].
The following expressions were obtained for γ and the respective apparent activation energies:
• for the A alumina (AF 997):

γ = 0.2162 exp(-1997/T) and Ea ≈ 16 kJ/mol below 1400 K
γ = 25.436 exp(-8747/T) and Ea ≈ 73 kJ/mol above 1400 K
• for the B alumina (Anderman):

γ = 0.1588 exp(-1140/T) and Ea ≈ 10 kJ/mol below 1400 K
γ = 3.4788 exp(-5423/T) and Ea ≈ 45 kJ/mol above 1400 K to 2030 K
γ = 0.26 in the range 2330-2430 K (fused zone)
• for the C alumina (coating on AlN):

γ = 1.896 exp(-5042/T) and Ea = 42 kJ/mol (880 ≤ T ≤ 2000 K)
The accuracy on the activation values is also calculated from several measurements. The two activation
energies obtained for sintered Al2O3 show that the recombination mechanism may be different for the two
temperature ranges around 1400 K. In the thermal study (Fig. 3), this change was also encountered around
1400 K.
RTO-EN-AVT-142 12 - 15
1 1
0,1 0,1
0,01 0,01
0,001 0,001
0,3 0,5 0,7 0,9 1,1 0,3 0,5 0,7 0,9 1,1
1000/T (1/K) 1000/T (1/K)
Figure 6: Atomic oxygen recombination coefficient γ for two sintered alumina samples
versus reciprocal temperature: A on the left and B on the right
0,1
0,01
0,001
0,3 0,5 0,7 0,9 1,1
1000/T (1/K)
Figure 7: Atomic oxygen recombination coefficient γ for alumina on AlN

versus reciprocal temperature
X-ray Photoelectron Spectroscopy (XPS) was used to analyze the surfaces of the tested samples after the
measurement of the recombination coefficient γ [34]. The alumina coating on AlN (C alumina) is
characterized by the presence of a photoelectron peak Y 3d due to the fact that Y2O3 is a sintering additive
of the AlN material present under the oxide layer of alumina. For this sample, N was not detected proving
that the oxide layer is more than 30 nm thick. This oxide layer was formed during a 24 h oxidation at 1300
K under atmospheric pressure in an electrical furnace. A comparison of the Al 2p spectra corresponding to
samples C and A was done. A broadening of the Al 2p peak for sample C can be observed in comparison
with the one of sample A without yttrium. Therefore, the Al 2p peak of sample C was fitted into two
peaks, one located at a binding energy of 74.5 eV corresponding to the Al 2p peak in Al2O3 and the other
located at 73.3 eV. This last peak can be attributed to one kind of Al present in the garnet YAG phase of
composition Y3Al2(AlO4)3 also detected by XRD, this phase having two different sites for aluminum: one
in tetrahedral coordination number (peak at 74.5 eV) and the other in octahedral coordination number
(peak at 73.3 eV). With this assumption, 83% of Al could be in tetrahedral sites and 17% in octahedral
sites on the analyzed surface of sample C. So, probably, the presence of this garnet phase in the surface of
sample C has an influence on the recombination coefficient and leads to one apparent activation energy
equivalent to one main recombination mechanism in all the temperature range.
12 - 16 RTO-EN-AVT-142
4.3 Energy accommodation coefficient β - mesoscopic approach

The chemical energy accommodation is an important parameter that characterizes the catalycity of a
material. Following the determination of the thermal flux of recombination qrec,acc transferred to the
surface, an analytical expression can be proposed to deduce the chemical energy accommodation
coefficient β.
First, it is supposed that the β coefficient can be represented as the ratio of the number of oxygen atoms
recombined at the material surface that transfer their recombination energy to the material N acc
O, rec to the total
number of oxygen atoms recombined at the surface N O, rec . Therefore, the number of oxygen atoms N acc O, rec is
linked to the thermal flux of recombination qrec,acc. Finally, the accommodation coefficient is expressed by:
q rec, acc .S eff. .N A (14)

β= O2
γ .N O, tot .VO .E rec.
with Seff the effective surface area, NA the Avogadro number, γ the recombination coefficient, NO,tot the
O
total number of O atoms, VO the atomic oxygen velocity, E rec2 the energy of recombination to form O2
molecules.
Figures 8 and 9 present respectively the results obtained for both γand
 β coefficients for β-cristobalite and
the A sintered alumina versus temperature. Similar to experimental data obtained on metallic surfaces by
Halpern [16], the β coefficient of cristobalite and alumina is a strong function of the surface temperature
and the resulted values are very different from 1, especially at high temperature.
0,8 silica
β
0,6
0,4
0,2
γ
0
800 1200 1600 2000 2400
T (K)
Figure 8: γ and β recombination coefficients versus temperature for cristobalite (silica)
RTO-EN-AVT-142 12 - 17
0,8 alumina β
0,6
0,4
0,2 γ
0
800 1200 1600 2000 2400
T (K)
Figure 9:γ and β recombination coefficients versus temperature for the A sintered alumina
This study is still in progress because, as shown in Eq. (14), the knowledge of the absolute concentration
of atomic oxygen is required to increase the accuracy on the experimental results. Several techniques are
now envisaged to reach this absolute concentration.
Another important point to increase also the accuracy on the measurement of the recombination
parameters is the evaluation of the gas temperature profile. Some preliminary results were recently
obtained by the measurement of the image of N2 bands in the plasma discharge to reach the rotational
temperature of N2 in the 2nd positive system (C3Πu – B3Πg). The comparison of experimental spectra is
performed with calculated spectra to finally obtain the rotational temperature comparable to the gas
temperature in such non-equilibrium plasmas. Finally, we will have, nearly at the same time, the profiles
of the gas temperature and of the concentration of atomic oxygen above the surface with a spatial
resolution of 270 µm.
5. MOLECULAR DYNAMICS SIMULATION

In collaboration with Cacciatore, to extract more information at the atomic scale level of such
heterogeneous reaction mechanisms, a semi-classical Molecular Dynamics simulation of the
heterogeneous oxygen molecule formation over β-quartz were performed according to the E-R mechanism
at a surface temperature of 1000 K during the PhD of Bedra [34] (co-tutelle Balat-Pichelin and
Cacciatore).
This chapter is not detailed as Cacciatore will present the theory and results in the same conference (cf his
paper). A few words on the method are just given here.
As a starting point, a 3D sample cluster is built up. Then, phonons dynamics is studied using an
appropriate interatomic potential taken from literature. The collision dynamics are performed according to
a semi-classical method taking into account the energy exchange processes between the surface atoms and
the O and O2 particles in the gas phase. The O2 formation probability is calculated at different collisional
energies of the incoming gas oxygen atoms, together with others possible reaction channels taking place at
the surface. Previous MD results obtained for the recombination coefficient γ on β-cristobalite surfaces
[32] were in good agreement with experiments and recently we have demonstrate the different catalytic
behaviour of these two silica polymorphs. The present MD calculations are encouraging since, similarly to
12 - 18 RTO-EN-AVT-142
the experiments, they show much smaller recombination probabilities of the oxygen atoms over quartz
than for cristobalite. Unlike experimental approach, MD provides intrinsic physico-chemical properties as
the vibrational state selected, the vibro-rotational energy distribution of the formed O2 molecules or the
exothermic energy fraction transferred to the material by heat flux that is also the definition of the
accommodation coefficient β.
6. COMPARISON WITH LITERATURE DATA

The comparison with literature data is difficult because most of the studies are realized at room
temperature or at temperature higher than 1400 K but only for RCG (Reaction Cured Glass) coating
(borosilicate glass with 94% silica, 4% B2O3 and 2% SiB4) and, moreover, without any information on the
surface morphology of the samples.
Nevertheless, our experimental results, obtained in microwave-induced air plasma, are consistent with the
results of Jumper on quartz silica (γ= 14. 10-3 at 900 K) [2], of Greaves and Linnett (γ= 14. 10-3 at 873 K)
[35], and of Cacciatore on β-cristobalite (γ= 29. 10-3 at 1000 K) [32]. Kim find lower value on quartz (γ =
29. 10-5 at 1030 K) [11] because of the low pressure used (27 Pa). The results obtained by Deutschmann
[14] and Nasuti [15] modelling on silica agree both with our values, giving γ=10-2 at 1000 K.
The RCG coating gives different values according to Rakich (γ= 5.10-4) [36] or Stewart (γ= 10-3) [37].
We have reported on Figure 10, in order to compare our experimental results with those obtained on arc-
jet plasma facilities, the effective recombination coefficient γ’ calculated as the product (γ.β and that is the
global recombination parameter obtained from heat flux measurements on arc-jet or plasmatron facilities.
The results obtained using actinometry and thermal flux measurements seem to be in relative good
agreement with those obtained using different methods such as those used in arc-jet facilities based on the
relative heating rate of a known catalytic body compared to the studied surface, taking into account the
accuracy of the results and the difficulty of such measurements.
s SiO2 β-cristobalite
γΟ
m SiO2 quartz
0,1
γ'Ο
0,01
RCG Nasa: Scott ( l ) Kolodziej ( n ) Stewart ( u ) Wiley ( ∆)

0,001
0,4 0,5 0,6 0,7 0,8 0,9 1 1,1 1,2 1,3
1000/T (1/K)
Figure 10: Comparison of our experimental results of γ (on β-cristobalite and quartz, in black) and
γ’ (on β-cristobalite, in red) with those of RCG Nasa from [1, 3-5]
RTO-EN-AVT-142 12 - 19
Figure 11 represents the effective recombination obtained for silica and alumina. The effective
recombination coefficient goes from 0.019 to 0.011 for β-cristobalite and from 0.050 to 0.012 for alumina
(sample A) in the temperature range 1000-1800K with a maximum of around 0.08 at 1400 K, also
showing a higher catalycity for alumina.
0,1
alumina
γ'
0,01
silica
0,001
0,4 0,5 0,6 0,7 0,8 0,9 1 1,1
1000/T (1/K)
Figure 11: Effective recombination coefficient γ’ versus reciprocal temperature

for β-cristobalite (silica) and the A sintered alumina
Another example is given by the results obtained for OREX materials in order to compare our results with
those obtained on plasmatron facilities (inductive plasma). Figure 12 presents the results of the effective
recombination coefficient on carbon/carbon composites with an oxidation resistant coating. This work was
done under a CNES-JAXA cooperation to compare the results obtained on ground facilities to the results
obtained during the flight of the OREX re-entry capsule that was lost in the sea. On Figure 12, we report
the values obtained on three inductive plasma facilities: CORIA (France), VKI (Belgium) and
TSNIIMASH (Russia) and compared to our results obtained on the microwave-induced plasma of the
MESOX set-up (PROMES, France). The nominal and minimal values of OREX flight given by JAXA are
also reported.
12 - 20 RTO-EN-AVT-142
1
γ.β VKI CORIA
TSNIIMASH PROMES
0,1
0,01
0,001
OREX flight (nom. - min.)
0,0001
0,4 0,6 0,8 1 1,2
1000/T(1/K)
Figure 12: Comparison of the effective recombination coefficient γ’ for OREX materials
The results obtained by direct (MESOX facilities) and indirect (plasmatron facilities) methods for the
effective recombination coefficient γ’ provide comparable values in all the temperature range when the
recombination coefficient γ from actinometry measurement is multiplied by the corresponding chemical
energy accommodation coefficient β. This normalization is necessary to have consistent data as the
catalycity model in plasmatron methodology assumes an arbitrary accommodation coefficient of unity.
7. CONCLUSIONS
The determination of the atomic oxygen recombination at the surface of different sintered ceramic
materials at high temperature under low air pressure plasma has been realized in the MESOX set-up using
a multi-scale approach.
At the mesoscopic scale, the thermal method is based on a heat balance on materials under air and argon
plasmas. This approach leads to the determination of the temperature increase due to the catalytic
recombination of atomic oxygen and to the evaluation of the thermal flux of recombination transferred to
the surface. A low influence of the pressure has been observed between 200 and 2000 Pa air, that can
come from a limitation of the recombination by the number of active sites and the prevalence of the Eley-
Rideal mechanism, with recombination reactions between an atom in the gaseous phase and one adsorbed,
this being in accordance with literature results. The temperature increase is more pronounced at low
temperature level and decreases rapidly with surface temperature. Thermal fluxes of recombination have
also been calculated from heat balance with a one-dimensional hypothesis. Catalytic activities for the
different materials move weakly between 1000-1400 K, except for Al2O3 for which recombination fluxes
vary strongly and reach a maximum at 1400 K. Most of the studied materials are weakly catalytic (except
Al2O3) and were classified according to a catalytic scale available between 1000 and 1800 K: SiC ≈
SiC+SiO2 ≤ Si3N4 ≤ AlN ≈ AlN+Al2O3 << Al2O3.
A heat transfer modelling at high temperature has been developed in order to identify the absorbed
microwave flux and finally the thermal flux of recombination and, also, to precise the validity of the
experimental assumptions. The catalytic scale is confirmed by using the bi-dimensional modelling and
give more accurate values of the thermal flux of recombination particularly for temperatures lower than
RTO-EN-AVT-142 12 - 21
1400 K; the discrepancy between modelling and experimental values of the thermal flux of recombination
being less important at higher temperature levels (1600 - 1800 K).
At the same scale, the evaluation of the chemical energy accommodation coefficient β has been carried out
to obtain all the parameters describing the atoms recombination on surfaces. Moreover, the measurement
of this coefficient lets possible the comparison with high enthalpy facilities (arc-jet and plasmatron).
At the microscopic level, new experimental results were obtained on the recombination coefficient of
atomic oxygen γ at the surface of ceramic materials at high temperature using relative atomic
concentration measurements (chemical approach). This work has followed the evaluation of the thermal
flux of recombination qrec,acc transferred to the surface material in the same temperature range.
Experimental results have shown the importance of the crystalline lattice (silica polymorphs) and of the
impurity content (polycrystalline α-alumina) on the recombination coefficient.
At the atomic scale, the study done in collaboration with Cacciatore by Molecular Dynamics simulation
has given first interesting results not only on the recombination coefficient but also on the several energy
transfers in the formed oxygen molecule leaving the surface after recombination of oxygen atoms.
Both these data γ and β are very useful for the calculation and design of the planetary entry of hypersonic
vehicles and for comparison with results obtained by indirect methods on arc-jet or plasmatron facilities.
8. REFERENCES
[1] Scott C.D., AIAA Paper 80-1477, July 1980.
[2] Jumper,E.J., Seward W.A., J. Thermophys.& Heat Transfer, 8 (3), 1994, 460-465.
[3] Kolodzeij P., Stewart D.A., AIAA Paper 87-1637, June 1987.
[4] Willey R.J., J. Thermophys. & Heat Transfer, 7 (1), 1993, 55-62.
[5] Stewart D.A., Rakich J.V., Lanfranco M.J., NASA CP-2283, Part 2, March 1983, 827-845.
[6] Zoby E.V., Gupta R.N., Simmonds A.L, Progr. Astron. Aeron., 96, 1985, 445-464.
[7] Clark R.K., Cunnington G.R., Wiedemann K.E., J. Spacecrafts & Rockets, 32 (1), 1995, 89-96.
[8] Cunnington G., Robinson J., Clark R., AIAA Paper 90-1742, June 1990.
[9] Vasil’evskii S.A., Kolesnikov A.F., Yakushin M.I., High Temp., 1991, 411-418.
[10] Gordeev A.N., Kolesnikov A.F., Yakushin M.I., SAMPE J., 28 (3), 1992, 29-33.
[11] Kim Y.C., Boudart M, Langmuir, 7, 1991, 2999-3005.
[12] Pallix J.B., Copeland R.A., J. Thermophysics & Heat Transfer, 10 (2), 1996, 224-233.
[13] N’Guyen-Xuan F., Hassouni K., Cavadias S., Amouroux J., Proceed. 11th Int. Symp. Plasma Chem.,
Loughborough (UK), 4, 1993, 1516-1521.
[14] Deutschmann D., Riedel U., Warnatz J., J. Heat Transfer, 117, 1995, 495-501.
12 - 22 RTO-EN-AVT-142
[15] Nasuti F., Barbato M., Bruno C., J. Thermophys. & Heat Transfer, 10 (1), 1996, 131-136.
[16] Halpern B., Rosner D.E., Trans. Faraday Soc., 74 (8), 1978, 1883-1912.
[17] Melin G.A., Madix R.J., Trans. Faraday Soc., 67, 1971, 198-211
[18] Carleton K.L., Marinelli W.J, J. Thermophysics & Heat Transfer, 6 (4), 1992, 650-655.
[19] Clark R.K., Cunnington G.R., Robinson J.C., J. Thermophysics, 1 (1), 1987, 28-34.
[20] Daiss A., Frühauf H.H., Messerschmid E.W., AIAA Paper 96-1903, June 1996.
[21] Balat-Pichelin M., Robert J.F., Sans J.L., Appl. Surf. Sci., under press, 2006.
[22] Balat M., J. Eur. Ceram. Soc., 16, 1996, 55-62.
[23] Balat-Pichelin M., Duqueroie F., Int. J. Thermal Sci., 40, 2001, 279-287.
[24] Balat-Pichelin M., Badie J.M., Berjoan R., Boubert P., Chem. Phys. 291, 2003, 181-194.
[25] Granier A., Chéreau D., Henda K., Safari R., Leprince P., J. Appl. Phys., 75 (1), 1994, 104-114.
[26] Booth J.P., Joubert O., Pelletier J., Sadeghi N, J. Appl. Phys., 69 (2), 1991, 618-626.
[27] Diamy A.M., Legrand J.C., Al Andari J., New. J. Chem. 21, 1997, 177-185.
[28] Walkup R.E., Saenger K.L., Selwyn G.S., J. Chem. Phys., 84, 1996, 2668-2674.
[29] Balat-Pichelin M., Badie J.M., Cacciatore M., Rutigliano M., 4th Eur. Work. Hot Structures and
Thermal Protection Systems for Space Vehicles Palermo (Italy), ESA SP-521, April 2003, 403-410.
[30] Bedra L., Balat-Pichelin M., Rutigliano M., Cacciatore M., Rarefied Gas Dynamics, AIP Conf.
Proceed Subseries Atomic, Molecular & Chemical Physics, Ed. M. Capitelli, 762, May 2005, 1037-
1042.
[31] Bedra L., Rutigliano M., Cacciatore M., Balat-Pichelin M., to be published in Langmuir, 2006.
[32] Cacciatore M., Rutigliano M., Billing G.D., J. Themophys. & Heat Transfer 13, 1999, 195-203.
[33] Guerra V., Rarefied Gas Dynamics, AIP Conf. Proceed Subseries Atomic, Molecular & Chemical
Physics, Ed. M. Capitelli, 762, May 2005, 981-986.
[34] Bedra L., PhD thesis, Perpignan University, 27 June 2005.
[35] Greaves J.C., Linnett J.W., Trans. Faraday Soc., 55, 1959, 1355-1361.
[36] Rakich J.V., Stewart D.A., Lanfranco M.J., Progr. Astron. & Aeron. 85, 1983, 97-122.
[37] Stewart D.A., Rakich J.V., Lanfranco M.J., NASA CP-2283, Part 2, 1983, 827-845.
RTO-EN-AVT-142 12 - 23
ACKNOWLEDGEMENTS
The author wants to thanks for their fruitful collaboration her colleagues, particularly J.M. Badie and R.
Berjoan, and PhD students from PROMES-CNRS laboratory and from outside P. Boubert (IUSTI
Marseille), M. Cacciatore and M. Rutigliano (IMIP-CNR Bari), O. Gerasimova for AFM (USU,
Ekaterinburg), A. Enzian (CNES, Paris), M. Mozetic and A. Vesel (JSI, Ljubljana).
NOMENCLATURE
C concentration, atom/m3
D binary diffusion coefficient, m2/s
Ea apparent activation energy, kJ/mol
E-R Eley-Rideal
I intensity (a.u)
kB Boltzmann constant
kw catalytic recombination rate, m/s
L thickness of the boundary layer, m
L-H Langmuir-Hinshelwood
M molar mass, g/mol
N number of atoms
NA Avogadro number
P total pressure, Pa
q thermal flux, W/m2
r radius, m
S area, m2
t time, s
T temperature, K
U mass transfer velocity, m/s
V mean square atomic velocity, m/s
x abscissa, m
Greek letters
α solar absorptivity coefficient
β chemical energy accommodation coefficient
∆T difference between two temperatures, K
ε total hemispherical emissivity
γ recombination coefficient
λ wavelength, µm
ω variation of concentration
σ Stefan-Boltzmann constant
Subscripts
acc accommodation
Ar argon
b back face of the sample
eff effective
f front face of the sample
g gas
lat lateral thermal losses at the limit of the reference cylinder
mw microwave (microwave-material interaction)
0 zero (at x=0)
12 - 24 RTO-EN-AVT-142
O atomic oxygen
rad radiative transfer
rec recombination
s surface
sol solar
tot total
Superscripts
air under air flow
arg under argon flow
* under plasma conditions
RTO-EN-AVT-142 12 - 25
12 - 26 RTO-EN-AVT-142
Experimental Studies on Hypersonic Stagnation Point
Chemical Environment
O. Chazot
VKI, 72 Chaussée de Waterloo
1640 Rhode-St-Genese
Belgium
chazot@vki.ac.be
1 Introduction ............................................................................................................................................................................. 1
2 Stagnation point testing configuration..................................................................................................................................... 2
2.1 Methodology................................................................................................................................................................. 2
2.2 Ground testing facility .................................................................................................................................................. 4
2.3 Probes and measurements techniques ........................................................................................................................... 7
2.4 Numerical tools........................................................................................................................................................... 11
2.5 Hypersonic versus Subsonic testing............................................................................................................................ 12
3 Stagnation point reactor......................................................................................................................................................... 14
3.1 BL general definitions................................................................................................................................................. 14
3.2 Chemical environment evaluation............................................................................................................................... 16
3.3 Damköhler probes...................................................................................................................................................... 19
4 Application to the determination of TPS properties .............................................................................................................. 21
4.1 Emissivity determination ............................................................................................................................................ 21
4.2 Catalycity determination ............................................................................................................................................. 22
4.3 Reference catalycity determination ............................................................................................................................. 24
4.4 The 0-model................................................................................................................................................................ 26
5 Conclusions ........................................................................................................................................................................... 27
6 Acknowledgements ............................................................................................................................................................... 28
7 References ............................................................................................................................................................................. 28
1 INTRODUCTION
Development of space transportation is a very challenging task for the aerospace research and industry.
Hypersonic flight should be investigated in details to allow designing spacecraft according to the severe
environment of their flight conditions. Typically during a planetary re-entry, when a capsule or a space
vehicle approaches the relatively dense atmosphere a strong bow shock takes place ahead of the vehicle
detached from its nose. It is subjected to wide range of pressure, heat transfer and shear levels. Several
features, specific of hypersonic regime appear, as thin shock layer, entropy layer, viscous interaction …
which are described in classical textbook [1-3]. Among those phenomena one could remark the high
temperature effects since they appear as one of the critical points in the design phase of the vehicle
mission. Indeed across the shock a large amount of kinetic energy is converted into thermal energy. This
large energy density leads to high temperature of the gas mixture where dissociation and ionization take
place. It results into a plasma flow which impinges on the vehicle wall. To sustain this important heat-
transfer the spacecraft must be equipped with suitable Thermal Protection System (TPS). Their role is
essential for the success of the re-entry manoeuvre, but their design is difficult due to the complexity of
the heat-transfer phenomena. This later can be described with two main contributions [4]: a conductive
heat-flux from the very high temperature flow reaching the wall, and a diffusion one due to chemical
recombination at the wall activated by the catalytic properties of the TPS. The radiative part is left aside in
this description, but one should remember that it could be important in certain condition.
From this situation it could be understood that ground testing appear as a strong requirement for TPS
design. Number one they represent a first convenient step at reduce cost before the launch of a mission.
Chazot, O. (2007) Experimental Studies on Hypersonic Stagnation Point Chemical Environment. In Experiment, Modeling and Simulation of
Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 13-1 – 13-32). Educational Notes RTO-EN-AVT-142, Paper 13.
RTO-EN-AVT-142 13 - 1
Experimental Studies on Hypersonic Stagnation Point Chemical Environment
Number two the tests allow a better control for the environment and for the measurement techniques to
investigate TPS properties. However the development of high–enthalpy facilities for hypersonic testing is
a challenge in itself. It involved the management of very high pressure levels to set-up extreme test
conditions. Moreover, as it is known, a complete experimental simulation of hypersonic condition is most
of the time impossible to achieve in a ground test facility [5]. If not all the conditions could be reproduced
at once a lot of complementary facilities have been set-up. They are of several kinds to address at best
these flight environments. They have been widely used to study the typical features of hypersonic flows
and assess the critical issues associated with aerospace flight [6].
When the investigations concern high temperature effects, and their consequences on TPS material,
specific ground facilities have to be used allowing long time range and chemical reacting flows.
Classically plasma wind tunnel are used: Arc-jet facilities in which supersonic testing in different
configuration is available at very high power [7] and very large scale [8], or Plasmatron facility offering
usually a better chemical environment with subsonic testing [9,10]. This lectures focus on this last type of
facility and attends to present the use of such plasma wind tunnel to study the chemical environment of a
re-entry flight as well as the catalytic properties of TPS material. It will deal with the methodology
adapted for the Plasmatron testing, some aspects of the facility design, the development of the specific
measurement techniques and the data processing in connection with appropriate CFD tools. It will finally
concentrate on the application of this ground testing methodology to the determination of TPS properties
in real flight conditions.
2 STAGNATION POINT TESTING CONFIGURATION
2.1 Methodology
In real flight condition the flow along the stagnation line present very different situations from the gas at
rest in front of the shock up to the wall of the vehicle downstream. Across the bow shock the flow gain a
large amount of thermal energy, it leads to high temperature and excites the different internal energy mode
of the gas particles. Immediately after the shock appears a region of non-equilibrium flow. The excitation
processes re-equilibrate downstream through a relaxation region and then reach an intermediate zone
where Local Thermodynamic Equilibrium (LTE) can be consider in certain conditions. At the wall the
flow enter the boundary layer area which can be mainly consider as a chemical non-equilibrium flow. This
brief description is very schematic and does not pretend to give a complete view of all the processes
occurring in reality. Figure 1 summarized this rough presentation. Many times the different regions
overlap and it becomes much more complex. Classical textbook about gasdynamics must be consulted for
more accurate details [11, 12].
Figure 1:real flight situation (stagnation line)
13 - 2 RTO-EN-AVT-142
The reproduction of this kind of flow environment requires specific facilities. As it has been mentioned in
the introduction the complete re-entry simulation, as it could interest the designer, is impossible or at best
impratical in a laboratory. A classical similarity analysis as it used to be practiced in a wind tunnel cannot
be carry out in this case. However the experimentalists can consider the problem in different parts and
provide sufficient simulation for the study of various aspects of the re-entry flights. For instance,
concerning the heat-transfer for the hypersonic stagnation point, an inspection of the Boundary Layer (BL)
equation as it could be written in those conditions illustrate this situation. Considering the treatment of the
compressible axisymmetric BL equations done by Lees [13] after Cohen and Reshotko [14] using the
Howarth-Doronitzin transformation with Mangler transformation and η as similitude parameter, it can be
read:
ue ∞ ∞
η= ∫ r0 ρdy with s = ∫ µ e ρ e u e r02 dx (1)

2s 0 0
(Only the final result is given here, the derivative of the function f corresponds to the non-dimensional
longitudinal velocity, the function g corresponds to the non-dimensional temperature and z to a non-
dimensional concentration.)
2s ∂ue ⎛ ρ e ⎞
x-momentum: (lf ′′)′ + ff ′′ + ⎜ − f ′2 ⎟ = 0 (2)
ue ∂s ⎝ ρ ⎠
total energy:
′
⎛ l ⎞ µ e2 ⎛ 1 ⎞ ⎛ l 1 c ⎞
fg ′ + ⎜ g ′ ⎟ + ⎜ 2l (1 − ) f ′f ′′ ⎟ + ⎜⎜ Le(1 − )∑ ie hi z i′ ⎟⎟ = 0 (3)
⎝ Pr ⎠ 2hse ⎝ Pr ⎠ ⎝ Pr Le hse ⎠
′ −1
⎛ l ⎞ ⎛ due ⎞ ωi
Atom mass conservation: fz ′ + ⎜ z ′ ⎟ − 2⎜ ⎟ =0 (4)
⎝ Sc ⎠ ⎝ dx ⎠ ρ cis
One can notice that these equations are not similar due to dimensional terms and extra terms accounting
for diffusion and production rate. In some case, as the cold wall at equilibrium with some approximation
the set of equations can be reduced to a similar one [13]. However in the general case, if the production
term vanishes (ωi = 0) by considering equilibrium it appears that He, Pe and du/dx are the values that have
to be reproduced at stagnation point to have identical situations from equations (2) and (3).
In this condition it has been shown that a complete duplication of real flight condition is possible in
ground facility, if the total enthalpy (He), the total pressure (Pe) and the velocity gradient (β = du/dx), of
the flight conditions, can be matched locally on the test sample [4, 13-15]. Experiments based on this
approach could be designed and the very first ground facility set up to study those aerothermodynamics
effects was the shock tube [16]. It allows an accurate duplication of the flow for the stagnation location
[15]. It is still used nowadays to study gas reaction rate in the hypervelocity conditions. Its simple
principle allows a good control of the test environment. But like most of the others hypersonic ground
facilities the test duration (1 ~ 100 ms) are too short to be suitable for TPS testing. These material tests
require high enthalpies associated with long time scale (in this order of the material thermal inertia) to be
representative of re-entry conditions. In this frame facilities like arcjet and Plasmatron have been
developed [17]. They are able to produce dissociated flows for a long time base which is suitable for tests
involving aero-thermochemistry. The tests are restricted to local area where the flight conditions could be
reproduced. It concerns specifically the stagnation point. The complete theory of stagnation point heat
transfer for dissociated flows have been elaborated for hypersonic boundary layers [4], the expression for
equilibrium flow is given below (Le = 1):
RTO-EN-AVT-142 13 - 3
1/ 2
⎛ du ⎞
Qw = 0.67(ρ e µ e / ρ w µ w ) ρwµw ⎜ e ⎟ (He − h ) / Pr
0.4
(5)
⎝ dx ⎠
w
The theory is well suited for hypersonic facilities and in the introduction of their paper Fay and Riddel [4]
make a short remark about the relevance of boundary layer in re-entry problems. Indeed the hypothesis of
the existence of a boundary layer should be preserved for the development of the theory. In the case of
plasma facility where the analysis involving non-dimensional numbers as Mach and Reynolds can no
longer be maintained this question had to be re-examined. In this purpose the theoretical frame has been
adapted for plasma wind tunnel by Kolesnikov, in a methodology called Local Heat Transfer Simulation
(LHTS) [18] and a very careful justification of this development is given in the PhD thesis of Barbante
[19]. The duplication of the flight condition at stagnation point is strictly reduced to the boundary layer
with its appropriate treatment. In the case of a subsonic plasma facility like the VKI Plasmatron [20] the
testing configuration could be presented as in figure 2.
Figure 2 : Testing in Plasmatron facility in LHTS conditions
Knowing the total enthalpy Hs and the total pressure Ps, the extrapolation to the freestream flight
conditions are achieved considering the conservation of enthalpy and momentum along the stagnation line
[18, 21].
H∞ + ½. V∞ = Hs (6)
P∞ + ½. ρ∞V∞ = Ps (7)
2.2 Ground testing facility

The knowledge of the working principles and the operating conditions of the ground facility used to carry
out the experimental study is essential for the interpretation and a relevant discussion about the results
obtained. This lecture presents developments achieved in the VKI Plasmatron and some details concerning
this type of facility is given here.
A plasmatron is a plasma wind tunnel using an Inductively Coupled Plasma (ICP) torch as plasma source.
ICP torches have been developed for 50 years, initially in Russia, in the context of electrical discharge in
13 - 4 RTO-EN-AVT-142
low pressure gases [22]. These devices are used in spectroscopy , material processing, spray coating,
powder synthesis, chemical vapor deposition and in many, generally called Thermal Plasma Processes
(TPP) [23]. Their applications to high enthalpy wind tunnel are mainly due to their high purity and high
energy density flow characteristics which are very beneficial for aerothermodynamic studies. The
stagnation configuration of TPS testing with ICP torch is sketch in figure 3.
Figure 3 : Sketch of an ICP torch for TPS tests in a Plasmatron facility
The basic concept of ICP torches is also schematically shown in this figure (fig. 3). A coil with a few
turns, connected to a high-frequency generator, surrounds a quartz tube, where gas is injected (argon, air,
CO2). Once an initial ionization is created by suitable means, the coil induces eddy currents in the
conducting gas, transferring energy and maintaining it into a plasma state.
The so-called “torch” is the chamber where the plasma discharge takes place. It is designed basically with
two coaxial flows inside the quartz tube surrounded by the coil. The plasma region can be divided in two
parts, the most sensitive part is the coil region called “the discharge region” where all the parameters
interact strongly and the downstream part named the "plasma jet" (fig.3). Historically this design has been
evolving during the last two decades, as a result of direct observation of the torch discharge and of its
behavior [24-28]. For the same purpose, measurements of emission intensity of samples introduced into
the plasma have been made. By these methods “empirical laws” have been established to match the
minimum flow rate [26], the tube configuration [27], and critical geometrical parameters in an ICP torch
[28]. All these criteria led to a torch design characterized by an easy ignition, a good stability and a low
consumption of plasma gas. A correct axi-symmetry of the of the plasma jet is also of major importance
for the stagnation point testing. This aspect has to be attentively checked and the torch properly designs to
achieve good testing conditions [29]. Torch design is a very important point for a Plasmatron facility it
could be investigate by careful experiments [30] joined with CFD computations [31].
The fundamental phenomenon occurring in an ICP torch is radiation/matter interaction. The plasma
discharge is heated by electrical currents induced by a time varying magnetic field and fed by a gas flow.
This leads to an interaction between the fluid flow field, the temperature field and the electromagnetic
field in the discharge region. As shown in figure 4, each field directly interacts with the other two.
RTO-EN-AVT-142 13 - 5
Figure 4 : Modes of interaction between the fluid flow field, the temperature field
and the electromagnetic field
The operation of an ICP torch can be described in a more practical way. Basically the discharge depends
on two parameters: the magnetic field imposed by the current intensity and the coil design, and the flow
field imposed by the gas mass flow rate and the injection system design. The temperature field appears as
a result of the interaction of these two parameters.
Unsteadiness of the discharge is a more complex problem because of all the interconnected parameters.
Nevertheless two main sources of unsteadiness can be distinguished: the electrical effects and the fluids
dynamic effects. The electrical fluctuations due to the time varying electro-magnetic fields represent a
very high frequency (1/f = τw ~ 10-6s) and can be average in comparison of the characteristic time of the
flow (D/U = δt ~ 10-3s). However some interactions exist: the modulation of the current after the rectifier,
in the power generator, is reflected in the energy deposition into the plasma discharge. This phenomenon
occurs on a time base of the order of the flow time (τe ~ 10-3s). It could be reduced by adjusting the plasma
setting. In addition to those phenomena one have to consider the influence of the confined recirculation
zone in the ICP torch which induce a vortex shedding and creates unsteadiness traveling in the plasma
flow [32].
The plasma facility used at VKI for the testing is a Plasmatron-type using an Inductively Coupled Plasma
(ICP) torch (fig. 5). Its basic working principle and operation, as well as the test configuration correspond
to what has been presented above (fig. 3). The main control parameters are the mass flow injected into the
torch, the pressure in the test chamber and the generator power imposed. In the context of TPS testing and
catalycity determination it could be relevant to mention that it is an appropriate facility for those purposes:
It allows testing in a subsonic plasma jet, where LTE conditions are more likely to be verified, produced
by a ICP torch which provide chemical purity. This last point has been checked by spectroscopic
measurements for the most powerful emission line of copper, silicon and iron [33].
13 - 6 RTO-EN-AVT-142
Figure 5: VKI Plasmatron facility
Beside the ICP torch the facility is composed of a test chamber in which the plasma jet exit. Measurement
probes as well as a sample holder are set in the lateral sides of the chamber, ready to be injected on the
centreline of the jet for intrusive measurements or testing purposes. A 3-axes displacement table allow an
accurate positioning for the stagnation point testing configuration (fig. 6). A powerful pumping device
allows to control the static pressure in the chamber during the tests.
Figure 6: Sketch of the Plasmatron test chamber instrumentation
2.3 Probes and measurements techniques

The testing procedure for TPS material has been developed at VKI with three kinds of probes: heat-flux
probes, Pitot probe and the sample holder itself. The heat-flux and Pitot probe have been firstly used for
the qualification of the Plasmatron facility and to assess its operating envelope 20]. All these experimental
tools have been presented with details in previous publications [20, 34]. Their main characteristics are just
mentioned here. The rest of the presentation will preferably deals with more recent developments for the
calibration procedure and the combined measurements.
The measurement probes have the same external shape to preserve the same characteristics for the
stagnation point boundary layer in term of dynamic pressure, stagnation point heat-flux and velocity
gradient. Their geometry is known as the ESA Euromodel: flat faced cylinder (50 mm in diameter) with
RTO-EN-AVT-142 13 - 7
rounded corners (10 mm radius) (fig. 7). The probes are water cooled and mounted on a foot to allow their
injection in the plasma jet. A cylindrical channel along the centre of the probe body is available for the
passage of the sensor device. At the front a sensing element of 14 mm in diameter is installed in the
cylinder. It corresponds to a water cooled calorimeter for heat-flux measurement (fig. 7) or a water cooled
pressure orifice for the total pressure measurement (fig. 8).
Figure 7: Sketch of the heat flux probe, with water-cooled calorimeter installed inside
Figure 8: Pitot probe mounted to 50mm diameter probe
The Sample holder is composed of two coaxial tubes in which water circulates for cooling. At the
extremity, a SiC cover is attached to the holder body by three metallic pins. The sample to be tested is
installed in this support, maintained at its periphery by the cover and at its back by an insulator (fig.9). The
sample holder has the same shape as the stagnation point probes to preserve the similitude in the test
conditions. The surface temperature is measured on the front part by a 2-colors pyrometer (λ1:0.75-1.1µm,
λ2:0.95-1.1µm). The conductive heat flux passing through the TPS sample is measured at the back, after
an insulating material, by a water-cooled calorimeter (fig. 9).
13 - 8 RTO-EN-AVT-142
Figure 9: Sketch of the Sample Holder
Stagnation point heat-flux calibration for plasma flow is not trivial to realize since it is difficult to
reproduce and control the thermal environment of the dissociated gas in a reference test bench. Several
calibration benches, involving radiative heat-transfer, have been tested at VKI but the problems in
determining the reference heat-flux were limitating to achieve accurate results [ref VKI reports]. They
end-up in a never ending loop to reach a relevant calibration procedure. To come over those difficulties
the method finally set-up at VKI proposes to compare in-situ the current measurement technique, using
water cooled calorimeter with a reference one, using a slug calorimeter, following the designation of the
American Society for Testing and Materials (ASTM) [35]. This procedure presents the advantage of direct
measurements in the same environment and the same configuration that the calorimeter to calibrate. The
calibration process requires careful and diligent experiments with an appropriated Data Acquisition
System to allow high enough sampling rate. In parallel to those a data processing method have been set-up
at VKI to exploit at best the transient measurements an evaluate the related losses. Only a schematic view
of the Copper Slug installation used for the calibration in the VKI plasmatron is shown in figure 10, the
complete study is reported in a VKI Technical Note [36]. The final accuracy for the current stagnation
point heat-flux probe is found to be 7 %.
Teflon ring
Plasma Jet Copper Slug
Type K
Thermocouple
Probe
Special Tape
Air Gap
Ice Bath BNC Connection
Figure 10: Schematic view of Copper Slug Calorimeter for heat-transfer calibration
A combined probe which permits to measure simultaneously heat-transfer and total pressure at stagnation
point has been designed at VKI [37-39]. It is based on a classical measurement probe in which the water
RTO-EN-AVT-142 13 - 9
cooled calorimeter is equipped with a pressure line for total pressure evaluation (fig. 11). A test campaign
has been carried out to assess its performance and to compare its measurement results with the classical
heat-flux probe ones; good agreement has been found (fig. 12). The integration of the combined probe is a
major advantage to determine the stagnation point characteristics in a single probe injection. This probe is
permanently installed in the Plasmatron test chamber. It allows checking the testing conditions just before
placing the test article into the plasma jet. This procedure avoids the problems link to repeatability issue
and lead to a large improvement in the test accuracy.
Figure 11: Combined calorimeter
Figure 12: Heat-flux and Combined probe comparison
13 - 10 RTO-EN-AVT-142
2.4 Numerical tools

The methodology as presented in the first section, adapted to the experimental stagnation point
configuration requires as well the use of CFD codes. It is usually presented as an hybrid methodology
where experiments and CFD are intimately linked. If the facility allows to set up the boundary layer
physically at the stagnation point, the CFD allows to describe it and determine its parameters with two
main numerical codes:
1- The ICP code simulates the experimental configuration. It computes the flow in the ICP torch and
around the sample in the test chamber. It solves the time averaged magneto-hydrodynamic equation at low
Mach and low magnetic Reynolds number. In this computation local thermodynamic equilibrium and
axisymmetric flow is assumed. It was developed by D. vanden Abeele [40] and T. Magin [41].
2- The BL code simulates the boundary layer that develops in the neighborough of the stagnation point. It
solves the laminar chemically reacting boundary layer for axisymmetric or 2D flows. It was developed by
P. Barbante [19].
These codes are interfaced with the Pegase library, written at VKI [42], which performs the computation
of thermodynamic and transport properties for arbitrary gas mixture as well as the chemistry of the flow.
ICP furnish a NS solution from which the external conditions for the boundary layer problem are
determined by Non Dimensional Parameter (NDP). They represent a non-dimensional axial velocity (Ve),
a non-dimensional radial velocity gradient (u1e), a non-dimensional derivative of the velocity gradient (u1y)
and a non-dimensional thickness for the boundary layer (∆). Some more details will be given about the
NDP in the next section (fig. 15).
The numerical data are transfered from ICP to the BL code which computes the stagnation point heat-flux.
By comparing the numerical heat-flux obtained with the experimental one measured in an iterative
procedure the dimensional conditions at the edge of the BL, assuming LTE, can be determined. The heat-
flux measurement is realised with the copper calorimeter which is assumed to be fully catalytic, in this
condition its value can be consider fairly close to what one should obtained for a equilibrium boundary
layer. The rebuilding procedure for the methodology is summarized in figure 13. It allows to determine the
enthalpy and the dimensional velocity gradient at the stagnation point.
RTO-EN-AVT-142 13 - 11
Figure 13: Rebuilding procedure
2.5 Hypersonic versus Subsonic testing

For aerospace ground testing, in one hand hypersonic facilities allow a good reproduction of stagnation
point condition but offer very short test time. In the other hand Plasmatron facilities are ideal with long
test run, for TPS design, but their duplication is strictly limited to stagnation point heat transfer and
require a proper treatment for the boundary layer (BL) model. According to the methodology presented all
these facilities generate analogous BL conditions with the one existing at the stagnation point in a real
flight. In order to compare their actual performance, in terms of testing condition, one could use empirical
stagnation point heat-transfer relations which have been developed to predict, with a simple and accurate
method the stagnation point heating rate for various gases in high enthalpy facilities [43]. It results in a
compilation of numerous experimental data and predicting methods for high enthalpy levels. Such
relations allow to compare results obtained in hypersonic testing as it can be found in many references
[43-45]. The simplest one is expressed by a linear relation between the heat flux rate (Qw), normalized
with an effective radius (Reff,H) divided by a stagnation pressure (ps) and the enthalpy difference at the
stagnation location (He-hw). This formulation is derived from a rearrangement of Fay & Riddell equation
at equilibrium (5), it reads:
Reff , H
Qw = K h ( He − hw ) (8)
Ps
13 - 12 RTO-EN-AVT-142
Kh is a constant deduced from the compilation of data for each gas mixtures [43]. The concept of effective
radius (Reff,H) is used for non-spherical bodies. It corresponds to the radius of an equivalent sphere leading
to the same velocity gradient at stagnation point. Reff,H is mostly useful for flight extrapolation. In
hypersonic, where the modified Newton theory applied, Reff,H is easily defined in the expression of the
velocity gradient (β) [46, 47]:
1 2 pt ⎛ p∞ ⎞
β= ⎜1 − ⎟⎟ (9)
Reff , H ρ t ⎜⎝ pt ⎠
Expression (8) provides a way to plot in a same graph experimental results obtained from different
hypersonic tests that can be found in the literature for equivalent conditions. In order to compare these
latter’s with the VKI Plasmatron results achieved in a subsonic air plasma one has to found a proper way
to normalized the subsonic results as it is known that the expression of velocity gradient (i.e effective
radius) is different in this case. Starting from the same equation (5) an empirical relation for the stagnation
point heat-transfer in subsonic could be found [47-49].
R eff , S
Qw = K s ( He − h w ) (10)
4 Pe ⋅ ∆Ps
∆Ps is the dynamic pressure and Reff,S is expressed in this case with the NPD defined at the edge of the BL
computed by ICP as it has been presented in the previous paragraph.
Rm
Reff , S = 1/ 2
(11)
⎛ ve u1 y ⎞⎟
u1e ⎜1 +
⎜ u12e ⎟
⎝ ⎠
Using expression (8) for hypersonic cases and (10) for the subsonic one experimental results taken from
the references [15, 43, 50-55] for the hypersonic testing can be compare in the same graph with VKI
Plasmatron data [56] (fig. 13).
RTO-EN-AVT-142 13 - 13
Figure 14: Comparison of hypersonic and subsonic results for stagnation-point heating rates
The hypersonic and subsonic results properly normalized compare reasonably well. From the graph the
constant Ks appearing in the relation (10) for the plasmatron tests is close to the constant Kh defined in (8).
The good matching of the results allow to establish the correspondence between hypersonic and subsonic
effective radius and to assess the flight extrapolation related to the subsonic case. This equivalence has
also been checked by numerical computation [21, 57]. Even if the points are fairly well aligned many
comments could be done on the scattering of the data. In all the rebuilding procedure the LTE hypothesis
is always invoke, but this point remains questionable. The predicting methods represented, from the
quoted references, reveal quite large dispersion for high enthalpies. The prediction of Onera [50] is limited
by two curves (Onera+ and Onera-) which represent the extreme values and illustrate the large scattering
of the data. One could remark that the enthalpy determination may bring a lot of differences in the final
results depending on the method used [17, 56]. However this method could be useful to compare testing
conditions in different plasma wind tunnel as well.
3 STAGNATION POINT REACTOR
3.1 BL general definitions

In the previous section it has been seen that thanks to a “local similarity” form real flight stagnation-point
boundary layer could be duplicated in a Plasmatron facility. Assuming LTE at the outer edge of the
boundary layer the enthalpy (He), the total pressure (Pe) and the velocity gradient (βe) are the three
parameters to reproduce for matching the conditions. In this case LTE was considered and equation (4)
13 - 14 RTO-EN-AVT-142
from the set of BL equations was ignored. Another way to disregard equation (4) is to assume a frozen BL
in which the production terms also vanish. This situation has been treated by Fay & Riddell as an extreme
case [4]. In the present section we would like to inspect more general feature of the reacting BL including
diffusion and reaction processes, taking into account the full set of BL equations.
The “big picture” of the BL, from where the NDP are determined is composed by an axial velocity, a
radial velocity gradient profile and a temperature profile. These normalized quantities on the stagnation
line in front of the calorimeter wall are represented in figure 15 with the mention of the different NDP.
Figure 15: General view of the stagnation-point BL
In top of this view species profiles of the gas mixture should appear as they are computed by the BL code.
From here one could see that, if we could check, in the previous section, the ability of a ground testing
facility to duplicate the outer conditions of the BL, one has now to investigate how the inner conditions
could be match as well by reproducing the term for chemistry production as it appears in equation (4). The
question could be put in this way, in which limit a plasma facility is able to generate the same chemical
environment as it exists in a real flight stagnation point. The question will not be address directly since it
is far beyond the scope of this lecture.(Such investigations would be based at least on heavy flight
experiments, modeling development and a number of optical diagnostics).
A first step for the assessment of the plasma testing is to set up tools to evaluate the chemical environment
at the stagnation point. The very first and more direct way is the laser based optical diagnostics in the
region of the stagnation point as they are currently developed at VKI or already realized in some facilities
[59]. In any ways the chemical phenomena happening in the BL will be inspect in fine through a model.
One can then take the problem by the other end, since different chemistry models are implemented in the
BL code, and explore which chemical situation can be produce with the VKI Plasmatron facility, in
stagnation point configuration.
RTO-EN-AVT-142 13 - 15
Usually a reacting BL is described by its dynamic BL, thermal BL and chemical BL in which different
diffusion processes take place and lead to the different profiles already mentioned (fig. 15). It can also be
describe with characteristic non-dimensional number as Pr (Prandtl number), Le (Lewis number) and Sc
(Schmidt number). This characterization is well suited when we consider the ideal diatomic gas [4, 13],
but it becomes less effective for gas mixture with a lot of species to take into account. Our approach is to
consider the stagnation point as a local chemical reactor (Stagnation Point Reactor: SPR) where gas
chemistry happen as well as surface chemistry. Its control parameters have been already mentioned, they
are:
At the BL outer edge: He, Pe, βe.

At the wall: γ (recombination probability coefficient, as it will be presented in the last section).
In the gas: Da (Damköhler number, to be defined in the following paragraph).
This description requires NDPs coming from ICP computations which have been organized at VKI in a
database for a large number of Plasmatron testing conditions [60]. Having this complete definition we will
focus on the chemical environment produce in the SPR.
3.2 Chemical environment evaluation

Flow chemistry involves a very large number of processes and microscopic phenomena, they are usually
summarized in a set of chemical reactions, with their own kinetics, which still constitute numerous
forward and backward reactions. To account for the chemical state of the mixture is a quite complex
problem. The only reference state is equilibrium when all reactions balance each other and end-up in a
stable composition. For chemical non-equilibrium different point of views could be adopted. A
characteristic non-dimensional number (Da) has been suggested by Damkölher [61] to account for the
chemical (non)equilibrium of a gas mixture. He proposed to compare a characteristic time for the
chemistry to happen (τc) with a characteristic time of the flow (τf) in the same region:
τf
Da = (12)
τc
This number could be regard as a chemistry scaling:
For chemical equilibrium flow: τf >> τc : Da → ∞
For frozen flow: τf << τc : Da → 0
For non-equilibrium flow Da will take a finite number that could be ranged as it will be deal with in the
following. The same non-dimensional quantity can be found if one considers the last term of the LHS of
equation (4). β−1 = (du/dx)-1 appears as a mechanical time when (wi/ρi) has a dimension of a time for the
chemical production. This last term of the equation can be interpreted as a Danköhler number.
To build a chemistry scaling for the SPR one has to define a relevant flow time and an appropriate
chemical time. By following the statement of Fay & Riddell [4] the inverse of the velocity gradient,
(du/dx)-1, could represent at first glance “the time for a particle in the free stream to move a distance equal
to the nose radius, and thus also the time for a particle to diffuse through the boundary layer at the
stagnation point” (sic). This term is convenient to choose for a characteristic flow time since it is one of
the parameters of the SPR and it is computed using a NDP in the rebuilding procedure, τf = (du/dx)-1.
A chemical time is more difficult to define since one has to face a large number of very different times
associated with all the chemical processes happening, and it is not easy to pick up just one representative
for the chemical state. Different strategies could be found in the literature. In a case of the ideal diatomic
gas Fay and Riddell proposed a recombination rate parameter (C1) :
13 - 16 RTO-EN-AVT-142
−1
⎛ du ⎞
C1 = K1 ρ Ts
2 − 3.5 −2
R ⎜ e⎟ (13)
⎝ dx ⎠ s
s
It can also be expressed as:
−1
⎛ du ⎞ ω
C1 = ⎜ e ⎟ ⋅ i (14)
⎝ dx ⎠ s ρ i
We recognize the term of equation (4) which is no more than a Damköhler number for the binary mixture.
This approach as also been followed by Baumgart at VKI to assess the design of an enthalpy probe [62].
To treat an air mixture an overall production rate (R) considering the dissociation-recombination process
can be defined for A2 molecules [63]. A2 being O2 or N2 and M a generic collision particle. With kf and kb
the reaction rate constant R reads:
R = Σ (-kf [A2][M] + kb [A]2[M]) (15)
A characteristic chemical time is formulated as the inverse of the ratio of R over the sum of all the species
concentrations in the mixture. The advantage is its validity for multi-elemental mixture but its physical
meaning is not straight forward.
A more general way for building a characteristic time scale for the chemistry can be found in the approach
of Barbante [19]. It is proposed as a first approximation to linearize the expression of the mass production
rate in function of the composition around a known state (ωi0). Instead of having as in the general case a
system of non-linear ODE it can be written:
ωi = ωi0 + [A].ρi [A ] = ⎡⎢ ∂ω ⎤
⎥
i (16)
⎣ ∂ρ i ⎦
ωi and ωi0 are vectors of dimension i and [A] is a matrix of dimension i*j. Each element (i,j) of the matrix
has a dimension of 1/t. They could be interpreted as the sensibility of species i production rate to an
evolution of species j quantity in the mixture. Barbante suggests to find a norm for [A] to compute a
characteristic chemical time.
Following this approach, Herpin in her report tackle the problem of establishing a chemical time from this
“sensibility matrix” [64]. This matrix is a jacobian matrix of the mass production rate made of 5 x 5
components if one considers a 5 species air mixture. It is computed by the BL code all along the
stagnation line in the boundary layer. To extract a single value representative of the chemical time in the
SPR from all these numbers physical hypothesis are take into account.
We first selected particular species, O2 and N2 molecules, which specify two lines on the matrix. In each
line we focus on the positive terms, because we assume that recombination is the main reaction occuring
in the SPR, and drop the negative on. On the components which are left we pick-up the one concerning the
major population. For doing so each term representing the sensitivity to a specie is weighted by the mass
fraction of this specie. With these considerations the characteristic time for a molecule i is expressed like:
RTO-EN-AVT-142 13 - 17
1 1
= ⋅ ∑ (a ∗ xj ) (17)
τi ∑ xj
ij
j ( a >0 )
j ( a >0 )
τi is computed for O2 and N2 along the stagnation line. The typical time is selected from the (1/ τi) profile
on the location corresponding to half of the final mass fraction for equilibrium. More details on these
developments are given in VKI publication [64] where the Da computation is included in the rebuilding
procedure [65]. An example extract from [64] is shown (fig. 16) for two chemical models, Park [66] and
Dunn-Kang [67]. The typical conditions for the SPR are: He = 26 MJ/kg, Pe = 202 mbar, β = 731 s -1.
Figure 16: Chemical characteristic time for O2 and N2 recombination

from different chemistry models
The Danköhler numbers for O2 and N2 computed in this situation give:

DaO2 = 6.83*10-2
DaN2 = 1.21*102
It could be interpreted using a classical scaling for Da that N2 recombination is just at equilibrium, when
O2 recombination is in non –equilibrium near to a frozen flow in the SPR considered (fig. 17).
Figure 17: General scaling for chemistry with Da
13 - 18 RTO-EN-AVT-142
3.3 Damköhler probes

The chemical state can be evaluated at the stagnation point for given parameters of the SPR by a proper
computation for Da number as presented above. On the experimental side the interest is to control the
chemical environment in order to be able to adjust it in order to reproduce real flight conditions.
Concerning ground testing for aero-thermochemistry it would be very useful to have the ability of
scanning a large range of Da numbers.
In the SPR chemistry is ruled by a model and is influenced by the parameters of the reactor like He, Pe
and β which are somehow interconnected. In this situation there is no direct control on the chemical time.
However the Da number could be changed independently by acting on the flow time given from the
velocity gradient. On that purpose several probes were designed based on a generic geometry. Their shape
is defined by two parameter: a body radius (Rb) and a corner radius (Rc) (fig. 18). It corresponds to the
geometry of the heat flux probe prescribed in the first section and to typical parameters used for probes in
hypersonic testing as well [46].
Figure 18: Generic geometry for the Da probes
These probes are called “Danköhler probes” since each of them addresses a particular domain on the Da
number scale. It has been decided to have, in addition to the standard probe used for typical non-
equilibrium BL, a “equilibrium” probe and a “frozen” probe. During the design process CFD
computations have been run on different geometries to determine their respective performance in term of
velocity gradient for given plasma jet condition [64]. Three probes were finally designed, manufacftured
and tested for the plasmatron facility.
The characteristic geometries of the Danköhler probes used at VKI are given below (in mm):
“Equilibrium” probe: Rb = 57.54 Rc = 5
Standard probe: Rb = 25 Rc = 10
“Frozen” probe: Rb = 15 Rc = 15
The classical water-cooled calorimeter can be installed in each of the Da-probe to allow stagnation point
heat-flux measurement in the test chamber (fig. 19-21). The “equilibrium” probe is well suited to generate
equilibrium BL, when the “frozen” one is efficient to produce frozen conditions. They give a larger
variability to the SPR and each of them could be used for specific applications as it will be shown in the
next section.
RTO-EN-AVT-142 13 - 19
Figure 19: Equilibrium probe in the VKI plasmatron
Figure 20: Standard probe in the VKI plasmatron
Figure 21: Frozen probe in the VKI plasmatron
13 - 20 RTO-EN-AVT-142
4 APPLICATION TO THE DETERMINATION OF TPS PROPERTIES

Having defined the external parameters for the stagnation point BL and set-up a method to control and
evaluate the chemical environment of this region we address in this section the conditions at the wall
which represent the other limit of the boundary layer. This limit is of particular interest since it
corresponds to the TPS location. Thus using the experimental apparatus to operate in stagnation point
configuration with the measurement techniques and the processing tools developed the TPS can be tested
in the VKI plasmatron facility for a specified real flight condition. In this representative situation one
could test and investigate the material properties.
At the wall the heat transfer balance reads, with a general form for the diffusion term and neglecting the
radiation from the gas:
∂T G G
−λ + ∑ hi ⋅ J i ⋅ n = σεTw4 (18)
∂n i
The properties of the surface are represented by the emissivity (ε) and the wall chemical activity, roughly
speaking, that it called “catalycity” in these notes that influences the diffusive flux Ji of the LHS. They are
the two properties that the designer needs to assess the performances of a TPS (disregarding the
mechanical strength).
4.1 Emissivity determination

If one considers TPS testing with an “equilibrium” probe type, it allows to impose high heat-flux
comparable to the one of a planetary re-entry without any chemistry involve. In these conditions the outer
edge enthalpy could be determined by heat-flux measurement and the rebuilding procedure, as presented
above, with the rest of the parameters for the SPR. In that case the heat-flux reaching the wall could be
computed depending on the surface temperature (Tw) by the BL code, it is also given by equation (5).
With a TPS sample mounted in an equilibrium-probe type and instrumented as it is shown in figure 9 the
heat transfer at the surface can be expressed like :
Qweq(Tw) = σ.ε.(Tw)4 + Qcond (19)
Qweq is the heat-flux measured by the water-cooled calorimeter for the equilibrium-probe and Qcond is the
conductive heat-flux passing through the insulating material at the back of the TPS sample and measured
by a calorimeter (fig. 9). One can plot in the same graph (Qweq(Tw) - Qcond) and a family of curves
fε(Tw)=σ.ε(Tw)4 for several ε values. They intercept for different Tw (fig. 22). The global emissivity
coefficient could be determined by measuring the surface temperature of the sample (Tw) in the testing
condition. A precise experimental set-up with several pyrometers, properly calibrated, is required to
achieve accurate surface measurements in this situation. At VKI it is proposed to proceed at a first
estimation with a 2-colors pyrometer, calibrated in its experimental environment.
RTO-EN-AVT-142 13 - 21
Figure 22: Procedure for a global TPS emissivity determination
4.2 Catalycity determination

In a chemical reacting boundary layer the surface can be highly involved. As a general statement one
knows that three bodies collisions are more efficient for chemical reactions. This combination allows a
better management of the quantum exchanges between particles, to say it short. In the same way the wall
is playing the role of a third body which makes easier the reactions to happen. Several situations can be
considered. The surface is “inert” and do not promote any particular reaction, it corresponds to a “non-
catalytic” wall, or the wall can activate any reactions at its surface and be called a “fully-catalytic” wall. In
between these two extreme cases one has to determine the definite catalytic properties of the surface.
In the case of re-entry conditions the determination of catalytic properties of TPS materials is a major
issue for the aerospace vehicles. These properties strongly affect the heat transfer to the materials with up
to a factor of two greater heat flux for a fully catalytic material compared to a non-catalytic material.
Knowledge of TPS catalytic properties is extremely important for designing aerospace vehicles that have
very stringent mass budget, for reusable launch vehicles the problem is even more critical.
Plenty of interactions happen at the microscopic scale close to the wall and many mechanisms are
proposed in the literature [68-70]. Different designation can also be found which do not necessarily come
under the same meaning, as catalytic recombination, catalytic properties, catalysis, catalysticity or
catalycity [71-74]. We choose, in this note, to call this “gas-surface interaction” by a neologism:
catalycity. It reflects the fact that we are considering a macroscopic behaviour for a collection of complex
microscopic phenomena, besides we also distinguish this global reaction from the catalysis more precisely
known by the chemists.
The basic phenomena that we consider in the situation of the SPR are the diffusion/reaction of species
through the BL and the probability of recombination at the wall. This probability of recombination at the
wall is defined as the catalycity of the wall (γi). In our model we called it effective catalycity (γ) since it is
assumed that the catalycity of each species are the same. Moreover it includes another characteristic of the
wall usually called β, accounting for the energy accommodation of the surface. γ is then the product of the
13 - 22 RTO-EN-AVT-142
recombination probability γi, assumed equal for all species, with β. It appears as a production term for the
boundary condition in the BL problem to compute. This term is built with the impinging mass flux
(mi.Mi↓) multiply by the probability coefficient γ:
ωicat = γ.mi.Mi↓ (20)
For a fully catalytic wall γ equal 1 and for a non-catalytic wall γ equal 0. For a finite catalycity γ takes a
value between 0 and 1 which represents the ability of the surface to enhance recombination. This boundary
condition is implemented in the BL code to compute the heat-transfer at the wall depending on γ.
The determination of the catalycity by a plasmatron testing in stagnation point configuration results in an
identification process. Heat-flux combined with pressure measurements with the standard probe are
primarily achieved on a reference copper water-cooled calorimeter for which γ is assumed to be 1. This
hypothesis is often done for convenience and its validity will be discussed in the following section. It
allows to consider that the measurement corresponds fairly to the equilibrium heat-flux. Nevertheless one
can point out that the equivalence between the equilibrium heat-flux and the heat-flux on a fully-catalytic
surface is strictly valid only if Le =1.
In any case starting with a heat-flux and a pressure measurement the rebuilding procedure allows to
determine the enthalpy, and the other parameters of the SPR, as soon as the corresponding NDP are given
with a known catalycity. Thus the enthalpy is obtained for the experimental conditions.
With this value and the parameters of the SPR the heat-transfer (Qw) can be computed varying the surface
temperature with γ as parameter. A series of curves are obtained which form a heat-flux abacus, in a Qw-
Tw plot, from the fully catalytic condition to non-catalytic one (fig. 23). The TPS sample to be tested is
placed in the same stagnation point condition (same plasma jet with same probe geometry) and the surface
temperature (Ts) with the total heat-flux at the surface (Qtot) are measured by a 2-colors pyrometer for the
radiative part and a water-cooled calorimeter for the conductive one (fig. 9). By plotting the experimental
results (Qtot, Ts) on the heat-flux abacus the effective catalycity can be identified by interpolation (fig.
23).
Figure 23: Example of heat-flux abacus for γ identification
RTO-EN-AVT-142 13 - 23
This γ determination method is based on several hypotheses and models that have to be furthered checked,
but it already allows to characterize a surface material for TPS design. It is a quite long procedure (fig. 24)
and one has to assess the accuracy and the validity of the hypotheses at each step. Such checking has been
initiated at VKI carrying out a critical approach on the numerical side [60] and a careful calibration on the
experimental side [36]. It has been found that the uncertainty on the catalycity determination is mainly due
to the heat-flux measurement in relation to the enthalpy rebuilding. A precise set-up operating with
successive probe injection for the testing is highly recommended to get rid of the repeatability issue of the
plasma testing.
Figure 24: Procedure for effective catalycity determination
4.3 Reference catalycity determination

By looking closer, the previous method for catalycity determination provides only a relative value for the
effective catalycity compared to the cold copper surface which is taken as reference. The determination
procedure in itself does not present major limitation except the fact that the reference catalycity is not
known and is assumed to be 1 a priori. This assumption is somehow not realistic and present some
physical inconsistency as it is often pointed out [19]. It is usually said that it can be viewed as an ideal
situation and because of the typical S-curve of Qw vs. γ and the uncertainty on calorimeter measurements
only the “highly catalytic” hypothesis for the copper surface is of importance. Taking γ=1 will not change
drastically the results and is in some way a conservative approach for TPS design. Nevertheless the
situation can be sometime more critical and a better accuracy on the value of γ is required. In this case one
should work with a known value of γ as reference or determine the reference catalycity. It could be
thought to carry out these measurements in another facility as a chemical reactor which allows an absolute
determination. But such experiments are not easy to realise due to the oxidation of the surface cause by the
low operating pressure of those devices. Moreover, as for the heat-flux calibration, the measurement
should be done in the proper experimental condition to do it at best.
In this path, a new determination process has been set-up at VKI. At first one can remark that the
rebuilding procedure, presented above (section 4.2), could function starting with any γ value, as well as the
heat-flux abacus computed by the BL code for the γ-determination. These two processing programmes
need only a reference catalycity. The first one rebuild an enthalpy (He) at the outer edge of the BL from
the reference measurement when the second one determines a catalycity from the test measurements. The
13 - 24 RTO-EN-AVT-142
idea of the Reference Determination Procedure is to operate with two identical water-cooled calorimeter
put in two different Danköhler probes. The first one in the standard probe (probe 1) is the reference
surface. The second one mounted in the frozen-probe (probe 2) is the test surface. The two probes are
placed successively in the same plasma jet. They imposed at their stagnation point two different velocity
gradients (β1 and β2 ) because of their geometry. The NDP of each one is computed by the ICP code
(NDP1 and NDP2) and both provide heat-flux and pressure measurement; (Qw1, ∆P) for probe 1 and
(Qw2, ∆P) for probe 2.
With all these experimental and numerical data one could run the He-rebuilding procedure and the γ–
determination one in a same loop. The enthalpy (He) is rebuilt with the reference measurement (probe 1)
assuming γ =1 as initial value. This enthalpy (He) is given to the BL code which computes a heat-flux
abacus for probe 2 with the same initial γ value (γ=1), since we consider identical calorimeters in the same
plasma jet. By placing Qw2 in the computed abacus one could check the validity of the initial guess for
the catalycity of the copper surface (γCu) and eventually determined a new value for γCu. This new value is
now taken as reference and re-injected in the rebuilding procedure for starting the same determination
loop. This iterative process stops when the computed heat-flux on the probe 2 is equal to the measured
one, as it is summarized in figure 25. In other term this procedure ends when the catalycity of the
reference surface is the same as the test surface. This final value will be the effective catalycity of the
reference surface, in this case the water cooled copper calorimeter (γCu). This new methodology is also
based on physico-chemical modelling but do not require the strong hypothesis of a fully-catalytic surface.
The assumption is that the surfaces have the same effective catalycity on their respective testing situation.
To be more precise the consistency of the methodology is also supported by the fact that the Prandtl
number remains near unity in our situation [75].
The reference determination procedure has been implemented and is currently running at VKI, the first
results indicate value like γCu = 0.01 but further investigations are carried out to complement and assess
these promising results [76].
Figure 25: Procedure for reference catalycity determination
RTO-EN-AVT-142 13 - 25
4.4 The 0-model

It has already been mentioned that innumerable microscopic phenomena are happening at the surface in a
reacting boundary layer. The effective catalycity (γ) that is defined in a previous section is only a global
coefficient that helps to characterize catalytic properties in the purpose of TPS design. The catalycity
property of a surface is implemented as a production term for the boundary condition of the BL problem to
solve. The formulation of this term comes under a proper modelling which is used for the data processing
of the experimental results. This reflects the fact that the methodology presented is a determination of
catalycity and not a direct measurement of it. The point here is not to list or discuss the large number of
models which can be found in the literature but rather to propose a simple view. As experimentalist in the
perspective of surface properties characterization for TPS design it is always an interest to adopt a global
approach for catalycity. Nevertheless the model used should be simple enough without being simplistic.
Considering one of the simplest models in use a first approach can be find, in the early work of Goulard,
where the gas-surface interaction is treated as an additional first order reaction [71]:
ωi = ki.ρi.Ci (21)
A second approach currently used is the one presented in this lecture which defines a recombination
probability (γ) for the flux impinging on the wall, given by (20).
Each of this approach is appealing by its simplicity but none of these formulation gives a wall chemical
composition that tends correctly to the local equilibrium composition for a fully catalytic wall as it has
often been remarked [19]. Some new formulations help to remove their inconsistencies as it is proposed by
Sarma [77] for the first approach and Rini for the second one [78]. But in one hand the first order reaction
imposed at the wall appears not really justified for high temperature surfaces and in the other hand
imposing γ = 1 is strictly speaking no physical since it should remain an asymptotic value.
If one wants to treat catalysis at the wall with a global approach, it can be simply consider that the
reactions are accelerated at the wall. Instead of looking at “gas-surface reactions” one could focus on “gas
reactions at the surface”. In this view, no specific model is needed for the reaction at the wall but the set of
equations for gas chemistry is conserved and their reaction rate are multiplied by a “booster-coefficient”
for the condition at the wall, to indicate that the reactions could happen faster at this location:
kfS = kf*10w kbS = kb*10w w : [0, ∞[ (22)
kfS, kbS are the forward and backward reaction rate for the gas chemistry at the surface and the catalytic
property of the surface is defined by:
w = 0 for non-catalytic surface.

w = ∞ for a fully catalytic surface.
We propose to call this simple model the “0-model” since no specific model are implemented for the wall
which is simply considered as a location where gas chemistry could happen faster. This approach has
several advantages:
The physical nature of the catalysis is preserved, the wall has a role of a catalyst which is not involved in
the reaction but helps them to happen. The wall chemical composition is no more a problem since the final
chemical state is ruled by the gas chemistry which is automatically coherent with the wall condition.
Finally the catalycity which is still defined as a probability of recombination is now an outcome of the BL
solution. The catalycity (γi) can be processed for each specie with the diffusion fluxes computed by the BL
solver. The same expression as before can be rearranged to give:
13 - 26 RTO-EN-AVT-142
Ji
γi = (23)
M i↓ ⋅ mi
γ = 1 remains an asymptotic situation and for w = 0 there is no catalycity and the wall behave chemically
as part of the gas at its surface temperature. The 0-model presents many benefits to deal with a simple and
coherent model. It is currently implemented on VKI code to test its efficiency for ground testing situation.
Further fine consideration can be made for its improvement but it is believed to be a efficient tool to deal
with the characterization of the TPS catalytic properties.
5 CONCLUSIONS
Experimental hypersonic stagnation point studies are at the confluence of many aspects of
aerothermochemistry. At this point the reproduction of the real flight conditions in a high enthalpy wind
tunnel requires a specific methodology, adapted to the ground testing facility. For a plasmatron facility the
LHTS methodology has been proved to be accurate and efficient. Nevertheless a great care has t be taken
to set-up the plasma facility and design the experiment environment because many disturbances can occur
to perturb the ideal stagnation point configuration. To ensure accurate flight duplication as well as to
support material testing that aim to determine fine properties of the TPS the quality of the measurements is
of first importance. To that purpose careful probe design together with extensive calibration campaign
must be carried out.
In these investigations experiments and computations are intimately linked on two levels. Firstly through a
rebuilding procedure which leads to the characterization of the flow parameters of the stagnation point,
and secondly because the measurements are interpreted with physico-chemical models implemented in a
proper boundary layer solver.
With all these tools we were able to treat the stagnation point as a local chemical reactor in which a
Damköhler number has been defined to scale the chemical environment produced in the ground testing
facility. Measurement probes have been designed to expand the testing capabilities of the facility in a wide
range of chemical non-equilibrium situation. Applications of these developments have been presented for
emissivity and catalycity determination of TPS. An innovative method for the determination of the
reference catalycity has been explained, it results in a definite improvement for the characterization of
TPS catalytic properties.
A new model for the surface catalycity has been proposed to keep a simple approach avoiding major
inconsistencies. It is called the “0-model”, since no specific models are considered for the wall, but
involves only gas chemistry which can be accelerated at the wall by a booster-coefficient for the chemical
rates.
Finally some questions should be raised. As a first one everybody could wonder which chemical model
has to be used for the hypersonic stagnation location. This study cannot reply directly since it is based on
existing models, at least a relative comparison could be done. But the experimental set-up in its stagnation
point configuration with its careful characterization methodology is a useful basis for laser based
diagnostics which could investigate such problems for these reacting flows.
More practically stagnation point as it is treated in these studies requires axisymetry and steadiness which
are situations not likely to happen easily in nature. At this point very diligent efforts must be done on the
experimental set-up to assess the testing conditions. Most of the attention should be focused on the ICP
torch design which represents the most sensitive part of the plasma wind tunnel. Very instructive
investigations are currently carried out at VKI on those problematic [32].
RTO-EN-AVT-142 13 - 27
At the end a more fundamental interrogation concerns the nature of the dissociated flow generated in a
ground testing facility. One could ask how the reacting flow present at the nose of an hypersonic vehicle
could be compared to the one produced by an electrical discharge on a ground based facility. Certainly
optical diagnostics should be involved for free stream characterization since LTE appears as a unique
reference which is important to check in the facilities. But flight experiments are also essentials to
consider since a lot still to be learnt concerning real hypersonic flights to address at best ground studies
from numeric and experimental sides.
6 ACKNOWLEDGEMENTS
The author would like to thank Sophie Herpin and Hyun Woo Krassilchikoff, VKI students, Cem Asma,
VKI research engineer and Pascal Collin, VKI Technician for their useful and efficient collaboration in the
VKI Plasmatron experimental studies.
7 REFERENCES
[1] AGARD, Aerothermodynamics of Hypersonic Vehicles, AGARD R-761, 1988.
[2] Dorrance WH, Viscous Hypersonic Flow, MacGraw-Hill, New York, 1962.
[3] Anderson, J.D. Jr., Hypersonic and high temperature gas dynamics, MacGraw-Hill Book
Compagny, New York, 1989.
[4] Fay, J.A., Riddell, F.R., Theory of stagnation point heat transfer in dissociated air, Journal of
Aeronautical Sciences, Vol. 25, No. 2, February 1958, p. 73-85.
[5] Neumann R.D., Experimental Methods for Hypersonics: Capabilities and Limitations, 2nd Joint
Europe-US Short Course on Hypersonic: GAMNI-SMAI and Uni. Of Texas at Austin, USAF
Academy, Colorado springs, CO 80840, January 1989.
[6] Ferri, A., Hypersonic flight testing, International Science and Technology, April 1964, pp 64-76.
[7] Winovich, W., Carlson, W.C.A., The 60-MW Shuttle interaction heating facility, 25th Int. Instr.
Symposium, Anaheim, May 1999.
[8] Caristia, S., De Fillipis, F., SCIROCCO Project a major plasma wind tunnel for the year 2000,
proceeding of International Symposium on Atmospheric Re-entry Vehicles and Systems, Arcachon
(F), 16-18 May 1999.
[9] Gordeev, A.N., Overview of characteristics and experiments in IPM plasmatrons, RTO-VKI Special
course. Measurement Techniques for High Enthalpy and Plasma Flows, October 1999
[10] Bottin, B., Carbonaro, M., Decré, M., Mazauric, S., Novelli, A. Design of a new inductively-coupled
plasma wind tunnel for reentry material testing at the von Kármán Institute, Wind tunnels and wind
tunnel test techniques, Cambridge (UK), 1997.
[11] Vincenti W.G., and Kruger, C.H., Introduction to Physical Gas Dynamics, John Wiley & Sons, 1965.
[12] Park, C. Nonequilibrium Hypersonic Aerothermodynamics. John Wiley & Sons, 1990.
13 - 28 RTO-EN-AVT-142
[13] Lees L., Laminar heat transfer over blunt-nosed bodies at hypersonic flight speed, Jet Propulsion,
Vol. 26, April 1956, pp. 259-269.
[14] Cohen, K, Reshotko, E., Similar solutions for the compressible laminar boundary layer with heat
transfer and pressure gradient, NACA TN 3325, 1955.
[15] Rose P.H., Stark W.I., Stagnation-point heat transfer measurements in dissociated air, J. Aeron. Sci.,
Vol. 25, No. 2, February 1958, pp. 86-97.
[16] Bleakney, W., Weimer, D.K., and Fletchner, C.H., The shock tube: A facility for investigations in
fluid dynamics, Review of Scientific Instruments, Vol. 20, pp. 807-815, November, 1949.
[17] Warren W.R. & Diaconis N.S., Air arc simulation of hypersonic environments, ARS Int. Hyp. Conf.
Cambridge, Massachusetts, Aug. 16-18, 1961.
[18] Kolesnikov, A. Extrapolation from high enthalpy tests to flight based on the concept of LHTS.
Belgium. RTO-VKI Special course. Measurement Techniques for High Enthalpy and Plasma Flows,
October 1999.
[19] Barbante, P.F. : Accurate and Efficient Modelling of High Temperature Nonequilibrium Air Flows.
PhD thesis, Université Libre de Bruxelles, May 2001.
[20] Bottin, B., Chazot, O., Carbonaro, M., Vander Haegen, V., Paris, S. : The VKI Plasmatron
Characteristics and Performances. Belgium. RTO-VKI Special Course. Measurement Techniques for
High Enthalpy and Plasma Flows, October 1999.
[21] Barbante, P.F., Chazot, O., Flight extrapolation of plasma wind tunnel of stagnation region flowfield,
J. Thermophys. Heat Transfer, (accepted for publication)
[22] Babat, G.I., Electrodeless discharges and some allied problems, J. Institution of Electrical Engineers,
part III, 94 (27), January 1947, pp 27-37
[23] Boulos M. I., The inductively coupled R.F.plasma, Pure & Appl. Chem., vol. 57, 1985, pp 1321-
1352
[24] Reed T. B., induction- coupled plasma torch, J. Appl. Phys., vol. 32, 1961, pp 821-824.
[25] Dunken H., Pforr G., Z. Chem., vol. 6, 1966
[26] Mermet J.M., Trassy C., A plasma torch configuration for inductively coupled plasma as a source in
optical emission spectroscopy and atomic absorbtion spectroscopy, Appl. Spectrosc. vol. 31, 1977,
pp 237-239
[27] Allemand C. D., Barnes R.M., A study of inductively coupled plasma torch configurations, Appl.
Spectrosc. vol. 31, 1977, pp 434-443
[28] Rezaaiyaan R., Hieftje G.M., Anderson H., Kaiser H., Meddings B., Design and construction of low-
flow, low-power torch for inductively coupled plasma spectrometry, Appl. Spectrosc. vol. 36, 1982,
pp 627-631
RTO-EN-AVT-142 13 - 29
[29] Chazot, O., Van Der Haegen,V., Carbonaro, M. Flow visualization with a cold model for ICP torch
design, Progress in Plasma Processing of Materials 1999, begell house, ,inc., New York,
Wallingford, u.k., pp 41-47.
[30] Chazot, O., Study of flow field topology in an ICP torch by surface pressure measurements and
numerical calculations, Progress in Plasma Processing of Materials 2001, begell house,inc., New
York, Wallingford, u.k., pp 241-248.
[31] Vanden Abeele D., and Degrez, G., Efficient computational model for inductive plasmas flows.
AIAA Journal, 38 (2), 2000, pp 234-242.
[32] Playez, M., VKI private communication
[33] Bottin, B., Carbonaro, M., Van Der Haegen, V., Paris, S. Predicted and measured capability of the
VKI 1.2 MW Plasmatron regarding re-entry simulation, 3rd European Symposium on
Aerothermodynamics for Space Vehicles, ESA SP-426, 1999, p.553-560.
[34] Bottin, B. Aerothermodynamic model of an inductively-coupled plasma wind tunnel. PhD Thesis,
Université de Liège, October 1999.
[35] Caimano D., Chazot O., Calibration of Heat Flux Measurement and Development of data
Acquisition System for Inductively Coupled Plasma Wind Tunnel, FLUCOME’03, 7th Triennial
International Symposium on Fluid Control, Measurement and Visualization, Sorrento, August 25-28,
2003.
[36] Caimano D., Chazot O, D.A.S and Heat Flux Measurements for Plasma Wind Tunnels, VKI
Technical Note, TN 202, December 2003.
[37] Sahin, P. Simultaneous pressure and heat-flux measurement in high-enthalpy facilities for aerospace
applications. 2002. VKI SR 2002-26.
[38] Bonnewijn, S. Combined measurements in Plasmatron facility for planetary entry simulation. VKI
SR. June 2003.
[39] Vera, I., Assessment of pressure and enthalpy rebuilding for simulation of planetary re-entry
conditions in a plasmatron facility, August 2004, VKI SR 2004-37.
[40] Vanden Abeele, D., An efficient computational model for inductively coupled air plasma flows
under thermal and chemical non-equilibrium. PhD Thesis, KU Leuven, November 2000.
[41] Magin, T., A model for inductive plasma wind tunnels, PhD Thesis, Université Libre de Bruxelles,
June 2004.
[42] Bottin, B., Vanden Abeele, D., Carbonaro, M., Degrez, G., and Sarma G.S.R., Thermodynamic and
transport properties for inductive plasma modelling. J. Thermophys. Heat Transfer, 13(3), July-
September 1999, pp 343-350.
[43] Zoby, E.V., Empirical stagnation-point heat-transfer relation in several gas mixtures at high enthalpy
levels, Langley Research Center, NASA TN D-4799, 1968.
[44] Marvin J.G., Deiwert G.S, Convective heat transfer in planetary gases, NASA TR R-224, 1965.
13 - 30 RTO-EN-AVT-142
[45] Sutton, K., Graves R. A., A general stagnation point convective-heating equation for arbitrary gas
mixtures, Langley Research Center, NASA TR R-376, 1971.
[46] Boison J.C. & Curtiss H.A., An Experimental Investigation of Blunt Body Stagnation Point Velocity
Gradient, ARS Journal, Vol. 29, Feb. 1959, pp 130-135.
[47] Chazot O., R. Régnier and A. Garcia Muñoz, Simulation Methodology in Plasmatron Facility and
Hypersonic Wind Tunnels, 12th International Conference on Method of Aerophysical Research,
Akademgorodok, Novosibirsk, Russia, June 28 – July 4, 2004.
[48] Pirard, B., Experimental simulation of planetary re-entry condition in the VKI Plasmatron,
September 2001, VKI SR 2001-29
[49] Regnier R., Rebuilding of re-entry conditions by combined measurements in a Plasmatron Facility,
VKI SR 16, 2003.
[50] Sagnier P., Vérant, J.-L., Flow characterization in the ONERA F4 high-enthalpy wind tunnel, AIAA
J., Vol. 36, No. 4, April 1998, pp. 522-531.
[51] De Fillipis F, Serpico M., Air high enthalpy stagnation point heat flux calculation, CIRA TN 96-014,
Jan. 1996.
[52] Horton T.E., Babineaux T.L., Influence of atmosphere composition on hypersonic stagnation-point
convective heating, AIAA J., Vol. 5, No. 1, January 1967, pp. 36-43.
[53] Rose P.H., Stankevics J.O., Stagnation-point heat transfer measurements in partially ionized air,
AIAA J., Vol. 1, No.12, December 1963, pp. 2752-2763.
[54] Yee L., Bailey H.E., Woodward H.T., Ballistic range measurements of stagnation-point hat transfer
in air and in carbon dioxide at velocities up to 18,000 feet per second, NASA TN D-777, 1961.
[55] Sagnier, P., Masson, A., Mohamed, A.K., Verant, J.L., and Devezeaux, D., Synthesis of mstp
calibration campaigns in onera F4 hot shot wind tunnel, ONERA TP, 86, 1995.
[56] Krassilchikoff, H, Weinquin, V., Etude des conditions de simulations de rentrees atmospherique pour
souffleries a haute enthalpie, VKI SR- 2005.
[57] Barbante P.F., Reacting flows simulation with application to ground to flight extrapolation. RTO-
VKI Special course. Experiments, modelling and simulation of gas-surface interactions for reactive
flows in hypersonic flights. February 2006.
[58] Pope R.B., Measurements of Enthalpy in Low-Density Arc-heated Flows, AIAA J., Vol. 6, No. 1,
Jan. 1968, pp 103-110.
[59] Fletcher, D.G., Nonintrusive diagnostic strategies for arcjet stream characterization, Belgium. RTO-
VKI Special course. Measurement Techniques for High Enthalpy and Plasma Flows, October 1999.
[60] Thömel, J., Local heat transfer simulation of air and CO2, July 2005, VKI PR 2005-01.
[61] Wicke, E. Gerhard Damköhler – Founder of chemical reaction engineering, International chemical
engineering, 25, n. 4, Oct. 1985, pp 770-773.
RTO-EN-AVT-142 13 - 31
[62] Baumgart, J., Design of an enthalpy probe, VKI SR 2002-12.
[63] Gokcen, T., Effects of flowfield nonequilibrium on convective heat transfer to blunt body, J.
Thermophys. Heat Transfer, vol. 10, n.2, April-June 1996, pp 234-241.
[64] Herpin, S., Chemical aspects of hypersonic stagnation point heat transfer, August 2005, VKI PR
2005-10.
[65] Sureda, M., del Campo, V., Determination of Damköhler number at stagnation point in experimental
simulation of re-entry conditions, August 2005, VKI SR 2005-36
[66] Park, C., Review of chemical-kinetics problems of future NASA Missions, I Earth Entries. J. of
Thermophysics and Heat Transfer 7, 3 (1993), pp 385-398.
[67] Gnoffo, P.A., Gupta, R.N., Shinn, J.L., Conservation equations and physical models for hypersonics
air flows in thermal and chemical nonequilibrium. Technical 2867, NASA (Langley), Virginia, 1989.
[68] Rosner, D.E., Recent advances in convective heat transfer with dissociation and atom recombination,
Jet Propulsion, July 1958, pp 445- 451.
[69] Deutschmann, O., Riedel, U., and Warnatz, J., Modeling of Nitrogen and Oxygen Recombination on
Partial Catalytic Surfaces, Journal of Heat Transfer, Vol. 117, 1995, pp. 495-501.
[70] Natsui, F., Barbato, M., and Bruno, C., Material-Dependent Catalytic Recombination Modeling for
Hypersonic Flows, Journal of Thermophysics and Heat Transfer, Vol. 10, 1996, pp. 131-136.
[71] Goulard, R., On catalytic recombination rates in hypersonic stagnation heat transfer. Jet Propulsion
28, November 1958, pp 737-745.
[72] Stewart, D.A., “Surface Catalysis and Characterization of Proposed Candidate TPS for Access-to-
Space Vehicles,” NASA TM-112206, July 1997.
[73] Neuman, R. D., Flight experiments for hypersonic vehicle development, RTO-AVT-VKI lectures
series 2006.
[74] Chazot O., S. Paris, P. Collin, M. Bickel, T. Ullman, TPS Testing and Catalycity Determination in
the VKI Plasmatron Facility, Atmospheric Reentry Vehicles and Systems Symposium, arcachon,
France, March 2003.
[75] Bottin, B., Pegase 4.4 Tables of thermodynamic and transport properties for air and argon plasmas,
December 1997, VKI IN 109.
[76] Krassilchikoff, H., VKI private communication.
[77] Sarma, G.S.R., Physico-Chemical Modelling in Hypersonic Flow simulation, Progress in Aerospace
Sciences, 36, 3-4 (2000), pp 281-349.
[78] Rini, P., Garcia, A, Magin, T., Degrez, G., Numerical simulation of nonequilibrium stagnation line
CO2 Flows with catalyzed surface reactions. J. of Thermophysics and Heat transfer vol. 18, n. 4,
October-december 2004, pp 511-518.
13 - 32 RTO-EN-AVT-142
Reacting flows simulation with applications to ground to flight extrapolation
P.F. Barbante
Researcher, Politecnico di Milano, Dept. of Mathematics
P.zza Leonardo da Vinci 32, 20133 Milano, Italy
barbante@mate.polimi.it
Abstract
The development of next generation reusable space vehicles requires a precise qualification
of their Thermal Protection System materials. The catalytic properties are usually determined
in plasma wind tunnels for sets of test conditions relevant to the planned flight mission program.
Therefore, for such a situation, it is important to have a methodology that allows for the correct
extrapolation of the ground test conditions to the real flight ones and vice-versa. The Local Heat
Transfer Simulation concept presented in this paper is a possible strategy to accomplish such
a task. The computational results show that the ground test conditions are indeed correctly ex-
trapolated to the flight ones and a simple method to account for possible discrepancies between
the two configurations is presented.
1 Introduction
Design and testing of Thermal Protection System (TPS) materials, that are used to protect space vehicles
from heat load during the (re-)entry phase, are a major issue for the definition of space missions [20, 17].
The determination of the catalytic properties of TPS materials is especially crucial for the design of an
optimal flight strategy: by a matter of fact the stagnation region heat flux for a fully catalytic wall can be
more than twice the heat flux for a noncatalytic one [1]. Such situation requires ground facilities able to
provide representative testing conditions for the evaluation of the material performances. It is clear that
TPS materials should be tested in real flight conditions, in order to safely rely on their catalytic properties.
However it is also well known that the complete real flight conditions cannot be exactly duplicated in a
ground test facility [19]. The usual strategy to overcome such a problem is to resort to some kind of partial
simulation: only some characteristics of the flight environment, that are of interest in the specific experiment,
are reproduced [21].
Plasma wind tunnels that can provide dissociated flows for large characteristic time have been widely
exploited, in particular for stagnation point testing configuration. Both supersonic arc-jet wind tunnels [22]
and subsonic plasma wind tunnels [16] (also known as Plasmatron facilities) have been demonstrated suitable
for TPS catalycity studies. The testing methodology in subsonic plasma flows has also been implemented
for the high enthalpy facilities at the von Karman Institute (VKI). This approach of catalycity determination
makes use of dedicated experimental procedures, combined with accurate CFD computations of the flowfield
inside the ground facility [9, 10, 7].
In this contribution we want to start from experimental test results pertinent to the subsonic test method-
ology and to extrapolate them to the corresponding real flight application. The results allow to verify the
hypothesis of the flight extrapolation methodology and to discuss detailed features of hypersonic flows.
2 Local Heat Transfer Simulation for Stagnation Point

One of the most critical parts of a (re-)entry vehicle is the stagnation point region, which is often subject
to the highest heat flux; it is therefore important that the testing conditions in the ground facility correctly
Barbante, P.F. (2007) Reacting Flows Simulation with Applications to Ground to Flight Extrapolation. In Experiment, Modeling
and Simulation of Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 14-1 – 14-20). Educational Notes
RTO-EN-AVT-142 14 - 1
Reacting Flows Simulation with
Applications to Ground to Flight Extrapolation
reproduce at least the real flight environment around the stagnation point of the aerospace vehicle. The
other way around, one has also to be able to correctly determine to which flight conditions the ground test
conditions correspond. When we talk about flight conditions we mean a set of freestream conditions, i.e.
pressure, temperature, Mach number, that will reproduce, in the boundary layer near the stagnation point of
the space vehicle, the same kind of environment found in the ground facility.
We consider the Fay and Riddell [11] and Goulard [13] formulae for the heat flux at the stagnation point
of a body immersed in a reacting flow.
The Fay and Riddell formula reads:

−0.6 0.4 0.1 0.5 α hD,e
qw = 0.76P r (ρe µe ) (ρw µw ) βe (He − hw ) 1 + (Le − 1) (1)
He
hD,e is the dissociation enthalpy (it is equal to N

P s 0 0
i=1 yi,e ∆hF,i , ∆hF,i being the enthalpy of formation of
species i and yi,e their mass fractions). The exponent α is equal to 0.52 for an equilibrium boundary layer
and to 0.63 for a frozen boundary layer with fully catalytic wall. βe is the velocity gradient at the boundary
layer outer edge.
The Goulard formula is valid for a frozen boundary layer with an arbitrarily catalytic wall and it reads:

hD,e ye
qw = 0.664P r−2/3 (βe ρe µe )0.5 He 1 + (Le2/3 φ − 1) (2)
He
ye is the atom mass fraction at the boundary layer edge and φ is a factor that takes into account the wall
catalycity.
These two very similar formulae state that the stagnation point heat flux is a function of velocity gradient,
flow enthalpy, density and chemical composition (which appears in the determination of the dissociation en-
thalpy hD,e ) at the boundary layer outer edge and of the wall enthalpy and catalycity. The ground simulation
is correct when the boundary layer conditions are the same in the wind tunnel and in flight: as a conse-
quence the heat flux is equal in the two cases if the wall conditions and the outer edge enthalpy, chemical
composition, density or pressure and velocity gradient are the same for wind tunnel and flight.
Under the somewhat restrictive assumption that the flow is near equilibrium, it follows that the heat flux
in the stagnation region is the same in flight and in the wind tunnel if the total enthalpy (He ), the pressure
(pe ) and the velocity gradient (βe ) at the outer edge of the boundary layer are the same:
Hef = Het pfe = pte βef = βet (3)
We assume the testing conditions are known and we determine the freestream flight conditions and the
flying body geometry that will ensure the needed values of He , pe and βe in the stagnation region. Such a
task is accomplished by means of the formulae:
1 f2
hf∞ + V∞ = Het (4)
2
f 2
pf∞ + ρf∞ V∞ = pte (5)
s
8 ρf∞ V∞f
= βet (6)
3 ρfe R∞f
Eqs. (4) and (5) are the conservation of energy and momentum for the monodimensional flow on the stag-
nation line respectively; Eq. (6) is an expression for the velocity gradient computed from the thin shock
f
layer theory [18] (where R∞ is the nose radius of the flying body). Although these equations provide only
an approximate description of the phenomena that are really happening on the stagnation line, they are still
accurate enough for the determination of the flight conditions, as it will be shown by the results. Eq. (4) can
14 - 2 RTO-EN-AVT-142
be simplified by taking into account that, at hypersonic flight speed, hf∞ 12 V∞ f 2

; similarly Eq. (5) is sim-
f f f 2
plified by means of the inequality p∞ ρ∞ V∞ . With these simplifications Eq. (4) gives the flight speed,
Eq. (5) the freestream density and therefore the altitude (for example using the US Standard Atmosphere
tables) and Eq. (6) the flying body nose radius, all the other quantities being known.
The Local Heat Transfer concept for stagnation region is validated with two series of computations each
one performed for different levels of wall catalycity. Ground facility boundary layer outer edge flow con-
ditions are taken from VKI plasma wind tunnel experiments. The first series of computations characterizes
the flowfield at the stagnation point of the TPS sample, the second series of computations defines the corre-
sponding real flight situation.
3 Governing Equations and Numerical Methods

We consider axisymmetric configurations and laminar flow. The fluid is made out of NS chemical species,
each one being a thermally perfect gas. Transport coefficients (viscosity µ, thermal conductivity λ, binary
diffusion coefficients Dij ) are computed with the Chapman-Enskog method [6]; thermodynamic properties
are computed by means of statistical mechanics [6]: in the calculations shown here, the rigid rotator, har-
monic oscillator model is chosen for the molecular species. An important point is that diffusion fluxes are
computed by means of the exact Stefan-Maxwell equations [2, 23] instead of the simplified and inconsis-
tent Fick’s law often found in literature. This choice is necessary if one wants to correctly compute the heat
flux [23, 2] and it is therefore mandatory for the correctness of the Local Heat Transfer Simulation technique.
The chemical forward reaction rates are computed from Arrhenius data fits given in Ref. [14]; the backward
reaction rates are computed from the forward ones and the equilibrium constant; the latter being computed
from statistical mechanics [6]. Wall chemical reactions are taken into account by means of a wall catalycity
model which allows a correct reproduction of the TPS material properties [4, 2].
As already mentioned, the first series of computations is performed in the stagnation region of the TPS
material sample; a boundary layer code developed by the author is used [4]. This choice is justified by the
fact that boundary layer equations are perfectly equivalent to Navier-Stokes ones in the stagnation point [24].
Once the Lees-Dorodnitsyn coordinate transformation is applied [4, 2] the boundary layer equations on the
stagnation line reads:
Continuity:
∂ Ṽ
+F =0 (7)
∂ η̂
Momentum:
1 ρe ve ∂βet

∂F 1 ρe ∂ ∂F
Ṽ = − F2 + + K2 l0 (8)
∂ η̂ 2 ρ 2 ρβet 2 ∂y ∂ η̂ ∂ η̂
Species continuity:
∂yi ∂Ji
Ṽ +K = ẇi (9)
∂ η̂ ∂ η̂
Energy:
Ns Ns
! !
∂g ∂ l0 ∂g ∂ l0 X ∂yi hi ∂ X hi
Ṽ = K2 − K2 −K Ji (10)
∂ η̂ ∂ η̂ P r ∂ η̂ ∂ η̂ Pr ∂ η̂ he ∂ η̂ he
i=1 i=1
Unlike standard stagnation line boundary layer equations written in Lees-Dorodnitsyn variables [2, 12] two
additional terms appear: K and the second term of the r.h.s. of the momentum equation. In the standard
Prandtl boundary layer theory the outer edge quantities are computed by extrapolating the external inviscid
solution to the wall. In the ground facility a low Mach, low Reynolds number jet impinges on the TPS
material sample and it is therefore more appropriate to match the boundary layer (which should be interpreted
as a zoom of the stagnation region) and the external flow at the real boundary layer outer edge. The two
RTO-EN-AVT-142 14 - 3
extra terms are due to such a matching procedure; K accounts for the real boundary layer thickness and
the second term on the r.h.s. of the momentum equation accounts for the fact that the quantity ∂ue /∂y
(computed at the boundary layer outer edge) has to be included in the expression of the boundary layer
pressure gradient [2]. The factor K is constant in the boundary layer and it is computed during the numerical
solution [2]. The second term on the r.h.s. of Eq. (8) is given by suitable processing of ground experiments
data. The procedure combines experimental results with numerical computations of the flowfield inside the
ground facility (interested readers are addressed to Refs. [9] and [10]).
The second series of computations, for the real flight condition, is performed by means of a Navier-Stokes
finite volume code for reacting flows (which was also developed by the author [3, 2]). The convective terms
are discretized with the Hybrid Upwind Splitting Riemann solver [8]; a MUSCL type reconstruction [15] is
applied to achieve second order accuracy: Van Albada’s limiter is used in the present work. The transport
terms are centrally discretized with second order accuracy.
We want to emphasize that the two codes make use of the same set of physico-chemical models, thus
allowing for a coherent and physically consistent comparison between the two series of computations; in
Ref. [4] it is shown that they provide results in excellent agreement when applied to the same configuration.
4 Flight Data Determination

The external flowfield conditions in the stagnation region of the TPS material sample are taken from data of
existing experiments [10, 9]. The first set of data, hereinafter called case (a), has been obtained from VKI
Minitorch (a small scale plasma wind tunnel) experiments [9]. The second set of data, hereinafter called case
(b), has been taken from experiments carried out in the VKI Plasmatron facility [5].
The following conditions are selected for case (a) (they have to be understood as conditions at the TPS
boundary layer outer edge):
MJ
Het = 9.8 pte = 10161 P a βet = 11670 s−1 (11)
Kg
Under the assumption of local equilibrium, they correspond to a temperature Tet = 4560 K and to a chemical
composition where molecular oxygen is almost fully dissociated and molecular nitrogen is just starting to
dissociate.
Extrapolation of these wind tunnel conditions to the real flight ones, by means of the procedure described
in Sec. (2), gives:
f m f f
V∞ = 4430 M∞ = 13.7 z∞ = 55.78 km
s
f
pf∞ = 38 P a T∞ = 258.6 K R∞f
= 17.72 cm (12)
f
(z∞ is the altitude, determined from the U.S. 1976 Standard Atmosphere). The computed speed value is
within upper 10% of the Space Shuttle speed at the same altitude [1]. For simplicity the shape of the (re-
f
)entry body has been taken to be a sphere with a radius equal to R∞ . The predicted body nose radius is too
small compared with the one of existing or planned (re-)entry vehicles. In this specific case the small value
depends from the diameter (10 mm) of the TPS sample that has been used in the experiments. In practice,
more realistic values are obtained by using samples of bigger diameter. In any event, for the purpose of the
present discussion, i.e. a demonstration of the validity of the Local Heat Transfer concept, the computed nose
radius is acceptable.
A more realistic (re-)entry vehicle nose radius is obtained for case (b), to whom the following conditions
apply:
MJ
Het = 22.8 pte = 12525 P a βet = 1502 s−1 (13)
Kg
14 - 4 RTO-EN-AVT-142
Under the assumption of local equilibrium, the boundary layer outer edge temperature is Tet = 6085 K,
molecular oxygen is fully dissociated and nearly 40% of molecular nitrogen is dissociated. The real flight
conditions are:
f m f f
V∞ = 6755 M∞ = 21.5 z∞ = 60.96 km
s
f
pf∞ = 19 P a T∞ = 244.4 K R∞ f
= 1.79 m (14)
The air mixture considered for both cases is a five component mixture; more precisely, molecular oxygen
O2 , molecular nitrogen N2 , nitric oxide N O, atomic oxygen O and atomic nitrogen N are used. Due to the
relatively low temperature there is no need to take into account ionized species.
As mentioned before, only some characteristics of the flight environment are reproduced in the ground
facility. It seems therefore interesting to check which nondimensional parameters are effectively reproduced.
We compare two classical nondimensional parameters, the Reynolds number and the Mach number. As
characteristic dimension for the flying body and the TPS sample we take the respective diameters. For case
(a) the jet speed in the ground facility is V∞ t = 149 m/s and the probe diameter is 10 mm. With these
values we have: Re = 49600, M = 13.7 for the flight and Ret = 62, M t = 0.1 for the ground facility.
f f
For case (b) the ground jet speed is V∞ t = 99 m/s and the probe diameter is 50 mm. The Reynolds and
Mach numbers are: Ref = 424000, M f = 21.5 for the flight and Ret = 123, M t = 0.06 for the ground.
Reynolds and Mach numbers are really different in flight and in the wind tunnel, this means that, in the
specific configuration, they are not the key parameters for the duplication of the stagnation point heat flux.
5 Flight Extrapolation Results

For the validation of the local heat transfer concept two series of computations have been performed for each
test case. The first series is performed along the stagnation line of the TPS material sample by means of the
boundary layer method. The second series, for the real flight condition, has been performed by means of the
Navier-Stokes finite volume technique.
Six different levels of wall catalytic activity have been taken into account. The wall reaction set is given
by: N + N → N2 and O + O → O2 , i.e. we take into account recombination of atomic nitrogen and of
atomic oxygen at the wall and we also assume that the wall has no effect on nitric oxide N O. For simplicity,
a single value of the catalytic recombination probability γ has been taken for the two reactions. The six
different computations correspond to the following values of γ:
γ = 0; 10−4 ; 10−3 ; 10−2 ; 10−1 ; 1

The wall temperature Tw has been taken uniformly equal to 1000 K both on the TPS sample and on the
equivalent (re-)entry geometry.
The boundary layer computations have been performed only in the stagnation point: 100 points in the
direction across the boundary layer have been taken for both case (a) and case (b). The Navier-Stokes
computations have been performed on the front part of the sphere and only half of it has been considered,
because of the symmetry with respect to the stagnation line. The grid has 3600 cells for case (a), 5680 for
case (b) and it is suitably refined in the boundary layer and in the proximity of the bow shock in order to
improve the computational accuracy.
We discuss now the results for case (a). In Fig. 1 the heat flux, computed for the different values of the
catalytic recombination probability, γ is shown (the heat flux value for γ = 0 is not shown because of the
logarithmic scale, but it is practically equal to the one for γ = 10−4 ). The difference in heat flux between
the Navier-Stokes and the boundary layer computations goes from 3% for the case of noncatalytic wall to
37% for the case of fully catalytic wall. Although the difference in heat flux for fully catalytic wall is high,
the difference in the case of noncatalytic or low catalytic (γ < 10−3 ) wall (which is the case of interest for
aerospace applications) is below 6%, i.e. in the range of the experimental uncertainties [16].
RTO-EN-AVT-142 14 - 5
Heat flux
1.5e+06
Flying body
TPS sample
1e+06
2
qw, W/m
5e+05 -4 -3 -2 -1 0
10 10 10 10 10
γ
Figure 1: Case (a): stagnation point heat flux for different catalycity levels
Temperature for noncatalytic wall

6000
5000
4000
T, K
3000
Flying Body
TPS Sample
2000
1000
0 0.001 0.002 0.003
Distance from the wall, m
Figure 2: Case (a): temperature profile for noncatalytic wall
The reason for the good agreement between the two computations in the cases of low or noncatalytic
wall can be explained by an examination of Figs. 2 and 3. The temperature profiles are similar in the
portion of the boundary layer with y < 0.001 m and the respective slopes at the wall are almost identical
(see Fig. 2); this leads to a good agreement on the conductive part (−λ∇T ) of the heat flux between the
two configurations. Looking at Fig. 3 one can notice
P sthat ~the boundary layer is practically frozen in both
computations; the diffusive part of the heat flux ( N
i=1 hi Ji ) is thus negligible in the two cases (and zero
for the noncatalytic case). The atomic oxygen mass fraction is higher for the TPS sample than for the
14 - 6 RTO-EN-AVT-142
Mass fractions for noncatalytic wall

0.25
0.2
mass fraction
0.15
O: Flying Body
0.1 O: TPS Sample
N: Flying Body
N: TPS Sample
0.05
0
0 0.001 0.002 0.003
Figure 3: Case (a): O and N mass fractions for noncatalytic wall
flying body, but the two profiles have an almost identical shape, differing only by a constant shift. The
same observations are valid for atomic nitrogen. This similar behaviour is not surprising, because Eq. (6)
implies the equality of the first Damköhler number in the stagnation region of both TPS sample and real
flight body. The inverse of velocity gradient is a time scale of the flow in the stagnation region and it is
the same for both configurations; the outer edge temperature and pressure being the same, also the chemical
characteristic time is the same and, therefore, the first Damköhler number. Another parameter that has to
be correctly reproduced is the second Damköhler number, which characterizes the heterogeneous chemistry-
diffusion coupling and therefore the interaction between the TPS material and the reacting gas. The second
Damköhler number is defined as: Da2 = kS l2 /D where ks is the inverse of a characteristic time of the wall
heterogeneous reactions, D a diffusion coefficient and l a characteristic length over which diffusion takes
place (as, for example, the boundary layer thickness). The surface material is the same and, therefore, ks
is the same too; in addition, because pressure, temperature and chemical composition are similar also D is
the same. Inspecting Figs. 2 and 3 we notice that the boundary layer thickness l is fairly close in the two
configurations: in effect stagnation point boundary layer thickness is a function of the square root of the
inverse of velocity gradient [24]. Therefore we deduce that also the second Damköhler number is acceptably
duplicated in the ground facility.
Temperature is higher and atomic species mass fractions are lower, at the boundary layer edge in the
flying body case than in the TPS sample case, because in the former configuration there is still a certain
degree of nonequilibrium in the shock layer, thus violating one of the assumptions on which the Local
Heat Transfer Simulation concept is based. However, the influence of the violation is negligible (at least
for heat flux determination) in the case of a low catalytic wall. A good reproduction of the behaviour of
atomic species, especially oxygen, in the ground facility is important to ensure that phenomena like aging
and oxidation of the TPS material are correctly simulated. In the specific case, the overestimation of atomic
oxygen mass fraction for the TPS sample in the test configuration gives a certain safety margin.
In Fig. 4 the temperature profiles for fully catalytic wall (γ = 1) are shown. The agreement is even better
than before and the conductive part of the heat flux is within 2% in the two configurations. Therefore, the
difference in total heat flux (see Fig. 1) is only due to the diffusive part. From Fig. 5 we notice that atomic
species are recombining at the wall because it is fully catalytic. The chemical composition in the boundary
RTO-EN-AVT-142 14 - 7
Temperature for fully catalytic wall

6000
5000
4000
T, K
3000
Flying Body
TPS Sample
2000
1000
0 0.001 0.002 0.003
Figure 4: Case (a): temperature profile for fully catalytic wall
Mass fractions for fully catalytic wall

0.25
0.2
mass fraction
0.15
O: Flying Body
0.1 O: TPS Sample
N: Flying Body
N: TPS Sample
0.05
0
0 0.001 0.002 0.003
Figure 5: Case (a): O and N mass fractions for fully catalytic wall
layer is dominated in both cases by the wall catalytic reactions O + O → O2 and N + N → N2 . Molecular
oxygen and nitrogen created at the wall diffuse toward the interior of the boundary layer, completely changing
the picture with respect to the noncatalytic case. We further notice that not only atomic oxygen mass fraction
is higher for the TPS sample, but also its slope at the wall. This implies that the diffusive heat flux is higher
in the TPS sample case than in the flying body one, thus explaining the observed differences in total heat flux
(the contribution of atomic nitrogen recombination to the heat flux is negligible).
Results for case (b) are discussed now. In Fig. 6 the stagnation point heat flux, computed for several
14 - 8 RTO-EN-AVT-142
values of catalytic recombination probability γ, is shown. (As in case (a) the heat flux value for γ = 0 is not
shown because of the logarithmic scale, but it is practically equal to the one for γ = 10−4 ). Computed heat
flux difference between Navier-Stokes and boundary layer goes from 2% for noncatalytic wall to 3% for fully
catalytic wall. We notice also that for noncatalytic wall heat flux is higher for boundary layer computations,
whilst for fully catalytic wall heat flux is higher for Navier-Stokes computations.
Heat flux
9e+05
8e+05 Flying body

TPS sample
7e+05
2
qw, W/m
6e+05
5e+05 -4 -3 -2 -1 0
10 10 10 10 10
γ
Figure 6: Case (b): stagnation point heat flux for different catalycity levels
In Fig. 7 temperature profiles for noncatalytic wall are shown; the agreement is excellent, in particular
close to the wall.

7000
6000
5000
T, K
4000
3000
Flying Body
TPS Sample
2000
1000
0 0.005 0.01 0.015
Figure 7: Case (b): temperature profile for noncatalytic wall
RTO-EN-AVT-142 14 - 9
In Fig. 8 atomic species profiles for noncatalytic wall are shown. Oxygen is fully dissociated at the
boundary layer outer edge and remains fully dissociated all along the boundary layer; agreement between
ground and flight profiles is excellent. Nitrogen is partially dissociated at boundary layer edge, the agreement
between TPS sample and flying body profiles is good, suggesting that the same physico-chemical phenomena
are happening in the two boundary layers. Compared with case (a) we notice a much improved agreement
in atomic species profiles. The main reason is the close matching of boundary layer outer edge chemical
composition between ground and flight. Chemistry is assumed to be in equilibrium at the TPS boundary
layer outer edge and for case (b) is close to equilibrium for flight too: the main reason is the fact that velocity
gradient, which is the inverse of the typical flow time in the stagnation region, is now six times smaller than
in case (a). We would also like to point out that, as in case (a), boundary layer thickness and first and second
Damköhler numbers are well matched for TPS sample and flying body.
0.3
0.25
mass fraction
0.2
O: Flying Body
0.15
O: TPS Sample
N: Flying Body
0.1 N: TPS Sample
0.05
0
0 0.005 0.01 0.015
Figure 8: Case (b): O and N mass fractions for noncatalytic wall
In Fig. 9 the temperature profiles for fully catalytic wall are shown. The two temperature profiles are
again very close to each other. One can notice that, at the same distance from the wall, temperature is higher
for fully catalytic wall than for noncatalytic one.
Atomic species profiles are shown in Fig. 10. Atomic oxygen profile is strongly affected by wall catalytic
recombination, because oxygen molecules created at the wall diffuse toward the interior of boundary layer.
Atomic nitrogen appreciably recombines in the bulk of boundary layer and its slope at the wall is consid-
erably shallow, although the wall is fully catalytic. Both phenomena, wall dominated recombination for O
and gas phase dominated recombination for N , are reproduced with good agreement between TPS sample
and flying body. The slope of atomic species profile at the wall is modestly larger in the real body boundary
layer, explaining the slightly higher flight heat flux.
From the previous discussion one could conclude that the Local Heat Transfer Simulation concept has
only a limited range of validity, i.e. when the conditions at the boundary layer outer edge are close to local
equilibrium. However both curves in Fig. 1 and in Fig. 6 are similar and show a strong dependence of the
heat flux from wall catalycity in the interval [10−3 − 10−1 ]; this suggests that the same physico-chemical
trends are reproduced in the ground facility and in flight. In addition, by looking at Fig. 11, we notice that
14 - 10 RTO-EN-AVT-142
Temperature for fully catalytic wall

7000
6000
5000
T, K
4000
3000
Flying Body
TPS Sample
2000
1000
0 0.005 0.01 0.015
Figure 9: Case (b): temperature profile for fully catalytic wall
Mass fractions for fully catalytic wall
0.3
0.25
mass fraction
0.2
O: Flying Body
0.15
O: TPS Sample
N: Flying Body
0.1 N: TPS Sample
0.05
0
0 0.005 0.01 0.015
Figure 10: Case (b): O and N mass fractions for fully catalytic wall
the normalized heat flux

NC
qw − qw
Q= F C − qN C
(15)
qw w
is very similar for ground and flight in both cases (a) and (b). The excellent agreement between normalized
heat fluxes is mainly due to the equality of Damköhler numbers for TPS sample and flying body. The flight
RTO-EN-AVT-142 14 - 11
heat flux for an arbitrary value of wall catalycity could therefore be deduced by means of the formula:

f
qw = Qt · qw f,F C f,N C
− qw f,N C
+ qw (16)
where Qt is computed by means of Eq. (15) with the heat flux values obtained for the TPS sample; these
values can also be experimentally measured ones. Eq. (16) is very accurate. Consider, for example, the heat
flux value for case (a) with γ = 10−1 : the TPS sample computation overpredicts the heat flux by 22% with
respect to the real flight one. If, on the opposite, Eq. (16) is used, the heat flux is underpredicted only by 1%,
a really dramatic improvement.
Normalized heat flux

1
0.8
)
NC
)/(qw -qw
0.6
FC
NC
0.4
(qw-qw
Flying body: case(a)

0.2 TPS sample: case (a)
Flying body: case (b)
TPS sample: case (b)
0 -4 -3 -2 -1 0
10 10 10 10 10
γ
Figure 11: Stagnation point normalized heat flux for different catalycity levels
The normalized heat fluxes presented above (Fig. 11) could be interpreted as the effect of boundary layer
chemical activity on the heat transfer. Since the heat flux is always increasing with the wall catalycity the
curves will be strictly monotonous. However a curve standing in the lower part of the graph indicates a
lower gas chemical activity, while a curve lying in the upper part reveals a higher chemical activity inside the
boundary layer. In effect, in the limit case of chemical equilibrium, the heat flux is the same for every possible
value of wall catalycity [11] and the normalized heat flux curve collapses into a straight line corresponding
to Q = 1. The relative position of the curves for case (a) and (b) could therefore be interpreted based on
these considerations. Case (a), for instance, denotes a situation where the gas phase chemical activity is
lower. This fact is due to the much higher velocity gradient for this case, which promotes a frozen flow at the
stagnation region by imposing a lower Damköhler number. This characteristic can be noticed in Fig. 11: the
curve for the case (a) appears below the curve for case (b). In addition the inspection of Figs. 3 and 8 clearly
points out that chemical activity in the boundary layer is higher in case (b) than in case (a).
6 Local Heat Transfer Simulation for a Flat Plate

The previous discussion has been on stagnation region flowfield; now we would like to present a preliminary
discussion of the Local Heat Transfer concept for a flat plate configuration. The main hypothesis we make is
that the outer edge boundary conditions are constant along the flat plate: it amounts practically to impose that
the boundary layer is everywhere self similar. Assuming also that the boundary layer is frozen we can write,
14 - 12 RTO-EN-AVT-142
by means of an approximate analytical solution of the boundary layer equations, a wall heat flux formula that
is quite similar to Eqs. 1 and 2:

−2/3 ρw µw
u 0.5
e ad 2/3 1 ye hD,e
qw = 0.332P r (He − hw ) 1 + (Le − 1) (17)
(ρe µe )0.5 x 1 + φw Head − hw
Where x is the distance from the flat plate leading edge and Head is the recovery enthalpy defined as: Head =
He + ru2e /2. The recovery factor r is mainly a function of the Prandtl number and is equal to zero for
P r = 1 and less than zero for P r < 1 (which is the case for air reacting mixtures). The factor φw takes
into account the wall catalycity. The formula states that the flat plate wall heat flux is a function of recovery
enthalpy, density, chemical composition at the boundary layer outer edge, of the term ue /x (that has the
dimensions of a velocity gradient) and of the wall enthalpy and catalycity. Therefore the heat flux is equal in
flight and in the wind tunnel if the previous quantities are the same in both cases. The equality of recovery
enthalpy, density and chemical composition implies the equality of outer edge velocity and also of the flat
plate dimensions (because ue /x has to be the same). It seems that it is possible to have the same heat flux
only with perfect duplication of real flight conditions and of body dimensions in the wind tunnel. However,
if we can tolerate a discrepancy in the heat flux, it is possible to relax the condition on recovery enthalpy and
to impose the equality of static enthalpy. The heat flux would be approximately the same in flight and in the
wind tunnel if (the assumption that the boundary layer is close to chemical equilibrium at the outer edge is
made):
ufe ute
hfe = hte pfe = pte = (18)
xf xt
It is clear that, imposing the equality of static enthalpy and not of recovery enthalpy, we cannot obtain an
exact duplication of heat flux as in the stagnation point, but only an approximate one; however it is the only
way to avoid a one to one duplication of the flight environment in the wind tunnel.
Heat flux for noncatalytic wall

Flat plate
2e+06
1.5e+06
Me = 0.1
1e+06
Me = 0.3
2
qw, W/m
Me = 0.9
Me = 2
5e+05
0
0 0.1 0.2 0.3 0.4 0.5
x*(ue,M =0.1/ue), m
e
Figure 12: Flat plate heat flux for noncatalytic wall
We present now a few computations that are meant to provide a preliminary assessment of the validity
of our assumptions. The boundary layer along a flat plate is computed; the outer edge boundary conditions
are as follows: pe = 12525 P a, Te = 6000 K, Me = 0.1, 0.3, 0.9, 2. The outer edge chemical composition
RTO-EN-AVT-142 14 - 13
is considered to be the equilibrium one. The wall is assumed to be at radiative equilibrium and the wall
emissivity coefficient is set to 0.85. The reacting mixture is five species air, the wall catalytic reactions are
N + N → N2 and O + O → O2 and the computations are made for two levels of catalytic recombination
probability, γ = 0 and γ = 1.
In Figs. 12 and 13 the wall heat flux for noncatalytic and fully catalytic wall respectively is shown for
all Mach numbers. The heat flux is expressed in function of the normalized coordinate x(ue,Me =0.1 /ue ) that
should be the same for all Mach numbers because of Eq. 18.
Heat flux for fully catalytic wall

Flat plate
4e+06
3e+06
Me = 0.1
2e+06
Me = 0.3
2
qw, W/m
Me = 0.9
Me = 2
1e+06
0
0 0.1 0.2 0.3 0.4 0.5
x*(ue,M =0.1/ue), m
e
Figure 13: Flat plate heat flux for fully catalytic wall
Supposing we want to simulate the real flight flow with a Mach 0.1 wind tunnel flow, the agreement in
heat flux (except for a small region near the flat plate tip, where the classical boundary layer theory is not
adequate) ranges from excellent for Me = 0.1, 0.3, 0.9 to acceptable for Me = 2. The maximum heat flux
difference is as follows: for a noncatalytic wall it is around 21 % between Me = 0.1 and Me = 2, for a fully
catalytic wall it is around 18 − 19 % for the same Mach number range. The heat flux difference is easily
explained: it is due to the viscous dissipation that naturally increases with Mach number.
Although the outer edge Mach number is different, the structure of the boundary layer is similar for equal
values of the normalized coordinate x(ue,Me =0.1 /ue ) for all cases. Temperature profile and atomic oxygen
and atomic nitrogen profiles for noncatalytic wall are shown at the station x(ue,Me =0.1 /ue ) = 0.5 m in
Figs. 14 and 15 respectively. Only the profiles for Me = 0.1 and Me = 2 are shown, the ones for the remain-
ing Mach numbers fall in between the two. The temperature profile agreement is good; the temperature for
Me = 2 is slightly higher and has a stronger slope at the wall, as it should be because of the heat generated by
viscous dissipation inside the boundary layer. Atomic oxygen and atomic nitrogen profiles match quite well
because the first Damköhler number at every x/ue station is the same for all boundary layers. We also notice
that the boundary layer thickness is equal and this is important because, as already discussed in Sec. 5, it
ensures that the second Damköhler number is the same and therefore the heterogeneous chemistry-diffusion
coupling is well reproduced.
14 - 14 RTO-EN-AVT-142

Flat plate
6000
5000
4000
T, K
3000
Me = 0.1
2000
Me = 2
1000
0 0.02 0.04 0.06 0.08

Figure 14: Flat plate temperature at x(ue,Me =0.1 /ue ) = 0.5 for noncatalytic wall

Flat plate
0.25
0.2
mass fraction
0.15
0.1 O: Me = 0.1
O: Me = 2
0.05 N: Me = 0.1
N: Me = 2
0
0 0.02 0.04 0.06 0.08
Figure 15: Flat plate O and N mass fraction at x(ue,Me =0.1 /ue ) = 0.5 for noncatalytic wall
7 Conclusions
The Local Heat Transfer Simulation concept for stagnation point configurations has been presented with its
application to plasma wind tunnel experiments. Two examples allowed to illustrate the different aspects of
the methodology and demonstrated that it correctly extrapolates ground test conditions to real flight ones. The
equality of stagnation enthalpy, pressure and velocity gradient, plus the fact that the flow is near equilibrium
in the stagnation region, are necessary for a correct reproduction of the flight thermochemical environment
RTO-EN-AVT-142 14 - 15
inside the ground facility. Deviations from the condition of local equilibrium in the stagnation region of the
real body are mainly responsible for the discrepancies in heat flux determination between the ground facility
and the flight. However, a normalized heat flux formulation can be used to account for these inconsistencies
and to obtain an improved heat flux prediction.
We have also shown a preliminary version of an approximate Local Heat Transfer concept for a flat plate
configuration. The heat flux is not duplicated as accurately as in the stagnation point configuration, however
first and second Damköhler numbers are well duplicated and therefore also the chemical composition in the
boundary layer and the wall heterogeneous chemical reactions phenomena.
14 - 16 RTO-EN-AVT-142
References
[1] A NDERSON , J. D. Hypersonic and High Temperature Gas Dynamics. McGraw-Hill, New York, 1989.
[2] BARBANTE , P. F. Accurate and Efficient Modeling of High Temperature Nonequilibrium Air Flows.
PhD thesis, Université Libre de Bruxelles-von Karman Institute, Bruxelles, Belgium, 2001.
[3] BARBANTE , P. F., AND D EGREZ , G. An Efficient Euler/Navier-Stokes Solver for Reacting Flows. In
16th IMACS World Congress 2000 (21-25 August 2000).
[4] BARBANTE , P. F., D EGREZ , G., AND S ARMA , G. S. R. Computation of Nonequilibirum High-
Temperature Axisymmetric Boundary-Layer Flows. Journal of Thermophysics and Heat Transfer 16,
4 (2002), 490–497.
[5] B OTTIN , B., C HAZOT, O., C ARBONARO , M., VAN D ER H AEGEN , V., AND PARIS , S. The VKI
Plasmatron Characteristics and Performance. In Measurement Techniques for High Temperature and
Plasma Flows, J. M. Charbonnier and G. S. R. Sarma, Eds. NATO-RTO, 1999.
[6] B OTTIN , B., VANDEN A BEELE , D., C ARBONARO , M., D EGREZ , G., AND S ARMA , G. S. R. Ther-
modynamic and Transport Properties for Inductive Plasma Modeling. Journal of Thermophysics and
Heat Transfer 13, 3 (1999), 343–350.
[7] C HAZOT, O., PARIS , S., C OLLIN , P., B ICKEL , M., AND U LLMAN , T. Tps Testing and Catalycity De-
termination in the vki Plasmatron Facility. In 3rd Atmospheric Reentry Vehicles and System Symposium
(24-27 March 2003).
[8] C OQUEL , F., AND L IOU , M. S. Hybrid Upwind Splitting Scheme by a Field by Field Decomposition.
TM 106843, NASA, Jan. 1995.
[9] DE LA L LAVE P LATA , M. Analysis and Application of a Methodology for the Determination of TPS
Materials Catalycity. Diploma Project Report 2000-4, von Karman Institute, June 2000.
[10] D EGREZ , G., BARBANTE , P. F., DE LA L LAVE P LATA , M., M AGIN , T., AND C HAZOT, O. De-
termination of the Catalytic Properties of TPS Materials in the VKI ICP Facilities. In ECCOMAS
Computational Fluid Dynamics Conference. Swansea UK (September 2001).
[11] FAY, J. A., AND R IDDELL , F. R. Theory of Stagnation Point Heat Transfer in Dissociating Air. Journal
of the Aeronautical Sciences 25, 2 (1958), 73–85.
[12] F LETCHER , C. A. J. Computational Techniques for Fluid Dynamics, 2nd ed., vol. 2. Springer-Verlag,
Heidelberg, 1991, ch. 15.
[13] G OULARD , R. On Catalytic Recombination Rates in Hypersonic Stagnation Heat Transfer. Jet Propul-
sion 28 (1958), 737–745.
[14] G UPTA , R. N., Y OS , J. M., T HOMPSON , R. A., AND L EE , K. P. A Review of Reaction Rates
and Thermodynamic and Transport Properties for an 11-Species Air Model for Chemical and Thermal
Nonequilibrium Calculations to 30000 K. RP 1232, NASA, Aug. 1990.
[15] H IRSCH , C. Numerical Computation of Internal and External Flows, vol. 2. John Wiley & Sons,
Chichester UK, 1990.
[16] KOLESNIKOV, A. F. Combined Measurements and Computations of High Enthalpy and Plasma Flows
for Determination of TPM Surface Catalycity. In Measurement Techniques for High Temperature and
Plasma Flows, J. M. Charbonnier and G. S. R. Sarma, Eds. NATO-RTO, 1999.
RTO-EN-AVT-142 14 - 17
[17] KOLODZIEJ , P. Strategy and Approach to tps Design. In Critical Technologies for Hypersonic Vehicle
Development, D. Gaitonde and D. Fletcher, Eds. NATO-RTO, 2004.
[18] L UNEV, V. V. Hypersonic Aerodynamics. Maschinostroenie, Moscow, 1975. (in Russian).
[19] N EUMANN , R. Experimental Methods for Hypersonics: Capabilities and Limitations. In The second
joint Europe/US Short Course in Hypersonics (Jan. 1989), US Air Force Academy Colorado Springs,
CO 80840.
[20] P RABHU , D. System Design Constraints - Trajectory Aerothermal Environments. In Critical Tech-
nologies for Hypersonic Vehicle Development, D. Gaitonde and D. Fletcher, Eds. NATO-RTO, 2004.
[21] ROSE , P., AND S TARK , W. Stagnation Point Heat Transfer Measurements in Dissociating Air. Journal
of the Aeronautical Sciences 25, 2 (1958), 86–97.
[22] S TEWART, D., C HEN , Y., BAMFORD , D., AND ROMANOVSKY, A. Predicting Material Surface Cat-
alytic Efficiency Using Arc-jet Tests. AIAA Paper 95-2013.
[23] S UTTON , K., AND G NOFFO , P. Multi-component Diffusion with Application to Computational
Aerothermodynamics. In 7th AIAA/ASME Joint Thermophysics and Heat Transfer Conference (15-
18 June 1998), AIAA Paper 98-2575.
[24] W HITE , F. M. Viscous Fluid Flows, 2nd ed. McGraw-Hill, New York, 1991, ch. 3.
14 - 18 RTO-EN-AVT-142
Appendix 1
Nomenclature
Dij = binary diffusion coefficient between species
i and j, m2 /s
F = u/ue , nondimensional tangential velocity
g = h/he , nondimensional mixture enthalpy
h = mixture static enthalpy, J/kg
H = mixture total enthalpy, J/kg
H ad = mixture recovery enthalpy, J/kg
hi = enthalpy of species i, J/kg
Le = Lewis number
Ji = nondimensional mass diffusion flux of species i
l0 = ρµ/ρe µe , Chapman-Rubesin parameter
Ns = number of species in the mixture
p = pressure, Pa
Pr = Prandtl number
q = heat flux, W/m2
Q = normalized heat flux
R = body nose radius, m
T = mixture temperature, K
u, v = tangential and normal velocity components, m/s
V = velocity, m/s
Ṽ = nondimensional transformed normal velocity
ẇi = nondimensional mass production rate of species i
x, y = Cartesian coordinates, m
yi = mass fraction of species i
β = ∂u/∂x, velocity gradient, s−1
γ = wall catalytic recombination probability
η̂ = transformed y coordinate
λ = mixture thermal conductivity, W/(m K)
µ = mixture viscosity, m2 /s
ρ = mixture density, kg/m3
Subscripts
w = wall surface
e = boundary layer outer edge
∞ = free stream
Superscripts
f = flight
FC = fully catalytic
NC = noncatalytic
t = ground facility
RTO-EN-AVT-142 14 - 19
14 - 20 RTO-EN-AVT-142
Determination of effective recombination probability: detailed
aspects of a macroscopic methodology
P. Rini∗and G. Degrez†
von Karman Institute for Fluid Dynamics, 1640, Rhode-St.-Genèse, Belgium
Université Libre de Bruxelles, 1050 Bruxelles, Belgium
1 Introduction
The purpose of the present Lecture Series is to provide an up-to-date review of the experimental techniques,
the theoretical models, as well as the numerical simulation strategies involved in the treatment of the chemical
character of high temperatures gases. To this end, aspects of both materials and high temperatures uid sciences
are discussed together, with the aim of enhancing the improvements in the understanding of the processes of
heat release on solid surfaces close to reactive gases taking place in re-usable hypersonic ight vehicles, bal-
listic missiles, and rockets exhaust nozzles. In addition, further details are given concerning the modeling and
the experimental investigation of gas-surface interactions. This latter aspect is strongly related to the diffusive
component of the heat ux experienced by a vehicle entering the atmosphere. In particular, the way in which
the Thermal Protection Material (TPM) in uences the gas particle recombination on its surface has a tremen-
dous effect on the wall heat ux. As will be shown later on, the heat ux measured in the stagnation point of
a probe held in a plasma wind tunnel can simply double when materials enhancing complete recombination
are used instead of inert material. Therefore, the role played by the modelling of gas/surface interactions is of
fundamental importance in the understanding of heat-transfer and as a consequence in the design of Thermal
Protection Systems (TPS).
When analyzing the interaction between a gas and a non ablating surface, at least two approaches can be put in
evidence. A microscopic and a macroscopic approach.
Recent years have seen a very important progress in the microscopic theoretical treatment of surfaces and pro-
cesses on surfaces described by the so called theoretical surface science [31]. The aim of theoretical surface
science is to contribute signi cantly to the fundamental understanding of the underlying principles governing
the geometric and electronic structure of surfaces, together with the processes taking place on these surfaces
such as growth of surface layers, gas-surface scattering, friction or reaction at surfaces [30]. In the framework
of atmospheric (re)entry, theoretical surface science could provide a very attractive tool for the improvement
of the current techniques used to estimate the catalytic activity of thermal protection materials. Some research
is ongoing in this direction and published data are already available providing values of recombination prob-
abilities of oxygen on Silica based materials computed using Molecular Dynamics techniques [54]. The main
advantage of this microscopic approach is that a variety of surface properties can be described from rst prin-
ciples, i.e. without invoking any empirical parameters. On the other hand, this approach, based on expensive
computations, has not yet been used to investigate complicated o w conditions like those typical of high en-
thalpy facilities. It will certainly be worth to investigate this possibility in the future.
In the eld of TPS design, a macroscopic approach is more widespread, which is closely coupled with the
use and the simulation of high enthalpy plasma o ws produced in suitable facilities. In this manuscript, our
attention is focused on this second way of analyzing gas-surface interactions. In particular, the methodology
∗
e-mail : rini@vki.ac.be
†
e-mail : gdegrez@ulb.ac.be
Rini, P.; Degrez, G. (2007) Determination of Effective Recombination Probability: Detailed Aspects of a Macroscopic Methodology.
In Experiment, Modeling and Simulation of Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 15-1 – 15-46).
Educational Notes RTO-EN-AVT-142, Paper 15. Neuilly-sur-Seine, France: RTO. Available from: http://www.rto.nato.int/abstracts.asp.
RTO-EN-AVT-142 15 - 1
Determination of Effective Recombination
Probability: Detailed Aspects of a Macroscopic Methodology
developed at the Institute for Problems in Mechanics (IPM) has been selected as a suitable instrument for the
determination of the catalytic activity of TPM.
1.1 IPM Methodology to determine TPS materials catalytic activity
Within the framework of the IPM methodology, the determination of TPS materials catalytic properties rely
on experimental testing on TPS materials samples in high enthalpy facilities producing o w conditions close
to ight conditions. Inductively Coupled Plasma (ICP) wind tunnels are particularly suited for this purpose
because they produce plasma jets of very high chemical purity, contrary to arc heated wind tunnels in which
the o w is polluted by metallic impurities originating from electrode erosion, which can contaminate the TPS
material sample. This was the motivation for the construction of two ICP facilities at VKI : a large scale ICP
facility (Plasmatron) [6] and a small scale facility (Minitorch) [13].
Kolesnikov et al. [39, 34] proposed a methodology to estimate TPS materials catalytic activity. This methodol-
ogy requires heat transfer and pitot pressure measurements on one hand, and numerical simulation of the o w
in the testing facility and along the stagnation line in front of the heat transfer probe on the other hand. In the
following, we present the main aspects of this methodology from the analysis of which the main motivations of
this contribution were inspired.
The IPM methodology relies on a combination of heat transfer and pitot pressure measurements in a subsonic
plasma jet produced in an ICP facility and numerical simulations of the o w. In this sense it represents a hybrid
methodology which indeed builds on both experiments [15, 27, 40, 12, 2] and numerical simulation of high
enthalpy o ws [75, 43]. Actually, the o w conditions in the ICP facility are related to ight conditions using
the Local Heat Transfer Simulation (LHTS) concept [35, 3] so that the measured heat ux es are equivalent to
the actual stagnation point heat ux es of a certain space vehicle at a point (altitude/velocity) of its (re)entry
trajectory. A detailed description of the IPM methodology is presented in Ref. [16] and hereafter we present
only its main characteristics.
This methodology, which is schematically represented in Fig. 1, is composed of the following essential building
blocks:
1. An LTE with constant elemental fraction viscous o w simulation of the o w in the ICP facility heating
chamber and in the downstream subsonic plasma jet. The purpose of this LTE viscous o w computation
is to compute the values of the non-dimensional parameters (Πi ) which characterize the o w near the
stagnation point of the heat ux probe in the plasma jet.
2. Experimental measurements of the stagnation point wall heat ux (qw ) and of the pitot pressure (p) on a
cold wall (T w ∼ 300K) reference heat ux probe assumed to be fully catalytic1 .
3. Reconstruction of the plasma jet enthalpy and velocity gradient at the edge of the stagnation line using
the measured stagnation point heat ux and pitot pressure.
4. Calculation of heat ux abacus (optional). Once the stagnation line edge conditions have been deter-
mined, the stagnation line solver can be run for various wall catalytic activity/temperature combinations
to produce a heat ux abacus (set of qw , T w curves at constant catalytic activity γw ).
1
For air o ws the material used is copper while for CO2 mixtures we use Silver.
15 - 2 RTO-EN-AVT-142
LTE viscous flow simulation Experiments
Set of non−dimensional Pitot pressure ( ∆p)

parameters Heat flux (qw)
Outer edge Πi
Reconstruction of the outer edge conditions
Enthalpy (hδ ) Velocity gradient (dv/dr δ )
Non equilibrium stagnation line computation
qw (h , dv/dr δ , pδ ,Π1 , Π2,γw,Tw) = qw exp

SL δ
Outer edge
elemental Hydrodynamics
composition
2
Π3 ρ δ (h δ ,p δ ) (Rm , dv/dr δ ) = ∆p exp
0.7
γw=1 Measurement on a
fully catalytic
γw=0.1
0.6 cold wall material
γw=0.01 (γw =1)
qw [MW/m ]
2
0.5
0.4
γw=0.001 Test measurement on a TPM sample
(Tw,qw)
Computation of the wall heat flux for various 0.3
catalytic efficiency ( γw) / wall temperatures
γw=0
using a non equilibrium stagnation line solver 0.2 γw
500 1000
Tw [K]
1500 2000 Identification of
Figure 1: Schematic representation of the IPM methodology for the determination of the catalytic activity of
TPS materials
5. Experimental measurement of the stagnation point heat ux to a TPS material sample and of its temper-
ature, and determination of the material catalytic activity2 .
The previous list shows clearly the highly hybrid character of the methodology. A particular attention should
be devoted to the third step, usually referred as the rebuilding of the outer edge conditions. The logic on which
this process is based follows from the investigation of the parameters from which the stagnation point heat ux
depends. Such investigation reveals that these parameters are as follows :
• the thermochemical state of the stagnation line outer edge. In the original version of the methodology
the outer edge is supposed to be under thermochemical equilibrium conditions de ned by pδ , hδ , and the
elemental fraction corresponding to the ICP torch inlet.
• The radial velocity gradient [(∂v/∂r)δ ] at the stagnation line outer edge.
• The stagnation line nite thickness (δ) and the product of the normal (i.e. axial) velocity by the normal
derivative of the radial velocity gradient at the stagnation line edge [uδ ∂/∂x(∂v/∂r)δ ].
• The wall temperature (T w ) and the catalytic activity of the sample surface with respect to each of the
kth possible recombination processes (γwk ). One of the major assumption of the original version of the
2
This can be done graphically using the heat ux abacus by identifying the γw contour on which the data point (qw , T w ) lies, or
alternatively numerically.
RTO-EN-AVT-142 15 - 3
methodology consists in assuming all these recombination probabilities to be equal to the same effective
probability γw .
Among these parameters, only two are unknown : the outer edge enthalpy (hδ ) and the radial velocity gra-
dient [(∂v/∂r)δ ]. Indeed, the stagnation pressure (pδ ) and the wall temperature (T w ) are known experimentally,
the recombination probability (γwk = γw ∀k) is equal to one according to the assumption of a fully catalytic heat
ux probe. In addition, the stagnation line thickness (δ) and uδ ∂/∂x(∂v/∂r)δ are related to the radial veloc-
ity gradient itself and to the probe radius Rm through the non-dimensional hydrodynamic parameters obtained
from the simulation of the plasma o wing in the ICP facility performed assuming thermochemical equilibrium
conditions with constant elemental fraction to be established both in the torch and in the test chamber. The rst
two non dimensional parameters3 read as follows :
!2
δ ∂ ∂v ∂v
!
Π1 = Π2 = uδ / , (1)
Rm ∂x ∂r δ ∂r
so that the stagnation point heat ux can be determined from the solution of the stagnation line problem (SL)
which implicitly reads
∂v
!
qw (hδ , , pδ , Π1 , Π2 , γw , T w ). (2)
∂r δ
| {z }
unknown
The additional equation needed to retrieve the two unknowns (hδ , (∂v/∂r)δ ) is given from the knowledge of the
measured pitot pressure. This measurement is related to ρδ and (∂v/∂r)δ through a third non-dimensional hy-
drodynamic parameter obtained from the LTE viscous o w simulation carried out assuming constant elemental
fraction, i. e.
∆p
Π3 = . (3)
ρδ (hδ , pδ )[Rm (∂v/∂r)2δ ]
The two unknown o w quantities are then determined by solving the system (2)-(3) by some suitable itera-
tive scheme, relying on the fact the functional form (2) is provided implicitly by running a non-equilibrium
stagnation line solver4 .
In this paper we wish to investigate and solve some of the issues previously described by merging and
exploring some of the ideas scattered in the literature, and at the same time present and recommend new
concepts. The outcome of this process consists in the proposition of a modi cation to the IPM methodology for
the determination of the catalytic properties of TPS materials, based on the investigation of its main weak points.
The determination of these weak points follows directly from the description presented in the previous section.
Indeed, we notice that several conditions must be satis ed in order for the methodology to be successful :
3
For the computation of the non-dimensional parameters it is necessary to de ne the location of the stagnation line outer edge. This
position is selected to be coincident with the in ection point of the axial velocity pro le along the stagnation line obtained from the
inductively coupled plasma o w simulation. Further details about the theoretical background on which this choice is based are given
in Ref. [37], while the low sensitivity of the Πi to the choice of δ has been veri ed in Ref. [71]
4
A slightly different approach, involving an additional dependent non-dimensional parameter, is proposed in the original version of
the IPM methodology [16] and implemented in the VKI-Rebuilding code [28].
15 - 4 RTO-EN-AVT-142
• the plasma o wing in the ICP facility should be under thermochemical equilibrium and elemental diffu-
sion should not affect the o w solution.
• The stagnation line o w should be under chemical non equilibrium to make the wall heat ux sensitive
to variations in the surface catalytic activity.
• The recombination probability of all heterogeneous recombination processes should be the same.
The investigation of the validity of some of the assumptions/conditions previously listed represented the main
motivation and objective of the present study and lead us to the following main contributions of our work.
• In the past, not sufficient credit has been given to the phenomenon of elemental diffusion within the
atmospheric (re)entry related literature. Indeed, in a wide variety of applications, this phenomenon has
been neglected. In particular, this is the case in the IPM methodology. The investigation of this aspect of
high temperature o w modeling leads us to the derivation of an explicitly closed form of the governing
equations of mixtures of reacting gases under LTE presented in Sec. 2. This was possible reducing
the equations of chemically reacting o ws to an elegant system consisting of the conventional Navier-
Stokes equations (mass, momentum, energy) complemented by an advection-diffusion equation for the
mass fraction of each chemical element in the mixture. The obtained formalism is in closed form in the
sense that diffusive ux es are directly expressed in terms of gradients of the solution unknowns, unlike
other formulations in which these ux es are obtained in an implicit manner, by solving the full system
of Stefan-Maxwell equations. The main contribution of this formalism consists in the introduction of
several new LTE transport coefficients.
• In Sec. 3 we present an application of the derived theory to the description of CO2 /N2 LTE mixtures
o wing along a stagnation line. This leads to the original computation of multicomponent elemental
diffusion coefficients for mixtures containing more than two elements.
• In Sec. 4 we investigate the effects of elemental diffusion on the behavior of air and CO2 plasmas o wing
within the VKI-Plasmatron in a certain range of operating conditions. In addition we provide a charac-
terization of hot air jet o wing into the test chamber of the VKI-Plasmatron facility. This study suggest
a modi cation of the original version of the IPM methodology based on the observation that, for suf-
ciently high operating pressures, chemical equilibrium conditions with variable elemental fraction are
established within the torch and the test chamber. As a consequence, the stagnation line outer edge ele-
mental fractions should be added to the set of non-dimensional parameters (Πi ) needed in the rebuilding
process.
• In Sec. 5 an alternative model for the phenomenological description of gas-surface interactions is pre-
sented and tested in the framework of TPM catalycity determination for Mars entry applications.
RTO-EN-AVT-142 15 - 5
2 Theory of LTE viscous o ws with variable elemental fractions
The physics of collision-dominated chemically reacting o ws is described by an extended Navier-Stokes sys-

tem, consisting of the following equations [67]: global continuity; momentum and total energy; a separate
continuity equation for each species, including nite-rate chemistry; if thermal non equilibrium occurs, an en-
ergy equation for each additional mode of freedom (vibrational, rotational and electronic energies).
This formalism has several drawbacks. Firstly, the equations are costly to implement and solve numerically.
Secondly, many physical parameters essential for the modeling of chemistry and energy relaxation processes
are usually missing. Thirdly, even when a numerical solution is obtained, it is by no means obvious to interpret
the vast amount of information obtained (e.g. concentration elds of a large number of chemical species). For
these reasons, when chemistry and energy exchanges are fast, it is usually preferable to solve the more elegant
and less uncertain Local Thermodynamic Equilibrium (LTE) form of the aforementioned set of equations.
A major breakthrough in the eld of LTE o w modeling was made by Butler and Brokaw [9, 7], who showed
that, assuming vanishing diffusive ux es of chemical elements, the diffusive transport of species reaction en-
thalpies in the energy equation could be incorporated in a straightforward manner by introducing a coefficient
of ‘thermal reactive conductivity’ λR :
N sp
X
W s h s = −λR ∇T. (4)
s=1
One often makes use of this result to reduce the full set of non equilibrium equations to a system formally
equivalent to the ‘conventional’ Navier-Stokes equations (continuity, momentum and energy), complemented
by a modi ed equation of state ρ(p, T ) computed from statistical mechanics assuming a x ed elemental com-
position in the o w. For instance, Vasil’evskii et al. [77] successfully used this classical LTE formalism to
simulate high-pressure air inductively coupled plasma o ws, imposing a 21/79 volumetric ratio of oxygen and
nitrogen elements throughout the o w eld. While appealing because of its simplicity, it is important to under-
stand that this approach is approximate at best, since in general the elemental composition varies signi cantly
in chemically reacting o ws. Over the past four decades, several LTE formulations accounting for (de)mixing
effects have been proposed in several contributions [48, 23, 24, 25, 26, 55, 70, 32, 22, 72, 73, 78, 68, 41, 45,
46, 48, 47, 18, 74, 59, 58, 64], a description of which is proposed in Ref. [57]. In the remaining part of the
section we will rstly present a preliminary analysis of the in uence of elemental demixing on the behavior
of a mixture of reacting gases by computing the composition for a x ed pressure and several elemental com-
positions, in the range of temperatures [300 K,15000 K] typical of applications to TPS testing. Secondly we
will recall the theory of LTE viscous o ws with variable elemental fractions for neutral mixtures derived in
Ref. [62], referring the extension to mixtures of ionized species to Ref. [57]. Although we will focus on the
particular case of air and carbon dioxide mixtures, well-suited for Earth and Mars entry applications, we wish
to point out that the presented LTE formulation is applicable to any chemically reacting o w near LTE and in
particular the hydrocarbon-air mixtures used for combustion.
2.1 In uence of elemental fractions on thermochemical equilibrium composition
It is well known that the composition of a mixture of reacting gases under thermochemical equilibrium can be
expressed as a function of pressure, temperature, and of the elemental fractions of the elements shared among
15 - 6 RTO-EN-AVT-142
the mixture species [79, 1]. The contents of this section deal mainly with the issue of elemental fraction vari-
ations in reacting o ws under LTE. To attract the readers’s interest, we present a preliminary analysis of the
in uence of elemental fraction variations on the mixture composition for both air and carbon dioxide mixtures.
As a result we show that a 10% variation of oxygen elemental fraction can induce a 50% difference in the
species concentration for CO2 mixtures.
(a) Mole fractions (X 0 = 0.189, X N = (b) Mole fractions (X O = 0.21, X N = (c) Mole fractions (X O = 0.231, X N =
0.811). 0.79). 0.769).
18 12
(d) Mole fractions (X O = 30
, XC = 30
). (e) Mole fractions (X O = 32 , XC = 31 ). (f) Mole fractions (X O = 22
30
, XC = 8
30
).
Figure 2: In uence of elemental fractions on thermochemical equilibrium composition for air and carbon
dioxide mixtures. (p=1 atm)
Consider an air mixture composed by the following 11 species : N2 , O2 , NO, N, O, N+2 , O+2 , NO+ , N+ , O+ , e- .
Let us de ne a reference mixture characterized by elemental fractions X O = 0.21, X N = 0.79, and two addi-
tional mixtures obtained by perturbing the oxygen fraction by an amount of ±10%, i.e. X O = 0.231, X N = 0.769
and X O = 0.189, X N = 0.811. In Figs. 2(a), 2(b), and 2(c) the composition of these mixtures is shown as a
function of temperature for a x ed pressure equal to 1 atm, assuming thermochemical equilibrium conditions.
From these gures appear that the species mole fraction is little affected by the elemental fraction, whose in-
uence is basically limited to a scaling of the various species according to the elemental fraction. The reason
is because in the air mixture, there are only 2 species including both N and O atoms, i.e. NO and NO+ , and in
addition their concentration always remains small.
For carbon dioxide mixtures, the sensitivity of the chemical composition to elemental fraction variations is
much higher than for air. Indeed, if one considers an 8 species mixture [59] in chemical equilibrium composed
by CO2 , O2 , CO, C, O, C+ , O+ , and e- , and computes the chemical composition as a function of temperature
for a x ed pressure, very different results are obtained depending on the elemental fraction used. Consider a
reference mixture characterized by elemental fractions X O = 23 and XC = 31 . Perturbing by ±10% the oxygen
22 8 18 12
fraction, two new mixtures are de ned: X O = 30 , XC = 30 and X O = 30 , XC = 30 . The mole fraction evolution
as a function of temperature, for a pressure of 1 atm, for these three cases is shown in Figs. 2(d), 2(e), and 2(f). A
RTO-EN-AVT-142 15 - 7
very strong in uence of the elemental fraction on the chemical composition is clearly visible. Moreover, from
the analysis of the previous results, it follows that, in the case of CO2 mixtures, the sensitivity to elemental
fraction variation is much higher than in the case of air.
This will result in an important in uence of elemental demixing on heat ux in thermochemical equilibrium
CO2 mixtures, as will be shown by the results to be presented later on.
2.2 Mixtures of neutral species
2.2.1 Species ordering and nomenclature
We represent mixtures of perfect gases by means of a nite set of N sp species S̃, amongst which we furthermore
distinguish between Nc ‘independent species’ Ẽ consisting of pure elements (as such as in their stable form)
and Nr ‘combined species’ R̃, for instance:
5-species air Ẽ = {O, N}, R̃ = {O2 , N2 , NO}, and S̃ = R̃ ∪ Ẽ,
valid for LTE mixtures at pressures above 0.01atm and for temperatures between 300 and 8000 K. We accord-
ingly de ne the three sets of indexes R = {1, . . . , Nr }, E = {Nr + 1, . . . , N sp }, and S = R ∪ E. We characterize
the chemical composition of the mixture in terms of mole fractions x s = n s /n, where n s and n stand for the
molar densities of individual species and of the entire mixture. Alternatively, we can also characterize the mix-
ture composition by means of the mass fraction y s = ρ s /ρ, where ρ s and ρ stand for the mass densities of the
individual species, respectively the full mixture.
We will indicate the number of atoms of element e contained in a species s by φes , for instance, for NO, φ53 = 1
while for O2 φ41 = 2. This enables us to de ne the mole fractions X e and mass fractions Y e of elements in the
mixture as follows:
φs xs
P e
s∈S
X Me
X = P P e ;
e
Ye = φes y s , (5)
φs xs Ms
e∈E s∈S s∈S
where M s is the molar mass of species s, related to the mixture molar mass M = s∈S x s M s . We introduce the
P
diffusion velocity V s with respect to the mass-averaged velocity of the mixture u. The molar and mass ux es
of species s are then given by respectively W s = n s V s and J s = M s n s V s .The mole and mass ux es of element
e are then given by X
Ne = φes W s ; J e = Me Ne (e ∈ E). (6)
s∈S
We use bold type fonts to indicate vectors in the physical space. With ‘bar notation’, we refer to arrays
containing species, elemental or reactive properties, with respective lengths of N sp , Nc and Nr .
15 - 8 RTO-EN-AVT-142
2.2.2 Computation of composition under LTE
Species continuity equations
We consider the commonly encountered o w regime in which chemical reactions are relatively rare w.r.t. elastic
collisions, such that they do not play an important role in the thermalization of species in the o w (unlike the
‘kinetic chemical equilibrium regime’ considered in the rst part of Ref. [18], for which chemical reactions
and elastic collisions are treated on the same level). The concentration of each species may then be determined
from a respective species continuity equation [67]:
∂t ( ρy s ) + ∇·( ρy s u) + ∇·(M s W s ) = ω s (7)
where u stands for the mass-averaged velocity of the mixture and ω s is the mass production/destruction
term [79] of species s due to chemical reactions. The number ux es of species respect the mass conserva-
tion constraint X
MsWs = 0 (8)
s∈S
and obey the Stefan-Maxwell equations [22, 32]

M X x s Wk − xk W s
= ∇x s (9)
ρ D sk f sk (L)
k∈S
for s ∈ S. To keep the analysis as simple as possible, we have neglected effects of pressure and thermal
diffusion; note however that these could be included without any particular problem as shown in Ref. [57].
Elemental continuity equations for neutral mixtures
As pointed out by Murphy, we need to solve additional element advection-diffusion equations to determine the
elemental composition of the mixture [46, 58]. The solution of these equations will then be used, together with
two independent thermodynamics variables, temperature and pressure for example, to compute the mixture
composition as the solution of a non-linear system [79]. To obtain these equations, we multiply Eqs. (7) by
φes Me /M s and sum over all species. Since no elements are created in the considered chemical reactions, the
mass fraction of any element e obeys the following equation:
∂t ( ρY e ) + ∇· ρuY e + ∇·J e = 0 (e ∈ E). (10)

In Ref. [62] we have shown that under LTE conditions, the mass diffusion ux of elements J e can be expressed
in terms of gradients of elemental mass fractions and temperature. For simplicity, we consider o ws at constant
pressure, consistent with the neglect of pressure diffusion in Eqs. (9). Once again, we remark that this does
not imply any fundamental limitation and pressure diffusion could be included easily if needed (see Ref. [57]).
Following the approach proposed in Ref. [62], the elemental diffusive ux es are expressed introducing the
RTO-EN-AVT-142 15 - 9
elemental multicomponent diffusion coefficients De f and thermal demixing coefficients DTe . This lead to the
following expression for the elemental diffusive ux es :
X
Je = − ρDe f ∇Y f − ρDTe ∇T. (11)
f ∈E
The second term in the right hand side of the above result will in general generate nonzero elemental diffusion
ux es even when the initial elemental composition is uniform.
2.2.3 Diffusive transport of enthalpy
In reacting o ws, the diffusion of species affects the mixture energy balance through the heat ux term:
X
qd = Wshs. (12)
s∈S
In a general non equilibrium case, to compute qd , one should determine all the W s as a solution of Eq. (9) and
then compute the above linear combination, as done in the methodology of Refs. [70],[18] and [58]. On the
other hand, under thermochemical equilibrium, Eq. (12) can be cast under a particular form which avoids the
computation of the W s . A rst step in this direction was made by Butler and Brokaw [9], who showed that,
under the assumption of vanishing elemental ux es, the diffusive heat ux takes the form of Eq. (4). In this
section, we extend the work of Butler and Brokaw to the more general case when elemental ux es are nonzero.
As shown in Ref. [62], for a mixture of reacting gases under thermochemical equilibrium, at constant pressure,
the diffusive heat ux is proportional to both temperature and elemental concentration gradients, and can be
expressed as: X X
W s h s = −(λR + λD )∇T − λeEL ∇Y e . (13)
s∈S e∈E
From Eq. (13), we see that in an LTE o w, the diffusive heat ux vector consists of three different parts:
(1) the ‘thermal reactive conductivity’ coefficient λR is identical to the well-known results by Butler and
Brokaw [9, 7] and takes into account diffusive transfer of species enthalpies in the absence of elemental
diffusion.
(2) The ‘thermal demixing conductivity’ coefficient λD corrects for the additional diffusive heat transfer that
occurs due to nonzero elemental diffusive ux es when elemental mass fraction gradients are zero.
(3) Finally, the ‘elemental heat transfer coefficients’ λeEL , take into account heat transfer due to elemental
demixing driven by gradients in elemental composition.
In the next section we will present an application of this formulation to investigate diffusion phenomena along
the stagnation line present in front of a probe placed in the test chamber of an inductively coupled plasma
facility.
15 - 10 RTO-EN-AVT-142
3 Diffusion phenomena in reacting mixtures : applications to stagnation line

o ws
This section is inspired by the research presented in Ref. [63], where the authors present a detailed analysis of
diffusion phenomena in o ws of reacting mixtures along a stagnation line present in front of a probe placed
in the test chamber of an inductively coupled plasma facility. In the following we limit our description to the
presentation of the results of the solution of the stagnation line differential problem and we refer the interested
reader to our previous publications [59, 58, 57] for a detailed presentation of the stagnation line equations. In
these references we also discuss the models used for the computation of the transport and thermodynamics
properties of reacting mixtures such as those used to obtain the following results.
The following mixture has been selected as representative for Mars entry :
Mars-8 species : E = {C, N, O}, R = {CO2 , CO, O2 , N2 , NO}.
The computations carried out using the Mars-8 mixture represent a direct application of the theory presented in
Sec. 2. This represents the rst attempt to simulate mixtures of hetero-nuclear molecules with more than two
elements using the closed form of the equations derived in Ref. [62]. The use of the theory presented in Sec. 2,
will moreover allow to show the improvements brought by the theory itself in terms of physical understanding
of the o w behavior. Although stagnation line o ws are often under chemical non equilibrium conditions, our
previous analysis [58] showed that for a CO2 mixture, assuming a fully catalytic wall provides an estimation of
wall heat ux close to the one obtained by assuming LTE conditions to be established along the stagnation line,
provided that elemental demixing is taken into account. In addition, the elemental fraction pro les obtained
under chemical non equilibrium follow closely those under chemical equilibrium. Thus we will present the
solution of the LTE stagnation line problem formulated following the equations presented in Ref. [57].
To present our analysis of the stagnation line o w we will start de ning the test case to be investigated and then
the solution of the stagnation line o w will be presented with strong emphasis on the description of diffusion
phenomena.
3.1 Test case de nition
The o w conditions considered, characteristic of Martian entry [66], are presented in Table 1.
The presented outer edge conditions and geometrical parameters, have been determined from a detailed numer-
ical [33] study of the o w eld inside the inductively coupled plasma generator and test chamber of the VKI’s
plasmatron wind tunnel [5]. This allows for the computation of some non-dimensional parameters related to
the stagnation line edge in terms of the boundary layer thickness, the axial velocity, and the velocity gradient
and further details about their de nition are available in previous publications [16].
In the following a single chemical regime is considered, corresponding to LTE with variable elemental fraction
(LTE-VEF), where the o w is in thermochemical equilibrium and the composition is computed as a function
of pressure, temperature, and local elemental fractions, obtained from the solution of the set of elemental con-
tinuity equations. Two formulations have been used.
RTO-EN-AVT-142 15 - 11
Table 1: Operating Conditions.

T w [K] 300
T δ [K] 5827
pδ [Pa] 7000
Rm [mm] 25
δ [mm] 9
2
vδ ∂/∂y(∂uδ /∂x)/(∂uδ /∂x) [-] 1.77
∂uδ /∂x [s−1 ] 913
LTE-VEF (1) we obtain the elemental concentration by solving Eqs. (10), where the element diffusive ux es
are computed as a linear combination of the species ones obtained as the solution of the Stefan-Maxwell
equations [Eq. (9)]. The computed species diffusive ux es are then used in the energy equation to evaluate
the diffusive transport of enthalpy [58] ( J s h s ).
P
LTE-VEF (2) we obtain the elemental concentration by solving Eqs. (10), where the thermal demixing and
multicomponent diffusion coefficients are used to compute the element diffusive ux es [Eq. (11)]. More-
over, an alternative form of the energy balance, based on the use of Eq. (13) is used, introducing the
correction to λR as well as the elemental heat transfer coefficients.
3.2 Stagnation line solution
In this section we present the solution of the stagnation line equations for the conditions speci ed in Tab. 1. We
start our analysis by discussing the temperature and enthalpy pro les shown in Fig. 3(a).
There we present the results obtained for the LTE-VEF regime using the two formulations LTE-VEF (1) and
LTE-VEF (2) previously de ned. As observed, the results obtained with the two formulations are identical.
The same match is observed for all o w variables along the stagnation line, supporting the correctness of the
alternative formulation recalled in Sec. 2.
In Fig. 3(b) we present the species concentration pro les. Starting from CO, O, C, and N at the outer edge, we
see how their concentration decreases providing a mixture of CO2 , O2 , N2 at the wall and a small amount of
NO. From the analysis of this picture, it appears that elemental demixing tends to reduce the amount of carbon
and nitrogen at the wall while it enhances the oxygen concentration. This is re ected by the presence of an ex-
cess of O2 in the low temperature region, which would have been zero if the wall elemental fractions had been
the same as at the outer edge. This behavior is con rmed by the pro les presented in Fig. 3(c), where we plot
the difference between the local elemental mass fractions and their outer edge values [Y e (y)-Y e |δ ]. In addition,
by looking to the element concentration pro les [Fig. 3(c)] we notice that the lack of carbon and nitrogen at the
wall is compensated by a higher concentration around y/δ = 0.4, while for oxygen we observe a minimum in
this position and a subsequent concentration rise as the wall is approached.
Going further with the analysis of the pro les presented in Fig. 3(c), we notice that they present a non-monotone
behavior and hereunder we give an alternative explanation for this observation. To discuss the elemental mass
fraction behavior we focus on the elemental continuity equations [Eq. (10)]. It is clear how the knowledge of
15 - 12 RTO-EN-AVT-142
(a) Temperature and enthalpy pro les. (b) Species mass fractions.
(c) Difference between the elemental mass fractions and

their outer edge values (Y e (y)-Y e |δ ).
Figure 3: Comparison of the computed results along the stagnation line obtained using the LTE-VEF (1) and
the LTE-VEF (2) formulations.
Je will help understanding the evolution of Y e knowing that the velocity is aligned with the stagnation line and
directed towards the wall.
For this purpose we present the pro les of Je in Fig. 4 for the three elements contained in the mixture. As
expected [Eq. (10)] we observe a correspondence between the local extrema of Je and those of Y e . In addition,
we notice that the diffusive ux es of elements are zero all over the rst 10% of the stagnation line and they
start to grow around y/δ ≈ 0.1. This shows that over the rst 10% of the stagnation line, Eq. (10) simpli es to
Je = 0, meaning that convection is by far negligible. Moving further towards the outer edge a local extremum
is encountered, followed by an increase for C and N, while for oxygen a decrease of JO is observed. From this
we notice that in the region de ned by y/δ > 0.1, the convective term will differ from zero and will counteract
the derivative of the pro les presented in Fig. 4. At this point, it is important to observe that, although the
pro les of Fig. 4 allow the interpretation of the elemental fraction pro les presented in Fig. 3(c), it is by no
RTO-EN-AVT-142 15 - 13
Figure 4: Elemental diffusive ux es (-Je ).
means evident to give an explanation for the behavior of the Je if they are computed as a linear combination of
the solutions of Eqs. (9) as done in the formulation LTE-VEF (1) [58, 64]. Now, to improve our understanding,
we use the theory introduced in Sec. 2 and applied in the LTE-VEF (2) formulation of the equations.
Indeed, to understand the behavior of the element diffusive ux es we discuss now the two components of Je .
The rst proportional to the temperature gradient (ρDTe ∇T ) and the second obtained as a linear combination
of the element mass fractions gradients (ρ Deq ∇Y q ). In Fig. 5(a), we present the rst part of the element
P
diffusive ux es. From Fig. 3(a) we notice that ∂T/∂y > 0 ∀y and therefore the sign of ρDTe ∇T will depend only
on DTe , ρ being obviously a positive quantity.
As shown in Ref. [62, 63] and con rmed in Fig. 5(b), the DTe change their sign as temperature rises, for x ed
pressure and elemental fractions. In the present case, the pressure is constant along the stagnation line but the
elemental fraction varies. Therefore, both temperature and elemental composition will in uence the transport
properties and therefore concur to the establishment of the pro les presented in Fig. 5(a).
Close to the wall, the thermal demixing coefficients are zero and do not contribute to the element diffusive
ux es. As temperature rises, we observe an increase in module of all contributions until a maximum is reached.
Then we notice a decrease and a subsequent change in sign typical of the thermal demixing coefficients behav-
ior.
Indeed, from Fig. 5(b), we see how the behavior of the thermal demixing coefficients is similar to the one of
the rst contribution to the element diffusive ux es.
In Fig. 5(c), we plot the second contribution to the elemental diffusive ux es. Also this contribution starts
from zero at the wall and follows a non monotone behavior passing through a change in sign for O and C. The
change in sign observed for the second contribution to the elemental diffusive ux es is not due to the elemental
multicomponent diffusion coefficients but to the elemental mass fractions. Indeed, as shown in Ref. [63] and
observed in Fig. 5(d), the multicomponent diffusion coefficients have the same sign in the temperature range of
interest. This allows for the following interpretation of diffusion phenomena along the stagnation line. Since
temperature decreases from the outer edge towards the wall, the temperature gradient is always positive along
the stagnation line. In sufficiently low temperature regions, where chemical reactions do not occur, the thermal
demixing coefficients are zero and do not induce demixing. As a consequence, since elemental multicomponent
diffusion coefficients are not zero at low temperatures, the elemental composition stays constant ensuring mass
15 - 14 RTO-EN-AVT-142
(a) First contribution to the element diffusive uxes (b) Thermal demixing diffusion coefficients.
(ρDTe ∇T ).
(c) Second contribution to the element diffusive uxes (d) Multicomponent elemental diffusion coefficients.
(ρ Deq ∇Y q ).
P
Figure 5: Elemental diffusive ux components, elemental multicomponent and thermal diffusion coefficients.
conservation ( Je |w = 0). As we move away from the wall (y < 0.05), the thermal demixing coefficients causes
a non zero contribution to the elemental diffusion ux which tends to decrease oxygen concentration (DTO > 0),
increase nitrogen concentration (DTN < 0), and slightly decrease but then increase again carbon concentration
(DCT > 0 and DCT < 0). As soon as variations in elemental composition are induced (∇Y e , 0), a counteracting
contribution to the element diffusive ux es appear which tends to smooth out elemental concentration pro les
until the outer edge is reached. In the central part of the stagnation line, the two contributions to element
ux es interact until the point in which the thermal demixing coefficients change sign. There, to counteract this
demixing term, the sign of element concentration gradients needs to change since the multicomponent diffusion
coefficients have a constant sign [Fig. 5(d)]. The counteracting character of the term ρDeq ∇Y q with respect to
P
ρDTe ∇T justify its label of ‘mixing term’ and the comparison of Figs. 5(a) and 5(c) clearly shows this behavior.
The mathematical origin of this diffusive character of the mixing term has to be searched in the non negativity
of the matrix containing the elemental multicomponent diffusion coefficients. To prove the latter property of the
RTO-EN-AVT-142 15 - 15
matrix Deq we computed its eigenvalues along the stagnation line. Since this matrix is singular [ q∈E Deq = 0],
P
one eigenvalue is k1 = 0 and the remaining two (k2 , k3 ) have been found to be positive and monotonically
increasing as a function of y [63]. This result shows the non negative character of the matrix Deq .
As a summary, we wish to point out that, from the knowledge of the set of transport properties including both
elemental thermal demixing and multicomponent diffusion coefficients, we can justify the evolution of the
diffusive ux es and therefore understand the shape of the elemental fractions pro les one can observe in the
solution of the stagnation line problem.
Figure 6: Heat ux components.
We now move to the analysis of the heat and mass transfer along the stagnation line by means of conduction
and diffusion. As shown in Sec. 2, under the assumption of local thermodynamic equilibrium, the diffusive
transport of enthalpy can be split into two parts related to ∇T and ∇Y q respectively. As a consequence, as
shown in Eq. (2), the computation of the heat ux involves the determination of two terms one proportional to
the temperature gradient [−(λ + λR + λD )∇T ] and the other proportional to the linear combination of element
P q
concentration gradients [− λEL ∇Y q ]. Moreover, the heat ux in a mixture of reacting gases is composed of a
conductive part (qc = −λ∇T ), to which the diffusive transport of enthalpy needs to be added (qd = h s J s ). To
P
investigate the heat ux along the stagnation line and highlight the relative importance of the various contribu-
tions we start presenting the evolution of q, qc , and qd along the stagnation line in Fig. 6. There we observe that
the total heat ux presents an increasing monotone behavior starting from the outer edge until the wall, acting
to balance the convection of enthalpy. At the same time we notice that this smooth evolution appears to be the
result of a quite complex shape of the two contributions qc and qd presented in the same gure. This highly
non-linear behavior of the two heat ux components is the result of the evolution of six transport coefficients
q
(λ, λR , λD , λEL ) and four gradients (∇T , ∇Y q ). To provide an exhaustive description of the heat ux along the
stagnation line, we start investigating the behavior of the three thermal (λ), thermal reactive (λR ) and demixing
(λD ) conductivities shown in Fig. 7(a).
As expected, we notice λR to be the major contribution to the rst part of the heat ux followed by λ. We also
notice that λD has a non-monotone behavior including changes in sign, highlighting the fact that demixing tends
to increase or decrease heat ux depending on the local temperature and elemental fractions as already observed
in Ref. [64]. Moreover, in terms of heat ux intensity, we notice that around y/δ = 0.3, λD ≈ λ showing how,
neglecting λD with respect to λ could lead to important errors. The knowledge of the evolution of the transport
15 - 16 RTO-EN-AVT-142
(a) Thermal, reactive, and demixing conductivities along (b) Heat ux components (λR ∇T , λD ∇T ) along the stag-
the stagnation line. nation line.
(c) Elemental heat transfer coefficients. (d) Heat ux components (λqEL ∇Y q ).
Figure 7: Heat ux components, thermal conductivities, and elemental heat transfer coefficients.
coefficients presented in Fig. 7(a) helps us understanding the behavior of the various heat ux contributions.
Indeed, the non-linear behavior of qc can be easily explained. λ is an increasing monotone function from the
wall towards the outer edge, while ∂T/∂y starting from the outer edge increases, decreases and then increases
again as shown in Fig. 3(a). Therefore qc follows mainly the evolution of ∂T/∂y. For what concerns qd , we
q
need to consider separately the ve contributions −λR ∇T , −λD ∇T , and −λEL ∇Y q .
In Fig. 7(b), we present the evolution of the rst two terms and their sum. The evolution of these two contri-
butions is more complex than the previous one since the two associated transport coefficients present a highly
non-linear behavior with extrema in different positions than the temperature gradient. As far as the elemental
heat transfer coefficients are concerned, we present their evolution in Fig. 7(c). There we notice that they all
have a non-monotone behavior re ecting the highly reacting character of the o w. These transport coefficients
lead to three contributions to the heat ux presented in Fig. 7(d), which are of the same order of magnitude for
the three elements. The sum of these contributions is also plotted in the same gure to show the importance of
RTO-EN-AVT-142 15 - 17
this term with respect to the total heat ux.

From the analysis of the pro les presented in Figs. 7(b)-7(d), we can justify the evolution of qd concurring to
the determination of the nal shape of the heat ux. As a nal step we analyze in detail the relative importance
of the various terms contributing to the heat ux. In Fig. 8, we present the results of this analysis showing the
evolution of the four following ratios along the stagnation line : −(λdT/dy)/q, −(λR dT/dy)/q, −(λD dT/dy)/q,
P q
and −( λEL dY q /dy)/q.
Figure 8: Normalized heat ux components.
The analysis of these pro les help us to quantify the importance of each term. Indeed, this shows that the
major contribution to the heat ux comes from the thermal reactive conductivity (λR ), the second comes from
conduction (λ) and the two remaining contributions are of the same order of magnitude. The analysis of the
curve −(λD dT/dy)/q reveals that the contribution related to elemental demixing is higher than the one due to
P q
thermal conductivity for 0.1 < y/δ < 0.3. The two contributions −λD dT/dy and − λEL dY q /dy present almost
P q
an opposite behavior for y/δ < 0.4. On the other hand, for y/δ > 0.4 we notice that − λEL dY q /dy becomes
higher than −λD dT/dy approaching the 10% of the total heat ux around y/δ ∼ 0.6 and decreasing to 5% at the
outer edge. The almost opposite behavior of the two latter contributions clearly shows that elemental diffusive
ux es have not a very important in uence on the total heat ux. Indeed if Je = 0 ∀e ∈ E, the Butler and
Brokaw thermal reactive conductivity is enough to compute qd . For the conditions previously analyzed, the
presence of non vanishing elemental ux es acts on the mixture energetic behavior with a heat ux contribution
which is at most of the order of 5% of the total heat ux.
4 Applications to ICP o w computation
In this section we present a detailed analysis of chemical equilibrium and non equilibrium plasma o ws both in
the torch and in the test chamber. In the remaining part of the section we limit our discussion to the presentation
of the results of several plasma o w simulations and we refer the interested reader to previous publications [64,
75, 17] for the description of the ICP model used to obtain these results. In the same references the reader
will nd an extensive description of the models used to compute the thermodynamic and transport properties
15 - 18 RTO-EN-AVT-142
of reacting mixtures such as those used to obtain the results to be presented shortly.
In Sec. 4.1 we will present results concerning only the torch domain for carbon dioxide plasmas, while in
Sec. 4.2 we extend our investigation to the hot air jet o wing into the chamber.
The two following mixtures will be considered in this section to simulate air and carbon dioxide plasmas :
Air-11 species : Ẽ = {O, N, e+ }, R̃ = {O2 , N2 , NO, NO+ , O+ , N + , O+2 , N2+ }, and S̃ = R̃ ∪ Ẽ,
CO2 -8 species : Ẽ = {O, C, e+ }, R̃ = {CO2 , O2 , CO, O+ , C + }, and S̃ = R̃ ∪ Ẽ.
The analysis presented in the rst part of this section is based on the study presented by Rini et al. in Ref. [65]
where a code-to-code validation is presented for carbon dioxide plasmas in the VKI plasmatron torch. To this
end the authors present the results of the simulation of CO2 plasma o ws within the VKI plasmatron torch
carried out obtained independently from the two groups with the VKI-ICP and the IPM-α codes. The results of
this comparison showed how the 1-D approach of the α code fails in the description of the Electric eld within
the torch. In addition, the importance of elemental diffusion on the behavior of CO2 plasma o ws is assessed
and only this part will be presented in the following whose motivation is recalled hereafter:
(1) accurate carbon dioxide plasma o w simulations are rare in the literature related to aerospace sciences.
Moreover, up to the best author knowledge, no simulation is available where elemental demixing is
considered under thermochemical equilibrium conditions. For this reason we investigate the phenomena
of elemental diffusion in LTE carbon dioxide plasmas to evaluate its effects on the o w behavior.
The second section is devoted to the physico-chemical characterization of the jet o wing in the test chamber
for air plasmas, whose motivation is twofold :
(1) the methodology proposed by IPM [35] to estimate the catalytic properties of thermal protection system
materials supposes thermochemical equilibrium conditions to be established in the torch and in the test
chamber. However, there is no proof that equilibrium conditions exist and a detailed analysis assessing the
quality of this assumption is not available. To supply this need we present both chemical non equilibrium
and chemical equilibrium computations to provide a range of pressure in which equilibrium computations
are as accurate as those obtained under chemical non equilibrium, using two nite rate chemistry models.
(2) As a result, we will be able to assess the in uence of the nite rate chemistry model and operating
pressure on the o w behavior both in the torch and especially in the test chamber.
4.1 Analysis of carbon dioxide plasma o ws within the torch
We now move to the analysis of the numerical results obtained for an operating condition of the VKI plasmatron
representative of Mars entry [33]. The analysis presented in this section is inspired on the study conducted by
Rini et al. in Ref [65], omitting the part concerning the code-to-code comparison. Therefore, in the following,
RTO-EN-AVT-142 15 - 19
we only present the investigation of elemental diffusion to assess the importance of the variations in carbon and
oxygen elemental concentrations on the o w behavior. In Tab. 2, we present the operating condition selected
as representative of Mars entry [33] used to produce the results to be presented shortly.
Table 2: Operating conditions for the VKI-Plasmatron operating with carbon dioxide.
Ambient and wall temperature [K] 300
Power injected into the plasma [kW] 90
Frequency [MHz] 0.37
Inlet Swirl [deg] 0
Mass o w [g/s] 8
Operating pressure [Pa] 7000
The o w has been simulated under LTE with both constant and variable elemental fractions and both com-
putations have been carried out on the same mesh of 102 by 92 cells for which grid convergence is assured
(Ref. [64]). All results presented in the following are converged at least 10 order of magnitude (based on the
drop of the L2 residual norm).
To investigate the in uence of the elemental fraction variation on the o w behavior, we used the upgraded ver-
sion of the VKI-ICP code [57] to simulate both the LTE-VEF and the LTE with Constant Elemental Fraction
(LTE-CEF) formalisms. For that purpose, the plasma in the torch has been considered both under LTE-VEF
and LTE-CEF conditions. For the operating conditions presented in Tab. 2 we obtain the elemental distribution
presented in Fig. 9(a) in terms of carbon elemental mole fraction (XC ).
In Fig. 9(b), we present the temperature contours and in Fig. 9(c) the o w pattern, both within the torch, ob-
tained under LTE-VEF conditions. In these gures, we observe the presence of a quite wide high-temperature
region where the maximum temperature is around ∼ 10 000 K, as expected for the speci ed power injected in
the o w. The mixture enters the torch with a reference elemental composition of 2/3-1/3 for oxygen and carbon
respectively. Entering the torch, the mixture is subject to very strong temperature gradients which produce a
non zero demixing contribution (−ρDTe ∇T ) to the elemental diffusion ux [62], inducing elemental composi-
tion variations. Once a non-uniform elemental composition eld is established, a counteracting mixing term
( ρDck ∇Y k ), proportional to the elemental composition gradients, tends to re-establish a uniform elemental
P
eld. These two mechanisms act together with convection to provide the nal steady elemental composition
pattern. From the analysis of Fig. 9(a), at least three zones can be observed. The rst one represents an almost
cylindrical zone delimited by 7.5 < r < 8.0 cm close to the inlet and 6.0 < r < 8.0 cm at the outlet. There,
we do not observe substantial variations in the elemental fraction and this is mainly due to the fact that, in that
region, the o w is continuously lled by a stream coming from the inlet with the reference elemental composi-
tion. In addition, the wall temperature being x ed, the axial temperature gradient is negligible close to the wall
leading to the absence of diffusion in the same direction.
A second zone is visible around the torch axis both inside and after the recirculation bubble. There, we ob-
serve that carbon has the tendency to diffuse out from hot regions towards lower temperatures, leading to an
axial zone poorer in carbon than in the case where a constant elemental composition is arbitrarily imposed.
A third region is visible at the intermediate radii and a bit downstream of the inlet. There, the carbon mole
fraction reaches the maximum value of 0.38, i.e. ∼ 13% more than the inlet value. This zone of high carbon
concentration is therefore the result of the diffusion of carbon from the high temperature region around the axis
accompanied by the difficulty for carbon to diffuse towards the quartz tube where the concentration is mainly
15 - 20 RTO-EN-AVT-142
0.33 300 1300 2300

0.3 0.37 6300
7 6300
0.3
0.33
5
9300 9300
10300
9
0.28
0.2
1 0.29 8300
0.3
0.36 7300 6300
0.38
0.33 300 1300
(a) Carbon elemental molar fraction. (b) Temperature (K).
(c) Stream function.
Figure 9: Carbon dioxide plasma o ws within the torch (LTE-VEF regime).
x ed by the cold mixture convected from the inlet to the torch exit.
The previous description of the carbon elemental eld is re ected by the pro le depicted in Fig. 10(a) which
corresponds to the torch exit. There we clearly see that the carbon pro le has a non-monotone behavior. As
we approach the exit, the o w eld became almost axial and radial convection can be neglected as shown in
Fig. 9(c). This behavior of the outlet carbon elemental fraction, observed for nitrogen in air mixtures [64], could
be predicted by observing the evolution of the elemental diffusion coefficients for the carbon element [59] as a
function of temperature. In Figs. 10(b) and 10(c), we present the in uence of elemental fraction variations on
the mixture composition and on the mixture temperature. From the analysis of Fig. 10(b), a big in uence of
elemental fraction variations on the species concentration is observed. Starting from the wall we observe that,
since the carbon molar fraction is slightly smaller than 1/3, a small amount of O2 is present close to the wall
as a result of a LTE-VEF computation. Moving further towards the axis the differences between the CEF and
the VEF computations become more evident. Indeed, the decrease of XC up to a local minimum around 6 cm
causes a slower CO2 dissociation accompanied by the displacement to lower radii of both the O2 and the CO
peaks. Then, because of the successive CO dissociation we observe the formation of both atomic carbon and
oxygen, where the second one is more affected by the elemental fraction variation because of the retardation of
both O2 and CO peaks. Finally, atomic species decrease their concentration close to the axis because of ioniza-
tion due to the high temperature around 10000 K. On the axis we therefore nd a carbon fraction of ∼ 0.275,
RTO-EN-AVT-142 15 - 21
(a) Carbon elemental mole fraction. (b) Species mole fractions.
(c) Temperature.
Figure 10: Radial pro les at the outlet of the torch.
i.e. ∼ 16% less than the reference value.

The nal in uence of elemental demixing is observed in Fig. 10(c), where the outlet temperature pro les for
both CEF and VEF are compared. We notice the temperature to be overestimated in the case of constant ele-
mental composition starting from r < 7 cm. By comparing Fig. 10(c) with the results presented in Ref. [64], we
easily realize that the in uence of elemental diffusion on the temperature pro le is much higher for CO2 than
for air mixtures.
4.2 Characterization of air plasma o w in the test chamber
We now move to the analysis of air plasma o ws. The results to be presented shortly concern both the in-
vestigation of the torch and the hot exhaust jet in the chamber. As discussed at the beginning of this section,
the forthcoming results will allow to verify for which operating conditions of the VKI plasmatron, the nu-
merical simulation of LTE plasma o ws is as accurate as the chemical non equilibrium one, both in the torch
15 - 22 RTO-EN-AVT-142
and especially in the test chamber. The consequences of this nding will have a direct impact on the range of
applicability of the methodology used for the estimation of catalytic properties of thermal protection system
materials [35].
The mesh used for the solution of the o w eld equations is divided into two blocks: the rst one (60 by 44
cells) is used to discretize the domain within the torch, while the second block (60 by 67 cells) reproduces a
part of the test chamber. All results presented in the following are converged at least 10 order of magnitude
(based on the drop in the L2 residual norm).
The operating conditions, selected as representative of Earth (re)entry [51], are recalled in Tab. 3.
Table 3: Plasmatron operating conditions.

Ambient and wall temperature [K] 350
Power injected into the plasma [kW] 80
Frequency [MHz] 0.37
Inlet Swirl [deg] 0
Mass o w [g/s] 8
Operating pressures [kPa] 5, 10, 20, 30
For each simulation, the three chemical regimes chemical non equilibrium (CNEQ), LTE-CEF, and LTE-VEF
have been considered.
In the following we present a qualitative analysis based on the investigation of the contour plots of the main
o w properties and we refer the interested reader to Ref. [57] for a more detailed description of the o w behav-
ior based on the analysis of temperature and elemental molar fraction pro les extracted in several crucial parts
of the computational domain.
We rstly address the issue of the in uence of nite rate chemistry on the results of the simulation. To this
end, two models have been selected as representative of Earth’s atmosphere chemistry (Park [53] and Dunn-
Kang [29]). Several numerical experiments have been conducted to assess how the operating pressure drives
the o w behavior when one of the two models is used. As a result, we notice that the in uence of the nite rate
chemistry is only limited to low pressures (around 5000 Pa). Indeed, as the operating pressure is set to higher
values, the results obtained under CNEQ with both models are in very good agreement.
We start presenting temperature and elements concentration both in the torch and in the chamber by means of
contour plots. Two of the four operating pressures speci ed in Tab. 3 have been selected to perform this study,
the lowest (5000 Pa) and the highest (30000 Pa).
Focusing on these two pressures, we will be able to easily show how the differences observed in the o w solu-
tion, due to differences in the reaction rates, are visible only in the low pressure case and disappear when the
operating pressure is increased up to 30000 Pa. The results to be presented shortly have been obtained under
LTE-VEF conditions and under CNEQ using two nite rate chemistry models: the Park model [53], indicated
as CNEQ-P, and the Dunn-Kang model [29] named as CNEQ-D&K.
Each non equilibrium result is compared with the respective one obtained with the other nite rate model but
also with the result of the simulation carried out under thermochemical equilibrium with variable elemental
fraction. This, at the same time, allows for the assessment of the in uence of the nite rate chemistry model
and to answer the question of whether or not equilibrium computations give the same results as non equilibrium
simulations for high pressures in the chamber.
In Figs. 11(a) and 11(b) we plot the temperature, respectively the oxygen elemental mass fraction both in the
RTO-EN-AVT-142 15 - 23
CNEQ-D&K CNEQ-D&K
1100
350 2000 0.21 0.23 0.24
6000 4000
5000
0.16
6000 0 8000 6000 0.18
7 00 8000 9230 7000 0.17 0.18 0.19
600 7000
8000 8000 7000 0.16
0.16
0.19
9230
0.18
0 6000 0.19
7000 6000 5000 4000
350 2000 0.21 0.23 0.24
1100
CNEQ-P CNEQ-P
(a) Temperature (K) contours obtained with two nit e (b) Oxygen elemental molar fraction obtained with two
rate models. nite rate models.
CNEQ-D&K CNEQ-P
1000 1100
350 2000 350 2000
6000 5000 4000 4000 6000 5000 4000 3000
5000 5000
8000 8000
6000 7000 8000
7000
6000 600
0
7000 8000 9000
7000
6000
9230
7000 9230 10000 9230 8000 7000 7000 9000 10000

9000 8000 7000
6000 6000 6000 6000
6000 6000
5000 4000 5000 4000
3000 3000
350 2000 350 2000
1000 1100
LTE-VEF LTE-VEF
(c) Comparison of the temperature (K) contours ob- (d) Comparison of the temperature (K) contours ob-
tained with the Dunn-Kang model (upper) and under tained with the Park model (upper) and under LTE-VEF
LTE-VEF (lower). (lower).
CNEQ-D&K CNEQ-P
0.21 0.23 0.21 0.24

0.22 0.23
0.16 0.16
0.18 0.19 0.18 0.19
0.17 0.19 0.17 0.19
0.17 0.17
0.19 0.19
0.18 0.18
0.16 0.17 0.16 0.17
0.21 0.24 0.21 0.24
0.24 0.24
LTE-VEF LTE-VEF
(e) Comparison of the oxygen elemental molar fraction (f) Comparison of the oxygen elemental molar fraction
contours obtained with the Dunn-Kang model (upper) contours obtained with the Park model (upper) and un-
and under LTE-VEF (lower). der LTE-VEF (lower).
Figure 11: Air plasma o w within the VKI plasmatron facility under CNEQ, LTE-CEF, and LTE-VEF condi-
tions (5000 Pa).
torch and in the chamber. The analysis of these gures reveals the presence of two zones in which the compar-
ison of the results bring us to different conclusions. Indeed, within the torch, we notice that the CNEQ-P and
CNEQ-D&P predictions differ considerably leading to different temperatures and, to a lesser extent, elemental
fractions. From Fig. 11(a) we observe that the Dunn-Kang model predicts a slightly wider hot (T≥ 6000 K)
region in the torch acting on the structure of the plasma ball. As soon as we move towards the torch exit, we
start noticing a rapid dwindling of the disparity between the two predictions which, at least from a qualitative
point of view, show the same temperature eld in the jet, corresponding to the most interesting part of the
15 - 24 RTO-EN-AVT-142
CNEQ-D&K CNEQ-D&K
1100
350 2000 0.21 0.24
3000 0.24
5000 4000 0.17
6000 0.17
0.18
700
0 0.18 0.19
8000 9000 10000 7000
8000 9000 10000 8000 7000 0.17 0.19
7000 0.17 0.1
8
6000 5000 0.17
4000
3000 0.24 0.24
350 2000 0.21
1100
CNEQ-P CNEQ-P
(a) Temperature (K) contours obtained with two nit e (b) Oxygen elemental molar fraction obtained with two
rate models. nite rate models.
CNEQ-D&K CNEQ-P
1100 1100
350 2000 350 2000
3000 3000
5000 4000 5000 4000
6000 6000
700 700
0 8000 0 8000
8000 9000 10000 7000 8000 9000 10000 7000
8000 9000 10000 8000 7000 8000 9000 10000 8000 7000
7000 7000
6000 5000 6000 5000
4000 4000
350 3000 2000 350 3000 2000
1100 1100
LTE-VEF LTE-VEF
(c) Comparison of the temperature (K) contours ob- (d) Comparison of the temperature (K) contours ob-
tained with the Dunn-Kang model (upper) and under tained with the Park model (upper) and under LTE-VEF
LTE-VEF (lower). (lower).
CNEQ-D&K CNEQ-P
0.21 0.23 0.24 0.21 0.24 0.24

0.19 0.19
0.17 0.17
8
0.18 0.18 0.17
0.1
0.18 0.18
0.18 0.18
0.17 0.18 0.17 0.18
0.17 0.17
0.24 0.19 0.24 0.19
0.21 0.24 0.21 0.24
LTE-VEF LTE-VEF
(e) Comparison of the oxygen elemental molar fraction (f) Comparison of the oxygen elemental molar fraction
obtained with the Dunn-Kang model (upper) and under obtained with the Park model (upper) and under LTE-
LTE-VEF (lower). VEF (lower).
Figure 12: Air plasma o w within the VKI plasmatron facility under CNEQ, LTE-CEF, and LTE-VEF condi-
tions (30000 Pa).
o w eld from the point of view of TPM testing.

For what concerns the oxygen elemental fraction, we notice that the nite rate chemistry model has little in-
uence on the elements concentration eld, showing once again that demixing is not a non equilibrium effect
[Fig. 11(b)]. In Figs. 11(c)-11(d) and 11(e)-11(f), we present a comparison between the results obtained under
chemical non equilibrium with the Park and Dunn-Kang models with those obtained under chemical equilib-
rium with variable elemental fractions. In Figs. 11(c)-11(d) we focus on the temperature contours while in
Figs. 11(e)-11(f), we investigate the oxygen elemental fraction. As expected [64], the temperature contours ob-
RTO-EN-AVT-142 15 - 25
tained under chemical non equilibrium differ considerably from those obtained with the equilibrium formalism
and this for both nite rate chemistry models. From the oxygen elemental fraction presented in Figs. 11(e)-
11(f) we notice a certain difference in the prediction of the elemental concentration eld between equilibrium
and non equilibrium even though the order of magnitude of the oxygen variations is the same.
We now move to the analysis of the higher pressure (30000 Pa) and we start our analysis presenting, as done
for the lower pressure case, the comparison between non equilibrium results. In Figs. 12(a)-12(b) we present
the temperature, respectively the elemental fraction contours both in the torch and in the chamber. From the
analysis of the temperature contours, we notice that at an operating pressure of 30000 Pa, the results obtained
with the two models are, at least from a qualitative point of view, equivalent. The same conclusion is drawn
after inspection of Fig. 12(b) where attention is focused on the oxygen elemental fraction. Differences between
the prediction with the Park and the Dunn-Kang models are hardly visible. In the following, we present the
comparison between the previous results with the respective equilibrium computations. The critical analysis
of these comparison will enhance the importance of the previous conclusions. In Figs. 12(c)-12(d) and 12(e)-
12(f) we present the direct comparison between the temperature, respectively the elemental mole fraction con-
tours obtained both under chemical non equilibrium and equilibrium with variable elemental fraction. From
Figs. 12(c)-12(d) we realize that the temperature contours predicted by the simulation of equilibrium condi-
tions coincide with those obtained under chemical non equilibrium for both nite rate chemistry models after
the last coil. Indeed, some small differences are visible within the torch until the last coil is reached. To further
decrease these differences in the torch we just need to increase the operating pressure. On the other hand, at
the torch exit, but most importantly in the test chamber, a perfect match is observed. Similar conclusions can
be drawn for the elemental composition after inspection of Figs. 12(e)-12(f).
The results just described have a direct impact on the methodology used for the estimation of the catalytic prop-
erties of thermal protection materials. Indeed, thanks to the previous analysis we can argue that for sufficiently
high pressures the equilibrium formalism allows for the determination of results, both in the torch and in the
chamber, characterized by an accuracy very close to the one we can achieve with non equilibrium simulations.
This conclusion will allow researchers performing inductively coupled plasma o w simulations for the charac-
terization of TPM, to obtain accurate results using an equilibrium formulation at a fraction of the cost required
to simulate the same conditions under non equilibrium.
5 Practical application to TPM testing
The von Karman Institute has been involved since 1997 in the determination of catalytic properties of ther-
mal protection-system materials. Recently, with the initiation of several Mars exploration projects, interest has
been attracted to Mars entry problems [19, 20, 11]. The two inductively coupled plasma (ICP) facilities of
the Institute have been equipped to work with CO2 mixtures [59]. The VKI high-enthalpy o w solvers are
interfaced with libraries to compute thermodynamic, transport, and non equilibrium chemistry properties. The
thermodynamic and transport libraries have been recently updated to incorporate CO2 mixtures. In this section
the physico-chemical description of a CO2 o w is completed, allowing for simulations in chemical non equilib-
rium, characterized by both bulk and surface reactions. A critical output of the LHTS methodology is the heat
load on the stagnation point of a ying body, where large gradients of temperature and mass concentrations are
present. These features make this point a suitable benchmark for testing thermochemical models. The o w eld
15 - 26 RTO-EN-AVT-142
in the vicinity of the stagnation point has been analyzed by means of stagnation-line o w calculations. These
computations are applied for the determination of surface catalytic efficiency in CO2 mixtures, leading to the
determination of so-called heat- ux maps. The analysis that ensues is based on the study presented by Rini et
al. in Ref. [59] whose motivation is recalled here after :
(1) the simulation of gas/surface interactions in CO2 o ws requires a special care because of the coupling
between wall reactions sharing oxygen atoms. This is correctly described by the model presented by Rini
et al. in Ref. [59], which will be successfully applied to describe heterogeneous recombination in carbon
dioxide mixtures.
(2) The model proposed in Ref. [59] for the description of gas/surface interactions is compared against the
one presented by Scott [69] showing that the latter one is not consistent with element mass conservation.
(3) In the aerospace related literature, the usual de nition of a ‘catalytic surface’ is often given as the prop-
erty of a certain material to promote recombination of atoms or molecules on the surface itself. As a
consequence, a fully catalytic surface is said to be the one that forces a complete recombination induc-
ing local conditions which tend to the equilibrium limit. Following this observation, several researchers
working in the eld of TPS design treat the gas/surface interactions for fully catalytic materials by com-
puting the local mixture composition as the one corresponding to equilibrium conditions de ned by the
local pressure and temperature but using a reference elemental fraction, usually corresponding to the free
stream. Thanks to the analysis of the results to be presented shortly, we will show that to de ne the
equilibrium conditions representing the limit toward which a fully catalytic condition should tend, the
local elemental fraction should be used instead of the free stream one.
(4) A crucial aspect of the methodology for the estimation of the catalytic properties of TPM is the heat ux
map expressing the relation between wall heat ux, wall temperature and the effective recombination
probability. As a nal summary of the previous investigation, we will show how the outer edge elemental
fraction affects the heat ux map. This will allow for the estimation of the in uence that elemental
demixing in the jet could have on the prediction of the effective recombination probability of thermal
protection materials.
5.1 Stagnation line o w computations
In this section, we present the solution of the stagnation line differential problem for a carbon dioxide mixture
under chemical non equilibrium conditions. As in Sec. 3, our attention will be focused mainly on the presen-
tation of the results while we refer the interested reader to our previous publications [59, 4] for a complete
description of the stagnation line equations. The test case de nition is based on several simulations and experi-
ments presented by Kolesnikov in Refs. [39] and [38] referring to the 100-kW IPG-4 plasmatron (IPM). Many
calculations have been carried out at VKI [56] covering a wide range of enthalpy conditions. To investigate
the importance of bulk and surface chemistry, we focus our attention on two operating conditions characterized
by different outer edge enthalpies presented in Tab. 4, where hδ is the mixture enthalpy at the outer edge, V s
the incoming o w velocity and p the operating pressure. This will lead to different chemical compositions
and temperatures at the outer edge. As a consequence, different chemical processes will be enhanced by the
RTO-EN-AVT-142 15 - 27
bulk recombination and the presence of atomic carbon at the outer edge in the second case, will in uence the
de nition of the set of wall reactions.
Table 4: Flow conditions.

Low enthalpy case High enthalpy case
hδ [MJ/kg] 15.3 39
pδ [atm] 5.8·10−2 0.1
Rm [mm] 25 25
δ [mm] 10 10
vδ ∂/∂y(∂uδ /∂x)/(∂uδ /∂x)2 [-] 2.09 2.09
∂uδ /∂x [s−1 ] 1794 2558
The o w is considered to be under equilibrium conditions at the boundary layer edge. Therefore, the outer edge
temperature is the result of an iterative calculation in which the thermodynamic modeling of the o w plays an
important role. The corresponding values of temperature are 5922 K and 8127 K for the low and high enthalpy
cases respectively.
Table 5: Chemical reactions used in the Stagnation Line computations. (M=CO2 , CO, O2 , O, C)
Dissociation reactions
CO2 + M CO + O + M
CO + M C + O + M
O2 + M O + O + M
Neutral exchange reactions
CO+OC+O2
CO2 +OCO+O2
For the given o w conditions at the boundary layer edge, the heat ux at the stagnation point of the model has
been computed as a function of the surface temperature T w and the surface effective catalycity γw , from the
solution of the stagnation line equations. The analysis of the non equilibrium dissociated subsonic o w around
the model is based on the assumptions used to derive the stagnation line equations [4, 59] and the following
ones :
• the gas is represented by a 5 species (CO2 ,CO,O2 ,O,C) viscous and heat conductive mixture.
• The reaction set describing the nite rate bulk chemistry of the o w is the one described in Tab. 5.
• The surface catalytic reactions proceed according to the Eley-Rideal mechanism [21] and the following
models are considered for the various cases analyzed.
2-surface reactions model :
(1) O + (O − S ) → O2 + S (2) CO + (O − S ) → CO2 + S
3-surface reactions model :
(1) O + (O − S ) → O2 + S (2) CO + (O − S ) → CO2 + S (3) C + 2(O − S ) → CO2 + 2S
15 - 28 RTO-EN-AVT-142
Another important assumption in the present model is that all the recombining species have the same effective
recombination probability, i.e. for the mentioned wall chemistry model we have
γCOw = γOw = γCw = γw . (14)
As the result of the computations performed along the stagnation line, for a 50 mm diameter probe, several
charts have been determined for the stagnation point heat ux as a function of T w and γw ∈ [0, 1], and will be
presented later in this section.
5.1.1 Gas surface interactions
The description of non equilibrium o ws requires the modeling of recombination phenomena happening at
the solid surface. To this end we have speci ed two sets of heterogeneous reactions. The next step consists
in modeling the way in which the species involved in these reactions exchange atoms in order to express the
mass production term ωi,cat as a function of known quantities. Scott [69] proposes a model of the boundary
conditions for the species equations [Eq. (7)] based on the following expression for the net mass ux of species
i at the surface
Nr
X X N sp
Nr X
↓
ωi,cat = Mi Mi νri γ −
r
µi jr γr M j M↓j , (15)
r=1 r=1 j=1
where γr is the reaction probability of the rth reaction, Mi the molar mass of the ith species, and M↓i is the
thermal agitation ux of i-particles towards the surface. The matrix νri indicates which ith reactants collide
upon the surface in the rth reaction, while the matrices µi jr de ne the jth incident reactants producing the ith
products for the rth reaction. Within the methodology followed for the determination of catalytic properties
of TPS materials, the reaction probability γr is supposed to be independent from the reaction and equal to an
effective recombination probability γw .
Therefore, the expression for the impinging ux M↓i , if the Chapman-Enskog perturbation term is considered,
reads r
↓ 2 kT w
Mi = ni .
2 − γw 2πMi
This formulation ensures a zero net total mass ux at the wall but does not satisfy in general the element
conservation, indeed, considering a catalytic surface model constituted by the two reactions O + O → O2 and
CO + O → CO2 , one notices a coupling between the two chemical processes. This in some way establishes a
link between the reaction probabilities of the two reactions which is not speci ed in the previous formulation.
Elaborating further on this consideration, an alternative approach is proposed. Let us de ne a rate of reaction
per unit surface and unit time for the wall reaction χk , which represents the frequency at which reactions take
place on the surface.
With this in mind, a formalism similar to the one used for bulk reactions is followed. The number of moles
produced per unit surface and unit time are given by
N
ωi X r
00 0
= νir − νir χr (16)
Mi r=1
RTO-EN-AVT-142 15 - 29
0
where, for νir , 0
M↓i
χr = γir . (17)
νir
0
The surface reaction rate χr is therefore related to the recombination probability of species i involved in the rth
reaction γir by the relation (17). As the reaction rate χr has to be independent of the species i, it results that
0
the species reaction probabilities γir are linked by the constraints γir M↓i /νir = constant = χr , for all species such
0
that νir , 0.
√
In this case, the ux of particle M↓i is given by M↓i = 2/(2−γi )ni kT w /(2πMi ), while γi = r=1,Nr γir represents
P
the recombination probability of species i. The computation of the boundary condition for the species continuity
equations will therefore consists in rstly calculate χr and then evaluating the production term by means of
Eq. (16). The purpose of this alternative formulation is not to better describe the physical phenomenon of
heterogeneous catalycity, which is hidden in the de nition of the different recombination probabilities. But at
least, it provides a model which is consistent with both the global mass and element conservation at the wall,
as will be demonstrated by numerical experiments. Moreover, it should be noticed that the quantity χr depends
only on the reaction considered and not on the species involved in it, which represents another similitude with
the bulk reaction mechanism.
5.1.2 Low enthalpy case
We start our analysis considering the low enthalpy case for which only the 2-surface reactions model has been
applied, using both Scott’s model [69] and the formulation presented in the previous section. This reaction
scheme is based on the recombination of atomic oxygen and carbon oxide to produce molecular oxygen, re-
spectively carbon dioxide. The success of these two processes is based on the simultaneous presence of O and
CO close to the surface. The atomic oxygen will then concur to both recombination processes while CO only
to one of them. The description of the kinetics of these two reactions requires the knowledge of the rates of
each chemical process. These frequencies will be computed on the basis of the effective probability that an
atom of oxygen has to be involved in the ith process (γOi ) and on the probability that a molecule of CO has to
j
be involved in the jth reaction (γCO ). Obviously these effective recombination probability are related by mass
conservation and constrained by Eq. (14). Nevertheless, a correct treatment of these issues is missed in several
gas/surface interaction models [69, 35], whereas they are correctly described by the formalism presented in
Sec. 5.1.1 which in the present case reads :
1 2 2 2 2
γO + γO = γO , γCO = γCO , γO M↓O = γCO ↓
MCO , (18)
1 , γ2 , γ2
where, thanks to Eq. (14), γO = γCO = γw . Eqs. (18) constitute a 3 × 3 system in the unknowns γO O CO
which must also satisfy the inequalities 0 ≤ γir ≤ γw . The solution is easily obtained
↓
2 2 2
MCO 1 2
γCO = γw , γO = γCO , γO = γw − γO . (19)
M↓O
2 ≤ γ . If it is not, then it means that one of the assumptions made does
This solution is valid as long as γO w
2 was always below γ for our
not hold (e.g. equal recombination probability for all species). In practice, γO w
15 - 30 RTO-EN-AVT-142
computations. Then, the reaction frequencies [Eq. (17)] are obtained as

1 M↓
γO O 2 ↓
χ1 = , χ2 = γCO MCO , (20)
2
leading to the following expressions for the species production terms
1 M↓
γO
ωCO2 2 ↓ ωCO 2 ↓ ωO2 ωC ωO 1 2
= γCO MCO , = −γCO MCO , = O
, = 0, = −(γO + γO )M↓O . (21)
MCO2 MCO MO2 2 MC MO
(a) McKenzie bulk chemistry model. (b) Park bulk chemistry model.
Figure 13: Species mass fraction and non dimensional temperature pro les [Surface reaction model presented
in Eq. (21), hδ =15.3MJ/kg, Tw =300K], [0: wall; 1: δ].
We now move to the presentation of the solution of the stagnation line problem [59]. All the computations
showed in the following are converged for a 100 points discretization of the stagnation line. A grid resolution
study was conducted to obtain reliable heating predictions. Species mass fraction and non-dimensional temper-
ature pro les along the stagnation line are shown in Fig. 13 for 100 and 200 equidistant point grids, for both the
McKenzie [Fig. 13(a)] and Park [Fig. 13(b)] bulk chemistry models. The solutions for both mesh re nements
are seen to be identical, showing the grid independence of the results. The analysis of Fig. 13 reveals a non
negligible in uence of the bulk chemistry model. The main difference in the predictions obtained with the
two sets of nite rate chemical processes lies in the evolution of atomic carbon along the stagnation line. The
McKenzie model does not enhance considerably C recombination, while for the Park model C is involved in
the formation of CO2 starting from 80% of the stagnation line and it is completely depleted when we reach the
40% of the stagnation line. As a consequence, in Fig. 13(a) we observe that, even if CO disappears completely,
a lower amount of CO2 is present at the wall with respect to the computations obtained with the Park model.
In addition, since less carbon is available for recombination, a higher amount of O2 is observed at the wall for
the prediction obtained with the McKenzie model. The difference in the prediction of atomic carbon chemistry
will be more evident for the high enthalpy case and some of the previous observations will be helpful for the
de nition of the set of wall reactions.
We now turn our attention to the analysis of the diffusion demixing phenomenon considering three different
wall boundary conditions :
RTO-EN-AVT-142 15 - 31
Figure 14: Mass fractions and non-dimensional temperature for three different boundary conditions (Park bulk
chemistry model) [hδ =15.3MJ/kg, T w =300K].
• fully catalytic wall (present wall chemistry model - FCW);
• equilibrium wall with the element fraction of the outer edge (LEW - Xδc );
• equilibrium wall with the element fraction corresponding the fully catalytic case
(LEW - XFCc ).
The wall heat ux corresponding to these cases is summarized in Tab. 6.
Table 6: Wall heat ux for several boundary conditions (Park chemistry model) [hδ =15.3MJ/kg, Tw =300K].
Boundary Condition qw [MW/m2 ]
FCW 0.856
LEW - XFCc 0.857
LEW - Xδc 0.918
Assuming a local equilibrium wall with the outer edge elemental fraction results in an overestimation of the
wall heat ux, which is explained as follows. The wall heat ux is primarily controlled by the recombination
processes taking place at the wall. Due to diffusion demixing in the boundary layer, there is more oxygen
and less carbon at the wall then at the outer edge (see Fig. 14 which shows the mass fraction pro les for the
3 boundary conditions). As a result, fewer CO2 molecules can be formed, the excess oxygen forming O2
molecules. Hence, less heat is released than if C and O elements were in the 1/3 − 2/3 proportion.
These results point out the importance of diffusion demixing effects, in particular regarding their in uence on
wall heat ux, and clearly establishes that the local equilibrium boundary condition based on outer edge (free
stream) elemental fractions, widely used in the framework of hypersonic o ws simulations, is incorrect [61, 60].
At this point, it is interesting to comment on the relation between elemental diffusion and surface chemistry.
Indeed, some approximate models used to describe stagnation line o ws are based on the constant elemental
15 - 32 RTO-EN-AVT-142
fraction assumption as for example in the γ-code developed at IPM [36]. Under chemical non equilibrium
conditions, using the 2-surface reactions model, both O2 and CO2 should be present at the wall under fully
catalytic conditions. On the other hand, if constant elemental fraction is assumed through the boundary layer,
equal to the outer edge one (1/3-2/3), the co-presence of these two species is not possible. Indeed, as shown in
Ref. [66], the results of the γ-code operating with CO2 mixtures under non equilibrium and for a fully catalytic
surface, based on the 2-surface reactions scheme, show the presence of only CO2 on the surface for cold wall
conditions. This is clearly in contradiction with the equal effective recombination probability which is one of
the major assumption of the IPM methodology. We wish to stress that the model presented in Sec. 5.1.1 is free
of these inconsistency.
Another illustration of the importance of the demixing phenomenon is provided by comparing Scott’s and the
present formulation for a fully catalytic wall. Computed elemental fraction pro les are presented in Fig. 15,
always for the low enthalpy case. Elemental fractions of carbon (XC ) and oxygen (X O ) de ned in Eq. (5), are
related to molar elemental concentration (ξi ) as follows :
ξO ξC
XO = and XC =
ξC + ξO ξC + ξO
where
p p
ξC = xCO2 + xCO + xC and ξO = 2xCO2 + xCO + 2xO2 + xO
Ru T Ru T
Ru being the universal gas constant. From the analysis of Fig. 15(a) we observe that for Scott’s formulation
X O > X O,δ throughout the boundary layer, which is in contradiction with the elemental balance [59]. Indeed,
in general only convective ux es are signi cant in the elemental balance (the wall diffusion ux is zero and the
edge diffusion ux is negligible) so that the elemental balance [59] reduces to
ηmax
∂uδ δ
Z
F(η̂)
ξ jδ Vδ + 2 ξj dη̂ = 0. (22)
∂x η̂Rmax dη̂ ρ
0
ρ
0
Hence, the elemental concentration (ξ j ) distribution must be such that some weighted average of it equals the
outer edge concentration, and similarly for the elemental fraction (X j ), which is not veri ed when X O > X O,δ
throughout the boundary layer. In contrast, the elemental fraction distributions obtained with the present wall
chemistry model are compatible with this condition [Fig. 15(b)].
In fact, the elemental fraction distribution obtained with Scott’s wall chemistry model violates this condition
because this model does not respect elemental conservation at the wall. For the present problem, it provides a
source of O and a sink of C at the wall. This was veri ed by computing the elemental diffusion ux es at the
wall and observing that
y y y y y y y
JCO2 w J J JCO2 w JCOw JO2 w JOw
C: + COw + Cw < 0 O: 2 + +2 + > 0.
MCO2 MCO MC MCO2 MCO MO2 MO
We now analyze the calculated heat ux maps for the present o w conditions. Heat ux maps consist of curves
of wall heat ux (qw ) as a function of wall temperature between 300 K and 2100 K for various values of the
recombination probability γw . Four different maps have been obtained for the present test case, corresponding
to all possible bulk chemistry (Park/McKenzie) and wall chemistry (Scott/present) model combinations. They
RTO-EN-AVT-142 15 - 33
(a) Scott wall chemistry model. (b) Present wall chemistry model.
Figure 15: Element fractions computed with the Scott’s and the present catalycity models (Park bulk chemistry
model) [0: wall; 1: δ].
Figure 16: Heat ux maps for the lower enthalpy case [hδ =15.3MJ/kg].
are displayed in Fig. 16.

The in uence of the bulk chemistry model is clearly visible. Park’s model reaction rates being much larger
than McKenzie’s, the wall heat ux is signi cantly larger for low wall temperatures and low recombination
probabilities. The in uence of the wall reaction model is also important for large recombination probabilities.
For high values of catalycity, the heat ux is seen to be signi cantly higher with the present model, which is
consistent with the fact that, because of the existence of a sink of C atoms at the wall in Scott’s model, there
are fewer C atoms available for recombination.
15 - 34 RTO-EN-AVT-142
5.1.3 High enthalpy case
Attention is now turned to the high enthalpy case. Because of the higher enthalpy level, the mass fraction of
C atom at the outer edge is much higher, so that the third wall reaction, C + 2(O − S ) → CO2 , may become
important. For this reason, we rst analyze the in uence of the wall reactions set using Scott’s formulation
for both bulk chemistry models. Species mass fractions and non-dimensional temperature pro les are shown
in Fig. 17 for a fully catalytic wall. Whereas for Park’s model there is no in uence of the wall reactions set
because the carbon atom entirely recombines inside the boundary layer (bulk chemistry), on the contrary the
wall reactions set has a large in uence when using McKenzie’s model, for which bulk reactions are much
slower.
Figure 17: Mass fractions and non-dimensional temperature pro le [hδ =39MJ/kg, Tw =300K][0: wall; 1: δ].
In particular, if the third wall reaction is ignored, there remains an important amount of atomic carbon at the
wall. As a result, the wall heat ux is considerably smaller (see Tab. 7 in which heat ux es are listed for all
bulk chemistry/wall reaction set combinations).
Table 7: Wall heat ux for several sets of surface reactions [hδ =39MJ/kg, Tw =300K].
Surface Stagnation Line qw [MW/m2 ]
Scott 2WR McKenzie 1.20
Scott 3WR McKenzie 2.14
Scott 3WR Park 2.23
Scott 2WR Park 2.24
The present formulation for the 3 wall reactions model reads as follows
3 M↓
γO
1 2 3 2
γO + γO + γO = γO , γCO = γCO , γC3 = γC , 2
γO 2
M↓O = γCO ↓
MCO , γC3 MC↓ = O
, (23)
2
RTO-EN-AVT-142 15 - 35
where γCO = γO = γC = γw . Eqs. (23) represent a system of 5 equations in 5 unknowns γO 1 , γ2 , γ3 , γ2 , and

O O CO
γC3 . For the present o w conditions, numerical experiments have shown that the solution does not satisfy the
constraint γO3 ≤ γ , which implies that the hypothesis of equal recombination probabilities for all species does
w
not hold [71]. The analysis of a suitable assumption is however left as an open subject for a further study.
We therefore switch our attention to the Park bulk chemistry/2 wall reaction model for the present high enthalpy
case.
(a) Mass fractions and non-dimensional temperature (b) Heat ux map.

pro le. (Tw =300K) (0: wall; 1:δ).
Figure 18: Stagnation line solution for the higher enthalpy case (Park bulk chemistry model) [hδ =39MJ/kg].
Species mass fraction and non-dimensional temperature pro les for a fully catalytic wall computed using the
present wall chemistry formulation are shown in Fig. 18(a). This is to be compared with Fig. 17(b) where
Scott’s formulation was used. The same conclusion as for the low enthalpy case is obtained, i.e. for high γw
Scott’s model acts as a sink of carbon atoms at the wall. As a result, the wall heat ux is smaller, as seen in
Fig. 18(b) which shows the heat ux map for the high enthalpy case.
5.1.4 In uence of outer edge elemental fractions on heat ux maps
As discussed in the introduction, the methodology developed at the IPM and currently used at VKI to estimate
the catalycity properties of TPS materials, is based upon Navier-Stokes simulations of an ICP facility supposed
to be under LTE conditions determined assuming constant elemental fractions.
The purpose of the Navier-Stokes plasma o w simulation is to provide the LTE outer edge inputs for the nite
thickness non equilibrium stagnation line o w computations, carried out to compute a heat ux map. Among
the quantities needed, the composition is implicitly given as a function of the outer edge pressure and rebuilt-
enthalpy, assuming LTE and using the torch inlet elemental fractions. In general, thanks to elements diffusion,
the elemental fractions vary within the ICP facility [76], as shown in Sec. 4, leading to an outer edge elemental
composition different from the torch inlet one. If this difference has an in uence on the heat ux map, by shift-
ing or stretching the iso-catalycity lines, the same in uence will be re ected to the estimation of TPS material
catalycity.
15 - 36 RTO-EN-AVT-142
(a) Heat ux map : CO2 mixture. (b) Heat ux map: air mixture.
(c) Species mass fraction : CO2 mixture (T w = 300K, γw = (d) Species mass fraction : air mixture (T w = 300K, γw =
1) (0:w;1:δ) 1) (0:w;1:δ)
Figure 19: In uence of outer edge elemental fraction on the stagnation line solution [The outer edge conditions
are speci ed in Ref. [58]].
One way to easily estimate this in uence is to use the results of the constant elemental fraction ICP simulation
and vary only the outer edge elemental fractions, used to determine the outer edge composition for the given
enthalpy and pressure. In Sec. 4 we found an oxygen elemental fraction lower than the inlet value (XδO = 0.21)
in the jet near the axis for air plasmas. On the other hand, for carbon dioxide o ws the opposite behavior was
observed, i.e. the oxygen elemental fraction reached values higher than at the torch inlet (XδO = 2/3). Following
this result, we have perturbed the oxygen outer edge elemental fraction by +10% and −10% of the reference
inlet value, for carbon dioxide and air plasmas respectively.
The results of this analysis are presented in Figs. 19(a) and 19(b), where are shown the heat ux maps corre-
sponding to the conditions de ned in Ref. [58], for both air and carbon dioxide. The results show an important
in uence on the stagnation point heat ux for carbon dioxide mixtures. The reason of these differences can be
explained from the analysis of species concentration showed in Fig. 19(c) for a fully catalytic cold wall. Indeed,
the increase of oxygen outer edge fraction reduces the amount of carbon available for CO2 recombination at the
RTO-EN-AVT-142 15 - 37
surface, this leading to a lower heat ux. This is more evident for high values of recombination probabilities,
where almost the same effect on wall heat ux is observed in the whole temperature range considered. As the
recombination probability is reduced the in uence of the outer edge elemental fraction decreases as well, and a
higher in uence is observed for cold wall conditions, where recombination is enhanced by the bulk chemistry
owing to the low temperature.
As far as air mixtures are concerned, the effects on wall heat ux are essentially negligible. Indeed the lines of
Fig. 19(b) are superposed for all values of catalycity in the temperature range investigated. The quite low in u-
ence on heat ux can be explained by the analysis of the species mass fraction pro les of Fig. 19(d). Indeed the
perturbation of the elements fraction shifts slightly the concentration of N2 , N, and O at the outer edge without
altering too much the stagnation line chemistry.
The analysis carried out in this section shows therefore that, especially for Mars entry applications, the outer
edge elemental fractions play an important role in the determination of the heat ux map and therefore in the
extrapolation of the TPS material catalycity.
6 Conclusions
In this section the main achievements of the present research are recalled following the structure of this
manuscript with the aim to summarize the contributions brought to the modeling of high enthalpy gases in
the framework of a macroscopic methodology used ti estimate the catalytic properties of thermal protection
materials.
6.1 Theory of reacting mixtures under Local Thermodynamic Equilibrium
A series of new elemental transport coefficients have been presented for mixtures under thermochemical equi-
librium conditions. They allow to reduce the equations of chemically reacting o ws in LTE to an elegant
system consisting of the conventional Navier-Stokes equations (mass, momentum, energy) complemented by
an advection-diffusion equation for the mass fraction of each chemical element in the mixture. The obtained
formalism is in closed form in the sense that diffusive ux es are directly expressed in terms of gradients of
the solution unknowns, unlike other formulations in which these ux es are obtained in an implicit manner, by
solving the full system of Stefan-Maxwell equations. In the case of mixtures containing neutral components,
assuming local neutrality and absence of current, we have introduced the following new transport properties :
the elemental multicomponent (Deq ) and thermal demixing (DTe ) diffusion coefficients, the thermal demixing
conductivity (λD ), the elemental heat transfer coefficients (λeEL ). We refer the interested reader to Ref. [57] for
further details about the extension of this formalism composed of both neutral ad charged components.
15 - 38 RTO-EN-AVT-142
6.2 Elemental diffusion along the stagnation line
To verify the importance of elemental diffusion phenomena and to assess the performances of the theory pre-
sented in Sec. 2, we focused on a 8-species CO2 /N2 mixture containing three elements (C, N, O). This analysis
was based on the use of the transport properties previously presented, which required the computation of nine
elemental multicomponent and three thermal demixing coefficients, the demixing thermal conductivity and
three elemental heat transfer coefficients, in addition to the ‘standard’ LTE transport properties.
The results of this study represent the rst attempt to compute thermal demixing and multicomponent diffusion
coefficients as well as elemental heat transfer coefficients and thermal demixing conductivity for a mixture of
heteronuclear components containing three elements. This proves the generality of the proposed formulation
and clearly represents an improvement with respect to previous formulations available in the literature. In ad-
dition, the results obtained with the proposed closed form of the equations coincide with those obtained using
an implicit formulation, indicating the correctness of the new formulation.
Finally, after a critical analysis of the results, we noticed how complex elemental concentration and heat ux
patterns can be easily explained by examining graphs of LTE transport coefficients that arise from the new
proposed LTE formulation.
6.3 Elemental diffusion in inductive coupled plasmas
In Sec. 4, we extended our analysis of elemental diffusion considering the investigation of air and carbon
dioxide inductively coupled plasma o ws for application to Earth and Mars entry. To verify the validity of
one of the major hypothesis of the IPM methodology, i.e. the existence of equilibrium conditions within ICP
facilities, we considered both the equilibrium and the non equilibrium form of the governing equations. The
numerical solution of these equations led to a detailed analysis of the o w behavior within the torch for air and
CO2 mixtures and then of the jet o wing into the test chamber for air mixtures.
Firstly, we presented original carbon dioxide plasma o w simulations under equilibrium considering both
constant and variable elemental fraction conditions. A critical analysis of these results shows that for an inlet
carbon volumetric fraction of ∼ 33% :
• the solution, under chemical equilibrium, of an additional set of elemental continuity equations for oxy-
gen and carbon shows the presence of important variations in the elemental composition.
• The amount of demixing observed for the operating conditions considered is characterized by a maximum
increase to ∼ 38% and a decrease to ∼ 28% of the carbon elemental fraction.
• The effects of these variations in the elemental fraction on the species evolution is much higher than for
air o ws. Indeed, at the outlet of the torch, important differences are observed between the species radial
pro les obtained under equilibrium with constant and variable elemental fractions.
• The in uence of demixing on the temperature prediction is much higher for CO2 o ws than for air. At the
torch exit, neglecting elemental diffusion a maximum relative error of ∼ 40% is observed for r ∼5.8 cm,
while on the axis it decreases sharply to ∼ 2%.
RTO-EN-AVT-142 15 - 39
Secondly, we turned our attention to the analysis of air plasma o ws both in the torch and in the chamber
of the VKI-Plasmatron. Two operating pressures have been investigated corresponding to 5 kPa and 30 kPa.
In addition, the three chemical regimes used to analyze air plasmas within the torch have been considered
and for the description of chemical non equilibrium, two nite rate chemistry models have been selected as
representative of Earth atmosphere chemistry.
The analysis of these results leads us to the following conclusions :
• for the low pressure case (5 kPa), we observe a certain in uence of the nite rate chemistry model on the
prediction of the o w behavior within the torch, while, as we move towards the test chamber, the o w
predictions rapidly match leading to almost negligible discrepancies ∼ 30 cm after the torch.
• Non equilibrium effects are observed within the torch for the 5 kPa operating pressure. In the jet, the non
equilibrium results are rather close to the equilibrium ones, provided that elemental demixing is taken
into account.
• The consequence of increasing the operating pressure is to sharply decrease the differences between the
equilibrium and non equilibrium predictions. These differences are very small for the low pressure case
(5 kPa) and they become negligible for the high pressure case (30 kPa), especially in the jet.
The major result of this analysis consists in the fact that, for the purpose of TPM testing, the costly sim-
ulation of chemical non equilibrium ICP o ws, both in the torch and in the test chamber, can be avoided for
sufficiently high pressures. Indeed the same accuracy can be reached relying on the numerical simulation of a
more compact thermochemical equilibrium formalism based on the solution of an additional set of elemental
advection diffusion equations, free of stiff source terms. In addition, because of the important elemental frac-
tion variations observed as a result of both air and carbon dioxide plasma o w simulation, it appears desirable
to add the elemental fractions at the stagnation line outer edge to the list of non dimensional parameters de ned
in the framework of the IPM methodology.
6.4 Practical application to TPM testing
Following the results obtained in the previous sections, we focused our interest on the analysis of stagnation
line o ws for application to TPS testing. To this end, both air and carbon dioxide mixtures have been analyzed
considering only neutral components.
In our simulations, two models (Park [52] and McKenzie [44]) for the description of nite rate chemistry for
a carbon dioxide mixtures of neutral species have been tested and important differences in the results have been
observed. Then, our attention has been attracted by the heterogeneous recombination processes taking place
on the surface of TPM samples. During our investigation, some unexpected a ws of the Scott [69] formulation
have been discovered. To overcome them, an alternative model has been proposed and successfully tested.
The main outcome of the analysis of the results obtained with this new gas/surface interaction model reads as
follows :
15 - 40 RTO-EN-AVT-142
• a deeper description of the theoretical relation between a fully catalytic surface and the related equilib-
rium conditions is presented;
• we have shown that the constant elemental fraction assumption may contradict the hypothesis of equal
effective recombination probability;
• we have shown that a non equilibrium model of a carbon dioxide stagnation line o w based on the
constant elemental fraction assumption is in contrast with the equal recombination probability for CO
and O involved in the formation of CO2 and O2 at the wall;
• we have shown that the equal recombination probability hypothesis may not be compatible with the local
wall conditions;
• we have veri ed the in uence of the outer edge elemental fractions on the stagnation point heat ux. This
in uence, rather important for carbon dioxide mixtures, con rms the necessity to add the them to the list
of parameters of the methodology developed at the IPM suited for the estimation of catalytic properties
of TPS materials, especially for Mars entry applications.
References
[1] J. D. Anderson. Hypersonic and high temperature gas dynamics. McGraw-Hill, New York, 1989.
[2] C. O. Asma and O. Chazot. Trajectory simulation and catalycity determination for expert vehicle. In Fifth
European Symposium on Aerothermodynamics for Space Vehicles, Cologne (Germany), 8-11 November
2004.
[3] P. Barbante, O. Chazot, and G. Degrez. Flight extrapolation of wind tunnel data for the determination of
TPS materials catalycity. In 2nd International Symposium on Atmospheric Reentry Vehicles and Systems,
Arcachon, France, March 26-29 2001.
[4] P. F. Barbante. Accurate and Efficient Modeling of High Temperature Nonequilibrium Air Flows. PhD
thesis, von Karman Institute, Rhode-Saint-Genèse, Belgium, 2001.
[5] B. Bottin, M. Carbonaro, O. Chazot, G. Degrez, D. Vanden Abeele, P. F. Barbante, S. Paris, V. Vander-
haegen, T. Magin, and M. Playez. The plasmatron and beyond: a decade of aerothermal plasma testing
at the von karman institute. In XXVI International Conference on Phenomena in Ionized Gases (ICPIG
2003), Greifswald, Germany, July 15- 20 2003.
[6] B. Bottin, O. Chazot, M. Carbonaro, V. Van der Haegen, and S. Paris. The VKI plasmatron characteristics
and performance. In J. M. Charbonnier and G. S. R. Sarma, editors, Measurement Techniques for High
Temperature and Plasma Flows, von Karman Institute for Fluid Dynamics, 1999. NATO-RTO-EN 8.
[7] R. S. Brokaw. Thermal conductivity of gas mixtures in chemical equilibrium. II. Journal of Chemical
Physics, 32(4):1005, 1960.
RTO-EN-AVT-142 15 - 41
[8] D. Bruno, M. Capitelli, E. Kustova, and E. Nagnibeda. Non-equilibrium vibrational distributions and
transport coefficients of N2(v)-N mixtures. Chemical Physics Letters, 308(5-6):463–472, July 1999.
[9] J. N. Butler and R. S. Brokaw. Thermal conductivity of gas mixtures in chemical equilibrium. Journal of
Chemical Physics, 26(6):1636, 1957.
[10] M. Capitelli, G. Colonna, D. Giordano, E. Kustova, E. Nagnibeda, M. Tuttafesta, and D. Bruno. Mathe-
matical Modelling, 11(2):45, 1999.
[11] J. M. Charbonnier, J. Couzi, W. Dieudonne, and J. L. Verant. Workshop 2003 ” radiation of high temper-
ature gas”, TC3 : De nition of an axially symmetric testcase for high temperature gas radiationprediction
in Mars atmosphere entry. Technical report, CNES, Centre de Toulouse, Toulouse, France, May 2003.
NG104-07-TF-001-CNES.
[12] O. Chazot, S. Paris, P. Collin, M. Bickel, and T. Ullman. TPS testing and catalycity determination in the
VKI plasmatron facility. In 3rd International Symposium on Atmospheric Reentry Vehicles and Systems,
Arcachon, France, March 2003.
[13] O. Chazot, J. M. Pereira Gomes, and M. Carbonaro. Characterization of a ’mini-plasmatron’ facility by

pitot probe measurements. In 29th AIAA Plasmadynamics and Lasers Conference, Albuquerque, NM,
June, 15-18 1998. AIAA-1998-2478.
[14] G. Colonna, M. Tuttafesta, M. Capitelli, and D. Giordano. In uence of dissociation rates on the state-
to-state vibrational kinetics in nozzle expansions. In CEPADUES, editor, Rarefied Gas Dynamics 21,
volume 2, page 281, Toulouse, France, 1999.
[15] Marta de la Llave PLATA, O. Chazot, T. Magin, G. Degrez, and A. F. Kolesnikov. Application and anal-
ysis of a methodology for the determination of TPS catalytic properties. In 2nd International Symposium
on Atmospheric Reentry Vehicles and Systems, Arcachon, France, March 26-29 2001.
[16] G. Degrez, P. Barbante, M. de la Llave, T. Magin, and O. Chazot. Determination of the catalytic properties
of TPS materials in the VKI ICP facilities. In ECCOMAS Computational Fluid Dynamics Conference
2001, pages 162–167, Swansea, Wales, UK, 4-7 September 2001.
[17] G. Degrez, D. Vanden Abeele, P. Barbante, and B. Bottin. Numerical simulation of inductively coupled
plasma o ws under chemical non equilibrium. International Journal of Numerical Methods in Heat and
Fluid Flow, 14(4):538–558, 2004.
[18] A. Ern and V. Giovangigli. The kinetic chemical equlibrium regime. Physica A, 260:49–72, 1998.
[19] P. Rini et al. Assessment of thermodynamics, transport, and chemical properties for CO2 o ws. Contract
report, von Karman Institute for Fluid Dynamics, Rhode-Saint-Genèse, Belgium, August 2002. MSR02-
TNP4.1-VKI.
[20] P. Rini et al. Evaluation of numerical codes for high enthalpy CO2 o ws. Technical report, von Karman
Institute for Fluid Dynamics, Rhode-Saint-Genèse, Belgium, August 2002. MSR02-TNP6.1-VKI.
[21] J. A. Fay and F. R. Riddell. Theory of stagnation point heat transfer in dissociated air. Journal of the
Aeronautical Science, 1958.
15 - 42 RTO-EN-AVT-142
[22] J. H. Ferziger and H. G. Kaper. Mathematical Theory of Transport Processes in Gases. North-Holland,
Amsterdam, 1972.
[23] W. Frie. Thermodynamics of irreversible processes and the force equations of schluter. Zeitschrift für
Physik, 162:61–68, 1961.
[24] W. Frie and H. Maecker. Mass separation by diffusion of reacting gases. Zeitschrift für Physik, 162:69–83,
1961.
[25] W. Frie and H. Maecker. Dynamics of reacting gases. Zeitschrift für Physik, 168:206–232, 1962.
[26] W. Frie and H. Maecker. Demixing effects in mixtures of ionizing atomic gases. Zeitschrift für Physik,
172:99–117, 1963.
[27] A. Garcia, O. Chazot, and D. Fletcher. Investigation in plasmatron facilities on catalycity determination.
In 4th European symposium for space vehicles, Capua, Italy, October 16-18 2001.
[28] A. M. Garcia. Catalycity effects in plasmatron tests. Technical report, von Karman Institute for Fluid
Dynamics, Rhode-Saint-Genèse, Belgium, June 2001. VKI PR 2001-12.
[29] P. A. Gnoffo, R. N. Gupta, and J. L. Shinn. Conservation equations and physical models for hypersonic air
o ws in thermal and chemical nonequilibrium. Technical Paper 2867, NASA (Langley), Virginia, 1989.
[30] A. Groß. Theoretical Surface Science - A microscopic perspective. Springer Verlag, 2002.
[31] A. Groß. Introduction to theoretical surface science. In Experiment, Modelings and Simulation of gas-
surface interactions for reactive flows in hypersonic flights, Rhode Saint Genèse, February 2006. von
Karman Institute for Fluid Dynamics. VKI-LS.
[32] J. O. Hirschfelder, C. F. Curtiss, and R. B. Bird. Molecular Theory of Gases and Liquids. Wiley, New
York, 1964.
[33] IPM. Surface catalysis determination for earth and mars atmospheres re-entry vehicles: Microscopic vs
macroscopic methods. Technical report, INTAS/CNES, 2005. Project 03-53-5117.
[34] A. F. Kolesnikov. Combined measurements and computations of high enthalpy and plasma o ws for
determination TPM surface catalycity. In J. M. Charbonnier and G. S. R. Sarma, editors, Measurements
Techniques for High Enthalpy and Plasma Flows, pages 8A 1–16, Rhode-St-Genèse Belgium, Oct. 1999.
NATO-RTO-EN.
[35] A. F. Kolesnikov. Extrapolation from high enthalpy tests to ight based on the concept of local heat
transfer simulation. In J. M. Charbonnier and G. S. R. Sarma, editors, Measurements Techniques for High
Enthalpy and Plasma Flows, pages 8A 1–14, Rhode-St-Genèse Belgium, Oct. 1999. NATO-RTO-EN.
[36] A. F. Kolesnikov. Calculation of heat transfer and effective atoms recombination probability γw in disso-
ciated air jet of the induction plasmatron.(user manual). Technical report, Research Institute of Mechanics
at Moscow State University, Moscow, March 2000.
[37] A. F. Kolesnikov. An efficient modeling stagnation point heat transfer for subsonic aerothermal test
conditions. In Technical note 200, Rhode-St-Genèse Belgium, January 2001. VKI-TN-200.
RTO-EN-AVT-142 15 - 43
[38] A. F. Kolesnikov and L. Marraffa. An analysis of stagnation point thermochemical simulation by plas-
matron for Mars probe. In 33rd Thermophysics Conference, Norfolk, VA, June-July 1999. AIAA Paper
99-3564.
[39] A. F. Kolesnikov, I. S. Pershin, S. A. Vasil’evskii, and M. I. Yakushin. Study of quartz surface catalycity
in dissociated carbon dioxide subsonic o ws. Journal of Spacecraft and Rockets, 37:573–579, September-
October 2000.
[40] A. F. Kolesnikov, M. I. Yakushin, I. S. Pershin, S. A. Vasil’evskii, O. Chazot, B. Vancrayenest, and J.

Muylaert. Comparative study of surface catalycity under subsonic air test conditions. In 4th European
symposium for space vehicles, Capua, Italy, October 16-18 2001.
[41] V. L. Kovalev and O. N. Suslov. Diffusion separation of chemical elements on a catalytic surface. Fluid
Dynamics, 23(4):579–585, 1988.
[42] E. Kustova and E. Nagnibeda. State-to-state approach in the transport kinetic theory. In CEPADUES,
editor, Rarefied Gas Dynamics 21, volume 1, page 231, Toulouse, France, 1999.
[43] T. Magin. A model for Inductive Plasma Wind Tunnels. PhD thesis, von Karman Institute for Fluid
Dynamics, St.-Genesius-Rode, Belgium, 2004.
[44] R. L. McKenzie. An estimate of the chemical kinetics behind normal shock waves in mixtures of carbon
dioxide. NASA TN D-3287, Ames Research Center, Moffett Field, CA, 1966.
[45] A. B. Murphy. Combined diffusion coefficients in equilibrium mixtures of dissociating gases. Journal of
Chemical Physics, 99(2):1340, 1993.
[46] A. B. Murphy. Diffusion in equilibrium mixtures of ionized gases. Physical Review E (Statistical, Non-
linear, and Soft Matter Physics), 48(5):3594, 1993.
[47] A. B. Murphy. Demixing due to frictional forces in an electric arc. Physical Review Letters, 73(13):1797,
1994.
[48] A. B. Murphy. Demixing in free-burning arcs. Physical Review E (Statistical, Nonlinear, and Soft Matter
Physics), 55:7473, 1997.
[49] E. Nagnibeda. The structure of the relaxation zone behind shock waves in the reacting gas o ws. In
J. Hunt, editor, Aerothermodynamics for Space Vehicles, Noordwijk, The Netherlands, 1995. ESTEC,
ESA Publication Division.
[50] E. Nagnibeda. Kinetic theory of transport processes and thermodynamics, polyatomic molecules. In
Fletcher et al., editor, Physico-chemical models for high enthalpy and plasma flows, Rhode-Saint-Genése,
Belgium, 2002. von Karman Institute for Fluid Dynamics. VKI LS 2002-07.
[51] M. Panesi. Numerical simulation of non equilibrium plasma o ws. Technical report, von Karman Institute
for Fluid Dynamics, Rhode-Saint-Genèse, Belgium, June 2005. VKI PR 2005-23.
[52] C. Park, J. T. Howe, and R. L. Jaffe. Review of chemical-kinetic problems of future nasa mission, II: Mars
entries. Journal of Thermophysics and Heat Transfer, 8(1):9–23, 1994.
15 - 44 RTO-EN-AVT-142
[53] C. Park, R. L. Jaffe, and H. Partridge. Chemical-kinetic parameters of hyperbolic earth entry. Journal of
Thermophysics and Heat Transfer, 15(1):76–90, 2001.
[54] M. Balat Pichelin et al. Molecular dynamic simulation and experimental determination of the recombi-
nation of atomic oxygen on silica polymorphs at 1000 K. In 4th International Symposium Atmospheric
Reentry Vehicles and Systems, Arcachon (France), 21-23 March 2005.
[55] J. Richter. Über diffusionsvorgaenge in lichtboegen. Zeitschrift für Astrophysik, 53:262–272, 1961.
[56] P. Rini. Numerical simulation of equilibrium and non equilibrium CO2 plasma o ws. Technical report,
von Karman Institute for Fluid Dynamics, Rhode-Saint-Genèse, Belgium, June 2002. VKI PR 2002-22.
[57] P. Rini. Analysis of differential diffusion phenomena in high enthalpy flows, with application to thermal
protection material testing in ICP facilities. PhD thesis, von Karman Institute for Fluid Dynamics, St.-
Genesius-Rode, Belgium, 2006.
[58] P. Rini and G. Degrez. Elemental demixing along both air and carbon dioxide stagnation line o ws.
Journal of Thermophysics and Heat Transfer, 18(4):511–518, 2004.
[59] P. Rini, A. Garcia, T. Magin, and G. Degrez. Numerical simulation of CO2 stagnation line o ws with
catalyzed surface reactions. Journal Thermophysics and Heat Transfer, 18(1):114–121, 2004.
[60] P. Rini, T. Magin, G. Degrez, and D. Fletcher. Numerical Simulation of Non Equilibrium Hypersonic
CO2 o ws for mars entry application. In 3rd International Symposium Atmospheric Reentry Vehicles and
Systems, Arcachon, France, March 2003.
[61] P. Rini, N. Nannipieri, O. Chazot, and G. Degrez. Numerical simulation of thermochemical equilibrium
boundary layers. In 6th National Congress on Theoretical and Applied Mechanics, Gent, Belgium, 26-27
May 2003.
[62] P. Rini, D. Vanden Abeele, and G. Degrez. Closed form for the equations of chemically reacting o ws un-
der local thermodynamic equilibrium. Physical Review E (Statistical, Nonlinear, and Soft Matter Physics),
72(1):011204, 2005.
[63] P. Rini, D. Vanden Abeele, and G. Degrez. Analysis of diffusion phenomena in lte mixtures with applica-
tions to CO2 /N2 mixtures. to be submitted to the Journal of Thermophysics and Heat Transfer, 2006.
[64] P. Rini, D. Vanden Abeele, and G. Degrez. Elemental demixing in inductively coupled air plasmas torches
at high pressures. Journal of Thermophysics and Heat Transfer, 20(1):31–40, 2006.
[65] P. Rini, S. A . Vasil’evskii, A. F. Kolesnikov, O. Chazot, and G. Degrez. Inductively coupled co2 plasma
o ws: a code-to-code comparison. In 4th International Symposium Atmospheric Reentry Vehicles and
Systems, Arcachon (France), 21-23 March 2005.
[66] P. Rini, S. A. Vasil’evskii, A. F. Kolesnikov, O. Chazot, and G. Degrez. CO2 stagnation line o w sim-
ulation for mars entry applications. In 38th AIAA Thermophysics Conference, Toronto (CA), 6-8 Jun
2005.
[67] G. S. R. Sarma. Physico-chemical modeling in hypersonic o w simulation. Progress in Aerospace Sci-
ences, 36(3-4):281–349, 2000.
RTO-EN-AVT-142 15 - 45
[68] S.A.Vasil’evskii. Calculation of o w and heat transfer at blunt body stagnation line with account for the
diffusion of elements and higher approximations for transport coefficients. In Investigations of hyper-
sonic aerodynamics and heat transfer with account for non equilibrium chemical reactions, pages 30–45,
Moscow, 1987. Moscow State University publishing.
[69] D. C. Scott. Catalytic boundary conditions in nonequilibrium o w. Proceeding of the IUTAM Symposium
Marseille, pages 298–305, 1992.
[70] O. N. Suslov, G. A. Tirsky, and V. V. Shchennikov. Flows of Multicomponent Ionized Mixtures in

Chemical Equilibrium. Description within the Framework of the Navier-Stokes and Prandtl Equations. In
Prikl. Mekh. and Tekhn. Fiz., 1971.
[71] J. Thömel. Local heat transfer simulation of air and CO2 . Technical report, von Karman Institute for Fluid
Dynamics, Rhode-Saint-Genèse, Belgium, June 2005. VKI PR 2005-01.
[72] G. A. Tirsky. Up-to-date gasdynamic models of hypersonic aerodynamics and heat transfer with real gas
properties. Annual Review of Fluid Mechanics, 25:151–181, 1993.
[73] G. A. Tirsky. The hydrodynamic equations for chemically equilibrium o ws of a multielement plasma
with exact transport coefficients. Journal of Applied Mathematics and Mechanics, 63(6):841–861, 1999.
[74] H. W. P. van der Heijden. Modeling of radiative transfer in light sources. PhD thesis, Technische Univer-
siteit Eindhoven, Eindhoven, the Netherlands, January 2003.
[75] D. Vanden Abeele. An Efficient Computational Model for Inductively Coupled Air Plasma Flows un-
der Thermal and Chemical Non-Equilibrium. PhD thesis, von Karman Institute, Rhode-Saint-Genèse,
Belgium, 2000.
[76] D. Vanden Abeele and G. Degrez. Efficient computational model for inductive plasma o ws. AIAA
Journal, 38(2):234–242, 2000.
[77] S. A. Vasil’evskii, A. F. Kolesnikov, and M. I. Yakushin. Mathematical models for plasma and gas o ws
in induction plasmatrons. In Molecular Physics and Hypersonic Flows (Ed. M. Capitelli), pages 495–504,
Kluwer, Dordrecht, the Netherlands, 1996.
[78] S.A. Vasil’evskii and G.A. Tirsky. In uence of multicomponent diffusion and higher approximations for
the transport coefficients on the heat ux to blunt body at hypersonic ight. In Applied aerodynamics of
vehicles, pages 100–103, Kiev, 1984. Naukova dumka publishing.
[79] W. G. Vincenti and C. H. Kruger. Introduction to Physical Gas Dynamics. John Wiley and Sons, New
York, 1965.
15 - 46 RTO-EN-AVT-142
NUMERICAL IMPLEMENTATION OF SURFACE CATALYSIS,
REACTION, AND SUBLIMATION
Chul Park
Department of Aerospace Engineering, Korea Advanced Institute of Science and Technology
373-1 Guseong-dong, Yuseong-gu, Daejeon, 305-701 SOUTH KOREA
cpark216@kaist.ac.kr
This lecture consists of three parts: 1) quantification of rates of gas-surface interaction, 2) formulation of
gas-surface balance conditions, and 3) survey of the gas-surface interaction problems in the entry flights of
various planets. The first part reviews the role of surface rates and flow parameters in Goulard’s theory,
the catalytic rates for Martian entry problem, the reaction rates for oxidation and nitridation of carbon,
rough surfaces, and the relationship between forward and reverse surface rates. The second part covers the
method of expressing the diffusion flux at wall, derivation of the mass balance condition at wall, and its
implementation in the boundary layer and computational-fluid-dynamics formulations. In the third part, a
brief survey is made of the gas-surface interaction problems occurring in entry flights into Earth, Mars,
Venus, Titan, and outer planets.
1.0 NOMENCLATURE
B : Rotational constant, cm−1 .
c : Speed of sound, m/s.
(c) : Crystalline (solid) state.
C : Average molecular speed ( 8kT/πm), m/s.
Cp : Specific heat at constant pressure, J/(kg-K).
Cv : Specific heat at constant volume, J/(kg-K).
D : Effective diffusion coefficient.
Da : Adsorption energy, J/mol.
Db : Diffusion barrier energy, J/mol.
Dg : Dissociation energy in gas phase, J/mol.
Dis : Multicomponent diffusion coefficient between species i and s, m2 /sec.
f : Stream function, Eq. (32).
Fs : A bifurcation parameter, Eq. (44), dimensionless.
g : Normalized energy, Eq. (36).
H : Enthalpy, J/kg.
∆H : Energy produced by reaction, ev or J/mol.
Js : Rate of mass diffusion of species s, kg/(m2 -sec).
k : Boltzmann constant, J/K.
kw : Catalytic velocity, Eq. (5), m/s.
m : Mass of one unspecified particle, kg.
m s : Mass of one particle of species s, kg.
ṁp : Pyrolysis gas mass flow rate, kg/(m2 -sec).
ṁp,s : Mass flow rate of species s in pyrolysis gas, kg/(m2 -sec).
ṁt,s : Mass flow rate of species s resulting from reactions and sublimation, kg/(m2 -sec).
ṁt : Mass flow rate of material removed by surface reaction and sublimation, kg/(m2 -sec).
ṁv,s : Mass flow rate of species s in sublimation product, kg/(m2 -sec).
M s : Molecular weight of species s, kg/mol.
M̄ : Average molecular weight, kg/mol.
M : V∞ /c , dimensionless.
Number density of adsorbed molecules, m−2 .
w w
n :
Number density of gas molecules, m−3 .
a
ng :
N : Number of species.
p : Pressure, Pascal.
Pr : Prandtl number, dimensionless.
q : Heat transfer rate, W/m2 .
Park, C. (2007) Numerical Implementation of Surface Catalysis, Reaction, and Sublimation. In Experiment, Modeling and Simulation of
Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 16-1 – 16-20). Educational Notes RTO-EN-AVT-142, Paper 16.
RTO-EN-AVT-142 16 - 1
Numerical Implementation of Surface Catalysis, Reaction, and Sublimation
Q : Partition function.
r : Radial distance from axis of symmetry, m.
R : Nose radius, m.
Re∞ : Freestream Reynolds number ρ∞ V∞ R/µ∞ .
S (c) : Unspecified solid surface.
Sc : Schmidt number, dimensionless.
T : Temperature, K.
u : Tangential velocity, m/s.
v : Normal velocity, m/s.
V∞ : Freestream velocity, m/s.
x : Distance along wall measured from the stagnation point, m.
W s : Rate of production of species s, sec−1 .
X s : Molar fraction of species s.
y : Distance normal to wall measured from wall, m.
Z s : Virtual species mass fraction in bifurcation model, Eq. (47).
α s : Mass fraction of species s, dimensionless.
γ : Cp/Cv .
φ : Roughness ratio (BET area/projected area), dimensionless.
η : Boundary layer coordinate normal to wall, Eq. (31b), dimensionless.
η a : Adsorption (sticking) coefficient, dimensionless.
η c : Catalytic recombination coefficient, dimensionless.
η r : Reaction coefficient, dimensionless.
η v : Evaporation coefficient, dimensionless.
µ : Viscosity, m2 /sec.
µ1 : A bifurcation parameter, Eq. (48a), dimensionless.
µ2 : A bifurcation parameter, Eq. (48b), dimensionless.
ρ : Density, kg/m3 .
ξ : Boundary layer coordinate along wall, Eq. (31a), dimensionless.
1.1. SUPERSCRIPT
: ∂/∂η.
1.2. SUBSCRIPTS
∞ : Freestream.
a : Adsorbed state.
e : Edge of boundary layer.
E : Equilibrium.
,f : Forward reaction.
g : Gas.
,r : Reverse reaction.
R : Rotational mode.
T : Translational mode.
v : Evaporation.
V : Vibrational mode.
w : Wall.
2.0 PHYSICS OF GAS-SURFACE INTERACTIONS
2.1. GOULARD’S WORK ON CATALYTIC EFFECTS
The problem of chemical reactions at a solid surface was made aware first by Goulard.1 He
pointed out that, in an environment where chemical reactions are frozen in the boundary layer, heat
transfer rate q to the wall can be reduced by making it noncatalytic to recombination. In most
practical flight regimes of hypersonic vehicles, boundary layer flows are indeed chemically frozen.
The flight experiments conducted with the Space Shuttle proved Goulard’s thesis.
Goulard considered the catalytic recombination process typified by ( 1 ev = 96.48 J/mol)
O + O + S (c) → O2 + S (c), ∆H = 5.08 ev = 490 kJ/mol (1)
16 - 2 RTO-EN-AVT-142
in the stagnation region shown schematically in Fig. 1. Here, S(c) signifies that the phenomenon
occurs at the surface of an unspecified material (c here signifies crystalline state). His work introduced
three important axioms. The first axiom is that the ratio of heat transfer rates between a fully
catalytic and noncatalytic surfaces is approximately the ratio of the total, i.e., translational (CpT )
plus chemical, enthalpy to the translation-only enthalpies.
q to fully catalytic wall ≈ CpT + energy contained in dissociation (2)
q to noncatalytic wall CpT
V
y,v,h Edge of
boundary layer
Shock wave
x,u,x
r
Wall
R
Figure 1. Flowfield considered by Goulard.1
The second axiom is that the heat transfer rate to the fully catalytic wall is approximately the
same as that in a nonreacting gas:
q to a fully catalytic wall ≈ q in a nonreacting gas. (3)
The third axiom is that the transition from the noncatalytic to catalytic surface occurs at around
the point where the quantity
ζ = Sc0.47 1 2ρ µ ( du ) e
(4)
2 3 ρ k /
dx e e
w w
becomes unity. Here k is the so-called catalytic velocity defined as

w
k = η 4 C = η 4 8πm
w
c,f
w
kT = η
c,f 8
4 πγ c ≈ 0.34η c
w c,f
w c,f (5)
w
where η is the recombination coefficient, i.e., the fraction of the colliding reactive species (e.g.,
atomic oxygen) that undergoes the recombination process. The subscript “,f ” signifies that the
c,f
coefficient is for the forward process, to distinguish it from the reverse process. The quantity γ is
the specific heat ratio, Cp/Cv, and c is the sound speed. γ is taken to be 1.4 to arrive at the factor
0.34 in Eq. (5). The quantity k has its origin in the kinetic theory. In kinetic theory, the number
w
of atoms or molecules hitting a surface is given by the so-called arrival rate
Arrival rate = n4 C = n4 8πm

kT m−2 sec−1 (6)
where n is the number density in m−3 . The subscripts e and w in Eq. (4) signify the edge of boundary
layer and wall, respectively.
The dependence of heat transfer rate on ζ is
q −q = 1 +1 ζ .
−q
noncatalytic
q f ullycatalytic noncatalytic
This produces a curve known as Goulard’s S-curve, which is shown in Fig. 2.
RTO-EN-AVT-142 16 - 3
1.0
[q - qnoncat]/[qfull cat- qnoncat] 0.8
Fully catalytic
0.6
0.4 Goulard's S-curve

showing dependence
of heat transfer rate
0.2 Non-catalytic on ς
0.0
1000 100 10 1 0.1 0.01 0.001
ς = M/(ηRe0.5)
Figure 2. Goulard’s S-curve dependence of heat transfer rate on ζ .

By assuming that the pressure distribution in the stagnation region is that given by the Newto-
nian hyersonic theory, the quantity du /dx is found approximately to be e
du = V∞ 2 ρ∞ .
e
(7)
dx R ρ e
A typical value of ζ can be found by taking Sc = 0.5, T∞ = 250 K, T = 6000 K, T = 3000 K,

ρ∞ /ρ = 1/10, µ ∝ T 3 4 , γ = 1.4, and defining the freestream Reynolds number
e w
/
e
Re∞ = ρ∞µV∞ R ,
∞
which leads to, to within about 30% accuracy,
ζ ≈ η 1 Vc∞ √ 1 . (8)
Re∞ c,f w
Writing the ratio V∞ /c as M , the freestream Mach number evaluated at the wall temperature,
w w
a fully catalytic surface and a noncatlaytic wall can be defined as

Re >> M Fully catalytic wall : η (8a)
∞

c,f w
Noncatalytic wall : η Re∞ << M . c,f w (8b)

Using arc-heated wind tunnels, η for the O + O → O and N + N → N
2 c,f 2 reactions have been
measured for several common heatshield materials. Generally, it is between 0.1 and 1 for a metal,
between 0.01 and 0.1 for a metal oxide, and less than 0.01 for silica and alumina, two most commonly
found glassy oxides.
It is to be noted here that the surface of most innoble metals, such as copper, turns quickly
into a metal oxide when it is exposed to a hot stream containing atomic oxygen. A systematic
study was made in Ref. 2 to find out how fast a copper surface turns into a copper oxide surface.
The experiment was made at a pressure of the order of 0.01 atm. Within 1 sec at this pressure,
copper became copper oxide when it was tested in an arc-heated wind tunnel. In Ref. 3, copper
calorimeter was used to measure the heat transfer rate at a pressure of 2/3 atmospheres. In this
case, enthalpy of the flow was determined spectroscopically. The measured heat transfer rate to the
copper calorimeter was consistent with the assumption that η is 0.01, which is a typical value for c,f
metal oxides.
Of particular interest is the η value for O-atoms on the surface of graphitic carbon, which is
c,f
used as an ablative heatshield. Efforts have been made in the past to measure this quantity in an
16 - 4 RTO-EN-AVT-142
atomic beam experiment (see, e.g., Ref. 4). Those experiments found no O2 in the product: the
O-atoms either formed CO by combining with carbon or reflected elastically as an O-atom, without
forming O2 . That is,
For O + O + C (c) → O2 + C (c) : η ≈ 0. c,f (9)
2.2. η FOR CO + O → CO2
c,f
When unmanned Mars entry missions were started, question was raised as to what is the impact
of surface catalytic recombination. Martian atmosphere consists mostly of CO2 . At the edge of
boundary layer, CO2 is fully dissociated into CO and O. CO and O could conceivably recombine as
CO + O + S (c) → CO2 + S (c) + 5.35 ev.
At a flight velocity of 6 km/s, a typical entry velocity of a Martian vehicle, the kinetic energy of
one CO2 molecule is 8.21 ev. Therefore the reaction energy 5.35 ev is more than half of the flow
enthalpy. If CO and O do not recombine catalytically at wall, Eq. (1) tells us that the heat transfer
rate will be less than half of the fully catalytic case. That is, reward to making the wall noncatalytic
is quite large. The same problem exists for Venusian entry because the atmospheric composition of
Venus is similar to that of Mars.
Automobile manufacturers have developed catalytic converters which burn CO with O2 to make
CO2 . They accomplished it by making CO molecules collide many times with platinum surfaces in
the presence of O2 . This means that the catalytic efficiency η is very small for CO + O2 . NASA
c,f
Ames Research Center carried out an experiment to measure η for the CO + O reaction. Because
c,f
the surface of the heatshield for Martian entry vehicle is usually fused silica, the experiment was
conducted for the silica surface.5 The experiment found that the reaction CO + O is too slow to be
detected.
That η for CO + O → CO2 is small can be deduced theoretically. A catalytic reaction involves:
c,f
1) adsorption (sticking) of the two component species, 2) migration of the component species in the
adsorbed phase, 3) recombination of the two partners, and 4) desorption of the product. The CO
and O particles that strike the wall may not be adsorbed, i.e., fail step 1), or, skip step 3) and desorb
without recombining. The rates of these four processes can be modeled. That work will be done by
author lecturers. Here, only the steps 1) and 4) will be given to explain why η for the process is
c,f
slow.
The η for the catalytic process cannot be larger than the adsorption coefficient of CO, because
c,f
the latter is a component of the former. For a monatomic gas, the adsorption coefficient, denoted
here η , is known to be a function of the ratio of the mass of the gas atom m to that of the atoms
a g
forming the solid m . Classical calculation of the surface processes6 predict that, when the mass
c
ratio is much smaller than unity, η behaves as

a
ηa ≈ m /m . g c
The existing experimental data with helium, neon, and argon7 agree roughly with this theory.
As mentioned, the surface of the heatshield for the Martian entry vehicles is usually SiO2 .
Because the molecular weight of Si is 28, collisions of the CO and O will have little effect of this
mass ratio. Thus, in the absence of other effects, η should be nearly 1 in the Martian case.
a
But there is what is known as entropy effect, which concerns the change in the internal degrees
of freedom during adsorption. (To change internal degrees of freedom, heat must be given or taken
away at a constant temperature. By the definition of entropy, this implies a change in entropy.)
Let us designate the gas phase quantities by a subscript g and the adsorbed state quantities by a
subscript a. The adsorbed gas species are held there by the attractive potential between the solid
surface and the gas, D . The numerical value of D is known for some cases8 . For CO on a metal,
a a
it is typically between 1.5 and 6 ev.

A molecule in the adsorbed state oscillates within the adsorption potential well approximately
with the same frequency as the solid lattice, which is ν ≈ kT /h on the average. The fraction of
w
the adsorbed CO molecules that has an energy exceeding the adsorption energy D is exp(-D /kT ).
a a w
RTO-EN-AVT-142 16 - 5
Therefore, the number of CO molecules in the adsorbed state that desorb from the adsorption
potential well within 1 sec per 1 m2 is
Desorption rate = kTh exp(− kT
D )n m−2 sec−1
w a
a
w
As given by Eq. (6), the flux of gas-phase molecules reaching the wall is C n /4. Only the η w g a
fraction of this flux will be adsorbed, so that

Adsorption rate = η C n m−2 sec−1 . a
4 w g
Under equilibrium, these two rates must be equal, i.e.,

kT exp(− D )n = η C n
w a a w
h kT 4 w
aE g
which leads to
η = C4 kTh exp( kT
a
D )n .
n
w a aE
w w g
The ratio n /n is given by the partition functions for translation Q , for vibration Q , and
aE g T v
for rotation Q , by
n = Q Q Q exp( D ).
r
aE Ta Va Ra a
n Q Q Q
g kTTg Vg Rg w
Therefore
η = C4 kTh Q
a
Q Q
Q Q Q .
w Ta Va Ra
w Tg Vg Rg
Here, all partition functions are evaluated at the wall temperature T . w
The translational partition function in the gas phase, Q , is given by Tg
Q = ( 2πmkTh2 )
3 2
Tg
w /
The largest possible value of the translational partition function in the adsorbed phase, Q , occurs Ta
when the adsorbed molecule has a full freedom of translation in two dimensions, in which case
Q (full 2D) = 2πmkT
Ta
h2 .
w
When there is an energy barrier of D that hinders the two-dimensional motion, so that only those
b
molecules with energy greater than D have the two-dimensional translational freedom, one must
multiply a factor exp(-D /kT ) to this value, and therefore
b
Q = 2πmkT
Ta
D
h2 exp(− kT ).
w
(10) b
This leads to
η = C4 kTh ( 2πmkT
a
h2 )1 2 Q Q exp(− D ) = Q Q exp(− D ).
w
Q Q
/ Va
kT
Ra
Q Q kT (11)
b Va Ra b
w w Vg Rg w Vg Rg w
For a monatomic gas, Q and Q are both unity. If D = 0, Eq. (11) gives η = 1. This agrees
v r b a
with what is known about η for monatomic species.

a
At best, the rotational motion of an adsorbed molecule occurs around only one axis, the axis
normal to the surface. By assuming that the vibrational motion of the adsorbed phase is excited to
the same extent as the gas phase, one obtains the largest possible value of the adsorption coefficient
to be
η = kTB exp(− kT
a
D ) (12)
b
w w
16 - 6 RTO-EN-AVT-142
Assuming D = 0, the numerical value of η for CO at the wall temperature of 2000 K is deduced
a a
to be 0.0373. If D is finite, or the rotational motion is hindered completely, or if the vibrational

a
motion is hindered in the adsorbed phase, η will be smaller. This explains at least partly why the
a
catalytic recombination coefficient for CO + O is small on the surface of SiO2 .

2.3. GAS-SOLID REACTIONS AND SUBLIMATION
At high heating rates, the heatshield surfaces ablate. For an ablating environment, carbonaceous
heatshield is commonly used. For this case, the gas species O, N, and H can combine with the carbon
atoms of the solid surface as
O + C (c) → CO, ∆H = 3.74 ev, (13)
N + C (c) → CN, ∆H = 0.35 ev, (14)
H + C (c) → CH, ∆H = −3.88 ev, (15)
H + 2C (c) → C2 H, ∆H = −2.67 ev. (16)
For reaction (13), the existing experimental data9 show that the reaction coefficient η can be
r,f
approximated by
O + C (c) → CO : η ≈ 0.63exp(−1160/T ).
r,f w
For reaction (14), Ref. 10 shows

N + C (c) → CN : η r,f ≈ 0.3.
Reactions (15) and (16) can occur only at high wall temperatures, and have not yet been studied.
However, the maximum possible values can be reasoned simply from the requirement that only the
particles with the required energies (3.88 ev for H + C(c) and 2.67 ev for H + 2C(c)) can produce
the change. This leads to
H + C (c) → CH : η < exp(−45, 000/T )
r,f w
H + 2C (c) → C2 H : η < exp(−31, 000/T ). r,f w
If the flow contains CH or C2 H at the boundary layer edge, as in the atmosphere of Titan, reactions
(15) and (16) will occur in the reverse direction at wall. That is, carbon will condense on the wall
from CH and C2 H.
At high wall temperatures, evaporation (sublimation if the surface is solid) occurs. Imagine
that a wall material is enclosed in a high temperature furnace and is heated. Evaporation will
occur and fill the inside volume with the vapor of the material. The equilibrium vapor pressure for
most materials are known, and are given in the JANAF Thermochemical Tables.11 Denoting the
equilibrium vapor number density by n , the flux of the vapor species hitting the wall is given by
E
the kinetic theory to be n C /4. Of this flux, in general an η fraction will be adsorbed on the wall.
E w v
Under equilibrium, this flux must equal the flux of evaporation. Thus the evaporation rate is given
by η n C /4. Under nonequilibrium, the vapor number density n will produce adsorption rate of
v E w
η nC /4. Therefore, the net mass flux of evaporation will be

v w
Evaporating flux = η C (n − n) m−2 sec−1

4
v
w E
where n and n are evaluated at the wall. The corresponding mass flux becomes, in terms of mass
E
fraction α,
J = η4 ρ C (α − α) kg/(m2 − sec)
v
w w E (17)
Graphitic carbon vaporizes into C, C2 , C3 , C4 , and C5 . Of these, the predominant reaction is
3C (c) → C3, ∆H = −7.51 ev (18)
RTO-EN-AVT-142 16 - 7
For this process, η has been measured in Ref. 12. The results can be fitted by
v
η (C3 ) = 30exp(−21490/T ).
v w
Glassy oxides liquefy before they vaporize. Vaporization of liquid oxides involves at least two
steps. For instance, when SiO2 is heated, its vapor consists of SiO and O2 . No quantitative infor-
mation is available for η for this class of materials.
v
2.4. ROUGH SURFCES

When the surface reaction of the type Eqs. (13) to (16) or vaporization, Eq. (18), occurs,
the surface becomes rough. The extent of roughness is measured by the ratio of the so-called BET
(Bennett-Emmons-Thomas) area to the projected area, named here φ
BET area .
Roughness ratio φ = projected (19)
area
BET area is the wet area of a substance, and is determined by an experiment shown schematically
in Fig. 3. One places the sample inside a sealed jar and evacuates the jar. The sample is heated
while pumping to purge all gas molecules adsorbed on it. Then one admits gas, typically nitrogen,
while measuring the mass flow rate that enters the jar and watching the pressure inside the jar. In
the beginning, pressure in the jar does not rise because all gas molecules are consumed in adsorbing
on the wet surface of the sample. After a while, pressure begins to rise, because the surface is fully
covered by a monolayer of the adsorbed gas. The surface area occupied by one molecule is typically
10−15 cm2 . By dividing the total number of gas molecules consumed up to this point by this value,
one obtains the BET area.
POINT OF
TOTAL
COVERAGE
Figure 3. Schematic of the procedure to determine the BET area of a solid surface.
The BET area has never been measured during ablation. After ablation, theoretically it can be
measured. However, a heatshield material becomes highly porous after ablation. The experiment
shown in Fig. 3 will measure the wet area of the pores as well as the area exposed to the boundary
layer flow. Therefore, the roughness ratio φ can only be guessed. For many such porous materials,
φ is of the order of 100. Perhaps we should consider φ to be between 10 and 100.
We can now conduct a thought experiment as to what will be the magnitude of the adsorption
coefficient η on a rough surface, as shown schematically in Fig. 4. Here, a swarm of atoms is hitting
a
the inside of a crevice. In the first collision, η fraction is adsorbed, and 1 - η fraction leaves. This
a a
remaining swarm of atoms hits the wall the second time. The atoms that leave are now (1 - η )2 a
fraction of the original swarm. After n number of collisions, there will be (1 - η ) fraction of the
a
n
original swarm left unadsorbed. Therefore, the fraction of the atoms adsorbed are
Effective adsorption coef = 1 − (1 − η ) .
a
n
(20)
The total number of collisions, n, will be of the order of φ. So, to this order of accuracy, n can be
replaced by φ. If η is very small and if φ is large, Eq. (20) can be approximated by
a
Effective adsorption coef ≈ 1 − exp(−η φ).

a (21)
16 - 8 RTO-EN-AVT-142
These show that, when a surface is rough, its effective η approaches unity. a
Figure 4. Schematic showing collisions of particles on a rough surface.

2.5. INTER-RELATIONSHIP AMONG η’s
The above reasoning can be applied to catalytic reactions and surface reactions to conclude that
not only η , but η , η , and η all approach unity when roughness ratio is high. However, all η’s
a c r v
cannot become unity together because, by definition, the sum of all η’s for one species must not be
larger than unity:
Ση < 1. (22)
Otherwise, mass conservation is violated.
Take an example with O-atoms hitting a carbon surface. O-atoms can possibly recombine on the
surface with a coefficient η and simultaneously can react with carbon to form CO with a coefficient
c,f
η . The sum, η + η , cannot be larger than unity. Because η is nearly zero and η is large,
r,f c,f r,f c,f r,f
the best practice may be to take the effective η to be unity and the effective η to be 0 for a
r,f c,f
rough carbon surface.

2.6. FORWARD VS. REVERSE RATES
In Eq. (17), the two terms in the parenthesis represent the forward and reverse rates in evapo-
ration. Likewise, for both the catalytic recombination processes typified by Eq. (1) and the surface
reactions, Eqs. (13) to (16), the reverse rates must be accounted for. The reverse rates are chosen so
as to satisfy the principle of detailed balance in equilibrium. Otherwise, equilibrium condition will
not be reached when it should be. This aspect of the problem is discussed in Ref. 13.
For example, let us take the recombination process, Eq. (1). There is an opposing (reverse)
reaction
O + O + S (c) ← O2 + S(c). (23)
The fraction of the O2 molecules that undergoes this process will be designated by η . The net c,r
reaction rate will be the difference between the forward and the reverse rates
Net rate of O removal = η C n − 2 η C n m−2 sec−1 .
c,f
4 w,O O
c,r
4 w,O2 O2
Here, the factor 2 in the second term signifies that two O-atoms are created by the dissociation of
one O2 molecule. The mass flux of the O-atms removed by this process can be written as
J = 14 ρ (η C α − 2η C 2 α 2 ).
O w c,f w,O O,w c,r w,O O ,w (24)
Under equilibrium, the forward and reverse rates must be equal, which leads to
η = η 2 CC nn = √1 η ( nn ) .
c,f w,O O ,E O
(25)
2
c,r
2 2 w,O2 O ,E
c,f
O
E
RTO-EN-AVT-142 16 - 9
The ratio n O,E /n O2 ,E is given by the equilibrium relationship

( nn ) = Q (QQ )Q D ),
exp(− kT
2 2
O T ,O g
E
2 O 2 2 T ,O V,O R,O2 w
from which there results

( nn ) = n1 Q (QQ )Q D ).
exp(− kT
2
O
E
T ,O g
(26)
2O 2 2 O T ,O V ,O R,O2 w
A similar relation holds for all other catalytic recombination processes.

Similarly, for the surface reaction (13), the net rate is
Net rate of O removal = η 4 C n − η4 C n m−2 sec−1 r,f
w,O O
r,r
w,CO CO
or
J = 14 ρ (η C
O w r,f w,O α O,w −η r,r C w,C O α CO ,w ) kg/(m2 − sec). (27)
Here
η = η CC ( nn ) . r,r r,f
w,O
(28) O
E
w,C O CO
The ratio (n /n ) is given by the equilibrium relationship

O CO E
( nn ) = n 1 Q QQ Q Q exp(− kT
O
E
D ). T ,O
(29)
T ,C g
CO C,E T ,C O V,C O R,C O w
where n is the equilibrium vapor number density of C3 . A similar relationship holds for all other
C,E
surface reactions.
2.7. HEAT TRANSFER RATES
All three surface phenomena, i.e. catalytic recombination, reaction, and sublimation, produce
heat corresponding to ∆H . All these heat enter into the heatshield wall. The heat transfer rate
caused by these phenomena are
event × ∆H in J/mol W/m2 .
q = Rate of reaction in m (30)
2 sec 6.0225 × 1023
Here, one mus not confuse event/(m2 sec) with rate of species removal or species production. For
instance, for the oxygen recombination, Eq. (9), removal of two atomic oxygen equals one event.
Therefore, event/(m2 sec) is half of the rate of removal of O-atoms. The above equation is valid
unconditionally, regardless of the magnitude of ablation rate.
2.8. EXTENTION OF GOULARD’S THIRD AXIOM
We can now ponder how much impact the coefficients η (in the surface reactions of the form r
(13) to (16)) and η (in sublimation of the form (18)) will have on heat transfer at the stagnation
v
point. In other word, we would like to extend Goulard’s analysis to surface reactions and sublima-
tion. Unlike catalytic recombination process, surface reaction and sublimation produce ablation. To
extend Goulard’s theory to an ablating surface, we have to start with the fundamental conservation
equations.
Fundamental conservation equations for the frozen boundary layer flow are given in Goulard’s
work1 , and so will not be repeated here. Those equations are valid even when surface reactions and
sublimation occur. Following Goulard, we write those conservation equations using the coordinates
x and y shown in Fig. 1. Then one introduces the coordinate transformation
ru
√
ξ = ρ µ u r dx, η = 2ξ ρdy, (31a,b)
x y
2 e
e e e
0 0
16 - 10 RTO-EN-AVT-142
and the stream function f that satisfies the conditions

∂ Ψ = ρur, ∂ Ψ = −ρvr
∂y ∂x

Ψ = 2ξf (ξ, η). (32)
The stream function f can be written as
f = − ρv (33)
2ρ µ (du /dx) .
e e e
Then, denoting the partial differentiation ∂/∂η by , the mass and momentum conservation equations
reduce to:
( ρρµµ f ) + ff + 21 ( ρρ − f 2) = 0. e
(34)
e e
Following Goulard, we assume ρ /ρ = f 2 , so that Eq. (34) reduces further to

e
( ρρµµ f ) + ff = 0. (35)

e e
The energy equation is transformed likewise by introducing a normalized energy function

g = H/H e (36)
into
1 g) + fg + u2 [(1 − 1 ) ρµ f f ] = 0.
( ρρµµ Pr e
(37)
e e H Pr ρ µ e e e
Assuming Prandtl number to be unity, Eq. (37) reduces further to

1 g) + fg = 0.
( ρρµµ Pr (38)
e e
The species equation is transformed likewise into

( ρρµµ Sc1 α ) + fα = 0
e e
s s
(39)
When there is no ablation, the wall value of f is zero according to Eq. (33). The numerical value
of 0.47 appearing in Goulard’s parameter ζ , Eq. (4), is the wall slope of species mass fraction α in
the solution of Eq (39), (∂α/∂η) , for Sc = 1. 0.47 also happens to be the wall slope of the energy
variable in Eq. (38), (∂g/∂η) , for Pr = 1.
w
When there is ablation, the wall value of f

f = − ρ v . w w
(40)
2ρ µ (du /dx)
w
e e e
is no longer zero, but is finite negative. When f has a finite negative value, numerical value of
(∂α/∂η) and (∂g/∂η) are different from 0.47. Numerical solution of those differential equations
w
w w
give these wall slope values as shown in Fig. 5. A two-parameter fit

∂g ) , = ( ∂α ) = 0.47exp(0.467f − 4.017f 2 )
( ∂η (41)
w
∂η w w w
is shown in the figure for comparison. The analytical fit is seen to be valid to f of up to about -1.2. w
RTO-EN-AVT-142 16 - 11
0
10
-1
10
{dg/dη)w, (dα/dη)w
-2
10
-3
10
-4 Wall slope of
10
energy and species variables
at stagnation point
-5 Numerical solution
10 2-parameter fit
-6
10
0.0 -0.4 -0.8 -1.2 -1.6
fw
Figure 5. Wall slope of energy and species equations at the stagnation point,
under Goulard’s simplifications.
The wall slope of the energy equation, (∂g/∂η ) , is a measure of heat transfer rate. As Fig.
w
5 shows, it decreases rapidly as -f increases. This phenomenon is known as convective blockage

w
effect. Using Eq.(7), f can be written as w
/

f = − ρ ṁV 2−3 4 ( ρρ∞ )1 4 Re∞ ,
/
w
∞ ∞ e
where ṁ is the rate of mass injection by ablation. Taking ρ∞ /ρ to be 0.1, this becomes e

f ≈ −0.33 ρ ṁV Re∞ . (42)
∞∞ w
The quantity f is the fundamental quantity in the so-called blowing parameter, which is a ratio of
w
the mass injection rate ṁ and the freestream flow rate ρ∞ V∞ and which is defined slightly differently
by various researchers.
By replacing 0.47 in Goulard’s ζ , Eq. (4), by the expression (41), one has Goulard’s third axiom
extended to the case of finite ablation rate. Generally, finite f pushes the S-curve in Fig. 2 toward
w
the right, so that the wall behaves more as a catalytic wall for the same catalytic coefficient.
The S-curve can be used also to gage the impacts of surface reactions or sublimation on heat
transfer. For instance, if there was surface reaction of the form Eq. (13) but no surface catalysis or
sublimation, heat transfer rate will increase due to the surface reaction. The S-curve in Fig. 2 is
applicable exactly as it is. However, the magnitude of the heat transfer rate will have to be changed
using the curve shown in Fig. 5.
3.0. KINETIC BOUNDARY CONDITIONS
3.1. BIFURCATION MODEL
When the heating rate is above about 500 kW/m2 , a heatshield material loses mass during
heating. That is, it is not truly reusable. Between about 500 kW/m2 and 1.5 MW/m2 , the mass loss
rate is so small that the material can be replenished after use, to be used the next time. When the
heating rate exceeds about 1.5 MW/m2 , the mass loss rate is so high that an ablating material must
be used. An ablating material consists generally of two components, a highly temperature-resistant,
highly tensile-resistant fibre matrix and a resin. The material is usually porous. When it is heated,
resin vaporizes at relatively low temperatures, e.g., 500 K. The vapor passes through the pores and
escapes from the surface (see Fig. 6). The vaporization process absorbs heat. While passing through
16 - 12 RTO-EN-AVT-142
the pores, the vapor molecule decomposes into smaller molecules. This decomposition process absorbs
heat and thereby cools the material further. This gas is called pyrolysis gas, and the process is called
pyrolysis process. In the region near the wall surface, the matrix is devoid of the resin. This region
is called char. The zone where the vaporization of the resin is occurring is called the pyrolysis zone.
Diffusion of O
Catalytic
O recombination
O CO
O2
Sublimation
Surface
Pyrolysis gas
chemical
reaction
Char
Pyrolysis zone
Virgin
Figure 6. Schematic of the phenomena occurring below, at, and above the wall surface of
an ablating heatshield.
The general principle of mass balance at wall states that whatever is produced or removed at
wall is taken away or supplied by diffusion in the boundary layer, i.e.
Rate of production at wall = rate of removal by diffusion. (43)
Therefore, before proceeding any further, we must know how to express the rate of diffusion.
The rate of diffusion of a species is given in general by the Chapman-Enskog theory using
the multi-component diffusion coefficients. According to this formulation, the rate of diffusion of
species s is a weighted sum of gradients, ∂X /∂y, of all species except s. The weights contain the
i
multi-component diffusion coefficients D as a factor. This complicates wall mass balance condition
is
greatly.
A considerable simplification is achieved by adopting the so-called bifurcation model. The model
is based on the assumption that the multipcomponent diffusion coefficient between species i and j
can be approximated by
D ≈ FD̄F ,
i,j (44)
i j
Here, D̄ is a reference diffusion coefficient, which is chosen in Ref. 14 to be the self diffusion coefficient
of N2 . Eq. (44) is theoretically rigorous if the interaction among species is purely by the induced
dipole-induced dipole interactions. At temperatures of up to about 3000 K in air, molecular potentials
are indeed almost those of induced dipole-induced dipole interactions. At higher temperatures, that
is not so, and so a very rough approximation must be made with D ’s in order to satisfy this
i,j
condition. The values of F ’s so selected are given in Ref. 14. F is proportional approximately to
i s
the square-root of its molecular weight relative to the reference species (which is 0.028 kg/mol for
N2 ):
F ≈ 0M
s
.028 .
s
(45)
If Eq. (44) holds, the bifurcation model states that the rate of mass diffusion of species s can
be written as
s
D̄µ2 ∂Z kg/(m2 − s),
J = ρM̄µ ∂y
s
(46)
1
RTO-EN-AVT-142 16 - 13
where Z is a virtual mass fraction

Z = MF µX ,
s
s
s s
(47)
1 s
and X F , M X .
µ1 = s s µ2 = F
s s
(48a,b)
s
s s
M̄ is the average molecular weight

M̄ =
X M .
s s
The rate of production of species s by chemical reaction will be denoted by W . The conservation s
equation for species s can be written as

ρu ∂α ∂α ∂ ρD̄µ2 ∂Z
∂x + ρv ∂y = ∂y ( M̄µ1 ∂y ) + ρW .
s s
(49) s
s
The quantity D̄µ2 /(M̄µ1 ) can be considered to be an effective diffusion coefficient:

D̄µ2 .
D = M̄µ (50)
1
Schmidt number can now be defined as
Sc = µ/ρD. (51)
Sc is a function of temperature, but is not a function of species. The virtual mass fraction Z in Eq. s
(47) can be written further as

MX MX
Z = M X F µ = α F µ = α FM̄µ
s
M X s s k k k
s
k k k
s
k 2 k 2k 2
s s s
One can introduce a parameter C : s
C = FM̄µ . s (52)
2 s
Using these notations, the species conservation equation, Eq. (49), can be written as
ρu ∂α ∂α ∂ µ ∂ (C α )
∂x + ρv ∂y = ∂y [ Sc ∂y ] + ρW . (53)
s s s s
s
3.2. MASS BALANCE

In order to describe the behavior of an ablating wall, what happens inside the heatshield material,
i.e., material response, needs to be calculated. This can be done by using a computer code such as
Super Charring Materials Ablation (SCMA) code given in Ref. 15. The calculation will yield the
mass flow rate of the pyrolysis gas issuing from the wall, ṁ . By invoking equilibrium among the p
various species constituting the pyrolysis gas, one can determine the mass flow rates of each of the
species ṁ . The mass fraction of species s in the pyrolysis gas will be represented by α .
p,s s,p
The total mass flow rate emerging from the wall is the sum of the pyrolysis mass flow rate ṁ p
and the mass flow rate produced at wall ṁ t
ρ v = ṁ + ṁ , w w p (54) t
The surface mass removal rate ṁ is t
ṁ = sum of mass removal rates by surface reaction

t
+sum of mass removal rates by sublimation

As can be seen from Eqs. (24) and (27), mass removal rates by surface reactions consist of terms
proportional to α . As can be seen from Eq. (17), the mass removal rates by sublimation consist
s,w
of terms proportional to α and terms independent of α . Thus, one can write

s,w

ṁ = (δ α + ). (55)
s,w
t s s,w s
The parameters δ and are determined by adding up contributions from all surface chemical and
s s
16 - 14 RTO-EN-AVT-142
sublimation processes. For instance, sublimation into C3 , Eq. (18), contributes

δ 3 = −0.25η
C v,C3 ρ C w w,C3 ,
3 = η 3m 3n 3C 3.C v,C C E ,C w,C
For a species s, the total mass flow rate at wall is the sum of the mass flow rate in the pyrolysis
gas and the mass flow rate due to surface removal
Species flow rate at wall = ṁ + ṁ . p,s t,s (56)
In the gas phase, at an infinitesimally small distance away from the wall, the species mass flow rate
is the sum of a convective term, ρ µ α , and a diffusion term
w w s,w
Species flow rate in gas phase = ρ v α − ( ρM̄µ

D̄µ2 ) ( ∂Z )
∂y .
s
(57)
w w s,w w w
1
µ ∂C
= ρ v α − Sc [α ( ∂y ) + C ( ∂y ) ] ∂α s s
w w s,w s,w w s,w w
Because expression (56) and expession (57) must be equal to each other, one has
ṁ + ṁ = ρ v α − Sc µ [α ( ∂C ) + C ( ∂α ) ]. s s
(58)
p,s t,s w
∂y w
∂y
s,w s,w w s,w w
On the other hand ṁ can be written as

γ
t,s
ṁ = t,s k,s α k,w + ṁ . v,s (59)

k
For example, γ by O + C(c) → CO is

k,s
γ
O,O = −0.25η ρ C . r,O w w,O
The mass flow ρ v can be written in turn as

(δ α + ).
w w
ρ v = ṁ +
w w p s s,w s (60)
s
By substituting these into Eq. (58), one obtains the equation governing the wall value of mass
fractions of species s as

[−(ṁ + (δ α + ) + Sc µ ( ∂C ) ]α + γ α w s
p k
∂y
k,w k w s,w k,s k,w
k
w
(61) k
µ ∂α
+ Sc C ( ∂y ) = −ṁ − ṁ . w s
k,w w p,s v,s
w
Eq. (61) forms a system of N simultaneous quadratic equations in α ’s. To determine α ’s

using this equation, one must determine the wall slope (∂α /∂y) . The wall slope is determined by
s,w w
s w
solving the flow field. Flow field can be solved for this purpose either by using the so-called two-layer
method, i.e., inviscid layer plus boundary layer, or by using computational-fluid-dynamics (CFD).
In the following, these two methods are reviewed.
3.3. APPLICATION TO BOUNDARY LAYER ANALYSIS
In the two-layer method, the inviscid flow field is solved by any available method. Then the
boundary layer is attached to the inner boundary of the inviscid solution. The outer boundary
condition of the boundary layer is thus given by the inner boundary value of the inviscid solution. It
is believed that several computer codes exist in the world presently in which this method is used. To
RTO-EN-AVT-142 16 - 15
a high accuracy, chemical equilibrium can be assumed at the edge of the boundary layer. However,
within the boundary layer, frozen flow assumption is valid to a quite high accuracy.
Boundary layer analysis was introduced in Section 2.8. Here, we have to generalize it into
off-stagnation points and to the case of Prandtl number different from 1, and, above all, must use
the bifurcation model introduced above. Coordinate transformation is unchanged. The resulting
differential equation becomes, for mass and momentum,
∂f
( ρρµµ f ) + ff + 2β( ρρ − f 2 ) = 2ξ(f ∂f
e
∂ξ − ∂ξ f ) (62)
e e
where
β = dd[[lnln((uξ )])] . e
The energy equation becomes

( ρρµµ Pr 1 g) + fg + u2 [(1 − 1 ) ρµ f f ] = 2ξ(f ∂g − ∂f g).
e
(63)
e e H Pr ρ µ
e ∂ξ ∂ξ
e e
The species equation is transformed into

( ρρµµ CSc α ) + (f + ρρµµ CSc )α + [( ρρµµ CSc ) − 2f β ]α
s s s
s s
(64)
s s
e e e e e e
= 2ξ(f ∂α −
∂ξ ∂ξ
∂f α ) s
s
where
β = d[dln[ln(α(ξ )])] . s
s,e
There are N (number of species) equations of the form Eq. (64).

Eqs. (62) to (64) form a system of nonlinear ordinary differential equations because a) Eq.
(62) contains terms consisting of products of dependent variables, and (b) ρ /ρ in Eq. (62) depends e
nonlinearly on g and α’s. This system of equations can be solved through approximate linearization
and iteration. In the end, at each ξ -point, the solution of the species equations can be written as
α (η ) = α Y 1 (η ) + α Y 2 (η ) + Y ( η ).
s s,w s,w
(65) p
where the homogeneous solutions Y1 , Y2 and the particular solution Y are chosen so that p
Y1 (0) = 1, Y1 (0) = 0,

Y2 (0) = 0, Y2 (0) = 1,
Y (0) = 0, Y (0) = 0.
p p
By demanding that α value in Eq. (65) becomes the boundary layer edge value at the boundary
s
layer edge, one has

α Y1 (s,η ) + α Y2 (s,η ) + Y (s,η ) = α .
s,w e s,w e (66)
p e s,e
There are N equations of the form (66).

Eq. (61) is converted into the boundary layer coordinates by introducing
∂α = dη α s
∂y dy s
which leads to a system of simultaneous nonlinear algebraic equations:

[−(ṁ + (δ α + ) + Sc µ ( ∂η ) C ]α + γ α w
p k k,w
∂y k w s,w s,w k,s k,w
k
w
(67) k
µ ∂η
+ Sc C ( ∂y ) α = −ṁ − ṁ .
w
k,w w s,w p,s v,s
w
16 - 16 RTO-EN-AVT-142
Eqs. (66) and (67) form 2N algebraic equations containing 2N unknowns α and α . These s,w
equations can be solved through approximate linearization and iteration. The values of α are first
s,w
s,w
assumed. These values are used in evaluating δ α . Then there results a system of linear equations
k k,w
of order 2N
[A]X = R.
Here,
For k = 1 to N : X = α , k,w
For k = N + 1 to 2N : X = α . k,w
This equation set is solved as

X = [A]−1 R,
to obtain improved values of α and α . For the next iteration,
s,w s,w
Next α s,w = × (New α ) + (1 − ) × (old α )

s,w s,w (68)
where is a number smaller than 1, is used in the evaluation of δ α , and so forth. This process is
k k,w
repeated until a converged set of solution vector is obtained. Experience has shown that the most
appropriate value of varies between 0.2 and 1 depending on the problem.
3.4. APPLICATION TO CFD
CFD codes are believed to exist worldwide in which the kinetic boundary conditions are applied
as outlined in this lecture (e.g. Ref. 16). In CFD, the relation (61) becomes the boundary condition
for species variables. The wall slope (∂α /∂y) is written in a difference form. Because Eq. (61) is
s w
nonlinear in α , approximate linearization must be carried out. It can be achieved by assuming

s,w
δ α to be known. In updating δ α , it would be wise to correct only partly, i.e. use Eq. (68),
k k,w k k,w
to ensure stability.
Applying the boundary condition of a linear form will not be of any difficulty in the CFD
community. If a two-point representation of wall slope is used, Eq. (61) will supply the diagonal
term and one off-diagonal term in the difference formulation. If a three-point representation of wall
slope is used, two off-diagonal terms will be specified. If a steady-state solution is sought and if
the solution is obtained by a time-asymptotic method, the approximate linearization scheme will be
implemented by updating δ α after each iteration. If a time-accurate solution is sought, a unique
k k,w
iterative procedure must be devised in which the iteration is performed while time is frozen.
4.0. SURFACE PROBLEMS IN PLANETS
4.1. EARTH ENTRIES
In Earth’s atmosphere, the gas-surface interaction problem was studied in the past mostly in
two areas: catalytic recombination of oxygen and ablation of carbonaceous heatshield. Goulard’s
pioneering work for the stagnation point was discussed already. His work was extended to off-
stagnation points, three dimensional flows, and to nitrogen recombination by various researchers.
Until recently, ablation phenomenon was treated by invoking gas-surface equilibrium. The as-
sumption of gas-surface equilibrium circumvents the need to know the η and η . In the kinetic
r,f v,f
approach, the equilibrium condition can be reached by assuming η and η to √ be a large number.
r,f
According to Eqs. (8a), η and η need only to be much larger than M / Re∞ to produce an
r,f v,f
v,f
effect of a gas-surface equilibrium. The equilibrium approach and the kinetic approach have been
compared in Ref. 13. It found that equilibrium approach gives always higher ablation rate than the
kinetic approach.
The recent data on nitridation10 , N + C(c) → CN, which gives η to be 0.3, has brought an
r,f
alarm and skepticism of the measurement13 . This is because, if the measured value of η is accurate, r,f
this process will be the major process of ablation in a hyperbolic Earth entry. More experiment is
desirable.
For wall heat transfer rates in the range from 5×105 and 1.5×106 W/m2 , such as in the stagnation
region of the nose cone of the Space Shuttle, carbon-carbon composite overcoated with silicon carbide,
RTO-EN-AVT-142 16 - 17
SiC, is used for the surface. When exposed to a stream containing atomic oxygen, SiC is converted
quickly to SiO2 . SiO2 is in an amorphous form. Oxygen atoms are dissolved in this amorphous SiO2 ,
diffuse, and reach the SiC substrate. Once they reach the SiC substrate, they combine with C to
form gaseous CO. These CO molecules are dissolved in SiO2 , diffuse, and escape from the surface,
leaving SiO2 . This phenomenon has been studied experimentally to a considerable extent in Europe
and in Japan. Yet, there are no theoretical model to describe the rate process quantitatively.
In the future, attempts will be made to construct an airbreathing hypersonic vehicle. High
heating rates will occur at the nose tip, wing leading edge, cowl lip, and in the combustion chamber
of such a vehicle. New material such as halfnium di-boride, HfB2 , may be used for such high heating
areas. Gas-surface interaction for such a new material will be a challenge.
The ablation rate of carbonaceous heatshield material in the nose region of a blunt body was
measured systematically in Ref. 17. The measured ablation rate was about 3 times higher than
calculated assuming gas-surface equilibrium in the stagnation region. This cannot be attributed to
roughness-caused turbulence because surface roughness cannot produce turbulence at the stagnation
point, because velocity is zero there. In Ref. 18, this large ablation rate is attributed to the turbulence
generated by ablation.
Various experiments showed that the highest heating rates occur at points away from the stag-
nation point, somewhere where the surface normal is about 40◦ with respect to the flow direction
(see, e.g, Ref. 19). This peak-heating region is usually the point where turbulent transition occurs
due to roughness. In many cases it coincides also with the point of sphere-cone juncture where an
adverse pressure gradient occurs. Heat transfer rate in the region downstream of this point is usually
higher than the heat transfer rate to the stagnation point, and therefore the ablation rate is also
higher there. Theoretical prediction of the ablation rate in this downstream region from the first
principles is at this time not yet achieved. To do so, one needs to correctly model injection-induced
turbulence and roughness-induced turbulence.
If one can model turbulence accounting for these effects, then the theory given in Chapter 3 can
be applied. The only part that needs modification is viscosity µ which will have to be modified to
include the effect of turbulence.
Because there is a large uncertainty about the behavior of turbulence in the downstream region,
the heatshield is usually over-designed. In the past, this philosophy was proven to be effective and
necessary.
4.2. MARTIAN ENTRIES
Martian entries occur typically at 6 km/s. The density of the atmosphere is about 1/70 of
that of the Earth. Heating rate does not exceed 4×106 W/m2 . A material made of silica matrix
impregnated with an organic resin is used as an ablating heatshield. As mentioned earlier, Martian
atmosphere consists mostly of CO2 . At the edge of boundary layer, CO2 is dissociated into CO and
O. CO and O diffuse through the boundary layer and reach the wall. At wall, CO + O → CO2 can
conceivably occur. The surface of the material during entry is pure silica. As mentioned, silica shows
a near-zero reaction coefficient for CO + O.
One uncertainty is about the effect of surface roughness on the effective recombination coefficient,
Eq. (21). Because we do not know the BET area of a Martian heatshield during an entry flight,
we cannot determine the effective recombination coefficient. In addition, we do not know where
turbulent transition occurs because we do not know the intensity of the injection-induced turbulence
or the roughness-induced turbulence.
4.3. VENUSIAN ENTRIES
The atmospheric composition of the planet Venus is similar to that of Mars. But the density
of the atmosphere is about 70 times that of the Earth, and the entry speed is more than 10 km/s.
At the peak-heating point, the pressure at the stagnation point is of the order of 10 atm. The
combination of the density and entry speed produces heating rates in the range of several million
W/m2 . Only carbonaceous ablation material can be used as heatshield. The comments about carbon
surface made in Chapters 1 and 2 are applicable.
In the Pioneer-Venus Probe mission, four probe vehicles entered into the Venusian atmosphere.
Each vehicles were equipped with two thermocouples in their heatshield, one near the stagnation
16 - 18 RTO-EN-AVT-142
point and the other near the frustum edge. The data from these thermocouples indicated that the
heat transfer rates were higher at the frustum edge than at the stagnation point. The flight data
are consistent with the assumption of a turbulent flow.20 21
,
In this entry flights, the edge of boundary layer will contain oxygen atoms and a small concen-
tration of nitrogen atoms. The reactions (13) and (14) will occur. Reaction (13) will contribute
considerably to the total heat transfer rate. Sublimation reaction (18) will occur also.
4.4. TITAN ENTRIES
The atmosphere of Titan consists mostly of N2 , with a small amount of CH4 . The density of the
atmosphere is nearly the same as that of the Earth. For the Huygens vehicle sent by the European
Space Agency, the entry velocity was about 5 km/s. However, in future missions, entry velocity of
up to about 14 km is possible. Radiation from the CN becomes a major component of heat transfer
rate. Both a silica-based and a carbon-based heatshield can be used for these entries. There will
be a small amount of dissociation of N2 , and full dissociation of CH4 . Reactions (15) and (16) will
occur in both directions.
4.5. OUTER PLANET ENTRIES
Atmosphere of the outer planets consists of hydrogen and helium. Entry into Jupiter has been
made in the Galileo Probe mission. Heat transfer was due overwhelmingly by radiation. The
heatshield was made with a carbon-based material. During this entry flight, progression of surface
recession was measured. The flight data showed a surprisingly low surface recession at the stagnation
region and surprisingly high recession at the frustum.
The low recession in the stagnatin region is attributed to thermal nonequilibrium.22 The high
recession in the frustum region is attributed to increase in radiation due to turbulence caused by
ablation.23 The dominant gas-surface interaction was sublimation, reaction (18). The sublimation
rate was driven essentially by the energy conservation law, rather than by surface kinetics.
In a futuristic mission, one can conceive entry flights into Saturn, Uranus, or Neptune. In such
entries, radiative heating will be much less than that to the Galileo Probe. There, not only the
sublimation but the eactions, Eqs. (15) and (16) could conceivably become important. Surface
kinetics may play a role.
5.0. CONCLUDING REMARKS
In the above, gas-surface interaction problem was studied from the kinetic approach. The kinetic
approach enables rational way of quantitatively evaluating the surface phenomena in determining
heat transfer rates and surface ablation rates. The study identifies areas of uncertainty also, which
are: 1) the effective coefficients of gas-surface interactions for rough surfaces, and 2) the intensity of
turbulence for a rough surface in the presence of ablation. As a minor and overcomable difficulty,
nonlinearity in the governing equations at wall is pointed out.
The problem of gas-surface interaction in hypersonic flight contains several different disciplines
that are tied together in a rather unexpected way. Because of the unusual nature of the problem, the
traditional discipline-based approach to the problem becomes difficult to apply. This is probably the
reason for the misunderstanding and mismanagement of the issues that existed in the past. To ensure
mission success in the midst of such uncertainties, the planetary entry vehicles had been designed in
the past with a rather large safety factor. In retrospect, we feel that such large safety factors were
justified, because several unexpected phenomena occurred. For a rational management of the future
space programs, all scientific issues about gas-surface interaction will have to be understood better
by all involved. Much more research on the subject is highly desirable.
6.0. REFERENCES
[1] Goulard, R. (1958). Catalytic Recombination Rates in Hypersonic Stagnation Heat Transfer. Jet
Propulsion, Vol. 28, No. 11, pp. 737-745.
[2] Winkler, E. L., & Sheldahl, R. E. (1966). Influence of Calorimeter Surface Treatment on Heat-
Transfer Measurements in Arc-Heated Test Streams. AIAA Journal, Vol. 4, No. 4, pp. 715
-716.
RTO-EN-AVT-142 16 - 19
[3] Park, C., Raiche, G. A., II, Driver, D. M., Olejniczak, J., Terrazas-Salinas, I., Hightower, T.
M. & Sakai, T. (2004). Comparison of Enthalpy Determination Methods for an Arc-jet Facility.
AIAA Paper 2004-0487, January 2004.
[4] Liu, G. N-K. (1973). High Temperature Oxidation of Graphite by a Dissociated Oxygen Beam.
Ph.D thesis, Massachusetts Institute of Technology, Department of Aeronautics and Astron-
autics.
[5] Sepka, S., Copeland, R. A., Marshall, J., & Chen, Y-K. (1999). Experimental Investigation of
Surface Reactions in Carbon Monoxide and Oxygen Mixtures. AIAA Paper 99-3629.
[6] Pagni, P. J. (1972). Comparison of Diffusion Theory Adsorption and Desorption Rates with Exper-
imental Lifetimes. University of California College of Engineering Report No. ME-72-9.
[7] Somorjai, G. A. (1972). Principles of Surface Chemistry. Prentice-Hall, Engldwood Cliffs, NJ.,
pp. 209-213.
[8] Somorjai, G. A. (1972). Principles of Surface Chemistry. Prentice-Hall, Engldwood Cliffs, NJ.,
p. 245.
[9] Park, C. (1990). Nonequilibrium Hypersonic Aerothermodynamics. John Wiley, p 351.
[10] Park, C. & and Bogdanoff, D. W. (2003). Shock Tube Measurement of Coefficient of Reaction of
Nitrogen Atoms with Solid Carbon: Preliminary Results. AIAA Paper 2003-0158.
[11] Chase, M. W., Davies, C. A., Downey, J. R., Jr., Frurip, D. J., McDonald, R. A., & Syverud,
A. N. (1985). JANAF Thermochemical Tables, Part 1. Al-Co, Part 2. Cr-Zr. American Chemical
Society.
[12] Baker, R. L., McDonaugh, J. M., Herr, K. C., Klingberg, R. A., Coffer, J. C., & Covington,
M. A. (1084). Carbon Vaporization and Condensation Effects. NASA TM 874300.
[13] Chen, Y. K. & Milos, F. S. (2004). Finite-Rate Ablation Boundary Conditions for a Carbon-Phen-
olic Heat-Shield, AIAA Paper 2004-2270.
[14] Park, C., Jaffe, R. L., & Partridge, H. (2001). Chemical-Kinetic Parameters of Hyperbolic Earth
Entry. Journal of Thermophysics and Heat Transfer, Vol. 15, No. 1, pp. 76-90.
[15] Ahn, H. K., Park, C., & Sawada, K. (2002). Response of Heatshield Material at Stagnation Point
of Pioneer-Venus Probes. Journal of Thermophysics and Heat Transfer, Vol. 16, No. 3, pp.
432-439.
[16] Park, C. (2004). Stagnation-Point Radiation for Apollo 4. Journal of Thermophysics and Heat
Transfer, Vol. 18, No. 3, pp. 349-357.
[17] Wool, M. R. (1975). Final Summary Report Passive Nosetip Technology (PANT) Program. SAMSO-
TR-75-250.
[18] Park, C. (1984). Injection-Induced Turbulence in Stagnation-Point Boundary Layers AIAA Journal,
Vol. 22, No. 2, pp.219-225.
[19] Park, C. & Balakrishnan, A. (1985). Ablation of Galileo Probe Heat-Shield Models in a Ballistic
Range. AIAA Journal, Vol. 23, No. 2, pp.301-308.
[20] Ahn, H. K., Park, C., & Sawada, K. (2002). Response of Heatshield Material at Stagnation Point
of Pioneer-Venus Probes. Journal of Thermophysics and Heat Transfer, Vol. 16, No. 3,pp. 432-
439.
[21] Takahashi, M. & Sawada, K. (2002). Simulation of Entry Flight Flowfield over Four Probe Vehicles
in Pioneer-Venus Mission. AIAA Paper 2002-0909.
[22] Park, C. (2005). Overview of Radiation Problems in Planetary Entries. Proceedings of the Intern-
ational Workshop on Radiation of High Temperature Gases in Atmospheric Entry - Part 2,
European Space Agency SP 583, pp. 3-14.
[23] Matsuyama, S., Shimogonya, Y., Ohnishi, N., Sawada, K. & Sasoh, A (2002) Numerical Simul-
ation of Galileo Proe Entry Flowfield with Radiation. AIAA Paper 2002-2994.
16 - 20 RTO-EN-AVT-142
A Risk-Based Approach for Aerothermal/
TPS Analysis and Testing
Michael J. Wright∗ and Jay H. Grinstead†

NASA Ames Research Center, MS 230-2
Moffett Field, CA 94035
USA
Deepak Bose‡
ELORET Corp.
Sunnyvale, CA 94089
USA
ABSTRACT
The current status of aerothermal and thermal protection system modeling for civilian entry missions is
reviewed. For most such missions, the accuracy of our simulations is limited not by the tools and
processes currently employed, but rather by reducible deficiencies in the underlying physical models.
Improving the accuracy of and reducing the uncertainties in these models will enable a greater
understanding of the system level impacts of a particular thermal protection system and of the system
operation and risk over the operational life of the system. A strategic plan will be laid out by which key
modeling deficiencies can be identified via mission-specific gap analysis. Once these gaps have been
identified, the driving component uncertainties are determined via sensitivity analyses. A Monte-Carlo
based methodology is presented for physics-based probabilistic uncertainty analysis of
aerothermodynamics and thermal protection system material response modeling. These data are then used
to advocate for and plan focused testing aimed at reducing key uncertainties. The results of these tests are
used to validate or modify existing physical models. Concurrently, a testing methodology is outlined for
thermal protection materials. The proposed approach is based on using the results of
uncertainty/sensitivity analyses discussed above to tailor ground testing so as to best identify and quantify
system performance and risk drivers. A key component of this testing is understanding the relationship
between the test and flight environments. No existing ground test facility can simultaneously replicate all
aspects of the flight environment, and therefore good models for traceability to flight are critical to ensure
a low risk, high reliability thermal protection system design. Finally, the role of flight testing in the overall
thermal protection system development strategy is discussed.
1.0 INTRODUCTION
Any Earth or planetary entry vehicle will be subjected to significant aerothermal heating as it dissipates its
kinetic energy at the destination planet (or moon). The primary purpose of the thermal protection system
(TPS) is to protect the payload (crew, cargo, or science) from this entry heating environment. The
performance of the TPS is determined by the efficiency and reliability of this system, typically measured
∗
Senior Research Scientist, Reacting Flow Environments Branch. e-mail address: Michael.J.Wright@nasa.gov
†
Senior Research Scientist, Reacting Flow Environments Branch. e-mail address: Jay.H.Grinstead@nasa.gov
‡
Senior Research Scientist. e-mail address: dbose@ionamerica.com
Wright, M.J.; Grinstead, J.H.; Bose, D. (2007) A Risk-Based Approach for Aerothermal/TPS Analysis and Testing. In Experiment,
Modeling and Simulation of Gas-Surface Interactions for Reactive Flows in Hypersonic Flights (pp. 17-1 – 17-24). Educational Notes
RTO-EN-AVT-142 17 - 1
A Risk-Based Approach for Aerothermal/TPS Analysis and Testing
in terms of the total mass of material and associated sub-structure required to protect the payload to a
prescribed level of risk-tolerance. Therefore, for a rigid or flexible aeroshell, the choice and design
thickness of the TPS material are key performance metrics. The choice of the TPS material is typically
governed by the peak heat flux, surface pressure, and shear stress encountered during the entry, with more
robust (higher density) materials required to protect the vehicle from more severe entry conditions. The
thickness of the chosen material is governed by the total integrated heat load during the entry (which can
be very large for aerocapture or lifting-body cruise missions, due to a long residence time in the
atmosphere [1]). The determination of these quantities relies on aerothermodynamics and material
response modeling as well as ground-based testing in flight-relevant environments to ensure adequate
material performance.
For most missions of interest, no ground test can simultaneously reproduce all aspects of the flight
environment. Therefore, a good understanding of the relevant physics is required to trace ground testing
results to predicted flight performance metrics. The design and operation of reliable TPS requires accurate
knowledge of the incident environment and resulting material response. Our ability to accurately and
conservatively simulate the environment and the material response directly affects TPS risk, mass and
reliability. Improving the accuracy of the underlying physical models will enable a greater understanding
of the system-level impacts of a particular thermal protection system and of the system operation and risk
over its operational life. Prior experience with both reusable and ablative TPS is that the accuracy of our
simulations is typically limited not by the tools and processes currently employed, but rather by reducible
deficiencies in the underlying physical models. Fortunately, while specific modeling deficiencies are very
mission dependent, in general they can be divided into several broad categories: nonequilibrium gas
kinetics, shock-layer radiation, transition and turbulent heating, afterbody heating, gas-surface
interactions, and coupling between the TPS material and the flow environment.
Although baseline predictions of the aerothermal environment and resulting material response are made
using high-fidelity modeling and simulation tools, the evaluation of associated uncertainties often involves
comparatively little or no rigor. However, an accurate determination of these uncertainties is critical in
ensuring selection of an appropriate material and in determining the resulting margin and factor of safety.
An accurate assessment of TPS risk is required to ensure that the sub-system design is consistent with the
risk posture of the program. The results can also be incorporated with probabilistic design techniques
[2][3] so that engineers can make informed decisions on ways to effectively balance mission risk between
the TPS and other sub-systems. Historical approaches for dealing with these uncertainties have
traditionally been somewhat ad-hoc, most commonly relying on expert judgment [4]. Even when rigorous
attempts at uncertainty estimation were attempted [5][6], computational resource limitations prevented the
sort of non-linear multivariate analysis that is truly required. Final “rolled-up” uncertainties are then
typically determined using either a stacked worst-case approach, which can be needlessly conservative, or
a root-sum-square approach, which is only appropriate for a set of small linearly independent errors.
Worse yet, both of these approaches may in fact be non-conservative if the underlying component
uncertainties are not estimated correctly [5]. This paper will briefly discuss a Monte-Carlo based
methodology for physics-based probabilistic uncertainty analysis of aerothermodynamics and material
response modeling, as detailed in Refs. [7] & [8]. A primary objective of this work is to quantify the
uncertainties in vehicle heating and the resulting TPS sizing based on inaccuracies in the knowledge of the
input parameters. In addition, the technique allows TPS designers to prioritize and target key input
uncertainties for further testing [9] in order to maximize the return from limited research funding. The
results of the analysis can also be used to assist in the identification of possible additional uncertainties
due to erroneous or overly simplified physical models (also known as structural uncertainties); if the
models employed are correct, the sensitivities predicted by the analysis should be reproducible via targeted
experiments. Identification of structural uncertainties is a key element of this type of analysis, because a
large undetected structural uncertainty can invalidate the design, resulting in a potentially non-
conservative TPS solution.
17 - 2 RTO-EN-AVT-142
Although high fidelity analysis is an essential element of TPS design, it is only one component of the
overall program. The analysis tools employed are fundamentally supported by ground tests in high
enthalpy test facilities that simulate the flight environment over a range of conditions and at appropriate
time and length scales. Ground testing in these facilities is used to validate the simulation tools; develop
and select TPS materials; and qualify the final material for flight [10]. For example, arc jet facilities have
been a workhorse for TPS testing, design, and qualification in the United States for over 40 years.
However, no single high enthalpy ground test facility can replicate all transatmospheric flight conditions
for most missions of interest. Therefore, successful test planning strategies combine the advantages of
different facilities with appropriate choices of test conditions to encompass, as thoroughly as possible, the
critical features of the flight environment [11]. A validation effort is broken down into discrete
components, with each component designed to stimulate a specific aerothermal or material response
process through parametric variation of test conditions. The capabilities and shortcomings of the facilities
employed must be understood when formulating test objectives, interpreting test results, and integrating
the results across the ground test envelope. This paper outlines a testing strategy, first discussed by
Grinstead et al. [10], intended to ensure that the testing performed is of maximum utility to the
computational modeler and TPS designer. Pre- and post-test computational analysis, as well as non-
intrusive experimental flowfield diagnostic techniques, play a critical role in ensuring ground to flight
traceability of the resulting test data.
Finally, the role of flight testing in the overall TPS design and certification process must be explored.
Flight testing, while certainly capable of fully exercising the TPS design, is prohibitively expensive in
most cases. In general flight testing should be reserved for final model and system-level validation, once
we have good physics-based models of the expected environment and resulting material performance.
Ideally, the purpose of a dedicated flight experiment should not be to learn anything new about the flight
environment encountered, but rather to validate that the models employed are adequate. While dedicated
flight tests are rare, it should be noted that valuable engineering information could be obtained at a much
lower cost by including engineering instrumentation on science missions. Such instrumentation has little
benefit to the mission that flies it, but the overall exploration program benefits greatly from the additional
knowledge gained about the flight environment, which can the be applied to improving the design of
future missions.
2.0 IDENTIFICATION OF MODELING AND SIMULATION GAPS

Modeling gaps are deficiencies in the current state of the art that impair the fidelity of aeroheating or TPS
sizing analysis. In general, civilian entry vehicles can be separated into three classes: reusable launch
vehicles (RLV’s), such as the Shuttle Orbiter, inflatable decelerators, such as ballutes, and entry capsules,
such as all previous planetary probes, the Apollo Command Module, and the proposed NASA Crew
Exploration Vehicle (CEV). Each of these vehicle classes has their own set of modeling gaps. For RLV
analysis many of the primary gaps are in process improvement (e.g. grid generation over steps, gaps and
control surfaces) [12][13]. The entry physics are generally well understood, although local effects around
deflected control surfaces and gas-surface interactions (primarily catalysis) on the reusable TPS materials
employed present significant challenges to the state of the art. Inflatable decelerator systems operate at a
low ballistic coefficient, and are by definition flexible. Although such vehicles fly in a regime where
heating rates are comparatively low, the materials employed have much lower performance limits than
traditional rigid TPS materials [14], and thus are potentially more sensitive to large uncertainties. Principle
modeling gaps for this class of entry systems include non-continuum and free molecular aeroheating as
well as unsteady aeroelastic/aerothermal interactions [15]. Finally, capsules typically have much higher
entry velocity than RLV’s and can enter a variety of planetary atmospheres, each with unique
thermochemistry. For this class of missions the accuracy of our simulations is typically limited not by the
tools and processes currently employed, but rather by reducible deficiencies in the underlying physical
models. These deficiencies, while very mission-specific, can be divided into several broad categories:
RTO-EN-AVT-142 17 - 3
nonequilibrium gas kinetics, shock-layer radiation, transition and turbulent heating, afterbody heating, gas-
surface interactions, and coupling between the TPS material and the flow environment.
In any case, the first step after identifying the key mission requirements is to perform a mission-specific
gap analysis. In this process we first seek to determine knowns and “known-unknowns” via literature
searches, analysis of existing ground test and flight data, and relevant previous mission studies. A baseline
set of physical models is then selected that captures the important aspects of the entry environment.
Sensitivity and uncertainty analyses are then used to quantify uncertainties in the physical models
employed and prioritize the uncertainty drivers. This process is discussed in more detail in Section 3.
However, it is important to note that such analyses, since they are looking only at the sensitivities in the
physical models employed in the simulation, cannot be used in and of themselves to identify fundamental
structural uncertainties, or “unknown-unknowns, in those models. These “unknown-unknowns” must be
exposed through dedicated ground or flight testing. However, the objectives of this testing can be tailored
by pre-test analysis, which can predict the trends and sensitivities that should be observed if the models
employed are correct. This sequence of events allows for much greater focus of (presumably) limited
testing resources, which can now be spent targeting only those uncertainties that are a) reducible and b)
have a large impact on the performance of the overall TPS system.
2.1 Example: Mars Entry Convective Heating

In the current section, we examine key modeling gaps for Mars entry simulations. It is well known that the
surface recombination of gas-phase atoms and radicals due to the catalytic activity of the wall is often a
primary source of convective heating during Earth or planetary entry. However, the physical mechanism
that governs the catalytic recombination process is poorly understood for non-Earth entries. In addition,
the surface condition due to past history and exposure to other adsorbing species can alter the catalytic
properties, making it further difficult to quantify and address flight traceability issues. Consequently,
modeling of catalytic reactions at the surface has been relatively primitive, and constitutes a significant
modeling gap for Mars entry simulations. In fact, for Mars entries the so-called supercatalytic wall model
[16], in which the gas composition at the surface is specified to be equal to that in the freestream, is
frequently employed as a design assumption [17], primarily because it is easy to implement in a CFD
code, requires no knowledge of the actual surface processes, and provides a conservative upper bound to
the predicted heating level.
In general, surface catalytic reaction models, such as the Langmuir-Hinshelwood (LH) or Eley-Rideal
(ER) mechanisms, solve a surface site balance equation coupled with gas phase densities. In the LH model
the recombining species are adsorbed on the surface prior to recombining, while in the ER mechanism a
gas phase species recombines directly with an adsorbed species. Several prior simulations of Mars entry
aeroheating have assumed an ER mechanism to describe wall catalysis because of its simplicity. For
example, in the Mitcheltree and Gnoffo model [18], CO + O recombination is assumed to occur on the
surface via a two-step ER process. However, the literature indicates that the LH mechanism is more
commonly observed than the ER mechanism [19]. In fact, direct experimental evidence of ER mechanism
under any condition is lacking.
While there have been some attempts to obtain experimental data and build corresponding theoretical
models [20]–[24], none of these have yet described the surface chemistry of actual TPS materials at
temperatures and conditions relevant to Mars entries. However, it is possible to bound the expected heat
transfer rates by using a simplified flux-based model. We know that for moderate velocity Mars entry, the
freestream CO2 is almost completely dissociated into CO and O. Among the several possible surface
recombination reactions, the following two have been proposed [18] [25] as the principal reactions that are
likely to control surface catalytic heating:
17 - 4 RTO-EN-AVT-142
γ1
CO + O→ CO 2 ,∆E = 5.51 eV (1)
γ2
O + O→ O 2 ,∆E = 5.16 eV (2)
The parameters γ1 and γ2 are the reaction probabilities for reactions (1) and (2) respectively. However,
these parameters do not vary independently, since both reactions consume O-atoms. After some algebra
[26], it is possible to recast reactions (1) and (2) in a form that accounts for both diffusion and rate limiting
using two independent parameters: γcat, which controls the rate at which O-atoms are consumed at the
surface via either reaction, and the preference factor, p2, which determines the fraction of recombining O
atoms that follow reaction (2) as opposed to (1). Each of γcat and p2 can independently vary between 0 and
1. For γcat, a value of 0 corresponds to a non-catalytic wall, while a value of 1 corresponds to a fully
catalytic wall at which all inbound O-atoms recombine via reactions (1) and (2). For p2, a value of 0
indicates that all O-atoms that recombine do so via reaction (1), while a value of 1 indicates that reaction
(2) is exclusively preferred. By varying these parameters, the entire window of catalycity can be spanned.
For example, setting γcat = 0 results in a non-catalytic surface regardless of the value of p2. The fully
catalytic limit of the Mitcheltree model [18] can be reproduced by setting γcat = 1 and p2 = 0. It should be
noted that, to date, experimental studies on low temperature quartz and metallic surfaces [21][24], have
shown no conclusive evidence of reaction (1), indicating that p2 may in fact be quite large in flight
applications.
Number densities ( × 10 17 cm-3 )

2.5
120
Supercatalytic Wall 115 γcat=1
Moderately 110 2
p 2= 0 qw
qw ( W/cm2 )
Catalytic
qw ( W/cm2 )
100 Wall
105
1.5
100
80 p 2= 1
1
Weakly Catalytic
Wall
95 CO
90 O 0.5
60 Highly
Catalytic
Wall 85 0
40 -5 -4 -3 -2 -1 0 10-4 10-3 10-2 10-1 100
10 10 10 10 10 10 p2
γcat
(a) (b)
Figure 1 (a) Stagnation point heat flux variation for Mars Pathfinder with different levels of wall catalycity and
preference factors and (b) variations of heat flux and wall number densities of gas phase CO and O with the
preference parameter p2, (γcat =1).
The drawback of this approach is that the parameters γcat and p2 have unknown dependencies on factors
like temperature, density, and surface condition. However, until sufficient experimental data are obtained
to build a higher fidelity model, this approach is a useful tool to permit a systematic exploration of the
parameter space and related sensitivities. For example, Fig. 1a shows the variation of the peak stagnation
point heat flux (qw) for Mars Pathfinder with γcat at the two extremes of the preference factor p2. The
curves in Fig. 1a have a distinctive “S” shape that outline three distinct heating regimes as indicated by
the vertical dashed lines. The first is the high catalytic efficiency region (γcat > 0.1), where the heat flux
weakly varies with the value of γcat due to diffusion limiting. The second region in is the moderate
catalytic efficiency regime (10-3 < γcat < 10-1), where the heat flux is highly sensitive to γcat as the
recombination rate is surface-process rate limited. The third region in Fig. 1a is the weakly catalytic
regime (γcat < 10-3), where the exact value of γcat is again unimportant to heating since wall catalycity
contributes little to the total heat flux. Figure 1b shows the variation of the stagnation point heat flux and
RTO-EN-AVT-142 17 - 5
the gas phase number densities of CO and O at the stagnation point versus the preference factor, p2,
assuming a fully catalytic wall (γcat =1). When p2 is low, CO+O recombination is dominant, and both CO
and O are efficiently utilized. Figure 1b shows that a slight excess of O atoms (caused by unequal
diffusion rates) is seen. The extra O atoms can be further utilized if a small amount of O+O recombination
is allowed to occur by raising p2. This trend is evident in Fig. 1b as a drop in O density and a further rise
of qw as p2 is increased. Once p2 is raised above this critical value, the catalytic heating begins to sharply
fall, because the stoichiometric requirement of the catalytic reactions no longer matches the available
number density ratio of CO and O. As a result a rise in p2 is accompanied by an under utilization of CO
manifested as a rise in its density.
3.0 TYPES OF MODELING UNCERTAINTY

Over the years the nonequilibrium CFD and material response codes used for TPS sizing of entry vehicles
have been calibrated against each other [27] and validated with a variety of experimental data from ground
and flight tests [25], [28]–[31]. Based on these analyses, it can be clearly demonstrated that the predictions
made by these tools are in most cases highly sensitive to the physical, chemical, and numerical models
employed, as well as the multitude of input parameters that these models introduce [4]. As a result, the net
uncertainty in heating and TPS sizing predictions is a result of a combined effect of the uncertainties in all
of the models and input parameters used in the analysis. Therefore, in order to place confidence levels on
an aeroheating or TPS sizing calculation, all of the chief sources of input uncertainty must be identified
and quantified. The propagation of these uncertainties through the model must then be tracked to make
probabilistic estimates of the resulting quantities, in the form of a most probable result and a probability
distribution characterizing the variability of the prediction. The primary sources of uncertainty in the
thermochemical models used in aerothermal and TPS analysis, like other physical models, can be
classified into three categories: stochastic variability, parametric uncertainty, structural uncertainty. Each
of these is discussed below:
1. Stochastic Variability. Arises due to natural fluctuations in the physical environment. Stochastic
variability is also known as irreducible uncertainty, because it can be characterized, but not reduced, by
analysis or testing. Examples in TPS design include fluctuations in atmospheric composition,
temperature, or density, as well as small changes in entry flight path angle and vehicle orientation.
2. Parametric Uncertainty. Arises from the uncertainties in the input model parameter estimates.
Parametric uncertainties can generally be reduced, but not eliminated, via focused testing or theoretical
analysis. In an aerothermal CFD code, input variables include such things as kinetic reaction rates,
vibration-dissociation coupling parameters, vibrational-translational relaxation times, binary interaction
collision integrals for transport property calculations (diffusion, viscosity, and thermal conductivity
coefficients), wall catalycity parameters, and possibly even freestream conditions. In a TPS material
response code the input variables include virgin and char thermal conductivity, specific heats of the
various components, surface emissivity, recession rate, initial cold soak temperature, and properties of
the underlying sub-structure and overall material stack-up.
3. Structural Uncertainty. All numerical simulations employ mathematical models of the underlying
physical processes. Structural uncertainties arise when the models employed, or their discretization, are
incorrect or insufficient to adequately describe the phenomenon under study. Obviously, the potential
for structural uncertainties in a given simulation is larger in regimes for which the models have not
been properly validated. Examples of structural uncertainties include poor grid resolution or flux
discretization scheme, using the Euler equations to simulate a viscous flowfield, or ignoring shock
layer radiation for Apollo reentry simulations.
17 - 6 RTO-EN-AVT-142
3.1 Example of a Structural Uncertainty: Radiative Heating at Titan

Saturn’s largest moon Titan is a high priority target for future solar system exploration, with interest
fueled in large part by the successful ESA Huygens probe entry in January 2005. Several ambitious
follow-on surface missions have been proposed, including rovers and balloons. However, aeroheating
predictions for both direct entry [32][33] and aerocapture [34] Titan entry missions to this point have
indicated that a large portion of the total heat load during the entry could come from shock layer radiation.
This radiative heating is due to the unique composition of the Titan atmosphere, which consists primarily
of N2 with a small amount (~2% by volume) of methane. This composition produces significant amounts
of the CN radical behind the bow shock wave due to gas-phase chemical reactions. CN radiates strongly in
the violet and red bands of the spectrum, even at fairly low entry velocities. Figure 2a shows predicted
[35] convective and radiative heating levels to a typical Titan aerocapture vehicle with an entry velocity of
6.5 km/s. From the figure we see that the predicted radiative heating is nearly a factor of three higher than
the convective heating on the aeroshell. In comparison, radiative heating rates at this entry velocity would
be negligible for an Earth or Mars entry. Clearly these predicted heating rates, if real, will govern the TPS
material selection and thickness for future Titan entry missions.
All of these analyses assumed that the radiative emission from the CN radical was governed by a Boltzmann
(equilibrium) distribution of the low-lying excited states. While it was recognized [34] that this simple model
was not validated for Titan entries, it was selected because there were no applicable ground test or flight data
with which to develop a better model, and it was deemed likely (but not certain) that the Boltzmann
assumption would provide a conservative estimate of the resulting radiative heat flux. The assumption of a
Boltzmann distribution thus constituted a potentially large structural uncertainty for this analysis.
a) b)
Figure 2 a) comparison of convective (left) and radiative (right) heating rates on a potential Titan aerocapture
vehicle, and b) comparison of the Boltzmann and Collisional-Radiative (CR) models to experimental data for
Titan shock layer radiation.
In order to address this concern, shock tube data were obtained recently in the Electric Arc Shock Tube
(EAST) facility at NASA Ames and analyzed to measure radiation behind shock waves at velocities,
pressures, and gas compositions compatible with the expected flight environment [36]. The results of this
testing clearly indicated that the radiation intensity measured in the experiments was inconsistent with (in
fact much lower than) that predicted by the Boltzmann model, as shown in Fig. 2b. In the figure, distance
is measured downstream of the normal shock in the tube, and the vertical line represents the end of the
usable test time (data at larger distances from the shock are contaminated by impurities from the driver
RTO-EN-AVT-142 17 - 7
gas). These data indicate that the Boltzmann model (red line) for Titan radiative heating is in fact incorrect
for the relatively low pressures encountered during flight. A new non-local collisional-radiative model
(green line) was developed based on the shock tube data [36], which is in much better agreement with the
experimental data. Similar models were also developed recently at Ecole Centrale Paris [37] and EADS
[38]. While more work remains to be done to fully validate these models, these results make it clear that
detailed analysis and uncertainty estimation based on the Boltzmann model would be invalidated by the
structural uncertainty inherent in the erroneous selection of this model to simulate the Titan radiative
heating environment.
4.0 MONTE-CARLO UNCERTAINTY ANALYSIS

For systems where the uncertainties are small, a linear analysis may be used. In such an analysis, linear
sensitivity coefficients are first computed. The uncertainty in the output parameter is then described by the
law of propagation of errors:
2
 ∂y 
σ ( y i ) = ∑ i  σ 2 ( x k )
2
(3)
k  ∂x k 
where σ 2 (y i ) is the variance (uncertainty) in the value of the output parameter y i and σ 2 (x k ) is the
corresponding variance in the input parameter x k . Although computationally efficient, the linear analysis
outlined here is purely local, i.e. the analysis yields sensitivity coefficients only in the neighborhood of the
baseline values. However, in practice the variability in input parameters can be quite large in aerothermal
and material response analysis, and the underlying models may be significantly non-linear. Therefore, a
global uncertainty analysis is necessary.
Apart from the ability to handle non-linearities and large variabilities in the input and output parameters, a
global model allows for the simultaneous variation of potentially large numbers of input parameters in
order to account for uncertainty and sensitivity interference effects. While there are many possible
statistical approaches available for the analysis of non-linear systems, we have chosen the Monte-Carlo
technique, which relies on probability distribution sampling techniques to determine sensitivities and
uncertainties in the numerical model under study. While Monte-Carlo analysis converges slower than
some other non-linear methods (the convergence rate is proportional to the square root of the number of
samples), it has the advantage that the rate of convergence is independent of the number of independent
input variables considered in the model [3]. This feature makes the Monte-Carlo approach attractive for
this type of aerothermal/TPS sizing analysis, where there can be literally hundreds of independent input
modeling parameters in a given simulation.
In the Monte-Carlo methodology used here, the entire set of input parameters is varied independently and
input/output correlation coefficients are computed via linear regression analysis. This process requires that
large numbers (hundreds to thousands) of computational fluid dynamics (CFD) and material response
calculations be performed in order to generate the necessary statistics with sufficient accuracy. This type
of statistical analysis is commonly employed for entry trajectory design in the presence of stochastic
aerodynamic, entry state, and atmospheric dispersions [39][40]. However, the methodology has only
rarely been applied to TPS design [8][41], due primarily to the large cost associated with generating the
number of solutions required. Fortunately, with the fairly recent advent of large, low cost parallel
computing platforms and improved numerical algorithms designed specifically for their efficiency on such
machines, this problem is now tractable.
4.1 Monte-Carlo Methodology

The Monte-Carlo technique outlined in this lecture can quantify and track the propagation of parametric
and stochastic uncertainties that arise in a given physical model. On the other hand, structural uncertainties
17 - 8 RTO-EN-AVT-142
cannot be quantified or even directly identified by Monte-Carlo based analysis alone; they can only be
exposed via targeted ground or flight tests. In fact, a significant structural uncertainty in the model
employed can completely invalidate the results of a Monte-Carlo based analysis, because the predicted
trends and correlations that result will be those of the chosen model, not necessarily reality [8]. However,
the analysis approach presented here can be used to design experiments that will expose structural
uncertainties by displaying trends and sensitivities that are inconsistent with those predicted. The major
steps in a typical Monte-Carlo uncertainty analysis are briefly summarized here. For more information see
Refs. [7] & [8].
1. Parameter Identification. Input variables that need to be varied are first identified. Input variables
include all modeling parameters required by the physical and numerical models employed in the
simulation.
2. Initial Uncertainty Estimation. Variability limits for all input parameters are chosen that roughly
represent their typical uncertainties. The variability limits need not represent the true estimated input
uncertainties at this point, since they will only be used for an initial sensitivity analysis. Each input
parameter is independently varied about its baseline value using an appropriate probability distribution
function (typically Gaussian).
3. Sensitivity Analysis. A global sensitivity analysis is then performed by randomly varying each of the
input parameters to generate CFD and/or material response simulations. It is important that sufficient
runs are made for each case to ensure statistical accuracy of the resulting sensitivities. Typically
hundreds to thousands of runs are required, depending on the desired accuracy of the output data.
4. Correlation Coefficient Computation. Input-output correlation coefficients are computed using linear
regression analysis and the fractional contribution of each input variability to the overall output
variability is obtained. For most cases the majority of the correlation is due to a small number of input
parameters; the vast majority display minimal sensitivity, and thus typically warrant no further
analysis. This step allows us to short list a small subset of the input parameters for a more detailed
investigation.
5. Final Input Uncertainty Estimation. A more accurate estimate of the associated uncertainties for the
reduced list of input parameters is then made. This can be one of the most time-consuming parts of the
entire process, which is why a detailed assessment is deferred until the input uncertainties have been
short listed via a sensitivity analysis. Methods of estimation for input parameter uncertainties are
discussed in more detail in Refs. [7] and [26]. Unfortunately, little if any quantitative data are available
for many of the parameters used in modern computational aerothermodynamics and material response
models, and frequently expert judgment is the only method available.
6. Uncertainty Analysis. A second set of CFD and/or material response calculations is then made using
the input uncertainties from the previous step. The function of this step is to create a database of
simulation data that can be statistically analyzed.
7. Apportionment of Output Uncertainty Into Input Parameters. At the completion of Step 6, the
variability in the output quantity of interest represents the true parametric uncertainty of the model to
the desired level of fidelity. Finally, input-output correlations are again computed and ranked to
apportion the output uncertainty into those of input parameters.
The most common approach for Step 7 (output uncertainty apportionment) is to use linear regression
analysis. However, more complex approaches, which account for multi-level interference effects, can be
employed if desired. Also, this step is only used in post-processing to identify the most important input
RTO-EN-AVT-142 17 - 9
sources of final model uncertainty; the Monte-Carlo analysis method employed automatically includes all
interference effects in the rolled up uncertainty value predicted.
Finally it should be noted that, although all input variables are typically assumed to vary independently,
the Monte-Carlo technique as formulated does not require this. Relationships, or correlations, between
input parameters can readily be modeled in the analysis if the functional forms of the relationships are
known a priori. Input correlations can also be detected using principal component analysis (PCA), by
which a large set of correlated input variables can be transformed into a smaller set of uncorrelated
variables by means of eigenvalue analysis of their correlation matrix [42].
4.2 Example: Mars Pathfinder Convective Heating

The Mars Pathfinder aeroshell was a 2.65 meter diameter 70° sphere-cone, which entered the atmosphere
of Mars on a ballistic (non-lifting) trajectory at a relative velocity of 7.5 km/s on July 4, 1997. Pre-flight
calculations predicted a peak convective stagnation point heat flux of 110 W/cm2 (Ref. [43]). As
demonstrated in Ref. [26] and discussed above, the largest uncertainty for Mars laminar convective
heating is the catalycity of the TPS material. A detailed uncertainty analysis of Mars Pathfinder laminar
entry convective heating at the peak heating point on the trajectory was performed by Bose and Wright
[26]. Since the actual catalytic properties of the surface are unknown, separate analyses were performed
for each of the catalytic regimes defined by the surface reaction model discussed above and shown in Fig.
1. In addition to γcat and p2, a total of 128 other independent parameters were varied, including chemical
reaction rates, vibration-dissociation coupling parameters, vibrational relaxation times, and the binary
collision integrals that make up mixture transport properties. A total of 3000 CFD runs were performed for
each analysis. The nominal heating rate (Table 1) varied from 121 W/cm2 for a supercatalytic wall to
about 47 W/cm2 for a weakly catalytic surface, a factor of 2.5 in the predicted heat flux. Uncertainty
estimates on the heat flux were also determined in each catalytic regime. The largest uncertainties by far
are for the moderately catalytic surface.
Table 1 Uncertainty results for the stagnation point heating of Mars Pathfinder.
Level of Catalycity γcat qw 95% confidence limits

(W/cm2) (%)
Supercatalytic – 120.6 +10.3 -9.9

Highly catalytic 10-1–100 106.7 +12.0 -17.2
Moder. catalytic 10-3–10-1 74.0 +41.0 -33.6
Weakly catalytic 10-5–10-3 47.2 +11.7 -10.6
Figure 3 shows the key input contributors to the total uncertainty as determined via linear regression
analysis. It is clear that only a small number of the 130 input parameters are significant contributors to the
uncertainty in heat flux. In Fig. 3 all parameters that individually contribute more than 5% to the total
uncertainty are labeled with the total contribution in parentheses. The binary collision integrals, which are
used to compute the species and mixture transport properties, are denoted by the two interacting species
separated by a dash (–). For the limiting case of a supercatalytic wall there is no variation in the catalytic
parameters (γcat and p2) and nearly all the uncertainty in the computed heat flux comes from a small
number of collision integrals that govern the rate of diffusion of the reactants to the surface. The highly
catalytic wall is in the diffusion limited regime, and thus the majority of the uncertainty again comes from
collision integrals, although the preference factor p2 is also important. For the moderately catalytic wall
nearly all of the uncertainty comes from γcat, indicating that we are in a rate-limited regime at these
17 - 10 RTO-EN-AVT-142
conditions. Finally, for the weakly catalytic wall significant uncertainty arises from collision integrals
which govern the thermal conductivity, as well as the rate of a single chemical reaction O2 + O ' 2O + O,
which affects heat release in the boundary layer.
CO2-CO ( 10% ) CO-O ( 18% ) CO2-O ( 12% )

CO-O ( 14% )
γcat ( 12% )
Other (13% )
p2 ( 17% )
Other (20% )
CO2-CO ( 32% )
CO2-O ( 59% )
a) b)
O2+O↔2O+O ( 27% )
γcat ( 12% )
Other (21% )
Other (6% )
γcat ( 94% )
CO-O ( 41% )
c) d)
Figure 3. Principal contributors to Pathfinder heat flux uncertainty for a) super-, b) highly, c) moderately, and
d) weakly catalytic wall assumption (from Ref. [26]).
From a design standpoint, it is clear that an improved understanding of surface catalysis for Mars entries
could have a significant impact on TPS selection and design for future missions. One of the strengths of
the technique presented here is that it can help to determine how much improvement is required. For
example, if focused testing (or flight data) determined that a given material performed either as a highly
catalytic or weakly catalytic surface, the current analysis indicates that further refinement in our
knowledge of γcat may not be necessary, and additional research monies could be targeted to other risk
drivers. However, if the material were determined to be moderately catalytic, the resulting heating
uncertainties could be greatly reduced if the input uncertainty of γcat were better refined.
5.0 TPS GROUND TESTING

Although CFD and material response analysis tools have become much more sophisticated and capable in
recent years, no TPS system is designed entirely with analysis. Ground testing in a relevant environment is
a vital part of the TPS design process, and is used to screen materials in order to determine their suitability
for a given mission, collect data to construct high-fidelity material response models, and qualify the
chosen material and specific installations (penetrations, struts, tile gaps, etc.) for the flight environment.
Ground testing is also used to validate the computational tools and methodologies employed in the flight
RTO-EN-AVT-142 17 - 11
predictions, and to identify and eliminate structural uncertainties in these models. In order for the high-
fidelity aeroheating and material response simulations to be used with confidence in TPS design and risk
analyses, they must be verified and validated (V&V) for the specific application. Verification addresses
whether the numerical models are implemented correctly [44], while validation addresses whether the
simulations model the physics accurately [45]. Although the specific objectives of a given test vary
considerably, the primary objectives of the overall test program should be to lower the risk profile and
maximize the performance for the mission the TPS supports.
Rate constants,
transport Material properties
properties
Thermochemical, Material,
transport, radiation thermostructural
models response models
Environment Flight Surface boundary Material,

inputs (trajectory, environment conditions ( heat flux, structure Risk
atmosphere) simulation pressure, species) response analysis
simulation
coupling
Vehicle design
Vehicle design
(structure,
(geometry)
material thickness)
Figure 4. Elements of an analytical simulation that define the aerothermal environment and material response
of a TPS. The output is used to assess performance and risk of a TPS subsystem.
The first step in developing a sound test strategy is to identify places where key validation gaps exist. A
diagram indicating some (but not all) elements of an analytical simulation of TPS loads and performance
is shown in Fig. 4. The analytical simulation in this context is a collection of specialized simulations
supported by phenomenological models. The output is used as a performance and risk assessment metric
for the TPS subsystem. TPS sizing margins and factors of safety, for example, are established based in
part on the estimated uncertainties of various input parameters and test conditions. Testing should focus on
reducing uncertainties that lead to the greatest risk, as determined by sensitivity and uncertainty analyses
as described in Section 4. Test plans should also be constructed to maximize the chance of exposing
suspected structural uncertainties in the physics models employed, and to minimize the need to extrapolate
a given model from its validated range.
5.1 Facility Characteristics

High enthalpy ground testing can address uncertainties of several elements in Fig. 4. The approach
depends on the features of the aerothermal, thermochemical, or thermostructural processes to be
stimulated through testing. In each case, the information obtained from a test program can contribute to
the validation of the overall analysis process. Improving phenomenological model accuracy,
understanding failure mechanisms, or characterizing variabilities in material performance are but three
examples of the ways in which analytical tools can be validated. The validation gaps and uncertainties in a
particular element will define test objectives. Fulfilling those test objectives raises issues concerning the
appropriate test facility, test conditions, and the test article design.
17 - 12 RTO-EN-AVT-142
Ideally the performance of a material or system should be evaluated in a ground test simulation that
closely matches the all of the significant flight environment parameters simultaneously. A well-designed
ground test program would then cover the entire range of expected conditions and would fully validate the
material and design for the chosen mission. However, the wide variety of length and time scales involved
presents difficulties in matching relevant aerothermodynamic similarity parameters in a high enthalpy
ground test simulation. Laminar stagnation point convective heat flux scales as R-1/2 while Reynolds
number scales as R, where R is the effective radius of curvature of a blunt-body test article. This limits
options for matching enthalpy and more than one other similarity parameter. The nonequilibrium
characteristics of high enthalpy facilities also present challenges for simulation. Time scales become
important due to finite reaction and diffusion rates. Time scales impact length scales in flows with very
high velocities and further complicate efforts to achieve similarity [46]. Additionally, the heat flux to a
partially catalytic surface is sensitive to the thermodynamic state at the boundary layer edge which, in
turn, can depend on the degree of thermochemical nonequilibrium in the freestream [47]. More
importantly, some mission-specific aspects can be simulated only partially or not at all [11]. For example,
time-varying conditions of an entry profile cannot be achieved in most facilities. Arc plasma facilities that
can accommodate non-terrestrial gas mixtures (e.g. CO2/N2 for Mars and Venus) at relevant enthalpies and
test article sizes may not be available. The capability to simulate combined radiative-convective heating
environments, such as would be encountered during a crewed lunar or Mars return entry, no longer exists
in the United States.
Figure 5. High enthalpy test facility capabilities and applications.
Figure 5 shows the types of facilities and the elements of TPS development they are capable of simulating.
Both the enthalpy range and the testing time influence their role in validation strategies. With their short
testing times, impulse facilities (e.g. shock tubes, shock tunnels, and ballistic ranges) are suitable primarily
for studying aeroheating as well as gas-kinetic and radiation processes. Material response mechanisms
have much slower timescales and can only be evaluated in facilities that operate for long durations.
Although several types of long-duration facilities, such as radiative lamps, lasers, and combustion-driven
flows, are routinely used to achieve specific test objectives, arc jets have been the workhorse facilities for
the past 40 years to support material response model development and subsystem-scale hardware
performance validation. Arc jets provide the means for TPS materials to reach flight temperatures under
conditions where surface chemistry effects approximate that of flight. Large-scale, low-pressure arc jet
RTO-EN-AVT-142 17 - 13
facilities in particular accommodate near flight-scale test articles, permitting simulations at appropriate
enthalpies and laminar Reynolds numbers. For this reason they are the primary test facilities that are used
to flight-qualify TPS materials for civilian entry vehicles within NASA. A survey of large-scale arc-heated
facilities worldwide can be found in Ref. [48].
5.2 Arc Jet Testing Methodology

In an arc jet, a series of electrodes are used to arc heat the gas prior to expansion. This heated gas is then
passed through a nozzle to expand the flow to reach temperatures and pressures of interest to flight. Arc
heating of the gas stream produces a significant amount of dissociation and ionization. In the plenum
region prior to the nozzle expansion, the gas is in thermo-chemical equilibrium at temperatures on the
order of 7000 K. During the expansion through the nozzle, the flow accelerates rapidly and the stream is
left in a state of thermo-chemical nonequilibrium as various kinetic process freeze at different times due to
the differing timescales of thermal and chemical relaxation in the flow. For air flows, a significant portion
of the total flow enthalpy remains in dissociated N2 and O2; the flow is chemically frozen at temperatures
for which much less dissociation would exist in chemical equilibrium. This increase in chemical enthalpy
reduces the velocity of the stream – energy locked in chemical nonequilibrium is unavailable as kinetic
energy. The dissociated free stream reduces the Mach number due to both a lower velocity and higher
sound speed. This affects the shock standoff distance observed in the test. Dissociation also increases the
density over what would result at equilibrium for the same total enthalpy. The composition at the
boundary layer edge is also affected, altering the amount of chemical energy available for surface
recombination. Therefore, matching total enthalpy and surface pressure in flight and arc jet environments
does not necessarily yield the same heat flux to a partially catalytic material [47]. This potential difference
in heat flux underscores one of the principal challenges in establishing ground-to-flight traceability.
In all cases, the confidence in any causal relationship between test conditions and test observables rests
upon the accuracy to which the test conditions are known. The selection of appropriate conditions for a
given test requires detailed knowledge of a facility’s operating characteristics, with enthalpy and pressure
being most important among the relevant parameters. Pressure is typically determined with pitot probe
surveys of the test section. Unfortunately, enthalpy can be particularly difficult to measure in an arc jet,
and significant discrepancies are frequently observed between different intrusive and non intrusive
measurement techniques. In complement to experimental techniques, high fidelity simulation techniques
for arc jet flows can provide unique insight into a particular facility’s nonequilibrium thermochemistry;
this type of information is typically unavailable via conventional flow diagnostics. Considerable attention
has been devoted in recent years to understanding the operating characteristics of arc jet facilities [49][50].
Applied to the test planning process, arc jet flow simulation is a tool that can be employed to rapidly
explore potential test configurations and optimize the return on investment in testing resources to meet
program objectives. The merits of a particular test design can be identified and understood early in the
planning process. Most importantly, by using the same CFD flow solver for both arc jet and flight
environment simulation, consistency between flight environment definition and test interpretation can be
maintained – a critical need in resolving ground-to-flight traceability.
An end-to-end simulation of an arc jet test involves 1) a nozzle calculation that defines the arc jet free
stream conditions, 2) a model calculation of the flow over the test article, and 3) a transient thermo-
structural analysis of the test article including material thermal response. Facility data (such as flow rates)
and test observables (such as pressure, heat flux, model temperatures, and emissivity) are incorporated as
boundary conditions. An example of a virtual arc jet simulation is depicted in Fig. 6. The simulation is of a
typical candidate TPS material “flat-face” coupon tested in a stagnation configuration. The overall test
simulation procedure is as follows. First, stream conditions are determined via a CFD simulation of the
nozzle flow, based on prescribed (from facility data) conditions at the nozzle throat. A second CFD
calculation is then performed of the flow over the test article with prescribed surface boundary conditions,
using freestream conditions extracted from the nozzle simulation. Environment parameters that cannot at
17 - 14 RTO-EN-AVT-142
present be measured (or are not measured routinely) such as shear stress, Reynolds number, or species
mass fractions, can then be extracted directly from solutions. The surface temperature time history is
typically a primary metric by which the performance of a TPS material is assessed. However, the surface
temperature is controlled in large part by the efficiency with which the incident heat flux is dissipated by
other loss mechanisms at the surface: re-radiation, conduction, and, in the case of ablators, surface
reactions, pyrolysis, and mass transfer. Also, the thermal properties of the test article and the model holder
also can have a large effect on the results. Therefore, thermal analysis of arc jet model designs is a crucial
part of any test planning process to ensure that relevant thermal losses are incorporated.
Figure 6. Virtual arc jet simulation. Facility operating parameters are inlet conditions for the CFD nozzle
calculation. Nonequilibrium free stream conditions from the nozzle solution become the upstream boundary
condition for the CFD solution of the test model. The thermal response of the model is then computed with a
finite element simulation using boundary conditions from the model solution.
Without validation of the arc jet CFD simulations, test observations may not be accurately – or correctly –
captured in the material response models. Detailed chemical and physical measurements of arc jet flow
properties and their spatial distribution could help reduce uncertainties in the simulations. Accounting for
all contributions to the enthalpy at a particular point requires several measurements, with the minimum
being velocity, temperature, and species densities. One of the more useful tools for performing
nonintrusive optical measurements in arc jet flows is laser-induced fluorescence (LIF), a spatially
resolved, species-selective probe of individual atomic and molecular states. LIF of atomic nitrogen has
been applied to directly measure the velocity, translational temperature, and nitrogen number density in
the free stream of arc jet facilities [51][52]. The kinetic, thermal, and chemical modes of the total enthalpy
can be quantified with the LIF data and other facility measurements [51]. Optical diagnostics have also
been developed for the shock layer region ahead of blunt body test articles. Emission spectra have been
analyzed in an effort to characterize flow properties within the shock layer and their relation to the free
stream. Spectra at multiple points along the stagnation streamline were compared to those from
simulations to assess the degree of chemical or thermal nonequilibrium within the shock layer [53]. Time
resolved emission spectra and optical attenuation ahead of an ablating test model have been recorded with
novel fiber optic- and laser-based instruments [54]. Other advanced non-intrusive techniques are also
under development, including methodologies for measuring time-dependent recession during testing.
These diagnostics provide insight on the transient performance of ablative TPS materials that can support
the development of improved gas-surface interaction and in-depth thermal response models.
5.3 Example: Mars Exploration Rover TIRS Cover Testing

The twin Mars Exploration Rover (MER) entry vehicles successfully landed on Mars in January 2004. The
MER aeroshell was initially intended to replicate that of Mars Pathfinder. However, one significant
change was the addition of three Transverse Impulse Rocket System (TIRS) motors on the backshell,
RTO-EN-AVT-142 17 - 15
intended to minimize lateral velocity while on parachute and at ground impact. Because these TIRS
motors (of necessity) protruded through the backshell, TPS covers were required to protect the motors
from the entry environment. These covers were designed to be jettisoned before the TIRS became active
on terminal descent. The design of the protective cover and its interface with the backshell was validated
through arc jet testing in the 20 Megawatt Panel Test Facility (PTF) at NASA Ames Research Center [55].
Heating environment predictions from flight CFD solutions including the TIRS covers defined the heat
flux, heat load, and pressure on the cover and at the interface seal. The maximum local heat flux to the
cover occurs on the aft end due to the large recirculation region on the backshell. Final sizing for the TIRS
cover TPS was based on these analyses and included margins due to uncertainties arising from transition
to turbulence and aerothermal environment variations.
a) b)
Figure 7. a) TIRS cover calibration model in the PTF arc jet. b) Comparison of measured and simulated heat
flux along the TIRS cover centerline for the calibration test. The origin (X = 0) begins at the nozzle exit.
Design validation of the TIRS cover seal at its interface with the backshell was a primary test objective. A
full-scale flight article was tested at flight-representative heat fluxes and pressures along the gap between
the cover and simulated backshell. A pressure differential across the seal was maintained, and the test
model design incorporated a witness plate below the cover attachment as a means to monitor seal
performance. The large, recirculating flow that heats the aft end of the cover could not be reproduced with
relevant similarity in an arc jet test. Instead, turning the cover to face the stream approximated the
predicted heating distribution. However, because of the differing complex 3-D flow features of both the
arc jet and flight environment, the heat flux and pressure distributions on the cover did not compare
favorably at all locations. This test case illustrates the difficulty in maintaining ground-to-flight
traceability when it is impossible to simultaneously reproduce all aspects of the flight environment. In this
case, the primary objective was to evaluate the thermostructural performance of the seal, and thus
necessary compromises were made in other areas. CFD simulations of the nozzle flow with the TIRS
cover test article were used to determine the test conditions (facility flow rate, enthalpy, and test panel
inclination) that best matched the predicted peak heat load and surface pressure along the gap region. A
calibration test article instrumented with calorimeters and thermocouples was used to validate these pre-
test analyses. The calibration model under test is shown in Fig. 7a. Centerline heat flux measurements
compare favorably with the CFD solution (Fig. 7b). Subsequent tests at the calibrated test condition were
used to assess the performance of the TIRS cover seal. The connection between the test and flight
conditions for the heat flux and pressure along the seal was defined through the CFD simulations. The
calibration test confirmed the pre-test prediction and demonstrated that CFD can be used successfully as a
test planning tool.
17 - 16 RTO-EN-AVT-142
5.4 Example: Wing Leading Edge Testing

A test series was recently performed in the 60 Megawatt Interaction Heating Facility (IHF) at NASA
Ames to evaluate the performance of a candidate RLV wing leading edge TPS design. Because of the
complex nature of this test series, three-dimensional arc jet CFD simulations were essential to establish
appropriate conditions and test article design for a representative arc jet test that would best simulate a
wing leading edge flight environment [56]. The primary objective of the test was to evaluate the
thermostructural performance of a candidate leading edge TPS assembly, including attachments, gaps and
seams. The IHF arc jet, when equipped with a semi-elliptic nozzle, provides a boundary layer flow
appropriate for reaching high Reynolds numbers with boundary layer edge characteristics comparable to
flight. A swept pylon test model was designed to simulate the flow along the curved leading edge and its
intersection with the incoming boundary layer. The arc jet operating parameters and final model
dimensions were chosen through the pre-test CFD analysis to match the flight predictions of heat flux,
boundary layer edge Mach number, boundary layer thickness, and Reynolds number as closely as possible.
Figure 8 is a photograph of the pylon leading edge model during an IHF arc jet test. The thickness of the
incoming boundary layer and the leading edge shock location can be inferred visually from the change in
emission intensity of the stream.
Leading edge shock
Boundary layer
Semi-elliptic nozzle Pylon test model
Figure 8. Arc jet test of a test model used to simulate wing leading edge heating. The boundary layer
thickness and leading edge shock location can be discerned by the change in the emission intensity of the
free stream.
The simulation and test results were used in a finite-element analysis to assess the thermostructural
response of a candidate leading edge system. The test model included features of a leading edge assembly
that required performance validation. The heat flux and pressure distributions from the CFD solution were
applied as boundary conditions for the finite element simulation. The CFD simulation was necessary to
optimize the test design and facility operating conditions, confirm measured material surface temperatures,
and define the boundary condition distributions for the finite element analysis.
6.0 FLIGHT TESTING

Although most TPS systems are designed exclusively with analysis and ground testing, flight testing
remains the only way to fully validate the performance of a full-scale thermal protection system in a new
environment. In addition, certain structural uncertainties in the aeroheating or material response model can
only be exposed through flight testing. Unfortunately, the prohibitively high cost of full-scale flight testing
RTO-EN-AVT-142 17 - 17
implies that such tests, if carried out at all, must be carefully designed and executed. Such testing should if
possible be reserved for final model and system-level validation, once we have good physics-based models
of the expected environment and resulting material performance based on ground-test and historical data.
However, valuable information on TPS flight performance can frequently be obtained via much less
expensive sub-scale tests, as long as these tests are well designed to ensure that they reasonably simulate
the full-scale flight environment.
6.1 Example: Apollo Flight Test Program

An excellent example of a well designed flight test program was that carried out in support of Apollo; the
primary test objectives relating to aerothermal analysis and TPS validation are briefly reviewed here. The
Apollo Command Module was the first human-rated lunar return entry vehicle, and it was also the first
spacecraft that employed a mid-density ablator (Avcoat) for primary thermal protection. As such, an
extensive series of flight tests were planned to obtain data for model development and to validate the
performance of a new TPS material for such an environment. A total of six large scale flight tests were
flown. The first two, called Project Fire, had a primary objective to collect data on shock layer radiative
heating at lunar return velocities [57]. The motivation for this program was the large disparity in radiative
heating design predictions for the Apollo return capsule. Several models for shock heating were employed
during the Apollo design phase, and predictions of the total heating rate from shock layer radiation varied
by more than a factor of three, [58] indicating the presence of a possibly large structural uncertainty in one
or all of the models. While some validation data were obtained in ground-test facilities (primarily shock
tubes), these data were not considered sufficient to resolve this issue to the level of confidence required for
a human-rated system. Unfortunately, flight data from the Fire-I flight were compromised by a stage
separation problem [59], but radiative heating measurements from Fire-II were generally in good
agreement with the lower bound of pre-flight predictions. These data have since been used for validation
and calibration of a variety of nonequilibrium air radiation models [60].
During the development phase of the low-density Avcoat TPS material, ground tests were conducted in
almost every arc jet facility in the United States. However, attempts to correlate these data into a single
material response model were frustrating, because the results from different facilities indicated vastly
different characteristics, ranging from catastrophic failure to excellent performance [61]. Early sub-scale
flight experiments showed similar variability in material performance. One notable example was the R-4
flight test [62], in which the measured ablation rate was much higher than predicted; in fact, the entire
ablative nosecap was eroded away during the entry. The root cause of the catastrophic failures observed
during ground and flight testing was eventually determined to be the extremely high dynamic pressures of
these tests. For example, the R-4 flight test was conducted at dynamic pressures three times that expected
during Apollo entry, in an attempt to increase the heat transfer rate. Although the root cause of the
problems was eventually determined, a great deal of effort was expended to prove this fact to the
satisfaction of all involved. One of the primary lessons learned from this experience was that ground or
flight testing in non-representative environments can be highly misleading and cause unwarranted concern
[62]. In other words, a poorly conceived test can be far worse than no test at all.
In part because of the concerns raised during ground testing of the TPS, a series of four flight tests of the
full scale Apollo Command Module were flown: two at orbital velocity [63], and two at superorbital
velocity representative of lunar return [64]. Some of the primary objectives of the flight tests were to
obtain aeroheating and TPS material response data to validate the analytical design tools and the overall
performance of the ablator system. Each of these capsules was heavily instrumented with calorimeters
(surface and in-depth), char sensors, pressure transducers, and shock-layer radiometers [62]. In addition,
the heatshields were recovered and core samples of the post-flight TPS material were analyzed extensively
[65][66]. The results of this analysis and flight data reduction indicated that the material performed
extremely well under lunar entry conditions. In fact, measured performance was much better than pre-
flight predictions. Among other things, the total TPS material recession more than an order of magnitude
17 - 18 RTO-EN-AVT-142
less than even the most conservative predictions. The reasons for this discrepancy were eventually traced
to a significant structural uncertainty in the existing material response models (none accounted for the char
coking phenomenon [61]). Once the models were updated to account for this effect, the agreement with
flight data was excellent [66].
Finally, it is important to note that there are other ways to obtain useful flight data other than full-scale
tests. Sub-scale testing, via sounding rockets or similar launch platforms, can provide valuable data for
component design and analysis for a (relatively) small cost. Placing instrumentation on the heatshield of
science (e.g. Mars Pathfinder [25]) and human exploration (e.g. Shuttle Orbiter [67]) missions once was
standard engineering practice. These data provide valuable model validation information at a relatively
low per-mission cost. Unfortunately, recent practice has been to eliminate such instrumentation as a cost
savings measure. For example, neither the ESA Huygens probe nor the NASA MER rovers carried any
active instrumentation on the heatshield. Each mission would have provided a prime opportunity to
investigate key modeling gaps for future Titan and Mars entries, respectively, as discussed in previous
sections. The decision to eliminate instrumentation on recent missions was primarily motivated by cost-
savings. This decision is somewhat understandable given that, in general, a given mission does not benefit
directly from the engineering instrumentation that it carries. Therefore, from the standpoint of the mission
manager, inclusion of heatshield instrumentation adds cost, complexity and risk to the entry without a
corresponding mission benefit. The benefits become apparent only when looked at from the viewpoint of
an overall exploration program, for which the engineering data from each individual mission are laying the
groundwork for the presumably more complex missions to follow.
7.0 SUMMARY AND CONCLUSIONS

A methodology was presented for the risk based design and analysis of thermal protection systems for
Earth and planetary entry vehicles. The first step in the process is the identification of key modeling
deficiencies via a mission-specific gap analysis. Once these modeling deficiencies are identified, Monte-
Carlo based sensitivity and uncertainty analyses are conducted to quantify parametric and stochastic
uncertainties in the models employed. It is important to note that structural uncertainties, or basic
deficiencies in the models employed for simulation, cannot be identified by this approach alone. However,
the results of a sensitivity analysis can be used to design tests to expose these structural uncertainties by
quantifying trends that should be reproducible by a well designed test series. These techniques also
facilitate a risk-based probabilistic design approach, whereby the TPS can be designed to a desired risk
tolerance level, and remaining risk can be effectively compared to and traded with that of other
subsystems via a system level risk mitigation analysis. Modeling sensitivities, which are a by-product of
the uncertainty analysis, can be used to rank input uncertainty drivers, which can then be prioritized and
targeted for further analysis or testing. Examples of structural uncertainty identification and parametric
uncertainty quantification are given for Mars and Titan entry problems. These examples demonstrate the
utility of the methodology to quantify the uncertainty levels, rank sources of input uncertainty, identify
structural uncertainties in the models employed, and probabilistically design a TPS system for a planetary
entry mission.
Although high-fidelity analysis is used increasingly earlier in the design cycle as the tools and methods
become faster and more accurate, ground-based testing remains a critical component of the design of any
TPS system. Given that it is seldom possible to simultaneously match all aspects of the flight environment
in any ground test facility, great care must be taken in the design of any test series to ensure traceability to
the flight environment. The various types of ground test facility commonly employed in TPS design are
discussed briefly, but the paper focuses on arc jets, which have been the primary TPS testing workhorse
facility for the past 40 years. This paper outlines a testing methodology designed to ensure that the tests
performed are flight-relevant and focused on driving down key risks and/or uncertainties in the flight
design. This methodology is based on experimental diagnostics (both intrusive and non-intrusive),
RTO-EN-AVT-142 17 - 19
augmented by high-fidelity simulations of the arc jet test environment. Examples are given for the design
of flight hardware on the Mars Exploration Rovers, and the preliminary design of an RLV wing leading
edge concept. Finally, the role of flight testing in TPS design is discussed. The high cost of full-scale
flight testing generally limits its role to one of final model validation in a new flight environment.
Examples are given from the Apollo program highlighting both the benefits of a well constructed flight
test program, and the dangers of poorly conceived flight experiments.
ACKNOWLEDGEMENTS
The authors would like to thank Y.-K. Chen (NASA Ames), Peter Gage (ELORET), Tahir Gökçen
(ELORET), Dean Kontinos (Ames), Bernard Laub (Ames), Ryan McDaniel (Ames), Grant Palmer
(ELORET), Dave Stewart (Ames), and Michael Wilder (Ames) for manuscript suggestions and
improvements. Technical discussions with Joe Marschall (SRI) were extremely helpful in understanding
Mars surface catalysis models. Finally, the authors thank Michelle Munk (NASA Marshall) and Bonnie
James (Marshall) of the NASA In-Space Propulsion Program for their unwavering support of the Monte-
Carlo uncertainty analysis work.
REFERENCES
[1] Hall, J.L., Noca, M.A., and Bailey, R.W., “Cost-Benefit Analysis of the Aerocapture Mission Set,”
Journal of Spacecraft and Rockets, Vol. 42, No. 2, 2005, pp. 309-320.
[2] Thacker, B., Riha, D., Millwater, H., and Enright, M., “Errors and Uncertainties in Probabilistic
Engineering Analysis,” AIAA Paper No. 2001-1239, Jan. 2001.
[3] Garzon, V. and Darmofal, D., “Using Computational Fluid Dynamics in Probabilistic Engineering
Design,” AIAA Paper No. 2001-2526, Jun. 2001.
[4] Gnoffo, P.A., Weilmuenster, K.J., Hamilton, H.H., Olynick, D.R., and Venkatapathy, E.V.,
“Computational Aerothermodynamic Design Issues for Hypersonic Vehicles,” Journal of Spacecraft
and Rockets, Vol. 36, No. 1, 1999, pp. 21-43.
[5] Avco Corporation, “Ablation Handbook, Entry Materials Data and Design,” Technical Report
AFML-TR-66-262, Sep. 1966.
[6] Coleman, W.D., Hearne, L.F., Lefferdo, J.M., Gallagher, L.W., and Vojvodich, N.S., “Effects of
Environmental and Ablator Performance Uncertainties on Heat-Shielding Requirements for
Hyperbolic Entry Vehicles,” Journal of Spacecraft and Rockets, Vol. 5, No. 11, 1968, pp. 1260-1270.
[7] Bose, D., Wright, M.J., and Gökçen, T., “Uncertainty and Sensitivity Analysis of Thermochemical
Modeling for Titan Atmospheric Entry,” AIAA Paper No. 2004-2455, Jun. 2004.
[8] Wright, M.J., Bose, D., and Chen, Y.-K., “Probabilistic Modeling of Aerothermal and Material
Response Uncertainties for Aerocapture Using Rigid Aeroshells,” 53rd JANNAF Joint Propulsion
Meeting, Dec. 2005.
[9] Park, C., “Laboratory Simulation of Aerothermodynamic Phenomena: A Review,” AIAA Paper No.
92-4025, Jul. 1992.
[10] Grinstead, J., Stewart, D., and Smith, C., “High Enthalpy Test Methodologies for Thermal Protection
Systems Development at NASA Ames Research Center,” AIAA Paper No. 2005-3326, May 2005.
17 - 20 RTO-EN-AVT-142
[11] Laub, B., Balboni, J., and Goldstein, H., “Ground Test Facilities for TPS Development,” NASA TM-
2002-211400, May 2002.
[12] Palmer, G.E., Kontinos, D.A., and Sherman, B., “Surface Heating Effects of X-33 Vehicle TPS Panel
Bowing, Steps, and Gaps,” AIAA Paper No. 98-0865, Jan. 1998.
[13] Reuther, J., Prabhu, D., Brown, J., Wright, M., and Saunders, D., “Computational Fluid Dynamics for
Winged Reentry Vehicles at Hypersonic Conditions,” AIAA Paper No. 2004-2537, Jun. 2004.
[14] Richardson, E., Munk, M., and James, B., “Review of NASA In-Space Propulsion Technology
Program Inflatable Decelerator Investments,” AIAA Paper No. 2005-1603, May 2005.
[15] Gulick, D., Lewis, J., Trochman, B., Stein, J., Lyons, D., and Wilmoth, R., “Trailing Ballute
Aerocapture: Concept and Feasibility Assessment,” AIAA Paper No. 2003-4655, Jul. 2003.
[16] Papadopoulos, P., Prabhu, D., Olynick, D., Chen, Y.-K., and Cheatwood, F.M., “CFD Code
Comparisons for Mars Entry Simulations,” AIAA Paper No. 98-0272, Jan. 1998.
[17] Edquist, K., Liechty, D., Hollis, B., Alter, S., and Loomis, M., “Aeroheating Environments for a
Mars Smart Lander,” AIAA Paper No. 2002-4505, Aug. 2002.
[18] Mitcheltree, R.A., and Gnoffo, P.A., “Wake Flow about the Mars Pathfinder Entry Vehicle,” Journal
of Spacecraft and Rockets, Vol. 32, No. 5, 1995, pp. 771-776.
[19] Christmann K., Introduction to Surface Physical Chemistry, Springer-Verlag, New York, 1991.
[20] Kolesnikov, A., Yakushin, M., Pershin, I., and Vasil’evskii, S., “Heat Transfer Simulation and
Surface Catalycity Prediction at the Martian Atmosphere Entry Conditions,” AIAA Paper 99-4892,
Aug. 1999.
[21] Sepka, S., Chen, Y.-K., Marschall, J., and Copeland, R., “Experimental Investigation of Surface
Reactions in Carbon Monoxide and Oxygen Mixtures,” Journal of Thermophysics and Heat
Transfer, Vol. 14, No. 1, 2000, pp. 45-52.
[22] Afonina, N.E., Gromov, V.G., and Kovalev, V.L., “Catalysis Modeling for Thermal Protection
Systems of Vehicles Entering into Martian Atmosphere,” AIAA Paper No. 2001-2832, Jun. 2001.
[23] Rini, P., Garcia, A., Magin, T., and Degrez, G., “Numerical Simulation of CO2 Nonequilibrium
Flows with Catalyzed Surface Reactions,” AIAA Paper No. 2003-4038, Jun. 2003.
[24] Marschall, J., Copeland, R.A., Hwang, H.H., and Wright, M.J., “Surface Catalysis Experiments on
Metal Surfaces in Oxygen and Carbon Monoxide Mixtures,” AIAA Paper No. 2006-0181, Jan. 2006.
[25] Milos, F.S., Chen, Y.-K., Congdon, W.M., and Thornton, J.M., “Mars Pathfinder Entry Temperature
Data, Aerothermal Heating, and Heatshield Material Response,” Journal of Spacecraft and Rockets,
Vol. 36, No. 3, 1999, pp. 380-391.
[26] Bose, D. and Wright, M.J., “Uncertainty Analysis of Laminar Aeroheating Predictions for Mars
Entries,” AIAA Paper No. 2005-4682, Jun. 2005.
[27] Wright, M., Olejniczak, J., Walpot, L., Raynaud, E., Magin, T., Caillaut, L., and Hollis, B., “A Code
Calibration Study for Huygens Entry Aeroheating,” AIAA Paper No. 2006-0382, Jan. 2006.
RTO-EN-AVT-142 17 - 21
[28] Hollis, B., Horvath, T., Berry, S., Hamilton, H., and Alter, S., “X33 Computational Aeroheating
Predictions and Comparisons with Experimental Data,” AIAA Paper No. 99-3559, Jun. 1999.
[29] Gnoffo, P.A., “CFD Validation Studies for Hypersonic Flow Predictions,” AIAA Paper No. 2001-
1025, Jan. 2001.
[30] Wright, M.J., Loomis, M.A., and Papadopoulos, P.E., “Aerothermal Analysis of the Project Fire II
Afterbody Flow,” Journal of Thermophysics and Heat Transfer, Vol. 17, No. 2, 2003, pp. 240-249.
[31] Wright, M.J., Prabhu, D.P., and Martinez, E.R., “Analysis of Apollo Command Module Afterbody
Heating, Part 1: AS-202,” AIAA Paper No. 2004-2456, Jun. 2004.
[32] Baillion, M., Taquin, G., and Soler, J., “Huygens Radiative Probe Environment,” Proceedings of the
19th International Symposium on Shock Waves, 1993, pp. 339-346.
[33] Baillion, M. and Taquin, G., “Radiative Heat Flux: Theoretical and Experimental Predictions for
Titan Entry Probe,” Capsule Aerothermodynamics, AGARD Report No. 808, pp. 9/1-9/30, May
1997.
[34] Olejniczak, J., Wright, M., Prabhu, D., Takashima, N., Hollis, B., Zoby, E., and Sutton, K., “An
Analysis of the Radiative Heating Environment for Aerocapture at Titan,” AIAA Paper No. 2003-
4953, Jul. 2003.
[35] Wright, M.J., Bose, D., and Olejniczak, J., “The Impact of Flowfield-Radiation Coupling on
Aeroheating for Titan Aerocapture,” Journal of Thermophysics and Heat Transfer, Vol. 19, No. 1,
2005, pp. 17-27.
[36] Bose, D., Wright, M.J., Raiche, G., Bogdanoff, D., and Allen, G.A., “Modeling and Experimental
Validation of CN Radiation Behind a Strong Shock Wave,” AIAA Paper No. 2005-0768, Jan. 2005.
[37] Magin, T.E., Caillaut, L., Bourdon, A., and Laux, C.O., “Nonequilibrium Radiation Modeling for
Huygens Entry,” Proceedings of the 3rd International Planetary Probe Workshop, Jul. 2005.
[38] Raynaud, E., Tran, P., Soler, J., and Baillion, M., “Huygens Aerothermal Environment: Radiative
Heating,” Proceedings of the 3rd International Planetary Probe Workshop, Jul. 2005.
[39] Spencer, D. and Braun, R., “Mars Pathfinder Atmospheric Entry: Trajectory Design and Dispersion
Analysis,” Journal of Spacecraft and Rockets, Vol. 33, No. 5, 1996, pp. 670-676.
[40] Desai, P. and Cheatwood, F.M., “Entry Dispersion Analysis for the Genesis Sample Return Capsule,”
Journal of Spacecraft and Rockets, Vol. 38, No. 3, 2001, pp. 345-350.
[41] Dec, J.A. and Mitcheltree, R.A., “Probabilistic Design of Mars Sample Return Earth Entry Vehicle
Thermal Protection System,” AIAA Paper No. 2002-0910, Jan. 2002.
[42] Kachigan, S.K., Multivariate Statistical Analysis: A Conceptual Introduction, Radius Press, New
York, NY, 1982.
[43] Chen, Y.-K., Henline, W.D., and Tauber, M.E., “Mars Pathfinder Trajectory Based Heating and
Ablation Calculations,” Journal of Spacecraft and Rockets, Vol. 32, No. 2, 1995, pp. 225-230.
17 - 22 RTO-EN-AVT-142
[44] Roy, C., Oberkampf, W., and McWherter-Payne, M., “Verification and Validation for Laminar
Hypersonic Flowfields, Part 1: Verification,” AIAA Journal, Vol. 41, No. 10, 2003, pp. 1934-1943.
[45] Roy, C., Oberkampf, W., and McWherter-Payne, M., “Verification and Validation for Laminar
Hypersonic Flowfields, Part 2: Validation,” AIAA Journal, Vol. 41, No. 10, 2003, pp. 1944-1954.
[46] Inger, G.R., “Scaling Nonequilibrium-Reacting Flows: The Legacy of Gerhard Damköhler,” Journal
of Spacecraft and Rockets, Vol. 38, No. 2, 2001, pp. 185-190.
[47] Gökçen, T., “Effects of Flowfield Nonequilibrium on Convective Heat Transfer to a Blunt Body,”
Journal of Thermophysics and Heat Transfer, Vol. 10, No. 2, 1996, pp. 234-241.
[48] Smith, R.K., Wagner, D.A., and Cunningham, J. “A Survey of Current and Future Plasma Arc-
Heated Test Facilities for Aerospace and Commercial Applications,” AIAA Paper No. 98-0146, Jan.
1998.
[49] Loomis, M., Hui, F., Polsky, S., Venkatapathy, E., and Prabhu, D., “Arc-Jet Semi-Elliptic Nozzle
Simulations and Validation in Support of X-33 TPS Testing,” AIAA Paper No. 98-0864, Jan. 1998.
[50] Olejniczak, J. and Fletcher, D.G., “An Experimental and Computational Study of the Freestream
Conditions in an Arc-Jet Facility,” AIAA Paper No. 2000-2567, Jun. 2000.
[51] Fletcher, D.G., “Arcjet Flow Properties Determined From Laser-Induced Fluorescence of Atomic
Nitrogen,” Applied Optics, Vol. 38, No. 9, 1999, pp. 1850-1858.
[52] Grinstead, J., Driver, D., and Raiche, G., “Radial Profiles of Arcjet Flow Properties Measured with
Laser-Induced Fluorescence of Atomic Nitrogen,” AIAA Paper No. 2003-0400, Jan. 2003.
[53] Fletcher, D.G., Raiche, G.A., and Prabhu, D.K., “A Method for Determining Species Mass Fractions
From Spectrally Resolved Emission Measurements,” AIAA Paper No. 2000-2699, Jun. 2000.
[54] Raiche, G. and Driver, D., “Shock Layer Optical Attenuation and Emission Spectroscopy
Measurements During Arc Jet Testing with Ablating Models,” AIAA Paper No. 2004-0825, Jan.
2004.
[55] Szalai, C., Chen, Y.-K., Loomis, M., Thoma, B., and Buck, S., “Mars Exploration Rover Transverse
Impulse Rocket Cover Thermal Protection System Design Verification,” Journal of Spacecraft and
Rockets, Vol. 42, No. 6, 2005, pp. 990-998.
[56] Gökçen T. and Stewart, D., “Computational Analysis of Semi-elliptical Nozzle Arc-jet Experiments,”
AIAA Paper No. 2005-4887, Jun. 2005.
[57] Cauchon, D.L., “Radiative Heating Results from the Fire II Flight Experiment at Reentry Velocity of
11.4 km/s,” NASA TM X-1402, Jul. 1967.
[58] Anderson, J.D., “An Engineering Survey of Radiating Shock Layers,” AIAA Paper 68-1151, Dec.
1968.
[59] Woodbury, G., “Angle of Attack Analysis for Project Fire I Reentry Flight,” NASA TN D-3366,
1966.
RTO-EN-AVT-142 17 - 23
[60] Olynick, D., Henline, W., Hartung-Chambers, L., and Candler, G., “Comparison of Coupled
Radiative Flow Solutions with Project Fire II Flight Data,” Journal of Thermophysics and Heat
Transfer, Vol. 9, No. 4, 1995, pp. 586-594.
[61] Laub, B., “The Apollo Heatshield: Why Performance Exceeded Expectations,” Proceedings of the
International Symposium on Atmospheric Reentry Vehicles and Systems, Arcachon, France, Mar.
1999.
[62] Pavlosky, J.E. and St. Leger, L.G., “Apollo Experience Report: Thermal Protection Subsystem,”
NASA TN D-7564, Jan. 1974.
[63] Lee, D., Bertin, J., and Goodrich, W., “Heat Transfer Rate and Pressure Measurements During
Apollo Orbital Entries,” NASA TN D-6028, Oct. 1970.
[64] Lee, D. and Goodrich, W., “Aerothermodynamic Environment of the Apollo Command Module
During Superorbital Entry,” NASA TN D-6792, Apr. 1972.
[65] Bartlett, E.P., “Improved Heat Shield Design Procedures for Manned Entry Systems, Part II:
Application to Apollo,” NASA CR108689, Jun. 1970.
[66] Bartlett, E.P., “An Evaluation of Flight Data for the Apollo Thermal Protection System,” Space
Simulation, pp. 749-768, Jan. 1972.
[67] Thompson, R., “Comparison of Nonequilibrium Viscous-Shock-Layer Solutions With Shuttle

Heating Measurement,” Journal of Thermophysics and Heat Transfer, Vol. 4, No. 2, 1990, pp. 162-
169.
17 - 24 RTO-EN-AVT-142

RTO of NATO. Chazot O., Rini P. (Eds) - Experiment, Modeling and Simulation of Gas-Surface Interactions For Reactive Flows in Hypersonic Flights

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

RTO of NATO. Chazot O., Rini P. (Eds) - Experiment, Modeling and Simulation of Gas-Surface Interactions For Reactive Flows in Hypersonic Flights

Uploaded by

Copyright:

Available Formats

NORTH ATLANTIC TREATY RESEARCH AND TECHNOLOGY

RTO EDUCATIONAL NOTES EN-AVT-142

Experiment, Modeling and Simulation of

Papers presented during the AVT-142 RTO AVT/VKI Lecture Series

Published July 2007

The content of this publication has been reproduced

Published July 2007

Copyright © RTO/NATO 2007

Mr. O. Chazot and Mr. P. Rini

Jutta Rogal and Karsten Reuter

2.0 SURFACES IN REALISTIC ENVIRONMENTS

2.1 Surface free energy

(1) G = Gsolid + Ggas + ∆Gsurf .

∆G ad (T , p ) = γclean (T , p,0, N ' M ) − γ (T ,p, N O , N M ) =

2.2 Calculating Gibbs free energies

2.2.1 Gas phase chemical potential

(6) µO (T , p) = 1 2 µO2 (gas) (T , p) = 1

(7) QOtot2 (gas) =

Inserting eq. (7) into eq. (6) we then arrive at

Translational free energy

Rotational free energy

Vibrational free energy

Evaluation of the geometric series yields

where the first term arises from the zero-point vibrations.

Electronic and nuclear free energy

this ground state into account, we end up with

(16) µ electr ≈ EOtotal

Bringing it all together

2.2.2 Gibbs free energy of solid bulk and surface

(20) G = Etotal + Fvib + Fconf + pV ,

(21) F vib = dω F vib (T , ω) σ (ω)

≈ − ⎨ [∫ dω(σ ( N (N M − N 'M )σ M (ω) )F vib (T , ω)]− N O EOZPE ⎫⎬

2.3 Free energy plots and surface phase diagrams

Realizing that the definition for the average binding energy at T = 0 K is

1 ⎛ total total N total ⎞

with Eb > 0 for exothermicity, we arrive at the expression

[2] W. Kohn and L. Sham, Phys. Rev. A 140, 1133 (1965).

[9] M. Scheffler, “Thermodynamic Aspects of Bulk and Surface Defects – First-Principles

[13] K. Reuter and M. Scheffler, Phys. Rev. B 65, 035406 (2002).

[15] K. Reuter and M. Scheffler, Phys. Rev. B 68, 045407 (2003).

[17] K. Reuter and M. Scheffler, Appl. Phys. A 78, 793 (2004).

[24] D. Loffreda, Surf. Sci. (in press).

[31] C.T. Campbell, Phys. Rev. Lett. (in press).

[35] M. Todorova, E. Lundgren, V. Blum, A. Mikkelsen, S. Gray, J. Gustafson, M. Borg, J. Rogal, K.

[36] E. Lundgren, J. Gustafson, A. Mikkelsen, J.N. Andersen, A. Stierle, H. Dosch, M. Todorova, J.

[40] J. Rogal, K. Reuter, and M. Scheffler (in preparation).

Figure 1: Simulation of surface structures and processes at surfaces on the computer.

H = Tnucl + Tel + Vnucl−nucl + Vnucl−el + Vel−el , (1)

electrons. Explicitly, the single terms are

HΦ(R, r) = EΦ(R, r) , (4)

Hel ({R}) = Tel + Vnucl−nucl + Vnucl−el + Vel−el . (5)

ψkk (r) = eikk ·r ukk (r) , (9)

where ukk (r) has the two-dimensional periodicity of the surface.

3 Electronic structure calculations

where vext is the external potential. The Hartree potential vH is given by

It is numerically very efficient to use a plane-wave expansion of the Kohn-Sham single-particle

4 Electronic and geometric structure of surfaces

Electron density (1/n)

Figure 5: Layer-resolved, local d-band density of states of Pd(210) determined by GGA-DFT

Furthermore, at a surface the local density of states (LDOS)

[001] direction (Å)

Surface Method γ (J/m2 ) ∆d12 ∆d23 ∆d34 d0 (hkl) (Å)

Cu(111) Theorya 1.30 -0.9 -0.3 2.10