You are on page 1of 11

Chemical Engineering Science 169 (2017) 140–150

Contents lists available at ScienceDirect

Chemical Engineering Science


journal homepage: www.elsevier.com/locate/ces

Process intensification writ large with microchannel absorption in nitric


acid production
Jessy J.L. Lee, Brian S. Haynes ⇑
School of Chemical and Biomolecular Engineering, University of Sydney, NSW 2006, Australia

h i g h l i g h t s

 Process microfluidics coupled with process innovation.


 Steam ballast in nitric acid production from ammonia provides profound process intensification.
 Annular flow model shows condensation of steam in microchannels is extremely rapid.
 Residual gas-phase NO in O2 is rapidly oxidised and absorbed to form acid.
 Experiments show 1000-fold process intensification with 99% absorption.

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents experimental studies on the absorption of nitrous gas in microchannels. Liquid and
Received 18 June 2016 gas phase products have been analysed for a wide range of nominal residence times (0.03–1.4 s), gas
Received in revised form 25 October 2016 compositions (5–10% NO, 5–49% O2, 46–82% H2O, balance argon), system pressures (2.0–10.0 bar), mass
Accepted 9 January 2017
fluxes (1.5–30 kg m2 s1), and coolant temperatures (23–51 °C) in circular tubes with internal diameters
Available online 11 January 2017
of 1.4 and 3.9 mm.
Condensation of water vapour is very rapid in the microchannel, thereby enhancing the concentrations
Keywords:
of NO and O2 and leading to highly intensified production of nitric acid. The presence of inerts inhibits the
Nitric acid
Process intensification
degree of enhancement. A simple model of condensing two-phase annular flow, in which vapour-liquid
Narrow passage interface is assumed to be smooth, describes the main features of the observations. The interfacial area
Microreactor between the phases is the single most sensitive parameter in the model.
Microchannel Under the best conditions studied experimentally, an absorption efficiency of 99% was achieved at
2 bar, with an intensification factor relative to a conventional tower absorber of the order of 1000. The
NOx concentration in the tail gas is of the order of a few percent under these conditions but the tail
gas flow is very small. While further work remains to develop a complete process with acceptable emis-
sions, these results indicate that a microstructured system employing passages 1 m long could replace
the 70 m absorption tower in a conventional nitric acid plant.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction whereby the microfluidic benefits of miniaturised devices can be


brought to bear on large production capacities – ‘‘microreactors”
The development of microreactor technology for chemical pro- could in fact be very large. However, there are relatively few pub-
cessing has been an active topic of research for 20 years and there lished articles that deal with application of microstructured sys-
are many examples of fine chemical syntheses that achieve greatly tems to large-scale chemical production despite the fact that
improved conversion and selectivity as a result of the mixing and these would enable process intensification and process integration
heat and mass transfer performance achieved with this approach that would bring enormous benefits in energy efficiency (Johnston,
(Ehrfeld et al., 2000; Hessel et al., 2005; Dietrich, 2009; Boodhoo 1986) as well as opening the door for distributed chemical manu-
and Harvey, 2013). It has long been recognised that microreactor facturing (Benson and Ponton, 1993). Examples include the cat-
concepts can be implemented at scale through microstructuring, alytic production of propene oxide (Markowz et al., 2005),
Fischer-Tropsch liquids (Deshmukh et al., 2010), hydrogen
(Haynes and Johnston, 2011) and hydrogen peroxide
⇑ Corresponding author.
E-mail address: brian.haynes@sydney.edu.au (B.S. Haynes).

http://dx.doi.org/10.1016/j.ces.2017.01.015
0009-2509/Ó 2017 Elsevier Ltd. All rights reserved.
J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150 141

Nomenclature
3
~r rate of production [mol m s1 ]
1
List of symbols R ideal gas constant; tube radius [J mol K1 ; m]
a area density [m1 ] Re Reynolds number [–]
A area [m2 ] Sc Schmidt number [–]
B surface suction rate, GG!L =GG [–] St Stanton number [–]
3 T temperature [K]
c concentration [mol m ]
D tube diameter [m] t time [s]
Dh hydraulic mean diameter [m] u velocity [m s1 ]
D mass diffusivity [m2 s1 ] U overall heat transfer coefficient [W m2 K1 ]
f fanning friction factor [–] V volume [m3 ]
2 x distance [m]
G mass flux based on superficial area [kg m s1 ]
2 X yield; conversion [–]
G0 mass flux based on phase area [kg m s1 ]
a heat transfer coefficient [W m2 K1 ]
Gz Graetz number, ðL=Dh ÞReSc [–]
1 b mass transfer coefficient [m s1 ]
H enthalpy [J mol ] d film thickness [m]
1 1
H Henry law constant [mol kg bar ] h area multiplier [–]
2 1
j diffusive mole flux [mol m s ] l viscosity [Pa s]
k thermal conductivity [W m1 K1 ] q density [kg m ]
3
~
k forward rate constant for reaction j s shear stress [Pa]
j;f
Kj equilibrium constant for reaction j
L length [m] Location subscripts
1
m mass flow [kg s ] C cooling medium
1
n mole flow [mol s ] G gas
2
N mole flux [mol m s1 ] I interface
p pressure [bar] L liquid
r radius [m] W tube wall

(Pennemann et al., 2004), as well as the synthesis of pigments and the combustor where it serves as a ballast gas to moderate the
polymers (Wille et al., 2004; Iwasaki et al., 2006). temperature such that this lies in the optimal range for NO selec-
The benefits of microstructured systems in gas-liquid process- tivity (750–950 °C).
ing are especially attractive due to the high rates of interfacial heat Substitution of N2 with steam offers a resolution to the dilemma
and mass transfer that can be achieved (Vankayala et al., 2007; since steam may act as ballast in the combustor and be removed by
Lam et al., 2013; Ghiaasiaan, 2007) and it is perhaps surprising that condensation in the absorber. Indeed, this has been practised in the
this has not been exploited more in practice. One of the largest ton- past to produce NO for the production of high strength nitric acid
nage gas-liquid reaction processes in industry is that of the oxida- (typically >90 wt%, Caro et al., 1935; Manning, 1943) and for
tion and absorption of nitrous gases (principally NO from the hydroxylamine and caprolactam manufacture (Krauss and
combustion of ammonia in air) to make nitric acid. The overall Diekmann, 1963). In these processes, ammonia was combusted
reaction is in minimal excess of oxygen in steam – condensation of the steam
produced a weak acid and a highly concentrated stream of NO for
2NOðgÞ þ O2 ðgÞ ! 2NO2 ðgÞ
further processing. This provides the motivation for the present
work, to investigate the production of nitric acid through the com-
3HNO2 ðaqÞ ! HNO3 ðaqÞ þ 2NOðgÞ þ H2 OðlÞ plete absorption of the nitrous gas (as the mixture of nitrogen oxi-
The industrial operation (Thiemann et al., 2012; Andrew, 1985) des formed in the combustor is known) from the combustor into
has hardly changed in 100 years, using large stage-wise absorption the condensed steam ballast. Specifically, we investigate the pro-
towers that provide sufficient volume to enable the slow, recursive cess microfluidics of a microchannel condenser-absorber, on the
third-order gas-phase oxidation of NO to proceed sufficiently to basis that the 5-fold concentrating effect of condensing the steam
bring about, typically, about 99% absorption of the nitrous gas feed. ballast provides a factor of approximately 53 enhancement in the
The process is enhanced by operating at pressure (up to 13 bar) rate of the termolecular oxidation of NO.
and sometimes by injecting oxygen into the absorber to increase
its partial pressure (Kankani et al., 2015). Use of ozone injection
has been suggested to enhance oxidation of nitrous acid which is 2. Experimental method
carried out in the bleacher section of the absorption train
(Chacuk et al., 2007). Experiments were conducted to study the oxidation and
In general, it would not be appropriate to use a microstructured absorption behaviour of nitrogen oxides in microchannels over a
volume for a slow reaction because creating large volumes of this wide range of conditions. The supply of oxygen is always sufficient
kind is unnecessarily difficult and expensive. In the case of nitric for the complete conversion of NO to nitric acid, which has a
acid production, it could only be contemplated if the reactions stoichiometric requirement of 0.75 mol oxygen per mol NO.
could be accelerated far beyond what is currently achieved
3 1
through raising the pressure or adding some oxygen. The main NOðgÞ þ O2 ðgÞ þ H2 OðlÞ ! HNO3 ðaqÞ
impediment to raising the concentrations of NO and O2 is the pres- 4 2
ence of a large N2 fraction (typically  80%) that is carried over The total extent of absorption and the composition of the pro-
from the air feed to the ammonia combustor. However, although duct were characterised. The experimental apparatus is depicted
this N2 is a nuisance in the absorber, it has a useful function in in Fig. 1. Most experiments were performed with the absorber
142 J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150

tubes oriented horizontally but a downflow orientation was also lengths as the intensity of heat release (due to water condensation
examined. For the experiments with downflow, only the first and nitrous gas oxidation and absorption) after the first section
straight absorber section was employed. was low, as confirmed by type-K thermocouples (±1 K accuracy)
Reactant (99.9% NO, Linde and 99.999% O2, Coregas) and diluent mounted at various points downstream of the cooling section.
(99.999% Ar, Coregas) gas flows were controlled using calibrated The process gas typically enters the absorber at 180 °C and exits
mass flow controllers. A High Performance Liquid Chromatography at 25 °C, while the inlet and outlet temperatures of the cooling
(HPLC) pump (Waters 515) was used to deliver de-ionised and jacket were generally 20.5 °C and 21.5 °C respectively. Data from
degassed water to a mixing point where it was combined with the thermocouples and pressure transducers were retrieved by a
the O2 stream. The resulting two-phase stream was passed into a data acquisition unit (Agilent 34970A).
heated line (M&C TechGroup Type 5-M) where the water was Circular holes with a diameter of 1 mm were drilled at regular
evaporated completely. NO and Ar were preheated separately intervals along the length of the second and third straight sections
using 2.0 X/m resistance wires encased in PTFE tubes and coiled to enable liquid samples to be withdrawn for analysis. The holes
around the stainless steel gas line. The O2/steam and NO/Ar were drilled only through one side of the wall and placed along
streams were combined and entered the absorber as a single- the same axial line on each tube. Swagelok tee pieces allowed
phase, superheated gas at a temperature in the range 150–180 °C. the liquid passing through the holes to be directed into stainless
Pressure gauges (Swagelok, Model: PGI-63B-BG16-LAQX) and steel capillary tubing (ID = 140 lm, OD 1.6 mm). The lengths of
calibrated pressure transducers (Swagelok, Model: PTI-S-AA16- the capillary (2–12 m) were chosen to control the sampling flows
26AQ) were used to indicate pressures at the inlet and outlet of (typically  0.02 ml/min, or 5% of the condensed liquid flow in
the absorber. The absorber outlet was fed into a separator compris- the absorber tube). Since the flow pattern in the absorber under
ing a 1 L stainless steel sample cylinder (Swagelok, Model: 304L- the prevailing conditions is annular (Triplett et al., 1999), with
HDF4-1000) to allow the liquid phase to collect at the bottom of the liquid flowing as a film along the wall, it was possible to extract
the vessel and separate from the gaseous products. A backpressure liquid only, without allowing any significant amount of gas into the
regulator (Swagelok, Model: KBP1G0A4B5A20000) was used to sample. It was essential to avoid such gas flow in order to ensure
control the system pressure while the liquid level in the separator that the composition of the acid sample did not change in passing
was controlled manually by releasing liquid to maintain the liquid through the sampling capillary.
level in the separator at approximately half full. The measured Because the liquid product may contain significant quantities of
pressure drop across the length of the absorber was generally small nitrous acid HNO2, as well as the desired nitric acid HNO3, it is
(<7  10–3 bar). important also to eliminate composition changes due to decompo-
The absorber consisted of three straight lengths of stainless sition of HNO2 during sampling and prior to analysis. Nitrous acid
steel tubing with an internal diameter (ID) of either 1.4 mm undergoes disproportionation as
(3.2 mm OD) or 3.9 mm (6.4 mm OD) as shown in Fig. 2. Tubes of
equal diameter were connected in series via tube bends of the 3HNO2 ðaqÞ  HNO3 ðaqÞ þ 2 NOðgÞ þ H2 OðlÞ
same material and diameters, using Swagelok fittings to give a
total length of up to 1.47 m. The first straight section was located which can not only distort the relative nitrous/nitric acid yields, but
within a cooling jacket in the form of a coaxial stainless steel tube can also lead to significant loss of total acid product (Lee, 2012).
with ID = 10.2 mm. Cooling water from a temperature-controlled Although this reaction does equilibrate in the sampling capillary
bath (Thermoline BTC-9090, 21–50 °C) was pumped counter- prior to discharge, the loss to NO here (equivalent to the dissolved
currently into the annulus between the two tubes. The inlet and gas with an equilibrium partial pressure of 1 bar) is negligible (Lee,
outlet temperatures of both the absorber and the cooling jacket 2012). However, if the sample were to be exposed to atmosphere,
were measured using type-K thermocouples with an accuracy of the loss (as NO) would become significant. Therefore, the liquid
±1 K. Water jackets were not used on the subsequent downstream samples exit the capillary tubes into containers pre-filled with deio-
nised water, with the outlet of the sampling lines immersed below
the water surface. Since the rate of disproportionation is fourth-
order in the concentration of HNO2 (Abel and Schmid, 1928a,
Gas to Vent 1928b), the reaction can be arrested by dilution. A further benefit
of this dilution derives from the concomitant reduction in the pro-
ton concentration, which favours more extensive dissociation of
FTIR HNO2 to relatively stable nitrite ions:
Valve
Ar GMFC HNO2 ðaqÞ  Hþ ðaqÞ þ NO2 ðaqÞ

NO GMFC In order to obtain absolute concentrations, the incremental


PCV
mass due to the liquid products was measured. After further dilu-
Ar GMFC tion to <100 mg L1, the concentrations of nitrite and nitrate ions
P in each sample were measured by ion chromatography (Dionex
P

ICS-2000). The carrier for this determination was 4.5–15 mM


O2 GMFC PRV KOH (Thermo Scientific Dionex EluGenÒ EGC III KOH Cartridge,
T T
Filter Model: 074532) which ensured complete dissociation of the acids
H2O Absorber in the anion-exchange column (Dionex IonPac AS11-HC). A sup-
HPLC Heater pressor (Dionex 2 mm ASRS 300) was used to reduce background
Pump Separator conductance of the eluent and enhance detection of sample ions
in the conductivity cell detector.
The gaseous products leaving the separator were diluted with
Titration/I.C. argon before being passed through the gas cell of a calibrated Four-
Fig. 1. Schematic of the experimental setup. GMFC = Gas Mass Flow Controller,
ier Transform Infrared Spectrometer (Bio-Rad FTS6000) for NO,
PCV = Pressure Control Valve; PRV = Pressure Relief Valve; FTIR = Fourier Transform NO2, N2O and H2O analyses. Measurements were made using 64
Infrared Spectrometer; I.C. = Ion Chromatography. scans at a resolution of 0.25 cm1. Since the NO supply contained
J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150 143

Water Chiller

Water Bath
Cooling Water

Thermocouple
Cooling Jacket
Inflow
0.25 m
Absorber Tube
0.5 m

0.5 m
Outflow

Capillary Sampling Tubes


Fig. 2. Schematic of the absorber reactor.

N2O (2.11 vol%), which is neither formed nor destroyed in the sys- Table 1
tem, the N2O was used as a gas tracer to determine the total volu- The range of experimental parameters studied for nitrous gas absorption.

metric flow rate of the product gas. Parameter Range


Assuming that essentially all the water in the incoming mixture Tube internal diameter 1.4, 3.9 mm
is condensed very rapidly (as occurs in the BASF process (Krauss Absorber length 0.23–1.4 m
and Diekmann, 1963), and as found also in the modelling to be Inlet gas mole fraction 5–10% NO
described below), apparent yields of the acids (including both dis- 5–49% O2
46–82% H2O
sociated and undissociated forms) obtained at different locations
Bal. Ar
in the absorption tube are: Inlet system pressure, ðP T Þin 2–10 bar (abs)
  Inlet mass flux, Gin 1.5–30 kg m2 s1
cHNO3 L cHNO2 L 0.2–8.5 m s1
X HNO3 ¼  ; X HNO2 ¼  ð1Þ Inlet velocity, uG;in
cHNO  3 1
cHNO 
3 1
Nominal residence time, t res;nom 0.03–1.4 s
Coolant mass flux, Gc 66, 341 kg m2 s1

where cHNO3 1 is the molar concentration of nitric acid that would Coolant temperature, T c 23–51 °C

be achieved if all the NO were to be converted to nitric acid and dis-


solved in all the condensed water. The overall absorption efficiency
is simply the sum X HNO3 þ X HNO2 .
The nominal residence time tres;nom at a given distance x based 1968; Hoftyzer and Kwanten, 1972; Carta and Pigford, 1983;
on the inlet flow conditions is: Sherwood et al., 1975; Miller, 1987; Hüpen and Kenig, 2005;
q  Pradhan et al., 1997) – here, we adopt accepted chemistry and
x develop a model for microchannel condensation, absorption and
t res;nom ¼ ¼x G
ð2Þ
uG;in G in reaction which takes into account the specific features of the small
length scales encountered.
where qG is the gas density.
The nominal residence time carries no fundamental significance
and is used only as a comparative indicator of run conditions 3.1. Reaction chemistry and kinetics
because the fluid density changes as a result of physical and chem-
ical changes through the absorber. A more meaningful two-phase In the gas phase, the reactive species are NO, NO2, N2O3, N2O4,
residence time will be defined later in this paper. and O2. The gas-phase reaction mechanism is assumed to be
The ranges of experimental conditions used in this study are
2NO þ O2  2NO2 ðR1Þ
summarised in Table 1. The flow conditions in both phases are
laminar throughout.
2NO2  N2 O4 ðR2Þ

3. Modelling
NO þ NO2  N2 O3 ðR3Þ
The absorption of nitrogen oxides in water is a highly complex Although oxidation of NO (R1) proceeds essentially to comple-
process, with reactions taking place in both gas and liquid phases, tion at equilibrium, the reaction is slow, especially at low concen-
and simultaneous mass transfer interchange occurring between trations of NO and O2. On the other hand, equilibrium is rapidly
the phases (Thiemann et al., 2012; Andrew, 1985). Modelling of established in the forward and backward directions for reactions
nitrous gas absorption towers in conventional nitric acid produc- (R2) and (R3) (Pradhan et al., 1997), with the reaction rate con-
tion has been studied over many years (Miles, 1961; Chilton, stants being several orders of magnitude higher than that for the
144 J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150

oxidation of NO. In these circumstances, the governing equations The partial pressure of water at the interface is assumed to be
become stiff. Therefore, a partial-equilibrium approximation its saturation pressure at the interface conditions.
(PEA) was used, with (R1) being kinetically controlled while (R2) Table 2 reports the kinetic and equilibrium constants employed
and (R3) are assumed always to be in equilibrium. The concentra- in this work.
tions of N2O3 and N2O4 are therefore determined algebraically as
3.2. Condenser/absorber model
pN2 O4 ¼ K 2 p2NO2 ð3Þ
The structure of gas-liquid flows in microchannels has been
pN2 O3 ¼ K 3 pNO pNO2 ð4Þ studied in great depth in recent years (Ghiaasiaan, 2007; Triplett
where K 2 and K 3 are the equilibrium constants for reactions (R2) et al., 1999; Ide et al., 2007). Flow regime maps indicate that, for
and (R3) respectively. total mass fluxes G < 500 kg m2 s1, the annular regime prevails
The PEA has negligible effect on the accuracy of the calculated for almost the entire quality regime down to 10% by mass of
concentrations if the establishment of the reaction equilibria is fast vapour (Triplett et al., 1999), below which slug-annular and even-
compared to the time-step of the integration, as discussed by Rein tually slug flow arises. Baird et al. (2003) developed a model for
(1992) and Mott (1999) and confirmed specifically for the present shear-driven condensation in microchannels, based on a vapour
model (Lee, 2012). Following Ramshaw (1980), the chemical spe- core and an annular liquid film with a smooth interface. The model
cies involved in the equilibrium reactions are now assigned to was validated against an experimental data set of >2000 points
two pools based on their nitrogen oxidation states of either +2 or from 0 to >90% extent of condensation. The assumption of annular
+4. Taking note that N2O3 can be written as NO.NO2, and that flow driven by shear (and augmented by gravity in downflow
one mole of N2O4 contains two moles of N(+4) species, the concen- arrangements) was therefore adopted here also to model the con-
trations of the pools are defined as follows: densation and absorption behaviour of nitrous gases in
microchannels.
pNðþ2Þ ¼ pNO þ pN2 O3 ð5Þ Condensation was assumed to start immediately at the inlet of
the absorber (Baird et al., 2003), where the internal wall surface
pNðþ4Þ ¼ pNO2 þ pN2 O3 þ 2pN2 O4 ð6Þ has been pre-coated with a thin layer of liquid water correspond-
ing to 0.1% of the inlet mass flux. This assumption allows the
Eqs. (3)–(6) constitute a nonlinear system of four algebraic equa-
core-annular model description to be applied throughout the
tions that determines the four species concentrations from the pool
domain without having to model explicitly the first liquid forma-
concentrations. The gas phase mass balance is then written for the
tion – there is no loss of accuracy overall with this initialisation
two pools rather than for each individual species.
 (Baird et al., 2003). The liquid film grows in thickness as vapour
The chemical rates of production of the pools, ~r Nðþ2Þ G and
 condenses and gases are absorbed along the length of the absorber.
~rNðþ4Þ  are given by:
G The problem is solved at steady state and is assumed to be locally
! 1D, i.e. the rate of change of film thickness along the tube is slow.
  1 p2NO2
~ p2 p
~r Nðþ2Þ G ¼ ~r Nðþ4Þ G ¼ 2k1;f NO O2 1  ð7Þ At a distance x along the tube, the liquid phase occupies an
K 1 p2NO pO2
annular region of physical thickness dL. Chemical reactions occur
In the liquid phase, the relevant irreversible reactions are the in the gas and liquid phases, and there is transfer of heat and mass
hydration of N2O3 and N2O4 and the disproportionation of HNO2. between the phases. Heat is lost from the cooled section by con-
Nitric and nitrous acid dissociation is equilibrated but dissociation duction through the tube wall to the coolant stream which is
of HNO2 can in fact be neglected because the mixture is very acidic assumed to remain isothermal at a specified temperature. The
(except perhaps at the earliest stages of condensation). gas-liquid interface is assumed to be smooth.
The differential mole balances for the gas phase components
N2 O4 ðaqÞ þ H2 OðlÞ ! HNO2 ðaqÞ þ HNO3 ðaqÞ ðR4Þ ði ¼ Nð2þÞ; Nð4þÞ; HNO2 ; HNO3 ; H2 OÞ are:
 
N2 O3 ðaqÞ þ H2 OðlÞ ! 2HNO2 ðaqÞ ðR5Þ dni jG d½pðR  dL Þ2 Ni jG  dAI
¼ ¼ jj þ ~ri jG ð10Þ
dV G dV G dV G i G!L
3HNO2 ðaqÞ ! HNO3 ðaqÞ þ 2NOðaqÞ þ H2 OðlÞ ðR6Þ
Neglecting the derivative of the film thickness, therefore
HNO3 ðaqÞ þ H2 OðlÞ  NO3 ðaqÞ þ
þ H3 O ðaqÞ ðR7Þ dNi jG 2
¼ jj þ ~r i jG ð11Þ
dx ðR  dL Þ i G!L
Reactions (R4) and (R5) are very fast and occur within the mass
transfer boundary layer (Hoftyzer and Kwanten, 1972). The rate Similarly for the liquid phase ði ¼ NO;NO2 ; N2 O3 ; N2 O4 ;
constant for absorption from the gas phase is then given by the HNO2 ; HNO3 ; H2 OÞ:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
species specific value of H k ~
hydration DL (Hoftyzer and Kwanten, dNi jL 2ðR  dL Þ
¼ 2 ji jG!L þ ~r i jL ð12Þ
1972). The disproportionation of HNO2 represented by (R6) is not dx R  ðR  dL Þ2
an elementary reaction – its rate is given (Abel and Schmid,
1928b; Wendel and Pigford, 1958) by For laminar film flow, the film thickness is determined
4 (Henstock and Hanratty, 1976) as
cHNO2  sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
~
~r 6 ¼ k L
ð8Þ mtot jL lL
6;eff
p2NO dL ¼ ð13Þ
pRqL sc
Chemical equilibrium is assumed at the interface between the
gas and liquid phases, expressed in terms of Henry Law constants where sc is a characteristic shear stress that includes the effect of
for the nitrous gas species. vapour shear and gravity
   
ci jL ¼ Hi qL pi at equilibrium ð9Þ 1 dL 1 1 dp
sc ¼ sw 1   qL  þ g dL ð14Þ
3 R 3 qL dz
J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150 145

Table 2
Values of parameters employed in the model.

Parameter Value Units Ref


~ þ530 2
k 1;f
1:73105
exp bar s 1 Atkinson et al. (2004)
T2 T
ln K 1 4:27ð1  ln TÞ þ 4:49  103 T þ 2:27  107 T 2  1:04  109 T 3 þ 3:36  1013 T 4 þ 1:27  104 T 1 þ 4:35 bar
1 Burcat and Ruscic
(2005)
ln K 2 2 5 2 9 3 13 4 3 1 1 Burcat and Ruscic
4:80ð1  ln TÞ þ 1:62  10 T  1:04  10 T þ 4:54  10 T  8:94  10 T þ 6:39  10 T  0:81 bar
(2005)
ln K 3 2:32ð1  ln TÞ þ 1:01  102 T  7:78  106 T 2 þ 3:88  109 T 3  8:44  1013 T 4 þ 4:49  103 T 1  7:29 bar
1 Burcat and Ruscic
(2005)
pffiffiffiffiffiffiffi 1750 2 1
½H kDN2 O4 2:95  102 exp mol m s1 bar Hoftyzer and Kwanten
T
(1972)
pffiffiffiffiffiffiffi 2 1
½H kDN2 O3 1:57 mol m s1 bar Corriveau (1971)
 
~
k 1:62  10 11
exp 14;280 2 9
bar m mol
3
s 1 Wendel and Pigford
6;eff T
(1958)


HNO 1:90  103 exp 1500 T1  298:15
1
mol kg
1
bar
1 Schwartz and White
(1981)
HNO2 0.012 mol kg
1
bar
1 Schwartz and White
(1981)
HN2 O4 1.4 mol kg
1
bar
1 Schwartz and White
(1981)
HN2 O3 0.6 mol kg
1
bar
1 Schwartz and White
(1981)


HHNO2 49 exp 4800 1T  298:15
1
mol kg
1
bar
1 Schwartz and White
(1981)
HHNO3 2:1  105 mol kg
1
bar
1 Schwartz and White
(1981)

Here we evaluate the pressure gradient from the interfacial Viewed from the liquid side, the interphase transport of the
shear stress as absorbing species is
dp 2si 
 ¼  qG g ð15Þ ji jG!L ¼ beff ;i L ðci jLI  ci jL Þ ð20Þ
dz Ri
  
where the interfacial shear stress is determined as that arising only We calculate beff ;i L ¼ Di jL =deff ;i L , where deff ;i L is the effective
from the gas side motion mass transfer thickness for species i. In confined laminar flow,
the flow is fully developed when the Graetz number satisfies
 G02 Gz K 50, under which circumstances the diffusion length scale is
sI ¼ f eff GI G
ð16Þ
2qG of the same order as the length scale of the confinement, i.e.

Baird et al. (2003) considered the effect of interphase mass deff ;i L  dL . With the hydraulic mean diameter Dh ¼ 4dL for trans-
transfer (condensation of H2O) in determining the effective friction port to one face of a film, we use the model to evaluate

factor f eff GI and showed that the effect of condensation mass trans-
16mtot jL dL
fer is correctly accounted for via (Hewitt et al., 1994): Gz ¼  ð21Þ
pDqL D L
 2BG
f eff GI ¼ ð17Þ
1  exp ð2BG =f 0 Þ and find that the fully-developed flow condition is always satisfied
where BG ¼ GG!L =G0G
is calculated from the interphase mass flux at within the first, cooled absorber stage. We therefore take the diffu-
sion thickness to be dL , unless the diffusion is enhanced by reaction
the gas-liquid interface. In the present work, the interphase mass
within the liquid film.
flux is taken to include both condensation (of H2O) and absorp-
The absorption of N2O4 and N2O3 is enhanced by their rapid
tion/desorption of the nitrous gas components, and the reference
hydration, with the result that the effective mass transfer coeffi-
(without mass transfer) friction factor f 0 is calculated for the gas
cient becomes (Wendel and Pigford, 1958):
flow in a conduit with the same diameter as the core. Following
Baird et al. (2003), a smooth conduit is assumed here. rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi


The interphase fluxes due to absorption/desorption of the beff ;i L ¼ ~ 
k Di jL ði ¼ N2 O3 ; N2 O4 Þ ð22Þ
i
hydration
nitrous gases are calculated as
  
beff G ci j In general, the effective film thickness corresponding to absorp-
 ffi
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ji jG!L ¼ pi  LI ð18Þ 
~ 
RT I Hi qL tion with fast reaction drxn;i L ¼ Di jL =k i is much less
hydration
where equilibrium between the gas and liquid phases is assumed to than the physical film thickness. However, at the earliest stages

occur at the interface. The effective mass transfer coefficient beff G in the condensation process, when the liquid film is extremely thin,
includes the effect of interphase mass transfer (Hewitt et al., 1994): the calculated film thickness for fast reaction is found to exceed the

 physical thickness ðdL < drxn;i L Þ. In order to ensure that the mass
BG
St eff G ¼ ð19Þ transfer coefficient is calculated approximately correctly through-
1  expðBG =St 0 Þ
out the range, we simply add mass-transfer coefficients calculated
The reference (without mass transfer) Stanton number St 0 is for diffusion only and diffusion with reaction. While this is not
that calculated for the gas flow in a conduit with the same diame- strictly accurate, it provides a smooth interpolation from the very
ter as the core. For laminar flow, the augmentation of the Stanton early stages of the condensation through to the region where
number is the same for heat and mass transfer. reaction enhancement determines the rate of absorption of these
146 J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150


species. Because the region where dL K drxn;i L is very limited, the Absorption Efficiency). Note that acid formation consumes water
accuracy of the final solution is not significantly affected by the substance in the liquid phase, which is responsible for the apparent
details of this smoothing. reduction in the degree of condensation after 0.007 s.
In order to determine the interphase mass flux due to conden- It is immediately apparent that the time-scale of the absorption
sation of water, the energy balance equations for the gas and liquid in Fig. 3 is vastly less than the 100 s residence time in a conven-
phases must be solved. For the gas phase tional absorption tower (Gutierrez-Cañas et al., 1989). This intensi-
n  fication comes both from the scale of the microchannel and from
d X  X  o
 i  ¼ 2
N i jG H aeff G ðT G  T I Þ þ  i
ji jG!L H ð23Þ the process improvement whereby condensation of the ballast
dx G ðR  dL Þ G
gas enhances the chemical reaction rates. The condensation of
For the liquid phase, the balance includes heat flow to the cool- steam and the absorption of nitrous gases also lead to significant
ing medium reductions in the volumetric gas flow rate along the absorber tube:
the actual gas-phase residence time relevant to the chemical kinet-
d X  n X  o
 i  ¼ 2ðR  dL Þ
N i jL H aL ðT I  T L Þ þ  i
ji jG!L H ð24Þ ics can be calculated as
dx L 2 2 G
R  ðR  dL Þ
Z x Z x
dx qG
2R tres jG ¼ ¼ dx ð25Þ
 U W ðT L  T C Þ 0 uG 0 G0G
R2  ðR  dL Þ2
For the case in Fig. 3, the gas velocity decreases fourfold as
The thermodynamic properties of gaseous species are calcu- steam condenses and by a further factor 2 as the nitrous gas is
lated using the NASA polynomials (Burcat and Ruscic, 2005). oxidised and absorbed – the extended gas residence time leads
Liquid-phase data are determined using OLI Analyzer Studio 3.1. to further enhancement of the extent of NO oxidation and its sub-
Once the reaction rate terms for the gas and liquid flows are sequent absorption.

evaluated (as described above), the rate of condensation jH2 O G!L In the following paragraphs and associated figures, results are
can be calculated and the evolution of the phase flow rates, the film presented for a number of process variables and compared with
thickness and the species concentration profiles are all determined the predictions of the model in order to understand the most
by numerical integration. The differential equations are solved by a important features of the microchannel condenser/absorber. The
forward Euler technique in Microsoft Office Excel 2007. The model basic model generally gives a qualitatively correct description of
contains many implicit dependencies and these are solved itera- the results but, in order to improve quantitative comparisons, the
tively at each step. A step size of 1 mm was shown to produce sensitivity of the model to its various parameters has been consid-
grid-independent solutions and is used for all the model results ered. This analysis shows (Lee, 2012) that the vapour-liquid inter-
presented here. facial area density aI is the most important single parameter in
controlling the kinetics of the system, as expressed in the concen-
4. Results and discussion trations of HNO2 and HNO3 produced. In the model, this parameter
is taken as that for a smooth interface between the gas core and the
Fig. 3 shows results from the model for typical conditions, annular liquid – this is clearly a conservative simplification as
focussing on the early stages of the absorption process in order interfacial waves and intermittent flow patterns often arise in
to reveal the rapidity of the processes of water condensation, NO gas-liquid flows (Triplett et al., 1999). Sensitivity to the interfacial
oxidation and N2O3 and N2O4 absorption that occur in the system. area density may also reflect sensitivity to heat and mass transfer
Condensation of water vapour (75% mol fraction in the inlet) rates, frictional drag and film thickness. We have therefore chosen
begins immediately and proceeds rapidly to completion, this parameter as the single adjustable quantity when comparing
tres;nom  0:007 s. As a result, the concentrations of the residual the model predictions with the experimental results, through the
gases (nitrous gases and oxygen) rise by a factor of 4, under introduction of an area multiplier hI on the geometrical value.
which conditions the homogeneous oxidation of NO is enhanced While values of hI determined in this way should therefore not
significantly (by 43); at the same time, the rates of formation be equated too precisely with a physical area adjustment, it is
and absorption of N2O3 and N2O4 are promoted, resulting in an noteworthy that the magnitude of hI for the different conditions
increasing rate of acid formation (as indicated by the nitrous gas studied here are generally in a narrow range, 1 < hI < 3.
Fig. 4 shows results for acid formation as a function of the nom-
inal residence time in tubes at 8 bar, with 1.4 mm (filled symbols)
and 3.9 mm internal diameter (open symbols). The data for each
tube were obtained from runs over a range of inlet mass fluxes.
For the 1.4 mm tube, the grey filled symbols show results obtained
with horizontal flow using the extended absorber as shown in
Fig. 2 and with samples taken at the exit – total absorber
lengths of 0.46 m and 1.47 m were used with mass fluxes
11–23 kg m2 s1; dashed lines in the Fig. show results of the zero
gravity model for these conditions. An additional series of runs
with identical mass fluxes, but in downflow (black filled symbols),
was conducted using only the primary 0.25 m absorber with sam-
ples collected at the exit – solid lines in the figure show results of
the model with gravitational forces. The model shows negligible
effect of mass flux at a given nominal residence time (because film
thickness in these partially condensed flows depends only weakly
on the inlet mass flux), so the predictions for a given flow orienta-
Fig. 3. Model predictions showing course of condensation and nitrous gas reaction
tion can be represented by a single line, with the differences
and absorption in a 1.4 mm ID tube at 2.0 bar with an inlet gas composition of 10% between downflow predictions (solid lines) and those for horizon-
NO, 15% O2 and 75% H2O, and an inlet mass flux of 9.6 kg m2 s1. tal flow (dashed) also being slight. Clearly, the model accurately
J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150 147

GL;min = 70 kg m2 s1, approximately 100 times greater than could


have arisen in the experiment. This comparison is sensitive to
properties (notably the contact angle which, if taken as 40°, gives
a reduction in minimum stable flow by a factor of 10) but it is
apparent that film flow in this larger tube is unlikely to be the con-
figuration with minimum energy. However, while water is still
condensing from the gas phase, it can be expected that the film
flow conditions will prevail and that breakdown of the film can
only occur further downstream – formation of a single rivulet at
this point would reduce the interfacial area for mass transfer by
a factor 50, after which the rate of absorption of nitrous gases
would effectively cease.
A similar comparison for the 1.4 mm tube is less conclusive,
with the predicted minimum stable film flow for an assumed
contact angle 60° now about 30 kg m2 s1 (compared with
7 kg m2 s1 for 40° contact angle and an actual condensed water
Fig. 4. Acid yield profiles for absorption at 8.0 bar with an inlet gas composition of
5% NO, 49% O2 and 46% H2O. Filled points show experimental results for 1.4 mm
flow of 5 kg m2 s1). However, the observed gas absorption goes
tube diameter with mass fluxes of 11–23 kg m2 s1; black symbols are for smoothly to completion in this and all other experiments reported
downflow, grey are for horizontal flow. Open points are for 3.9 mm tube with here for the 1.4 mm tube and there is no evidence of a sudden
downflow mass flux 1.5–3.0 kg m2 s1. Solid and dashed lines show predictions of change in absorption rate. Two-phase horizontal flow regime maps
the model (hI ¼ 1) with and without gravitational effects respectively.
for narrow passages (Shao et al., 2009) barely extend to the low
values of superficial liquid velocity encountered here
(0.003 m s1) but, based either on velocity or Weber number plots,
captures the details across the whole length of the absorber with- the prevailing flow condition is expected to be in the Taylor-
out requiring any adjustment (hI ¼ 1Þ. Early rapid formation of annular to annular flow regime, consistent with the application
HNO2 and HNO3 sees the overall conversion rate exceed 90% by of the annular flow model.
t res;nom  0.2 s. The yield of HNO2 peaks at 7%, after which the In the following discussion, emphasis is placed on the perfor-
main apparent change is the gradual conversion of HNO2 to mance of the 1.4 mm microchannel absorber in the earlier phases
HNO3, controlled by the diminishing rate of the disproportionation of the absorption where mass transfer effects and process intensi-
reaction (R6) following fourth-order kinetics. The final absorption fication are greatest. Fig. 5 presents results for the yields of HNO3
efficiency is in excess of 99%, comparable with requirements for and HNO2 as a function of pressure for fixed nominal residence
a conventional tower absorber. time (t res;nom = 0.08 s) – the inlet mass flux in these runs
For the 3.9 mm tube in Fig. 4, the results, which were obtained therefore increases with pressure, from 4.3 kg m2 s1 at 2 bar to
in downflow using inlet mass fluxes in the range of 1.5– 22 kg m2 s–1 at 10.0 bar. The experimental results show a strong
3.0 kg m2 s1, are very different. Absorption apparently ceases enhancing effect of pressure up to 6.0 bar, above which the total
after 68% completion with disproportionation of the nitrous acid acid yield barely changes. The results of the model are qualitatively
that is formed apparently suppressed, perhaps corresponding to similar, but now with the maximum acid yield at this residence
the inhibiting effect of residual NO on the kinetics of dispropor- time being achieved at 5.0 bar, and with the yield passing through
tionation. The mass flows in the two series of experiments are the maximum and falling significantly at higher pressures. Interro-
the same and the velocities in the larger tube are therefore consid- gation of the model results (Lee, 2012) reveals that the increased
erably less than in the smaller tube. With the lower velocities com- thermal load (greater mass fluxes) at the higher pressures means
bined with the usual effects of duct size in process intensification, that the water condensation process is increasingly delayed. For
the model predicts lower heat and mass transfer coefficients and
therefore a lower extent of absorption in a given nominal residence
time. However, the model does show that the absorption should go
near to completion and that HNO2 should disproportionate as
much as in the smaller tube.
We believe that the apparent cessation of absorption in the
3.9 mm tube may be due to a breakdown of the annular film flow
at the low velocity ðuL K 0:001 m s1 Þ arising in this case. Down-
ward flowing film stability has been studied in some depth because
of its relevance to falling-film contact devices and it is known that,
at sufficiently low flow rates (i.e. for sufficiently thin films), the
flow is more stable as one or more rivulets than as a thin film
(Mikielewicz and Moszynskl, 1976; El-Genk and Saber, 2001).
The minimum stable film flow depends on the fluid properties
and on the contact angle of the fluid with the wall (Mikielewicz
and Moszynskl, 1976; El-Genk and Saber, 2001). Hughes and Bott
(1991) considered the effect in tube flow where the curvature
extends the range of stable film flow; co-flowing vapour shear also
promotes film stability (Saber and El-Genk, 2004; Wilson et al.,
2011). Applying the method of Hughes and Bott (1991) to account
for curvature effects for the conditions in Fig. 4, we calculate Fig. 5. Effects of system pressure on product yields for downflow absorption in a
1.4 mm ID tube with an inlet gas composition of 5% NO, 24% O2, 25% Ar and 46%
(assuming the properties of water at 25 °C and a contact angle of H2O, inlet mass fluxes of 4.3–22 kg m2 s1 (t res;nom = 0.08 s). Points show experi-
60° (Li et al., 2012)) that the minimum stable film flow in the mental results while lines show predictions of the model with hI ¼ 1; -and-line
3.9 mm tube corresponds to a condensed liquid mass flux symbols are for different values of area multiplier (hI ¼ 2 and hI ¼ 3).
148 J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150

the same reason, the gas core temperatures remain >100 °C (high
enough to suppress the formation of the acid precursors N2O3
and N2O4) until further into the tube (<0.005 s at 2.0 bar; 0.05 s
at 10.0 bar). The reason for the failure of the model to capture
the turning point in the yield curves is not known – however,
the results are very sensitive to the balance between the cooling
rate and the rate of N2O4 absorption. In terms of the adjustment
parameter, changing from hI ¼ 1 as in Fig. 5 to hI ¼ 2 brings the
prediction of HNO3 yield at 5.0 bar from 29% to 48%, thus bracket-
ing the experimentally observed value.
Direct evidence for the importance of O2 concentration is shown
in Fig. 6 which compares the course of the absorption processes for
10% NO with either 10% or 24% O2. In these experiments, at 2.0 bar
in a 1.4 mm ID tube, the water vapour mole fraction is maintained
constant at 66%, with argon making up the difference (14% or 0%, Fig. 7. Effects of NO partial pressure on product yields for downflow absorption in a
respectively) – the overall mass flux is the same in both cases, 1.4 mm ID tube with an inlet gas composition of 5–10% NO, 24% O2, 20–25% Ar and
G = 10 kg m2 s1. The concentrating effect of steam condensation 46% H2O, inlet mass fluxes of 4.3–22 kg m2 s1 (tres;nom = 0.08 s) and hI ¼ 1. Filled
is the same in the two experiments, as is the flow of noncondensi- points show experimental results for 5% NO; open points are for 10% NO. Point-and-
line symbols show predictions of the model described in the text.
ble gas (O2 + Ar). The formation of acid is clearly accelerated at the
higher O2 concentration (O2:NO = 2.4 versus 1.0) – the model
reproduces these effects with hI ¼ 3 and confirms that the effect
of oxygen is simply to accelerate the oxidation of NO in the gas
phase.
The oxidation of NO is second order in the NO concentration
and is expected to proceed faster at higher NO concentrations.
Fig. 7 confirms this effect by comparing results for acid yield from
10% NO with that from 5% NO, under conditions close to those in
Fig. 6. For the run with 10% NO, the argon fraction was reduced
to 20% in order to maintain comparable inlet velocities. The results
show that the yields of HNO3 are significantly higher with 10% NO
than with 5% NO, as expected from the kinetics of reaction (R1). At
the same time the yield of HNO2 is slightly reduced, this effect
being more pronounced when the ratio HNO2:HNO3 is considered
– this is also a kinetic effect, reflecting now the greater rate of dis-
proportionation of HNO2 at the higher concentration in the solu-
tion formed from 10% NO. These trends are reproduced by the
model with hI ¼ 1 but, as discussed in relation to Fig. 5, the model Fig. 8. Product yields (left axis) and calculated true residence times (dotted line,
right axis) for horizontal absorption in a 1.4 mm ID tube at 2.0 bar with an inlet gas
predictions appear not to capture temperature effects at these composition of 10% NO, 10% O2, 66–80% H2O, and balance Ar (14–0%). The inlet
higher pressures. mass flux is maintained at 10 kg m2 s1 (t res;nom = 0.16 s). Points show experimental
Given the importance of NO oxidation, especially at later stages results while solid lines show predictions of the model with hI ¼ 3.
of the absorption process, it is clear that the actual residence time
of the gas in the absorber impacts the extent of oxidation and
absorption. Fig. 8 shows results for absorption of 10% NO at to 0% in order to maintain the overall mass flux constant at
2.0 bar in a 1.4 mm ID tube with 10% O2 and varying amounts of 10 kg m2 s1. The extent of absorption increases as the argon is
steam (66–80%). Argon was added in amounts ranging from 14% progressively replaced by steam. Two main factors contribute to
this effect of reducing argon content – the concentrating effect of
steam condensation is enhanced; and the actual residence time
in the reactor increases, by approximately the same factor. Both
effects favour oxidation of NO and the production of acid. In addi-
tion, the longer residence time of the liquid phase allows a greater
extent of disproportionation of HNO2. The model captures these
trends accurately with an area enhancement factor hI ¼ 3.
The stoichiometric oxygen requirement for conversion of NO to
HNO3 is 0.75 mol mol1, corresponding to H2O:NO = 8.25 in Fig. 8.
Any oxygen in excess of stoichiometric is itself a noncondensible
diluent and it might be expected that further enhancement of
the process could be achieved by operating closer to the stoichio-
metric requirement. However, reducing the oxygen concentration
also reduces the rate of NO oxidation, as shown above. Therefore,
there is likely to be an optimal O2 excess that takes advantage of
the increased residence time while also providing adequate oxygen
for rapid oxidation of NO. This remains a topic for further
Fig. 6. Effects of oxygen partial pressure on product yields for horizontal absorption investigation.
in a 1.4 mm ID tube at 2.0 bar with an inlet gas composition of 10% NO + 66% H2O
+ 24% (Ar + O2) and a total inlet mass flux of 10 kg m2 s1. The filled symbols are
Fig. 9 shows the effects of cooling water temperature on the
for experimental data with Ar = 0% (O2 = 24%); open symbols for Ar = 14% performance of the absorber at short residence times
(O2 = 10%). Lines show predictions of the model with hI ¼ 3. (tres;nom = 0.03 s) when operating at 2.0 bar with a small excess of
J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150 149

example, the intensification achieved at relatively low pressures


(2 bar) is such that high-pressure absorption might be
unnecessary.

5. Conclusions

Absorption of nitrous gases in a 1.4 mm tube allows profound


process intensification relative to typical absorption towers
employed in conventional nitric acid plants. Intensification factors
of the order of 1000 are achievable when the nitrous gas is pro-
duced in a steam ballast – condensation of the steam is extremely
rapid in the microchannel, thereby leading to significant increases
in the concentration and rates of reaction and absorption of the
nitrous gases and oxygen. Overall, the intensification demon-
strated in this work derives from a combination of process modifi-
Fig. 9. Effects of cooling water temperature on product yields for horizontal cation and the inherent intensification of heat and mass transfer
absorption in a 1.4 mm ID tube at 2.0 bar with an inlet gas composition of 10% NO, rates in microchannel systems.
8% O2 and 82% H2O, and inlet mass flux of 9.2 kg m2 s1 (tres;nom = 0.03 s). Points
While the interactions between reaction kinetics and heat and
show experimental results while lines show predictions of the model with hI ¼ 3.
mass transfer effects are complex, a simple model of microchannel
condensation in annular flow, augmented to account for chemical
oxygen (O2:NO = 0.80). The yields of both acids are clearly reduced reactions in the gas and liquid phases, describes the qualitative fea-
when the cooling temperature is raised, with the trends being cap- tures of the system over a wide range of conditions. Quantitative
tured by the model. Overall, there is a reduction in the condensa- comparisons can generally be made upon adjustment of a single
tion rate and the persistence of higher gas temperatures – as a parameter, the interfacial area between the phases, by a factor
result, the rate of oxidation of NO and the concentrations of the 1 K hI 6 3.
acid-forming intermediates N2O3 and N2O4 are suppressed. The
yield of HNO2 is additionally reduced because its disproportiona- Acknowledgements
tion is enhanced when the liquid film temperature is increased.
In summary, this work has demonstrated that very high levels The authors acknowledge the financial support and permission
of process intensification in the production of nitric acid can be to publish provided by Orica Limited. Jessy Lee thanks the Aus-
achieved through a combination of process modification and the tralian Government for an Australian Postgraduate Award.
use of a microchannel condenser/absorber. For the conditions of
Fig. 8 with molar feed rate ratios H2O:NO = 8 and O2:NO = 1, 99%
acid yield is achieved within a tube length of 0.85 m, correspond- References
ing to a volumetric intensity of 13 tonne HNO3 h1 m3. This is Abel, E., Schmid, H., 1928a. The kinetics of nitrous acids. I. Introduction and
approximately three orders of magnitude greater than for conven- overview. II. Orientational experiments. Zeitschrift für Physikalische Chemie-
tional plate columns achieving 99% absorption at 2 bar, for which Stochiometrie Und Verwandtschaftslehre 132, 55–63.
Abel, E., Schmid, H., 1928b. The kinetics of nitrous acids. III. Kinetics of nitrous acid
values of 0.005 tonne HNO3 h1 m3 (Andrew, 1985) and
decomposition. Zeitschrift für Physikalische Chemie-Stochiometrie Und
0.014 tonne HNO3 h1 m3 (Hoftyzer and Kwanten, 1972) are esti- Verwandtschaftslehre 134, 279–300.
mated. While operation of a plate column at higher pressure Andrew, S.P.S., 1985. The absorption and oxidation system. In: Keleti, C. (Ed.), Nitric
Acid and Fertilizer Nitrates. Marcel Dekker, New York (Chapter 4).
increases its productivity / p2:5 , the intensity achievable at Atkinson, R., Baulch, D.L., Cox, R.A., Crowley, J.N., Hampson, R.F., Hynes, R.G., Jenkin,
10 bar is still vastly less than for low-pressure absorption in the M.E., Rossi, M.J., Troe, J., 2004. Evaluated kinetic and photochemical data for
steam-ballasted microchannel process. However, there are some atmospheric chemistry: Volume I - gas phase reactions of Ox, HOx, NOx and
SOx species. Atmos. Chem. Phys. 4, 1461–1738.
significant differences between the products obtained from the
Baird, J.R., Fletcher, D.F., Haynes, B.S., 2003. Local condensation heat transfer rates in
conditions in the microchannel example and those from the con- fine passages. Int. J. Heat Mass Transf. 46 (23), 4453–4466.
ventional operations. Benson, R.S., Ponton, J.W., 1993. Process miniaturization - a route to total
Firstly, the NOx concentration remaining after 99% absorption in environmental acceptability? Chem. Eng. Res. Des. 71 (2), 160–168.
Boodhoo, K., Harvey, A., 2013. Process Intensification Technologies for Green
the microchannel is 37,000 ppm whereas the conventional plant Chemistry: Engineering Solutions for Sustainable Chemical Processing. John
has 1000 ppm, even though the actual flow of NOx is the same Wiley & Sons, Chichester.
in the two cases. To bring the microchannel absorber tail gas down Burcat, A., Ruscic, B., 2005. Third Millennium Ideal Gas and Condensed Phase
Thermochemical Database for Combustion with Updates From Active
to 1000 ppm by absorption would require a tower not much smal- Thermochemical Tables. Argonne National Laboratory, Argonne, Illinois.
ler than a conventional tower because the volumetric requirement Caro, N., Frank, A.R. Wendlandt, R., Fischer, T., 1935. Production of Nitric Acid and
at the lower concentrations is determined entirely by the kinetics Liquid Nitrogen Tetroxide. US Patent 1,989,267.
Carta, G., Pigford, R.L., 1983. Absorption of Nitric Oxide in Nitric Acid and Water. Ind.
of NO oxidation. Therefore, other tail gas treatment options would Eng. Chem. Fundam. 22 (3), 329–335.
need to be considered, for which the small gas volume and high Chacuk, A., Miller, J.S., Wilk, M., Ledakowicz, S., 2007. Intensification of nitrous acid
concentrations would offer many advantages. oxidation. Chem. Eng. Sci. 62 (24), 7446–7453.
Chilton, T.H., 1968. Strong Water: Nitric Acid: Sources, Method of Manufacture, and
The second important difference between the microchannel and Uses. The M.I.T Press.
conventional products is that the acid concentration in the former Corriveau, C.E., 1971. The Absorption of N2O3 into Water. (Master’s thesis,
is 32 wt% whereas it is standard industry practice to produce 55– University of California, Berkeley).
Deshmukh, S.R., Tonkovich, A.L.Y., Jarosch, K.T., Schrader, L., Fitzgerald, S.P.,
65 wt%. To the extent that the concentrated form is actually
Kilanowski, D.R., Lerou, J.J., Mazanec, T.J., 2010. Scale-up of microchannel
required, the tower, typically 70 m tall, may still be needed. In reactors for FischerTropsch synthesis. Ind. Eng. Chem. Res. 49 (21), 10883–
other respects, however, the <1 m ‘‘height” of the microchannel 10888.
system is a very attractive alternative when coupled with ammo- Dietrich, T.R. (Ed.), 2009. Microchemical Engineering in Practice. John Wiley & Sons
Inc., Hoboken.
nia combustion in a steam ballast. The visual aspects of eliminating Ehrfeld, W., Hessel, V., Löwe, H., 2000. Microreactors – New Technology for Modern
the tower are obvious but there are process benefits also – for Chemistry. Wiley-VCH, Weinheim.
150 J.J.L. Lee, B.S. Haynes / Chemical Engineering Science 169 (2017) 140–150

El-Genk, M.S., Saber, H.H., 2001. Minimum thickness of a flowing down liquid film Mikielewicz, J., Moszynskl, J.R., 1976. Minimum thickness of a liquid film flowing
on a vertical surface. Int. J. Heat Mass Transf. 44 (15), 2809–2825. vertically down a solid surface. Int. J. Heat Mass Transf. 19 (7), 771–776.
Ghiaasiaan, S.M., 2007. Two-Phase Flow, Boiling, and Condensation: In Miles, F.D., 1961. Nitric Acid: Manufacture and Uses. Oxford University Press,
Conventional and Miniature Systems. Cambridge. Cambridge University Press. London.
Gutierrez-Cañas, C., Arias, P.L., Legarreta, J.A., 1989. Industrial nitrogen oxides Miller, D.N., 1987. Mass transfer in nitric acid absorption. AIChE J. 33 (8), 1351–
absorption simulation. Comput. Chem. Eng. 13 (9), 985–1002. 1358.
Haynes, B.S., Johnston, A.M., 2011. Process design and performance of a Mott, D.R., 1999. New Quasi-Steady-State and Partial-Equilibrium Methods for
microstructured convective steam-methane reformer. Catal. Today 178 (1), Integrating Chemically Reacting Systems (Doctoral dissertation, The University
34–41. of Michigan).
Henstock, W.H., Hanratty, T.J., 1976. The interfacial drag and height of the wall layer Pennemann, H., Hessel, V., Löwe, H., 2004. Chemical microprocess technology - from
in annular flows. AIChE J. 22 (6), 990–1000. laboratory-scale to production. Chem. Eng. Sci. 59 (22), 4789–4794.
Hessel, V., Löwe, H., Müller, A., Kolb, G., 2005. Chemical Micro Process Engineering: Pradhan, M.P., Suchak, N.J., Walse, P.R., Joshi, J.B., 1997. Multicomponent gas
Processing and Plants. Wiley-VCH, Weinheim. absorption with multiple reactions: modelling and simulation of NOx
Hewitt, G.F., Shires, G.L., Bott, T.R., 1994. Process Heat Transfer. CRC Press, Boca absorption in nitric acid manufacture. Chem. Eng. Sci. 52 (24), 4569–4591.
Raton. Ramshaw, J.D., 1980. Partial chemical equilibrium in fluid dynamics. Phys. Fluids 23
Hoftyzer, P.J., Kwanten, F.J.G., 1972. In: Nonhebel, G. (Ed.), Gas Purification (4), 675–680.
Processes for Air Pollution Control. Butterworths & Co, London, pp. 164–187. Rein, M., 1992. The partial-equilibrium approximation in reacting flows. Phys.
Hughes, D.T., Bott, T.R., 1991. The breakup of falling films inside small diameter Fluids A 4 (5), 873–886.
tubes. Chem. Eng. Sci. 46 (7), 1795–1805. Saber, H.H., El-Genk, M.S., 2004. On the breakup of a thin liquid film subject to
Hüpen, B., Kenig, E.Y., 2005. Rigorous modelling of NOx absorption in tray and interfacial shear. J. Fluid Mech. 500, 113–133.
packed columns. Chem. Eng. Sci. 60 (22), 6462–6471. Schwartz, S.E., White, W.H., 1981. Solubility equilibria of the nitrogen oxides and
Ide, H., Kariyasaki, A., Fukano, T., 2007. Fundamental data on the gas–liquid two- oxyacids in dilute aqueous solution. Adv. Environ. Sci. Eng. 4, 1–45.
phase flow in minichannels. Int. J. Therm. Sci. 46 (6), 519–530. Shao, N., Gavriilidis, A., Angeli, P., 2009. Flow regimes for adiabatic gas–liquid flow
Iwasaki, T., Kawano, N., Yoshida, J.I., 2006. Radical polymerization using microflow in microchannels. Chem. Eng. Sci. 64 (11), 2749–2761.
system: numbering-up of microreactors and continuous operation. Org. Process Sherwood, T.K., Pigford, R.L., Wilke, C.R., 1975. Mass Transfer. McGraw-Hill, New
Res. Dev. 10 (6), 1126–1131. York.
Johnston, A.M., 1986. Miniaturized heat exchangers for chemical processing. Chem. Thiemann, M., Scheibler, E., Wiegand, K.W., 2012. Nitric Acid, Nitrous Acid, and
Eng. 431, 36–38. Nitrogen Oxides. Ullmann’s Encyclopedia of Industrial Chemistry. Wiley-VCH
Kankani, V.G., Chatterjee, I.B., Joshi, J.B., Suchak, N.J., 2015. Process intensification in Verlag GmbH & Co. KGaA, Weinheim. http://dx.doi.org/10.1002/14356007.
manufacture of nitric acid: NOx absorption using enriched and pure oxygen. a17_293.
Chem. Eng. J. 278, 430–446. Triplett, K.A., Ghiaasiaan, S.M., Abdel-Khalik, S.I., Sadowski, D.L., 1999. Gas-liquid
Krauss, K., Diekmann, H., 1963. Process for the Production of High Percentage Nitric two-phase flow in microchannels Part I: two-phase flow patterns. Int. J.
Oxide. US Patent 3,110,563. Multiph. Flow 25 (3), 377–394.
Lam, K.F., Sorensen, E., Gavriilidis, A., 2013. Review on gas-liquid separations in Vankayala, B.K., Löb, P., Hessel, V., Menges, G., Hofmann, C., Metzke, D., Krtschil, U.,
microchannel devices. Chem. Eng. Res. Des. 91 (10), 1941–1953. Kost, H.J., 2007. Scale-up of process intensifying falling film microreactors to
Lee, J.J.L., 2012. Process Intensification of Nitrous Gas Absorption (Doctoral thesis, pilot production scale. Int. J. Chem. Reactor Eng. 5 (1), A91.
The University of Sydney). Wendel, M.M., Pigford, R.L., 1958. Kinetics of nitrogen tetroxide absorption in water.
Li, L., Breedveld, V., Hess, D.W., 2012. Creation of superhydrophobic stainless steel AIChE J. 4 (3), 249–256.
surfaces by acid treatments and hydrophobic film deposition. ACS Appl. Mater. Wille, C., Gabski, H.P., Haller, T., Kim, H., Unverdorben, L., Winter, R., 2004. Synthesis
Interfaces 4 (9), 4549–4556. of pigments in a three-stage microreactor pilot plant - an experimental
Manning, A.H., 1943. The production of concentrated nitric acid. J. Soc. Chem. Ind. technical report. Chem. Eng. J. 101 (1), 179–185.
62, 97–103. Wilson, S.K., Sullivan, J.M., Duffy, B.R., 2011. The energetics of the breakup of a sheet
Markowz, G., Schirrmeister, S., Albrecht, J., Becker, F., Schütte, R., Caspary, K.J., and of a rivulet on a vertical substrate in the presence of a uniform surface shear
Klemm, E., 2005. Microstructured reactors for heterogeneously catalyzed gas- stress. J. Fluid Mech. 674, 281–306.
phase reactions on an industrial scale. Chem. Eng. Technol. 28 (4), 459–464.

You might also like