You are on page 1of 415

Al'l'l.

ICATIONS OF FUNDAMENTAL THERMODYNAMICS


TO METALLURGICAL PROCESSES

APPLICATIONS OF FUNDAMENTAL THERMODYNAMICS


TO METALLURGICAL PROCESSES

of the First Conference on the


Thermodynamic Properties of Materials
held at the University of Pittsburgh

Procee~ings

Edited by G.R.Fitterer

111 lllllON /\Nil llltli!A< II f1C; J11:NOJJ: l'IJlll. ISIUiJHS


J ,ONI HlN

PAR rfl

CONTENTS
Page

'''

ix

word

l'ANEL I : THE CURRENT STATUS AND

VAILABILITY OF FUNDAMENTAL
I lllmMODYNAMIC DATA

Ji,,, 1 mrm :

N11 f 111

I,

Dr. Frederick D. Rossini, Dean , College of Science ,


Unme Un iversity

Add i tional Data Needed to Establish the


'l'h r modynamic Properties of Alloying Systems
h,y llr. Ra lph Hultgren, Professor of Metallurgy,
llnlv rsity of California

alc ulation of Phase Boundaries in


Mt Lnllic Systems
11.Y Dr . Oswald Kubaschewski, Senior Principal Scientific
111'1'1 e r , Nat ional Physical Laboratory, Teddington,
Mllcllosex , England
'l'h1

i1

ll:rl1 L of Temperature on Thermodynamic Interactions

I 11 ll I Lute Alloys
Ii v llr . John Chi pman, Emeritus Professor of Metallurgy

11111( II . A. Corrigan, Graduade Student,


111 11uhuset ts Institute of Te~hnology

23

111 1111.1 fl lutions in Metals and Alloys with


1111 I. I< u nr R fe r ence to Oxygen and Sulphur
l v 111 . 1:. 11 . Alcock, Professor of Metallurgical
1'h1111 try, I mperia l College of Science and Technology,
llu I 111 I t.y of London; currently Visiting Professor,
N11 1 I h 1:11 rn I Ina
tate Uni ve rs'i ty

39

Nill i 11111~ I 141.i LHdn.rd Refere nce Data System


l1v lr . l:v1111it. t. It. Johns on , Offi ce Of Standard Reference
I If 1 , NIf I 111m I nurcau of Standards

53

111

1111

I llll

67

PANEL II: APPLICATIONS OF THERMODYNAMIC


PRINCIPLES TO METALLURGICAL SYSTEMS
Chairman : Dr.W.E.Wallace, Chairman, Chemistry
Department, University of Pittsburgh
1.

2.

3.

4.

5.

6.

Electrode Reactions in the Electrolysis


of Fused Salts
by Dr.G.W.Mellors, Research Scientist and
Dr.Senderoff, Senior Scientist,
Parma Research Laboratories, Union Carbide Company
The Gaseous Reduction of Metals from Aqueous Solutions
by Dr.B.Meddings, Assistant Head, Research Division
and Dr.V.N.Mackiw, Director, Research and Development
Division, Sherritt-Gordon Mines, Alberta, Canada

105

Partition of Metals between Liquid Metal Solutions


and Fused Salts
by Dr.Irving Johnson, Associate Chemist, Chemical
Engineering Division, Argonne National Laboratory

153

Sulphate Stability and Thermodynamic Phase Diagrams


with Particular Reference to Roasting
by Dr.T.R.Ingraham, Head, Research Section,
Canadian Mines Branch, Ottawa

179

The Preparation of Pure Tungsten Metal by Reaction


of Tungsten Hexacarbonyl with Steam or Carbon-Dioxide
by Dr.J.J.Ward, Research Associate, Dr.J.H.Oxley,
Assistant Division Chief, Battelle Memorial Institute,
Columbus, Ohio, ~nd Mr.A.D . Coon, Graduate Assistant,
The Ohta State University, Columbus, Ohio

197

Discussion

221

...

'

PANEL III: APPLICATIONS OF THERMODYNAMIC


PRINCIPLES TO METALLURGICAL SYSTEMS(Cont.)
Chairman: Dr.J.Alfred Berger, Chairman, Metallurgical
Engineering Department, University of Pittsburgh
1.

81

The Use of Free Energy Balanc in Pro HR Ano. l yHiR


by Dr . O oritc It. l'it. Pi r1n1 , l'rof11HH11r Mt t.1d hat'Klf-1 I
En1tlr11tl'i 11K,

'l'lu~

Ohio HHll11 ll11lv111

vI

..

App lication of Thermochemical Data During


tainless Steel Melting
by Dr.C.W.McCoy, Senior Research Scientist,
Mr. A. F. Kolek, Metallurgical Technician, and
llr.F.C. Langenberg, Director, Process Research
1uHI Development Laboratory, Crucible Steel
l:rnn1lany of America

249

'I'll rmodynamics of Materials at High Temperature

Low Pressure
11.Y llr. Paul Gilles, Professor, Department of
l'hmn1stry, University of Kansas
1111d

281

A ll111111nary of Experiences in Applying The11modynamics


Io Mc tallurgical Processes
hv llr.G.R.Fitterer, Distinguished Professor,

I'

Mt 11 llurgical Engineering and Director of the


'l'ht11111oclynamics Center, University of Pittsburgh

299

Ill , l ' l lllH! On

325

1,1. IV 1 NEEDS FOR FUTURE


tt11 IC: ATI ON OF THERMODYNAMICS
11
11

111 ..J. F. Elliott, Professor of Metallurgy,


1111 1 11 Institute of Technology, Cambridge,
1111 t I f,

1111111I1111111 11L H for Space Materials


I ~ Ht , 1'h111'11iH E. May, Head, High Temperature
1111 1111 I ry Sc Lion , Lewis Research Center, National
1 11111111 I I H lllld Space Administration, Cleveland,
llh I 11

337

111 Al1111i11111111 Industry


I 111 I'. 'I', Hl.rn1111 , Assistant Director of Research,
1111N 11111111 l'o1111m11y of America, New Kensington, Pa.

343

t ' 111111 111t 111 1.J11 Non - F rrous Industries and a Note
1111 I I l 11111 t. I011 nf' II ILL apaci ty
! 1' 11111
111 11.11. Kt I log g , F.xtractive Metallurgy,
1 11 I11111111 It 1111 I v111 lt.y, Ntw York City

357

v 11

Unfinished Business in the Thermodynamics of the


Basic Oxygen Steel Making Processes
by Professor William O.Philbrook, Metallurgical
Engineering Department, Carnegie Institute of
Technology, Pittsburgh, Pa.

367

Vacuum Degassing of Liquid Steel


by Mr.J.E.Fogarty, Assistant to the Superintendent,
Smith Division, Republic Steel Co., Canton, Ohio

385

The Problems in Continuous Casting


by Dr.T.B.Winkler, Assistant General Manager,
Homer Research Laboratories, Bethlehem Steel
Company, Bethlehem, Pa.

391

7.

Discussion

399

8.

Participants

409

4.

5.

6.

'

v I 11

FOREWORD

1'1111 M1 Ll\llurg ical Engineering Department of the University of


l'I 11 1111 1 h hns f or many years given courses and conducted research
1111 11 l11vo lvod t he application of thermodynamics to process
11111 1111111.Y und , i n 1963, established a formal Center for the Study
, 1 1111 '1'h1 rm dynami c Properties of Materials.
Shortly thereafter,
1 111 I l11Lt n10.tionally known thermodynamicists served as visiting
11111 1 ml advis ors to the Center which is dedicated to the
f t he t hermodynamic properties of materials and to
tit 11 11111PLI nl applications to industry.
1111

I'1111I01'111

which is r ecorded in these Proceedings was

111 d11 l 111 I l.o oln id wit h a visit by Dr.O.Kubaschewski who is a


11l1i1 11111< IL>1Ll
ci ntific Off i cer of the National Physical
1111 11 1111111 1',V In T ddington , Midd l esex, England.
1111

I 111.rn l oh.I

ti v

of th

Conference was to bring together

n ll o v1r Lho world who were eit her c urrently con1111w t.h1 r 1110<1.v1mmI dnLn LLlld pr i ncipl s or who were
l.11 1L1H1l ,Y Liu 111 Ln 1io 111111t r 1111 m t ullurgical pr c ss

1111111

Ix

The two-day Conference involved four panel sessions, each of which


consisted of four or five papers by distinguished panelists together
with intra-panel discussions as well as some questions from the
audience.
The First Panel discussed 'The Current Status and Availability of
Fundamental Thermodynamic Data". The Second and Third Panels involved papers on the 'Applications of Thermodynamic Principles to
Metallurgical Systems", and the Fourth Panel was concerned with the
'Needs for Future Applications of Thermodynamics' .
The Conference, which was one of the first of its kind which had
the stated objective of the reduction of this fundamental science
to industrial practice, was attended by some 350 scientists and
engineers. The high degree of interest in this practical approach
to the subject was expressed by manyof the conferees, both during
and after the meeting. Many suggested that another conference o~ a
similar type should be held in another year or two.
The staff members of the Metallurgical Engineering Department of
the University of Pittsburgh and especially those of the Center wish
to take this opportunity to express their sincere appreciation to
the panelists and to the conferees, all of whom contributed
immeasurably to the success of the Conference.
The assistance of the National Aeronautics and Space Administration
in making this work possible through research grant NsG-416 is
greatly appreciated .

G.R.Fitterer
Director of the Center

March 1967

'

PANEL I
THE CURRENT STATUS AND AVAILABILITY
OF FUNDAMENTAL THERMODYNAMIC DATA
Chairman
Dr.Frederick D.Rossini
Dean, College of Science, Notre Dame University

Acl(! Uonal Data Needed to Establish the

.r modynamic Properties of Alloying Systems

'l' ht

h.v llr. Ralph Hultgren, Professor of


llnlv rsity of California

~1etallurgy,

'1'111
Ltlc ulation of Phase Boundaries in
Mo I.id lie Systems

l1y ur. Oswald Kubaschewski, Senior Principal Scientific


0111 c11r , National Physical Laboratory, Teddington,
Ml1hl11 HCX , England
11:1 !'1 11 1, f Temperature on Thermodynamic Interactions
111 fl I l 111, Alloys

l1v 111.. John Chipman, Emeritus Professor of Metallurgy


1111cl II . A. Corrigan, Graduate Student,
Mn

tt

tdm

Institute of Technology

111111!,11 Hn lntions in Metals and Alloys with


1'111 f lc11l1u Refere nce to Oxygen and Sulfur

l1v 111 . I'. II. Alcock, Professor of Metallurgical


I h1111d tr.v , Imperial College of Science and Technology,
lhilv111 11,y of l.omlon; currently Visiting Professor,

"' t I l'11r11I hm State University


N11 I I 11111 I

~1LLu 1d n.rd

Rof re nce Data System

I II , 1: vlll'11tt. It. ,Johnson, Offic e of Standard Reference


lh I 11, N11 t I mm I 1111 f'l'llU of Atandards
llJ

Ill

11111

PANEL I
Dr. Frederick D.Rossini
1,1 dies and Gentlemen, it is a great pleasure for me to open this
I I 1rt l. Hess ion of the Symposium on Applications of Fundamental Thermoilv1111m I s t o Metallurgical Processes . As Dr.Fitterer has said, this
I 1111tainly one of the most important subjects facing us today.
11!1 l.n11ic for discussion under Panel I in this first session is the
1 111 ,.,111, Status and Availability of Fundamenta1 Thermodynamic Data.
' 11 I I lcnow from experience how seemingly little thermodynamic data
I l.11 I.hat we would like to have available for us, but at the same
I 11111 w11 recognize that there does in fact exist an enormous quantity
111 I ht rrnodynami c data of all kinds scattered throughout the literal 111 11 I' Lhe world. our big problem is to know precisely what data
I I ILll<l how to put them in their proper frameworks so that they
1 111 111 11 H d easily and efficiently.
Next we need to evaluate these
11111 1111d t hen to determine what data are missing from the literature
1Ii111111 hont t he world and what new data must be obtained. This brings
f 11 Lilt big pr oblem facing science today - the storage and
I 1I1 vr I f quantitative scientific information. Many simple facts
11 II c oct d, stored and retrieved by machines, but the problem of
1111d ret rieving what we may call quantitative scientific
1111111111 I. I on is not so simple, because careful appraisal of much
1111 I ~ I tm I materi al is required to recover the full significance and
of such data.
these ki nds of information, thermodynamic data are
111 st i mportant. Data on the thermodynamic properties of
t (:ul materi als and material systems are becoming more and
I1 l in our modern science and technology. I think we can
t.h1 ndvance of science r ests heavily upon quantitative
The better and more accurately we can
I 1 1111!1 me uHur , t he better and more precise l y we can develop
11 1 11 I 1 1 Lo 1 x.1>lain the nat ure of t hi ngs in our world.
To give us
111 11 111. Into t h current status and availability of fundamental
'"'"tlv11111111i' rlL n, wo have a pane l of five eminent scientists with
1 w1111 ld 11111 now to introduce t hese gentlemen to you in the
r I 1 I 11 wll I1 h I. hoy will present t heir r eports. Beginning on my
ltt 111111 v1111 r 1n l'L Aid Dr. Ralph Hultgr en, Profes sor of
f California at Berkeley; Dr .Oswal d
1 111 w h I , Ile 11I11r l'r I 11 c I pn l
c I c nL ifi Offi r at t he National
t 11 l 1,11l111111t 111,v l11'1'11dcll11 "l.011 , l:1w. l1L11cl , who IH urr ntly visiting
I
1al lh1 1111lv1111 1t ,v 111 I' ll.I. ll11 r11h: l)1..Joh11 C:h11J1n11n,
t ht
11111
1 1111
I hnt

:1

DR.FREDERICK D.ROSSINI
Professor (Emeritus) of Metallurgy at the Massachusetts Institute of
Technology; Dr.C.B.Alcock, Professor of Chemical Metallurgy at the
Imperial College of Science and Technology in London, England, who
is currently visiting professor at the North Carolina state
University; Dr.Everett R.Johnson, of the Office of Standard Reference
Data at the National Bureau of standards. our session this morning
will consist of presentations of about 15 or more minutes in length
by each of our five panelists . This will be .followed by an intra- .
panel discussion of approximately 20 or 25 minutes . This will leave
about a 30 minute period in which we can have questions and answers
involving audience participation. For this last question and
answer period, there will be cards distributed to you so that you
can write down your questions and the name of the panelist to whom
you wish to address your question . These cards will be collected in
time to be activated at the final question and answer period. l
hope very much that every member of the audience will feel free to
participate actively in this question and answer period.
our first paper is by Professor Ralph Hultgren, who will discuss
Additional Data Needed to Establish the Thermodynamic Properties of
Alloying Systems .

'

ADDITIONAL DATA NEEDED TO ESTABLISH THE


THERMODYNAMIC PROPERTIES OF ALLOYING SYSTEMS
by

Ralph Hultgren
Professor of Metallurgy , University of California
It would be much easier to describe the data we have than what we
because we have so little and need so much . The data can be
ii 1rwussed under three categories :
111111<1

(1 )
(2 )
(3 )

Elements (unary systems)


Alloys (binary, polynary systems)
Interstitial alloys (B, C, 0, H. N, with transition metals)

Wlt h your indulgence, the statistics on the first two categories


w11 I be derived from the book, 'Selected Values of Thermodynamic
1111)).rt ies of Metals and Alloys' . This book contains substantially
11 I I Lhermodynamic data for unary and binary metallic systems up to
I Ii <:los i ng dates which are, on the average, about 1961. Systems
w11 11 Incl ud ed only i f they had thermodynamic measurements . Systems
t 111 which there were phase diagrams alone, for example , were not
11 I11uted, although phase diagrams were used to evaluate and expand
1111 1111 11 r ed data.
M Lo.Il i c systems were defined as those which conducted electricity
11111Llon of electrons; about 80 elements including 11 trans - uranic
I 111111111 Ls wer e included. Metallic and semi metallic phases formed by
LI.
N, B in interstitial compounds or solid solutions were not
I w I 11d1 ct in the first edition _

It

Ill t h 80 e l ements 64 were included. Those omitted were 10


n 1111 11ranic e lements (all except Pu). and Tc, Po, Fr , Ra, Ac, Pa
11d 1'111 . The omissions , as can be seen are very natural in view of
1111 ii I f'f'ic ulty of obtaining samples of the ra:rer elements.

11 11. In nt meas ur ements of elements are the low- and high' 1111111111L1,ur heat capaciti es and enthalpies and heat capacities of
" 11
11il ulat ed from spectroscopic data. From these measurements
111 1111 01il u1ated th e ent halpies , entropies, and Gibbs energy
111111111111 1 of t h
1
111
"

111111 I. c1L111 ti Lh no l'IH ILl'Y m 1.1.1:1ur m nts hav been made wi th


mt 11 1'1L('.Y !'01 I. ho 1o 11 (J 111) l.o 1,11 1 m Hing point or . for

11 111111 1! I 1

RALPH HULTGREN
higher melting metals, to temperatures well over 1000c. Techniques
for heat content measurements at truly high temperatures need to be
developed.
The paucity of liquid data for the elements is, however,
astonishing. It would seem that these measurements went out of
fashion forty or more years ago . Data on even simple and important
metals date mainly from that time and are often seriously in error.
It is probably true that liquid heat capacities decline with
temperature increase; the measurements are sufficiently accurate to
show this in only a few cases. As an example, our laboratory had
occasion to measure liquid tin and liquid silver recently and found
notably different latent heats of fusion and heat contents from the
best values found in the literature . The state of affairs is shown
in Figure 1.
Binary alloys are much more inadequately studied. Of 3160
possible combinations, only 166 are listed. When it is remembered
that the data on some of the listed systems are very sketchy indeed,
the magnitude of the work to be done can be appreciated. Trans
uranic metals and rare ones (Tc, Po, Fr, Ra, Ac, Pa) are not
represented at all and only five systems involving lanthanides were
found. Of the remaining 49 metals less than 6% of the systems are
covered . This distribution of the systems is shown in Figure 2.
Most of the systems are between the more common, low-melting
metals. The transitionmetals, except for iron, are a vast wasteland. The intensity of coverage of the 12 systems involving iron is
usually low. It would seem that in this area data badly needed for
modern technology is almost entirely lacking.
The reasons for the neglect are technical . Transition metals are
high-melting and easily contaminated especially by elements difficult
to detect as C, H, ...N, 0, etc:' Acid solution calorimetry is
inaccurate and most of the metals will not dissolve with sufficient
rapidity in solvents such as tin. The products of combustion bomb
calorimetry may be complex and difficult to quantitatively identify.
Equilibrium emf measurements have to be carried out a high temperatures where techniques ..have not been well developed; moreover,
electropositivities may not differ enough for easy measurement s .
vapor pressure measu r ements lo.ok t he most promising, but t chn iques
fl

::
--~ - -

H I
I

IO

12

II

13

14

15

16

17

18

He

I
I

He Li
2

Ne
3

Ar

Be

Si

Na

Mg

Al

:-Mg : Al

: 0

: 0

ca

SC

Ti

Cr

Mn

Fe

co

Ni

cu

Zn

Ga

Sr

Ne

Cl

Ar

Kr

Rb

Zr

Nb

Mo

Ma

Ru

Rh

Pd

Ag

Cd

Ge

As

Se

Br

::a
0

In

>--3

Sn

Sb

Te

cs

Bo

LO

Hf

Ta

Re

Os

Ir

Pt

Au

Hg

Tl

Pb

Bi

Po

At

Rn

Fr

Ra

Ac

Th

Pa

Np

Pu

Rn

Arr. Cm

7
0

Kr

!
.....
(")

.....

l:r:l
r:tJ.

>
~
~
.....
~
~

r:tJ.

f}3
Ce
Ga

Fig.1.

Pr

Nd

11

Sm

Eu

Gd

Tb

status of Heat Content Data for Liquid Metals.

Dy

Ho

Er

Tm Yb
0

Lu
0

X =satisfactory, O =unsatisfactory

- 0n

-'H-,

10

12

II

13

14

15

16

17

18

He

I
I

He Li

Be

:-Mg : Al

Ne

Si

Cl

Ar

Ge

AS

Se

Br

Kr

2
g

Ne

No

Mg

Al

Ar

10

16

15

ca

SC

: 16 : 15

Ti

er

Mn

Fe

co

Ni

Cu

Zn

GO

11

Rh

Pd

Ag

Cd

15

17

13

21

Ir

Pt

Au

Hg

Tl

Pb

16

17

13

Rb

Sr

Zr

Nb

Mo

MO

cs

Bo

LO

l;lf

To

Re

Rn

Fr

RO

Ac

Th

Po

Np

Pu

. 0

Ce

Pr

Nd

11

Sm

Eu

Gd

12

Ru

16

16

In

Sn

Sb

Te

5I
0

Os

19

Bi

Po

At

18

16

Tm Yb

Lu

Am Cm

Tb

Dy

Ho

Er

60I
0

::r:

Cl

Fig.2.

;g

Kr

==

Number of Binary Systems with Significant Thermodynamic Data

Rn

THERMODYNAMIC PROPERTIES OF ALLOYING SYSTEMS


1110 dif ficult . The problem of concentration depletion of the more
vo Int il e component on the surface has not been adequately handled.
I nte rstitial solid solutions and compounds are becoming
1 11ormous l y important to our technology and have been scarcely
I 11 11 hed, except for the iron-carbon system. These systems occur
Iii I. ween B, C, 0, H and N with the transition metals . (The columns
11 t Lhe periodic table headed by Sc to Ni which contain incompletely
ti 11 d ct-shells of electrons.)
Many phasefi in these syst ems may vary in composition over wi de
1 lm1ts; a fact which is incompletely recognized. Two investigators
1111 y i>t udy the same carbide phase, for example, but get widely
d I f' f' orent results because their samples differed in composition.
1'li rn11 ic al analyses are difficult and, because of the widely different
1it rnnlc weights of the components, small differences of weight
'" r ntages may reflect large differences in the more significant
111 llmlc pe rcentages . Techniques of measurement are even more
I 1111(! quate than in the case of the transition metal alloys with one
111111!. her .
I It mcuts

Binary Sys tems


No .

With Data

Total

'l'r 1111 H urani c

11

814

1,11 11!. hani des

14

1015

459

111111 r M tals

49

166

2695

111 111 1

80

166

3160

1t11rn M t als
( I f , ~ I Po , Fr, Ra, Ac, . Pa)

r ~ l~ . 3 .

Binary Systems with Notable Thermodynamic Data Classified


By Elements Composing Them

1:v m Lh m t hod of calculating and recordin g the r esults is often,


111v oplul
ar i ly complex and confus ing, though I think
I I 1111 Id I(( I.
Us ually the attempt is made to
il1111111t.11 1 l i;.1 u llht ~H t hy u omvo Aj l on Axily wh r x and y are small
1111111 1'1. '1'111 1 101111)01 11, ln n mn.,y h1 111 lucl d In th o ro.n g
f
I.lit 11h1111, nr 111 11.v ht 1111111 1,v 111 ur to1111101 I l nrw whl h
111

RALPH HULTGREN
are stable . Composition variance of the phase is then treated as
solid solubility with, perhaps, assumptions made as to the ideality
of the solution .
It would take too much time to show how this leads to confusion
in some cases . In my opinion, these phases could be treated as are
alloy phases of variable composition . Partial and integral quantities
should then be tabulated as functions of composition.
The amount of data in this field is small but rapidly increasing.
I have at present no way of citing statistics of coverage, but it is
certainly small. We hope at my laboratory to develop these
fragmentary ideas I have mentioned within the next few years .

...

.1

10

THE CALCULATION OF PHASE BOUNDARIES


IN METALLIC SYSTEMS*
by

O.Kubaschewski
Senior Principal Scientific Officer
National Physical Laboratory, Teddington, Middlesex

All applied chemical thermodynamics is essentially phase diagram


wnrk. Given the constituents of a system and the external condi1.lons, the phases present and the concentrations in equilibrium are
worked out. The conventional methods of analysis, such as chemical
micr oscopic, thermal and x-ray analyses, have been devised to give
l,hese very answers. Considering the difficulties often involved in
th rmochernical measurements as compared with the conventional
111 t hods. where does the advantage of using thermochemical data for
l. h calculation of phase boundaries lie?
I n this paper the author will try to answer this question on the
l1111-1is of a few simple examples. The main reason for the frequent
f'utlure of the conventional methods to give the exact positions of
uhnse boundaries is kinetic in nature.

The determination of a solidus curve by thermal analysis, for


l m1tance, involves diffusion in the solid state. The concentration
111' t he solid phase initially precipitated on cooling differs from
1.111 f inal equilibrium concentration.
The time required to attain
111111 l li brium is difficult to estimate and is frequently under"' I. I mated by the investigators.
A case in point, what we may now
1 1t l I a classic example, is C. Wagner's calculation of the solidus1111 11 lrtus gap in the system silver-gold from known thermochemical
ilnl.iL. 11 l For the width of the gap at 50 at.%, Wagner obtained 6 T'
I :1llc while the then known experimental value for the interval

Jud W n beginning and end of solidification was 6 T = 10c. Very


I li111ough thermal analyses by White < 2 l and by Cooke and Hume11111 111 ry < 3 l later confirmed the calculated value within a few tenth
" ' It d gree .
One wonders how many of the solidus curves in the
1
1 L1ihlished'
phase diagrams (e.g. Hansen 1" l) can really be relied
11 1111 11 .

LI rJuidus curves on the othe r hand obtained by thermal analysis


1111 nnrmnlly b r gard d as r liable and as accurate as claimed by
l' h11 1011L1nl1-1

1,, - 1,M111 ,

or
u:1,

l, hl r1 Pll1 r hlV l hi II p11lll1Hh1d 111 pul't t n


:i~w
IOflfo) .

It

O.KUBASCHEWSKI
the investigators, but even in certain liquid alloys equilibration
may be delayed by slow diffusion. This applies in particular to
metals that exhibit pronounced short-range order in the liquir. state
(e.g. Bi, Ga). One can assess the partial heats and entropies of
solution in the system gallium-zinc, for example, from the
calorimetric work of Kleppa (SJ and the e.m.f. measurements of
Svirbely and Read ( 6 J and Genta, Fiorani and Valenti ( 7 l and
calculate the liquidus curve by means of the equation
(1)

where .6.Hi and .6.si are the partial heat and entropy of solution of
1 g-atom i (= Ga or Zn) in an alloy of given composition, and Li
and cri the heat and entropy of fusion of pure i. The small solid
solubilities may be disregarded .
The resulting liquidus curve was found to disagree near the
eutectic composition with the earlier experimental results:
Figure 1. An attempt to check the results by normal thermal
analysis by Chart (BJ produced similar results to the earlier ones,
the alloys showing a very strong tendency towards super-cooling.
200;--------------~

Put.chin

c t . al

Valent i

and

.6

Heumann

Fioroni

and

Prcdcl

C.olculotcd

Prcunt
Zn

work

..

cryatal found

150

...

.u

100

/;

.1/

...

>!

so

/0
.

G'l

I
"

I' I IC . I .

Ot - 7.11 11:11L1 r

10

20

(, !ti f;(lffi(loil
( '.

I I, I rm Corre

50

Li d

THE CALCULATION OF PHASE BOUNDARIES IN METALLIC SYSTEMS


Only after annealing for many hours near the liquidus could the
thermal arrests indicated by crosses in Figure 1 be obtained. These
r sults agree well with the calculated curve and also with those of
Lhe thermal analyses of Heumann and Predel < 9 > which were, however,
11ot extended to below 10 at. % Zn.
The advantage of the thermochemical method in such a case is
obviously that the measurements can be made at some convenient,
hlgher, temperature where the rate of equilibration is sufficiently
t"O.pid not to interfere with the measurements, and then used for the
nalculation of the phase boundaries at the lower temperature.
Similarly, the boundary of a miscibility gap in the solid state
temperatures where diffusion rates are slow may be obtained by
1111 as uring the thermochemical properties of the solid solutions at
1 me higher temperature. Where the excess free energies are positive,
t.h so lutions are expected to form a miscibility gap at some lower
l.1mperature. For regular solutions, the integral entropy of
t'ormation is given by the equation for the positional entropy
11t

(2)

wh re N1 and N2 are the atomic fractions of the component metals,


1111d the integral heat of formation by the following equation
(3)
wli11r o b is a constant. Figure 2 (upper diagram) shows a T6 Sf curve
111Ht a 6 Hf curve which have been constructed in accordance with
11m tions (2) and (3) respectively. It is the characteristic in
1l111no near Ni = O and Ni = r that determines the shape of the free
1 111 1gy curves at lower temperatures.
Free energies of formation are
11 ht.1Llnod by means of the Gibbs-Helmholtz equation: ( 4)
11 1111

I.It

I 11w1

r sulting curve has been drawn on an enlarged scale in the


diagram of Figure 2. It is eas ily seen that the mutual
1il11hl lt ty n.nn t b com z ro at fi nite temp ratur s.
I 1111 r

I 11

11

ti. 111 1 l wt .YM

h1

Dh 111

11111 l'IU t.11111. 111 11 1.1h11,

I.he

nr mtxL11r1

o f' Dhnfll H wi t h Lh!

t.1111"1 nt. t.n I.Ill /\ Or <:11 1v1

low1 HI.

hml 1Ll1H1

O.KUBASCHEWSKI

ATOM IC

FRACTION

6.G
ATOMIC

Fig.2.

FRACTION

Atomic Fraction

been drawn, the points of contact representing the phase boundaries


at temperature T. Using the 6 Hf and 6 Sf values, the phase
boundaries can now be calculated at any temperature, assuming that
the heat capacities are additive. We normally use the graphical
method of evaluation.
In this manner, the sys tem chromium-molybdenum was studied which
shows complete mutual solid solubility at high temperature.
T. G. Chari < 81 measured the heats of formation at 1400c in a hightemperature calorimeter des~ gned by W.A.Dench, and M.Laffitte t lOl
the dissociation ~pressures ' at 1200c by an effusion technique
employing radioactive chromium. From these results , the 6 Hf and
6 Sf curves vs. composition could be assessed and the boundary of
the miscibility gap calculated in the indicated manner - see
Figure 3.
In ord er to check t he r esults of t he calc ulation , homog nous
im ns ontaining 40 and 0 nL. % wor e 1U1 11on. h <I o.L v11r1ou
I.< rnn< rn l.11 r11 1 !'or 1ir vt ntl 111<111l.h11, l. ht1 1100 C npc <: !mt n 1 Vt 11 r nr mflr<
p

THE CALCULATION OF PHASE BOUNDARIES IN METALLIC SYSTEMS


Lhan a year. In Figure 3, the specimens showing x-ray line separaare indicated by half-filled circles, those that remained
homogenous by plain circles. The very good agreement with the
c:1Llculated phase boundary is coincidental because the accuracy of
l. h calculated curve is estimated to be not better than 40c.
1, I on

The advantage of the thermochemical method of phase boundary


valuation is quite clear in this case. If these alloys were to be
0
11 H d in practice at say 75o c the equilibration would take too long
!'or conventional methods to be employed for the determination of the
11o lubilities. With regard to the alloys annealed at 650c, it can
hi estimated from the time taken at soo 0 c and the diffusion rates
I.hat a distinct separation of the x-ray lines would be attained
1Lfter 250 years, that is, for most practical purposes the alloys
will remain homogenous at that temperature.
1

or

The thermochemical method may even be useful for the calculation


miscibility gaps in the liquid state. Figure 4 shows the great

oc
1000

40

60

900

800

700

600
Mo

20

Cr

At 0 /o Cr

l'lv. . a.

80

Cr - Mo n11anv1r.v ol' 1i Ml 11c1bil1Ly


Ill

OtLl)

i n th

Solid State

O.KUBASCHEWSKI
Bi-Zn

6 Sprin9 and Romanoff


X Hoss and Jtllinek
'\

700

O Seith and Johnen


- Calculated

600

500

400

300

Bi

02

0 4

06

08

Zn

-Nzn

Fig.4.

Bi-Zn Decision between Divergent Experimental Results

discrepancies in the results concerning the gap in the system


bismuth-zinc. The reason for this probably lies in the small
differences in free energy between the mixture and the component
solutions, that is, in the small 'pulling force'. The calculation
clearly confirms the phase disgram work of Seith and Johnson. There
are still many uncertainties concerning the limits of liquid
miscibility gaps in the 'established' phase diagrams.
In view of the pract i cal significance of u phase formation in
steels , the present author' s group investigated the chromium-iron
system. Dench (lll measured the heats of formation at 1138 - 1313c
calorimetrical!J. He als~ measured heats of transformation of u .
to a, and his results are in good agreement with preliminar y -calorimetric studies by Backhurst < 12 l and Martens and. Duwez (l 3 l.
Heymer < 14 l determined the dissociation pressures in an effusion
cell at 1341 - 1370c.
The results differ only slightly from those
of McCabe, Hudson .and Paxton (lS>, but are supported by those of
J eannin, Mannerskantz and Richard son (l 6 l who investigated equili bria
bet w n Cr - F alloys , Cr20 3, H2 at 1040 - 1300c.

THE CALCULATION 01' Pll ASE BOUNDARIES IN METALi..IC SYSTEMS


Wi th all this information, it has been possible to evaluate the
Only the heats at the
' ' - a- transformation (43-49 at. %Cr) have been measured; in order to
nl> t ai n also its entropies, the transformation temperatures near
11 00 c determined by Cook and Jones <1 7 > have been used. The data
I or the transformation together with those for the corresponding a
11 l l oys then give the heat and entropy of the a- phase between 43 and
110 at. %Cr.
IH nts and entropies of the a Cr-Fe alloys.

The mode of calculating the phase boundaries is illustrated in


l'lgure 5. At 600c, for instance, a- is in equilibrium with iron
l'I ch and chromium rich a alloys.
Above 805c the t::. G curv e for ow111 be found to lie above that for a ; in other words, only a is
11 1,ab le . It may also be seen from Figure 5 that at 460c, the t::. G
c:urves for (1) iron rich a, (2) o-, and (3) chromium rich a have a
c:ommon tangent, that is, below this temperature, a- disappears as a
11 Lable phase and a miscibility gap a + a ' takes its place. The
1a1lculated equilibrium diagram <8 > in the temperature range
:100 - 900c is shown in Figure 6. It supersedes an earlier
11
1 Lt empt<
> to calculate this diagrram before the calorimetric
111 asurements had been completed.
There is some independent evidence that the a/a + a- phase
houndary suggested by Cook and Jones should lie further to the l ofL,
t1nd a phase diagram based on direct studies proposed by Williams ru 11
l'uxton <18 > ~esembles the present one in most respects.
The calculated eutectoid temperature may be in error by 50 ,
1111d it seems to be possible, or even probable, that the so-call (!
'11 75 brittleness' in Cr-Fe alloys is not due to the format ion of' u
hu t t o the separation of the a solution into a chromium-rich and
I n m- rich ,phase.
Williams < 181 found the eutectoid temperatu r ul,
120c which is possibly more accurate.
Sti ll , the advantage of applying thermochemical data to th
Measurements t hat
ll t1vc been made between soo 0 c and 140o0 c where equilibrium is
1 cmsonably rapid can be used to calculate the boundaries at much
low r temperatures where equilibrium is attained much more s l owl y
li ul. at which temperatures chromium steels are actually used.

de L r mination of phase boundaries is evident.

17

O.KUBASCHEWSKI
E

...
0
0

+100

I
C7'

c:
0

-100

....0
0

-200

>C7'

L..
~

c:
~

-300

~
~
L..

u..

-400

20

Fe

40

60

80

Cr

Atomic percent Cr

Fig . 5.

Cr-Fe Demonstration of Phase Boundary Calculations in a


System Forming a er Phase as well as a Miscibility Gap
- - Cook and Jonn
Colculotcd
----- Colculotcd - "1cta1toblc
800

ot.

I'

I ..
/.
600

o<.'

I / 1a

700

I
I

.(

. /
:. I

ol'..+

3+ct.'

500

400

ol+oL'
JOOL-~---'~~~~~~~~~~~

Fe

20

60

40

80

Cr

Atomic /o Cr

Fig. 6.

Cr-Fe Thermoc hemica l Ass ssm nt of a T hn ical.ly


T111porLa. 11 L YH ( m
I ll

THE CALCULATION OF PHASE BOUNDARIES IN METALLIC SYSTEMS


At the present stage it will hardly be possible to derive complete
11quil ibrium diagrams, except for the very simplest, entirely from
1.11 t hermochemical properties. In the case of the assessment of the
('r - l~e diagram, for instance, the maximum transformation temperature
r11 r a- - a obtained by direct evidence has been incorporated. For
11 quid -solid equilibria it appears to be generally safer to adjust
I ht available thermochemical data for the liquid and solid phases in
r 11 h a way that the experimental liquidus curve is obtained and then
l.n a l culate the corresponding solidus curve(s).
n fo r e one can rely on thermochemical data alone, the methods of
111111LH urement must be further improved. A glance at Figure 5 shows
1111 orde r of the free energy differences to be expected . If one
I li!l ll adds that few of the available calorimetric and free-energy
1111 1!.hods produce results with accuracies much better than 100
1 11 I I rr-atom the point should become clear.

IL is significant that assessments of the thermochemical


1111111o"tie s of alloys, such as that by Hultgren and his colleagues ( 19 >
r 11q 11 nt l y use established phase boundaries as part of the assess1111 111. rat he r than vice versa.
l 11 Prt.unat

e ly, more accurate thermochemical data are gradually


111'!. he mi ng, and more reliabie calculations can be made .

11:vo n so, the thermochemical method cannot fully replace the con1111L I 011 al methods, for the simple reason that all the stable phases
I 11 IL 11,YHtem must first be known before the thermochemical investi1111 I 1111 an begin, and, of course, structures, densities and other
11111111 rLl s of the various phases are also required in addition to
1111 11huH boundaries.
lf 11w1 v r, t he restricti on that the thermochemical method depends
Lht Aupport by conventional methods is overcompensated by the
1111111 1111, of ad dit ion a l information it presents. The equilibrium
'd I 11 111i11 1.1; part of the free en ergy diagram and not vice versa. In
11 y nrllh t of process metallurgy it will be necessary to know the
11111 11 111 t. ho.t will be pr esent i n equilibrium, but the knowledge of
. t h1 r111t
n rgy r lations hips gi ves additional information on the
r 1111 I 1 <.1111 llbrium con ntratiom1 and pressures . Th is appli es , of
11111 1, t.n 1 11 t.ypoA of mt allurgi 1 r n 1011 1::1 wh r
quilibri um
Hll ,

0.KUBASCHEWSKI
is approached or attained, be it the production and/or refining of
metals and alloys, be it contamination of metals by refractories or
similar problems. The thermodynamic assessment of the purification
of 'spent' reactor fuels by the method of partition between a liquid
metal and a fused salt presented by I.Johnson to this conference is
a good example that demonstrates the usefulness of thermodynamic
data of alloys.
The method of evaluating phase boundaries from such data is, of
course, not confined to alloys : salt mixtures may be similarly
treated. Examples of this application were given in an earlier
paper ( 2 0 >.
The thermochemical method may eventually acquire some considerable
significance for the estimation of phase boundaries in ternary and
multicomponent systems. In a ternary system, the free energy curves
of the binaries assume the form of free energy surfaces and the
tangents become tangent planes. Graphical evaluation would be too
cumbersome, and Gomputers will have to be used.
It is envisaged that the heats and entropies of the ternary
phases may often be estimated from the thermochemical properties of
the corresponding binary phases. Much more experimental work on
ternary alloy systems is thus called for in order to establish
these relationships.
Figure 7 shows an attempt to calculate the phase diagram of the
iron-chromium-nickel system at 65o0 c. No claim is made that the
diagram is even approximately correct; it is supposed merely to
serve as an example of . what one might aim at. The thermochemical
properties of the binary diagrams are known. For the sake of
simplification it was assumed that the excess free energies of the
ternary alloys can be obta~~ ed additively from those of the
isostructural birt"ary phases and that the entropies of the ternary
alloys can be calculated by means of the equation

Of course, it will fake a long time of f urther experi m ntal work


b fore som such m t hod of yaluati.on can b xp ct d t giv
r n:; nLLbl y 11. c:11n1t llllflW r:; for Wh l
I, I ll H f I, rn11ry <I I rtv.r 01n A.

THE CALCULATION OF PHASE BOUNDARIES IN METALLIC SYSTEMS


Cr

F~-cr-Ni

650

- - -

c
Cook

and

Brown

0 8

~LJ.<~--''---;o~-8;;-~~----;;0~6:--_____
~~~0~4~~~~0~2=--~~---lNi
N Fe

Fig.7
REFERENCES

C. Wa gner, Acta.Met. 2, 242 (1954) .


l . I..White, Trans.Met.Soc.A.I.M.E. 215, 178 (1959).
t: ..J.Cooke and W.Hume-Rothery,

Acta.Met. 9, 982 (1961).

M. llnnsen and K.Anderko, Constitution of Binary Alloys,


MdJ raw-Hill:
New York (1958).

0 . .J . Kl e ppa, Acta.Met. 6, 225 (1958).


W. .J. virb ly and S.M.Read, I.Phys.Chem. 66, 658 (1962).
V. (hl 11Lu, M.Fiorani and V.Valenti,

Gazz.Chim.Ital. 85, 103

( l ll!1!i) .

11

u . K11lmt(l lt11w 11 lll 1rn<I '1 '. Cl.Clmrt.,


'! l

.J . lr1 .~ t.M11t.

!J:'J, 32!1 (H)fi!1) .

O.KUBASCHEWSKI
9.

T.Heumann and B.Predel, Z.Metallkunde, 49, 90 (1958).

10.

M.Laffitte and O.Kubaschewski, Trans.Farad.Soc. 57, 932 (1961).

11.

W.A.Dench, Trans.Farad.Soc. 59, 1279 (1963).

12.

I.Backhurst, I.Iron and Ste e l Inst . 189, 124 (1958).

13.

H.Martens and P.Duwez, I.Metals 8, 614 (1956).

14.

O.Kubaschewski and G. Heymer, Acta.Met. 8, 416 (1960).

15 .

L.McCabe, R.G.Hudson and H.W.Paxton, Trans.Met.Soc.A.I.M.E.


212, 702 (1958).

16.

Y.Jeannin, C.Mannerskantz and F.D.Richardson, Trans .Met.Soc.


A.I.M.E. 227, 300 (1963).

17.

A.F. Cook and F.W.Jones, I.Iron and S te e l Ins t. 148, 217 (1943).

18.

R. O.Williams, Tr ans. Met. S oc .A.I. M.E. 212, 497 (1958) .

19 .

R.Hultgren, R.L.Orr, P.D.Anderson and K.K.Kelley, Selected


Values of Ther mody namic Properties of Metals and Alloys,
J . Wiley : New York (1963).

20.

O.Kubaschewski, Tr ans .Brit .Ceram .Soc. 60, 67 (1961).

'

'

..

EFFECT OF TEMPERATURE ON THERMODYNAMIC


INTERACTIONS IN DILUTE ALLOYS
by

John Chipman* and Donald A.Corrigan**


Massachusetts Institute of Technology
INTRODUCTION
In a dilute, multicomponent solution of several elements in the
solvent, component 1, the activity coefficient of component 2 is
expressed < 1 > by therequation
Zn y 2 = Zn y
where

E ( 3)
2

+ E2 ( 2 ) x

= (o Zn y~ Cl x 3 )

+ E 2 ( 3 ) x3 + E 2 ( 4 ) x 11 +

xl .....

1.

(1)

etc., and the derivatives are

taken for the limiting case of the infinitely dilute solution. It


has been found in many systems that the derivative Cl Zn yi/Cl xj
retains the value characteristic of infinite dilution throughout a
finite, sometimes large, range of concentration. If we neglect
secondary effects which are generally small, and adopt the
infinitely dilute solution as the reference state so that y~ = 1,
Zn

-v.
1

= 2.:.1 [E~1 x.]


J

(2)

When a third component is added to the dilute binary solution of


component 2 in the solvent 1 the chemical potential of component 2
is increased by an amount,
F x3
2

= RT

E ( 3 )x
2

(3)

The quantity F 2 x3 is the extra partial molal free energy of component 2 occasioned by the addition of component 3. This extra
free energy is a linear function of concentration within a finite
range, at least within the limits of present-day experimental
accuracy. For example in the case of sulfur or oxygen or nitrogen
in liquid iron, the slope Cl Zn yi/Cl xj= E{ is constant up to 10
percent or more of nickel or manganese or chromium; for other
solutes such as silicon or carbon, deviations occur at lower
concentrations .
* Emeritus Prof ssor of M tallurgy

"'* G,raduat Stud nL

JOHN CHIPMAN and DONALD A.CORRIGAN


It is convenient to define a corresponding enthalpy term, the
extra partial molal enthalpy of component 2 due to the presence of
component 3. Within a limited concentration range this quantity is
also proportional to the concentration. For this range we write
~2(3)

= (oHx3/Clx
2

and Hx 3 = ~( 3 )x
2

(4)

Similarly for the extra partial molal entropy,


a-(3)
2

= (osx23;Clx

x -. 1

and sx 3 = a-( 3 )x
2
2 3

(5)

Variations in F;3 and H;3 covering a wider concentration range


extending as far as pure component 3 have been discussed by Alcock
and Richardson. ( 2 >
The 'extra' quantities are related as follows:.
(6)

The effect of temperature on the interaction coefficient E(2 3 ) is


determined when H; 3 is known, thus
dE ( 3 )

Rx 3

2
---

(7)

d( 1/T)

It is evident that if the heat of solution of component 2 in the


alloy is the same as in the solvent metal, E~3) is independent of
temperature. On the other hand if the entropy change is the same in
the two cases so that sX2 ,3 = O then by Eq 6 ~(3)
= RTE(3)
and
12
2
further from Eq.7 it follows that
.(8)

'

The condition s;-3 = 0, which leads to the direct proportionality


between E;3) and 1/T shown in Eq. 8, might be regarded as expressing
a second-order regularity in the solution in which the electronic
and spatial configurations of the atoms of solute i are the same
with respect to the atoms of solute j as to atoms of the solvent.
Despi t e the endorse~e nt of this i dea-by Al cock and Richardson, ( 2 )
it R m unli k Jy t hnt An h a Rit uation an xi1:1t xc pt wh n t h

THERMODYNAMIC INTERACTIONS IN DILUTE ALLOYS


of solute j are very similar to those of the solvent, i.e.
E) are very small.

I 11 111 H

e:{ and

111 11

I 11 s ome cases it is convenient to use weight percent and common


11111t1it hms. For this purpose we employ the activity coefficient
3)
a / [% 2], and the interaction coefficient
(o log
ii r% 3]) [% l] -+
corrent
relation
between
the two
The
.
100
I 11 1' r1L ti on coefficients derived by Schenck, Frohberg and Steinmetz< 3 1

E; =

(:I) =

230~ e(3) +Ml - M3


Ml

(9)

Ml

wit I ch the M's are the atomic weights. Corresponding terms for
111 11 1.tn,y and entropy are hp) = oH; 3;0[33] and s;3) = os;3;o[% 3].
11t1 I 1 n lat ions to the corresponding mole-fraction terms are:
111

100~ h( 3)
M

,I :I

100 ::

( 10)

'~3)

[ M,

~' M,

(11)

El

Wh I I n the literature contains many ~eterminations of


at fixed
111 111 1.111 s , data from which
and oi_ can be evaluated are very
111 ti .
The effects of alloy content and of temperature have been
I I 111. d experimentally only in a few series of alloy systems.
1111 purpose of this paper to review the limited data available
I It ' ' v I ow to disclosing any general or empirical relations which
1 I I. 1unong the quantities defined above.
Since the review is
dilute solutions the considerable amount of data on the
f carbon in iron alloys is omitted from consideration.

7rl

I 11111111N JN LlQ.UID IRON ALLOYS


11 111111 111

I
I

11r
"'

i n iron and in all of the alloys to be considered here

l.o ob y Sievert's Law.


Hence the activity of Nin an
I I x d omposition is proportional to its concentration,

0.

Ii 1n l111 1

Th o ncti vi ty of N in the binary Fe-N alloy at any


(! fi n d as
QUal to its wei ght percent. Its activity
111 1L 1 lautd a. ll oy iA
IH

.,
"'

JOHN CHIPMAN and DONALD A.CORRIGAN


fN = ~ %N = [% N in Fe]T

/ [% N in alloy]T

(12)

the comparison being made at equal values of T and ~ . It is


convenient to express log fN as a function of composition thus :
log f N = e~ [% j] + e~ [% k] +

(13)

this equation being restricted to compositions in which the e' s h.ave


values characteristic of the infinitely dilute solution.
The standard free energy of solution of nitrogen in iron in the
hypothetical 1 pct. solution according to Pehlke and Elliott(~) is

tN

= !':l_

.6.F = 86 0

'+ 5.

(14)

71 T

This is also the standard free energy of solution in any liquid iron
alloy .
In a given iron alloy let .6.Fa represent the free energy of
solution of half a mole of nitrogen in the hypothetical 1 pct .
solution of N in the alloy. This partakes of the nature of a
'standard' free energy but is a function of alloy composition.
write

} N2

~ (in alloy) ; c.Fa

[% N] a
- RT Zn - - -

We

(15)

N2

Correspondingly clla and L'Sa ref er to 1 atm pressure of N2 and


1 pct N in the given alloy. The former is calculable from the
temperature coefficient of solubility by the equation:
...

d(c.Fa/T)
d(l/ T)

all

clla

,J

= -

R( 8 Zn [% N]a \
o (1/ T)

)p

(16)
N2

n rgy of nitrogen due to the presence of t he


The extra fre
j 11;
yj n

'I

THERMODYNAMIC INTERACTIONS IN DILUTE ALLOYS

1 1 ~j

RT Zn [3 N] a/ [3 N in Fe]
+ RT Zn fN [in alloy]

11111

(17)

Th e effect of temperature on the activity coefficient is given by


quation

d(l/T )

d Zn fN
d(l/T)

R---

4. 575

d ej

[% j] __
N
d(l/T)

( 18)

Vnl ues of the heat of solution of nitrogen in iron alloys Lilia are
lv111 by Pehlke and Elliott. (lll Their Fig.22 gives the heat effect
111 1 IL atom of N rather than per mole of N2 as stated. It was
1 d rnwn from the original data in Pehlke' s thesis< 5 1 and is shown
I 11 11' I H. 1. Here the best average straight line is shown in preference
I 11 11 rtain curved lines in the original.
Results of Humbert and
1,l I lnl.t <6 1 for Fe-Cr-N are in good agreement. Data of El Tayeb and
111 1In< ( 7 1 for Fe-V-N and of Nelson ( 8 1 for Fe-C-N are added.
1

V1 lu es of h~. along with estimated limits of error, read from


1111 I I ues of Fig .1 are shc7tvn in Table I (see page 35 ). Also tabulated ar
vu I rn H of e ~ taken principally from the work of Pehlke and Elliott and
I 111 I 1Hl l ng earlier values quoted in their paper.
Values of e:~ and

'~ 111mDnted from the foregoing by means of Equations 9 and 10 ar


l.nhulated.

11 I 11

11' 11-1 . 2 contains a comparison of the values of ~ with E~ at o.boul.


Ir 1111 11 ln all the liquid Fe-N-j systems for which data are availubl1 ,
11111 1 l.r1king feature of Fig.2 is the approximately linear r Ja1,fo11
111 I w1, 11 ~ and ~ This is not at all unexpected if we cons 1ch r
I l11d I.ho chemical interactions which lead to increased soluh'll L.Y
11 I 11 111on1ote a larger negative value of the heat of solution.
Thi
111 I 1Lv( rage line of Fig. 2, which represents mo st of th data
11 11111 11mtts of experi mental uncertai nty, passes throu gh t h :r.1 10
p1il 11 1, nnd j s adequately represented by the equation

1~ {l:ul) 7.0 ~

( Ill )

111 willnh , .N
1 IH th< v11l1Hl at Lit

,/{

J OHN CHIPMAN and DONALD A.CORRIGAN

+4

Co

+2
~Ho

Ni

'\7.

'\1-

-2
-4

-0

-6

<l

-8
+ C- Nelson

-10

x V- El Toyeb and Parlee


-Mn-Beer

-12

A 11 Others Peh Ike and


El I Iott

-14

.. 2

10

12

Alloying Element, Wt pct

l'' b(. 1 .

II n.t, nf so1ut1

of' Nlt. rngo n 1n r.1 qu 1d Tron Alloys


'111

14

THERMODYNAMIC INTERACTIONS IN DILUTE ALLOYS

-50
0

-100

~
~

i::-

-150

-200

1' 11( ,

2.

250

-25

-20

-15

-5

+5

Comparison of Enthalpy and Interaction Coefficients;


Nitrogen in Liquid Iron Alloys at 1600

'1'111 1 c I ose correspondence between extra enthalpy and extra :t' r oo


11111 f .v 11 liown in Eq. 19 suggests some kind of regularity with r HP( nt.
111 1111 t. rnpy, and provides a test for Eq. 8.
The data of El Tay h o.11d
1
1 11111 1( > on the Fe-V-N system yield a linear relation bet w 11 1 ~
111d I l111L not a direct proportionality. Moreover, a valu or
~I I
O fo1:Lds to the relation:

11 ~

It'!'

~ ( for ~j

"' 1111111ox lm(l.Loly ?')~ / ~

(',() )

0) ...

3: 73 kc al inst ad of th

II h1 1 I from 1'11<. 2 n 11d EQ. 19.

xp ri mouto. I

v1

l11 n

JOHN CHIPMAN and DONALD A.CORRIGAN


It is now clear that neither the heat of solution of nitrogen
nor the accompanying en~ropy change is independent of the alloy
composition. The data of Fig.2 and the average result expressed in .
Equation 19 lead to the result:

~ =

1. 75

E~

(21)

The activity coefficient of nitrogen, fN= ~/% N at any temperature in any of the iron solutions studied, within the concentration
range in which it is a linear function
of x.,
may now be expressed
.
J
in terms of a single parameter E~ (at about 1600C) and the
concentration of the alloying element.

(lnfN)T=(E~)T

xj=

[3:20_0.ss]

(Ei)1873xj

(22)

Application of Eq.22 to the prediction of nitrogen solubility in


molten iron alloys is discussed elsewhere. <23 >
NITROGEN IN SOLID IRON ALLOYS
The solubility of nitrogen in pure gamma iron is known from the
work of Darken, Smith and Filer, <lZ> of Corney and Turkdogan, <13 >
and of Fountain and Chipman. <111 > These results agree well with
earlier data of Sieverts, Zapf and Moritz< 15 > and are used here in
preference to the more recent values reported by Wriedt and
Gonzales. <16 > It seems possible that the results of the latter were
affected by the 1 pct H2 in their gas stream. For this reason also
the data on Fe-Ni alloys mentioned below are preferred.
The effects of several alloying elements on the solubility of
nitrogen in gamma iron have been determined by Schenck, Frohberg and
Reinders. <17 > The lines which they have drawn depicting solubility
as a function of temperature are not those which would be obtained
' and the deduced heat effects are. not
by the method of ~least squares
entirely convincing. These are shown however in comparison with
values fore~ in Fig.3 for Co, Ni, Mn, C and Si. The two latter are
based on an insufficient range of concentration. For this reason
also the data on Mo and P are omitted. Similarly the Fe-V alloys
used by Fountain and Chipman <111 > did not exce d 0. 25 pct V x ept at
on
l'or

t mp ratur .

r~ h l'tlln I 11111 W<

Th
'111w o

p i 11t Hhown is t h rofor


llll(ld I.II;

l ' OHll I LH ()

:10

1'11

hiRhly nn c rtain.
IUld Tm11it.nw1 11 .

rkdoi.11ui

(la)

THERMODYNAMIC INTERACTIONS IN DILUTE ALLOYS

results, summarized in Fig. 3, show only the same general trend


was found in Fig.2 with the same notable exception of carbon.
111111 Lting carbon and vanadium from consideration the line drawn has a
lop h~ / e~ = 20 which, since the alloying. elements have nearly the
11 l 11inic weight of iron, corresponds to ~/E~ = 8. 7.
11i1111e

11 11

Ill.FUR IN MOLTEN COPPER ALLOYS

1111

Alcock and Richardson 12 > equilibrated copper al~oys with H2 S-H 2


xi.ores at 1300 - 150o0 c and reported values of E~ derived from

e~ ( YFel1200<:

-o.~2~~--~o~.1~~__.o~~--07.1~~-+-"""o.2

oC

-2

cu

-4

~
~Z

-6

-8

-10

-0.2

v ( ?)
- 0 .1

eJ

tO . I

+0.2

l'IK.:J.

J:11 thulpy un<i tnL< rucUon oufJ'i io nLH; Nil.rov,1 n


l 11 GI mmn Tr 0 11 LLL l ~0 0C

JOHN CHIPMAN and DONALD A.CORRIGAN

-w

Fig.4.

Enthalpy and Interaction Coefficients; Sulfur in


Molten Copper at 1400

measurements of D, the ratio of sulfur in copper to sulfur in the


alloy for the same gas at the same temperature. Their plotted
values permit approximate evaluation of ~ for concentrations up to
xj = 0.10 for the two alloy series shown in Fig.4. The ratio

771/Ej
s s is 11 kcal instead of 4.1 as suggested by Eq. (8).
HYDROGEN IN LIQUID IRON ALLOYS
The data on heat of solution of hydrogen in liquid iFon alloys
are neither extensive nor convincing. The work of Liang, Bever and
Floe< 19 > indicates essentially constant heat of solution in Fe-Si
alloys up to 21.7 pct Si. Weinstein and Elliott( 20 > determined the
interaction coefficients and heats of solution for a number of
liquid iron alloys at 1592C. Fig.5 contains the results of those
experiments which gave concordant results for h~. The relation is
of the same character as that observed in other groups of systems,
with the notable exception of Fe-Si-H.
HYDROGEN IN ALUMINIUM AIJLOYS
Opie and Grant< 21 > made accurate measurements of the solubility
of hydrogen in molten Al-Cu and Al-Si alloys at 700 - 1000c and
calculated the peat of soi'ution in each alloy. Precise evaluation
of r{iu and ~i is not possible because of uncertainty in the base
line for pure Al, especially the points at 700 and 800.
Accepting the 700 point as correct, and redrawing the line for pure
Al, we find very roughly, ~u = + 70 kcal and ~i = + 18 kcal.

THERMODYNAMIC INTERACTIONS IN DILUTE ALLOYG

0.2

c(..)
::i:::
~

0.1

:r

..c.

0.1

fi'ig. 5.

1'1 11 1 nll P

11v

1 11KC

Enthalpy and Interaction Coefficients; Hydrogen


in Liquid Iron Alloys at 1592C

ndi ng values* at 9oo0 c are E~u = + 5. 6, E~i


ra.t io ~/E~ is about 10 kcal.

2.34.

The

111 1'llHN lON

fo regoing sections we have reviewed the data on five


of systems in which the activity coefficient of a non-metallic
11 l11t.11 LA marltedly affected by addition of metallic solutes. It has
111 1111 11hown that in each group an approximate relationship exists
111 I Wt t ill Lil
interaction coefficient and the extra enthalpy of the
11 I1!1.11. HI) tfically the rat io 7?{ /E{ is nearly constant for several
I t 11111 wt t. hin on. h group, notably for systems in which both 7?{ and
111

f.h

1 1 111 11111

Wt

11111 1111 [

hl11 t.o

to 11fl f'lll

Lhn v11l11t H 1IHLtI by noaly nnd p hllt .1 2 ?

JOHN CHIPMAN and DONALD A.CORRIGAN


E{ have negative values. The exceptions are found among systems in
which these quantities are positive or zero.
Average values of 'T?f/E{ are summarized in Table II. The similarity
among the several groups, despite their markedly different physical
nature, is rather striking. The ratio of extra free energy to extra
enthalpy, RTE{/'T?f and the ratio Ta{/77{ are also tabulated. In no
case is the entropy term negligible, in fact in most groups it
amounts to between half and two-thirds of the enthalpy. Thus the
portion of the enthalpy available as extra free energy is only from
one-third to one-half.
The marked increase in the partial molal entropy of the nonmetallic solute upon addition of a metallic element which strongly
diminishes its activity coefficient is interpreted as the effect of
clustering of metallic atoms around those of the non-metal. This
would appear to be the controlling influence in systems of large
negative E{ and a{. In systems of large positive E{ and u{, such as
Fe-H-Al, the clustering involves the two metallic species to the
exclusion of the non-metal. When E{ is small u{ ifi also small and
may be of opposite sign; this may result from two opposing
tendencies or from other kinds of forces not here considered.
Finally it is observed that the question of the effect of
temperature on the interaction coefficient is one which deserves
more study than it has had in the past. All too many curves have
been drawn in accordance with preconceived ideas or unwarranted
assumptions. More dependable data on extra enthalpy and entropy
would be useful for their practical metallurgical application; they
would also contribute ~o a better understanding of the structure of
metallic solutions.

ACKNOWLEDGEMENT

The authors wish to express their thanks to Claude Lupis for


helpful discussion, especially for the derivation of equations
9, 10 and 11.

THERMODYNAMIC INTERACTIONS IN DILUTE ALLOYS


TABLE I

Enthalpy and Interaction Coefficients;


Nitrogen in Liquid Iron at about 1600c
(Data of Pehlke and Elliott' ij 1 except as indicated)

hj
N
kcal

ej
N
P. and E.

ej
N
others

Ref.

1.0 0.2
0.62 0.25
0.13 0.01
0.10 0.04
-0.04 0.04
-0.09 ?
-0. 13 0.04
-0. 31 0.10
-0.40 ?
-0.60 0.09
-0.70 0.05
-1. 34 0. 10
-1. 44 0.10

+0.2 (?)
+0.047
+0. 011
+O . 010
-0.002
+0.006
+0.009
-0. 011
-0.02
-0.034
-0.045
-Q.067
-0.100

+O. 13
+0.065
+O. 007
+0. 010

8
9
9
9

'l'hird
11: 1 ment

c
Si
0

Ni

w
'n
Cn

Mo
Mn
To,

Cr

Cb

+0. 002
-0 . 013
-0.023

9
11
10

-0.045

-0.094

~
22
31
11

10

Ej
N

4
12
1
4
13

-13
-19 ?
-15 5
-53 17
-40
-195 29
-68 5
-220 16
-132 9

7.0
5.9
2.7
2. 5
-3.8
+1. 8
+1

-5.0
-5.2
-27.2
-9.6
-26. 3
-20.9

TABLE II

Summary

<I rnnp

11 I N
1'1 I N
I 11 I 11
I 1 I ll
I
II

l quid
HO lid

11 Qnid
l I qn 1 d
J I au1d

Temp.
oc

77if E{
kcal

Toil17i

RTE{ /77{

1600
1200
1400
1592
900

7.0
8.7
11. 0
10. 8
10?

. 47
. 66
. 69
. 66
. 76

. 53
. 34
. 31
. 34
. 24?

--

:I I

JOHN CHIPMAN and DONALD A.CORRIGAN


REFERENCES
1.

C.Wagner , Thermodynamics of Alloys 51 , Addison Wesley Press,


Cambridge, Mass. (1952).

2.

C.B.Alcock and F. D.Richardson, Acta .Met. 6, 385-395 (1958) .

3.

H.Schenck, M. G.Frohberg and E. Steinmetz, Arch iv


Eisenhuttenwesen 31, 671-676 (1960) .

4.

R.D.Pehlke and J.F.Elliott, Trans.Met.Soc.A.I.M.E. 218,


1088-1101 (1960).

5.

R.D . Pehlke, Sc.D.Thesis, Mass.Inst. of Technology, Cambridge,


Mass. 286 (1960).

6.

J.C.Humbert and J.F.Elliott, Trans.Met.Soc.A.I.M.E . 218,


1076-1088 (1960).

7.

N.M. El Tayeb and N.A.D.Parlee , Trans.Met .Soc.A.I.M.E. 227,


929-939 (1963).

8.

E.C.Nelson,

9.

H.Schenck, M.G.Frohberg and H.Graf, Arch.Eisenhuttenwesen,


29, 673( 1958); 30, 533(1959).

10.

S.Z.Beer, Trans.Met.Soc.A.I.M.E. 221, 2-8 (1961).

11.

V.C.Kashyap and N.Parlee, Trans.Met.Soc. A.I.M.E .


86-91 (1958) .

12.

L.S.Darken, R.P.Smith and E.W.Filer, Trans.A.I.M.E. 191,


1174-1179 (1951) .

13.

N.S.Corney and E.T.Turkdogan, ,J.Iron Steel Inst. 180, 344(1955).

14.

R.W.Fountain and J.Chipman, Trans.Met.Soc.A.I.M.E. 212,


737-748 (1958).

Trans.Met.So~.A.I.M.E.

...

227, 189-191 (1963).

'

:111

THERMODYNAMIC INTERACTIONS IN DILUTE ALLOYS

I 1.

A.Sieverts, G.Zapf and H.Moritz, Z.physik.Chem. 183A, 19-37

(1938).
f !J.

A.A.Wriedt and 0.0.Gonzales, Trans.Met.Soc.A.I.M.E. 221,

532-535 (1961).
17.

A.Schenck, M.G.Frohberg and F.Reinders, Stahl und Eisen

83, 93-99 (1963).


I ll.

E.T.Turkdogan and S.Ignatowicz, The Physical Chemistry of


Metallic Solutions and Intermetallic Compounds, 2, Paper
No.6C A.M.S.O.London, (1959).

l fl .

A.Liang, M.B.Bever and C.F.Floe, Trans.A.I.M.E. 167, 395-403

(1946).

n.

M. Weinstein and J .F.Elliott, Trans .Met .Soc.A.I.M.E. 227,

382 -393 (1963).


' l.

W.R.Opie and N.J.Grant, Trans .A.I.M.E. 188, 1237-1241 (1950).

:~!.

J.M. Dealy and R.M.Pehlke, Trans.Met.Soc.A.I .M.E. 227, 88-94

(1963 ).

' :1.

J. Chipman and D. A. Corrigan, Paper submitted to Trans .Met .Soc.

A.LM.E .

:I'/

DILUTE SOLUTIONS IN METALS AND ALLOYS WITH


PARTICULAR REFERENCE TO OXYGEN AND SULFUR
by

C.B.Alcock
Professor of Metallurgical Chemistry, Imperial College, London
i\l11-1tract
Although there exists a considerable amount of experimental in/11r mation concerning the thermodynamics of dilute metallic solutions,
11 sa tisfactory theoretical description is still lacking, even for
l1inary systems. For the more complex solutions of the metallurgist,
i.mp le models can be applied to make semiquantitative calculations
using experimental data from binary systems. The extent to which
rnr:ces s can be achieved in this way is reviewed, and a comparison is
mad between calculated and measured activities in ternary solutions.
In the special case of dilute solutions of sulfur and oxygen in
I i'tuid metals, which are of particular interest to the process
111rtnllurgist 1 experimental techniques are still required for the
tndy of systems in which there are strong interactions. Progress
111. this direction is discussed especially with reference to
r le troc hemical 'and mass spectrometric techniques.

The st udy of dilute solutions is a particularly attractive branch


Firstly, data are
11111v lded for theoretical consideration under the most favorable
1. I 1 um tance.
Only the interaction between the solute and solvent
1d.1111rn needs to be considered, reducing the number of parameters to
11 111l!l!mum.
Secondly, and from a practical standpoint, data are
11111.nl n d for the assessment of refining processes, which are an
l111p11 rLant part of metal extraction, and for the chemical behaviour
11f l111purities under conditions of service, a growing branch of
1111 vr 1 al metallurgy.
111 met allurgical thermodynamics for two reasons.
1

111i.Alcally we would hope that the data of the experimentalist


1111Iti b i nt erpreted in terms of the wave mechanics of free electron
11111n r,y, but progress in this field is understandably slow. It is
111111u~h1y t ru e to say that the thermodynamics of solutions containing
1111 1111 l.1~1 u e l ments of major interest as structural components can
111 mo r( ro.pi dly m as ured and tabulated, with the modern experimental

C.B.ALCOCK
techniques at our command, than can the problem be solved
theoretically, given an equal involvement of man-years.
Descending from the fundamental theoretical analysis, there exist
however, a number of semi-empirical treatments which enable us to
make reasonably quantitative predictions of the behaviour of, for
example, ternary systems, given data for the composite binary
systems. One procedure is entirely algebraic, and another draws on
the concept of constant pair-wise chemical bonding.
THE ALGEBRAIC ANALYSIS
It is well known that for any binary system at constant temperature, . or for one having an ideal entropy of mixing over a range of
temperatures, the excess partial molar free energy of a component
can be expressed by a power series in the mole fraction thus,
L-Gxs
B

RTX A2

a+bX8 + cx8 2 etc .... where x8 is the mole fraction of


component B etc.

When b = c = O the solution is strictly regular, and when b is finite


but c = 0, the solution has been termed 'sub-regular' by Hardy' 11
He writes

The two chemical potentials are now


.c:.GA

RTZn XA + X~(2A 1 -A2 ) . + X~(2A 2 -2A 1 )

.c:.GB

RT Zn XB + X~(2A2-A1) + x1(2A1-A2)

Following Darken< 2 > the exc~~s integral free energy of the


pseudo-binary section of the A+B+C ternary system, with B and a
constant molar ratio of A/C as the pseudo-binary components, is
given by

tJO

DILUTE SOLUTIONS IN METALS AND ALLOYS

axs
/\ B+A/C

(1-XB)

j" [~"
(1~XB)2 d~

X/Xe

-XA

(1~XB)2
J [~"

dx8 ]

-Xe

[-"

GB
(l-X8 )

dX8

=0

XA

=0

1111d

Xe

(i)

Oarken showed that i f these pseudo-binary solutions were strictly


ar for all ratios XA/ Xe, then we obtain the equation applying
Lho dilute solutions of B in A + C as a function of X/Xe

1 11i.1 11 l

I 11

CI . 11. X - 0)
8

I 11 y

wl11 110

I\

1 11

(ii)

A+e

RT

'Y ~ is the activity coerficient of the dilute solution of


A tc.

Nnw writing equation (i) in terms of the series expansion for


po.rtial molar free energy, and integrating between the limits
1 and X8 = 0 and at constant XA/Xc we find that

1111 )1

11

N .~+A/C

- (a l

+ bl + cl

' ' ' '

XA/Y. .. + X
~

+ Xe ( a3 + b3 +
2

ii I

(a

A 2

+ b2 + c2
--;

2 ... . )X
3

=O

-S ) Xe=O

C.B.ALCOCK
where subscripts for the a, b, etc. are 1 when the A+C mixture is
the solvent, 2 when pure A is the solvent, 3 when pure C is the
solvent for the dilute solute B.
Now if we use the fact that

b~

cl
3

XA{b~ + c~ .. }~=O

-} XA/XB

~r~ + c~

(iii)

... }xA=O

This shows that the dilute constituent B dissolved in A+C will


follow equation (ii) in respect of the dependence of the activity
coefficient of B on the A/ C ratio, not only for the strictly
regular case, but also when the deviations from regularity in the
concentrated ternary solutions are the mole fraction weighted
averages of the deviations in the concentrated binary solutions
(Equation (iii)).
It also follows that in the concentrated ternary solution for a
constant value of X8
_B
XA
_B
~ _C
2
XS
f'IGA+B+C = XA+Xc L'lGA+B XA+Xc f'IGC+B - (1 - XB) f'IGA+c

. .
wh ere th e excess f ree energy o f mixing

Gxs
A+C

(iv)

is that for the pure

binary A+C system with mole fractions XA = XA


, .I

~/XA + ~

+XA and

,..

The empirical fact that equation (ii) is obeyed in a number of


ternary metallic systems where deviations from regularity are known
to occur in the binary component systems suggests that equation (iv)
might be useful in concentrated ternary alloys . The author has
tested this equation and fi nd s quite good agreement with experiment

....
A IJ

DILUTE SOLUTIONS IN METALS AND ALLOYS


In, for example, the cadmium-lead-antimony system, using the results
nf Ell iott and Chipman ( 3 a~ 3 b l.
_xs

Fig . la shows the a function of cadmium, ~GCd/(l - xcd) 2 in these


ull oys in the binary Cd-Pb, Cd-Sb and two ternary solutions with
Xrb/X 8b = 2 and 0.5. It can be seen that the deviations from the
11 Lri ctly regular behaviour, when the a function should be independent
of composition, are quite marked in each case. Fig. lb shows,
IHJwever, that the partial molar free energies of cadmium conform
quite closely to the values calculated from equation (iv) . This is
IJ cause the measured values only deviate from linearity, for constant
XGd' in going from pure lead to pure antimony as solvent, by an
11mount approximately equal to the excess free energy of mixing of
Ioo.d-antimony at the appropriate XPb/XSb"
'1'111\ CHEMICAL BOND ANALYSIS

Equation (ii) can be obtained by an analysis of the binding


' n rgy of a dilute solute to its immediate environment using the
nnn s tant pairwise bonding model and assuming a random distribution ( 11 l.
/\ J1nmber of ternary solutions, especially those containing non1111 1.allic dilute solutes such as oxygen and sulfur, only conform
111111 1-quantitatively to this regular approximation. The next step
111 lear ly then , to apply the quasi-chemical approximation. This
1111 11 L11::i t hat the number of A and C atoms immediately surrounding the
1 o lute B atom, nA and n
8 respectively, is given by
nA

XA

nB

XB

.0.E
e

RT instead of

nB

= A as in the strictly regular


XB

11 l11 Lion.
lh r

.0. E is the exchange energy per mole for the process


A-B + C

C-B + A

11 11l ng this e quation as a starting point, Alcock and Richardson (s l


l111w1 rt t hat

C.B.ALCOCK

+3.000

Cd - Pb

:Pb-Sb=2:1

0
Pb

-2.000

Sb

= O. 500

Ob - Sb

-3.000

'
1.. 0

o.s

Fig. la.

o.q.

0.6

o. 2.

Cd

The a Function for Solutions of Cadmium in


Pb, Sb and Pb+Sb Alloys at 500C

0.0

-DILUfE SOLUTIONS IN METALS AND ALLOYS

+2.0
+1. 75
+1. 5
+1.25
+1.0
+0, 75

-xs
b.GCd

+0.50
K Cal
+0.25
0

-0.25
-0.50

-o. 75
-1.0
-1.25
-1.50

0 Xcd = 0
El Xcd 0.25

-1. 75

Icd

-2.0

Xcd

= o. 50
= 0.75

Pb - Sb

-2 . 25
-2.50
Binary System

-2. 75
-3.0
0

-- -

--e-- ---e- ..._


0,33

0.66

1.0

!Pb + 1 sb

I'' I . lb .

Excess Partial Molar Free Energies of Solution of Cadmium


i n Pb+Sb Alloys at 500c as a Function of Cadmium Mole
Fraction
tJ fj

C.B.ALCOCK

1/z
+

XA (yA~C)
B
~

1/z
.. (V)

Ye

where z is the co-ordination number, assumed constant, of the B


solute atom in the solution. This equation, although an improvement
on equation (ii) was found only to agree with the experimental
variation of the activity coefficient of oxygen and sulfur in
metallic alloys when a value of z somewhere between O and 2 was used.
These chemical bond equations take no account of the fact that in
making alloys of metals, the quality of the bonds, which may be
measured by the valence electron-to-atom ratio, would then be a
function of composition. It might therefore be suggested that in
those systems showing significant departures from equations (ii)
and (v) above, the principal source of error is in the neglect of
this factor.
In an attempt to consider this possibility Cheng and the author
made measurements of the variation of the activity coefficient of
sulfur in lead and tin and silver+ copper alloys( 61 Here, at
least formally, the electron/atom ratio of the solvent was always
constant across the range of solvent composition. It was found,
again, that equation (v) could only be applied to give quantitative
agreement with z between 0 and 2. For the larger values, z = 8-10,
which might seem more appropriate to metallic solutions the equation
offers only a small improvement over the regular solution. However,
recent precise measurements by Leach and co-workers( 7 l on the system
Au-Sn-Bi indicate that equation (v) may be valuable for such
metallic systems when
is very much larger thany'3c

yBA

One other possibility which must be borne in mind is that the


constancy of the pairwise bond e~~rgy may be the fallacious
assumption in this treltfment. Thus we could retain 'sensible'
values of z such as 8-10 under all circumstances if we would allow
the values of the bond energies to be a function of the binary
solvent composition. This is tantamount to what is done in the
binary systems when a sub-regular model is used in place of the
strictly regular model. Riehardson and the author ( 11 > discussed the

46

DILUTE SOLUTIONS IN METALS AND ALLOYS


Iron + nickel + oxygen system in this respect. The exchange process
f'or the breaking and making of one metal-OXYgen bond is represented
Lhus
Fe-0 + Ni(Fe+Ni)- Ni-0 + Fe(Fe+Ni)
nnd according to the constant pairwise bonding approximation the
txchange energy for the process is given by

AE

= &:lo(Ni)

- &:!Ni(Fe+Ni)

- &:lo(Fe) + &:!Fe(Fe+Ni)

z
wh re Z is the nearest neighbour co-ordination number throughout the
111nge of compositions. Approximating the excess free energy of each
lllnary system to the heat of mixing when we then have

1
YNi
0 YFe+Ni
Fe
- . log
Z
0
Ni
[
Ype Ype+Ni

Nnw i t would seem reasonable to use y~i for the binary OXYgen +
11 I c ke l system, in this expression for the exchange energy, at the
11! kel- rich end of the ternary system, and similarly the binary
rt)( ffic ient for J{!e at the iron-rich end of the system. Any
tit purt ure of the measured exchange energy in the ternary system from
Ch1LL calculated using binary systems data may be reasonably ascribed
I 11 Lhe activity coefficient involving the nickel at the iron-rich
11 11!1 and vice versa.
noth of the possibilities suggested above for accommodating the
data within the framework of chemical bond theory have
,,fiv Ious weaknesses. They show the fact that the theory in essence
l~1 i rLs a product of bond energies and co-ordination number which
. 111111ot be separated into its components from consideration of
I ht nnodynamic data alone.
1

~ porimental

Another aspect which should be considered however is the overall


rn of b haviour wh ich has been obtained so far. The
1 111 1: 1 l 1~L1011 botwo n t h
partial heats of so lution of OXYgen and
1111 LLi

'17

C.B. ALCOCK
sulfur in liquid metals and the heats of formation of the corresponding oxides and sulfides suggested by Richardson<e> has been
borne out by experiment. The following table demonstrates the
present state of knowledge for the oxygen solutions.
TABLE I
Solvent

-&Io cal/ g. atom oxygen


for the !owe.st oxide

-di

%0 2 cal.

Ag

7,300

2,700

Cu

40,000

9,900

Ni

57 , 500 (22 , 500)

Co

57 , 100 (31, 100)

11 900
14 700( 9 )

Fe

63,500 (43,200)

28,000

Pb

52,400

28,500

Sn

68 , 400

43,800

Note :

191

the figures in parenthes is after th~ heats of formation of


iron, cobalt and nickel mono xides are the values obtained
when the crystal field contribution has been subtracted .

It would clearly be valuable to widen our knowledge of the


usefulness of this correlation by further measurement.
EXPERIMENTAL TECHNIQUES
It thus appears that although considerable progress can be made
with complex alloys, the cornerstone, at the present time is our
knowledge of dilute binary systems , and the important ones of oxygen
and sulfur in metals are theoretically ~ifficult' . Until recently
the experimental study of dilute solutions of sulfur and oxygen in
' the use of H/H2s or H/ H20 ga s
the elements was made" largely with
equilibrium methods. The well-known limitations of this technique
have meant that interest was , perforce, concentrated on solutions
of moderate stability such as those in iron, coppe r , ni ckel et c. as
solvent.
New experiment al met hods offer the possibility that w can break

his 'stability barri r' and b gi: to gath r knowl QdK nhcmt mo_t__ __,
11

DILUTE SOLUTIONS IN METALS AND ALLOYS


1 table systems . The techniques which will be discussed here are
I.ho e of electromotive force measurement using solid oxide
11 I ctro lytes , and mass spectrometry.
The first procedure is now relatively well known and the
lit r ature on the application of lime-zirconia solid solutions has

rnwn rapi dly 1 However, the failure of these particular solid


o 1ut ions to conduct electricity only by anion migration at oxygen
11 rnssures below about 10- 20 atmos. at 1000c means that they offer
vn ry little advantage over the gas equilibrium to the metallurgists.
~

The solid solutions of lime or yttria in thoria are better in


Lid A respect since it has been shown that they may be used success!'11 l l y as electrolytes for cells where the electrodes have oxygen
111u"tial pressures as low as 10- 35 atmos. at 1000c. Thus the free
11(rgy of formation of NbO and MnO which are quite stable oxides
111 vo been measured successfully in the author' s laboratory, using
I ho r ia-based electrolytes. Apart from stability another interesting
111 t l)t ct of this technique is that oxygen can be transferred to or
11w11,y f rom a metal simply by passing a current through the cell.
ri ln
the electrolyte conducts electricity only by anion migration,
I lt1;n t he passage of current represents a coulometric ti tr at ion of
11x.YR n between the cell electrodes .
Currents up to 500 amps have been used in the experiments
d out so far, at temperatures around 600c , and higher currents
111 11 Id of course be employed as the temperature at which the titration
I 1 11111de is increased . In fact the use of lime-zirconia as an
n I m:l. ro lyte material for a high temperature fuel cell operating at
I 11111 rH rature s around 1200C is a very practical consideration.
l'lii'f'i

Matis fac tory measurements have been obtained of the oxygen


'" 11hil ity and solution thermodynamics in liquid lead with the use
11 1_ ! Im -zirconia and thoria-yttria electrolyte.3(lo) , and in liquid
I l 11 11 Ri ng only the thoria-yttria electrolyte ( 11 >, The limits of
111I11 technique with respect to oxygen potential and temperature have
wit. y t been r eached and measurements on most of the Group IA, IIB
1111 1 I UB e lement s s eem to be now well within our grasp.
Such an
I 111 1nu
of ,t he scope of our knowledge may prove valuable to the
11iiv1111
f t heory.

C. B. ALCOCK
With respect to mass spectrometry, it cannot be claimed at the
moment that the technique is superior to the older gas equilibrium
except in two respects. Firstly the spectrometer is most conveniently
used in conjunction with the Knudsen cell in a high vacuum system.
This means that higher temperatures of operation, and therefore
greater dissociation pressures for a given system can be contemplated
than in gas-solid or gas-liquid equilibrium where the porosity of the
reaction tube often sets a limit to the temperature range. Secondly,
it is possible to measure the vapour pressure of one specie in the
presence of a number of other, and even higher, partial pressures.
This is a distinct advantage over the gas equilibrium technique and
means that when the technique of leaking controlled partial pressures
of gases into Knudsen cells< 12 > is firmly established , we may look
to an extension of our knowledge concerning dilute solutions of
sulfur in a number of metals, e.g. Ga, In, Zn .
The lower and upper limits of pressure measurement using the
Knudsen technique in the mass spectrometer seem to be about 10- 11
and 10- 4 atmos. The lower limit could be pushed down by another
order by individual ion counting techniques at the spectrometer
detector, but the upper limit is the practical limit of applicability
of the Knudsen condition of collision-free effusion. Higher partial
pressures might be handled by the use of heated capillary inlets to
the ion source of the spectrometer similar to the cold capillary
inlets at present used for permanent gases. The author is not aware
of any reports of such a technique being used at present, although
the measurement of partial pressure by capillary flow techniques is
by no means nove1< 13 >.
CONCLUSION

It can be said in summary that the present state of knowledge is


clearly not sufficient in the dilute solution field to enable us _to
dispose with further m~asurement . and rely on theoretical calculation:
The classical experimental methods have, in many instances, yielded
all the experimental data of which they are capable, but fortunately
new methods are emerging, and will emerge , to enable new and valuabl e
results to be obtained in hitherto unexplored areas.

DILUTE SOLUTIONS IN METALS AND ALLOYS

REFERENCES
I.

H.K. Hardy, Acta.Met., 1, 202 (1953).

2.

L.S.Darken, J.Amer, Chem.Soc., 72. 2909 (1950).

:10.. J.F.Elliott and J.Chipman, Trans.Far.Soc., 47, 138 (1951).


:lb.
ibid
J.Amer.Chem.Soc . , 73, 2682, (1951).
~.

C.B.Alcock and E.D.Richardson, Acta.Met., 6, 385, (1958).

Acta.Met., 8, 882, (1960).

ibid

I> .

o.

C.B.Alcock and L.L.Cheng.

'I.

J.S . Ll.Leach.

II .

F.D.Richardson, J.Iron Steel Inst.(London), 166, 144 (1950).

II.

G. R. Belton.

10 .
II .

Unpublished work.

Private communication.

Private communication.

C.B.Alcock and T.N.Belford, Trans.Far.Soc., 60, 822 (1964).


ibid

Trans.Far.Soc., 61, 443 (1965).

I :i.

J. Berkowitz, Thermodynamics of Nuclear Materials, p. 505,


I .A.E.A., Vienna (1962).

1:1.

P. Gross, C.S.Campbell, R.J.C.Kent and D.L.Levi, Disc.Far.Soc.,


4. 206 (1948).

THE NATIONAL STANDARD REFERENCE DATA SYSTEM*


by

Everett R.Johnson
Off ice of Standard Reference Data
National Bureau of Standards
In order to ensure that scientists and engineers might have
access to evaluated data, the Federal Council for Science and
'I' ohnology recommended to the Director of the President's Office of
t<' l1nce and Technology that a government-wide, co-ordinated, compre11 11 lve effort be established in the field of physical science.
l'tH pti ng this recommendation , Dr. Jerome Wiesner promulgated a
Ir 111 ral policy in June 1963, creating a National Standard Reference
llnl.1t System and assigning to the National Bureau of Standards the
11 111onsibility for its administration. The general objective of this
v t, m is to co-ordinate and integrate existing data compilation and
'v11luation activities into a systematic program, supplementing and
nm1di ng technical coverage when necessary, establishing and main1nI11 l ng standards for the output of the various groups, and providing
1111w h1tnisms for the dissemination of the output as required. The full
f 11xf. f the federal policy statement is given as Appendix A o this
1 p11rt and the press release announcing the establishment of the
NI ltll ' is given as Appendix B.

"l' LI mum

'l'h

e official actions constituted an endorsement of a proposed


of approach which had been developed over a period of several
1111111 l.hH by an ad hoc committee of staff members of the National Bureau
1rl 1 L1U1dards appointed for the purpose by the Director of NBS. The
11111 rnl organ ization of the program at the National Bureau of
I 1111durds to be described later is very similar to that developed by
Iii p lanning committee. This plan envisions a decentralized opera1 I 1111
nducted across the country in many organizations by many
I 11111 v I duals and groups, with central co-ordination managed by the
Nn l lnnal Bureau of Standards. Account is to be taken of all existing
111 flvlties , and no attempt is to be made to take over direct
11111rr1ttional management of existing groups, nor are new activities to
ltrr In Lti ated in fields now adequately covered. The needs of the
Ii h11fcal community are taken into consideration by calling upon
11111 11

1'11l 11 material is an excerpt from 'National Standard Reference Data


1 v11 I, in - Plan of Operation' by E. L . Brady and M. B. Wallenstein,
II ll, l) P!!rtm nt of Comm rce , NSRDS-NBS #1.

EVERETT R.JOHNSON
groups of specialists for advice and for active assistance in
planning the program. The program is envisioned as permanent but
with ever-changing internal emphasis in the many fields of knowledge
as the requirements of the users of quantitative data change.
The assignment that was given to the National Bureau of Standards
by the Office of Science and Technology includes the following tasks:
1.

Operation of a National Standard Reference Data Center at the


Bureau of Standards.

2.

Co-ordination of standard reference data activities of NBS,


DOD, AEC, NASA, NSF, and other governmental agencies, all of
which may operate components of the National Standard
Reference Data System, if this is mutually decided to be
appropriate.

3.

Establishment of standards of quality for various products


of the NSRDS.

4.

Establishment of standards of methodology, including machine


processing.

5.

Establishment of standards for such other functions as are


required to ensure the compatability of all units of the
NSRDS.

This directive is very broad, indeed; obviously, some of the


items listed above will require considerable thought, consultation,
and technical progress before conclusions can be reached and action,
if determined to be appro~riate, can be taken. It is perhaps
desirable at this point to reassure the readers of this report that
the management of the National Bureau of Standards does not interpret
this directive as a license,_1to make arbitrary, unilateral decisions
and attempt to impose them on groups outside its own operational
jurisdiction. The manner in which the Director of the National
Bureau of Standards proposes to fulfil the responsibilities assigned
by the Office of Science and Technology will now be described in
some detai 1.

r '1

NATIONAL STANDARD REFERENCE DATA SYSTEM

SCOPE OF THE NSRDS


I n order to develop a comprehensive plan of operation, it is
ces sary to decide which fields of science and technology should be
l 11cl uded in the scope and, in those fields, which kinds of data are
1L1>propriate for collection and evaluation. These decisions have
hoen made in general terms, recognizing that in many specific
l 11s tances the question of appropriateness of inclusion of a particular
le lnd of data will not be easy to answer. First, it has been decided
f. hat the program will be concerned only with physical phenomena.
Ph enomena in which biological effects are a significant factor will
11ot be included. Second, it has been decided that the term 'Standard
It f erence Data' would be defined as follows : Standard Reference Da_ta
111 ans critically evaluated quantitative information relating to a
pro perty of a definable substance or system.
11

To obtain complete precision of meaning, most of the terms used


111 t he above definition would themselves require definition. However,
l. 11 purposes of the National Standard Reference Data Program do not
n qui re such precision. General, flexible guidelines on the scope
111' the program and appropriateness of specific activities are quite
1uff i cient. Decisions will be made taking into consideration all
11 l vant circumstances.
Operationally, these guidelines mean that the NSRDS shall not be
11111 erned with the compilation of data relating to systems of
1111c: rt ain, variable, or uncontrollable composition nor of data that
1111 s ensitive to unknowable details of the structure of the material.
'I'll I H principle carries with it the corollary that a system or
1111 I. ri al may be well-defined for one property but not for another.
111 11utting this principle into practice, the program of the NSRDS
111111 I. i nclude careful examinations of the available data in a variety
t1 r fie lds in order to determine whether the data are appropriate for
y11tematic compilation activities; that is, a critical review of the
l.iLL of quantitative knowledge is first required.
Application of the general guidelines also leads to the exclusion
t he NSRDS of data whose value depends upon both the system or
1 1~1Htance being measured and the measuring technique iteself - in
11.l. hor words, data which are not characteristic of intrinsic properties
t1 I . t, h
syst em or s ubs tan ce. It i s recognized that many such types of
f rom

EVERETT R.JOHNSON
data have important applications and that the development of a comprehensive program for their compilation is worthy of serious
consideration.
Additional types of data not considered appropriate for systematic
activities in the System are those that relate to proprietary
materials, those that have validity for only a limited time, and
those that relate to materials which are of interest only for highly
specialized commercial purposes.
TYPES OF ACTIVITIES APPROPRIATE FOR NSRDS
The achievement of the objectives of the NSRDS require both the
compilation of 'standard reference data' and the 'processing' of
data - that is, gathering, sifting, and evaluating raw data and storina
retrieving, and disseminating evaluated data. For the compilation of
standard reference data, four types of activities are considered to
be approrriate in the program of the NSRDS:
1.

Data collection and evaluation - scanning the literature to


locate relevant data and background information, exercisin~
critical judgment , and compiling the evaluated data in an
orderly format.

2.

Preparation of critical reviews - examination of the status


of quantitative knowledge in a special aspect of a technical
field, including a survey of experimental uncertainties and
limitations and theoretical background, and development of
new relationships between properties, as may be appropriate.

3.

Computation of useful functions - computation of functions


derived from standard reference data (e.g. thermodynamic
functions), of values of theoretical or semi-empirica~ functions used in the int~rpretation of quantitative exper!mepts,
or of estintlited- values of unmeasured properties of important
substances.

4.

Experimental measurements - determination of experimental


values of properties needed to fill gaps in tables , to extend
the range of a paramet er over which a property is meas ur ed,
or to obtai n a mar pr cis figur
(It would n t b
di

NATIONAL SfANDARD REFERENCE DATA SYSTEM


considered appropriate to support research to develop a new
or better measurement technique; experimental work would be
undertaken only if a technique of known uncertainties would
give useful information. ) _
I n the data-processing type of activity it is planned to maintain
hoth a design and research function and an operations function
1tHsociated with the central office. The task of the research group
IH to identify and formulate t he problems associated with the pro<"< si ng of the raw data and other information that constitute the
working material of the data compilation groups and with the
1nllection, storage, retrieval and dissemination of the standard
I'll fere nce data.
The research group also designs data handling
11.vHtems as required to solve these pr oblems, using manpower and
fn11 ities from the National Bureau of Standards or other groups, as
111mropriate. The task of the group given the operational responsi111 l lty is to provide the services to the technical community that
1110 dete rmined to be useful and to maintain the collection of data
Umt wi ll constitute the data center at the National Bureau of
ll t.1mdards-indexing, filing, storing and retrieving data as required .
l'IUIGRAM MANAGEMENT

In order to plan the activities to be undertaken by the NSRDP,


fie ld of physical science has been divided into a number of
111,t.ogor ies, each covering a field broad enough to include a large
1111111ber of standard reference data compilation activities while at
llli s ame time retaining logical cohesiveness. In each field one
pc nrnn of high general competence can be expected to have a good
wn l'king knowledge of the subject matter in the individual task areas
I 11 which standard reference data activities would be undertaken.
1'11 1 t he initial phases of the program the following technical areas
l111v been selected: (1) nuclear data, (2) atomic and molecu.lar data.
:I) s olid state data, (4) thermodynamics and transport data,
1) chemical kinetics, (6) colloid and surface properties, and
(' I) mechanical properties of materials.
Other categories will be
1u lcl d as the financial and manpower resources available to the
tll Lee of the NSRDS increase.
flt(

1'h
I

planning and i mplementation of a program in each of these


direct responsibility Of a member Of the Staff Of

IL< U.Ori 8 iS th

'T

EVERETT R.JOHNSON
the Office of the NSRDS at the National Bureau of Standards. This
responsibility involves the following steps, many of which are being
taken concurrently over an extended period of time. Each category
is to be further subdivided into areas of smaller scope, listing and
organizing in a logical fashion the subfields which in the opinion
of specialists comprise the broad category. In effect, this step
constitutes a definition of the field. No two scientists seem to
agree completely on the exact way in which this categorization should
be carried out nor even on the names of the broad subdivisions.
Certainly there is no uniquely ideal set of principles upon which to
base such a classification. For practical purposes the ideal of
unanimous agreement is not necessary; the goal is to obtain a system
which can help to identify the activities which should be undertaken,
to indicate the completeness of coverage of the various fields of
science, and to judge the development of program activities from
year to year.
Another step in the development of the program is a survey of the
operations of e!isting data compilation and evaluation groups. This
survey is to determine the technical scope covered, the nature of
the output. the number and types of staff required, the financial
arrangements, plans for the future, and any other relevant
information.
Still another essential feature of the program is a determination
of the needs of the technical community for activities of the type
appropriate to the NSRDS. A great deal of information has been
obtained during the survey of existing groups now under way.
Additional information is obtained by consultation with individual
specialists who use the data, with representatives of industrial and '
professional organizations Xe.g. the Manufacturing Chemists'
Association, The American Society for Testing and Materials, etc.),
with committees of the National Academy of Sciences-National Research
Council, with program office~~ in other government agencies, and with
any other sources -ehat seem appropriate.
As a result of the surveys and consultations a list of desirable
activities can be prepared, with a priority assigned to each. No
attempt will be made to develop a detailed scale of priorities;
perhaps three levels at -most will be quite sufficient. The area
program leaders at the Nationa~ Burea of Standards hav th

NATIONAL STANDARD REFERENCE DATA SYSTEM


,. s ponsibility for allocating the available funds to the various
li1di victuals and groups who are prepared to undertake act iv it ies,
1111Lking the financial arrangements required, and monitoring the
Lt chn ical output of the various groups . Some of these groups are
111 ( mbers of the technical divisions of the National Bureau of
111.o.ndards , but the majority of the standard reference data compilat. Ion activities will be conducted outside the National Bureau of
/ ll,1tnda rd s - in academic inst it ut ions, in other government labora1.ories, and in industrial laboratories. No attempt will be made to
r 1.l mulat e new activities at the National Bureau of Standards unless
I ho s ub j ect matter is one in which special competence already
xlsts . I n all cases direct operational supervision is the renons ib ility of the NBS division management or, if outside the NBS,
t.lm t; of t he local technical supervision . The responsibility of the
Nt I. tona l Bureau of standards' Office of Standard Reference Data is
I 11 ( nsure that quality standards are met and that the individual
111 !. lvit ies are well co -ordinated with each other.
The authority of
11111 NBS ext ends to control of the funds directly available to the
NlllWP and to control of the designation of data compilation activi1 t1 11 and products as components of the National Standard Reference
1111t.1L Syste m. Advice and recommendations will, of course , be prov 1111 d to t he s pons ors of dat a activities supported by other
11vt rnmental and non - governmental agencies, upon request.
'l'li' funds available initially arE;J not sufficient to undertake

r:v task identified as desirable nor is it possible in all cases


I 11 Io at e indiv iduals both c ompetent and will to undertake some of
I Ii highest pr iority items. The mixture of activities which is
11 c:L d and the allocation of funds to each will be the result of
11111 11x rc ise of judgment on the part of the program leaders, taking
I11111 IL c ount a ll te chnica l , financial , and personnel consideration~
wlt111t :,; e m re levant.
V

AILhough a single individual might be expected to have a good '


11M1i l know l edge in the broad scientific categories selected for
1li i I 11 1 ~ ia l stages of operation, no one person can be expected to
l111v11 t. h detailed knowledge necessary to accomplish all of the tasks
111 1lncJ i n the progr am planning stage s just outlined . Hence, the
11111 11 111 1 ad r i n each ar ea wor ks closely with a panel of consultants,
1 111 Ir o l' wit m is a s p cialis t
in an appropri ate field. The partici l"l't 11141 ol' out HL1L11cl1np; 1 ad rs
f t h tec hnic a l community of the
1

EVERETT R.JOHNSON
United States is being sought for these panels. They will be asked
to make significant contributions during all of the planning and
implementation of the program - to the definition of the area, to
the appraisal of existing activities, to the assessment of priorities,
to the recruitment of individuals and groups for specific data
compilation and evaluation activities, to the evaluation of the
technical output of the various groups under the cognizance of the
NSRDS, and to any other tasks to which they are prepared to contribute.
The recommendations of these panels will be perhaps the most
important factor in the operational decisions made in the NSRDS.
In addition to seeking the aid of the panels of consultants just
described, it is planned to confer at intervals with advisory
committees established under the auspices of the National Academy of
Sciences-National Research Council . Arrangements have been made
with the Office of Critical Tables of the NAS-NRC to provide both
the channel of communication with these advisory committees and also
advice on the over- all operation of the NSRDS. In their fields of
specialization, the NAS-NRC committees will be requested to discuss
the general and specific problems encountered in the NSRDS , to
review and criticize the NSRDS program, and in general to serve as a
means of keeping the activities of the NSRDS in tune with the needs
of the technical community.

An essential feature of the program is integration of existing


government-sponsored activities together with new programs into a
coherent whole. This will require the establishment of an effective
co-ordination mechanism, consisting partly of the formation of
committees of appropriate representatives of the agencies concerned
and partly of extensive liaison activities on the part of the members of the staff of the Office of the NSRDP.
Although the systematic planning operation described above is only
in its early stages in each of the technical categories. it has been
found desirable to begin a num'ber of new and expanded standard ~
reference data activities immediately, some within the National
Bureau of Standards and some in outside organizations. These
activities have all been considered by ad hoc panels assembled for
the purpose and have been judged to be of high priority and interest.

(j()

NATIONAL STANDARD REFERENCE DATA SYSTEM


Many, if not most, of the activities of existing data compilation
1L11d evaluation centers will be judged to be appropriate to the
program of the National Standard Reference Data System. In
10 ordance with the federal policy establishing the NSRDS (Appendix
1), s uch data centers which are sponsored and financed by any
v, overnment agency may be designated as components of the NSRDS and
c1t lled National Standard Reference Data Centers . The Office of the
NHRDP will secure the agreement and co-operation of. the agency
1oncerned before entering into any arrangement with a center regarding
11,H operation, designation as component of NSRDS, or financial
1111pport. Data centers not sponsored by any government agency are
1LIHO eligible to become part of the NSRDS; the Office of NSRDS has
1uit,hori ty to make mutually agreeable and desirable arrangements with
'111c:h centers. It should be noted that the Office of the National
rl1,1mdard Reference Data Program plans to have no operational responsi111 I ity or authority over these data centers; its function will be
111lvisory only. Opinions and recommendations regarding the operations
111 bot h government-supported and nongovernment-supported data centers
11 11d t he ir des i gnation as components of the National Standard Reference
flit Li\ System will be submitted in the form of reports to the sponsoring
11 1 11cy, upon request.
By mutual agreement with the sponsoring
111111mizat ion, the administrative office of the NSRDS may assume a
p11r"Lion of the responsibility for the financial support or for the
I 11ohni cal monitoring of the operations of the data center .
Tile procedures by which it is planned to make the standard

r nee data produced under the cognizance of the NSRDS conveniently


1vi I !able to the technical public of the United States will now be
As already pointed out, these data will be produced by a
I 11 1, number of indi victuals and groups associated with laboratories
I 11 t 11 are as of the United States and probably in several other
'11 1111 ries as well.
Many of the institutions involved will have
' d lfL 1 ng publication and distribution systems which are quite adequate
111111 which the technical community has already learned how to use.
t 1wh xisting systems will be encouraged to continue their operations
111111 L<J xpand to new fields and activities, i f such expansion is
d1 1,11 r"mincd to be appropriate.
Hopefully, all products which satisfy
f Ito c l"! Lcria of t he NSRDS will ultimately be designated as such on
"It 1111hl 1 at ions , regardless of the extent of financial support
' 1111L1 I hnt ct by t he administrative office of the NSRDS at the National
11111111t11 uf tnndards .
1 1 1'1

0I

EVERETT R.JOHNSON
The word 'product' has been used in the preceding paragraph
because of its broad, all-inclusive connotations. This word is
intended to imply that many physical forms are considered appropriate
for recording the results of compilation and evaluation activities.
The product may be a separate monograph with a hard cover or a paper
cover; it may be a critical review published in a recognized journal;
it may be a set of IBM or other data-processing cards; it may be a
computer tape or a computer print-out. In short, the product may
take any form that the users of the data find to be most convenient .
.In the planning of the detailed program, considerable study is
being given to the nature of the services to be provided by the
'Information Services Operation' of the Off ice of Standard Reference
Data at the National Bureau of Standards. The guiding principles
for the introduction of a service are that it shall be of value to
the technical community and that it shall be concerned with compilations of evaluated data. The initial planning envisions the following
activities. The data in storage will consist of a complete collection
of the world output of 'standard reference data', regardless of place
of origin or physical form, appropriately indexed and stored for
rapid retrieval. The file and compilations shall be accessible to
all qualified persons who wish to use them, and information on their
contents shall be made readily available to the technical community.
An editorial and publication service shall be operated which will
have the capability of turning the raw output from a data compilation
group into a finished product, whether it be a monograph, a deck of
IBM cards, a computer tape, or something else.
BUDGETARY PLANS
The funds made available to the NSRDP are to be used for new and
supplementary activities of the standard reference data type.
Financial responsibility for existing activities now under the
sponsorship of other agencies will not be assumed by the NSRDP except in unusual CiPcumstances~ and certainly not1until the most .
urgent needs for new projects have been satisfied. It has already
become apparent during the earliest stages of establishment of the
NSRDS that this point must be reiterated often. The program at NBS
is not intended to supplant existing activities and sources of
financial support but ta augment them.

(I'

NATIONAL srANDARD REFERENCE DATA SYSTEM


One mechanism that will be used to obtain increased financial
111pport for needed projects is to enter into co-operative arrangements
with industrial organizations and professional societies. By this
1111 nns not only financial support but the guidance of industrial
111 lentists familiar with the needs of their industries will be
1il1tained.
In order to launch the NSRDP during Fiscal Year 1964, funds were
I 1unsferred to NBS from the Atomic Energy Commission, the Department
111' Defense, the National Aeronautics and Space Administration, and
1111 Nat ional Science Foundation. Responsible officials in these

1111<0.nizations have expressed great interest in the objectives of the


NllltDS and the mechanics of the relationships between appropriate
t.11,ff members of these agencies and of NBS are now being worked out.
l''11r FY 1965 and subsequent years it is planned to operate almost
1 rn I usi vely with funds included in the normal budget of the National
1111 1'( au of Standards. When a specific task is determined to be of
1111 I Que interest to the mission of one government agency, recommendaf tern s will be made to that agency for its financial support.
Only very rough estimates can be made at the present time of the
t 11t.1Ll financial and manpower resources being applied to standard
1

r nee data compilation and dissemination activities throughout


t 1111 United States.
A reasonable figure for the present expenditure
"' f'nnds is probably $3M to $5M per year. Various estimates have
111 1 11 made of the desirable expansion, ranging up to a level four to
r I v11 times as great as the present. The eventual level will, of
11111111 , depend largely ~n the usefulness of this kind of activity in
p1rnnoti ng the efficiency and progress of science and technology in
f h1 IJntted States.
111f1

APPENDIX A

Federal Council for Science and Technology


Committee on Scientific Information
28 May 1963
' 111

n I

will be established a National Standard Reference Data


(N RDS) to provide on a national basis critically evaluated
111I11 111 Lh< phyHt al H j nc
The NSRDS will consist of a National
'1'111

f1J11

Policy on _N ational Standard Reference Data System

EVERE.Tl' R.JOHNSON
Standard Reference Data Center (NSRDC) at the National Bureau of
Standards and such other Standard Reference Data Centers as may be
required.
The National Bureau of Standards will be charged with the
administration of the National Standard Reference Data System. This
assignment will include the establishment of standards of quality,
methodology including machine processing formats, and such other
functions as are required to ensure the compatibility of all units
of the NSRDS.
The National Bureau of Standards will be charged with funding and
administering the National Standard Reference Data Center. This
Center will be an identifiable part of the National Scientific and
Technical Information System (NSTIS).
Standard Reference Data Centers covering certain specific areas
of effort may be established by or be assigned to the various
Departments and Agencies in accordance with their specific desires
and capabilities. Such Centers will be financed and administered
by the Department to which assigned but will meet the quality
standards and other requirements of the NSRDS. Such Centers will be
included as identifiable components of the NSTIS.
The NSRDS may also include Standard Reference Data Centers at
universities, research institutes, and other appropriate nonGovernment activities. To be included in the NSRDS, such Centers
will meet the quality standards and other requirements of the NSRDS
and will be included as identifiable components of the NSTIS.
There will be an Advisory Board to review and recommend policy
relative to the operation of the NSRDS. It will include among
others, representation from the National Academy of Sciences ,
National Science Foundation, Fed ~ ral Agencies engaged in research
and development. and Such other ' representatives of the scientific~
and technical community as the Director of the National Bureau of
Standards may determine .
In establishing the NSRDS, the intent is to provide an articulat ed
system of Centers and acti~ities under -such co- ordi nat i on and
di r ect i on as to ensure an output meeting quality standards for
(Jtl

NATIONAL STANDARD REFERENCE DATA SYSTEM


nat ional reference data in t~e physical sciences. The establishment
of the System should not be construed .a s preventing the establishment
of such Federal or Departmental Data Systems as are required for the
ollection of . raw or evaluated data to serve engineering or operating
11 eds of the Federal Government or various Federal agencie,s.

APPENDIX B
Office of Science and Technology
Executive Office of the President
7 June 1963
Dr.Jerome B. Wiesner, Director of the Office of Science and
'I' chnology in the Executive Office of the President, today announced

1.h establishment of a National Standard Reference Data System.


This national effort will be administered by the National Bureau of
Htandards, which is already compiling standard data. The System
I utegrates to a single point of responsibility the present datanompiling activities of the National Bureau of Standards, Department
11 f' Defense, Atomic Energy Commission, National Aeronautics and Space
Administration, the National Science Foundation, and several other
1 fl ncies.
The announcement follows the action taken by the Federal Council
f'or Science and technology, a government-wide group of top level

ncy officials in science and technology, based upon recommendations


its Committee on Scientific Information. According to Adm. Charles
II. Marte ll of the Office of the Secretary of Defense , Chairman of the
r:ommi ttee on Scientific Information, the intent is to provide an
1 rticulated system of activities under such co-ordination and
111 r ction as to ensur.e a compilation of data meeting quality
L1u1datds, and also to maintain continuous policy guidance of the
.vHL m at the level of the Executive Office. The need to improve
1 I nti fic and technical information in the Federal government has
liiitn i ncreasingly apparent during recent years.
rLK

11r

fn making the announcement, Dr.Wiesner, who is also Chairman of


I.ho Federal Council for Science and Technology, stated that the
l.1 nclard Reference Data System will greatly benefit United States
111111 11 rs and scientists by giving them easy access to the data
111 rUn nt to th ir work. . The standard source for scientific data

EVERETT R. JOHNSON
has been the now almost totally out-dated International Critical
Tables, an eight-volume publication last issued in 1933. That data
source was compiled under the auspices of the National Academy of
Sciences, Office of Critical Tables, and published by the McGraw-Hill
Book Company.
Under the new system the National Academy of Sciences will act as
advisor to the National Bureau of Standards, providing the guidelines
to the fields of science and technology having greatest demand for
data.
In recent years the mass of available data resulting from the
stepped-up Federal research and development programs, has become
much too large to be handled by a private publishing venture. Also,
the flow of new data requires continuing efforts to update collections.
Only a national system is able to cope with this information flow.
Under the plan for the new system, the data will be published in
much smaller collections than in the old International Critical
Tables. In this form, the data will be more responsive to the
dynamics of modern research and development.
Detailed plans for the system components are presently being
developed by the several participating agencies and the National
Academy of Sciences. One set of these plans will be the the subject
of an all-day meeting on June 20, 1963, sponsored by the Office of
Critical Tables, National Academy of Sciences.

...

'

(HJ

PANEL I

DISCUSSION
Ur.Rossini - We will now proceed to our intra-panel discussion, and
[ wi ll invite our panelists to make comments on any matters which
have been so far recorded.
!Ir. Chipman - Why don't I start it off by making a comment on somehing Dr.Johnson said. There's a little story going about our
A r onautics Information Office. There was a man who went to a
11 ychi atrist. He said, "Doctor, I'm afraid I'm losing my mind. I
<;o.n' t remember anything. I can't remember what I did last night.
r can't remember what I had for breakfast this morning. I can't
r member whether I kissed my wife when I left home. In fact, I
< 1tn' t remember leaving home.
What can I do?" The Doctor said,
rhis is pretty serious. How long has this been going on?" The
1110.n said, "How long has what been going on?"
llr. nossini - That's a pretty good start John.

Well, perhaps I have


Question to ask Dr.Hultgren which might start the discussion. He
1111 utioned that there were several thousand systems on which informal Ion is much needed, and my question is, Is it possible in some ways
I 11 measure a few key systems and calculate some of the other systems
wlLh sufficient accuracy?
1

lit'. Hult gren - I don't think we' re far enough along to answer that
11111 Hti on. The metallic chemical bond is much more complex than the
111111 - metallic bonds, because it involves a resonance that is very
I111portant and not much understood. We can calculate bond energies
I 11 tl great extent in organic compounds. For instance the bond
11111rgies. are generally additive except in recognized cases, and we
' 1111 alculate the properties quite well for many organic compounds
Ii I t;h have never been made. However, I don't think we' re far
111111gh along in metallic systems to do very much on this. Maybe
Ill' . Al ock will disagree.

111 /\I cock_ - No, I'm not going to disagree, but I was merely going

111 nHlt a question of what was meant by sufficient accuracy.


111 . fto1-1Hi ni

11 1 Lh
11111 1111H

- I'm sure that depends a great deal on who is going to


information, but I would think sufficient for general
R as distinguished _from some rather precise needs of a few

- I

l!ll!' f.ll ,

07

I.

PANEL I
Dr.Alcock - If I could speak as a one-time extractive metallurgist,
I think that there are many problems in this field to which we can
get approximate solutions, and the thermodynamic content of these
problems largely revolves around heats of mixing. There are many
examples in the industrial applications of thermodynamics where we
can be satisfied with a modest degree of accuracy. Of course, when
one considers atomic model building for solution theory, then the
more accurate data are vital, but here again I suppose it is prudent
not to pursue accuFacy too far, because the more accurate data one
gets, the more embarrassing it becomes to try to fit them to a model.
Dr.Hultgren - This brings up a real war between the scientifically
minded and the engineering minded person. A scientifically minded
person wants to pick out something he can measure accurately and he
avoids everything else, and concentrates on this. The engineers
problem presented here is something we've got to do and ask if you
have any data? If not, a guess comes forth and is used. If it is
correct within the factor of four, maybe it's all right. So
everyone has to choose whether he shall interpret a lot of guesses,
or shall wait for more accurate experimental data. In our books we
stuck to purity. But, you certainly have to have the guessing. But
often these things get mixed up. We just measured some heat contents
of germanium the importance of which no one would deny, and it
develops that there are tables on the heat contents of germanium
which were once originated in an idle afternoon by Dr.Kelley because
someone wanted a guess. Those have been copied and copied. I just
made the statement that if you copy a thing three times it's true.
Our measured values are considerably different, and Dr.Kelley is more
amused at this than anyone else.
Dr.Rossini - Any other comments?
Dr.Kubaschewski - I agree with Ralph Hultgren, and this conference
is concerned with the ApplicationI of Thermodynamic Data and Principles
to Practical Problems. ~ I've dealt with quite a few such practical
problems, and I don't think there was a single case when I had all
the data I needed of sufficient accuracy. Mostly, 'half of them I
have not. So, we have to supply quite a lot of guess work, and I
think it is very important for a thermochemist to have experience in
estimating data. This intFoduces, of course, substantial errors in
the calculations, but this is acc~ptable as long as the final answer

OU

DISCUSSION
18 given with a value. Even if this is large, it is better than
not hing. Now, the ternary systems which have been discussed today
I 11 particular by Dr.' s Chipman and Alcock, are examples. We usually
11 og lect the inter-action between the metal and oxygen or another
11 n-metal in a solvent metal, and we get quite good results because
I.he two effects partly compensate each other. For instance , you
q 1m calculate the solubility of alumina in nickel, or iron, and get
wlthin 50% of the correct answer. But when you are working at a
I( ve l of . 002 per cent,. this is completely sufficient and no
1u1alysis can do better than that. I can't think of any further
11xamp les at the moment, but there are many of them. Of course ,
w1' ve got to pursue the accurate measurements, and as I've said
IH fo r e, we must find simplified rules such as are indicated in these
l.u.lks. But the problem still remains that we've got to supply data
11 11 t of the blue.
Sometimes we are really impudent, and mostly, of
1ourse, we are inaccurate, but not so inaccurate that it isn't
worthwhile.
!Jr.Rossini - Thank you. I think we might turn now to some of the
1J11 s t ions from the audience. E.W. Dewing of Aluminium Laboratories,
1,1.d. asks Dr. Hultgren: By what percentage did your new heat
<'I LPac i ty values fo r silver a11d zinc differ froln the old ones based
1111 heat contents?
111. llultg ren

- I might say that we made a bad judgment in our book.


111 11 values which Kelley had taken and which we adopted were based
1111 Home old measurements of Wust and others, and it gave a fusion
11 1' Hilver at 2855. Now, three people had measured the heat of
111 11 lon of silver since then and got values near 2600 instead of
'illhfi. We discussed this at the time, because the heat of fusion is
I 11 J>nrt a matter of definition, that is, there is a pre-melting
11111111 me non which raises the heat content just before melting.
,ii loy had discounted the pre-melting data and extrapolated the
+.1 1w1 1 temperature values. We thought perhaps this accounted for
I 1111 Cli s agreement and that really the heat contents of Wust might be
11 I I right, and so we adopted them. However, we put them in
1111 1n11thesis t o indicate that they were questionnable. Now we have
1111m1111r d them i n our laboratory; therefore, we believe they are
1 I Ii L.
It seems t hat t he value of the heat fusion is closer to
1 110 ihan 28!15.
Th e s ame is true of the heat content for zinc. I
11 11" I. quot
fr 111 m m ry just how far we we r e off, but it was
11111til dc rn hlP, 1.1. 1' tho oni r of hundr ds of calories pe r gram atom.
(II)

PANEL I
Dr.Chipman - Can you express the pre-melting effect in terms of

calories and temperatures?


Dr.Hultgren - We don't know for sure that there is one.

In practice
with samples that are not too pure, there is quite a large effect,
but with absolutely pure samples, we are not sure. There is, we
think, a small effect, but not nearly as much as is measured
ordinarily.
Dr.Chipman - Do you include anything about pre-melting, or whether

to assume the heat capacity up to a point?


Dr.Hultgren - We state what we chose, and if we think the values

are good, we will include the pre-melting. But otherwise, we


ignore this effect and suggest that it is due to impurities.
Dr.Chipman - In that case you must include the pre-melting in the

heat value.
Dr.Hultgren - This is included in the latent heat, . because you must

finally come to the heat content of the liquid regardless of what


you've assumed.
Dr.Kubaschewski - The chemist will always be interested in the

total heat of fusion and not in the pre-melting value.


Dr.Hultgren - That's right.
Dr.Chipman - How much of a range of temperature does the pre-melting

effect cover?
Dr.Hultgren - A couple of hundred degrees sometimes.

It is quite

large.
I

Dr.Kubaschewski - WelP, take the transformation of iron; i.e. the

magnetic transformation of iron, you usually condense the corresponding heat of transformation to one temperature, otherwise, you
would have calculations of great length.
Dr.Hultgren - Dr.Kelley has done that. I don't think that's right.
-Our tabulation, of course, gives the whole curv
Th
cond ordor
70

DISCUSSION
r eaction takes place over many hundreds of degrees; melting is
supposed to be a first order reaction.
Dr.Chipman - Dr.Kubaschewski has suggested that this is immaterial

t o the chemist, but I don't suggest it is to the metallurgist.


Dr .Rossini - Dr.Wallace of the University of Pittsburgh asks

Dr . Hultgren: What is the basis of your assertion that heat capacities


fo r liquid metals should decrease with increasing temperature?
Dr.Hultgren - There are a few papers out on some of the low melting

metals which indicate this. We have also measured true heat


capacities for five to six metals but unfortunately our publication
rat e is not following our experimental rate. We have found this
eff ect in tin and several of the other low melting metals. The
highest heat capacity is found just above the melting point , and it
goes down rather steadily for hundreds of degrees. This has been
known for quite a while for mercury for instance where it goes down
and later comes up at a temperature approximately 2.3 times the
melting point . I think it is probably true for water and some of
the other liquids .
Dr.Kubaschews ki - If you melt it and measure the heat adiabatically

with increasing temperature and then come back to near the melting
poin t , would you get a different curve?
llr. Hul tgren - No, it reproduces, and we did this not by drop
1;alo rimetry but we set up our liquid metal in a crucible and dropped
c ld metal into it, whose heat content we knew. That gave us our
noolirig effect . We know that heat absorbed by the cold metal, and
wo divi de that by the temperature drop to find the heat capacity.
We can go up and down in temperature and reproduce the measurements .
I hope we ' 11 have these experiments published within a year.
llr. Rossi ni - One final question from J. W. Linhart of Ohio Steel

l'nundries to Dr.Hultgren has to do with the classification as satisl 1i Lory or unsatisfactory of data of the elements .
The question is:
\\\t111L
t he difference in accuracy approximately?
Well , a lot of unsatisfactory data is sheer guess.
n.t t h on tr~py of melting as 2. 3 and multiplies by the

Ur . llu I tgren 0111 ;

1-11~<

HHOA

'1 1

PANEL I
temperature of melting, and that is the heat of fusion. Then it is
guessed that the liquid will have the same heat capaci ty as the
solid at the melting point, and that the heat capacity is constant
with temperature ; so it is very hard to say what you think is the
per cent accuracy. The specific heat may be off by 20% at least.
With the heat contents , of course, the percentage accuracy looks
much better, and I can' t tell you what the answer would be.
Dr.Rossini - This is a question from Dr.Wallace of the University
of Pittsburgh for Dr.Kubaschewski regarding the chromium, molybdenum
and chromium- iron systems. How do the measured~ S's correspond
with ~ S for random mixing?
Dr. Kubaschewski - Very poorly. Actually, for us, this is sort of a
disaster. We' ve always calculated or assumed for many years that
alloy systems with complete mutual solubility and activity not
deviating too much from Raoult would be. essentially regular . We
first measured the free energies of chromium-molybdenum, chromiumiron and other systems and assumed that to use excess free energies
in place of the heats would be quite sufficient. When we actually
measured the heats, we found we had substantial excess entropies,
the maximum being of the order of one entropy unit , plus , or half
an entropy unit minus. We suspected our measurements and checked
them again and we compared the .results with those of others who used
independent methods , and we have now resigned ourselves that at
least in the transition metal systems there are considerable heats
and excess entropies . With the phase diagram calculations , this is
not necessarily too much of a drawback as long as the calculation
is not too far away from the temperature of measurement, and in
this sense the method is rather fool-proof at least for superficial
calculations.
Dr. Hultgren - I could add a little, too, for non-transition metals .
In the liquid alloys in our book 1 , we went through those that we
thought had a reasonabf y good entropy determination. We plotted
those entropies of formation and they vary as Dr.Kubaschewski said,
from one unit less than ideal to one unit entropy more than ideal;
more or less uniformly scattered. One could say the ideal entropy
was the average of them all, but it certainly was not a Boltzman
di stribution about the average . It' s rat her un ifo rm excess entropy
varyi ng from pl us on and mi nus on~.
7''

DISCUSSION
Dr.Rossini - Dr.Wallace also asks Dr.Kubaschewski about the same
systems. The presumption is that the A H's and A S's were measured
at a fixed temperature or over a restricted range of temperature.
The question is: How are values obtained for the A H's and AS' s
needed at these temperatures? Was A Cp assumed equal to zero?
Dr.Kubaschewski - Yes, that is true. I admit the fact that if AH
and A sxs are finite you will also find a non-zero A Cp. This
correction is relatively small, and if it amounts to say 2 calories
per degree, it will not affect our calculations unduly, if we don't
go away too far from the actual temperature of measurements. But I
agree that for really good assessments we should also measure the
A Cp' s.
Dr.Rossini - Thank you. This is a question for .Dr.Chipman. In your
correlation of interaction coefficients, what can be done to estimate
the values for other unmeasured elements in a series? Mr.M.Parkman
of Aerojet General wants to know.
Dr.Chipman - There have been a number of things done about this. If
you read some of Schenck's papers in the last few years, you will
ind rather elaborate systems of estimating the inter-action
oefficient for various elements on nitrogen in iron, both solid
ond liquid iron. A plot of the inter-action coefficient against the
utomic number of alloying element gives regular periodic variations
li ke many other periodic functions, and with enough spots identified
by experiment, one can estimate the others by drawing this sort of
urve. This is about the only recourse we have in estimating, where
the re is no measurement.
fir.Ros sini - Thank you. This is a question for Dr.Johnson. The
~1nes tion is for the program you described.
Why are ceramics
1 liminated since many of these systems are well definable?
~ IL , L.i ttlewood of the Steel Co. of Canada asked.
Ill'. Johnson - We have eliminated the tabulation of mechanical
11tlcu lations of ceramics because unless one has a knowledge of the
u1I ro-st ructure for the particular measurement, the measurement
ll.11 lf is not of, say, standard reference data. That is, one manurr 'turer will produce the so-called same ceramic material, and yet
11111( htmi al m asurements on this ceramic, for instance, would vary
lll)l) I' illbly,
73

PANEL I
Dr.Chipman - You can, Dr.Johnson, on the same basis eliminate tabu-

lations of mechanical properties of steels.


the past history of the steel.

These are dependent on

Dr.Johnson - This is true.

We would eliminate any system which


depends on the way you make the measurement, from our current tabulations for standard reference statements. However, I don' t wish
to give the impression that we don't recognize the importance of
this. In our first review there is such an enormous amount of
information to be tabulated, critically evaluated and stored, that
for the time being we are eliminating this type of system or substance
from standard reference data.
Dr.Chipman - You're eliminating the mechanical properties of steel .

Dr.Johnson - I wouldn't say that.

I'm not that informed to make


that statement . Whether or not you can, I' m sure that steels vary
in composition from the same manufacturer , in minor amounts; that
is, that t3e percentages of the various alloying agents vary , say, by
plus or minus a few per cent. And I'm sure that some of the
mechanical properties, they would mostly macroscopica1ly be probably
good, but maybe as you refine to measurements this would then become
more and more difficult .
Dr.Rossini - Thank you.

I might add just one word to that, and that


is there are some efforts underway to bring up to date Volume II of
the International Critical Tables which had a lot of this kind of
information in it.
Now, the final question from Mr. Littlewood to me . It says, how
are efforts being co-ordinated i~ data evaluation with other
countries?
I might answer this by saying first that in addition to the
enormous amount of work ..being sponsored and now co-ordinated under
government offices and the program which Dr . Johnson has described
to you, there are significant amounts of other work on data going
on in this country under private auspices. Some 7 years ago, we
established under the National Research Council an Office of
Critical Tables, to co-ordi.n ate private and other ent erpri ses , to
encourage them, and to help them in what ever ways wer e possi ble.
74

_ _ __

DISCUSSION
Some of these now actually are rece1v1ng some support from the
program that Dr.Johnson has mentioned. This Office of Critical
Tables that we have in our country serves as the agency through the
Nati onal Academy of Sciences whereby we can join with efforts on
data in other countries.
A movement is under way at the present time to have the International Council of Scientific Unions! which encompasses all the
Sc ientific Unions of the world. to consider broadly this important
problem of critical tables and standard reference data. We hope
that there will be established in each country something corresponding to the Office of Critical Tables and the National Standard
Reference Data Program. Then through this International Council of
Sc ientific Unions some mechanism may be set up to co-ordinate the
fforts in the various countries, because clearly there is so much
to do that it would be most unfortunate if there were any serious
dup lication. Most of all, we want to bring to light the knowledge
of all the work that is going on.
In closing I. would like to say that I think all of you will agree
ihat our first session this morning may be declared successful
b cause I think we have raised more questions than we have answered.
'l'h is problem of making thermodynamic data available is a world-wide
one. Work needs to be done in each country, with some appropriate
International organization to co-ordinate the work and to arrange
I' r unnecessary duplication of effort. I think we can see that for
ttuanti tati ve information of the kind which we are discussing, the
llltire communication process may be thought of as occurring in
rour steps:

1)

the generation and recording of the data by the original


investigator from actual experiment,

2)

the abstracting and appraisal of the numerical data and the


production therefrom of selected best values by competent
expe rts to produce what we may call critical tables of
st andard reference data,

:l)

the dissemination of these tables in appropriate form all


ov r the world,

11) LIH

1tc:L 11 1tl 11 1-1

of' t ho data. by the scie ntific community.

PANEL I
I think we may say that the aim of our system of communication in
science must be to make the proper information available to the user,
quickly and efficiently.
Now, in closing this session, I want to extend on behalf of
Dr.Fitterer and the management of the Symposium, thanks to all of
the panelists of our first session for their reports and discussion,
and to the audience for patient listening and active participation
through their questions.

,J

70

PANEL II
APPLICATIONS OF THERMODYNAMIC PRINCIPLES
TO METALLURGICAL SYSTEMS
Chairman

Dr. W.E. Wallace


Chairman, Chemistry Department, University of Pittsburgh

1.

Electrode Reactions in the Electrolysis


of Fused Salts
by Dr.G.W.Mellors, Research Scientist and
Dr.Senderoff, Senior Scientist,
Parma Research Laboratories, Union Carbide Company

2.

The Gaseous Reduction of Metals from Aqueous Solutions


by Dr.B.Meddings, Assistant Head, Research Division
and Dr.V.N.Mackiw, Director, Research and Development
Division, Sherritt-Gordon Mines, Alberta, Canada

3.

Partition of Metals between Liquid Metal Solutions


and Fused Salts
by Dr.Irving Johnson, Associate Chemist, Chemical
Engineering Division, Argonne National Laboratory

4.

Sulphate, Stability and Thermodynamic Phase Diagrams


with Particular Reference to Roasting
by Dr.T.R.Ingraham, Head, Research Section,
Canadian Mines Branch, Ottawa

f?.

The Preparation of Pure Tungsten Metal by Reaction


of Tungsten Hexacarbonyl with Steam or Carbon-Dioxide
by Dr.J.J.Ward, Research Associate, Dr.J.H.Oxley,
Assistant Division Chief, Battelle Memorial Institute,
Columbus, Ohio, and Mr.A.D.Coon, Graduate Assistant,
The Ohio State University, Columbus, Ohio

fl .

Disc ussion
Tl

PANEL I I
Dr.Wallace
Ladies and Gentlemen, I am very happy to welcome you to the
second session of the Conference devoted to the Application of
Fundamental Thermodynamics to Metallurgical Processes.
It was rather in the nature of things that the distinguished group
of panelists that we had this morning were forced to accentuate the
negative. They, in this aspect, performed, I think, quite a useful
function, but there is going to be a change of pace, I hope this
afternoon. The equally distinguished group of panelists that we
have here this afternoon and which I have the privilege to introduce
to you will have a better chance to accentuate the positive. They
will be telling us of the applicability of the thermodynamic
principles to certain metallurgical problems. The first paper this
afternoon is by Dr.G.W.Mellors and Dr.Senderoff of the Parma Research
Laboratories of the Union Carbide Company .. It's entitled 'Electrode
Reactions in the Electrolysis of Fused Salts.' The paper will be
presented by Dr.Senderoff.

711

ELECTRODE REACTIONS IN
THE ELECTROLYSIS OF FUSED SALTS
by

G.W.Mellors and S.Senderoff


Union Carbide Corporation
Development Department
Parma Research Laboratory
Parma, Ohio

INTRODUCTION
It has .been found that of the nine refractory metals of Groups
IVB, VB and VIB, all but titanium can be electrodeposited in the
form of coherent, dense pore-free deposits on a large number of
electrically conducting substrates from electrolytes consisting of
molten fluoride mixtures ( 1). The appl'tcation to electrocladding
with tantalum, or molybdenum, for example for corrosion or wear
resistance. is obvious. Also, the substrate may be removed and a freestanding electroformed object may be produced. In addition,
consolidation by electrorefining is in some cases a more economical
techR ique than the usual powder metallurgy procedures and frequently
produces a product of superior propePties. Electrorefining, both
instead of, and in addition to, electron beam melting have provided
extremely pure metals for electrical and other applications which
could not be produced in any other way, Finally, electrowinning of
these metals may be accomplished by this process. It has been of
consideraqle interest, therefore, to study the thermodynamics of the
lectrode reactions of these systems.
The study of the thermodynamics of fused salt systems has been
uccomplished by the usual thermodynamic techniques such as vapor
pressure measurements, cryoscopy, mass spectrometry, and particularly
ff ectively by static EMF measurement of .reversible cells. However,
t he corrosive nature of molten flu9rides have rendered these methods
very difficult and in many cases unworkable. This factor, together
with the ambiguities generated by questions of reversibility and
tdentity of the reactions encountered with the polyvalent transition
ni tal ions, have in some cases proved pitfalls for the unwary. For
Lhose reasons it was decided to seek the thermodynamic information
l'r m the st udy of electrochemical transients, a type of experiment
wh Loh. hnti bo n ui:1od for studying electrochemical kinetics more than
l'or

h l'lnodyn11111I H.
Ill

G.W.MELLORS AND S.SENDEROFF


A brief review of the theory of the method wi ll be given, followed
by a detailed description of its application to the study of the
reduction of fluotantalate to metallic tantalum as an illustration
of the method . Finally the chronopotentiometric results on a few
other electrode reactions in molten alkali fluoride solvents will be
described and compared with the results of static EMF measurements
in similar systems.
THEORY
The chronopotentiometric (sometimes called 'galvanostatic') method
was chosen tfor this work. The method has been described in detail
elsewhere<2) and is essentially a constant current electrolysis at a
current density in excess of the limiting current density for the
r eaction under study, so that the potential - time relationship during
the electrolysis is as shown in Fig.1 . The end of the plateau .occurs
when the concentration of the diffusing reactant at the electrode
surface falls to zero and a current density is chosen so that this
occurs within a few milliseconds as a minimum to a few seconds as a
maximum. The lower limit is that in which the time to charge the
double layer is a significant portion of the total and the upper
limit is the time when convection at the electrode becomes a significant process in mass transfer . These time limits make an oscilloscope
and camera the most convenient method for observi ng the potential-time
relationship.

-1.3

~ -I.I
...J

>-0.9

_...,. / :
i--

.r

/
~

0.2 SEC

-OJ It
0

Fig.1.

Cathodic Chronopotentiogram for K2TaF 7 Dissolved in


KF-LiF-NaF Eutectic Mixture (Flinak)
Cone. = 8 x 10- 5 moles/cm 3
Temp. = 75o 0 c
Cat hode - Plat i num
Current density= 0, 113 amp/ cm 2 .

A fundamental equation of chr nopot nti m try uppli o.b l


r

l'Hibl

HyH f.( lllH

to

'Iii:

..

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS

~7T1/2 Fnol/2

(1)

where i is the current density (amp/cm2), Tis the transition time


(sec) and is equal to the duration of the plateau in Fig.1, C is the
bulk concentration of the reacting species (moles/cm 3 ), n is the
number of electrons involved in the electrode reaction, D is the
diffusion coefficient of the reacting species (cm 2/sec) and F the
Faraday constant.
Another fundamental equation for reversible processes with soluble
reactants and products is:
RT
Tl/2_tl/2
(2)
E = E1; 4 + ln
tl/2
nF
where E is the potential (volt) at time t (sec), E114 is the potential
at t = T/4. rhe experimental parameters which are measured are T and
E114 . Eq. (2) is the Nernst equation with the usual concentration
ratio in the logarithmic term replaced by time terms calculated from
the time dependence of the concentration during the chronopotentiometric run. The logarithmic term is analogous to the polarographic
term (ia-i)/i and the E1; 4 is analogous to the polarographic halfwave potential and bears the same relationship to the thermodynamic
standard electrode potential. The left side of Eq. (1) is called
the transition time constant. Its invariance with current density
and concentration is an extremely sensitive and accurate criterion
for d~ffusion control, e . g. the absence of chemical kinetic factors,
in an electrochemical reaction. Further, the plot of E vs log
(Tl/2_t 112);tl/2 gives a straight line of slope 2.3 RT/nF and
provides a value for n, the number of electrons involved in the
electrode reaction.* Finally in a reversible reaction E114 must be
inv:ariant with current density. There.fore, by running chronopotentiog ~ams over a range of concentration and current density, one may
establish unambiguously the reversibility of an electrode reaction
If the product of the reaction is insoluble in the electrolyte
(i.e., a metal) the appropriate equation for this plot is
RT
I: = E
+ _ ln (Tl/2_tl/2). (3)
i/4_ nP
11:1

G.W. MELLORS AND S.SENDEROFF


and the number of electrons involved in it by plotting i T1/2 against
C, noting its linearity and its extrapolation to zero at zero
concentration and plotting E versus the appropriate function of t
and noting its linearity and slope. The behaviour of E114 as current
and concentration are varied is a further diagnostic criterion for
the reversibility of the react ionC 3).
When an electrode reaction occurs in a stepwise fashion then each
step may be treated in the same manner , its reversibility and n
numbe r established, and the E114 associated with each step determined.
In addition, the n numbers of each step may be found by the
relationship(2).
2

~=n2 +2n2
n 2

(3)

a very sensitive indicator for the value of n.


Another direct check on reversibility can be obtained by reversing
the current at or before the end of the plateau and observing the
nature of the reverse process. In a reversible system the E114
(forward) and E1; 4 (backward) differ only by twice the IR drop of the
system and the value of the ratio Tb/Tf is also a diagnostic criterion
for kinetic phenomenaC3).
It can be shown that when transition times of a few tenths of a
second are used (as is the case in the work described below), and
when the criteria for diffusion control and reversibility are met,
that no electrode reaction with an exchange current slower than
approximately 10 3 amp/cm 2 can be involved in the process. This is
a far greater rate than is required to establish thermodynamic
reversibility of an electrode reaction.

EXPERIMENTAL

'

Solvent Preparation
The eutectic mixture LiF-NaF-KF* (46.5-11.5-42 mole per cent,
M.P. 454 C) was used as t~e solvent electrolyte. Approximat e ly

Referred.to as 'Flinak' .

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


1200 grams of the eutectic mixture was prepared from analytical
reagent grade salts and fused in a graphite crucible with 15 weight
per cent NH 4HF 2 to prevent hydrolysis. After the fused mixture had
cooled to a temperature just above its freezing point it was poured
int o a nickel crucible, transferred to a nickel purification
apparatus an.d sparged with H2 for 48 hours at 75o 0 c to remove the
NH 3 and HF remaining from the decomposition of NH 4HF 2 and to reduce
to metallic nickel the NiF 2 formed by attack on the crucible. During
t he last four hours of sparging the melt was electrolyzed between two
gr aphite electrodes at 5 ma/cm 2 with a total current of 100 ma. With
a nickel dip cup 200 g samples were then removed , transferred to the
chronopotentiometric cell, and remelted under vacuum. This purification procedure resulted in iT 112 values for the solvent of less than
1 x 10 - 3 amp cm- 2 sec 112 at - 1.2 volts vs a Ni/Ni+ 2 reference electrode.
Thi s reference electrode is described below.
Furnace

The furnace for the chronopotentiometric cell was constructed


ar ound two 12" long, 5" ID Hevi-Duty semicylindrical heating elements
type 5712-KSP. The elements were surrounded by a 6" layer of
i nsulating firebrick on all sides except the cell port and enclosed
ln an aluminium sheath . The temperature was controlled with a chromel
ul umel thermocouple and a Model 402 Wheelco controller.
Chronopotentiometric Circuit

A conventional chronopotentiometric circuit was used with a pulsing


rn chanism and voltage cut-off switch similar to that described by
H nder off, et al. ( 4 ), except that an automatic current reversing
- ;1witch was added. The constant current source was a 200 - 375 volt
t>ower supply, and its current output was measured by feeding the
voltage from a precision resistor to a .Sargent Recorder No.S-72150 .
1:11 pot ential between the indicating electrode and reference electrode
wuti fed i nto a Tektronix 535A oscilloscope with type D plug-in. The
r<H Ulting chronopotentiograms were photographed With an attached
l'olaroid camera.

G.W.MELLORS AND S.SENDEROFF


Chronopotentiometric Cell
The chronopotentiometric cell (Fig.2) consisted of two parts.
The cell bottom (K), was an 18" high, 4-7/ 8" OD can constructed from
1/ 16" Hastelloy X sheet and a 1/ 2" steel flange. The cell top (H)
was a 12" high, 4-7/ 8" OD cylinder constructed from 1/ 32~ stainless
steel sheet and two steel flanges . The flange of the can (J) , and
the upper flange of the cell top (G), were fitted with 'O' ring
grooves, and the cell top and flanges of the cell bottom were water
cooled with 1/ 8" copper tubing to protect the 'O' rings and Plexiglas
cover (F). Fittings on the side of the cell top provided for a
vacuum line (C), vacuum ~auge (B), and a gas exit tube (A) , through
a paraffin bubbler (D) ._

1~~{~"'
/-

Fig.2.

Chronopotentiomettic Cell
G. Steel flange
H. Water-cooled stainless steel can
Vacuum gauge
I. Steel flange
Vacuum or gas inlet line
Gas exit bubbler tip
J. St ee 1 flange
Steel Flange
. ~ Hastelloy X can
Plexiglas cover di sk

A. Gas exit line

B.

c.
D.
E.
F.

Cell Components
The chronopotentiometric apparatus is depicted in Fig.3. Containers for the test melt wer e molybdenum cups (I ). 6 cm wi de by 4
i n deep , No.Q14792-A from the Fa ~ t el M tall ur gical Corporation.
l Lr d npH (C) . w n mo.de l)y Mr1 ng out 1" l n ths.
~

IHI

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


1" diameter nickel rod with a flat bottom 7/8" drill. The addition
(D), and sampling cup (E), were also made of 1" nickel rod but
wlLh a normal 3/4" drill. The addition cup had a small hole drilled
I llr ugh the bottom to permit molten salts (but not solids) to pass
lhl"OUgh. The molybdenum cup was suspended by means of a nickel
l11mger (J); the reference electrode cups by means of nickel rods
In I ned by welding.

11 1

1 111>

E
-Ht-- - F
IHl1t-W--- - G

~-illiiilil!l~-- H

c- -- -

Fig.3.

Internal Components of Chronopotentiometric Cell

A. Gas inlet tube


11. Salt bridge, usually alumina
l ). Nickel crucible for reference
electrode
I>. t)ampling cup
11: . Addition cup

F. Counter electrode
G. Thermowell
H. Indicator electrode
I. Crucible (nickel or molybdenum)
for electrolyte under study
J. Nickel cradle

Insulation Problems
The use of a metal electrolysis can in conjunction with metal
ontainers for both the reference electrode and test melt led to
i;orious e1ectrical insulation problems. In the first few experiments
t ns rilation was attempted through the use of recrystallized alumina
ror' base supports, but at 700-800C this refractory acquired
1tppreCiable conductivity in the presence of the vapors of the
fluor ides and led to electrical leaks from the anode through the
t st melt container to the grounded metal electrolysis can giving
11se to' abnormally long transition times. The only effective means
or 1 trt nl jHol atjon has b n to upport all of the cell
I

117

G.W.MELLORS AND S.SENDEROFF


components from separate nickel rods leading from the cold zone of
the cell where they are held by Swagelok fittings on the Plexiglas
cover. There are no contacts among any of the components of the
cell or with the walls of the can below the Plexiglas cover.
Electrodes
Disks for use as indicating electrodes were fabricated on a
blanking punch and die from 0. 005" tantalum or platinum sheet and
spot welded to either 0.015" tantalum or platinum wire . Counter
electrodes were strips cut from 1/8" tantalum sheet. All of the
electrodes were cleaned with concentrated HN0 3 and distilled H20.
Both platinum and tantalum indicating electrodes with areas ranging
from 0.49-1 . 67 cm 2 were successfully used, although the majority of
the experiments were carried out with platinum electrodes having an
area of about 0. 9 cm2.
Reference Electrodes
A reference electrode suitable for use in molten fluorides
represents a problem not yet completely solved. Gr jotheim (S)
described a Ni/NiF 2, alkali fluoride system which was only moderately
constant for about two hours and then drifted badly. Manning ( 6)
used a platinum wire immersed directly in the fluoride melt under
study which he called a 'quasi-reference'. CoriouC 7 ) used a Ag-AgCl
reference with a fluoride salt bridge contained in a graphite
vessel . Crouthamel'sC 8) description of the Li 3Bi(s), Bi(l), Li+
electrode seemed like the best approach to us but we found it to be
very unstable, probably due to the reduction of K+ to K0 by the
Li 3Bi. Since the solvent of, our plating baths includes potassium ion
we were loath to eliminate it from our reference electrolyte.
Our most successful appr oach was a half - cel l similar to Grjotheim' s
in which a solution..,of 1 mole .iJer cent NiF 2 in Flinak was containf?d. i n
a nickel crucible which acted as the electrode . The NiF 2 solutions
were obtained as dip samples from a 1000 gram, l mole per cent
NiF 2/Flinak bath. The NiF 2 used for the refer ence electrodes ~as
prepared using the method of H. F.PriestC 9 ) except that CIF 3 was us ed
instead of F 2. Contact _between the r ef er ence and experimental
el ectrolyte was made wit h t he porous al umina br i dge de cribed below.
/I ll

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


Salt Bridge
The salt bridges (Fig.28), were cut on a diamond saw to a 'U'
shape from Norton Company alumina filter disks 8" in diameter, 1/4"
thick of RA-98 coarse particle size. The bridges were leached in
concentrated HN0 3 for 24 hours, continuously washed with distilled
H2o for 1/2 hour, immersed in boiling H20 for 1 hour, and vacuum
dried at 400C. They were then immersed in a purified molten Flinak
bat h under argon, cooled and stored in closed bottles in a dry box.
Bridges so prepared had a resistance of less than 10.0at 750C, but
on long contact with the molten fluorides they introduced impurities
i nto the melt which were reduced at -1.5 V and at -0.5 V. Introduction
of the impurity at -0.5 V could be avoided by immersing the bridge in
the melt only during the recording of the chronopotentiogram. The
impurity reduced at -0.5 Vis probably iron and the less noble one
reduced at -1.5 V is aluminium.
Procedure
Only moderate cleaning of the cell constituents was necessary
ufter each experiment, unless, as occasionally happened, an air leak
d veloped around the epoxy sealed Swageloks in the Plexiglas top.
The Hastelloy can was cleaned by sandblasting, the nickel rods, cups
ind hangers by dipping in concentrated HN0 3 followed by water
Immersion and a distilled H2o wash. Replacement was necessary only
whon the nickel rods became so annealed by exposure to high
t. mperat ure th'.at they would no longer remain reasonably straight.
The salt bridge, purified solvent slug, and NiF 2/alkali fluoride
l'erence electrode slug had been prepared previously and stored
1111cl r cover in a dry box.
The remainder of the apparatus was
cleaned, oven dried, assembled and the system evacuated overnight
wlLh the furnace temperature set at 600c. If a vacuum of less than
,,, had not been attained by morning, the cell was removed for exami1111LI. on with a helium leak detector.
The cell was then cooled, vented
1. 11 LLrgon, and the reference electrode slug, salt bridge and purified
1111 IL sl ug inserted, followed by vacuum evacuation overnight at 400C
Lo r move any adsorbed moisture. After fusing the salts under
v1 t 1u1rn at 6oo 0 c for 2-3 hours the cell was vented to argon, the
111run
temperature raised to 750C and the transition time of the
o Iv1 uL r
rd d aft r :t mperat ure equilibrium had been reached.
11

Ill>

G. W.MELLORS AND S.SENDEROFF


The solute (i.e. K2TaF 7) contained in a nickel addition cup in the
cold zone of the cell was then plunged into the melt.
The usual
procedure for recording chronopotentiograms was followed. At a
given concentration and temperature runs at a number of current
densities were recorded. The temperature was then changed, allowed
to become constant, and recordings at another group of current
densities were made. When the appropriate temperature and current
density range had been completed a sample for analysis was taken an?
the run terminated. Another experiment was then performed in the
same way with a new solution at another concentration .
THE REDUCTION OF FLUOTANTALATE
Results
Chronopotentiograms were run at concentrations of K2TaF 7 ranging
from 0. 02 to 0. 1 millimoles cm- 3, at current densities which varied
from 0.05 to 0.25 amp/cm 2 and at temperatures of 640 to 86o 0 c. The
chronopotentiogram of Fig.1 is typical of those obtained in this
study. It shows two reduction steps with E114 values at about 1.0 _.
and 1.2 volts. (The third step at 1.5 Vis due to an impurity of Al
from the salt bridge.) In Fig.4 is shown the relationship ~mong
transition time, r, current density, i, and concentration, C, at
750C for the first plateau. Each of the points represents the
average of about 10 chronopotentiograms and includes the maximum
practical current density range for that concentration. The average
value for [E 1 J114 for this set is 1.03 V 0.05 V for the first
plateau. The average value of the [E2J 114 for the second plateau ,.
is 1.26 0.06 V. The transition time of the second plateau is on
the average about 80 per cent longer than that of the first plateau.
Tables I and II summarize toe data obtained at 75o 0 c.
010

0.08

...

!:~

0 .0<

0 .04

'

~
~, 0.0 2

CONC.

Fig.4.

r:. t01 MOLES Ct1i1

10

Transition Tiine, Current, Concentration Relationship fro m


Cathodic Chronopotentiograms of K2TaF 7 i n Flinak at 7r:o 0c

no

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


In Fig.5 is shown the chronopotentiogram for the solvent only,
indicating the absence of anything significant below the alkali
metal reduction potential near 2 V.

-2.0
-1.5

V>
f-

,.
-

-LO

>

-0.5

0.05 SEC

Fig.5.

Cathodic Chronopotentiogram of Flinak Solvent Only


Current density= 0.0336 amp/cm 2
Temp. = 750c
Cathode - Tantalum

In Fig.6 is shown the result for successive cathodic and anodic


1:hronopot entiometry on the system. The ratio Tanodic/Tcathodic =
!I ,

83.

-1.0
(/)

I...J

>

-0.8
-0.6

=~

---

I SEC

'

-0.4

r---

-- '

0
1'11(. 6.

Cathodic-Anodic Chronopotentiogram of K2TaF 7 Dissolved


in Flinak.
Cathodic portion on left, anodic on right
Current density = 0.00964 amp/cm 2
Temp. = 75o 0 c
Cathode - Platinum

. Ill

G.W.MEILORS AND S.SENDEROFF


Discussion
1.

PRECISION OF DATA

An overall precision of about 5 per cent in the value of [E 1J114


for the first plateau is indicated by the data in Table I. This
corresponds to a standard deviation of about 50 mv in the potential
and at least 20 mv of this can be ascribed to the instability of the
reference electrode. Additional work on a reference electrode for
molten fluoride systems is required. No dependence of E114 on
concentration or on current density has been observed.
The transition time constant, ir 111 2;c, for the first plateau
also has a standard deviation of about 5 per cent, with the major
portion of the imprecision arising from the irreversibility of the
second reduction step. Because of this, the graphical determination
of r 1, the transition time of the first step, is made imprecise by
the slope of the transition region between the first and second step
shown in Fig.1. No dependence of ir 11/2;c on either current density
or concentration is noted.
In T~ble II it is seen that the [E 2J114 values are only slightly
less precise than the [E 1J114 values but the quantities associated
with the transition times are very much less precise. Also, except ,
for the highest current density, the transition time constant for
the second step, ir 211 2;c, is not constant, but de~reases with
current density. This lack of precision and constancy in transition
time data arises from the fact that the second step is not diffusion
controlled but has rather complex kinetics and is irreversible.
It will be noted in TabLes I anG II that the standard deviation
of the E114 values for each concentration is generally less than ,
that of the overall average. This suggests that the stability of
the reference electrode during the time it takes to complete the
experiment on one alectrolyte' is better than its reproducibility.
from experiment to experiment. In the case of the transition time
constants, however, the per cent standard deviations for the
individual experiments is not consistently better than the overall
value, but improves with increasing ir 1112 values. The absolut e
values, however, show - ~o consisrent variation with concent rat ion r
ir1112 and suggest, therefore, that . ther e is an absol ut
rr r n

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


TABLE I

Chronopotentiometric Data for the Reduction of K2TaF 7 in


Flinak at 750c

Cone.
K2TaF 7
C.D.
(moles cm- 3 ) (amp/cm 2)
x 10 5
Min. Max.

2.0
3.0
4.0
4.4
4.5
5.9
5.9
7.0
8.0
9.6

0. 056
0. 075
0.065
0.092
0.092
0.050
0. 045
0.050
0.061
0.113

0.110
o. 230
0.115
0.142
0.158
0.250
0. 246
0.180
0.240
0. 151

Overall Average

iT 1/2
No.
[El] 1/4
[E2] 1/4
1
(amp sec 112cm- 2) of
(Volt)
(Volt)
Avg. std.Dev. Avg. Std.Dev.
Avg. Std.Dev. Runs

1. 06
1.05
1.08

0.025
0. 024
0.056
0. 025
o. 033

1. 24
1. 28
1. 29
1. 34

0.057
0.031
0.052
0.040
0.022

o. 97

0. 038

1.18

0.020

0.97
1.07

0.052
0. 017

1.18
1. 31

0.064
0.021

1. 04
1. 02

1. 03

0.054

03

1. 24

1. 26

0.060

0. 020
0.031
0.045

0.060
0.057
0.071
0.081
0. 093

0.0025
0. 0029
0.0015

10
11
11

0.0015
0.0028
0. 0030
0.0024
0.0036

3
4
5
11
12
7
7

iT 1/2/C
1
(amp sec 112cm mole- 1)
x 10- 3

Avg.

Std.Dev.

1. 01

0.05

G.W.MELLORS AND S.SENDEROFF


TABLE I I

Transition Time of the Second Plateau for the Reduction of


K2TaF 7 in Flinak at 75o 0 c

Cone.
1 2 /T
1
(mole cm- 3)Avg. Std.Dev.
2. oxrn- 5 2.9
3.0
2.3
4.0
0.91
5.9
1. 5
8.0
2.1

1. 1
0.8
0. 07
0.02
0.02

Overall
Average

0. 7

1. 8

iT 1/2
No.
2
i1 2 1 1 2;c
of
(amp sec 112cm- 2)
1
2
1
(amp
sec
l
cm
mole)
Runs
Avg. Std.Dev.
.036
.044
. 044
. 068
. 118

.010
.005
.006
.013
.052

1. 80
1. 47
1.10
1.15
1. 47

TABLE III

Transition Time Ratios for Various Values of n

...

T2/ 1 1

Ill

n2

0.56
1. 8,
5.3
24.0

4
3
2

1
2
3
4

'

9
5
11
11

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


t he measurement which is independent of the concentration, current,
or any of the variables. This absolute error is most likely the
graphical measurement of the transition time which becomes smaller
percentagewise as the transition time increases .
2.

THE ELECTRODE REACTIONS

Figure 1, when compared with Fig.5, shows three steps at 1, 1.2


und 1.5 volts not associated with the solvent. The last of these
arises from aluminium introduced from the salt bridge , and i s
l iminated when a boron nitride salt bridge is used i nstead of the
al umina. The two steps then, at 1 and 1. 2 volts , are associated with
t he reduction of the fluotantalate to tantalum metal. This
i mmediately illustrates the difficulty in obtaining thermodynamic
lnformation from static potential measurements for if one inserts a
Lantalum sheet into a fluotantalate solution such as this, one must
know which of these reactions (i"f either) is determining the
ni asured static potential.
Such an experiment was done and a potential
of 0. 73 volt was observed. Clearly neither of these reactions is
Involved . in the static electrode potential. The properties of the
ol ctrode reaction associated with each step will now be evaluated
h:V the criteria of chronopotentiometry discussed above.
a. The First Step - Examination of Fig . 4 and Table I confi rms
Lhnt Eq. (1) is rigidly satisfied by the chronopotentiogram of the
!Ir t step. Figure 7 shows that Eq.(2) is also satisfied except
ll<>H::lib ly near the high potential end of the plateau. In addition,
11t1/2;c is seen to be invariant with concentration and [E 1] 114 is
rLIHo invariant with concentration and with current density. Thus
rtll t he criteria for a reversible diffusion controlled reaction are
ru~ L h1fied by the reaction associated with the first plateau of the
1 J11onopotent iogram.
The value of n obtained from the slope of the straight
I I11 0 i n Fig. 7 is 3. 3 O. 8, for the average value taken from 38
1 lll'n nopotentiograms.
The precision is not as good as would be
rlui I rCd; mainly beoause E varies rather slowly with t on the plateau,
1111!. l.h alte rnative method of obtaining n given by Eq. (3) may be
11111 d.
I n Table III are shown the values of T 2/T1 for various values
if' . 11 :~ tind nl .
onsid r~ng t hat n 2 + n 1 = 5. The average value of
fl

G.W.MELLORS AND S.SENDEROFF


T /T , from the 43 chronopotentiograms of Table II is 1.8
2 1
thus confirming a value of 3 for the n of the first step.
0.4

0.7,

a f(t) .. LOG [<r 112-t112112 )/t 112}


o

f(t) LOG (T

112
-t

0
-0.4

-0.8
-1.2
-1.6

1.05

1.00

1.10

VOLTS

Fig.7.

Instantaneous Cathode Potential as a Function


of Time During a Chronopotentiometric Run
Cone. = 5. 9 x 10- 5 mole/cm 3
Current density = 0.045 amp/cm 2
Temp. = 75o 0 c

Since the initial reactant in this step can only be a


species containing pentavalent tantalum, most likely [TaF 7] 2-, the
potential associated with the first step is that for the reversible
couple, Ta5+ + 3e !:; Ta 2+. It might be added for further confirmation
of the initial valence state that numerous analyses of the
K2TaF 7-KF-LiF-NaF solutions, even after long periods of electrolysis,
always show a valence for the tantalum in solution of 5.0.
There is some question regarding the nature of the divalent
species. The linearity of the, plot of Eq. (2) and the invariance of
E1; 4 with concentration point to the divalent material as a soluble
product. However, if it were soluble, Tanode/Tcathode in Fig.6
should be 0.33( 3). It is found to be 0.83 at 9.64 ma/cm 2, 0.88 at
45. 4 ma/cm 2 and 0. 90 at 79. 5 ma/cm 2. For a totally insoluble
product the ratio of anodic to cathodic transition times should be
unity since all that is cathodically produced at t .he electrode
remains there to be oxidized in the anodic cycle <3). This would.
indicate that the product of the first reduction is only slightly
soluble, that some of it .dissolves and diffuses away, but that the
. faster the cathodic reducfion is accomplished the less of it
escB.t>es fro m th el ctr d b for th r oxidati n At p, This is
onf1rm d bY th
nfli no. pr iun 1 of 1u1 1mrnlub(( Hludl(t

..

1)(1

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


consisting mostly of divalent tantalum fluorides in macroscopic
electrolyzed solutions. The behavior of the E vs t relationship
and the constancy of E1; 4 can be explained by suggesting that
through most of the time, T, and certainly well beyond T 114 , one
is dissolving the product in the cathode layer, and only near the
end of time, T, is saturation reached. The departure of the last
two points from the straight line of Fig.7 may be a result of this
saturation effect. The anodic cycle then oxidizes what is finally
deposited as a solid on the electrode as well as whatever is
available of the material in solution. It appears, therefore, that
the final product of the first reduction step is slightly soluble
TaF 2.
b. The Second Step - The product of the second reduction
is tantalum metal. This can be identified on the electrode and in
fact is known to be produced macroscopically at high efficiency over
a wide range of operating conditions in the plating process. The
react'ant is clearly the product of the first step and is almost
certainly solid TaF 2 or a saturated solution of this compound in the
molten eutectic mixture. The reaction is not diffusion controlled
as is shown by the decrease of iT 211 2;c with concentration. The
study of the details of the kinetics of this reduction is made
difficult by the imprecision of the data for this step. It is
be lieved that the lack of reproducibility results from varying
degrees of saturation of the electrolyte around the cathode and from
uncontrollable partition of the current between the solid and dissolved species. It is probable that potentiostatic measurements may
be more suitable than chronopotentiometric studies for th~
elucidation of the .mechanism of the final reduction step. Further,
whil e Fig.6 shows that the first step is highly reversible, cathodica.nodic chronopotentiometry on the second step shows a wide divergence
between the anodic and cathodic potentials. This probably indicates
u.ddi tional electrochemical complications requiring further study.
Macroscopic observations which emphasize the importance of
Lh sparingly soluble intermediate, TaF 2, are (a) tantalum metal is
sse ntial ly unattacked in contact with K2TaF 7 in molten Flinak
h t ween 650 and 850C. This is in contradistinction to the rapid
ILLtack of niobium metal in contact with K:fibF 7 in the same solvent;
(h) the sol ut ion of fluotantalate in contact with the metal or after
11 trolysis always hai:i an apparent valence of 5.0 while that of
97

G. W. MELLORS AND S.SENDEROFF


fluoniobate has a valence less than five and after prolonged
electrolysis approaches 4.0. This demonstrates the existence of a
soluble reduced niobium fluoride and shows that the tantalum is
inert in its pentavalent salt only because it is protected by the
insoluble difluoride. This also suggests that the reason that
immersion of a tantalum strip in a fluotantalate solution gives a
potential not identifiable with either step in the reaction is that
the strip is passivated with a coating of TaF 2 and that the electrod~
under observation is actually Ta, TaF 2 (s) IF. and irreversible. It
might additionally be noted that the potential value of 1.26 volts
found for the second step is of interest in the operation of the
plating solution but can have no thermodynamic significance since it
refers to an irreversible process.
THE REDUCTION OF FLUONIBATE
Figure 8 is a chronopotentiogram for the reduction of a solution
of K:fibF 7* in KF-NaF-LiF eutectic solvent at 75o 0 c. The first two
plateaus at 0.75 volt and 1.03 volts are steps in the reduction of
niobium and preliminary results indicate that they represent successive 3-electron and 1-electron reductions . The third plateau
near 1. 4 volts is due to an aluminium impurity and the fourth plateau
near 1. 8 volts is due to reduction of alkali metal of the solvent .
- 2.0
-1.5

....... ~

(/)

~ - 1.0
0

> - 0 .5

_v

Fig.8.

--0.05 SEC

Cathodic Chronopotentiogram for Reduction of


Niobium Dissolved in Flinak
... Current f1ensity = 0. 300 amp/cm2
Temp . = 650c
Cathode - Platinum

* The niobium had been reduced to a valence close to 4 by elect roly


In the pentavalent solution the Nb5t.Nb 4+ r eduction can be observ d
at o. 11 V at low current density (0. o3o amp/cm2).

fin

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


The same cri teria as used for the fluotantalate indicate that the
first step, Nb 4 + + 3e ..... Nb 1 +is reversible, but that the monovalent product
is soluble as shown by the Tb/Tf ratio of a~proximately 0.4. This
latter is seen in Fig.9A which also shows the near identity of the
forward and backward E114 values, thus confirming the reversibility
of the first step. Figure 9B, however, which is the cathodicanodic chronopotentiogram for the second step, shows the complete
irreversibility of this step in which niobium metal is produced.

0.5SEC

1--

-1.5

~ -1.0

g-o.5
0

II
7 l

:,.- i-- ~

...J

A. First Step of
Niobium Reduction
Fig.9.

0.5 SEC

I'\

B. Second Step of
Niobium Reduction

Cathodic-Anodic Chronopotentiogram

REDUCTION OF CHROMIUM FLUORIDE


Figure 10 is the chronopotentiogram for the reduction of CrF 3 in
the same solvent. At 750C it is also a 2-step reduction with the
fir st plateau 0.69 volt and the second at 1.13 volts. The ratio
12/T 1 indicates that it represents a 1-electron reduction followed
by a 2-electron reduction. The first step is then Cr 3+ + le ..... Cr 2+
1u1d the second step Cr 2 + + 2e ..... cr 0 . It is interesting to note that
IL chromium strip immersed in a solution of CrF
3 in molten NaF-KF
1utectic gave a potential of 0.70 volt vs Ni/NiF 2 at 85o 0 c< 5 ). This
uotential was assigned by Grjotheim to the Cr 0 /cr 3 +. A chromium
11Lri p .immersed in a solution C1f CrF 2 in molten NaF-ZrF 4 (53-47 mole
11< l' cent) mixture gives a potential of 0. 75 volt vs Ni/NiF 2 at 70o 0 c(lO).
JI, was shown later that CrF
disproportionates ( 11 ) under these con2
<11 Lions according to the reaction 3 CrF 2 => 2 CrF 3 + cr 0 The
1 uui libri um constant for this reaction has been measured and is such
cr 2 +
I.hit!, 1tt 800C the r at io - - in the LiF-NaF eutectic mixture is 1. 02
Cr 3 +
11 11<1 Ir~ Lil~- N uJ1'-Zr F 4 (47 mol pr cent ) it is 14.8.
It is clear,
0)

G.W.MELLORS AND S.SENDEROFF

-2.0
(/)
.....

-1.5

-1.0

..J

> -0.5

j.

--

0.05SEC

Fig.10.

100

Cathodic Chronopotentiogram for the Reduction


of CrF3 in Flinak
Current Density= 0.143 amp/cm2
Temp. = 75o 0 c
Cathode - Platinum

therefore, that immersion of a chromium strip into either


CrF 2 or CrF 3 solution in these molten fluorides gives a potential
related to the cr3+;cr 2+ couple and the value obtained will depend
on the equilibrium constant for the disproportionation in the
particular medium. The chronopotentiometric method clearly indicates
the correct equation for the assignment of the 0.69 volt reduction.
Preliminary results indicate that the second step is not reversible.
REDUCTION OF FERRIC FLUORIDE
The retluction of ferric fluoride dissolved in molten KF-LiF-NaF
eutectic mixture produces a chronopotentiogram very similar in
appearance to that for CrF3 with the two reduction steps at 0.21 volt
and 0.512 volt at 750C. The ratio r 2/r 1 indicates successive
1-electron and 2-electron reactions so that the first step represents
the Fe 3+ + le __, Fe 2+ and the second step the Fe 2+ + 2e __, Fe 0 reacti~ns.
Plots of ir11 2 vs concentration show that both steps are reve~sible
and a strip -of iron immersed in a solution of FeF 2 in the same
' volt in good agreement with ttie solvent gives a pote ~tial of 0.50
value found in the second step of the chronopotentiogram. Grjotheim(5)
reported a static potential of 0.12 volt for iron in contact with
FeF3 in NaF-KF solution. This is in fair agreement with the value
found here for the Fe 3+/Fe 2+ couple . It is interesting to note that
FeF 2 is reported to be stable and not -dispropor tionate ( l l) under
these conditions .
' 1 ()()

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS

TABLE IV
EMF Series for Transition Metal Ions in Molten
NaF-KF-LiF Eutectic Mixture at 1023K

E1; 4 (Volt) [vs NiJNiF 2


(1 mole%), Flinakl I
Reference Electrode]
Al III _,Alo
Ta II
crII
TaV
Nb 1

- 1. 5

_, Ta 0

- 1. 26

cr 0
_, TaII

- 1. 13 *
- 1. 03
- 1. 03

-+

_, Nb 0
Nb IV _, Nb 1

- 0.754

crIII _, crII

- 0.69

Fe II

_, Fe 0

- 0.512

Fe III _, FeII
NbV _, NbIV
Ni II

_,Ni

- 0.21
- 0.11
- 0

known to be potentials of irreversible couples

101

G.W.MELLORS AND S.SENDEROFF


SUMMARY

In Table IV the results of these and other measurements are


collected in the form of an EMF series for the transition metals in
the molten NaFLiF-KF eutectic mixture as solvent. The value of
this fundamental thermodynamic jnformation in electroplating and
electrorefining is obvious. Conclusions from this may be derived
regarding the electrolytic separation and purification of mixtures.
Information on the effects of impurities and possibilities of
alloying is easily deduced from this, and as it is extended with
additional meits it will become even more valuable. The reproducibility of the method is still not good enough, however, to
obtain adequate entropy information from the temperature coefficients
of these EMF's. The standard deviation of about 5 per cent in the
EMF values gives a probable error of about 25 per cent of their
temperature coefficients when measured over a range of about 150.
The small temperature coefficients (i.e . . . . , 0.8 mv/C 0 for the
Ta 5+/Ta 2+ couple) requires that precision in measuring EMF be
increased considerably if the results are to be valuable in entropy
calculations. , This improvement must start with a more stable
reference electrode.

...

10 2

ELECTRODE REACTIONS IN THE ELECTROLYSIS OF FUSED SALTS


REFERENCES
1.

G.W.Mellors and S.Senderoff, Can.Pat. 688,546 (1964).

2.

P.Delahay, New Instrumental Methods in Electrochemistry ,


Ch.8, Interscience: New York (1954).

3.

W.H.Reinmuth, Analyt.Chem. 32, 1514 (1960).

4.

S.Senderoff, E.M.Klopp and M.L.Kronenberg, Final Report


Contract NOrd-18240, Part II (Sept.15, 1962).

5.

K.Grjotheim, Zeit.Physik.Chemie, Neue Falge 11, 150 (1957).

6.

D.L.Manning, J.Electroanalytical Chem. 6, 227 (1963).

7.

H.Coriou, J.Dirian and J.Hure, J .,de Chemie Physique 52, 479


(1955).

8.

M.S.Foster, C.E.Crouthamel, D.M.Gruen and R.L.McBeth, J.Phys.


Chem. 68, 980 (1964). CAlso M.S.Foster, et al. Inorg.Chem.
3, 1428 (1964).]

9.

H.F. Priest, Inorganic Synthesis , 3, Ed. L.F.Audrieth,


McGraw-Hill: New York (1950).

10.

W.R.Grimes, Private Communication.

11 .

J.D.Redman, O.R.N.L. 2626 M.S.R.P., Q.P.R. 94-96 (1958).

IO:J

THE GASEOUS REDUCTION OF METALS


FROM AQUEOUS SOLUTIONS
by
B.Meddings
Assistant Head, Research Department and
V.N.Mackiw
Director, Research and Development Di vision
Sherritt Gdrdon Mines Limited, Fort Saskatchewan, Alberta

Abstract

This paper deals with the use of gases to reduce aqueous


solutions of metal salts. While the main emphasis is on systems
where the reduced state is the metallic element the scope of the
paper also embraces those systems where gaseous reduction produces
a precipitate of the metal in a lowered valency state.
The physical-chemical parameters controlli ~g the gaseous reduction of such dissolved metal species, thermodynamic feasibility,
t hermodynamic equilibria, reaction kinetics, reaction mechanism and
catalysis, are discussed within a broad general framework. More
de tailed, specific, discussion is given for the hydrogen reduction
of solutions of metals such as cobalt, copper, molybdenum, nickel
and tungsten.
The state of knowledge of the subjec't is reviewed and particular
areas where future research work would probably be most rewarding
are indicated and discussed.

1Oh

B.MEDDINGS AND V.N.MACKIW


INTRODUCTION
Until recent times metals have been obtained either by direct
mining or by pyrometallurgical processes. It is only comparatively
recently that metal recovery processes based upon the properties of
metal ions in aqueous solution have achieved commercial importance.
The first of these processes was electrolysis and this was introduced as a final purification step appended to an old-established
pyrometallurgical operation. During the last decade advances in the
hydrometallurgical industry have been made by which the metals are
brought into solution, separated, and precipitated out of solution
by more extensive and specific utilization of the chemical properties
of the metal ions in aqueous solutions. In general these reactions
have been carried out in aqueous solutions and temperatures above
100c have been used to achieve useful reaction rates, as a result
pressurized equipment has been mandatory.
The details of the leaching processes by which useful metals contained in ores are brought into solution and the solutions purified
have been well described by Mackiw and others. 47,17,23
The main emphasis of this paper will be on the use of gaseous reducing agents to reduce aqueous solutions of metal ions to precipitate
metallic powder. However, some consideration will also be given to
those cases where gaseous reducing agents cause the precipitation of
a chemical species of reduced valency which can then be converted to
metal in no more than one further processing step.
These reactions will be considered in terms of thermodynamic
probabilities, their kinetic characteristics, their feasibility on a
commercial scale, and the probable extension of such hydrometallurgical
operations.
THERMODYNAMICS

A study of the thermodynamics of a reaction system can decide how


far the reaction may proceed, it cannot decide how fast the reaction

will proceed. Such a statement is a truism but it is still worth


emphasising that although thermodynamics may indicate that gas A can
reduce Men+ to metallic Me it -does not necessarily f ollow that the
reduction can be made to proc d at 9: us f ul r at und x- av'i l abl
ondHJOllH.
1011

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


The Equilibrium Cons tant
The extent to which a reaction such as
A + B - - - - - - ? C +D

(2)

will proceed is given by the equilibrium constant K which is defined


as

[cJ [o]
K

[A] [BJ

(3)

Very often the square bracket terms are equated t o the analyti cal
concentrations of the various species but from the thermodynamic
vi ewpoint, with which we are now conce rned, the square brackets refer
to the effective chemica l concentrations, i.e. the activities of the
species. Equation (3) should therefore be re - written as
K

(3a)

whe re ac signifies the activity of C. The thermodynamic equilibrium


constant K is related to the standard free energy change , LF0 , of the
reaction by the van' t Hoff isochore
LF0 = -RT , ln K

(4)

The standard free energy change, L F0 , for the reaction can also be
xpressed in another form. If the reaction in (2) is considered to
b the basis of a perfectly reversible electrochemical cell then tlte
maximum net work obtainable from it is n.E . F. where n is the number
of Faradays passing through the cell to complete the reaction, E is
t. he potential of the cell in volts and F is the Faraday (96, 487
oulombs). This maximum net work must equal the standard free energy
hange of the reaction, hence

-.Cw0 = n.F. E.

(5)

'l'huH alt hough t he ext ent t o which a reaction will proceed is given
d r ctly by t h equ ili brium constant K there are three
1 t111 I v 1t l 11t ways in wh i h this tendency to completion can be assigned
1 11 ll1 1H~1 nl va.lu, lh H ar
i nt rr lat d by
10 7

11101-1L.

B.MEDDINGS AND V.N.MACKIW


f:iFO = -n.F.E. = - RT ln K
Care must be taken to express the constants F and R in compatible
units. If R is expressed as 1.987 cal. deg. -l mole-i, as is usual,
then F which is defined as 96,487 abs. coul. g. equiv. -l must be
expressed as 96,487 x 0.2390 since one international volt-coulomb
is equivalent to 0.2390 of a standard calorie. Using compatible
units,
f:iF

-n.E. 23,060

2. 303 RT log 10 K

(6)

The interdependence of these three expressions of the thermodynamic


probability of a reaction has also been given by Schaufelberger. 1
The Use of Electrode Potentials

It will be seen later that the most widely used reducing gas for
precipitating metals from solutions is hydrogen. For this reason
most attention will be given to the thermodynamics of the general
reaction
n
Men+ + - H
2

Me + n.H+

(7)

The equilibrium position of this reaction could be calculated from


use of the standard Gibbs Free Energy values but it is found to be
more convenient to use electrode potentials. In any case much of
the tabulated free energy data has been calculated from
electrochemical potentials.
The reaction described by equation (7) can be viewed as two
opposing reactions, each with their own potential,
Me
and
H2

M1;1.n+ + n. e. 1

Potential = EMe

f7a~

Potential = E

(7b)

H2

Clearly if EMe exceeds ~H 2 then the metal will liberate hydrogen fro m
water and pass into solutlon but if EH 2 exceeds EMe the hydrogen wil l

108

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


reverse reaction (7a) and precipitate metal Me from solution. The
potentials denoted by EMe and EH are single electrode potentials
and are defined in the usual ele~trochemical nomenclature by,
EH

Eo
,H2

2.303 RT
F

log 10 ~+

(Sa)

loglOaMen+

(Sb)

and
EMe

(i)

= EoMe

2.303 RT
n.F

THE HYDROOEN ELECTRODE POTENTIAL

The potential of the hydrogen electrode is governed by the concent ration of hydrogen ions and hydrogen molecules in solution in contact
wi th the electrode and is therefore a function of both pH and hydrogen
gas pressure. Equation (Sa) ,should therefore be re-written as

~2

= EoH2 -

2. 303 RT
2. 303 RT
x
. log1o'ii+ +
. loglOPH
F
2F
2

(Sc)

where PH x is the fugacity of the hydrogen. However, between 1 and


100 atmo~pheres the effect of using fugacity in place of hydrogen
par tial pressure is negligible , at 100 Ats. hydrogen pressure the
f ugacity is 106.1 atmospheres 2 and this only alters the last term in
(Sc ) by 1%. By definition pH is defined as - log 10aH+ and since the
st andard electrode potential of hydrogen, E0 H , is defined as zero,
.
2
equat i on (Sc) reduces to
EH

= 0. 05916

pH +- 0.0295S loglOPH

(Sd)

nt 25C where PH is the partial pressure of hydrogen expressed in


2
ntmospheres .
(ii )

THE METAL ELECTRODE POTENTIAL

The single electrode potential for a metal ion in acqueous


i on is controlled by the standard electrode potential for that
tun nnd t h activi ty of that ion i n the solution

H<J l ut

100

B.MEDDINGS AND V.N.MACKIW

By definition

where fMen+ is the activity coefficient of the ion Men+ and [Mell+]
is the analytical concentration of the ion. Hence at 25C.
EoMen+ -

0. 05916
n

. loglO [Men+] -

0.05916
n

At low concentrations the activity coefficient approaches unity and


the last term in (8d) vanishes. But in 1 molal solutions, which are
commonly used in hydrometallurgical operations, it is seldom
possible to use EMen+ = E0 Men+ since although the log 10 [Men+J term
in (8d) becomes zero under these conditions the activity coefficient
can be so much less than unity that the last term in (8d) becomes
appreciable. This is shown by Table 1 which list the variation of
fMen+ with [Men+] for some typical divalent metal sulphates and
chlorides, the values being taken from Latimer 4.

TABLE I
Activity Coefficients in Some Metal Salt Solutions
Activity Coefficients in solutions of molality
Salt

0.001

MgS04

0.002 0.005

0.01 o. 02 0.05

o. 1

0.40 0.32

0.18 0.13 0.088 0 . 06 ~


0.18 0.13 0.075 0.051
0.16 0.11 0.068 0.047
0. 15 0.11 0. 065 0.045

Niso 4
cuso 4 0.74
znso 4 o. 70

o.1f

o. 41 0.31 0. 21

0. 61

0.48

0.39

MgC1 2
FeC1 2 0.89
CuC1 2 0.89
ZnC1 2 0.88

0.86
0. 85
0.84

0.80
0.78
0.77

0.75
0. 72
0. 71

,,-

0.22

0.2

- 0. 56 0.53
0.70 0.62 0.58 0.55
0.66 o. 58 0. 52 0. 47
0. 64 0.56 0. 50 0. 45
110

0.5

1. 0

0.52 0.62
0.59 . 0.67
0. 42 0.43
o. ::!8 0.33

GASEOUS REDOCTION OF METALS FROM AQUEOUS SOLUTIONS


Thus if the activity coefficient of a 1 molal solution of divalent
metal sulphate is taken to be 0.05 one obtains
E
Me++

= Eo

Me

_ 2 .. 303 RT
l
2
. og100.05

which at 25c gives


EMe ++ = Eo Me + 0_. 04

(9)

The value of EMen+ clearly depends upon the value taken for the
standard electrode potential and in all that follows the values of
E0 Men+ will be those quoted by Latimer 4 except for those of Ni++
and co++. The standard electrode potentials used by Schaufelberger 1
were also those of Latimer except for Ni++ and co++. But whereas
Schaufelberger 1 used the values of Haring 5 6 for NiT+ and co++ the
present discussion uses the values for Ni++ and co++ taken from the
National Bureau of Standards circular 'Selected Values of Chemical
Thermodynamic Properties' . The values of these potentials for
comparison are;

Haring
N.B.S.
Latimer

Ni/Ni++

Co/co++

0. 231 volt
0.241
0.250

0.278 volt
0.267
0.277

(iii) THE USE OF EH AND EMen+ POTENTIALS


2

In Figure 1 the hydrogen potential at 25C and 1 and 100 Atmospheres is plotted against pH while the potential ordinates are also
used to plot the values of EMen+ in 1. O x 10 3 molal solutions.
' This figure is essentially the same as that used by Schaufelberger 1
hut differs from it in that the metal potential values have been
culculated taking activities into account. Because these corrections
tuw a much larger effect in 1. O molal solutions the slopes of the
r1:M 0 n+ iines from 1. O to 1. O x 10 3 molal are much flatter than
1uctt ated by Schaufelberger.

111

B.MEDDINGS AND V.N.MACKIW


-r

l0.11 10 r.ilolol

l OMol I

+ 10

!----------------------------<

.,,_
..I

0
~o

...;::

..I

=Ci------------------------1c.,++
0

Q.

-10+--r--.------,.------.---.----.---r--r---r---.--....----.---'
10
12
4

Fig.1.

pH-

H.vdrogen and Metal Electrode Potentials

It is thermodynamically possible to reduce a metal from solution


by hydrogen if the EH line in Figure 1 lies above the EMen+ line.
Thus it is clearly po~sible to reduce cupric ions to copper at all
normal pH values but to reduce cobaltous ions to cobalt it is
necessary to maintain a pH above 5.6 using 1 atmosphere of hydrogen
or above pH 4.4 with 100 atmospheres of hydrogen. It should be noted
that Figure 1 ignores effects such as hydrogen overvoltage . Although
the hydrogen overvoltage will have a definite influence on reduction
by hydrogen in heterogeneous systems the available information is too
sparse for a full thermodynamic treatment at present.
From Figure 1 and equation (8d) it is possible to calculate ,the
hydrogen pressure necessary to rduce a given metal at a given pH.
Thus if the potential S" required to reduce zn++, Fe++ and ca++
~ .
are taken to be 0.83, 0.50 and 0.47 volts respectively then it
follows that these are 0.47, 0. 14 and 0. 11 volts above the hydrogen
potential at pH 6 and 1 atmosphere at 25C. This difference in potential can be made up _by increasing the hydrogen pressure but the
pressure required is given. by
112

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS

Hence to reduce zn++ at pH= 6 where6E = 0.47, log 10pH = 15.9, a


hydrogen partial pressure of 8 x 10 15 atmospheres would2te required.
The corresponding pressures for reduction of Fe++ and Cd++ at a pH
of 6 would be 5.4 x 10 4 atmospheres and 5.2 x 10 3 atmospheres. If
the same reductions were attempted at pH ~ 10 it would be found that
EH~> EFe++ or Eca++ but that Ezn++ still exceeded EH by 0.283 volts
which corresponds to a required hydrogen pressure of f.1 x 10 8
atmospheres. Clearly it is most improbable that Zn .can ever be
reduced from aqueous solutions by hydrogen. It may be possible to
reduce ferrous solutions with hydrogen above pH 8.5 but it must be
emphasized that this is more of a thermodynamic possibility than a
practical chemical possibility. At pH 8.5 ferrous ions are unstable
with respect to hydrolysis and it will probably not be possible to
reduce such a system even by attempting to requce the small amount of
ferr ous ion in equilibrium with the hydroxide and so displacing the
equilibrium. When the ferrous ion concentration becomes very low
the potential of the Fe/Fe++ couple is rapidly raised by the term 0.02958 log 10 [Fe++J and this could only be counterbalanced by a
corresponding increase in the pressure term+ 0.02958 log 10PH 2 for
the hydrogen potential. The ~olubility product of ferrous hydroxide
is reported to be 1. 64 x 10- 14 and hence at pH 9 (pOH = 5) the ferrous
ion concentration must be about 10-9 molar, this would increase the
m tal potential to the point where about 105 - 10 6 atmospheres of
hydrogen would be required for reduction;
From a combination of Figure 1 and the expressions for EMen+
nnd EH
2

E Me -

0.05916

loglO~en+

0. 05916 pH+ 0.02958 loglOPH

IL 11-1 possible to derive a gr eat deal of information about the


t.lt1 1'01

dynamic p ssibili ty of various r educt ion reactions.

B.MEDDINGS AND V.N.MACKIW


The pH at which reduction is carried out determines the minimum
concentration of metallic ions which can be obtained since at
equilibrium EH must equal EMen+.
2

Thus at 1 atmosphere H2 and 25C

Since we are now considering equilibria in which the metal ion


concentration is low we may replace aMen+ by [Men+], rearrangement
then gives,
n. EoMen+

( 11)

n. pH - - - - -

0.05916

This identity is also given by Dobrochotov 8 who plots log 10 [Men+J


against pH down to values of log 10 [Men+J = -14. However in commercial
hydromettallurgy metal concentrations below 10- 2 molar are not o.f
significant interest and hence it is equally useful to solve (11) for
log 10 [Men+J = -2 and tabulate the pH values so obtained. There are
given in Table 2.
TABLE 2
Equilibrium pH Values at Men+

Ion
zn++
Fe++
Cd++
co++
Ni~ +

cu++
Ag++

Eo

.0.762
0.440
0.403
o..r267
0.241
-0.337
-0.799

10 2 Molar

pH
13.9
8.5
7.8
5.5
5. 1
-4.7
-11. 5

From Table 2 it is obvious that if a 1.0 molal solution is to be


reduced to 10 2 molal theri 1Jrecautions must be taken to prevent tho
n moles of H+ produced from decreas ing th pH b low t h qu i l j bri um
11 ti

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


values shown in Table 2. In general it is clear that metals with
negative E0 values can be reduced without such precautions but
that metals with positive E0 values will require the addition of
neutralizing agents to maintain the pH values above the equilibrium
va lue . This requirement becomes increasingly severe as E0 increases
and once again it seems that the reduction of zn++ to Zn by hydrogen from aqueous solutions is a practical impossibility. From the
practical viewpoint, solutions of uncomplexed zn++ or Fe++ ions in
so lutions of pH greater than 6 - 7 are not possible and the
mathematical solution of (11) does not correspond to chemical reality.
It is also possible to calculate how far hydrogen will reduce a
solution of metal salt.
Consider the following reaction at 25C and a constant hydrogen
pressure of 1 atmosphere.
Men+ + -n H

-'_ _ ____,

Let the fraction of Men+ reduced be x, then

(1-x)

x + n.x

Hence
E0 Me

0. 05916
-

log 10 [1-x] (neglecting activity coefficients)

u.nd

EH

= - 0.05916 log 10 (n.x)

trnd for equilibrium

re

0. 05916
M

. log 10 [1-x]

0.04916 log 10 nx

11

(12)

B.MEDDINGS AND V. N.MACKIW


Equation (12) can then be solved to give a quadratic in x from which
x can be determined. Solution of (12) for cu++ gives x = 1 which
agrees with Figure 1 and Table 2. Solution for the case of Ni++
proceeds as follows:
E0 Me

+ 0.241, n

Hence
0.241 - 0.02958 loglO (1-X)
Q.241
0.02958
..

8. 15

10 8 x 1. 41

.. 5.64x10 8

0. 05916 log 10 . 2x

log 10 (1-x) - 2 log 10 2x

(1-x)

loglO - 4x2

log 10 P

1-x
4x 2

x 2 +x-1= 0,

.. x=4.21x10- 5

This value of x corresponds to a hydrogen ion concentration of 8. 4 x


10- 5 i.e . a pH of 4.075 and from Figure 1 it will be seen that at
this pH the hydrogen potential is 0.-24 i.e. the same as E0 for
nickel . However, Schaufelberger and Roy 9 have shown that nickel can
be reduced by hydrogen from acid-sulphate systems down to a pH of
2.75. This is very probably another result of the influence of
activity coefficients, the activity coefficient of H2so 4 is markedly
low, being only 0.54 in 0.1 molar solutions 4, hence the production
of H+ ions by the reduction of Ni++ does not decrease the hydrogen
potential by the amount indica~ed by (8d). It is of interest that
Schaufelberger and Roy 1 9 also found that cobalt could be reduced
down to a pH of 3.3. This difference between pH 2.75 and 3.3
corresponds to a difference in hydrogen potential , f r om Figure 1, of
0.033 volt and this sl!Uuld then b'e equal to the difference between the E0 values for nickel and cobalt . The N.B . S. values 10 used in
this paper give a difference for E0 co - E0 Ni of 0.026 v. , the values
of Latimer4 give a difference of 0.027 v. while the values of
Haring5 6 give a difference of 0.048 v. Hence the r esult s . of
Schaufelberger and Roy9 seem to indicat e some i nacc uracy in t he
results of Haring. most pr obably i.n t he val ue of E0 for nick 1.
116

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


It is clear from Figure 1 that the driving force for reduction
i ncreases as the pH increases and on this basis the conditions for
reduction become more favourable as the solution becomes more
alkaline. However, above pH 7 the rate of hydrolysis of transition
metal cations becomes appreciable and in order to retain homogenous
solutions it is necessary to complex these ions. But the potentials
of the metal-bearing species in such complexed solutions are not
t hose shown in Figure 1, those were calculated on the basis of the
bare metal cations, Men+ (actually, hydrated Men+). As a result
Figure 1 must be modified to take account of the altered metal
potentials arising from the use of complexing agents.
For metals such as cobalt, nickel, copper and zinc, the simplest
method of both raising the pH and complexing the metal ions is the
addition of ammonia which causes the formation of the metal ammines
[Me(NH 3)x]++. This effect is best exemplified by the case of nickel.
The addition of ammonia to a solution of a nickel salt results in
t he sequential formation of ammines as follows;
Nl. ++ + NH 3

For each of these step-equilibria there is a corresponding


quilibrium constant given by:
f.Ni (NH ) ] ++
k

3 x

The standard electrode potential for the couple Ni 0 /f.Ni(NH 3)x]++


IH given by the sum of the standard electrode potentials for the two
11< po.ut
st ages

11 7

B.MEDDINGS AND V.N.MACKIW


Ni++ + 2. e
[Ni (NH 3 ) x] ++

x.NH 3 + Ni++

The first is merely the standard electrode potential for nickel


while the other is the standard electrode potential for the complex
[Ni(NH 3)x]++ and is derived from the equilibrium constant by the use
of
n. E. 23,060 = 2. 303 RT log 1ok
The equilibrium constant for the reaction
x. NH 3 + Ni++
is the product of all the equilibrium constants k 1, K2 . ... . kx.
The equilibrium constants for the Ni++ - NH 3 system have been determined by
Bjerrumll and are given below together with the standard electrode
potentials calculated for each ammine. To avoid confusion E0 Me
will always refer to the standard electrode potential of the metal Me
while xEoMe will refer to the standard electrode potential of the
arnmine [Me(NH 3)x] ++.

TABLE 3
Potentials f or the Nickel Ammines
E0 Ni "' 0. 241 v. 10
Equilibrium Constantll

..=

2.80
Log10K1
K2 = 2. 24

..
..

..
,,

..

XEoNi

'

1Eo
Ni =0.324v.
2EoNi
3Eo
Ni
4Eo .
N1
5EONi

K3 = 1. 73
K4 = 1..19
0.75
K5
0. 03
K5

F. N I

1111

0.390 v .
0. 441
0. 477
0. 499
0. 490

v.
v.
v.
v.

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


For any given ammoniacal nickel solution, in which all the
n ckel ammine species are necessarily in equilibrium, there will be
t> nly one nickel ammine potential but the absolute value will depend
upon both the absolute nickel concentration and the ratio of ammonia
to nickel present in the solution.
The potential of a solution of any ammine, such as [Ni(NH 3) 4]++,
ls given by
aammine
xEoNi - O. 02958 log 10 - - -

Where aammine represents the activity of the ammine [Ni(NH 3)x]++ and
~H 3 represents the activity of ammonia in solution. Since no activity
oefficient data are available for the nickel ammines the usual
1Lpproximation of substituting concentrations for activities must be
made and hence
[Ni (NH 3) x] ++
ENi = E0 Ni - 0.02958 log 10

Fig.2.

Potential of Ni++ - NH 3

TA BLE 4
Concentrations and Potentials for the Ni/NH3 System at 25c
Total Nickel Concentration = 1. 0 Molal

NH 3/.Ni
Rat io
:-'"
0

o. 064.
0. 184
0. 50
1. 12
1. 50
1. 99
2. 52
2. 97
3. 56
4.01
4. 73
5. 02
6.22
15 . 91

Ammine Concentrations
Ni++
0. 940
0. 826
0. 573
0.233
0.166
0.040
0.010
0. 002
0

Potential

Ni (NH 3) ++ Ni(NH 3 ) 2++ Ni(NH 3) 3++ Ni(NH 3 ) 4 ++ Ni(NH 3 ) 5 ++ Ni (NH 3 ) 6++

..0. 059

o. 165
o. 361
0.466
0.393
0.251
0.115
0.047
0. 011
0.003
0

.
I

0. 001
0.009
0.063
0.256
0.372
0. 437
0.375
0. 258
0. 114
0.050
0. 009
0.003
0

0
0
0. 003
0.043
o. 109
0.234
0.379
0.438
0. 379
0. 270
0.108
0.057
0. 006
0

0
0.009
0.036
0. 110
0.215
0. 362
0. 418
0.356
0.278
0.092
0. 001

0
0. 002
0. 012
0.038
o. 126
0.235
0.429
0.494
0.461
0.085

0
0.001
0. 008
0.025
0. 099
0. 168
0.440
0. 913

XENi

0.242
0.242
0.248
0.260
0. 269
0.282
0.300
o. 318
0.346
0. 370
0. 411
0.460
0.497
0. 517

D'

~
~
;<:
~

~
~
H

GASEOUS REDUCTION CF METALS FROM AQUEOUS SOLUTIONS


Bjerrum1 1 has determined the variation of
ummine species with the overall NH 3/Ni ratio
i t is possible to calculate the potential of
~ol ution as ammonia is added.
These results
1 and shown in Figure 2.

concentration of each
and from these results
a 1 molal nickel
are tabulated in Table

Figure 2 also shows the hydrogen potential line, this being der ived from the observed pH of ammoniacal nickel solutions. From
t his figure it is clear that the greatest driving force for the
r eduction occurs at about an ammonia/nickel ratio of 2. 0 - 2. 5.
Thi s is in agreement with the findings of Wadsworth 12 that in the
hydrogen reduction of cobaltous sulphate solutions the maximum rate
of reduction occurs at a ratio of NH 3/co = 2, it is also the ratio
f N~ 3 /Nichosen in commercial hydrometallurgy.
Bjerrum has also determined the equilibrium constants for the
ca++ - NH 3 system and the same calculations of the change of metal
potential with ammonia/metal ratio have been carried out and tabulated
t n Tables 5 and 6. These results are very similar to those for the
Ni ++ - NH 3 system.
TABLE 5
Potentials for Cobaltous Ammines
E0 Co = 0. 267 v. lO

XEO

Equilibrium Constant
log 10K1

2.11

K2

.
,,
.

K3
K4

,,
Ks
,,

K6

lEo

Co
0.329 v.

1. 63

Co
2Eo Co

0.378 v.

1. 05

3Eo

Co

0.409 v.

0.76

4Eo

Co

0.431 v.

0.18

5Eo

Co = 0.436 v.
6Eoco = o. 418 v.

-0.62

Ll

TABLE 6
Concentrations and Potentials for the Co/NH 3 System at 25C
Total Copalt Concentration = 1. 0 Molal

NH 3/Co
Ratio

.....
~

t-:1

0.124
0.509
0.901
1. 520
2.571
2.879
3.643
3. 687
4.429
4.998
5. 288
5. 663
7 .042
8.428
10.466

.
co++
0.882
0.575
0.345
0.126
0.013
0.006
0.001
0.001
0

Ammine Concentration

Co(NH )++
3
0.113
" 0.350
0.444
0.380
0. 123
0.075
0.017
0. 016
. 0. 003
0.001
0

.
,

Co(NH 3) 2++
0.005
0.071
0.189
0.380
0.388
0.320
0.148
0.140
0.044
0.016
0.010
0.005
0. 001
0

Co(NH 3) 3++
0
0.004
o. 021
0.100
0.323
0.359
0.332
0.326
0. 198
0.114
0.084
0.057
0.016
0.006
0.002

Co(NH ) ++
3 4
0
0
0.001
0.014
0. 138
0.207
0. 381
0.389
0.455
0.414
0.377
0.326
0.180
0.106
0.560

Potential
Co(NH 3) 5
0
0
0
0.0005
o. 016
0.031
0.115
0.122
0.274
0.395
0.445
0.494
0.544
0.505
0.428

++

Co(NH 3) 6++
0
0
0
0
0
0.0007
0.0055
0.0061
0.026
0.060
0. 083
0.118
0.260
0.383
0.514

Eco

0.269
0.275
0.280
0.293
0.323
0.333
0.360
0.362
0.394
0.419
0.430
0.445
0.488
0. 519
0.550

C7:l

~t::l

zc:'l

Ul

~
<
z

:><:

GASEOUS REDUCTION OF METALS FRCJd AQUEOUS SOLUTIONS

In general the use of alkaline solutions for hydrogen reduction of


metals from such solutions is thermodynamically justifiable only
for those metals whose standard electrode potentials exceed 0.25
vo lt. The use of complexing agents to suppress hydrolysis above pH 7
or 8 can also prevent hydrogen reduction by increasing the metal
pot ential to too high a level. Thus if ethylenediamminetetracetic
was used to complex nickel salts in order to operate in alkaline
solutions it is clear from the Ni++ - EDTA equilibrium constant,
log 1 oK = 18.62, that the metal potential would be increased to 0.77 v.
and this is only exceeded by the hydrogen potential at a pH of 12
and a pressure of 100 At.
Also, the corrosive conditions encountered when using hydrogen
atmospheres under acid conditions in stainless steel equipment are
severe enough that in commercial practice alkaline conditions are
always preferred. The solutions are usually made alkaline with
ammonia. For metals with negative potentials, such as copper, the
use of a base such as ammonia is solely to counteract the acid
produced during reducticn so as to maintain non-corrosive conditions
wi thin the equipment,
From the foregoing thermodynamic considerations it is clear that
hydrogen reduction of metals such as cu++, Ni++ and co++ should
proceed at 25c under a pressure of one atmosphere of hydrogen. That
~ hese reactions do not proceed under these conditions shows that
'Lhermodynamic arguments cannot indicate how fast these reactions will
pr ceed and that kinetic factors must now be considered.
l{lNETICS

Consideration of the reaction


M

1 111

n+

+ - H2
2

(13)

bl Ol:i the equilibrium constant to be formulated as


(l.M

"M

n+

H+n
(14)

H22
J 2:1

B.MEDDINGS AND V.N.MACKIW


and the value of the K can be calculated from a knowledge of the
free energies of formation of the species involved or the heats of
formation and entropy values. This is thermodynamics.
The rate at which the reaction proceeds is given by the rate
constant k,
Rate

(some function of reactants)

k .

which is not calculable from equation (13).

( 15)

This is kinetics.

The value of the rate constant k is determined by experiment and


is expressed as a function of the factors and concentrations
involved in the rate-determining step. For the reaction summarized
in equation 13 the rate constant might be expressed by any of the
following;
d [Men+]
Rate
dt

d [H2]

dt

= k [Men+]

or = k [Men+]

[H2]

k [Men+] [H 2]
or
[H+]

or -=

k f.Men+J [H 2J .i
[tt+2J

In general
Rate = k [Men+] a

'
[H 2Jb

[H+J c . [Me]d

where the values of a, b, c, and d must be determined by experiment


and usually lie within the range 2 and may , less often, include
s uch fractional terms such a 1/2 or 3/2. Th hydr g n reduction
!2'1

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS

1Jf metal ions from solution to produce elemental metal often involves
hi ghly specific factors, such as surface area, which cannot be
v. neralised. For this reason some of the kinetic factors in the
hydr ogen reduction of Ag+, cu++, Ni++ and co++ species will be
ns idered separately and compared and contrasted.

Silver Salts
I t has been shown by Halpern and Webster 13 that aqueous solutions
of s ilver salts can be reduced to metallic silver by hydrogen. Using
11 l utions of silver acetate it was found that the rate was first
order with- respect to silver concentration. The rate with respect
Lo hydrogen was not determined but was assumed to be first order and
1.h r ate law was therefore expressed as

d [H2]
dt

-~

d [AgI]

- - = k [Ag ] LH 2]
dt

(16)

Th us e of Ag 1 indicates that univalent silver is involved in the


t'LLte determining step but the precise nature of the species involved
I 11 not specified. When using acetate solutions the possible Ag(I)
11poc ies are Ag+, AgOAC and Ag(OAC) 2 -. Halpern and Webster 13 found
I.lint t he rate of reduction to silver was increased by the addition
of sodi um acetate but not by the addition of the very weakly dis11noiate d acetic acid. They therefore conclude that acetate ions
111 reased the rate by making the acetate complexing of the Ag+ ions
mpr complete , the AgOAC and/or Ag(OAC) 2- species being more readily
rt ducibl e than Ag+.

Copper Sal ts
Like t he Ag+ ion, cupric ions have a negative electrode potential
J 0. 33 7 v. ) and should therefore be reducible by hydrogen in bot!}
11 ldi and al kaline solutions.
'l'h
111 11 11

hydrogen reduction of cupric salts in aqueous solutions has


Ht udi ed by a number of workers.1, i 4 , 15 , 16

rn and Macgregor 15 studied the reduction of cupric


11r11oh l rnt j n p rchloric acid so lut i on for the overall reaction
Il l I p

125

B.MEDDINGS AND V.N.MACKIW

and derived a rate expression given by

dt
and deduced the following reaction mechanism:
cu++ + H

(17)

( 18)
2cu+ - - - - . Cu + Cu++
''"1

(19)

In this scheme (19) is rapid while (17) and (18) are rate controlling.
This reaction path and rate equation is thus able to account for the
observed fact that the reduction slows down very rapidly and virtually
ceases at little more than 50% reduction even although thermodynamic
considerations, as given in the preceding section, clearly indicate
that the equilibrium position corresponds essentially to 100%
reduction . These workers also studied the reduction of cupric sulphate in sulphuric acid solution and obtained a rate expression of
the same form as above but in this case the reduction was both more
rapid and more complete. Addition of sodium sulphate to the system
pushed the equilibrium position even further towards complete reaction
and it was concluded that the ' hydrogen ions produced in the reduction
were being bound up by the sulphate ions to form the weakly dissociated
bisulphate ion and this therefore retarded the back reaction of
equation (17). In these acidic,,conditions it wa~ also found that tjJ.e rate of reaction was 'independent of the amount of copper powder in
_s uspension in the system and the reduction was therefore a
homogeneous reduction as opposed to a heterogeneous one.
The reduction of acidic cupric sulphate sol utions has al so bee n
described by Schaufelberger 1 in some det ail. No rate expression was
126

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


derived but it was determined that the rate of reduction was increased
by an increase in temperature, hydrogen partial pressure and ammonium
s ulphate concentration and was independent of the available surface
are a of copper powder. These results are in full qualitative
agr eement with those of Halpern and Macgregor described above.
The reduction of cupric sulphate solutions by hydrogen under
ammoniacal conditions has been described by Evans, Romanchuk and
Mackiw 16 . Under these conditions various insoluble species are
precipitated from the reacting solution at various times and the
whole system becomes too complex for detailed kinetic interpretation.
However the effects of certain variables on the rate of reduction and
fi nal equilibrium position can be studied in isolation.
As in the acidic system the rate of reduction was increased by an
increase in temperature and in sulphate salt concentration, ammonium
8Ulphate in this case. It was also found that after an initial fast
leduction, which reduced approximately half the copper, the rate of
reaction slowed considerably and became independent of the copper
oncentration and proportional to the hydrogen pressure.
The amount of ammonia added to the system also influenced the
rate of reduction. High NH 3/Cu molar ratios i.e. 2. 5 - 3. 0/1 resulted
In slow reductions and high equilibrium concentrations of copper in
11olution. Ratios of NH 3 /Cu below 2/1 resulted in faster reductions
nnd complete reduction of the cupric ions to copper metal. However
1.11 most marked difference between the acidic and ammoniacal systems
WO.R that in the ammoniacal solutions the rate of reduction was
1c lerated by an increase in the amount of copper powder in suspen11lon i n the solution. It was determined that nucleation of copper
p1uLicles also occurred throughout the reduction and that therefore
I. hi r duct ion was both homogeneous and hetergeneous.
Cobalt Salts
'l'ho hydrog en reduction of cobaltous salt solutions has been prac1.1111 d on a co mmercial scal e by several companies i.e. National Lead
f 1111111m1,y f F'r d ricktown, Calera Mining Company of Garfield, Sherritt
I 111do11 M n R of Canada and the Freeport Nickel Company. The
11111 1'1 t I m1H or Liu :;o companies hav
been s ur veyed in a paper by Benz
1111d M1u:l1I w17 hut. v ry 11LL1
d Lo. lod w rk appears to have been done
I ;17

B.MEDDINGS AND V.N.MACKIW


on this system. It is usually assumed that it is so similar to the
nickel/hydrogen reduction system that the results for nickel can be
applied to cobalt with very little inaccuracy. Some detailed kinetic
studies of the cobalt system have, however, been carried out by
Wadsworth and co-workers1 2 18 and Courtney 19 , apart from the work
reported by Schaufelberger and Roy 9 .

:::
I' I

Kaneko and Wadsworth 12 reported on the hydrogen reduction of


ammoniacal cobaltous sulphate solutions using colloidal graphite
as a nucleation catalyst. It is found experimentally that in
ammoniacal solutions nickel and cobalt will only reduce at a solid
surface. This pre-requisite for reduction means that either some
solid nuclei must be provided in the reaction system at the start
or else the system must be caused to nucleate by some chemical means.
Kaneko and Wadsworth used colloidal graphite as a source of nuclei
for reduction. These workers found that the maximum rate of
reduction occurred when a molar ratio of NH 3/co = 2/1 was employed
and therefore concluded that the cobaltous diammine sulphate complex
was the most readily reducible. The rate of reduction in any
experiment was constant at a constant hydrogen pressure, indicating
a reaction of zero order with respect to cobalt. The dependence of
rate on hydrogen pressure was unusual and was an exponential function
of the gas pressure.
The rate of reduction was proportional to the amount of colloidal
graphite present and a reaction mechanism was proposed in which a
cobalt species was absorbed on to a phenolic site on the graphite
and then transformed to a cobalt species containing two added
hydrogeij atoms which decomposed to cobalt metal.
The results of Kaneko and Wadsworth were commented upon by
Courtney 19 .
Wimber and Wadsworth 18 later studied the hydrogen reduction of
I
cobalt sulphate - alflmonium acetate solution and obtained results .
indicating catalysis of the reduction reaction by stainless steel
and pyrex glass.

12R

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


Nickel Salts

Nickel is produced in larger quantities and greater purity than


any other metal by hydrogen reduction and a great deal of knowledge
has been accumulated about this reaction. l, 9, 20, 21, 22, 23, 24, 25, 26
Schaufelberger and Roy have shown 9 that nickel can be reduced
from slightly acidic solutions but because of the relatively unf avourable thermodynamics of acid reduction for nickel most work on
th is reduction has been carried out in ammoniacal solution and it is
this work which will now be discussed.
The hydrogen reduction of nickel, like that of cobalt, is strictly
heterogeneous and solid nuclei must be present for the reduction and
deposition of nickel from solution. Accordingly it is possible to
differentiate three cases of reduction according to whether the
initial reduction solution contains pre-existing nickel nuclei, nuclei
ot her than nickel, or is induced to nucleate during the reduction
reaction. The latter case will be dealt with later and because the
econd case reverts to the first case at some time during the
re duction/deposition reaction only the first case will be considered
i n some detail.
When an ammoniacal solution of a nickel salt is reduced by
hydrogen in the presence of nickel nuclei the overall reaction is
Ni surface

Ni + 2NH 4 + (x-2)NH 3

(20)

fl'l'o m equation (20) it is clear that i f x 2 the molar ratio NH 3/Ni


wtll not change during reduction. In normal commercial practice x
18 maintained in the range 1.9 - 2.2 and in all subsequent discussion
iL will be assumed that x ~ 2 unless otherwise specified.
If s uch ammoniacal solutions are reduced by hydrogen in the
se nce of nickel powder suspended in the agitated solution it is
l'ound that the reaction is strictly heterogeneous, the reduced metal
. 'lH Ing deposited upon the surface of the pre-existing nickel seed.
'I'll( !tin tics of this relatively simple reduction-deposition reaction
wl ll n w b con id r ed.
Lil'

1. n

B.MEDDINGS AND V.N.MACKIW


The rate of reduction is found to be independent of the nickel
concentration, at least down to 85% reduction, and first order with
respect to both the hydrogen pressure and nickel seed surface area. 22
The reaction is therefore fitted by the rate expression
Rate

d [Ni++]

dt

where A is the surface area of the nickel seed and [H 2] is a measure


of the hydrogen concentration, the partial pressure of hydrogen in
the gas phase being almost invariably used, while k is the specific
rate constant.
The variation of k with temperature has also been determined 27
and the apparent activation energy found to be 10. 2 K.cal. mole- 1.
It should be noted that in this determination of the activation
energy, EA the variation of Henry's Law constant, relating hs:drogen
partial pressure to the solubility of hydrogen, with temperature was
taken into account. The data for this correction were taken from the
paper by Pray, Schweickert and Minnich 46 on the solubilities of
hydrogen, oxygen, nitrogen and helium in water .
From the rate expression
Rate
it follows that if the reduction is carried out within a closed and
isolated autoclave the pressure within the autoclave will fall
according to the above expression. Integration of that expression
gives
k.A.t.

= 2.303

log 10 [H 2 Jiniti~} - 2.303 log 10 [H 2]


.
~

where [H 2]initialis the starting hydrogen pressure. From this a


plot of log 1 [H 2] against t will will give a straight line whose
slope, log ~ ]/t, gives kA./2.303 from which k can be evaluated
i f A is known.

10 2

1:10

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


It therefore follows that no further analysis of the kinetic data
is possible until the surface area, A, of the nickel seed is known.
Unfortunately the area A refers to the surface area on which
deposition actually occurs and this is not necessarily the same as
t he surface area of the nickel powder measured by some arbitrary
means such as the B.E.T. method . It is in this context that the
di scovery of a class of catalysts which accelerate the hydrogen
reduction of nickel offers hope of circumventing the difficulty of
evaluating A.
It has been found that anthraquinone, and a number of structurally
related compounds, will accelerate the rate of hydrogen reduction of
ni ckel solutions. This is true for both soluti,ans containing nickel
nuclei and solutions containing 'foreign' nuclei. Moreover, it is
fo und that the deposition of reduced nickel on to both classes of
nuclei is remarkably uniform, the whole surface of the nuclei being
evenly covered w~h freshly reduced nickel.
No Anthrqquinone

Anthraquinone Used

Fig. 4. Nickel
Powder x 160

Fig. 3. Nickel
Powder x 160

Th difference oetween the deposition of nickel in the absence or


pr sence of anthraquinone is shown by Figures 3 and 4. Figure 3
Nh ws a cross-section through a nickel particle which has been
t>roduced by approximately forty successive depositions of nickel by
h,ydrogen reduction of nickel solutions. Figure 4 shows a nickel
vu. icle produc ed in the same manner but with anthraquinone present
111111 ng each deposition. Clearly the use of anthraquinone has led
to a v ry uniform rate of nickel deposition over the whole surface of
I.ht pu.rt icl shown in Figure 4. Rather surprisingly the use of

1:11

B.MEDDINGS AND V.N.MACKIW


anthraquinone does not alter either the rate expression or the
appaxent activation energy, which remains at 10.2 Kcal. mole- 1 even
although the rate may be increased by a factor of twenty when using
anthraquinone.
On the basis of these observations the following crude physical
model may be proposed. It is assumed that the surface of the nickel
nuclei is covered by a strongly adsorbed layer of nickel ions (or
nickel ammine ions) and that these are so strongly adsorbed that the
number of adsorbed ions only begins to fall off appreciably when
more than 85% of the nickel ions originally in solution has been
reduced. Under these conditions the rate of reduction will be
governed by the rate, at which hydrogen molecules can be adsorbed
on to the nickel nuclei surface and this will be proportional to
both the hydrogen partial pressure and the nickel nuclei surface
area. The form of the rate equation fits such a model. It is
further assumed that the actual rate determining step, probably
between an adsorbed nickel ion and an adsorbed hydrogen molecule,
can only take place at certain unusually reactive sites i.e. at
'active centres'. The ~unction of the anthraquinone is then to
increase the number of these active centres and thereby increase the
observed rate of reduction without altering either the rate expression
or the apparent energy of activation.
In conformity with this model it has been found that the acceleration in reduction rate brought about by anthraquinone varies with the
nickel powder used as nuclei. For some very fine and highly active
nickel powders the ratio 'Rate with anthraquinone' /'Rate without
anthr~quinone' can be as low as 1.04 while the same ratio can be as
high as twenty for some relatively coarse and poorly active nickel
powders. If it is assumed tha~ 100% of the exposed nickel surface
acts as 'active centres' when sufficient anthraquinone is present
then the above range of ratios would indicate that the fraction of
total surface covered by active centres is about 95% for very fine,
active powders and decreases to dbout 5% for poorly active powders.
Such a range seems reasonable.
The above model also suggests that
nickel powder there should be a range
which the rate of reductio~ increases
is increased but that above a certain

for a relatively inactive


of anthraquinone additions in
as the anthraquinone addition
level, corresponding to

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS

complete activation 6f the nickel nuclei surface, the addition of


f urther anthraquinone should fail to increase the reduction rate.
Such a situation has been found .
Figures 5 and 6 show some typical reduction plots obtained by
fo llowing the fall in pressure within a sealed autoclave and plotting
log 10 [H 2] against time. Figure 6 then shows the Arrhenius plot
obtained from these results ,

...

- - o- - - - -- -

-0 -

...
0

TIME

Fi g.5 .

3
(M INI.I TU ) -

Reduction Rates at Various Temperatures

..

'

.- - ...,...
. -.. -----::. , - - - - - - ----,1

Pig. IL

Anh n111

PloL ~A 10.2 K.cal. mole- 1


1:1:1 .

B.MEDDINGS AND V.N.MACKIW


Figures 7 and 8 show some results obtained with a special nickel
powder as nuclei. Figure 7 shows the rate plots, log 10 [H 2] versus
t, for fixed conditions of temperature, initial hydrogen pressure,
initial nickel ion concentration and weight of nickel powder nuclei
while the amount of anthraquinone catalyst is varied. From this
plot the slopes are used to obtain values of k.A. which are then
plotted against the anthraquinone addition in Figure 8. By making
the assumption that the maximum value of k.A. in Figure 8 corresponds
to a value of A equal to the exposed surface area of the nickel
powder it is -now possible to proceed further with the kinetic analysis
of the hydrogen reduction of nickel solutions.

...t

I.
1 5 0---~~,o~--~.-o~---'--.-0---.0----,00

TIME ( oMutu) -

Fig.7.

Reduction Curves at Various Anthraquinone Concentrations

Under normal conditions even' the value of k. A. at the plateau in


Figure 8 would be insufficient to fix the effective . surfa~e - area since
the area measured by a method such as the B.E.T. ads9rption would
probably include fine pores and yjl.Pillaries which would not be
available to nickel amliiine ions in solution. However the nickel
powder used _as nuclei in the experiments represented in Figure 7 and 8 consisted of the very smooth, spherical powder whose cross-section
is shown in Figure 4. The particle size in this powder was very
uniform and the average diameter was 198 microns (1.98 x 10- 2 cm.)
giving a surface area of 34: 0 cm. 2 g. -l .- It seems improbable from
the appearance of this powder i n Fi gur e 4 that t he 'Anrfuc roug hn SH'

GASEOUS REDUCTION CF METALS FROM AQUEOUS SOLUTIONS


correction factor would be large enough to alter significantly this
value of the surface area per gram of powder. The amount of this
powder used as nuclei was 100 g, per litre and the amount of nickel
nuclei surface in the solution was therefore 3.4 cm. 2 per c.c.
However, in the presence of anthraquinone the stainless steel
interior surfaces of an autoclave are uniformly plated with nickel
during a reduction and the total surface area for reduction and
deposition of nickel in the experiments shown in Figures 7 and 8
was therefore 3.9 cm. 2 per c.c. of nickel solution .

I
2
-0 4
-06
ANTHRAQUINONE CONCENTRATION (gm /

Fig.8.

.os
LITRE)-

Values of k.A Versus Anthraquinone Concentration

Initial [Ni++] = 30 gpl


Initial lli2]

100 gpl Spherical Nickel Powder

= 300 psi Temperature = 350F

From Figure 8th~ maximum value of k.A. was 3. 4 x 10- 3 seconds- 1


mid if A = 3.9 cm. -l it follows that k = 0.872 x 10- 3 cm. seconds- 1 .
From the relationship
Rate
llows that the r ate of r eaction of hydrogen molecules, under
I. ht HLaudar d conditions of 1 gm. mol e of hydrogen per c.c . and 1
3 gm. moles per
1111. :.'. oi: nic k l surfac e p r c. c ., would be 0. 872 x 10 <

ta r

110

ond.

Expr $Sing this rate in molec ul es r eacting per

B.MEDDINGS AND V.N.MACKIW


cm. 2 of surface per second the rate is 5.25 x 10 20 molecules per
second . Making the assumption that one surface nickel atom can act
as a site for the adsorption of one hydrogen molecule the number of
adsorption sites per square centimetre of nickel surface is 2.07
x 10 15 since the radius of the nickel atom is 1.24 x 10-8 cm. The
rate of reaction is therefore 2.54 x 105 molecules site- 1 second- 1.
By using equation (178) in Chapter IV and equation (114) in Chapter
VII of Galsstone, Laidler and Eyring's 'Theorgy of Rate Processes' 30
the above specific rate constant can be equated to

e .

RT
Nh

By substituting k 1 = 2.54 x 10 5 , EA= 10.2 K. cal. mole- 1 and


T = 450K (350F) the entropy of activation can be evaluated and is
found to bes= 13.9 e.u.
The dependence of the specific rate constant on temperature can
be expressed in three ways, two of which are equivalent.
-EA
k

= Ax . e RT

where Ax = Frequency factor

-EA

where Z

P.Z. e RT

P
6 s#

e .

RT
Nh

e R

-EA
e RT

Collision number
'Probability' factor

where 6s;o! = Entropy of activation

These relationships can b~ used to discuss the effect of


anthraquinone on the hydrogen reduction of nickel solutions. Using
frequency factors, the rates derived from Figure 8 for the reductions
without anthraquinone and w1th sufficient anthraquinone to give the
maximum rate yield frequency factors of 5. 78 x 10 2 and 0. 93 x 10 2
respectively. The corresponding probability factors are 1.06 x 10- 4
and 1. 7 x 10- 3 while if the effective surface area, A, is assumed to
1:lfl

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS

be constant irrespective of anthraquinone concentration then the


entropies of activation are -19.5 e.u. and -13.9 e.u.
The acceleration of the rate of nickel reduction by anthraquinone
can therefore be correlated either to an increase in the Probability
factor : using the collisional theory of chemical kinetics, or by an
increase in the entropy of activation using the absolute theory of
reaction rates. The former correlation can be interpreted as an
alternative formulation of the crude physical model in which
anthraquinone is envisaged as rendering the whole of the nickel seed
sur face active for reduction-deposition of nickel. The correlation
of rate with entropy of activation is, however, deficient in this
i nstance in that there is no immediately obvious reason why
i ncreasing amounts of anthraquinone should increase 6 si up to a
cer tain level and then cease t o have any f urther effect upon the
activation entropy. It seems therefore that the most satis factory
explanation of the effect of anthraquinone is that it increases the
effective surface area of the nickel seed surface . How this increase
in effective surface is achieved is not clear from the data at
present available. , From Figur e 8 the maximum rate of reduction occurs
at 0. 030 gpl of anthraquinone for 3.9 x _10 3 cm. 2/ litre of seed
sur f ace . The amount of seed surface is therefore only about 0. 5
s q. Angstrom per anthraquinone molecules and it therefore seems unlikely that adsorption of the anthraquinone on to the seed surface
is responsible for the catalysis . Adsorption of anthraquinone could
account for the catalysis if the area of seed surface was increased
to give about 50 sq. Angstroms per catalyst molecule but it seems
unlikely that the observed surface of the very smooth nickel spheres
used as seed is only one per cent of the true surface area. It should
1ilso be noted that O. 030 gpl of anthraquinone for 3. 9 x 103 cm. 2/litre
o seed surface corresponds to about ten anthraquinone molecules per
1td1:1orption site. However one of the most remarkable aspects of the
t:LLt.alytic effect of anthraquinone is that it appears to be without
1 ff ct on the hydrogen reduction of cobalt solutions .
Apart from
l,h possibility of a very specific interaction between anthraquinone
mid nickel the s implest e~planation would be that a cobalt surface
111 nl ways 100% effective for the reduction deposition of cobalt .
'l'hlH 8 ms rather unli kely even i f only because hydrogen reduction
o l' <:obalt sol utions is appreciably s l ower than that for nickel
1 o l 11t.lorn1.
It s ms probable that an el ucidation of the detailed
11111 0 111111 11111 lly whic h 1L11(. brnqui non , and si milar catal ysts , accelerate
11l olu 1 n d11 ci t.lo 1111 wo u1cl b< moH L UH! ful.

1:17

B.MEDDINGS AND V.N.MACKIW


It has already been stressed that the hydrogen reduction of nickel
is heterogeneous and proceeds at a solid surface, this solid surface
being nickel powder in normal commercial production of nickel metal
by hydrogen reduction. However, ammoniacal solutions of nickel can
be reduced by hydrogen if the solutions can be induced to nucleate,
the reduction then proceeding heterogeneously on these nuclei. One
of the simplest and most efficient nucleating agents for the
nucleation of nickel solutions is ferrous sulphate and the kinetics
of this system have been described by Mackiw, Lin and Kunda. 22
When ferrous sulphate is used as the nucleating agent and hydrogen
reduction of nickel solutions carried out it is found that the
reduction curves i.e . gpl Ni++ versus time, are parabolic in shape.
This was explained by these workers on the hypothesis that during
these reductions the rate of formation of nuclei was constant and the
rate of nickel reduction was proportional to the number of nuclei
present.
Thus
dN

dt
where N

-d [Ni++]
and

dt

number of nuclei present.


and

-d [Ni++]
dt

Integration of this and substitution of [Ni++] initial at t

O gives

and a plot of Ni++ vel"sus t wid therefore be parabolic, as is found


by experiment.
Studies of various nucleating agents for the reduction of nickel
solutions have been carried out by Schaufelberger and Courtney25
who conclude, 'nucleation . of the Ni (II) - H2 react ion usually se ms
to involve a mixed salt of Ni (Ip and the r educing agent wit h onl y
138

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


transient nucleation effectiveness , and the elucidation of the
nuc l eation mechanisms will be difficult' . The over-riding difficulty
in any study of kinetics of the nucleation of the Ni (II) - H2
reduction lies in the difficulty of differentiating between reduction
of nickel to form nuclei and the reduction of nickel on to existing
nuclei. Until an adequate experimental technique is 'developed very
f ew definite conclusions can be drawn about the nucleation of
ammoniacal nickel solutions under hydrogen pressure .
In contract to the heterogeneous nature of ammoniacal reductions,
it has been found1, 9 that acidic solutions of nickel salts can be
r educed by hydrogen without the necessity of adding specific
nucleating agents to induce the formation of nuclei. This may
indicate that the reluctant nucleation of ammoniacal nickel solutions
is related to the lower concentration of uncomplexed Ni++ ions in
these solutions compared to acidic solutions . However the extreme
cor rosion problems encountered with acidic solutions under hydrogen
press ure make commercial operation under these conditions much less
favourable than the use of ammoniacal solutions even though special
nucl eation techniques must be employed.
So far the hydrogen reduction of aqueous solutions to precipitate
metallic powders has been considered. In the next section the use of
gaseous reducing gases other than hydrogen and the recovery of
reduced metallic compounds other than elemental metal will be
ons idered .
Reducing Gases Other Than Hydrogen and Indirect Metal Production
For all practical purposes the only alternatives to . hydrogen on a
tiommerc ial scale are carbon monoxide and sulphur dioxide . Of these
I t, is normally easier to produce carbon monoxide pure, or mixed with
ot h r gase s , normally hydrogen and nitrogen. One of the first
pr esses outlined for the recovery of metal values from solution by
1' ( ducing gases under pressure was described in patents issued to
M< 111 l r, Schlect and Schubardt in 1927 and assigned to
l .G.Farbi ni ndustr i e. 31 Thi s patent claimed that silver, copper,
11 I c:k 1, co balt and zinc could be precipitated from solution as the
111 ( f. n l H by water gas or car bon monoxide or hydrogen or mixtures thereof
1111d < r
ndition of l evat ed t emperat ure and pressure. While it is
hi hl y doub t ul t ho; zin was v r r duc ed by this method or nickel
1:I!)

B.MEDDINGS AND V.N.MACKIW


or cobalt metal produced by gases containing carbon monoxide the
patent examples do exemplify the reduction of silver and copper
solutions, to metal, by the use of water gas under 130 atmospheres
and various temperatures.
The use of carbon monoxide, and the kinetics involved, in tne
reduction of sil:ver and copper solutions have been described b;'
F.Pawlek and his co-workers. 32 33 These workers 33 found that the
use of solutions buffered by 5% ammonium acetate resulted in a much
faster reduction and lower activation energy than the use of unbuffered solutions for the carbon monoxide reduction of silver
sulphate solutions. However the form of the rate expression was
unchanged and in each case was
- d [Ag]

dt

k .

[Ag+]2 . [co]
-EA

=A

[Ag+J2 . [co] . e R

where A 12.8 x 10 5 for unbuffered solutions;


6.02 x 10 4 for
1
buffered solutions. EA = 14. 1 K cal. mole- for unbuffered solutions;
9.3 kcal. mole- 1 for buffered solutions.
Unfortunately a lack of knowledge of the solubility of carbon
monoxide in the silver solutions employed precludes any estimations
of any of the more fundamental equantities associated with the rate
constant k, activation energy EA. and rate of reaction. All that
can be done is to ctesignate the ' numerical value of k after
specifying [Ag+], usually.in gm. moles per litre, and [co] usually
as the partial pressure in atmospheres, from this the rate is then
known for all other combinations .!ff silver concentration, gas
pressure and temperature. It should be noted that there is no
indication that Pawlek, Bauch and Plieth 33 found the rate of reaction.
to be increased by the silver metal produced, the reaction is
therefore presumably a true homogeneous reduction as was the case
for the hydrogen reduction of silver solutions as studied by Halpern
and Webster 13 .

MO

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


Pawlek, Bauch and Plieth 33 also studied the use of carbon
monoxide to reduce solutions of cupric sulphate but found that
technical difficulties, such as the deposition of cupric oxide and
copper deposits and corrosion of agitators, compelled them to use
dilute unbuffered solutions of CuS0 4 . Under these conditions they
found the rate expression to be
-33.500
-d [cu]

dt

= 2. 56 x

10 13 .

[cu] 2 [co] . e

RT

where - d[Cu]/dt represents the rate of decrease of ionic copper in


solution and [cu] represents the concentration of ionic copper
in solution. Again the solubility of CO in the solutions is unknown
and hence k can only be expressed in arbitrary units. These workers
also found that reproducible kinetic measurements could not be
obtained unless the autoclave contained a sheet of etched copper
sheet throughout the reduction. This strongly suggests that, as in
the hydrogen reduction of ammoniacal Cuso 4 investigated by Evans et
alia, 16 the reduction of copper solutions by carbon monoxide is both
homogeneous and heterogeneous and that the variable results obtained
in the absence of the copper sheet were due to differences in the
rate and extent of nucleation in each experiment.
The relative rates of reduction of silver and copper solutions
using carbon monoxide, hydrogen and mixtures of the two gases were
also determined by Pawlek and his co-workers. 33 For both metals the
rate of reduction with carbon monoxide was only about 20% of the
rate with hydrogen but the decrease of rate with increasing CO
ontent in hydrogen was much more rapid for copper than for silver.
Thus a gas mixture containing 50 volume % CO still gave a rate of
r duction 85% of that for pure hydrogen in silver reduction but in
coppe r reduction a 50 volume % CO gas mixture gave a rate of
r duction only 22% of that with pure hydrogen. These observations
orre l ate well with some unpublished work carried out by Sherritt
Uord n Mines on the effect of carbon monoxide on the hydrogen
r du ti on of nickel solutions in which it was found that only 1% of
CO 1u t h hydro gen was sufficient to halt the reduction reaction
1111 11! r normul
omm r c i al conditions.

141

B.MEDDINGS AND V.N.MACKIW


Carbon monoxide can be utilized to obtain metallic nickel from
aqueous ammoniacal solutions of nickel salts by an indirect route
involving nickel tetracarbonyl. The very thorough researches of
Reppe on the chemistry of acetylene and carbon monoxide during the
Second World War have been summarized by Reppe. 34 In that summary,
and a subsequent reply to a criticism by Prof.Fierz-David, 35 Reppe
described the production of nickel tetracarbonyl by the action of
carbon monoxide under 100 atmospheres pressure, on highly ammoniacal
solutions and obtained quantitative conversion to Ni(C0) 4 according
to the following equation

Reppe was convinced that if the carbonyl formation proceeded via


metallic nickel as postulated by Prof.Fierz-David, 36 this intermediate
would have been noticed by himself and his co-workers. Since metallic
nickel was not detected in these experiments Reppe believed that the
conversion to Ni(C0) 4 did not involve the formation of metallic nickel
as an intermediate. Clearly this use of carbon monoxide to convert
solutions of nickelous salts to nickel tetracarbonyl could form the
basis of a method for winning nickel metal from solution by
decomposing the tetracarbonyl to metal and recycling the carbon
monoxide. Rather similar reactions with iron and cobalt salts were
also described by Reppe. 34 35 Both iron and cobalt.salts in aqueous
30% ammonia solutions react at so 0 c ~nder 100 At. of CO to give the
dihydride of the tetracarbonyl or its ammonium salt.

However the iron carbonyl decomposes in excess CO to yield hydrogen


and volatile iron pentacarbonyl while the cobalt analogue is stable
in alkaline solutions. The cobalt derivative can be recovered from
the reaction solution by cautious acidification followed by vacui.im
distillation to colleci;,.... the tetradarbonyl hydride in a col~ trap. ~ If
this condensate is aJlowed to warm up slowly it melts and below Gc
it decomposes to yield hydrogen and the solid dimer of cobalt
tetracarbonyl

142

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


Attention should also be drawn to a note published by Harkness and
Halpern 37 on the reduction of various cationic species by CO in
aqueous solution although these reductions do not result in metal
powder production.
Comparatively little work has been carried out on the use of
sulphur dioxide to reduce solutions of metals and it is not proposed
here to do more than indicate some typical patents utilizing so 2 as
a reducing gas.3a, 39 , 4 o, 4i, 4 2 Early unpublished work in the
laboratories of Sherritt Gordon showed that copper solutions could
be completely reduced to the metal by so 2 from ammoniacal solutions
of controlled composition. 23
In some cases the reduction of solutions of some metal species
results in the formation of insoluble oxide or hydroxide species of
lower valencies. Solutions of hexavalent molybdenum and tungsten
exhibit this form of reduction and it is possible that a hydrometallurgical process for the separation and recovery ot these metals
can be based upon this reaction. It was first shown by Schaufelberger
and his co-workers of Chemical Construction Company that hydrogen
would reduce solutions of hexavalent uranium and pentavalent vanadium.
Some studies on the kinetics of these reduction were subsequently
published by Forward and Halpern 43 and O'Brien, Forward and Halpern. 44
In both these cases the reduction was heterogeneous and the rate of
re duction was proportional to the amount of heterogeneous catalyst
ndded, this catalyst being nickel powder. The uranium solutions
ontained the uranium carbonate complex U0 2 (co 3) 34 - and were buffered
with sodium carbonate . The vanadium solutions contained sodium
111 tavanadate and were also buffered by sodium carbonate.
For both
m!tals the rate of reduction was independent of metal concentration
lu solution but while the reduction of uranium was first order with
11Hpect to hydrogen the vanadium reduction was proportional to the
11;1 m:i.re root of the hydrogen pressure. The rate expressions obtained
f'or these reductions were;
-9800

1111d

B.MEDDINGS AND V.N.MACKIW


-7850
-d [v]

dt

k2. [Ni]. [Hz]~. e RT

where [u 3o 8] and [v] are the concentrations of uranium and vanadium


in solution expressed as grams per litre of u3o8 or vanadium
respectively and [Ni] represents the surface area of the nickel
catalyst used in cm. 2 per litre. It should be noted that the surface
area of the nickel catalyst is the surface area as measured by the
B.E.T. method and the effective catalytic area may differ from this
so that any given value of the rate constants k 1 or k 2 may be valid
for only one sample of nickel powder.
The essential characteristic of these two reductions is that
relatively dilute leach solutions can be used and the metal values
obtained in a convenient and pure form by the simple application of
heat and hydrogen pressure . The uranium reduction yielded uranium
dioxide according to the following equation,

while the vanadium reduction was claimed to produce

v2o3

by;

The vanadium solutions were observed to change from colourless to a


deep reddish brown colour before precipitation occurred and it was
postulated that the red~ction of pentavalent vanadium to the trivalent
state proceeded in two stages, the first reduction from vV to yIV
being fast and accounting for the initial colour changes and the final
step, yIV to v 111 , being rate determining .
Following this ea~ ly work OJ) uranium and vanadium by Halpern and
his-co-workers the same type of hydrogen reduction was applied t o
molybdenum and tungsten solutions by Soviet workers . The use of
hydrogen reduction to precipitate molybdenum di oxide from sodi um
molybdate solutions has been studied by Lyapina and Zelikman. 45
These workers used solutions of sodium molybdenum made acid by the
addition of sulphuric ac i d in order to neut r alize t he alkali
produced dur ing t he reaction

GASEOUS REDUCTION CF METALS FROM AQUEOUS SOLUTIONS


Moo 42- + H 2

----~

Moo 2 + 20H -

No precise kinetic data were presented for the reaction but it was
shown that the reduction was accelerated by the addition of
molybdenum or molybdenum dioxide powder to the solution before
reduction. In other respects the rate of reduction varied with
conditions in a manner to be expected from the overall equation
given above, the rate was increased by an increase in hydrogen
pressure and temperature and by a decrease in the pH of the solution.
This work has been largely confirmed and extended to ammoniacal
solutions within the research laboratories of Sherritt Gordon Mines
and fuller details of the hydrogen reduction stage will be given in
a later paper. The molybdenum dioxide produced by this hydrometallurgical process can be reduced to metallic molybdenum very
conveniently by reduction with hydrogen in a furnace.
CONCLUSION
At the end of a survey of this nature it is possible to look back
and see that the most commonly used reducing gas in modern hydrometallurgy is hydrogen . The reasons for this are not difficult to
discover. Hydrogen is simple and cheap to produce on a large scale
and can be readily purified. Because of its molecular weight a
reduction can be carried out with less weight of reducing gas using
hydrogen than any other reducing gas and this is reflected in all the
mass transfer machinery within a plant designed to utilize gaseous
reduction of aqueous solutions to recover metals. From the chemical
po int of view the reaction products from aqueous hydrogen reductions
are either hydrogen or hydroxyl ions so that in effect there are no
by-products to contaminate the product or solution. Other reducing
gases such as CO or so 2 either produce gaseous or aqueous reaction
products which must be recovered or separated and in specific cases
11111y tend to contaminate the product.
The great advantage of gaseous reduction of metal solutions lies
111 t he high reaction rate with which solutions of even low metal
<:011tonts can be rapidly depleted of metal values and processed. The
1111 Lhod also offers the very attractive possibility of separating two
01 111or
m tals from the same soluqon if their reduction potentials
ILi'( 11 111'1' I I 11L l y far apart.
It should be noted that the pure nickel
14

B.MEDDINGS AND V.N.MACKIW


powder produced by Sherritt and which contains only 0.07% cobalt is
obtained by hydrogen reduction of solutions in which the cobalt
content is 2.5% of the total metal concentration ,
The future potential of gaseous reduction of aqueous solutions
in hydrometallurgical operations stems from its inherent amenability
to the methods of process control, methods of analysis, research and
modification which are employed in the chemical industry.
One of the major limitations of gaseous reduction of metal
solutions is that it is not applicable to all metals. The prospects
of obtaining iron powder by hydrogen reduction of aqueous ferrous or
ferric solutions seem remote. On the other hand metals such as
copper, silver, gold, platinum and palladium which can easily be
reduced to the metal by hydrogen reduction from aqueous solutions
are not produced in this manner to any significant extent at the
present. It is reasonable to expect that the application of gaseous
metal reduction will find more extensive application in this field
of metallurgy.
It is also probable that some of the more amphoteric transition
metals, such as molybdenum and tungsten, will be produced by processes
involving hydrometalturgy and hydrogen reduction techniques.

'

110

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS

ACKNOWLEDGEMENTS
The authors wish to express their appreciation to Mr.E.L.Brown,
President, and to the management of Sherritt Gordon Mines Limited
for permission to publish this paper.
The assistance of their colleagues in discussions, particular jy
Dr.D. J.I.Evans, is also gratefully acknowledged.
Thanks are also due to Professor H.H.Kellogg of Columbia
University for helpful discussions upon the potentials of solutions
of nickel ammines.

147

B.MEDDINGS AND V.N.MACKIW

REFERENCES
1.

F.A.Schaufelberger, Transactions A.I.M.E. 695 (May 1956).

2.

S.Glasstone,

3.

Ipatiev et.al., see Trans.A.I.M.E. 704, Ref.12 (May 1956).

4.

W.M.Latimer,
New York.

5.

Haring, Journal of Physical Chemistry 33, 161 (1929).

6.

Haring, Transactions of the Electrochemical Society 64, 235


(1934).

7.

A.G.Sharpe, Principles of Oxidation and Reduction , Royal


Institute of Chemistry Monograph (1959).

8.

H.N.Dobrochotov, Cvetnyie Metally. No.2 (1956).

9.

F.A.Schaufelberger and T.K. Roy, Trans.Inst.Mining Met . 64,


375 (1955).

10.

National Bureau of Standards, Circular 500,


Chemical Thermodynamic Properties .

11.

J.Bjerrum, Metal ammine formation in aqueous solutions


P.Haase and Son: Copenhagen (1957).

12.

T.M.Kaneko and M.E.WadswQrth, ].Phys.Chemistry 60, 457 (1956).

13.

J.Halpern and A.G.Webster, Journal of Phys.Chemistry 61, 1245


(1957).

14.

J.Halpern and R.F.Dakers, Can .J.Chem : 32, 969 (2954).

15.

J.Halpern and E.G.Macgregor, Trans.Met.Soc.A.I.M.E. 212, 244


(1958).

16.

D.Evans, S.Romanchuk and V.N.Mackiw, Can.Mi n .Met .Bu ll .


(July 1961) .

Thermodynamics for Chemists , New York (1947).

Oxidation Potentials

2nd Ed . Prentice--Hall Inc.,

Selected Values of

'

11111

GASEOUS REDUCTION OF METALS FROM AQUEOUS SOLUTIONS


17.

T.Benz and V.N.Mackiw, Application of Pressure Hydrometallurgy


to the Production of Metallic Cobalt, annual meeting of the
A.I.M.E. New York (February 1960). Extractive Metallurgy of
Copper, Nickel and Cobalt, A.I.M.E. International Symposium, 503
Interscience: New York.

18.

R.T.Wimber and M.E.Wadsworth.

19.

W.G.Courtney, ].Phys.Chem. 61, 693 (1957).

20.

Mackiw, Lin, Benoit and Benz, Nickel-Cobalt Separation at


Sherritt Gordon Mines, A.I.M.E. Meeting, New York (February
'1958).

21.

Cockburn, Loree and Haworth, The Production and Characteristics


of Chemically Precipitated Nickel Powder, proceedings of 13th
Annual Meeting, Metal Powder Association 1, 10, Chicago
(April 1957).

22.

Mackiw, Lin and Kunda, 'Reduction of Nickel by Hydrogen from


Ammoniacal Nickel Sulphate Solutions', J.Metals 209 (1957).

23.

Mackiw, Benz and Evans, The Practice and Potential of Pressure


Hydrometallurgy, XVIIIth I.U.P.A.C. Conference, Montreal
(1961). Subsequently published in condensed form in Chemie
Ingenieur Technik 34, No.6, 441-445 (June 1962).

24.

Sircar and Wiles., Trans .A.I.M.E. 218, 891 (1960).

25.

Courtney and Schaufelberger, Physical Chemistry of Process


Metallurgy, Part 2, Interscience: New York (1961).

26.

Tronev, Bondin and Khrenova, Academy of Sciences of U.S.S.R.


Transactions of Platinum Section, No.23, 123-131 (1949).

27.

Meddings, Kunda and Mackiw, 'The preparation of nickel-coated


powders', Powder Metallurgy, 775, Interscience Publishers
(1961).

28.

M dd i ng1;1 , Lund and Mackiw,


V11r/m1ulp11l 1Jl' r .

Di e Metal lurgie Nickel Ueberzogener

B.MEDDINGS AND V.N.MACKIW


29.

N.K.Adam,
(1956).

Physical Chemistry

30.

Glasstone, Laidler and Eyring,


McGraw-Hill: New York (1941).

31.

German Patent 463,913; British Patent 269,164; U.S.Patent


1,686,391.

32.

O.Knacke, F.Pawlek and E.Sussmuth, Zeit.f .Erzbergbau u.


Metallhuttenwessen 9 (1956).

33.

G.Bauch, F.Pawlek and K.Plieth, Erzmetall. 11, 520 (1958).

34.

W.Reppe, Experienta 5, 93 (1949).

35.

W.Reppe, Experienta 6, 68 (1950).

36.

H.E.Fierz-David, Experienta 5, 487 (1949).

37.

Harkness and Halpern, J.A.C.S. 83, 1258 (1961).

38.

Clark, U.S.Patent 1,503,229.

39.

Jumeau, U.S.Patent 930,967.

40.

Laist, U.S.Patent 1,461,918.

41.

Potter, U.S.Patent 894,902.

42.

Wilcox, U.S.Re.12,815.

43.

Forward and Halpern, Can.Inst.Min.Met. 56, 365 (1953).

44.

O'Brien, Forward and Halpern, Can.Inst.Min.Met. 56, 369 (1953).

45.

LYapina and Zelikman, Moscow Insti~ute of Non-Ferrous Metals


and Gold, Department of the Metallurgy of Rare Metals.

46.

Pray, Schwiechert and Minnich, Ind.Eng.Chem. 44, 1146 (1952).

106-109, Oxford University Press

Theory of Rate Processes ,

"
u

l hO

GASEOUS REDU<:I'ION OF METALS FROM AQUEOUS SOLUTIONS


47. F.A.Forward and V.N.Mackiw , 'Chemistry of the Ammonia
Pressure Process for Leaching Nickel, Cobalt and Copper from
Sherritt Gordon Sulphide Concentrates', J.Metals 457 (1955).

I I

PARTITION OF METALS BETWEEN LIQUID


METAL SOLUTIONS AND FUSED SALTS
by
Irving Johnson
Argonne National Laboratory, Argonne, Illinois

Abstract

The basic thermodynamic relations applicable to the partition of


a metal between a multicomponent liquid metal and a fused salt
mixture are presented. It is shown that the logarithm of the
dis tribution coefficient may be expressed as the sum of three terms:
1ne first term depends on the standard free energy change for thereaction, the second depends on the composition of the liquid metal,
and the third term depends on the composition of the fused salt.
The dependence of each of these terms -on the appropriate variables
is discussed and approximate computational methods suggested.
Examp les taken from the field of nuclear fuel reprocessing are used
to illustrate the use of the equations for practical process
rL ve lopment. The very large influence of the liquid metal composition on the partition of a rare-earth or an actinide metal between
n f used chloride mixture and a reductant-solvent liquid metal
~ lution is emphasized.
The use of partition studies for the
mas urement of the thermodynamic properties of dilute liquid metal
~ nlu tions is discussed.

1~

I.JOHNSON
The purification of 'spent' nuclear reactor fuel involves the
separation of fission product elements from uranium, plutonium,
their alloys or compounds. These fission products range from zinc
(at No.30) through gadolinium (at No.64) and hence elements from
all the families in the periodic table are involved. Nuclear fuels
are usually clad in metallic alloys during use; often it may not be
practical to remove these clads prior to purification and hence one
may al$O need to separate the cladding metals during the purif ication process. Thus the purification of nuclear reactor fuels
presents rather formidable separations problems. The large variety
of different elements which must be separated makes it unlikely
that a single step process will be adequate . It would be convenient
to develop a single type of process which by suitable control of
conditions or reagents could be made to achieve different separations.
Separation by partition between a liquid metal and a fused salt
offers the possibility of achieving this objective. The large
differences in stability of different fused salts offer the
possibility that substantial differences in partition will exist
for different substances. The large variety of fused salts and
liquid metals which might be used allow considerable latitude in
the choice of container materials . . It is the purpose of this paper
to present the thermodynamic basis for these processes and to
indicate the ways in which practical objectives may be achieved .
The partition of a metal between a liquid metal solvent and a
fused salt involves the existence of the metal in two or more
oxidation states. It is conventional to consider the oxidation
number of the metal when. dissolved in a liquid metal as zero but to
consider it some positive integar when the metal is dissolved in a
fused salt. The transfer of the metal from liquid metal phase to
the. fused salt phase invol v.es the oxidation (i.e. , loss of
electrons) of the metal. Since electrical neutrality is maintained,
this oxidation must be balanced by the reduction of some substance.
In the discussion which follows, the second reduction will be
assumed to involve .the transfh of a second metal from the salt
phase to the metal phase. The balanced oxidation-reduction
equilibrium may be written

MXm (fused salt) + -r R (liq.metal);: M (liq.metal) + - R~


r

(fused salt)
154

(1)

PARTITION OF METALS
where M is the metal whose partition is of principal interest and R
i s the 'reductant'. X is some univalent anion.
The practical systems which will be discussed generally involve
t wo or more 'M' metals. In such systems all possible pairs of
redox couples must be balanced if the system is in equilibrium.
This condition is obtained when each 'M' couple is in balance with
t he 'reductant' couple .
The thermodynamic equili~rium constant , Ka, is related to the
st andard free energy change for the reaction by the equation:
_ m

- RT In Ka - ~ AGf R~ - C:.Gf M~

( 2)

The distribution coefficient, D, of M between the fused salt and the


li quid metal sol vent is related to the equilibrium constant through
t he relation
m/ r
(3)

where

xM~

and xM are the atom fractions of M"m and M in the fused


"M~

rrnlt and liquid metal respectively,


11

ti vity coefficients and

8Rx

and 'YM are the corresponding

and aR the activities of RXr and R.

J1;quation 3 indicates that the distribution coefficient depends on a


l' r e ener gy factor , Ka-l a 'mass-action' factor, (~X /8R_)m/r, and
1u1

activi ty coefficient factor ,

yM/yM~'

While the significance of

I. I! s fi r st two factors have long been appreciated, the v'ery large


110L ntial influence of the activity coefficient factor has not been
1 11 l ly r eali zed.
Before discussing the very strong dependence of
ll 0 11 'YM and "M~ which has been found in c.ertain systems it is
1

c onv

nient to rewri t e Equation 3 in the following form :

log D -log Ka

+ (-

-r log

8R

log 'YM) - (- - log


r

8Rxr

(4)
1

I.JOHNSON

in which the terms depending on the free energy difference , the


liquid metal composition, and the fused salt composition have been
grouped together. In analogy with the practice in aqueous chemistry
we define the reaction potential,. reduction potential,. and
oxidation potential as follows:
PKa
PR
pO

- log Ka

(5)

m
- - log ~+ log 'YM
r
m
log
r
~~ + log 'YM~

( 6,,)
(7)

therefore,
log D = PKa

PR - pO

(8)

Equation 8 expresses the logarithm of the distribution coefficient


as the sum of three potentials which are independent of each other .
Thus the influence of the free energy difference, the liquid metal
composition, and the fused salt composition on the distribution
coefficient may be discussed separately.
REACTION POTENTIAL, PKa
The reaction potential depends on a difference of free energies
of formation :
m 6GfoRX
f

- 6Gf oMX

(9)

2.3 RT

For systems of practical importance PKa is neither a very large


positive or negative number. Therefore , the difference between the
free energy values is ..generally ' f rom one to two orders of magnitu'c:le
smaller than the individual free energy values. Thus accurate
computation of PKa requires very accurate free energy data.
The first st ep in the desi gn of a separation process based on a
l iqui d met al -fused ~alt part ition is t o choose t he salt sys t em and
the pri ncipal reducing agents. Consideration of such practical
lli6

PARTITION-OF METALS
matters as corrosion properties, melting point, cost, and stability
usually limits the choice of the salt system to a halide or possibly
a chalcogenide. If relatively low temperatures are desired, the
halides are limited to the chlorides, bromides, or iodides. The
available thermodynamic data are not extensive or good enough to
make a very critical choice between these three systems. The
chlorides were chosen for the examples to be considered in this
paper.
Bo

20

..

Rb,C1

~o
p,

Z
In

,,

Go

AQ

Bo

____

Cd

r~C
V

Pu

..'"
Z

~ -10

Go
Mo
Pd

..

z~

Cd

Sb

-20

..
G

'"
T<

-40

-oo~-----

Reaction Potentials for Uranium and Plutonium and Principal


Fission Product Elements. Reductants are Zinc and Magnesium.
Temperature 1000K

11' 1g.1.

tteaction potentials for fission product metals and the actinide


mntlils uranium and plutonium are shown in Figure 1. These
not ntials at 1000K were computed using the free energy data
oomp 1 d by Glassner. 1 The PK a values given are for zinc and for
fll tL n si um as the reduct ant.
Positive values of pKa would indicate
tH rti tion in favor of the salt phase while negative values indicate
imrl,1 tion in favor of the metal phase.
With zinc as the reductant,
~ood H po.rntioni; of uranium or plutonium from most of the 4-d
l. rn11H1Llon m< La) A ( xc pt zirconium) would be expected. This
1!i7

I.JOHNSON
'noble-metal leach' (if the fuel-fission product mixture is
initially in the salt phase) or 'salt oxidation' by the use of zinc
chloride (if the fuel-fission product mixture is initially in the
metal phase) might be considered the first step in a process. If
the salt phase from this first step is then contacted with a liquid
metal phase containing magnesium as the principal reductant,, a
separation of uranium from plutonium and the more active metals
would seem possible. However, the influence of the two other terms,
pR and pO, on the distribution coefficient must be studied before
the utility of the proposed process can be decided. It is unlikely
that the reduction or oxidation potential terms will be large
enough (either positive or negative) to change the phase to which a
metal will partition when the pKa values are greater than +5 or less
than -5. For values of PKa within this range, it is necessary to
evaluate the reduction and oxidation potentials before a reliable
prediction can be made of the phase to which the metal will
concentrate.
Since the reaction potential is proportional to a free energy
difference divided by the absolute temperature, a large dependence
on temperature is indicated. The free energy of formation of the
compounds of interest decreases with increasing temperature. This
temperature dependence is remarkably similar for compounds of the
same type. Thus, the free energy of formation of ionic chlorides
decrease by about 8 kilocalories per mole of chlorine between 500
and 1000K. The decrease ranges between about 6 and 9 kcal over
this temperature interval. Therefore, the difference between two
free energies of formation would be expected to be only slightly
dependent on temperature. This fact permits one to estimate PKa
values at a series of temperatures by the simple proportional
relation

(10)
,I

where PKa T and pKa T are the reaction potentials .at absolute
1

. 2

temperatures T1 and T2. The reaction potentials for UC1 3 and CeC1 3
when reduced with magnesium (to form U, Ce and MgC1 2) are -5.15
and 5. 70 at 1000K and -10,32 and 10. 97 at 500K, respectively, when
computed from the actual vaiues 2 of th e free ener gi es at the two
t mp ratur s .
15 13

PARTITION OF METALS
The inverse temperature dependence of the reaction potential
l eads to an increase in absolute value as the temperature is lowered.
This would be expected to spread out the distribution coefficients
11.nd generally lead to better separations. As will be noted below.
I.his improved separation is often almost completely counterbalanced
hy t he change in the reduction potential.

lllo:OUCTION POTENTIAL, pR
The reduction potential depends on the composition of the liquid
rnot al phase and the temperature. To investigate this dependence we
nons ider a liquid metal containing metals M, R and a sol vent metal
II at atom fractions xM' xR and x , respectively. The 'solvent'
8
111 tal S is chosen so that its compound SX
is of low enough relative
8
11(,abil ity so that essentially all S is present in the liquid metal
phase .
The dependence of the reduction potential on the liquid metal
nompos i tion would be too complex for calculation but for the fact
Llmt several simplifications are possible in systems of practical
Im portance. The concentration of M in the reductant-solvent system
111 gener ally low, being limited by its solubility to a few atom
t1111 ent. Thus, the effect of M on the activity of the reductant is
r 1111111 and to a very good approximation the activity of the reductant
1111 y be computed from the experimental data for the binary
1'11<luct ant-solvent system.
Th

dependence of the term log 'YM on the composition of the


tant-solvent system may be very large. Fortunately, when there
I 1~ large difference between the values of log 'YM in the two pure
1111 1.r h1 R and S, a simple relation may be used to estimate log 'YM in
111 I xt. ur
of the t wo metals . To a good first approximation we have
I 11 1111d t hat t he activity coefficient of a solute in a reductantrw I vo ut solut ion may be computed using the equation
11 !111

R
S
IOI{ 'Yu .. XR log 'Y + Xs l og 'Y
u
M
M

( 11)

I 11 will h 'Y R u.nd /' ar t h acti vity coefficients of M in pure liquid


M
M
It 11 1HI I, ro111H e ll v 1y.
11 ri Lu1i 1 11y1;t, mH t his equation tends to
1 1H

I.JOHNSON
under-estimate the effect of the metal in which M has the larger
activity coefficient. Thus when magnesium is added to zinc, the
activity coefficients of the actinide and rare-earth metals in the
solution are greater than would be computed using Equation 11. In
this case ~g is several orders of magnitude greater than yzn.
Darken 3 and Alcock and Richardson 4 independently suggested M
equations which are equivalent to Equation 11 with the addition of
a term (-~a;~x/2 . 3 RT) proportional to the excess free energy of
mixing ~G!~x of the reductant-solvent system. Since the excess free
energy of mixing of magnesium-zinc solutions is negative, the
correction would make log yM larger and closer to the values
actually observed. This correction would be the same for all
solutes in a given solvent system. The limited experimental data
available indicate that the deviations from Equation 11 depend in
part on the specific solute being considered. Nevertheless,
Equation 11 has proved to be extremely useful in making nearly
quantitative predictions of the effect of the reductant-solvent
composition on log yM. It must be used with care., keeping in mind
that errors as large as 0.5 in log yM may be made . In those few
cases where actual data on log yM as a function of reductant-solvent
composi,t ion are available for one solute. one may use the deviations
from Equation 11 as an approximate correction for other similar
solutes. It should be noted that Equation 11 is limited to
relatively low concentrations of M in the reductant-solvent systems.
We have generally limited the concentration of M to about 5 atom
percent .
The reduction potential may be represented by the approximate
equation

PR

m
s
r log ~ + log yM

(12)
,I

An example of the dependence of pR on the reductant concentration


is given in Figure 2. In the example shown, the reductant-soLvent
system exhibits negative deviations from ideal behavior and the
metal M has a negative. departure from ideality in the pure solvent S,
and a pos itive departure i n t he pure reductant R. .The r eduction
potential is seen to pass t hrough a mi ni mum value (at about
xii 0. l~). In th
XMIP
shown, pR is actually nogative over a
1(10

PARTITION OF MIITALS

-2

-3
0.2

0 .4

0.6

0 .8

1.0

l'lg.2.

Reduction Potential, pR =~log~+ log ?'M' as a Function

xR.

m/r = 1. 5 and R-S System Exhibits Negative Departures from


Ideal Behavior.

111'

l~nif icant range of reductant composition.


The existence of a
111l ni mum depends upon the ratio ~/')'~; if this ratio is unity or less

minimum will exist. The greater the difference between log '{
log '{. the greater the depth of the minimum. The more negative
lht value of log'{ . the larger the range of values of xR over which
11 1 is ne gative. Negative deviations from ideality in the reductant11 11 l v nt system t end to displace the minimum to larger values of xR
111ul to decrease the range of values of xR (if any) over which PR
11 n gat ive.
1111

11 11<1

'l'h o most significant result of this analysis is the indication


I lllLL it is possible for a dilute solution of a reductant in a liquid
1111 Llll s ol vent, which its elf is a poor reduct ant, to be a more
" t'1 ti ve reducing agent than the pure reductartt . As will be seen
lln l. w,. it is possi bl e to compensate for an unfavorable reaction
1111f.11nt1n.l (i.e., a l ar ge pos it i ve value for PKa) to and reduce very
l.1 hl c ompounds using a reduct ant which would not usually be

1'1 11,1111<1 r1 d Atr ng

n ugh .
101

I.JOHNSON
The temperature dependence of pR depends on the temperature
dependence of log yR and log yM. Activity coefficients in metallic
systems generally tend toward unity as the temperature increases.
Large differences in the temperature coefficient are found for
different systems; for typical reductant-solvent systems the effect
of temperature on log yR is gene:ally small. On the other hand, the
extensive data now available 5 on solutions of actinide metals in
liquid metals indicate large temperature coefficients for log yM.
Thus, the temperature dependence of pR is largely dependent on the
temperature dependence of log yM. (Since in practical cases a
separation of two solutes is under consideration, only the difference
between values of PR is significant. In this case the terms
involving log YR may cancel each other.) Some typical examples of
the temperature dependence of log yM will be given below when the
temperature dependence of the separation coefficient is discussed.
OXIDATION POTENTIAL, pO

The oxidation potential depends on the composition of the fused


salt phase and the temperature. In general the fused salt will
consist of a mixture of M~, RXr, and one or more other salts. The
additional salts that are chosen have a high enough stability so .
that they are not appreciably reduced by the reductant-solvent
system; i.e., the value of log D would be a large positive number
for the added salt. These additional salts are usually added to
lower the freezing point of the salt mixture. To compute the
oxidation potential requires knowledge of the activity of the
oxidant R~ and the activity coefficient of the oxidized form of the
metal, i.e., M~, in the salt mixture. A large amount of thermodynamic data on binary molten salt mixtures are available and for
many common oxidants activity data either is available or may be
closely estimated. On the other hand, values of log yMX for the
m

actinide halides and rare-earth halides (which are of particular


significance in the development ,9f methods for nuclear fuel
reprocessing) are not generally available. The data which are
available indicate that significant deviations from ideal behavior
may be fairly common. Values of y~ comparable to the extreme '

values of yM in liquid metals have not been reported .


The dilution of the oxidant by an inert salt will reduce the
activity of the oxidant. This will increase the value of pO and

162

PARTITION OF METALS
hence tend to shift the equilibrium of Equation 1 to the right. If
t he 'inert' salt forms a very stable complex with the oxidant, then
t he shift of the equilibrium will be significantly larger than
would be expected from the dilution alone . If the 'inert' salt also
f orms a stable complex with the oxidized form of the metal, then the
activity coefficient yM~ would be expected to be significantly less
t han unity; hence log

yM~

would be negative and the oxidation

potential would be decreased and might even become negative . This


would cause the equilibrium of Equation 1 to shift to the l eft. If
t he complexes formed have different formation constants a rather
omplex variation of the oxidation potential with salt composition
ould result due to the competition of the two substances for the
added anion.
A good example of the effect of salt composition on the dist r i buti on coefficient being large and unusual is found on the
dist ribution of uranium between liquid aluminum and molten aluminum
hloride-potassium chloride solutions. Moore, Morrey, and Voiland6
ound that the distribution coefficients (UC1 3 in fused salt/U in
li quid Al) passed through a minimum as the AlC1 3 content of the
fused salt was varied from xAlCl = 0.28 to 0.66 . The minimum
3

occurred near xAlCl = 0.5. They found a two order of magnitude


3
ohange in the distribution coefficient with most of the change
o cur ring between xAlCl = O. 44 and 0. 54. They were able to
3

<inanti tatively account for the changes by the assumption of a number


of equilibria in the fused salt. In a subsequent paper 7 , they
11howed that the same type of equilibria were applicable to the
dltttri bution of thorium, protactinium and plutonium between the same
Lw li quid phases. It is of interest to note that even in this
1yHtem where very large salt effects are found, due to deviation of
hoLh t he oxidant and metal chlorides from ideal behavior, the
t ff cts of the log yM term on the reduction potential are of
oornparable magnitude. Thus, for xAlCl = O. 28 one would estimate
I

101< D c 1. 34 on the assumption that both yAlCl and Yucl are unity
3
3
(~>Kn 4. 76 and pR = -2. 78 at 725C). The experimental value of
I OK D is 0. 85. At xAlC l = 0. 5, log D would be 1. 59 if no salt
3
1 r 11 t.H wor pr A nt; t he expe r i ment al value is -1.01. Thus, the
v111.v 11f.1i11108 KAl 14 ompound has led to a large decrease in yAlCl
3

I.JOHNSON
It should be noted that the large changes in yAlCl

in this system
3
on addition of KCl would be independent of the metal chloride (MClm)
as long as the concentration is not comparable to that of AlC1 3 .
Differences in the extent that KCl complexed the various MClm
compounds would lead to changes in pO and in the separations possible.
Thus if one wished to reduce the tendency of a metal to transfer to
the metal phase by the modification of the salt composition, it
would be necessary to add a salt which formed a strong complex with
the particular metal halide and which did not strongly complex the
other metal halides in the system.
THE DISTRIBUTION COEFFICIENT

Having discussed in the previous sections the composition and


temperature dependence of the reaction, reduction, and oxidation
potentials, we are now able to examine the dependence of the distribution coefficient on these same variables. The experimental work
in our laboratory has concentrated on the effects of changes in the
liquid metal composition. Therefore this aspect will be discussed
in most detail.
The appli~ation of Equation 8 to the computation of the distribution coefficient of uranium between liquid magnesium chloride and
magnesium-zinc solutions is illustrated in Table I. In this example
all of the necessary thermodynamic data are available or can be
estimated with good accuracy. The equilibrium reaction involved may
be written
UC1 3 (liq.MgC1 2)

zMg (Mg-Zn soln)

U (Mg-Zn Soln)
3

2 MgC1 2

(liq.)

(13)

The reaction potential involves the difference between the free


energies of formation ..-of ~ moles' of MgC1 2 and one mole of uc1 3 . we
2
have used the free energy data reported by Egan, McCoy and Bracken 2 .
for this computation. Their free energies are based on measurements
of the emf of high temperature solid electrolyte galvanic cells.
Their free energy value6 are in good agreement with the older
literature values in the case of MgC1 2 but differ significantly in
164

PARTITION OF METALS

TABLE I
Computation of Distribution of Uranium Between
Magnesium-Zinc Solutions and Liquid Magnesium
Chloride (800C)
pO

PR
XMg

PK a

log D

3 log
3
15
8Mg log 'Yu - 2 log 8MgC1 log 'Yuc1 Cale. Obs
2
2
3

0.0063 -4.45

4.60

-0.74

0.00

0. 00

-0.59 -0.51

0. 059

-4.45

2.92

-0.31

0. 00

0.00

- 1.84 -1.87

0.144

-4.45

2.03

0.26

0.00

0.00

-2. 16 -2. 08

0.290

-4.45

1. 24

1. 10

0.00

0.00

-2.11 -2.03

0. 46 5

-4. 45

0.87

2. 02

0. 00

0.00

-1. 56 -1. 66

0. 606

-4. 45

0.43

2.74

0.00

0.00

-1. 28 -1. 30

0.698

-4.45

0.30

3. 19

0.00

0.00

-0.96 -0.99

0. 779

-4.45

0.21

3.60

0.00

0.00

-0.64 -0.63

0 . 802

-4.45

0. 18

3.72

0.00

0.00

-0.55 -0.51

case of uc1 . The standard enthalpy of formation of UC1 3 at


3
derived from their high temperature cell data by the third law
111othod is abot ten kilocalories less negative than the value
rnported in 'Bulletin 500' 9 , thus cl'lf 0 298 = -203. 7 kcal/mole
1:11: an , et al. 2 ), -213 kcal/ mole (Bulletin 500). In view of the
1111 rtainties in the older data we have preferred the newer direct
Ii I l{h temperature free energy measurements .
t,h

:moK

Th reduction potential was computed from experimental data on


t.hn activities of magnesium and uranium in magnesium-zinc solutions.
11t1 nctivity data reported by TerpilowskylO were used to compute
:' IOR 8Mg Te rpilowsky's data are in good agreement with those
11 . The activity coefficient of
c norted by Eremenko and Lukashenko
1111t11 i um was comput ed from high temperature galvanic cell data on
t.lt1 uranium-zinc system12 and from unpublished solubility data 13
nr unin um in magnesi um-zinc so lutions. At 80o 0 c, between about
IO. wt. p i m11p;n Ri um and pur magnesium, pure solid uranium is the
11lu11111 l 11 qn'll 1br1 um with .th Anturated so lutions. Hence the

Hrn

I.JOHNSON
activity coefficients are the reciprocal of the atom fraction
solubilities. The maximum error which would have been made if
Equation 11 had been used to estimate log Yu would be about -0.3
(at XMg = 0. 4) .
The oxidation potential involves only log Yucl

in pure liquid
3

magnesium chloride since in these dilute solutions,

~gCl

1.

Experimental data are not available on the thermodynamics of the


UC1 3-MgC1 2 system. From examination of the data which have been
reported for uc1 3 in other salt systems, 14 we conclude that Yucl
in liquid MgC1 2 at 800c is approximately unity.

The calculated values of log D (Table I) are in excellent agree:


ment with the experimental values reported by Knighton and
steunenberg. 15 In this example the variation of log D with metal
solution composition which has been found to be typical of the
actinide and rare-earth metals by Knighton and Steunenberg 15 is
illustrated. A minimum in the distribution coefficient at relatively
low magnesium concentrations is found. The large, two to four orders
of magnitude effect of pR on the distribution coefficient is
observed. The effect of pR is to bring the distribution coefficient
closer to unity, i.e., the actual partition is much less in favor of
the metal phase than one would predict from PKa alone. In this
particular example, since log Yu is positive above about 10 atom pct
magnesium, the reducing power of the metallic solution is actually
less than would be predicted if Yu were assumed to be unity.
The distributions of cerium, plutonium, and uranium between
liquid magnesium chloride and magnesium-zinc solutions are compared
in Figure 3. These curves are based partially on the experimental
measurements of Knighton and Steunenberg 15 and partially on th,e
results of computations using Equation 8. At 80o 0 c the values of
PKa are 4.114, -1.478, and -4.447 for cerium. plutonium. and
uranium., respectively. The oxidation potential in these systems is
close to zero (aMgCl ~ 1 and yMCl '::'. 1). The results indicate
2
3
'
that the reduction potential has a large influence. ConsideratLon
of the activity of magnesium in zinc indicates that in all three
cases the log yM term is of significance. Thus at xM = 0.05,
- ~ log ~g = + 3.084 hence; i..f log yM were not signi~icant, log D
](16

PARTITIOO OF METALS

Fig.3.

Distribution of cerium, Plutonium and Uranium Between Liquid


Magnesium Chloride and Magnesium-Zinc Solutions, 800c

would be 7. 198, 1.606, and -1 . 363 for cerium, plutonium and uranium
r espectively. The values of log D given in Figure 3 are 0. 44, -1.27
and -1.80 for Ce, Pu, and U. These latter results are understandable
when it is noted that the logarithm of the activity coefficients of
Ce, Pu, and U in pure zinc are -6.18, -3.12, and -0 . 80, respectively.
The large leveling effect of the liquid metal solvent, zinc, in
bringing the distribution coefficients for Ce, Pu, and U closer
together in value is rather typical of the other B-subgroup lowmel ting metals when combined with magnesium as the reductant.
Approximate values for the re~uction potential in zinc, aluminum,
admium, and lead for magnesium at xMg = 0.05, computed using
t:quation 12 are given in Table II. The ideal solution value was
omputed assuming YMg = 1 and yM = 1. It is seen that both positive
und negative deviations from the ideal solution valae are found in
the case of uranium; thus. Mg-Zn and Mg-Pb are relatively poor
r ducti on solutions for uranium while Mg-Al and Mg-Cd are good
' duction solutions. All of the solvents listed are excellent
r duction solutions for cerium.
Vulu s for log D, f or the di stribution of uranium and cerium
iQu d mo.gn sium chloride and each magnesium-solvent

h1 Lw1on

Hl7

I.JOHNSON
solution (xMg = 0.05), are tabulated in Table III. A significant
difference in the separation which can be achieved with the different
solvents is shown. Thus reduction with magnesium-cadmium would give
the greatest separation and reduction with magnesium-lead the
smallest separation of uranium from cerium. The values in Table Ill
illustrate the large shifts in distribution coefficient values which
may be produced by changing the metallic solvent used to dilute the
reductant metal.
TABLE II

Reduction Potential for Magnesium Solutions


XMg

= 0.05, t = 800C
pR

"'ij;

Reductant-Solvent
Mg-Zn
Mg-Al
Mg-Cd
Mg-Pb
Ideal Solution

Uranium

Cerium

2.55
0.44
0.08
4.62
1. 95

-2.83
-3.71
-1. 70
-2.21
1. 95

TABLE III

Estimated Distribution Coefficients of Uranium and


Cerium Between Liquid MgC1 2 and Magnesium Solutions
XMg

= 0.05, t = 800C
log D

Reductant - Solvent Solution

Uranium

Cerium

Mg-Zn
Mg-Al
Mg-Cd
Mg-Pb
Ideal Solution

-1. 90

1. 28
0.40
2.41
1. 90
6. 06

'

-4. 01
-4.37
-0.17
-2.50

The effect of temperature on the partition between a liquid


,
metal and a fused salt is best discussed after the consideration of
separation coefficient.

168

PARTITION OF METALS
SEPARATION COEFFICIENT

The practical objective of the processes under consideration is


the separation of two substances by arranging conditions such that
the ratio of the two substances in the salt phase is different from
the ratio in the metal phase. One can define the separation
coefficient, Q, as:

(14)

Therefore in terms of the potentials for the two metals one may
write
(15)

In the case where a=b=m, i.e., where the two metals have the same
oxidation number (valence) in the salt phase., Equation 15 may be
written in the form

(16)

where 6Gf 0 A~ and 6Gf 0 8Xm are the free energies of formation of A~
and B~. YA and y8 the activity coefficients of A and B in the metal
phase and y~ and Ys~ the activity coefficients of A~ and B~ in
the fused salt phase. Thus the separation coefficient depends on a
fre e energy term, a term involving only the metallic solvent and a
term involving the fused salt. Althoygh the composition of neither
t he liquid metal nor the fused salt appears explicitly in
Equat ion 16 it must be remembered that the activity coefficients
are dependent on these compositions.
If Equation 11 is used to compute log YA and y8 , log Q will be
When

n l t n ar function of the atom fraction of the reductant.


1011

H
II

ls gr nt r than

og -~ the largest separation will occur as

Yu

11'10

I.JOHNSON
xR __, O. When the reverse is true, the largest separation will occur
as xR __, 1. For example , in the separation of plutonium and cerium

,.zn

by the use of a liquid magnesium-zinc solution log ~n = -3 . 06 and


,.Mg
3pu
log ~g = -3. 17 at soo 0 c. One should therefore expect the .
,.Pu

coefficient to become larger as xMg __, o. The separation factor i L


pure magnesium should be about 803 of the value approached in pure
zinc (since both distribution coefficients become extremely large
as xMg __, 0 it is inconvenient to measure values close to zero
magnesium concentration). Knighton and Steunenberg 15 report that
when magnesium-zinc solutions are used, the separation coefficients
of a rare-earth from plutonium become larger as xMg __, O. However,
theY. indicate the existence of a maximum in the value of Q at about
10 wt. pct magnesium in the case of the praseodymium-plutonium
separation. Such a maximum would indicate a deviation from Equation
11 for either or both of the activity coefficients (Pr or Pu) in
magnesium-zinc solutions. As discussed above in the section on the
reduction potential , estimates of log yM using Equation 11 tend to
be low. The composition of the reductant at which the maximum error
occurs is dependent on the specific solute (Ce, Pu, etc . ) being
considered , i.e., the deviations from Equation 11 are not symmetrical about xR = 0. 5 and differ from solute to solute. Such
deviations could lead to a maximum in the separation coefficient ,
~specially when the separation coefficient differs by only a small
amount in the two solvents . The separation coefficient for cerium
and plutonium 16 using magnesium-zinc solutions and liquid magnesium
chloride has a maximum value at xM = 0. 5 of about 63, compared to
a value of 41 computed using Equation 11 . This difference is typical
of the maximum errors which have been found when Equation 11 has
been used.
The influence of the solvent metal on the separation coeffi cient
may be seen by examination of the data (given in Table III)' for the
' between liquid magnesium chlorfde
distribution of uranium" and cerium
and binary metal systems using magnesium as the reductant. The Ce-U
separation coefficient varies from 6 x 10 6 for Mg-Cd to 1 x 10 2 0~
Mg-Pb. Thus , significant changes in the separation coefficient can
be expected when the metallic solvent is changed.
To effect large changes i n sepa~atio n coefficients by changing the
fus d alt compo i tion , it must b posRibl t mnr lc d l y uffoct t h
170

PARTITION OF METALS
ratio of the activity coefficients 'YAX and 'Ysx Knighton and
m
m
Steunenberg 16 have attempted to increase the separation factor of
plutonium and praseodymium by changes in the salt composition. While
the changes in the individual distribution coefficient were often
significant, their ratio was not greatly affected. Apparently
plutonium and praseodynium chlorides are similar in their complex
formation properties .
The effect of temperature on the separation coefficient depends
mainly on the effect 0f temperature on the difference in the reaction
potentials and the difference in the reduction potentials . The
temperature ranges of practical use are not large enough to cause
large changes in the oxidation potentials. Equation 10 indicates
that. pKa is inversely proportional to the absolute temperature
and therefore will be markedly affected by temperature. The terms
log 'YA and log 'YB in Equation 16 are both linear in the reciprocal
of the absolute temperature and are found in practice to have co efficients for the l/T term comparable in magnitude to the term
pKa T T2 in Equation 10. Thus , to a first approximation, the
effect of temperature on the separation coefficient is expected to
be small . The effect of temperature on each particular separation
coefficient must be examined separately. For example , changing the
t emperature from 1000 to 800K lowers the separation coefficient for
plutonium-uranium but raises the separation coefficient for ceriumplutonium. Since the temperature dependence of the activity
coefficients in the pure solvent differs from that in the pure
reductant metals (i . e., log 'Y~ and log 'Y~ in Equation 11), the effect
of temperature on the separation coefficient will vary with the
liquid metal composition. If the temperature dependence of the
act ivity coefficients of the solvents in the two pure liquid metals
is known, good predictions can be made of the effect of temperature
on the separation coefficient.
APPLICATI ON OF PARTITION STUDIES TO THERMODYNAMIC MEASUREMENTS
The ease which distribution coefficients can be measured to good
accuracy s uggests the i r use for the determination of thermodynamic
properties, Equati on 8 indicates that from knowledge of D and any two
f tho potentials, t he t hi rd can be computed. On the other hand, if
num rioo.l values of t wo of t he potential s are not known but the
poi, ntiuls o.r known tb ' be constant , the changes i n the value of the
71

I.JOHNSON
third potential can be computed .from measurements of the distribution
coefficient. This scheme has been particularly useful for the study
of the variation of the reduction potential with metal composition.
Since the activity of the reductant is usually known as a function
of reductant-solvent composition the variation of log yM with
composition can be studied. If Equation 8 is written in the form
log D +

m
;

log aR

= pKa

- pO

log yM

(17)

a plot of log D + ~ log aR as a function of xR will reflect the


variations of log yM with xR except for the constant term pKa - pO
(if the salt composition is maintained constant). If log yM is
known at some concentration of R, the value at other concentrations
can be computed. We have used this method to study the variation of
log yM for the actinide and rare-earth metals in magnesium-zinc
solutions with magnesium concentration. Independent measurements of
YM in pure zinc solutions (using high temperature galvanic cells)
allowed ' the determination of yM in pure magnesium and in various
magnesium-zinc solutions; this method only requires knowledge of D
as a function of xR, a single value of yM, and knowledge of aR as a
function of xR. It is based on the assumption that aR does not
depend upon the concentration of M in the metallic solution; which
is a valid assumption if xM is not comparable in magnitude with xR.
The variation of log YMX

with fused salt composition can be

studied in a similar fashion. For such studies the metal composition


is held constant and the salt composition is varied. If independent
measurements permit the evaluation of aRX , then yMX may be obtained
r
m
as a function of salt composition. Essentially such a study was
reported by Moore, Morrey and Voiland 6; they measured the distribution
of uranium between liquid aluminium and aluminium chloride-potassium
chloride solutions. Since data on the activity of A1C1 3 in the salt
solutions were not available, th~ attempted to estimate it by the
assumption of a number~of equilibria in the salt phase. Their
~
treatment emphasized the existence of complexes in the salt solutions.
This treatment is proper for salt systems, where one can obtain '
spectrographic evidence for the existence of complexes, but would not
be meaningful for metallic systems.

172

PARTITION OF METALS
The use of distribution studies for the determination of values
of the reaction potential and hence values of free energies depends
on the availability of accurate values of both pR and pO. Probably
pR and pO can be estimated well enough to indicate gross errors in
free energy values but such estimates do not seem particularly
attractive for accurate determinations.

17:1

I.JOHNSON
Symbols

;11:

M, A, B

Solute metal whose partition is of interest

Reductant metal in liquid metal phase, also gas


constant

Solvent metal in liquid metal phase

A univalent anion, chloride in most examples

m, r , s, a, b

Oxidation number of M, R, S, etc in fused salt


phase

XM, XR, etc.

Atom fraction of M, R, etc. in metal phase

Ka

Thermodynamic (activity) Equilibrium constant

Distribution coefficient (M in salt phase/M in


metal phase)

Separation coefficient

)llliii

Activity coefficient of M, R, etc. in metal


phase
Activity of M, R, etc. in metal phase
?Mxm YRXr etc.

Activity coefficient of MXm RXr in salt phase

{. ~

Activity coefficient of M in liquid R and S

8Rxr

Activity of RXr in salt phase

L'GfOMXm.

Standard, free energy of formation of MXm at


temp. TK
Reaction potential: - log Ka

pR

'

~Reduction potential:-~ log aR +log yM

pO

Oxidation potential: - ~ log aRX r

Absolute temperature , K

..

17'1

log -v.:v
.
, IYJAm

PARTITION OF METALS
ACKNOWLEDGEMENTS
The thermodynamic treatment of liquid metal - fused salt
partition presented in this paper originated from the need to
furnish a rational basis for process development work. Mr.James
B.Knighton and Dr . Robert K.Stuenenberg have worked closely with the
author. They have made all their experimental data available and
often extended the range of their experiments to obtain data
helpful in the testing of theoretical predictions. Their help was
invaluable and greatly appreciated. Finally, special thanks are due
to Dr.H.M.Feder and Professor Scott Wood for many valuable
discussions.
This study was performed under the auspices of the U.S.Atomic
Energy Commission.

17h

I.JOHNSON
REFERENCES

1.

Alvin Glassner, The Thermoche~ical Properties of the Oxides,


Fluorides and Chlorides to 2500K. Argonne National
Laboratory, Report ANL-5750.

2.

J.J.Egan, W.McCoy and J.Bracken, Thermodynamics of Nuclear


Materials, International Atomic Energy Agency, Vienna, Austria,
163 (1962).

3.

L.S.Darken, Jour.Amer.Chem.Soc. 72, 2909 (1950).

4.

C.B.Alco~k

5.

I.Johnson, Compounds of Interest in Nuclear Reactor Technology,


Met.Soc.A.I.M.E. IMD Special Report 13, 171 et seq. (1964).

6.

R.H.Moore, J.R.Morrey and E.E.Voiland, Jour.Phys.Chem. 67,


744 (1963).

7.

J.R.Morrey and R.H.Moore, Jour.Phys.Chem. 67, 748 (1963).

8.

I. S. Morozov and D. Ya Toptygin, Izv.Akad.Nauk.S.S.S.R., Otd.


Khim.Nauk 11, 1920-27 (1959).

9.

F.D.Rossini et al, Selected Values of Chemical Thermodynamic


Properties, NBS Bulletin 500, U.S.G.P.O. Washington (1952).

10.

J.Terpilowsky, Bull.Acad .Polon.Sci. 10, 221 (1962).

11.

V.N.Eremenko and G.M.Lukashenko, UKr.Khim.Zh. 28, 570 (1962).

12.

I.Johnson and H.M.Feder, reference 2, 319.

13.

A.E.Martin, Argonne

14.

M.H.Rand and O.Kubaschewski, The Thermochemical Properties of


Uranium Compounds, 19, Interscience Publishers, New York (1963).

15.

J.B.Knighton and R.K.Steunenberg, Trans. of the Amer. Nuclear


Soc. 5, 460 (1962).

and F.D.Richardson, Acta.Met. 6, 385 (1958).

...

Nationa.~

Laboratory, Private Communication .

170

PARTITION OF METALS
16.

J.B.Knighton and R.K.Steunenberg, Argonne National Laboratory,


Private Communications.

177

SULPHATE STABILITY AND THERMODYNAMIC PHASE


DIAGRAMS WITH PARTICULAR R.EFERENCE TO ROASTING
by
T. R. Ingraham
Head, Research Section, Extraction Metallurgy Division, Mines
Branch, Department of Mines and Technical Surveys, Ottawa, Canada

Abstract

New decomposition pressure data obtained using a corrosionresistant pyrex bellows mercury manometer are reviewed for the
cobalt, copper, iron, nickel and zinc sulphate systems. The data
are combined with published results for the metal-oxygen and metalsulphur systems to construct predominance area (log Pso vs log P0 )
and predominance volume (log

Ps02 vs log P02 vs 104/T) Jiagrams to

outline the boundaries of the stable phases under a variety of


conditions. The diagrams are used to explain several industrial
separations made by selective sulphation, and to suggest the
conditions under which other possible separations should be tried.

170

T.R.INGRAHAM
INTROOUCTION

Selective sulphation of metallic sulphides and oxides has many


attractive aspects when considered for the commercial sepan.tion of
the oxides or sulphates of one metal from those of another in a
mixture of oxides or sulphides. As generally applied, the required
metal is either selectively sulphated, or selectively retained as a
sulphate, while the other metallic compounds are converted to oxides.
A simple leaching of the cold calcine with water can often effect the
required separation of the water-soluble sulphate from the insoluble
oxide. This method has been used industrially for many years for the
separation of copper sulphate from iron oxide, when both metals occijr
together as sulphides (l). In a similar application, copper and
cobalt sulphates, together, may be selectively separated from other
metals in a mixed sulphide concentrateC 2). Under very carefully
controlled conditions it is possible to make a reasonably good
separation of cobalt sulphate from copper oxide( 3). Another recent
development is the application of the selective roasting process by
the St.Joseph Lead Co. to the separation of cadmium from zinc, lead
and iron ( 4). In each of the examples cited, not all of the factors
influencing the separation are understood. Although there have been
a number of extensive publications on the acquisition of data on the
thermal decomposition of sulphatesC 5 6 7), insufficient information
is available for the necessary calculations to determine the
feasibility of separati9ns. There are also obvious inconsistencies
in the available data which require checking. The outstanding contributions to the thoery of selective sulphation are: (1) the
compilation of data by Kelley(B), (2) the discussion of the advantages
of precise temperature control and gas composition for fluid bed
roasting by Stephens (1), and (3) the recent compilat'ion of sulphate
data by Kellogg (9)_

In this paper a description will be given of some of the experimental techniques used recent!~ to establish the thermodynamic
properties of cupri~ sulphate r 1o), cupric oxysulphateClO), cobaltbus
sulphate<ll), tricobalt tetroxide(ll), ferric sulpliate< 12 ), nickelous
sulphate< 12 >, zinc sulphate<l3) and zinc oxysulphate< 13 ). The

techniques include (a) the use o.f differential thermal analysis and
effluent gas analysis for identifying reactions and phase changes and
for estimating the enthalpies of the . phase changes, and (b) the use
of a corrosion-resistant pyrex-.bellows mercury manomet er for

mo

SULPHATE STABILITY AND THERMODYNAMIC PHASE DIAGRAMS


measuring the decomposition pressures, Predominance area
(log Pse vs log Po ) and predominance volume (log Pso vs log Po
2

vs 104 /T) diagrams are used to illustrate the thermodynamically


stable phases in the systems, to illustrate the effects of temperature
and gas composition changes, and to predict possible separations on
the assumption that all of the solid phases are present at unit
activity.
Because of the ready availability of the recent review of thermodynamic data on sulphates by Kellogg< 9 ), no attempt will be made in
this paper to compare our data with those of previous authors.
Prior to the preparation of this paper, all of the data, except those
for the nickel system, were made available for correlation by
Kel logg( 9 ). His correlations generally favour those obtained in this
study0 4 ).

IDENTIFICATION OF REACTIONS AND ESTIMATION


OF THE HEAT OF PHASE CHANGES
For many years the characteristic patterns of exothermic and
endothermic heat requirements, as detected by differential thermal
analysis, have been used for resolution of some of the changes
occur ring in minerals. In recent years, the technique has been
r efi ned and an instrument has been developed by the Robert L.Stone
Co. Ltd. for the combination of differential thermal analysis and
affl uent gas analysis simultaneously on the same sample. This combinat ion of techniques is particularly useful for the study of
s ulphate decompositions, because it permits the association of gas
vol ution with the appropriate endotherm. It also permits the ob~e rvation of phase changes and secondary reactions, even when these
ccur during decomposition. Three examples observed during the
course of this work were: the reversible a to ~transition in zinc
s ulphate< 13 ), which occurs in the midst of its decomposition range;
th irreve rsible phase change occurring late in the decomposition of
upric s ulphate or early in the decomposition of cupric oxysulphate (lO~
und the oxidation of tricobalt tetroxide to cobalt monoxide during
Lh d c mposition of cobaltous sulphate0 1 ). The frequency of occurn 11c;1 o f phns changes in t hese sys tems suggests that some of the
I L It of on l uUon i n available s ulphate data may be because of as
Y1 I. 1111!! l H oV'{ r d phns
hunl( H.
1111

T.R.INGRAHAM
During the course of this work it was observed that the DTA
patterns were highly reproducible from run to run, and that, by
alteration of the paper speed and amplification, it was possible to
draw peaks having areas of the order of 150 cm 2. This made it
possible to make interesting comparisons of the heats of the various
processes from the areas under the peaks. For example, by accepting
the established heat of transition of potassium sulphate and assuming
that the thermal conductivities of potassium sulphate and zinc
sulphate are identical, and then substituting potassium sulphate for
the usual alumina reference in DTA from the areas under the respective
curves, Ingraham and Kellogg< 13 ) made an estimate of the enthalpy of
a to ft transition for zinc sulphate which was in good agreement with '
an estimate made on the basis of the difference in slopes of a log p
versus reciprocal temperature relationship for the a form and for the
ft form. The limiting factors in this technique of estimation are
the validity of the assumption of identical thermal conductivities,
and the reproducibility of packing the DTA cell. The error due to
packing is probably less than 10%.
APPARATUS
The apparatus used in this program has been described in detail
in previous papers< 15 16 >. The essential features consist of a
resistance-wound tube furnace fitted with an inconel sleeve for
reducing temperature gradients, and a u-tube mercury manometer. The
temperature of the furnace can be maintained with 1. 5C at 950C
with a gradient of about 0.2C/cm. The exposed end of the mercury
manometer is fitted with a corrosion-resistant, flexible pyrex
bellows. The mercury column is deflected about 30 cm for a 1
atmosphere change in pressure. within the apparatus. The manometer
response is linear and the manometer is maintained at 100.0 0.5C
in a silicone oil bath to prevent condensation of the sulphur trioxide
liberated during decomposition. No effects attributable to thermal
diffusion have been observed ih1 this apparatus< 13 >. The open end of
the manometer is vented to a container in which the pressure is
maintained at 76.00 cm of Hg. When the readings of the sample
pressure are to be made, the air pressure in this container is alternately increased and decreased to minimize any errors caused by
sticking of the mercury column.

Ill :'.

SULPHATE STABILITY AND THERMODYNAMIC PHASE DIAGRAMS


PREPARATION OF SAMPLES
Al l materials used in the experiments were prepared from reagentgrade chemicals . The cobalt , copper, iron, nickel and zinc sulphates
were available as hydrates, and were dried for 24 hours at 400c
befor e use. The final traces of water were removed from the samples
in the pressure-measuring apparatus, where, just prior to the
beg inning of a run, tne sample temperature was increased for ten
mi nutes to a temperature at which incipient decomposition was evident.
The last traces of water and the evolved sulphur trioxide were expel led from the apparatus by heating just prior to sealing . The
presence of a small amount of the product of decomposition promoted
t he reversibility of the reaction.
The oxides and oxysulphates were prepared from the corresponding
s ul phates. The cupric oxysulphate was prepared by roasting cupric
s ulphate, with frequent rabbling, in a muffle furnace at 725C for
a period of 48 hours. Tricobalt tetroxide was prepared by roasting
cobaltous sulphate for 48 hours at 80o 0 c. The cobalt monoxide was
prepared by heating the tricobalt tetroxide under vacuum at soo 0 c for
10 hours. Nickelous oxide was prepared by decomposition of the
sulphate in air at 90o 0 c. The zinc oxysulphate was prepared by
intimately mixing stoichiometric amounts of zinc oxide and zinc
s ulphate over a period of 24 hours in a pebble mill, and then heattreating pellets of the mixture at 700C for 24 hours. All sample
compositions were confirmed by X-ray diffraction patterns.

EXPERIMENTAL PROCEDURE
The sulphate samples were placed in a platinum boat in the furnace.
As a catalyst, the platinum ensured the equilibration of the sulphur
tri oxide, sulphur dioxide and oxygen at each experimental temperature.
Re adings of the total pressure generated by these gases were made at
regular intervals to establish the equilibrium pressure. The
t emperature was then raised (often about 20C) and another series of
re adings made. The procedure was repeated with a decrease of
t emperature.
The time required to attain equilibrium depended on the compound
and varie d within the limits of 6 to 30 hours. Cobalt sulphate was
t he mo t difficult material to r e- f orm . It was found that nickel
sul phut , whi h wuH qual l y dif icult tor -form us ing a small
1tt~

T. R. INGRAHAM

interval of temperature decrease, could be rapidly re-formed using


a rapid temperature decrease of about 25c for a few minutes,
followed by readjustment to the more usual decrease of 10c. It
seems reasonable to believe that the protective nickel sulphate
coating spalls readily in a steep temperature gradient. This observation has interesting implications in suggesting that a
single-stage fluid bed, with the precise temperature control for
which it is noted, may not be the most effective apparatus for
sulphating a nickel mineral. The irregular temperature gradients
in a Wedge-type roaster may, in this instance, serve a useful
purpose .

RESULTS
The total decomposition pressures measured in the experiments
were resolved for the partial pressure of sulphur trioxide on the
basis of the argument by KelleyCl 7) that: a) the total pressure is
the sum of the partial pressures , b) the sulphur trioxide, sulphur
dioxide and oxygen are present throughout the app~ratus in their
equilibrium proportions at the sample temperature , and c) there has
been no reaction to alter the 2-to-1 ratio of sulphur dioxide to
oxygen in the system. These three assumptions may be combined to
develop a cubic equation. In this work, an IBM 1620 computer
program was used to obtain solutions of the equation.
For extrapolation to room temperature to check the reasonableness
of the values, the partial pressures of sulphur trioxide were incorporated, with the appropriate specific heat terms, in the type of
sigma equation suggested by KelleyC 18 ).
The results of the experiments with copper, cobalt, iron, nickel
and zinc sulphates are shown in Table 1 as constants for the
three-term equation,

6G 0 = a

-'
+ bT l ; g T + cT.

It will be noted that the table includes measurements for the


tricobalt tetroxide - cobalt monoxide system, which was studied
because of a major co.nflict in the existing dataC1 9, 20. 20.
Equations are given for both the a and /3 forms of zinc s ulphate .

L~-104

___

SULPHATE STABILITY AND THERMODYNAMIC PHASE DIAGRAMS

TABLE I
Free Energy Relationships
6G 0 =a + b T log T + cT

Reaction
CoSO4 ;:::: CoO + S0 3
Co 3o 4 ;:::: 3 coo + 1/2 o 2_
2 CUS04 ;:::: cuo CuS04 + S03 .
.uOcuso 4 ;:::: 2 CuO + S0 3

Range K

62,362* 15.000

-92.784

950-1250

36,153

-29.306

950-1250

50,945

3.519

-54.913

750-1000

49,914

3.324

-50. 121

750-1000

1/3 Fe 2(S0 4) 3 ;:::: 1/3 Fe 2o 3 + S0 3 44,178


N1S0 4 ;:::: NiO + S03
56, 250

-0.056

-42.314

800-1100

2.900

-53. 666

1000-1200

53,733

-45.210

800-1007

39,281

-30. 866

1007-1200

:1 ZnS0 3 (a) := ZnO _2ZnS0 4 + S03


:1 ZnS0 4 (,8) ~ ZnO 2ZnS0 4 + S0 3
1/2 ZnO 2ZnS0 4 := 3/2 ZnO + S0 3

110 3

;:::: so 2 + 1/2 02

57, 189

7.598

-65.629

800-1200

25,009

5.563

-40.522

700-1300

Alt hough not justified by the accuracy of the data, a sufficient


11umber of significant figures were retained to prevent the
acc umulation of truncation errors.

111

T. R. INGRAHAM

For convenience in rewriting the equations in terms of sulphur


dioxide and oxygen, a three-term free energy equation based on the
data of Evans and Wagman< 22 ) is included.
A much enhanced appreciation of the thermodynamically stable
phases in each of the sulphate systems can be obtained by preparing
the type of predominance area (log Pso vs log Po ) diagram suggested
2

for the lead-sulphur-oxygen system by Kellogg and Basu<2 3 ). The construction of these diagrams requires information on the free energies
for the metal - sulphur systems and for the metal-oxygen systems.
With this information, the free-energy change and corresponding
equilibrium constant are calculated for the various reactions in
the system at the temperature of interest. The essential features
of the calculation can be illustrated with two examples:
For the system:
(1)

K = Pso 2/Po 2 .

(2)

and

log

Pso

log K + log

Po

(3)

Equation 3 is linear when log

P80

is plotted against log

Po .
2

The slope of the line representing the thermodynamic boundary between


MS and M has the sign of the log Po term and a value corresponding
2

to the ratio of the multiple of the log P0 term to that of the log
2
P80 term; in this instance +1.
2

For the system:

'

MS0 4 ~ MO + sti2 + 1/2 o2


K

Pso 2

log

Pso

(4)

Po1/2
.
2
K - 1;2 1og

(5)

Po.

(6)

mo

SULPHATE STABILITY AND THERMODYNAMIC PHASE DIAGRAMS


In this instance the slope of the phase boundary is -1 /2 .
Since the log pressure versus reciprocal temperature relationship
is normally very close to linear over the temperature range in which
t he stability of the various metal-sulphur-oxygen compounds are of
i nterest, a third dimension may be added to make a predominance-volume
diagram, by plotting log Pso vs log Po vs 10 4/T.
2

Samples of each of these types of diagrams, and some of the


interesting information which may be elicited from them, will be
illustrated in Figures 1-5.
IOOO"K

+4

NIS

"'

Ni*

- 4

N IO

-12
-20

111 11{.1.

-16

-12

-4

Predominance area diagram for the nickel-sulphur-oxygen


system at lOOOK

Th

predominance area diagram for the nickel-sulphur-oxygen


Information for calculating
l,IH Ni-NiO sy~ tem was obtained from Wicks and Block <24 ). Information
!'or I.ho calc ulation of the Ni-S equilibrium was that reported by
1trn11 11Qv hlt ( 25). The area enclosed by the square between log Pso .
.v11L m, at 1000K, is shown in Figure 1.

.
I rtrnl -2 and log Po

= -1 and -2 represents gas compositions between


2

11,;, ttnd JO% t\ h f s ulphur dioxide and oxygen . It is evident from


t.111 1 (1 l1t1.t rnm Lhrtt n1, Ir 1 Rulphate woul d be the thermodynamically stable
ph11111 1tf. 1OOO"K f'or
ttH compowl t 1ons fro m 3% to 10% each of sulphur
II lox lrh 111111 11x.rno 11 . llow ov< 1, !'or it 1:1 n
ntn.i n ng only 1% each of

1117

T. R. INGRAHAM

sulphur dioxide and oxygen, nickel monoxide would be the stable


phase. It is clear, from the position of the univariant equilibrium
A at log Pso = -4.6 (Pso = 2.5 x 10- 5 atm) and log P0 = -15.3
2

(Po
2

=5

10-16 atm), that it is unlikely that either the sulphur

dioxide or the oxygen pressure would be sufficiently low under


normal roasting conditions for metallic nickel to form. By a
similar argument it can be determined, from the position of the
univariant equilibrium point C, that the direct conversion from
nickel monosulphide to nickelous sulphate is improbable at 1000K.
On the basis of the diagram it seems that the most probable
intermediate phases occurring in the conversion of NiS to NiS0 4 are
in those areas through which a horizontal line at log Pso = -2
(P80
2

= 1%)

would pass, e.g. Ni 3s 2 and NiO.


+

16~-----------.,-----,

10501<

FoS

- 16

Fe

FeO

-24'-L----L..-'-----'---~~-~~--~

-40

Fig.2.

-32

-24

-1 6

Predominance area diagram for the iron-sulphur-oxygen system


, 1 at 1050K

In Figure 2, the various thermodynamically stable phases are


illustrated at 1050K for the iron-sulphur-oxygen system. It is of
interest to note that this diagram has been constructed for the
temperature at which ~he _600 ton/day fluid bed roaster is operated
at th Int ernat i onal Ni ckel Company of Canada plant i n Sudbury f or
on ntrntos ( 26). Fr m t h di agram
t h\ ct ml ro11sting o pyrrh tit
11111

SULPHATE STABILITY AND THERMODYNAMIC PHASE DIAGRAMS


It is evident that, using the same type of argument illustrated for
t he nickel system, ferric s~lphate would not be a stable phase in
the roaster. Similarly, it is reasonable to expect that neither
metallic iron nor wiistite would be formed during the oxidation of
t he pyrrhotite. Since the roaster is automatically controlled by
magnetic susceptibility to produce a calcine containing 80% hematite
and 20% magnetite, the oxygen pressure is maintained in the roaster
at about 10- 10 atmospheres (log Po
-10). At equilibrium, contrary
2
26
to earlier reportsC ), trace amounts of oxygen could not exist in
the 13% sulphur dioxide off-gas.

J'lg. 3.

Predominance volume diagram for the iron-sulphur-oxygen


system for the temperature range 800-lOOOK

As ment ioned earlier in this paper, one of the basic principles


s ulphation roasting may involve the selective decomposition of a
prov1ously fo rmed sulphate by either a change in gas composition or
1L ' hange in tem
pe'r ature . The effects of gas composition on phase
11 t.ub1l 1ty have been illustrated in the two previous examples. In
l'l v. nr 3 th effects of t emperature change are shown in a prerlo111l111tn v lume diagram for t he iron-sulphur-oxygen system. At
10 - 4/ 1' 12."0 (800K), the stabl e phas e corresponding with gas
t~ o111uo11ll.lo 11 H b tw
n 1% and 10% oach of s ulphur diox ide and oxygen
111 11 r'l'lt 11 11 l1,Jh1Lt.1 . Who11 o. volum
rr sponding to t hese gas
I 11

11111

T. R. INGRAHAM

compositions is projected through the diagram to a temperature


corresponding to 10 4/T = 10.00 (1000K), it is evident that the
volume passes through the plane separating Fe 2o 3 and Fe 2 (S0 4) 3 at
about 900K. Thus, at temperatures above about 900K, ferric
sulphate is no longer a stable phase and it would disappear as
equilibrium was approached in a roasting operation.
Earlier in this paper, it was mentioned that sulphide concentrates
of copper and cobalt are treated industrially in a fluid bed roasting
operation at 950K to sulphate both the copper and the cobalt. The
roaster gas composition in this operation is 8% sulphur dioxide and
4% oxygen. From Figure 4, in which the predominance area diagram
for the Cu-S-0 system i~ superimposed on that for the Co-S-0 system
at 950K, it is evident that the particular gas composition used in
this operation lies well within the areas of stability of cupric
sulphate and cobaltous sulphate. At equilibrium, one would expect
a quantitative conversion of both metals to sulphates . The results
reported for the commercial operation are 97.0% soluble copper and
93.5% soluble cobalt(27) .

c"s

/
/
/
/

/
/
/
/
/
/

c"

8
/
/

Co

Fig.4.

Superimposed predominance area diagrams at 950K for


the cobalt-sulphur -oxygen system (solid lines) and.
the copper-sulphur-oxygen syst em (broke n lines).
Wit hin t he area A, Coso 4 and.CuO are the stable phases.
l!l()

SULPHATE STABILITY AND THERMODYNAMIC PHASE DIAGRAMS


Close examination of this diagram indicates that there exists a
range of gas compositions for which cupric oxide and cobaltous
sulphate are the stable phases. These are in the area marked A on
t he diagram. If control of the roasting operation could be
established at equilibrium in this area , it should be possible to
make a separation of the copper and oobalt as insoluble cupric
oxide and soluble cobaltous sulphate. This separation has been
r eported industrially in a Nichols-Herreshoff furnace(!) .
+. 5

~---------------.

0
0

POSSIBL E SEPARATIONS

BASED
-

ONLY

1000K

OF

SOz ANO 02 Coso

Fe-Cu

3 - 10

3 - 10

1 -10

1-10

I - 10

3 - IO

Ft -Co

AT

PRE SSURES
-t0 S02

Ft - NI

ON

METALS

e -Zn

1 -!0

1-10

Cu- Co

1 -2

1-2

Cu - Zn

1- 2

1- 2

I -Hi

,_,,,

.,,,. CoJ'Of

1-2

NI - Co
Ni - Zri

Zn -Co
N! -Cu

Fig .5.

- 10

LOG

,,,

-2

Section of predominance area diagram at lOOOK for the


cobalt, copper, iron, nickel and zinc systems,
illustrating the ranges of partial pressures of sulphur
dioxide and oxygen at which the various sulphate.
oxysulphate and oxide phases are stable.

Ju Table II , the data have been assembled for comparison of the


t y of t he various sulphate systems included in this paper.
l'li u l'r'rn - norgy equations shown in Table I were used to calculate
1.111 .( (J t1 111brium constants for each of the reactions at 1000K. This
l.t 1111H rn t.m wil.H A l t d b cause of it s clos e approximation to that
111 11 d 111 11uL11,y 1nduAir1o. I. 1rnlphatjon operations .
From the equilibrium
P o111 t. 11 11 t.11, LIH v1 111( 11 01 l r>V. PHo, w r
n. l c nl at d for log Po = Oand -10 .
11 t.1~bll

If) J

T. R. INGRAHAM
TABLE II

Values of Log

Pso

Calculated for each Reaction at

1000K for log

Po

0 and -10

log P80
2
Reaction
at log P0
2
CoS0 4

so 2 + 1/2 o2
Co 3o4 + 80 2 + 1/3 o2

CoO +

CoS0 4 ; 1/3

2 Cu80 4 ~ CuOCu80 + S0 2 + 1/2 0 2


4
CuOCuS0 4 ~ 2 CuO + so 2 + 1/2 o2
1/3 Fe 2(S0 4) 3 ~ Fe 2o3 + so + 1/2 02
2
Ni80 4 ~ NiO + S0 2 + 1/2 0 2
3 ZnS0 4 (a) ~ Zn02ZnS0 4 + 80 2 + 1/2 02
l/2(Zn02ZnS0 4)

3/2 ZnO + S0 2 + 1/2 02

See note, Table I.

'

=o

at log P0
2

-3.4438*

+1. 5562

-2.9449

+O. 3884

-1.6966

+3.3034

-2.3912

+2 .6088

-o. 6277

+4.3723

-2.7230

+2.2277

-2. 0960

+2.9040

-3.3940

+1. 6060

-10

SULPHA1E STABILITY AND THERMODYNAMIC PHASE DIAGRAMS


TABLE I II
Possible Separations at 1000K Based Only on
Partial Pressures of S0 2 and 0 2

Metals
Fe-Cu

3-10

3-10

Fe-Co

1-10

1-10

Fe-Ni

1-10

3-10

Fe-Zn

1-10

1-10

Cu-Co

1-2

1-2

Cu-Zn

1-2

1-2

Ni-Co

1-2

1-2

Ni-Zn

1-1. 5

1-1. 5

Zn-Co

*
*

Ni-Cu

..

%02

o/t.S02

Separations improbable on basis of partial pressures of so 2 and


o2 at 1000K.

10:1

T.R.INGRAHAM
Their linear relationships for each system are shown in Figure 5.
The shaded square corresponds with gas compositions in the range
1-10% each of sulphur dioxide and oxygen. For any specific system,
the area below the line represents the region of oxide or oxysulphate
stability and that above the line represents the area of stability of
the corresponding oxysulphate or normal sulphate.
Study of the diagram, on the assumption that the solid phases are
at unit activity and equilibrium can be approximated, indicates that
at l000K the sulphate separations shown in Table III might be feasible
at the suggested gas compositions .
Assessment of the possibility of making these separations at
other temperatures can be made by repeating the calculations using
the free-energy relationships shown in Table I.

ACKNOWLEDGEMENTS
The author is indebted to H.H.Kellogg of Columbia ' University, who
co-authored the zinc work reviewed in this paper, for much helpful
advice and guidance, and to R.A.Charlebois for most of the experimental work, P.Marier for the DTA and EGA experiments, and S.Kaiman
for the X-ray analyses.
104

SULPHATESTABILITY AND THERMODYNAMIC PHASE DIAGRAMS

REFERENCES
1.

F.M.Stephens, Jr. Chem.Eng.Progress 49, 455 (1953).

2.

L.F.Theys J.Metals 10, 46 (1958).

3.

'Cobalt', No.148, Amer.Chem.Soc.Monograph Series, 43, R.S. Young,


Ed . , Reinhold Publishing Corporation: New York (1960).

~.

R.E.Lund and D.E.Warnes Trans.of A.I. M.E. 209, 608 (1957) .

ri .

G.Marchal,

<J.

Wohler, Pliiddemann and Wohler, Dert.Chem .Ges. 1 , 703


(1908).

'{.

K. Friedrich and A.Blicke, Metallurgie 7, 323 (1910) .

II,

K. K. Kelley, United States Department of the Interior, Bureau


of Mines Bulletin, 406 (1937).

I,

].Chim.Phys~

H.H. Kellogg,

22 , 559 (1925).

'A Critical Review of Sulphation Equilibria' ,

Trans . of A.I.M.E. 230, 1622 (1964).


Ill .

T. R.Ingraham, 'Thermodynamics of the Thermal Decomposition of


Cupr ic Sulphate and Cupric Oxysulphate', Trans. of A.I.M.E.
233, 35 9 (1965) .

tt ,

T. R.I ngraham, Canadian Metallurgical Quarterly, 3, 221 (1964) .

I '!.

'!'. R. I ngraham, unpublished results.

1 1,

'I'. It. Ingraham and H.H.Kellogg, Trans. of A.I.M.E. 227, 1419


(I !)63).

lI.

11 . 11. K llogg , pr ivat e communication (1964).

N. A. W11rn ra nd T. R.I ngraha m, Can.J.Chem. 38, 196 (1960).

II ,

n. N.l11 l 11 11 f11iH

IL !ICI

1'.R . Jngraham, Can .J .Chem. 38, 813 (1960).

1111.

T.R.INGRAHAM
17.

K.K.Kelley, United States Department of the Interior, Bureau


of Mines Bulletin, 406, 84 (1937).

18.

K.K.Kelley, United States Department of the Interior, Bureau


of Mines Bulletin, 383, 3 (1935).

19.

O.Kubaschewski and E.Ll.Evans, Metallurgical Thermochemistry,


337, Pergamon Press: New York (1958).

20.

M.Watanabe, Tohoku Imperial University Sci.Repts., XXIII, 1,


89 (1934).

21.

B.D.Roiter and A.E.Paladina, J.Am.Chem.Soc., 45, 128 (1962).

22.

W.Evans and D.Wagman,J.Res.Nat.Bur.Stds.49, 141 (1952).

23.

H.H.Kellogg and S.K.Basu, Trans. of A.I.M.E. 218, 70 (1960).

24.

C.E.Wicks and F.E.Block, United States Department of the


Interior, Bureau of Mines Bulletin 605, 84 (1963).

25.

T.Rosenqvist, J.Iron and Steel Inst., 176, 37 (1954).

26.

Extractive Metallurgy of Copper, Nickel and Cobalt, 28,


Paul Queneau Ed., Intersc ience Publishers: New York (1961).

27.

Ibid., 24.

...

'

106

THE PREPARATION OF PURE TUNGSTEN METAL BY


REACTION OF TUNGSTEN HEXACARBONYL WITH
HYDROGEN AND STEAM OR CARBON DIOXIDE*
by
J.J.Ward, A.D.Coon and J.H.Oxley
Battelle Memorial Institute

Abstract

The thermochemical conditions necessary to deposit high purity


tungsten from the hexacarbonyl have been calculated. At temperat ures above about 1150K, there is a wide range of concentration
leve ls in which steam and hydrogen can be mixed with the carbonyl
to avoid formation of tungsten oxide and carbide. At lower
't e mperatures, thermodynamic considerations indicate that such high
dilu tions are necessary that the deposition of pure tungsten
&ecomes impractical. The thermodynamic results were verified by
1xper iments - tungsten deposits containing less than 10 ppm carbon
W're prepared. The oxygen contents were considerably higher, i.e.,
i,n t he 500-1000 ppm range, presumably due to atmospheric oxygen
Nurp tion on the fine-grained porous coating obtained .

.. Thi work, p rformed at Battelle Memorial Institute, was sponsored


h.Y t he Na Liono.l A rona utics and Space Administration, Lewis
Ill I I ILl'(: h C nL l'.
tn7

J.J.WARD, A.D.COON AND J.H.OXLEY


INTRODUCTION
The object of this work was to develop a process for depositing
an adherent coating of high purity tungsten metal on a surface.
The surface could be tungsten or another material which might react
undesirably with a halide coating atmosphere. A suggested method
was the pyrolysis of tungsten hexacarbonyl, W(C0) 6. However, the
pyrolysis is complicated by the simultaneous formation of metal and/
or oxide, carbide and oxycarbide of tungsten. In order to avoid
the formation of such undesirable products and to stabilize the
metal phase, a mixture of hydrogen and steam can be introduced into
the reactor along with the tungsten hexacarbonyl.
A description is given of a thermochemical calculational model
for finding the conditions of temperature and initial reactant ratio
to give pure metal at one atmosphere pressure. The calculated
results were used to select experimental conditions in subsequent .
experimental work. A comparison is made of the results predicted
by calculation and obtained in the laboratory.
BACKGROUND INFORMATION
Lander and Germer(l) presented a comprehensive treatment of
plating of the metals, molybdenum, chromium and tungsten, from their
carbonyls in 1947. Their paper contains a complete review of the
literatyre beginning with the early works of Langer and Mand to
1947. They point out that this early work in 1888 with iron and
nickel carbonyls rapidly developed into a highly successful process
for obtaining pure nickel.
The classic paper by Lander and Germer is required reading for
one interested in the plating of metals from their carbonyls.
Powell, Campbell and GonserC 2) have summarized other vapor deposition
processes for the preparation of tungsten up to about 1955. Since
then Caves has obtained tungsten ~eposits from the bromide( 2) , and ~
Brenner and ReidC 4- 6) a~d Neiberlein and KenworthyC 7), have investigated the fluoride process. Recent work on the vapor deposition of
tungsten in fluidized beds has also been reported. (B-lO)
The results summarized in the present paper are an extension of
the Lander and Germer studi~s. The major improvement is the
correction of the process conditions to eliminat e the car bide as the
198

PREPARATION OF PURE TUNGSTEN METAL


most stable form of carbon in the deposit and to permit the preparat ion of the pure metal at atmospheric pressure by adjusting feed gas
c omposition. These corrections result in a s.ignificant increase in
t he minimum temperature level for the preparation of carbon- (carbide)
f ree deposits over that predicted by Lander and Germer assuming
equilibrium behavior.
The free energy values for the various gaseous compounds were
obtained from the JANAF tables(ll). The data for the tungsten
carbides were obtained from the study of Cunningham, Ward and
AlexanderC 12 ), who recommended the best values for these components
f rom those previously presented by OrtonC1 3 ), Gleiser and ChipmanC1 4 ),
McGraw , Seltz and SnyderC 15 ). Recent work by Alekseyev and
Shvartsman ( 16 ) was not available at the time this study was carried
out, and although they believed the subcarbide (W 2c) rather and the
monocarbide (WC) is most stable, their free energies are so close
Lhat the final calculations are essentially unaffected. The oxide
l're e energy values were obtained from the work of Huff, Squiteri and
,'nyder (l 7). However, in using these data, it was ~ecessary to assume
that t he activity of the solid phase would be unity, even when only
IL t race of carbide or oxide was present in the tungsten deposit. On
I. he bas is of solubility data, it appears that this should be a valid
1qmrox imation. The solubilities of oxygen and carbon in tungsten are
1111pare ntly in the very low parts-per - million range at the temperature
nl' i nterest in this study(18-21).

'l'll!o:ORETICAL STWIES
The purpose of the calculational work was to define the temperaand the initial amounts of tungsten hexacarbonyl, steam and
h,vdr oge n to yield a final plating gas in equilibrium with tungsten
11111t.1 Ll at one atmosphere pressure.
The final plating gas was to be
11 1111c1uburizing and nonoxidizing to tungsten metal. The condition of
i11 I 11 l111um vol ume of plating gas per unit weight of deposited tungsten
w1 11 1 l AO desir ed in order to maximize deposition rates .
A detailed
11111. 11111 of t he ca l c ula tional model is given in the Appendix to this
1. tll' ll

IN Ill

I '.

ll11w<v<l' . n Rummary of the pr ocedure i s as follows.

First, since

I 1111111 L11 11 1:1L1' bo11y l 111 highly un table at high t emperature, it is fully
11111 111111 1111 1 d

i'111<l ur Li u

cmcJJ t

1111 of int r s t an d may be ignored as a


1 f)f)

J.J.WARD, A.D.COON AND J.H.OXLEY


separate species in the equilibrium calculations. However, it does
establish the total moles of carbon monoxide, carbon dioxide, and
methane formed per mole of carbonyl feed on the basis of a carbon
balance. By an oxygen balance the total moles of carbon monoxide,
carbon dioxide and steam at equilibrium can be related to the feed.
A hydrogen balance provides a third relationship between the
equilibrium moles of hydrogen, steam and methane to the feed composition. Finally, by Dalton's law, the sum of the mole fractions of
all the components in the feed, as well as those in the equilibrium
mixture must be equal to unity. With a five gaseous component system,
i.e. CO, co 2, H2, H20 and CH 4, assuming that the equilibrium composition of carbonyl is zero, and that there is no methane, carbon
monoxide or carbon dioxide in the feed, then for a mole of carbonyl
feed we have seven unknown composition terms and an unknown value
for the total final volume. By mass balances and Dalton's law, we
have five conditions of restraint. The other three conditions. are
obtained from equilibrium constants for the reactions, such as:
H20 + co <=! co 2 + H2

(A)

co + 3H 2

CH 4 + H20

(B)

WC + co 2

w+

(C)

and
2co

or
W0 2 + 2H 2 t= W + 2H 20

(D)

It can be shown that methane formation is negligible over the conditions of interest and the problem in actual practice is simplified to
one less unknown and one less restraint.

...

'

RESULTS OF THEORETICAL STUDIES


The calculated results are shown in Tables I-V and the data are
summarized in graphical form in Figure 1. In this figure temperature
. is given as an independent .parameter. The abscissa is the initial
200

PREPARATION OF PURE TUNGSTEN METAL


35
30
i-

"'

3:
0
I

~o~~ ,,c:P
d--WC

25

'

20

"'

15

"'
<1>

:a:
0

.;::

:s

10

\200K

we
w
WC W
10

15

20

1300K

WC

1400
25

30

35

40

45

50

55

60

H2
Initial Moles, /W (C0) 6

Fig.1.

Thermochemical Conditions for the Preparation of Pure Tungsten


in the System (Initial) ofW(C0) 6, H2, H20

molar ratio of H2/W(C0) 6.


H20/W(CO) 6.

The ordinate is shown as the molar ratio

At llOOK, it can be seen that carbonyl must be highly diluted


with bydrogen and steam to avoid contamination. However, as the
process temperature is increased, the amount of bydrogen and steam
uocessary to stabilize the tungsten can be drastically reduced. At
1400K, an equal molar mixture should lead to the formation of a
pure tungsten.
A few calculations were also performed on the use of carbon
dioxide as a diluent. It was found that tungsten metal should be
<I< vositable from mixtures of the carbonyl and carbon dioxide without
l'orruation of either the carbide or oxide. For example at 1300K
t. h fo llowing ratio limits were established:
Minimum C0 2/W(C0) 6 (to avoid carbide)

1.5

Mitximum co2/ W(C0) 6 (to avoid oxide)

28. 0

It, <tLll b H <11 t hnt at 1300K t hese limits are wide and allow for
<011111<1< rnbl1 J' r 1 1 <10111 111 rn1 xing W( 0) 0 nrnl co 2 to give metall ic tungsten .

'flt

TABLE I
Equilibrium Gas Composition over Tungsten-Tungsten Carbide Mixtures
After Reaction of Tungsten Hexacarbonyl at 1100K and 1200K
Initial Molar Ratios
for W-WC Equilibrium
Reactants

1200K

ll00K*

Moles, H 2/W (CO) 6


oles H2.0/W(C0) 6

2. 7011 *
4.603

54.621
25.586
;..

N
0
N

.....

4.062
4.938

45.330
8.670

20. 790
3.210

?"
0

Final Equilibrium Gas


Components, (X 1)

~
~

Gas Components

co
co 2
H2
H20
CH 4

.L

Mole
Moles Xi
Fraction per Mole
W(C0) 6
xi
0.0492
0.0204
0.6540
0.2764
(0.0002)
1.0000

Mole
Mole
Mole
Mole
Moles Xi
Moles Xi
Moles Xi
Moles Xi
Fraction per Mole Fraction per Mole Fraction per Mole Fraction per Mole
W(C0) 6
W(C0) 6
W(C0) 6
W(C0) 6
xi
xi
xi
xi

4.241
1.759
56:379
23.828

0. 313
o. 138
0.341
0.208

4.164
1. 836
4.537
2.767

0.2853
o. 1147
0.3855
0.2145

4.280
1. 720
5.783
3.217

0.1627
0.0373
0. 7303
0.0697

4.881
1.119
21.909
2.091

0.0888
0. 0112
0.7672
0.1328

5.328
0.672
46.032
7.968

86.207 II

1. 000

13.304

1.0000

15. 000

1. 0000

30.000

1.0000

60 . 000

Poi nt at which t~ree solid phases,

w-wc-wo 2,

are in equilibrium with each other and the gaseous phase.

::i::

TABLE II

Equilibrium Gas Composition over Tungsten-Tungsten Carbide Mixtures


After Reaction of Tungsten Hexacarbonyl at 1300!<
Initial Molar Ratios
for W-WC Equilibrium

;g
Reactants
ol es H2/ W(C0) 6
ol es H20/ W(C0) 6

1300K**
0.06
1. 44

6.885
2.115

2.28
1. 72

21 . 39
2. 61

>"'3

51.06
2.94

>-i

.,,~

Final Equilibrium Gas


Components , (Xi)

:...i

>"'3

Gas Components

co
co2
H2
H20

Mole
Mole
Mole
Mole
Moles Xi
Moles Xi Mole
Moles Xi
Moles Xi
Moles Xi
Fraction per Mole Fraction per Mole Fraction per Mole Fraction per Mole Fraction per Mole
x.l
X.l
W(C0) 6
W(C0) 6
W(C0) 6
W(C0) 6
W{C0)6
xi
xi
xi
0.662
0. 138
o. 146
0.054

4.965
1. 035
1. 100
0. 400

0.518
0. 082
0. 310
0. 090

5. 18
0. 82
3.10
0. 90

0. 359
0. 041
0.500
0. 100

5. 39
0.61
7. 50
1. 50

0. 189
0. 011
0.724
0. 076

5.67
0.33
21. 72
2. 28

0. 097
o. 003
0.854
0.046

5.82
o. 18
51. 24
2. 76

1. 000

7. 500

1 . 000

10. 00

1. 000

15.00

1. 000

30.00

1. 000

60.00

CH4

2:

At this or higher temperatures and one atmosphere pressure, three solid phases and a gas phase cannot
coexist.

Cl

LE III
~-------

::..5

-- - ""- ..,.a:- :'1=65 : e:: - lt::.gst:~ e::.:-:::::e


ii:l:gsre:: Hexa.carbon;rl ar; !WO

x.:r.:~e s

: :oc o:

Initi al Molar Rat ios


f or W-ll{: Equil ibr ium
Reactants

c...
c...
=s

1400 K

oles H:/W(C0 ) 6
oles H20/W( CO ) 6 \

0.942

4.246
0.754

0. 558

23 . 028
0.972

)>

Final Equilibrium Gas


Components, (Xi)

Gas

Component s
co
co2
H2

H20

Mole
Fraction
x 1.

Moles Xi
per Mole
W(C0) 6

, Mole
Fraction

Mole
Fraction

xi

Moles Xi
per Mole
W(C0) 6

xi

Moles Xi
per Mole
W(C0) 6

0. 7509
0. 0491
0.1747
0. 0253

5. 632
0.368
1. 310
0. 190

0.4795
0. 0205
0. 4577
0. 0423

5.754
0.246
5.492
0.508

o. 1968
0.0032
0.7708
0. 0292

5.904
0.096
23 . 124
0.876

1. 0000

7.500

1.0000

12.000

1. 0000

30.000

CH4

c...

r:

ABLI;. H

Equilibrium Gas Composition over Tungsten-Tungsten Dioxide Mixtures


After Reaction of Tungsten Hexacarbonyl at 1200K
Initial Molar Reaction
for W-W0 2 Equilibrium
"ti

1200K

Reactant s

"ti

13. 053
10.947

3.744
5.256

:es Hi'f (CO) 6


:es H20/W(CO) 6

Si;

31. 667
22.333

>
>-3
H

Final Equilibrium Gas


Components, (Xi)

"lj

"ti

::::::

Mole
Fraction

x1

Moles Xi
per Mole
W(C0) 6

0.2774
0.1226
0.3723
0.2277
1.0000

4.161
1.839
5.585
3.415
15.000

Mole
Fraction
Gas Components

co
co2
H2
H20

.L

Mole
Fraction

xi

Moles Xi
per Mole
W(C0) 6

xi

Moles Xi
per Mole
W(C0)6

0.1387
0.0613
0.4964
o. 3036
1. 0000

4.161
1. 839
14.892
9.108
30.000

0.0693
0.0307
0.5585
0.3415
1. 0000

4.158
1.842
33. 510
20. 490
60.000

t.:i:.:1

>-3

11
c;:l

Ul
>-3

zt.:i:.:1
~

>-3

TABLE V
Equilibrium Gas Composition over Tungsten-Tungsten Dioxide Mixtures
After React ion of Tungsten Hexacarbonyl at 1300K and 1400K
Initial Molar Ratios
for W-W0 2 Equilibrium
Reactants
Moles H2/W(C0) 6
Moles H20/W(C0) 6

11.22
12.78

0. 28
3.72

..

co
co 2
H2

H 2o

2:

c...
c...

1400K
0.949
5. 051

II

24.3
29;7

0
C')

II

Mole
Fraction
x.l
0.409
o. 191
0.219
0.181
1.000

Moles Xi
per Mole
W(C0) 6
4.09
1. 91
2. 19
1.81
10.00

Mole
Fraction

xi
0. 136
0.064
0.438
0.362
1.000

Moles Xi
per Mole
W(CO) 6
4.08
1. 92
13.14
10.86
30.00

~0
?'"

Final Equilibrium Gas


Components, (Xi)

Gas Components

II

1300K

8:.;;:
Mole
Fraction

xi

Moles Xi
per Mole
W(C0) 6

xi

Moles X1
per Mole
W(C0) 6

0.2425
0.2575
0.3366
0. 1634
1. 0000

2. 910
3.090
4.039
1. 961
12.000

0.0686
0.0314
o. 4363
0.4637
1. 0000

4.116
1.884
26.178
27.822
60.000

Mole
Fraction

~
~
::i::
0

~
~

PREPARATION OF PURE TUNGSTEN METAL


EXPERIMENTAL STUDIES

Exploratory tests were then made to determine if pure tungsten


could be deposited by thermal decomposition of tungsten hexacarbonyl
with controlled H2/H 20 additions as indicated by the t~ermodynamic
studies. A sketch of the experimental apparatus is shown in Figure
2. Hydrogen was passed through a water bubbler and then over the
surface of heated solid W(C0) 6; both the bubbler and sublimer were
maintained at appropriate temperatures to obtain the desired inlet
gas composition . The feed was then introduced into a fluidized bed
of either Al 2o3 or tungsten powder maintained at elevated temperatures. After depositing a reasonable amount of metal the reactor
was cooled, the bed discharged and the carbon contents determined
by a Leco conductometric analysis. Oxygen contents on samples
deposited on a tungsten bed were obtained by vacuum fusion. These
resu lts were then converted to impurities in the deposited metal by
a material. balance.

Water vaporizer

Manometer

Hydrogen or
helium supply

Vent

r oler

eoctor

3 [Heater2
>
_) ("~'

Dus t traps

Oibutyl phthalate
bubbler

Tungsten

hexocorbonyl
vaporizer

I'ig. 2.

Schematic Diagram of Apparatus for Applying Coatings of


Tungsten from Tungsten Hexacarbonyl

lmSULTS OF EXPERIMENTAL STUDIES

'l'n >i al experiment al data are shown in Table VI and a sample of

coat d particle is shown in Figure 3. Carbon contents


v ry l ow part s-p r-million range could be obtained.
llowt Vt r, th< ox.rn< n ont ntR w r always at a considerably higher
l 11v1 I , pl'Olmht .v duo L< 1Lt11101-1p h<r1c oxyg n s rption and /or r eaction
IL L1111 gHto n

111 th

J.J.WARD, A.D.COON AND J.H.OXLEY

Fig.3.

Photomicrograph of a Sectioned Tungsten-Coated Tungsten


Particle (l-3 micron coating on 105-210 micron particles)

as the deposits were very fine grained and porous. Germer and
Lander have reported on the pyrophoricity of some of their
products(l) which are believed to be similar to these reported
herein. Oxygen contents in the 500 to 1000 ppm range were usually
obtained. However, this range of oxygen content is somewhat less
than the range of oxygen contents for most fine particle size
commercial grades of tungsten powder< 22 ) . Caves has pointed out in
a private communication that high purity tungsten particles of a
commercial grade of much larger size than reported in this paper
have three to five times the oxygen content of tungsten prepared in
the experimental investigation< 23).
A limited number of experiments using mixtures of carbonyl, carbon
dioxide and helium were also carried out. The deposits for these
runs also had considerable porosity . However, material with carbon
contents of less than 10 ppm could be readily prepared at temperatures of about 1000c. At temperatures around 500c, Lander and
Germer were unable to deposit carbon-free tungsten with carbon
dioxide additions (1) ,,..
'
The porous nature of the deposits is believed to be due to gas
phase nucleation of tungsten , tungsten carbide, and/or tungsten
oxide particles which are subsequently reacted, s inter ed, and/or
cemented by subsequent co~ting, to the surface of the larger particles in the fluidized bed. Further work to avoid this ga phas
nucl eation was conduct d in a part i o. l vacuum uH In~~ 11 li (ltt ct Lungl'lt n
20fl

TABLE VI

Typical Data for Deposition of Tungsten From W(C0) 6


on Powder in a Fluid Bed

1-1

Sm:.ber

Initial
Bed,
g.

W(C0)5
Fed,
g./hr.

Cfi07 -6

40.0

20707 -9
20707-10

Run

Temperature
Bed W(C0) 6,

Tungsten
H20,
Final
H2/
W(C0) 6 W(C0) 6 Bed, g.

Deposited~

Efficiency,
%

Carbon
Content,
ppm

Reactor
Diameter,
in.

c.

c.

7.33

1000

129

50.3

744

47.3

2.66

35.0

1000

80.0

13.85

1000

138

44 . 3

736

97.8

9.62

83.4

75

80.0

15.20

1000

136

60.6

57

101. 8

10.90

72.0

g./hr.

;;g
~
>-3

2:

~
~

J.J.WARD, A.D.COON AND' J.H.OXLEY


wire. The deposits obtained on this tu~gsten substrate appeared to
be completely dense when examined metallographically. A
photomicrograph of this deposit is shown in Figure 4.

Fig . 4.

Photomicrograph of Tungsten-Coated Tungsten Filament (Etched)


Showing Apparent Dense structure

CONCLUSIONS
The overall conclusion that can be made from the computational
and experimental work of gas plating from tungsten hexacarbofiyl is
that:
(1)

A calculational procedure has been developed that relates a


final plating gas composition in equilibrium with tungsten
metal to the initial amounts of tungsten hexacarbonyl and
selected addition agents.

(2)

A high purity tungsten metal containing < 10 ppm carbon can


be prepared following experimental conditions predicted by
the calculational procedure.

(3)

A low oxygen oontent met'al can probably only be maintained


in an oxygen-free condition by operating under deposition
conditions whereby dense deposits are obtained.

PREPARATION OF PURE TUNGSTEN METAL


ACKNOWLEDGEMENTS

Special appreciation is expressed to the NASA Lewis Research


Center for financial support of this study and permission to publish
the results and to the NASA Technical Manager, Mr.Robert M.Caves
for his interest and timely suggestions during this program. The
continuing advice and encouragement of Dr.John M.Blocher, Jr., in
this program has also been of significant value. We also wish to
acknowledge the help of Mt.Herbert E.Carlton in carrying out the
low-pressure experiments, and to Dr.G.W.Cunningham for his many
helpful discussions.

l1

J.J.WARD, A.D.COON AND J.H.OXLEY


REFERENCES

1.

J.J.Lander and L.H.Germer, Am.Inst.Mining Met.Engrs., Inst.


Metals Div., Metals Technol. 14, 6, Tech.Pub.2259 (1947);
Metal Ind. (London) 71, 459,487 (1947).

2.

C.F.Powell, I.E.Campbell and B.W.Gonser, Vapor Plating, John


Wiley & Sons, Inc: New York (1955).

3.

R.M.Caves, Trans.Met.Soc.A.I.M.E. 224, 267 (1962).

4.

W.E.Reid and A.Brenner, NBS Tech. News STR-2450 (Jan.1960).

5.

A.Brenner, Trans.Inst.Metal Finishing 38,123 (1961).

6.

A.Brenner and W.E.Reid, U.S.Pat. 3,072,983 (Jan.15 1963).

7.

V.A.Nieberlein and H.Kenworthy, U.S.B.M. Rpt. Invest. 5539


(1959).

8.

J.H.Oxley, J.F.Hannah, J.M.Blocher, Jr. and I.E.Campbell,


U.S.A.E.C. Rpt. BMI-1297 (Oct.8 1958); Ind.Eng.Chem. 51, 1391
(1959).

9.

J.M.Blocher, Jr., N.D.Veigel, J.H.Oxley, V.M.Secrest and


E.E.Rose, U.S.A.E.C. Rpt. BMI-1440 (May 25 1960).

10.

J.H.Oxley, E.A.Beidler, J.M.Blocher, Jr., C.J.J..urons, R.S.Park


and J.H.Pearson, Paper presented at 5th Plansee Seminar,
Reutte-Tirol (June 22-26 1964).

11.

Dow Chemical Company, J.A.N.F. Thermochemical Tables,


prepared under the auspecies of the Joint Army-Navy-Air Force
Thermochemical Panel, Midland, Michigan (1960).
,I

12.

G.W.Cunningham, J.J.Ward and C.A.Alexander, U.S.A.E.C. Rpt.


BMI-1601 (Oct.30 1962).

13.

G.W.Orton, Ph.D., Thesis, The Ohio State University (1961).

14,

M. Gleiser and J.Chipman, Trans .A.I.M.E. 224, 1278 (1962).

212

PREPARATION OF PURE TUNGSTEN METAL


15.

L.D.McGraw, H.Seltz and P.E.Snyder, J.Am .Chem.Soc. 69,


329-331 (1947).

16 .

V.I.Alekseyev and L.A.Shvartsman, Akedemiya nauk S.S.S.R.,


Izvestiza, Otdeleniye tekhnicheskikh nauk., Metallurgiya i
gornoye Delo. (1), (1), 91W (1963).

17.

G.Huff, E.Aquiteri and P.E.Snyder, J.Am.Chem.Soc. 70, 3380


(1948).

18.

M.R.Andres and S.Dushman, J.Phys.Chem. 29, 462 (1925).

19.

B. C.Allen, D.J.Maykuth and R.I . Jaffee, J.Inst.Metals 90, 120


(1961).

20.

R.H. Atkinson, G.H.Keith and R.C.Koo, Chapter in Refractory


Metals and Alloys, Interscience Publishers, Inc : New York
(1961).

21.

H.J.Goldschmidt and J.A.Brand, J.Less-Common Metals, 5, 181


(1963).

' 2.

B.C.Allen, U.S.A.F. Rpt. DMIC-Memo 90 (March 13 1961) .

2 ~.

R.M.Caves, Private Communication N.A.S.A. Lewis Research


Center (November 19 1964).

J.J.WARD, A.D.COON AND J.H.OXLEY


APPENDIX
Detailed Outline of Method of Calculation
The following calculational method for determining initial and
final conditions in the w[co) 6-H 2-H 2o system is derived on the
basis of a mass balance and equilibrium relationship derived from
known thermochemical data. The thermochemical data required in the
calculations are given in Tables A-1 and A-2.
Mass Balance
The mass balance is derived as follows based on 1 mole of
and 'a' moles of H2 and 'b' moles of H20.

w[co) 6

Carbon Element Balance


A carbon balance is made by equating the total number of carbon
atoms added to the system to the number of moles of carbon products
in the gas phase, as:
6 = nco + nco

(1)

where n 1

= gaseous

moles at equilibrium, or:

1 = - Cnco + nco ]
6

(2)

Dividing thru by the tota} gaseous moles, :&lg at equilibrium:


1

Lllg

~ = - [xco + Xco ]

where Xco and Xco

...
2

(3)

'

= mole fraction in equilibrium gas .

ABLE A- I

Standard Free Energy and Logarithms of Equilibrium


Constants of Formations of Compounds of Interest (1l-l 7 l

=
Temperature
1300K

1200K

llOOK

1400K

1500K

"'Cl

f':.F 0f
Cal.

6F f0
Cal.

6F f0
Cal.

6F f0
Cal.

f':. FO
f
Cal.

log 10Kf

C02(g)

-94, 617

18. 798

-94,638

17.235

-94,653

15.912

-94,663

14.777

-94, 670

+13.793

OO(g )

-49,956

9.925

-52,046

9.478

-54, 121

9.098

-56, 185

8.270

-58, 240

+8. 485

"'Cl

320(g)

-44,708

8.882

-43,367

7.898

-42 , 017

7.063

-40,657

6.347

-39,292

+5.724

C(c )

-9,950

+1. 977

-10,270

+l.871

-10,600

+1. 782

-10, 950

+1. 710

-11,630

+1. 695

2C(c)
I0 2 (c)

+ll, 100

-2.206

+7,890

-1. 437

+4,670

-0.785

-1,460

-0.228

-1, 750

0.255

-93, 179

18,512

-89,078

16.223

-85,012

14.291

-80, 984

12.642

-76,989

11. 215

f()3 ( C)

-134,162

26 . 654

-128, 411

23.386

-122,697

20.626

- 117, 021

18.267

-111,380

+16.227

0 3 (c)

-55,927

11.111

-54,655

9.954

-53,382

8.974

-52, 109

8.134

-50,833

+7.404

CH4(g)

+7,257

-1. 442

+9,897

-1. 802

+12,547

-2. 109

+15,206

-2, 374

+17,871

-2.604

log11ff

log 10Kf

log 10Kf

log 10Kf

;:..
o-3

......

!.;

o-3

rA

TABLE A-II
standard Free Energy Change and Logarithms of
Equilibrium Constants of React ions of Interest

1200K

llOOK

t,pO

t,pO
Reaction

~
:::::

Cal.

I H2 (g) + W(c) +CO(!~):=


WC(c} + H20(g)
-4. 702
II 2CO(g) ;:= Co 2 (g) + C(..g) +5,295
III CO + 1/20 2 ..... C~ 2
-44,660
. I'll C0 2 (g) + WC(c) ;:= W +
2CO(g)
+4,655
V 2H 2 + W0 2 (c) ;:.W(C) +
+3,763
2H20(g)
VI H2 (g) + 1/20 2 (g) ;:=
H20(g)
-44,710
\TII C02(g) + H2(g) ;:=
CO(g) + H20(g) .
-5o
ITII H2 + CO := H20 + C
+5, 250
IX
+ CH 4 (g) ;:wc(c) +

t,pO

loglOKR

Cal.

1300K

t,pO

loglOKR

0.934 . -1,591 0 . 290


+9, 454 -1. 722
-1.052
+8.876 -42,592 7,760

Cal.

1400K

loglOKR

+l,504 -0.253
+13,589 -2.285
-40,532 6.816

Cal.

'

1500K

6F 0f
loglOKR

+4, 578 -0. 715


+17,707 -2.765
-38, 478 6.009

Cal.

log 10KR

+7. 318 -1. 069


+21, 810 -3.177
-36,, 430 +5, 300

~
>
!='
(')

0
0

:z
-0.925

+816 -0.149

-2,990 +0.503

-6, 760 +l, 056

-10, 180 +1. 482

-0.748

+2. 344 - 0. 427

+980 -0.165

-330 +0.052

-1,595 +0.231

+8.882

-43,370 +7 . 898

-42,020 +7.063

-40,660 +6.347

-40,657 +6.347

+o. 006
-1. 043

-780 +0.138
+8,680 -1.980

-1 , 490 +0.247
+12, 100 -2 . 035

-2. 180 +o. 338


+15,528 -2.423

-2,862 +0.416
+18,948 -2.761

-17,207 +3.419
2H2(g)
-49,956 +9.925
+ 1/202 := co
XI ~ + lf0 2 (c) ;: 2C0 +
2
{c )

-20,617 +3,673
-52, 046 +9. 476

-23, 147 +3. 891


-54, 121 9.098

-26, 156 +4. 085


-56, 185 8.770

-29, 900 +4.299


-58,240 +8.485

~
0

!'--<

?=
~-

xc

c....
c....

-. 709

-0.663

-0.628

t;;l

><

PREPARATION OF PURE TUNGSTEN METAL


OXygen Element Balance
The oxygen balance is made by equating the total atoms of oxygen
added to the number of moles of oxygen in the gas phase, as:
(4)

Dividing through by 6 + b and L:ng as in carbon element balance:


1

- -=-Tu g 6 + b

Cxco + xH o + 2Xco ]

(5)

Hydrogen Element Balance


On the basis of stoichiometry:
(2a + 2b) = 2nH

+ 2nH 0 ,
2

(6)

dividing as above by L:ng and 2a + 2b:

(7)

11:1Jll11tions 3, 5 and 7 may be equated.

11111 ton' s Law

there are only four gases present in significant amounts:

111 n

x0

+ Xco

+ XH

+ XH 0

(8)

=1

'l'h< four conservation equations for use in the solution are


11:1J111

Lion

3, 5, 7 and 8, respectively.

1:1mt librin Relationships


'1'111 l.w<J 1 <J.11 l brium oQuations of interest i n calculation of W-WC
111111 1 lhrtum 11ri 11:1,1u1Lt onA 1 nnd 4 , from Tabl A-2 . With the valid
Hri llllllll, 11111 1.lml,;
:~

17

J .J.WARD, A. D. COON AND J . H. OXLEY

where fi is fugacity, Pi = partial pressure, and Xi is the mole


fraction and P the total pressure .
Then:

(9)

And :

x2co

( 10)

. p .

"'
jri

The equilibrium expression of interest in studying the W-W0 2


equilibrium are:

(~,"r

aw
Kv

awo2

( 11)

and:

aw
Kxr = - -

awo 2

Method of

(~~
Xco

Calcul ~tion

(12)

'

A value for the sum Xco + Xco is assumed.


for determining if this is a possible value.

A test is given later

For the W-WC equilibr ~ a a .quadrat ic equat i on i s obt ained from


equations 3 and 10.
1 may be solved numerically wit h t he above
ass umption by :
2 111

PREPARATION OF PURE TUNGSTEN METAL


p

KIV

+ Xco - - - = 06
2:ng

- x 2co

CO is obtained, then
1 -

the sum of XH

co 2

by substitution in Equation 10.

(13)

Since:

co - co 2

(14)

+ XH 0 is obtained.
2

Then by Equation 9, the ratio XH 0;xH is calculated. Now the


2
2
val ues for XH and XH 0 are obtained. The value of a + b, that is
2
2
moles of initial H2 and H2o. may be obtained as a sum from Equation
6 and 2:ng from Equation 3. The value for 'b' may be obtained now
from Equation 5 and 'a' obtained by difference. A test for a unique
H lut ion to the initial assumption of [co + co ] is that a + b ~ b.
2
Now the initial moles of H2/W(C0) 6 and H20/W(C0) 6 and the final
The mole/W(CO) 6 of
product are obtained by multiplying:

mole fraction composition of gas are known.

(15)

In studying the simultaneous equilibria of WC-WC 2-W gas , the same


d may be used. In practice, this point was obtained from the
1111t11H balance equation and the following equilibria expression of
'l'1 Ll>l e A-2, K1 and Kxr
11111l. h

'l'ho mole fraction of CH , XcH , was calculated as a second!\rY


4
4
111111po11 nt by means of Equation 9 of Table A-2. It was always less
f.lm11 0. 0002 mole fraction and was therefore disregarded. It could
h1 LLLlt n into account by a reiterative method, i f so desired .

. l!1tlculation of WW0 2 Equilibria

nl u1 ati ng the w-wo 2 equilibria, the same mass balance


UH d. now ver, in calculating the equilibrium,
1:11111d. lo111 II uncl 12 o.r Ul:l dint ad of Equations 9 and 10.
111

11q111 d . 10111-1 itT'<

J.J.WARD, A.D.COON AND J.H.OXLEY


The calculation of the simultaneous equilibria of gas is made at a temperature, T from the equation:
W02(c) + 2WC(c);:: 3W + 2CO

W-Wc-wo 2
(16)

and:
(17)

'

' 20

PANEL II
DISCUSSION
Dr.Wallace - I wish to thank the panel for their interesting
presentations which we have just heard. I am sure that the audience
will wish to ask some questions. However, we will first have some
intra-panel discussion. To start the discussion I should like to ask Dr.Senderoff, what
material he used for the containers for his fluoride melts.
Dr.Senderoff - Most of the containers which we used in working with
the tantalate solutions were nickel. Platinum and molybdenum were
used in a few cases. With columbium we used platinum since nickel
was unsuitable. As we go to more noble metals, it looks as though
platinum or gold will be the most suitable containers. For the
nickel-nickel fluoride, Flinak reference electrode, the nickle
ruc ible can itself be the electrode. In order to maintain an inert
atmosphere, all compoqents were contained in a Hastelloy can filled
with argon.
Dr.Ingraham - May I add one more question? You are well known for
your work on fused salt potential series in chlorides. Have you
done enough work on the various fluorides to be abl~ to say how
lhe potentials agree between the chloride melts and the fluoride
111 Its.
nr.Senderoff - I have not studied the same systems in them, so I
nn' t say. It is pretty clear that if we interpolate among the
vurious elements that we have looked at in both systems, that in the
fluor ide system, the EMF series is spread out more. That is, there
mlght be a total of four volts from fluorine to sodium in the alkali
t'luoride solvent, but it is nearer 3 volts from chlorine to sodium
In the alkali chloride solvent. Nickel is extremely noble in both
HyHt ms compared to its place in water systems, and in general the
1 1 rn nts take their equivalent places in chlorides and fluorides.
l'd like to ask Dr.Johnson a question. He pointed out that a very
wid spread of activity coefficients is observed as one varies the
mt l1tl so lvent.
I would expect a similar effect if one varied the
rm It, HOlv nt.
Do you have any information or opinions along those
t

I Ill H?

PANEL II

Dr.Johnson - We have studied the effect of salt composition on the


distribution coefficient by the use of salt mixtures composed of
various alkali and alkaline-earth halides and magnesium chloride.
Changes in the distribution coefficients which could be attributed
to changes in activity coefficients of the solute in the salt
mixture were much smaller than the changes caused by variation in
the liquid metal solvent composition . The largest change was of the
order of a factor of 100. However, the effect of salt composition on
the separation coefficient was never found to be large enough to be
of practical use - apparently the effect of changes in salt
composition on the activity coefficients is about the same for the
actinide and rare-earth metal halides which we studied . Workers at
Hanford (References 6 and 7, main paper) have reported much larger
effects of salt composition on distribution coefficients for a fused
salt solvent in which aluminium chloride was a component . Part of
the effect they report is due to changes in the act i vity coefficient
of the solute chloride in the fused salt; however , the gr eater part
of the effect is probably due to the for mation of a stable complex
between the oxidant, aluminium chlor ide in their case , and the
potassium chloride used as a solvent .
Dr.Senderoff ~ I have not done anything is this area. I am
speculating , but the result with aluminium chl oride indicates the
direction where the big differences might be observed. I f one tries
to apply some general concepts of basicity of t he solvents r el ative
to that of the solutes one may effect major separations . If we
define alkali halides as basic solvents (halide ion donors) and
AlC1 3 as an acid solvent (halide ion acceptor) then a solvent of
intermediate basicity between the two solutes to be separated may be
very effective, while one much more strongly acidic or basic than
both solutes would be less effective . This may apply not only to
the activity coefficient term, but actually would affect your
reducing potential term, since it has been shown that in some multiple
valence systems the equilibriugi ratio of the upper and lower valence
states may differ by as much ~s 100-fold whether they be in NaF~KF
solvent or a NaF -ZrF 4 soJvent.
Dr.Johnson - Your speculations are interesting - the re s ult s
obtained by the Hanford worker s have been expla ined by t hem i n t er ms
of differe nces in ac i dity between t he oxic:ant (al uminium chloride ) .
th
Jute chlorid e (an act i nide chl orid ) and t h solv nt (potassi um

.,.,.
' ' '"

..f

DISCUSSION
chloride). You are correct in suggesting that both the activity of
the oxidant and the activity coefficient of the solute halide can
be modified by changing the solvent composition. However, only
changes in the activity coefficient of the solute halide will affect
the separation coefficient. It would be interesting to determine
whether sufficient differences exist in the acidity of the actinide
and rare-earth halides to markedly change their activity coefficients
by the addition of a basic component to the fused salt mixture.
Dr.Wallace - I have a question to ask you Dr.Johnson. I was concerned about one point in connection with your paper. I assume I
misunderstood something. Perhaps you could clarify this for me and
others, as well. I thought I picked up the impression from your
sl ide that magnesium was more effective in displacing certain species
f rom the fused salt solutions when in reduced concentration, (mole
~r action a half or two-tenths) than it was in the pure material.
I
can't understand this and it seems anti-thermodynamic to me . It
~ee ms to me that you're saying that the activity of the magnesium is
higher in the solution of mole fraction 0.2 or thereabouts, than it
is in the purer material. If this were the case, you should have a
t wo-phase' system.
Ur.Johnson - Nol I hope I have not given that impression. The
1Lct ivity of magnesium in a 20 atom per cent mixture, for example,
I~ certainly less than in pure magnesium.
The greater reducing
t>otential of a dilute magnesium-zinc solution compared to pure
mugnes tJrn is due to the very small activity coefficient of the metal
b i ng reduced in this metallic solution. Thus the low activity of
I.ho reducing agent in a dilute metallic solvent may be compensated
t or by a small activity coefficient of the metal being reduced.
'l'h se two factors are taken into account in the reduction potential
t. rm. The neglect of the activity coefficient factor can lead to
v ry erroneous predictions of the reduction potential of a metallic
11ol ution. When activity coefficient data are lacking a qualitative
Ind ication can be obtained by examination of the phase diagrams of
t.h metal being reduced with each of the metallic constituents of
t.h , r duction solution. Low solubilities of the solute metal comh Ind with s table intermediate phases may be taken as an indication
t.l11tL t h activity coefficient of the solute is small.

PANEL II

Dr.Wallace - So, it's the activity coefficients of the material in


the metallic form that makes the difference. It's not the reactants
activity, it's the products activity. I see. This clears up very
nicely.
Dr.Senderoff - Regarding Dr.Ward's discussion I would like to point
out that those of us who are interested in thermodynamics admire his
success in establishing the conditions for obtaining the desired
purity of tungsten by his excellent thermodynamic calculations.
Dr.Abner Brenner of the Bureau of Standards, however, used a nonthermodynamic method for obtaining the same result. For tungsten
hexacarbonyl he substituted tungsten hexafluoride, which he says has
no stable subflorides or other significant impurities as reaction
products under a wide range of temperatures and pressures, and so
obtained high purity tungsten with relative ease. I believe that
most gas-plated tungsten is now produced by reduction of the
hexafluoride.
Dr.Ward - I haven't any particular preference for the carbonyl
method myself, but there are other people at Battelle who do prefer
this method of plating out tungsten. I was interested more in the
design of the thermochemical calculational procedure. That is about
all that I can discuss. We have Dr.Cunningham here who is the head
of our Materials Thermodynamics Division and he may want to make a
comment.
Dr.Cunningham - The fluoride technique is well known to the people
at Battelle and is generally preferred. However, in this case it
was desirable to develop a technique which eliminated the
possibility of halide contamination and therefore the carbonyl
process was investigated. Dt:.Oxley and his co-workers reported vapor
coating of tungsten from the halides in previous publications.
Dr.Senderoff - May I put one question to Dr.Meddings. In his discussion of the effe,g-t of ammon'ia and of the hydrogen, he pointed out
that his information was mainly obtained from 25 degrees, but the
operation was done at a much higher temperature. In work that I
did some years ago in a bomb reaction with aqueous systems indicated
that as we went to about 300 degrees centigrade, many of the complex
compounds which normally:_ are stable,. hydrolize to form oxides or

L_.- _ _
224

_ __

DISCUSSION
hydroxides rather than soluble complexes. The reduction of these
compounds with hydrogen might be by an entirely different process
not related to the EMF considerations discussed. Does one have any
information on stability of these ammonia complexes at the
temperature of operation?
Dr.Meddings - Yes, indeed one does. In the first case, I should
perhaps have mentioned that the nickel ammine solutions used for
the commercial production of nickel by hydrogen reduction are
buffered with ammonium sulphate. The molar ratio of ammonium sulphate to nickel sulphate is about three to one. As a result there
is no discernible hydrolysis at the usual operating temperature of
350 F.

As to the question of the stability of these nickel ammine species


at higher temperatures, there have been no available measurements
until the last twelve months. But Professor Peters of U.B.C. has
recently published a paper which includes some studies on the
for mation of this nickel ammines at higher temperatures. The early
work of Bjerrum showed that at 25c the average number of ammonia
m lecules bound to nickel increases as the molar ratio of ammonia
t.o nickel in solution is increased. This average number, n, rises
l,o a maximum of 6 at an ammonia to nickel ratio of about ten. Now
Prof .Peters has shown that as the temperature is increased this
I mi ting value of n gradually decreases. Speaking from memory, I
holieve that at 175C the upper limit on n is between 2 and 2.5.
011 this basis one can only say that the distribution of the various
111unine species found by Bjerrum at 25C is not directly applicable
1 t, 175C.

To revert to the query regarding hydroxides. Although the commercial


01ic mt ion is carried out with buffered solutions which do not hydrolyze
It, ls possible to carry out reduction from unbuffered solutions. As

xtre me example it is possible to reduce a solution of nickel


11lphnte, with a two to one ammonia to nickel ratio. In this case
Ll1< , r acti on solution becomes a highly viscous paste at reduction
L1mpornt ure - but reduction is still as rapid as in the normal
l,1111' r d s olutions. However, there is no evidence as to whether
Liu r du tion mec hanism in these two extreme cases is the same or
1 11
1

ilot.

PANEL II
Dr.Wallace - I have one other question for Dr.Johnson. As many_of
us know, there has been work done primarily by Bredig that shows
that the alkali halides and the alkali metals are miscible in all
proportions at elevated temperatures. Your partitions, of course,
depend on the fact that you're dealing with two phases. Are there
instances in which things go along for a little while, and then you
achieve miscibility of your metallic phase and salt phase, so that
the two phase character disappears? Or is this an effect which is
unique with the alkali halides?
Dr.Johnson - We have not seen evidence for complete miscibility of
the metal and salt phases in the systems which we have studied.
Since these systems consisted of a binary magnesium-zinc liquid
metal phase and a magnesium chloride-alkali metal chloride fused
salt phase miscibility is not to be expected. It is generally
found that extensive miscibility of a liquid metal and a fused salt
is confined to the pure metal and the pure halide of the metal.
The addition of a foreign salt to the fused salt phase greatly
decreases the solubility of the metal.
Dr.Wallace - I have a question for Dr.Ward. In cracking or
pyrolizing this carbonyl since you want to keep the tungsten free of
oxygen , I would think that you would want to avoid putting any
other oxygen in the system. Why put steam in? What is the purpose
of the steam?
Dr.Ward - That's to decarbonize the tungsten monocarbide that's
being formed. See, there's a balance drawn between the deoxidation
and decarburization.
Dr.Wallace - Is it a kinetic 'proposition then? Would it not be
possible to decarbonize through the methane formation?
Dr.Ward - At the temperatures ,imvolved, we .couldn't, because the
equilibrium constant for the formation of methane is unfavorable.
We only have about 0.008 of a mole fraction of methane formed at
equilibrium.
~

Dr.Johnson - I have a question for Dr.Senderoff. In your paper


you present a method for a etermining the n, or the number of

DISCUSSION
faradays, of reduction reactions. Do you believe that agreement
within 5% of an integral value is about as good as can be expected
using your method?
Dr.Senderoff - There are a number of ways of getting n, depending
upon what reactions you are dealing with. The most general way is
the least accurate. That is the plot of E as a function of t that
we have shown in the paper as Fig.7. The fact that we have a
plateau in which E varies slowly with t (i.e., Fig.1) tells us
that this will not be a sensitive measure of n, but it is
nevertheless useful. However, where there is a step-wise reaction,
obtaining n by the ratio of the lengths of the two plateaus is an
extremely sensitive measure of n as shown in Table III. If r 2/r 1
can be obtained with even moderate accuracy, it provides a valuable
criterion for evaluating the n for each step. A third way of getting
n requires an independent knowledge of the diffusion coefficient.
If one knows the diffusion coefficient of the ion under study, the n
.for the reaction may be obtained from the equation,

11

= 2ir 1/ 2;c7T1/ 2 FDl/2 (Equation (1) in the paper), the accuracy of

which is limited only by the accuracy of measuring r.


llr. Wallace - Thank you. If there are no further questions or
nomments among the panelists, perhaps we could precede to the
1J11ost ions that have been relayed to us from the audience. I have,
I' lr1:1t of all, a question for Dr. Senderoff here. How is the
mimmercial hot bath for plating tantalum cleaned up to remove iron
(which would plate out if not removed)?

Acnderoff - In normal practices, as we operate now, the iron is


from the electrolyte before it is used. If iron does get
lnLo the electrolyte (and that has happened), it is removed by low
1 11r nt density electrolysis.
Ill'.

11 m ved

((i11< st i on from the floor) - How do you prevent iron from entering the
11 trolyt e when plating on ferrous alloys?

111, Hn111 lnrnl1 f - That is part of the, shall we say, art of plating which
1 cl 1!111' i xp ct t
up in this discussion. The art consists of getting
.v111 11 11l11LJ111<
1,1111

I'll!'

Lh< lro11 t.o

1LH f1 Liit 1.11:1

<ll1111<~lv

pos1:1ibl , and giving very little


In
jron is nobl to tantalum, it
'l:.l'l

PANEL II
wi ll not dissolve rapidly, and what does dissolve plates out
l'r ferentially. It is a serious problem only when there is accldental
Kross contamination which may be caused by anodic dissolution of
f rrous alloys or their long exposure to the electrolyte in an
nnplated state.
llL

th

Wallace - We have another question for Dr. Senderoff. What caused


potential of your nickel NiF 2 reference electrode to vary?

Dr.Senderoff - The problem of a suitable reference electrode is one


which we have been studying for some time and for which we have as
~tt no satis factory solution.
The instability of the Ni, NiF 2
I' fur nee electrode has also been noted by Professor Grojtheim who
Indicates that after about two hours of moderate fluctuation it
h Kins to change very rapidly. At present we can only say that we
ti n t know the cause of the variability of this electrode.
llr. W11llace - There is a question here that we have for Dr. Ingraham.
011 y ur predominance area diagram for the Fe-s-o system there
n11v1 nr d to be a point at which five condensed phases and one gaseous
11h1LH1

w re in equilibrium.

Is this in accordance with the phase

r u1 ?
llr. lniiraham - On close examination of the diagram you can see four
r n 1 cl phnses and one gas phase.
This is an invariant point in the
r .YHt.um and it occurs near a temperature of 1050K. When the

ttlu.l(rctms are calculated for a slightly higher or lower temperature,


1ml y th three solid phases and one gas phase appear in equilibrium,
lu
rdance with a univariant equilibria.
111. W1ll l1Loc - Dr. Ingraham, do you have any information on the
m1Lll 1i1J.11 H

oxide-manganese sulfide diagram?

Ill' . lnarraham - Ye s , but it is not published yet. It is of particular


l11L1 Mt to notice that the promin,rtnt oxide involved during roastin~
'

I 11 Lill mo.nga.nese system in Mn o rather than Mri 2o 3


3 4
111,

W1ll llL

Also Dr. Ingraham, is the iron oxide - iron sulfide

<I In1tr1u11 uvnilublo?

~211

DISCUSSION
Dr.Ingraham - Not yet. The first copies will be available when the
proceedings of this conference are published.
Dr.Wallace - The session this afternoon has involved quite wide
thermodynamic coverage. You will recall that there are several
techniques recommended from text books for obtaining thermodynamic
information. There are electrical methods, methods which depend
upon chemical reactivity, partition coefficient methods, freezing
point depressions or boiling point elevation methods for obtaining
thermodynamic data which are nothing more than specific kinds of
phase equilibria. All of these classic methods for obtaining
reliable thermodynamic information have been involved in the
discussion here this afternoon.

'.'II

PANEL III
APPLICATIONS OF THERMODYNAMIC PRINCIPLES
TO METALLURGICAL SYSTEMS(continued)
Chairman

Dr.J.Alfred Berger
Chairman, Metallurgical Engineering Department
University of Pittsburgh

1.

The Use of Free Energy Balance in Process Analysis


by Dr.George R.St.Pierre, Professor Metallurgical
Engineering, The Ohio State University

2.

Application of Thermochemical Data During


Stainless Steel Melting
by Dr.C.W.McCoy, Senior Research Scientist,
Mr.A.F.Kolek, Metallurgical Technician, and
Dr.F.C.Langenberg, Director, Process Research
and Development Laboratory, Crucible Steel Company
of America

3.

Thermodynamics of Materials at High Temperature


and Low Pressure
by Dr.Paul Gilles, Professor, Department of
Chemistry, University of Kansas

4.

A Summary of Experiences in Applying Thermodynamics


to Metallurgical Processes
by Dr.G.R.Fitterer, Distinguished Professor,
Metallurgical Engineering and Director of the
Thermodynamics Center, University of Pittsburgh

,o.

Discussion

PANEL III
Dr.Berger, Chairma

It is my pleasure to open this the third panel of the Conference


on Applications of Fundamental Thermodynamics to Metallurgical
Processes and introduce the first speaker. He is Dr.George
R.St.Pierre, Professor, Metallurgical Engineering, Ohio State
University who will present a paper entitled 'Free Energy Balance
in Process Analysis' . Dr.St.Pierre:

THE

USE OF FREE ENERGY BALANCES


IN PROCESS ANALYSIS
by

G.R.St.Pierre
The Ohio State University
INTRODUCTION
At the 1964 Symposium on Process Simulation and Control held in
conjunction with the Annual Meeting of AIME, the author suggested
that free energy balances could be as useful to the process analyst
as enthalpy balances 1. The intent of this paper is to extend the
discussion of the free energy balance as a representation of
metallurgical processes. The initial part of the paper will review
the criteria for spontaneous changes within reacting systems. The
latter part of the presentation will outline the methods of computing
free energy and entropy balances for processes .
Although the Second Law of Thermodynamics can be stated in a
number of equivalent forms, the best known of which are the Clausius,
Kelvin-Planck and Caratheodory statements 2, the following choice has
been made for the purpose of this presentation: 'Within an isolated
system, internal energy must remain constant and entropy must tend
toward a maximum value' . The previous statement is a combined statement of both the First and Second Laws. To use the principles
contained in it, the definitions of isolated systems, internal energy
and entropy have to be fully understood. An isolated system is one
which can interact in no manner whatsoever with its surroundings.
Inte rnal energy, E, is all of the energy within a system other than
its stored potential energy relative to a force field and its motion
as a body. Entropy, S, is defined in terms of the addition of entropy
to a system:

dS = - (reversible).
T

Thereare other possible means of

defining entropy which are equivalent to the given definition 3.


Furthermore, the Third Law allows for the calculation of the absolute
ntFopy of a system. Since the entropy of all equilibrium systems at
1t temperat ure of absolute zero is zero, the entropy of a system in
1t11y particular state is the sum of the q/T' s required to move that
r~yl'!t m fr m absolute zero to the final state along a reversible path.
Tt H l ar, th r for , that 1/T is the integrating factor for the
m ntH. Th first quantitative combination of
wutilly wrttt n in the form,

G.R.ST.PIERRE
dE = TdS - PdV + 2,pdn .

(1)

<(/

Equation (1) allows for the calculation of the change of the internal
energy of a system along a reversible path. The final term 24dn.
includes all mass and work terms other than work of expansion. If
the system is open with respect to addition or removal of mass, then
24dn would include a series of terms:

for each component of the system.* Similarly, electrical and magnetic


work terms must be included in equation (1) when they are involved in
the system. For a system closed with respect to mass and for which
only expansion work is being done, obtains,
dE = TdS - PdV

(2)

As previously mentioned, equations (1) and (2) can be applied to


compute the change in internal energy of a system moving along a
reversible path. That is:
(3)

However, internal energy is a state function and (E 2- E1) is in no


way dependent upon the manner in which the system moves through
intermediate states between state 1 and state 2. Although the
integrals in equation (3) must be taken along reversible paths, the
obtained value of (E 2 - E1) holds for any path along which the
system moves between the two states.
In a similar manner, it follows that
*The partial derivat"ive, (oE,/clni)S,V,N' is designated as the
chemical potential of component i and usually given the symbol, i.
It can readily be shown that:

"i (~ 0

S,V,N

~~i)

S,P,N -T

":w

~:,)

E,V,N

~:J.P.N

USE OF FREE ENERGY BALANCES IN PROCESS ANALYSIS


S2 -

21

s 1 = f - dE
1 T

F2 - F1 =

A2 - A1

VdP -

=- f2
1

,.2P
+ j

1 T

dV

s21 SdT

PdV -

.r21 SdT

(4)

(5)

(6)

The Gibbs Free Energy, (E + PV - TS) , has been designated by F and


t he Helmholtz Free Energy , (E - TS), has been designated by A.
Again, all of the integrals in equations (4) through (6) must be
t aken along reversible paths and the equations are written for
changes closed with respect to mass transfer and limited to
expansion work only . If mass transfers and other work terms are
involved, each equation must include 2:.Ji dn as indicated in
equation (1) .
CRITERIA FOR SPONTANEOUS CHANGE
Included within the statement of the Second Law already given is
a cr i terion for spontaneous change--within an isolated system,
entropy must tend towards a maximum. From equation (4) , it is clear
that. within an isolated system entropy cannot vary along a reversible
path through which the system may move. Along any spantaneous
irrever sible path , the entropy must increase in order to satisfy
the Second Law restriction. The interrelationships of the other
defined state functions with entropy allow for the establishment of
n whole set of criteria for ' spontaneous change within systems
HUbjected to restraints other than constant volume, energy and mass.
This can be illustrated by considering the characteristics of Gibbs
fr e ener gy change within a system which undergoes no net change in
pr ssure , t emperature , or mass. Such a system is not necessarily
IHolated; i n fact, generally, it exchanges heat and work with its
Hurroul'.ldings in passing from one state to another. One can allow
11u h I\ system to go t hrough a dt'O-step i nfinitesimal change which
11 pr H nts no net change i n pressure and temperature.
Step A is a
11p nl.1Ln na chnng at constant E, V and mass ; Step B is a reversible
11 t, urn o t ht HY t m t
its initial T and P at const ant mass. Then,

lii

G.R. ST. PIERRE


(7)
(8)

Since,

the sum of dFA and dF 8 is


(9)

From the Second Law, dSA must be >o and disallowing negative
values of T it follows that dFA + dF 8 <o for any spontaneous change
which does not result in a net change in P and T4 Similar arguments
may be made to establish that :

(DE)s,v <o

(lOa)

(d.H)P,S <o

(lOb)

(dA)T,V <o

(lOc)

(dV)E,S

<o

(lOd) .

Therefore~ to consider the spontaneity of a proposed process one


must consider the constraints placed upon the system. In many real
processes, the entropy of the d~fined system may decrease during a
spontaneous change or the free energy may increase. Such Qbservations are entirely consistent with the Second Law since it is only
when the system is subjected to ~rticular constraints that the
'
spontaneity criteria pr~viously deduced
may be applied,

THE FREE ENERGY BALANCE

To illustrate the calculation of 6 F for a particular change in


state of a system, the following case has been se l ected:

USE OF FREE ENERGY BALANCES IN PROCESS ANALYSIS


Fe0(1200) + co(1200) _, (1 - X)Feo(1200) + (1 - ,X)CO(l200) +
XFe(1200) + Xco2(1200)

(11)

Equation (11) represer,ts a change in state of a closed system


originally containing one mole of CO and one mole of FeO at 1200K
and 1 atm pressure and finally in a state containing (1 mole
FeO, (1 - A) mole CO, Xmole Fe, .and Xmole co 2 all at 1200K and
1 atm total pressure. The free energy change may be expressed in
terms of the extent of the reaction,
as follows:

x)

X.

l'iF =

(1

-A) Fe0( 2 ) +

(l -

x)

Co( 2 ) + ~e( 2 ) + ;t'.co ( 2) 2


Feo(l) - co(l) (l 2)

The superscripts (1) and (2) designate the initial and final state,
respectively. From the definition of activity, 1 1 +RT ln a 1,
equation (12) becomes
6F =

x {i4e + go2 - ~eo - go}

+ RT { X1nX + (1 -

x)

ln ( 1 - A)}
and

f~r

(13)

the particular temperature:

6F = +1940,X + 5500 {

X1ogX

+ (1 - ,X) log (1 -

x)}

(14)

Figure 1 shows the values of 6F for Xequal to values between 0 and


1. As can be seen, 6F exhibits a minimum at about X = 0.3. The
minimum can be determined analytically by setting o6F/oX equal to
:r. ro:

~~

+1940 + 5500 {log I

Arn=

0.31

\l

(15)

G.R.ST.PIERRE
+2000

+ 1800
+ 1600

F.O +C0-(1-x)FeO +(1 -x)CO +xFe +xC0 1

+ 1400

6F + 1940x + 5500{ xiov+U-x)loo(lx)}

(llOO) (1100,

(llOO)

(llOO) (1100)

(1100)

+1200

+ 1000
+800
+600

.,

+400

ii:
0
..J

+ 200

<I

<(

..:-200

~il 1

-400
-600

1;;::

''"

-800

~,

1000

'iii
0

.2

.4

.6

X, EXTENT of

REAC~ON

.8

1.0

Fig.1
The value of X at the minimum satisfies equilibrium within such a
system at the particular temperature and pressure.* However, it is
important to note that there are many states other than the
equilibrium state for which6F is negative and the criterion of
spontaneous change is satisfied. For example at X = 0. 2, 6F = -812
and at X = O. 4, 6F = -849. Both of these states are energetically
more favorable than the initial state; however, the system could
never move to X = 0.4 from its initial state because that would
require that the system pass through its equilibrium position. This
illustration serves t Q.. point out ' the existence of lower free energy .
states for the system other than the equilibrium state. To determine
whether or not a particular change is possible it is necessary to
establish that the final state has lower free energy than the initial
* That is, it satisfies solution of the equilibrium constant
expresstion, K = Pc 02/ Pco. 5
2110

USE OF FREE ENERGY BALANCES IN PROCESS ANALYSIS


state and that the system does not pass through its equilibrium state
in moving from the initial to the final state. In the given
illustration , the change to = 0.2 is possible; however, the change
to X = 0.4 is not possible.

The general procedure for calculating the ~F of a system may be


given as follows. For any reactor system for which input mass equals
output mass, the ~F over any time interval is given by

(16)

where the first two summations must be taken for each component
within each input phase and the second two summations must be made
over each component of each output phase. Although equation (16)
appears complicated, it is quite simple to apply. One of the most
import ant ways of simplifying the computations is to adopt standard
states as follows:
( 1 ) For condensed phases, the standard state of each component is
the pure component in its equilibrium structure at one atmosphere
pressure and the temper ature in question;

( 2) For gaseous phases, the standard state of each component is


the component at unit fugacity at the temperature in question.

The terms,

(F~ ~ H~)

, have been tabulated for most of the

uhsto.nc s of interest to metall urgists.


11or

m st

uur

1(118

The following expansion

nts this can bo considered equivalent to the


utlll{) ph r pr Asur
Wll

G.R.ST.PIERRE
Illustrates the simplicity of this function:

r cop

-T

cpdT
0

J [J
0

-dT
T

T
Cp0 dT

dT

(17)

T2

A11..lloul{h C~ f or a given substance may undergo one or more discon-

l, l1111lt.l1H between o and T, the function,

F T

tt0)
0

does not

lmn1 w 1 Lh free energy change for each phase transformation is zero.


'1'111 1' 1 for it is only necessary to know C~ for the substance over the
whe1l1 L< mp rat ure range and it is not necessary that the enphalpy (or
1 11!.rn~1y)
hanges of the transformations be known.
Thi L r ms, (FT - F) i in the summations are also simple to
v1 I 111 L in principle, although in this case the data are not so
111111plo l,o. l~rom the definition of activity, it follows that:
1

(18)

111111' I.hi

l{lLl:I

OU

phases, a 1 may be replaced by P 1 , the partial pressure

1>mp n nt. For the condensed phases, a 1 must be known from


1 K1~ 1 I11111111, or
stimat ed in some reasonable fashion.
11! I.lit

A11.

,I

'1'111 l.1 r11111 ,


1 , are the standard enthalpies of formation of the
1111111u111111l.11 ut Lh r f rence temperature, oK in this case . . The data
r'111 1. 111 1111 t,1 J'IJJH 11r available for most of the substances of interest.

( 11,

111 rn 11.1t1 11 l11Htu11 H, th op rational calculations of equation


1111 11 ht r1 clu <I by ro
gni:t.ing
rtain convenient groupings of

USE OF FREE ENERGY BALANCES IN PROCESS ANALYSIS


terms . However, equation (16) is the most general operatfonal
procedure and, when uncertainty exists, it should be follow~d.
Table I presents input-output statements for a simple reactor
system which might be part of an iron oxide reduction system or
characterize in some way the properties of the 'isothermal zone' of
an iron blast furnace. 6 The equations for 6F and 6H of the system
are given as follows:
6F = 1940 - 9640n 1 + 5500 { (2n 1 + n 2 - l)log Pco +
(1 - n 1) log Pc
6H

02

(19)

-4050 + 40, 290n1

(20)

where
2n 1 + n 2 - 1

1 - n1

If it is proposed that 6H

O (no heat addition to the reactor


sys tem), .then from equation (20), n 1 = O. 100. Then 6F becomes a
f unction of n 2 only,

+0.9 log-_9_)

n 2 + .1

n 2 +. 1J

6 F = 976 + 5500 { (n 2 - .8)log n


and 6 F can
Figure 1.
'
conditions
values may

(21)

be plotted against n 2 in the same manner as was done in


However, only values of n 2> 0.8 are possible by the
>
of the problem because (2n 1 + n 2 - 1) - 0. The following
be calculated from equation (21) taking n 1 = O. 1:
0.8

0.9

1. 0

1. 45

1. 90

2.80

+976

+201

-274

-1544

-2324

-3324

l'or nn 111 l t1 0. l har


ll ow v<r, I 'L

:Iii rn.

' L 11

of n 1 - 0.1 and n 2 = O. 95, both 6H and 6F are


n t b on l ud d t ho.t the process is feasible

G.R.ST.PIERRE
TABLE I

Input-Output Description of a Simple Reactor System


Output

Input
1 mole Fe

1 mole FeO at 1200K


n 1 mole C

at 1200K

n 2 mole CO at 1200K

at 1200K

(2n 1

n 2 - 1) mole CO at 1200K

(1 -

n 1) mole
2n 1

Pco

atm

Pco

co 2

at 1200K

n2 - 1

n1 + n2

1 - n1
Pco 2 = - nl + n2

llF 1940 - 9640n 1 ' 5500 { (2n 1 + n 2 - 1) log Pco + (1 - n 1)


log

l'iH = -4050 + 40, 290n 1

'

co,}

USE OF FREE ENERGY BALANCES IN PROCESS ANALYSIS


for such an initial charge or even feasible when n 2 = 1.0 for which
6F = -274 until it is established that the system does not have to
pass through lower free energy states to arrive at the final state.
Consider the final state for n 1 = 0.1 and n 2 = 0.9 as the initial
state for the following transformation :
mole Fe (1200)
0.1 mole co (1200)
0.9 mole co 2

(1 - y) mole Fe(T)
y mole FeO(T)
(0.1 + y) mole CO(T)
(0.9 y) mole C0 2 (T)

(1200)

When 6F is written as a function of y for 6H = 0, it is found that


6F is negative for positive y and that the system would settle into
a minimum near y = 0.6. Not until the initial charge to the reactor
contains n 2 ? 2. 80 does the process become feasible . This can be
solved analytically by writing out the reactor system in terms of
c

Maonetite (s) +Gas

F g. 2.

Phn3

Oiugrnm of the System Iron -Carbon-Oxygen at


1 atm

tooo 0 c

and

G. R. ST. PIERRE

general reaction parameters. An alternate method can be based on


Figure 2 which shows the phase boundaries of the ternary Fe-C-0
system in the temperature range of interest7. A charge mixture of
1 mole FeO, 0.1 mole C and 0.9 mole CO corresponds to NFe = 0.263,
Ne= 0.263 and N0 = 0.474 atomic fractions of each component.
Reference to Figure 2 shows that such a composition comes to
equilibrium in the three-phase field, Wiistite-Fe-Gas, as marked by
point X on the diagram. Additional CO in the charge moves the system
into the two-phase field, Fe-Gas, at the point Y.
It is of interest to compare the previous illustration of the
reduction of FeO in the absence of charged carbon to the present case .
Without C, 3,22 moles of CO were required and the free energy dissipation per mole of FeO reduced was 2840 calories. With 0.1 mole C
charged, 2.80 moles of CO were required and the free energy dissipation was 3324 calories. Clearly, there seems to be a marked
advantage indicated by this simple calculation.
DISCUSSION AND CONCLUSIONS
The availability of the thermodynamic properties of metallurgical
substances makes it quite feasible to compute free energy balances
for even the most complicated metallurgical processes. Although the
calculations are more involved than enthalpy balance calculations,
they can be performed with considerable speed especially if the
arithmetic is done by machine. The present paper is an attempt to
point out some of the interpretations that may be possible when such
balances are prepared. There is no doubt that processes which
dissipate large amounts of free energy should be faster than those
which dissipate smaller amounts of free energy. Furthermore, it is
possible that when such balances are prepared for real processes,
glaring inconsistencies in the process descriptions may be revealed.
Finally, the consideration of proposed process changes could be
enhanced by examining
the eff~qt the change might have on the free
.....
....:
energy terms of the system. Unfortunately, the free energy balance
remains untried in metallurgical analysis and its possible value is
unknown.
.._

:,IJ(J

USE OF FREE ENERGY BALANCES IN PROCESS ANALYSIS


REFERENCES
1.

G.R.St.Pierre, 'Inequality Statements for Some Metallurgical


Processes', Proceedings of A.I.M.E. Symposium on Process
Simulation and Control in Iron and Steelmaking, 63 F.dited
by J.M.Ays and H.L.Bishop, Gordon and Breach (1966).
0

2.

M. W.Zema.nsky, Heat and Thermodynamics, 148, McGraw-Hill Book


Co. (1957).

3.

J.D.Fast, Entropy, McGraw-Hill Book Co. (1962).

4.

L. S. Darken and R.W . Gurry, The Physical Chemistry of Metals,


187, McGraw-HU 1 Book Co. ( 1953).

5.

J.F.Elliott and M. Gleiser, Thermochemistry for Steelmaking,


1, Addison-Wesley Publishing Co. (1960) .

6.

Yu Shu-P'ei , 'A New Representation of the Blast Furnace Process',


Contemporary Problems of Me ta l lurgy, 98, Consultants Bureau:
New York (1960) .

7.

Based on calculations of A.Ghosh and G.R.St.Pierre .

47

APPLICATION OF THERMOCHEMICAL DATA DURING


STAINLESS STEEL MELTING
by

C.W.McCoy
Senior Research Scientist
A.F.Kolek
Metallurgical Technician and
F.C.Langenberg
Director
Process Research and Development Laboratory,
Crucible Steel Company of America, Pittsburgh, Pa.
I NTRODUCTION
Stainless steels are iron-base alloys containing 12 to 30%
chromium. These steels are normally classified as austenitic,
ferritic or martensitic. The major divisions are the 300 series
with nickel-stabilized austenite, the 400 series that are either
ni ckel-free or contain up to 2. 5% nickel, and the 200 series in
which some of the nickel has been replaced with manganese and
ni trogen. The carbon content of stainless steels is usually less
t han 0.12% and in many grades is less than 0.07%. Silicon is
normally present up to 1% maximum. Supplementary alloying elements
us ed in some stainless grades would include molybdenum, columbium,
titanium, copper and aluminium.
In 1964, over 1,500,000 tons
have been cast. The production
over two-thirds of the chromium
St ainless steels also represent

of stainless steel ingots will


of these ingots will have required
sold in the United States.
the largest single use for nickel.

Stainless steels are melted almost exclusively in electric


f urnaces. The most common melting practice is an oxidationreduction process using gaseous oxygen as the decarbu~izing agent.
A typical charge would be scrap (stainless and plain carbon), the
proper ferroalloys and slag-making materials. The charge is partially
or ,cpmpletely melted and oxygen is injected to remove carbon. The
xidation of carbon is accompanied by the oxidation of silicon,

chr om i um, manganese and iron. The oxides of these metals become
pnrt of t h slag. The partial recovery of chromium and manganese
. fl'O m t h R]ng by silicon during the reduction period (before slag-off)
l 11 11 . iH1ttry fol'
on mi u l op ration. Aft er slag-off or re ladling,
I. ht IH 1LI, tu 'fl11l 11 l11 !1' 1111 <1 1 o. R ond Hl o.g and t; apped.
VIII

C.W.McCOY, A.F.KOIEK AND F.C.LANGENIERG


The technical literature contains a great deal of information
concerning the conditions and factors affecting both the oxidation
and reduction periods of the two-slag stainless steel process. The
excellent work of the Union Carbide Metals Company deserves special
mention. The paper by Hilty, Rassbach and Crafts 1 is a good summary
of their work. This paper will not attemp to review all of the
technical literature on stainless steel melting but will offer some
industrial applications of thermochemical data in this area.
OXIDATION REACTION PERIOD
The thermodynamic relationship among carbon, temperature and the
other oxidizable elements in solution in the liquid iron during the
oxidation period of stainless steel melting is of great economic
importance. It is the guide for determining practical limits on
charge compositions to the furnace and blowing practices consistent
with practical temperature limits. Consider the two reactions
taking place. First, the primary reason for the oxidation period
is to remove carbon in solution down to some low level, i.e.,

Q + 1/2 02 (g) !:+CO(g)


Concurrently, other metallic elements,
slag components,

1
2

- M +-

y-

o2 (g)

(1)

M.

are also oxidized to form

~ - M 0
y
xy

(2)

In both the above equations, the underlined chemical symbol indicates


an element in solution in liquid iron.
The relationship between the concentrations of carbon and the other
oxidizable metallic elements at any given temperature can be obtained
by use of simple forms of the vant Hoff equation. Equilibrium
constants for the above react l'ons in terms of the activities, a, of
the elements in solution and the oxides formed and .the partial
pressures, P, of the gases are as follows:
(3)

p
02

1/2

APPLICATION OF

THERM~HEMICAL

DATA

and

K2

(4)

=----a X/Yp
M

1/2

o2

Substituting Henrian activity coefficients, f, and concentrations,


[ ], for activities of the elements in solution in the liquid iron,

(5)

[%] f

02

1/2

and

K2 = - - - - - - - -

(6)

[o/J,1] x/y f x/yp 1 /2


~

02

Under the particular conditions in actual stainless steel furnace


essentially the pure oxide is formed in the oxygen impingement reaction zone and~ 0 'ii' 1. P0 2 is the same for both reactions
x y
and during the rapid decarburization reaction, Pco is also assumed
to be approximately unity, neglecting any small amounts of co 2 that
may be present. Dividing K1 by K2, the ratio of metal to carbon
concentrations can be written as:
pr~ctice,

(7)
K2 f

x/y
M

or

ln (

fV
x/y

M.
_',b l

(8)

C.W.McCOY. A.F.KOIEK AND F.C.LANGENBERG

= -RTlnK,

But 6 F0

ln

ln

or

r~x/')
- -

+ ln

!'II

and by substituting,

(9)

re&

r~xl~
--

(6
=

H~

- 6

H~)

RT

C.fM
.f0)
( fCCG

(6
+

s~

- 6

s~)

C 0)

- -x ln ( f M.f. f
y
MMM

(10)

If the activity coefficients or their ratios are essentially constant


in the range of concentrations studied, the second, third and fourth
terms on the right side of equation (10) combine as a constant and
equation (10) reduces to the general form:

log

o/<Mx/~

-o/

(Constant) 1
T

+ (Constant) 2

(11)

This is the form of the original work done by et az 1 3 on the


relationship among chromium, carbon and temperature . Fortuitously,
the interplay among the activity coefficients and/or the metal-oxygen
atomic ratio are such that the simplified version of their results
is given by the equation:
log

%CR) = -13, 800


( %
T

a.76

(12)

where the equilibrium concentration of carbon and chromium in the


steels at oxygen saturation i~, expressed in weight per cent ancLT .
is the absolute temperature of the bath in degrees Kelvin .

. 252

TABLE I
Stmunary of Chemical Analyses and Measured and calculated Temperatures

for the Carbon-Chromium-Nickel System

Temperature (K)
Alloy Analysis (%)
Tes t

:..>

Sample

Cr

Ni

--

Tm
Cale. by
Eq. ( 12)

Cale . by
Eq. (13)

Tc
Cale. by
Eq . (15)

1863
1898

1861
1895

1861
1895

1861
1895

2
3

2
3

2
3

Tm,
Measured

Tc'

T -T
m c

T -T '
m c

T -T ''
m c

A-1

1
2

0.42
0.30

9. 16
8.96

B-1

1
2
3

0,33
0.23
0.16

8.95
8.48
8.06

4.13
4.18
4. 22

1879
1914
1951

1883
1919
1955

1871
1905
1941

1867
1901
1937

-4
-5
-4

8
9
10

12
13
14

1
2
3
4
5
6
7
8

0. 24
0.22
0 . 21
0.18
0.17
0.17
0.13
0.11

9 . 86
9.14
9. 12
8.58
8. 45
8. 42
7. 73
7.61

8.36
8. 42
8.37
8.40
8. 42
8.47
8. 68
8.61

1895
1892
1894
1928
1926
1927
1956
1955

1933
1933
1938
1949
1954
1953
1975
1994

1904
1905
1911
1921
1926
1925
1946
1965

1895
1896
1902
1913
1917
1916
1937
1955

-38
-41
-44
- 21
-28
-26
-19
-39

-9
-13
-15
7
0
2
10
-10

0
-4
-8
15
9

1
2
3
4
5
6
7

o. 18

9 . 50
9.16
8.97
8.59
8.57
8. 56
8.50

12.25
12.44
11.90
12. 16
12.33
12.37
12.09

1914
1916
1919
1939
1946
1947
1952

1961
1957
1954
1979
1988
1988
1997

1920
1916
1915
1939
1947
1946
1956

1908
1903
1903
1926
1933
1933
1943

-47
-41
-35
-40
-42
-41
-45

-6
0
4
0
-1
1
-4

6
13
16
13
13
14
9

B- 3

B-4

0. 18
0.18
0. 14
0.13
0. 13
0. 12

11
19
0

~n

1-C

1-C

~
o-:3

II
1-C

n
~
0

~
>

TABLE I (continued)

Temperature (OK)
Alloy Analysis (%)
Test

Sample

Cr

Ni

Tm
Measured

Tc,
Cale. by
F,q.

B-5

:"
.:.J.,

.:::.

B-6

0.19
9. 11
2
0.19
8.97
3 \ 0.16
8. 97
4
0.14
8.53
5
0.13
8.42
6
o. 12 8. 38
0.098 7.85
7
8 , 0. 095 1.75
9 . 0.090 7.58
1
2
3
4
5
6
7
8

o. 18
0.14
0.12
0.11
0.11
0.082
0.078
0.076

9.32
9.14
8.65
8.55
8.49
7.83
7.69
7.73

( 12)

Tc
Cale. by

Tc
Cale. by

(13)

Eq. (15)

Eq.

Tm-Tc

16.46
16.27
16.54
16.40
16. 50
16.56
16.62
16. 63
16. 52

1877
1882
1882
1907
1910
1909
1955
1955
1950

1949
1948
1968
1978
1986
1995
2013
2015
2019

1897
1895
1914
1924
1931
1939
1956
1958
1962

1880
1879
1896
1907
1913
1922
1938
1940
1944

-72
-66
-86

20.29
20.18
20.07
20. 02
20.05
20. 27
20. 21
20.11

1883
1885
1921
1921
1923
1948
1949
1950

1961
1987
1999
2009
2008
2035
2040
2044

1893
1920
1932
1941
1940
1965
1969
1974

1873
1898
1911
1920
1919
1943
1947
1951

-78
-102
-78
-88
-85
-87
-91
-94

-71

-76
-86
-58
-60
-69

Tm-Tc'

Tm-Tc''

-20
-13
-32
-17
-21
-30
-1
-3
-12

-3
3
-14
0
-3
-13
17
5
6

-10
-35

10
-13
10
1
4
5
2
-1

-11
-20
-17
-17
-20
-24

C')

"';;;:::"
0

.><

?"
:;i

;
~
:;i
C')

I
Q

APPLICATION OF THERMOCHEMICAL DATA


TABLE II

Average Difference between Calculated


and Measured Temperatures for the
Carbon-Chromium-Nickel System

Average of Tm-Tc (K)


Average Nickel Content (%)
Eq. ( 12)

2
-4
-32
- 42
-73
-89

0
4.18
8.47
12. 36
16.51
20.27

Eq . (13)

2
9
-3
-1
-18

-16

Eq . (15)

2
13
5
12
0
2

Alt hough equation (12) was derived from data obtained from
xper i ments with plain chromium steels, it was also applied to
Htainl ess steels containing nickel. Richardson 4 speculated that
high concent rations of nickel in steel would increase the carbon
nctivity coefficient. Fuwa and Chipman 5 later published data in
which t hey meas ured the effect of nickel on the activity coefficient
. o carbon for nic ke l contents between zero and 30%. Simkovich and
MoCoy 6 r 1l atod the Hilty experiments in systems containing up to
'0 u r 1 ut, 11 i It 1. Tnblos I and II show the results. The column

., 11>

C.W.McCOY, A.F.KOLEK AND F.C.LANGENBERG


entitled Tc shows the temperature calculated by equation (12). The
difference from the experimentally-measured temperature increases
as the nickel content increases, with the calculated temperature
always higher as shown also in Table II. The experimental measurements in both systems already incorporate the effect of the oxygen
on both the chromium and carbon activjties, the effect of chromium
on the carbon activity, and the effect of carbon on the chromium
activity. That is
f~.
f~, f~ and f~ are part of the
experimentally determined constant. The only difference between
the two is the presence of nickel in varying amounts. Therefore,
an', additional term, f~i, must be added to the ln fQ expression and
f~; added to the ln fer express~on. Fuwa and Chipman showed that
the interaction coefficient e ~ 1 = +0.012. Therefore:

tg,

tg,

log f~i = +0.012[%Ni]


and f~i represents the increase in the carbon activity coefficient
caused by nickel. Referring back to equations (10) and (12), this
relationship was used to correct for the change in the activity
coefficient to give the following final form: -

log

o/tCrj

--

( %

-13,800
T

(13)

+ 8. 76 + 0. 012 [%Ni]

Temperatures calculated by means of equation (13) are listed under


I
Tc in Tables I and II and are in good agreement with measured
temperatures for nickel concentrations up to about 12 per cent. At
16 per cent to 20 per cent nickel the difference between Tm and T0
increases noticeably. It i~ felt that at these high concentrations
of nickel, additional atomic interactions between nickel and chromium
become important, and corrections must be made for them.
The data in Tabl~ I can be -\Jsed to estimate this interaction-coefficient. Equation (10) can be rewritten as:
-13,800
T

+ 8.76 + o.012[o;.Ni]

x
y

log tNi

Cr
(14)

:,r n

APPLICATION OF THERMOCHEMICAL DATA

o/r
Differences between the experimentally measured values of log %Q
and those calculated by equation (13) are -x/y log f~i and the
slope of these values versus [o;.,Ni] is ye ~i = -0.004 as shown in
Fig.1 so that the final form of the relationship is:
-13,800
+8.76 +0.016[o;,Ni]

(15)

0
~

.,.01101.,.

L'...._j

-0.050

"'

.!:!

<I

-0.100

- 0. 140
0

10.0

5.0

15.0

20.0

WEIGHT PERCEN T NI
Fi~ .1 .

Effect of Nickel on the Activity Coefficient of Chromium

'l'ho est imated value of e ~~ depends of course on the value of -.


l''or t he formation of CR 3o4 proposed by Hilty, Forgeng, and
l~olkman 7, e ~~ = - 0,005 .

Tables I and II show the good agreement of the measured


t. mporat ur es and those calculated by Equation (15,) at all nickel
1< v ls.
J\11 ~h

r way of expressing the nickel effect observed on the chromium


1rb n relati onship i s by the temperature equivalency. It was found
Lhitt 'Lhe addition of 1 per cent nickel resulted in the same change
t n 1.hu log of t he chromium-carbon ratio as a 4. 21 K increase in
t.c mu ro.t ur in the experi ment a l range so that an equally useful
l'orm 0 th r la.ti nship was :

. 11'/

C.W.McCOY, A.F.KOLEK AND F.C.LANGENBERG

log

%Cr)
( %Q

-13,800
8
= T + 4. 21 [o/JU] + ' 76

(16)

The authors have recently begun similar work on the manganesecarbon-temperature and the manganese-nickel-carbon-temperature
relationships which are important in the Type 200 stainless steels
where manganese is also oxidized concurrently with carbon during the
oxidation period.

EXPERIMENTAL PROCEDURE FOR Mn-C-Ni TEMPERATURE STUDIES


Laboratory heats were made in a 30-pound capacity induction
furnace using magnesia crucibles. Charge materials consisted of
electrolytic iron, manganese, and nickel, and graphite.

!ii
f[

A series of heats with random combination of carbon levels up to


1.0 per cent and manganese levels up to 12.0 per cent was made and
blown with oxygen as in the chromium experiments. In the second
series, the carbon and manganese levels were again chosen randomly
and nickel was varied up to 6%.
Temperatures were measured with platinum-platinum + 10% rhodium
thermocouples which are estimated to be accurate to 7C.
The procedure in each heat was to melt the charge and hold
temperature constant for 1/2 hour to permit equilibration between
slag and metal. Oxygen was then injected into the bath through a
3/8 inch diameter silica tube at a pressure of 4 psig for
approximately 10 seconds. The temperature of the bath increased
due to oxidation and was held constant at this new level for at
least 15 minutes. The bath w~s then sampled with a silica tube
and aspirator bulb.
This procedure of holding a~9 blowing with oxygen was usually~ .
repeated several times for each heat. Temperature was held constant
between oxygen injections by thermocouple readings every 2 minutes
and minor power input variations to the furnace.
Table III and Fig.2 show the results of the study of the
manganese-carbon system where MnO is the oxide formed. According
258

TABLE III

SU=::ary of Chemical Ana lyses and Meas ur ed and Calculated Temperatures


f or t he Carbon-Manganese System
Temperature ( uK)
Alloy Analysis %
: est

~mp le

Mn

Tm
Measured

Tc
Cale. by
Eq. (17)

Temperature
Deviation (K)
Tm-T c

~~

Y.A

1
2

0.30
0.23

3.99
3.47

1916
1927

1901
1933

15
-6

1i!B

1
2
3

0.77
0.63
0. 44

10.70
10. 01
9.59

1908
1958
1983

1912
1945
2029

-4
13
-46

MC

1
2
3

0. 59
0. 46
0.32

7. 81
6.97
6.77

1926
1972
2002

1900
1933
2021

26
39
-19

1-i .

MD

1
2
3
4

0.95
0.79
0. 71
0.59

10.60
9.88
9.64
9.09

1841
1880
1911
1927

1859
1886
1906
1937

-18
-6
5
-10

ME

1
2

0.17
0.11

2.22
1.49

1889
1908

1896
1906

-7
2

MF

1
2
3
4

0.51
0.47
0.37
0.30

6.26
5. 88
5 . 51
5.09

1880
1908
1930
1941

1881
1886
1929
1962

-1
22
1
-21

MG

1
2
3
4

0.59
0.54
0.46
0.41

7. 26
6.97
6.42
5.82

1Cl72
1902
1916
1922

1882
189.4
1913
1917

-10
8
3
5

~
~
~
::z::

H
(")

~
0

~
>

IUli IT
~

1 07 c_: C;:~~~ i::;.:.._TSBS ~ Ji?G-~e:'. ~ :z:: ~~=- -

:a=- r;;:e ca::-X:: - Ma.-.e;c.eese-~::.cte :

Te~

'Ie;:perattire ( :K}
A:lqy Ana lys is ( %}

Sac11le

Mn

Ni

Ta
Measured

"

Tc
Cale . by

T ,
Ca le~ by

( 17)

Eq. (18 )

F.q .

Sys::eTemperature Dev iation


Tm-Tc

Tm-Tc'

~
~

1
2
3

0.67
0. 59
0. 35

9.07
8 . 57
4 . 58

2.19
2.1 9
2.21

1878
1894
1890

1905
1922.
1897

1891
1906
1885

-27
-28
-7

-13
-12

l\

2
3

0.60
0. 53
0.43

7.09
6.76
6.27

2.10
2.11
2.15

1853
1883
1900

1872
1891
1924

1857
187'7
1907

-19
-8
-24

-4
6
-7

>
'.'9

1
2
3 .,.,

0.62
o. 45
0. 27

7. 35
6.48
4.12

3. 67
3.80
4.14

1847
1894
1905

1874
1921
1935

1848
1894
1906

-27
-27
-30

-1
0
-1

1
2

0.18
0.14

2. 53
2.16

4.07
4.02

1889
1909

1915
1938

1889
1900

-26
-29

0
9

MN

1
, 2
3

0.62
0.57
0.40

7.65
7.14
5.80

6. 10
6.31
6. 38

1835
1858
1884

1833
1886
1922

1840
1845
1875

-48
-28
-38

-5
13
9

MO

1
2
3

0.20
0 . 19
0.18

2.86
2. 82
2.48

6.08
6.10
6. 08

1878
1883
1878

1919
1928
1910

1878
1887
1869

-41
-45
-32

0
-4
9

1
2

0. 44
0.38

5.59
5.31

6.19
6. 24

1848
1878

1889
1913

1850
1872

-41
-35

-2
6

MK

ML

g
0

MM

MP

(')

~
~

::>=:

>
~
~
(')

I
c;'l

APPLICATION OF THERMQCHE},fICAL DATA


100 , - - - , - - - - , - - - - , - - - - , - - . - - ,
80
60

%Mn -6440

40

loQ " g_

=-r-+1.51

20

11~10
..

6
4

1.___._____._ __.._ __.._


4.9

5.0

5.1

5.2

5.3

_.__~

5.4

5.5

I
4
TKELVIN X IO

Fig . 2. Effect of Temperature on the Manganese-Carbon Ratio - K

Pct. M!!
Pct.

to equation (11) the results can be expressed in its most simple


for m as:

log (:)

-6,449

(17)

- - - + 4. 51
T

Table III also shows the good agreement between the experimentallymoasured temperatures and those calculated by equation (17) .
Preliminary experiments in the manganese-carbon-nickel system are
s hown in Tables IV and V. As in the chromium-carbon-nickel system,
H gnificant deviations from the behavior of the nickel-free system
wor noted, as seen by comparing the measured temperatures versus
Lh temperatures calculated by equation (17). Again in this study,
111:1
of the Fuwa and Chipman data 17 for the correction of the carbon
. u L i vi~y coefficient caused by nickel according to equation (10)
m1Lcl the two sets of data consistent:

o/.tMTI) --6,440
( .%f
T

4.51 +

o. 012

[o/<.Ni]

(18)

C.W.McCOY, A.F.KOLEK AND F.C.LANGENBERG


TABLE V
Average Difference between Calculated and
Measured Temperatures for the
Carbon-Manganese-Nickel System

Average Nickel Content %


Equation (17)

Equation (18)

2.16

-19

3.94
6.18

-28
-38

-4
+l
+3

To date, the data are neither precise nor complete enough to


show any interaction effects between nickel and manganese as in the
chromium study, Equation (18) can also be expressed in a useful
form using the temperature equivalent of th~ added nickel as follows:

4.51

(19)

In both the preceding studies, the authors recognize the presence


of manganese volatilization in the oxidation reaction zone. Such
behavior has been reported by Linchevskii and Samarin 8 in their
study on the solubility OL oxygen in manganese containing liquid
iron, and by Fuwa and Chipman5 in their work on the activity of
carbon in liquid-iron alloys. However, this mechanism for manganese
removal from the bath is an integral part of the actual stainless
steel manufacturing process which is apparently reproducible, at
least on the laboratory scale.
REDUCTION AND FINISHI.NG

REACTIO~

PERIOD

Two additional types of slag-metal systems exist in the reduction


and finishing period of stainless steel melting. The first is that
developed during the oxidation period and altered by reduction of
the oxides during the reducing period up to slag-off or reladle,
and the second is the slag-metal system existing during the finishing
(1 2

APPLICATION OF THERMOCHEMICAL DATA


period up to tap. Very little study has been given to the slag-metal
equilibrium conditions during these periods with the exception of
studies on chromium distribution.
The partial recovery of chromium before slag-off from the slag
formed during the oxidation period is absolutely necessary for an
economic operation. This is accomplished by reduction of the
chromium oxide with silicon. The commonly used silicon additions
contain chromium and silicon in varying concentrations. Data from
small-scale experiments and production fUrnaces indicate that the
sig nificant factors influencing the :eduction of chromium from the
slag are temperature, slag basicity, slag fluidity, chromium and
silicon content of the metal, and mechanical factors such as reladling
and other methods of stirring1, 9,1o,11. 12. Hilty and co-workers 1
developed the following approximation from theoretical considerations:
[%Cr) 3
Wt . pct. Cr in slag after reduction a [ [%Si] 1. 75
(20)

where the brackets indicate the equilibrium weight per cent of


chromium and silicon in the metal and V is the slag basicity defined
at wt % Cao + wt % MgO I wt % Si0 2.
EXPERIMENTAL PROGRAM ON SLAG-METAL EQUILIBRIUM STUDIES

A laboratory experimental program was conducted at the Crucible


Steel Company to determine the equilibrium chromium concentration in
slags as influenced by temperature, slag basicity, and the silicon
and chromium contents in the metal in those ranges applicable to
low-carbon stainless steel reduction period conditions. 13 . The
general relationship between the weight per cent chromium in the
Hlag and the independent variables was developed by expressing the
chromium-reduction reaction by silicon without specific reference
t .tne nature of the chromium oxides in the slag.
(CrlCOY) slag+

[Si] metal::; x[cr] metal +

'.O:!

2 (Si0 2)

slag

(21)

\
Variable Speed Elect ric Molar (Riohl Ar191e Drivel

Quartz

Tube

()

:El

==
(')

Pyrex. Gloss

Sholl

1ef >

>:g

Ceramic

Top

:'.:>

,.;:..

.L'
94

t"'

~
>:g
()

~
Tronsile

L7
--..;:,,,r

Pedestal Soldered to Sholl

>--- - - -- -- - -- --

Fig. 3.

- - -- - - -- 4 2 - f -

Rotating Furnace Construction

I
Cl

Framini;i Members lSleel)

- -- - - -- -- - - -- --

--I

APPLICATION OF THERMOCHEMICAL DATA


The equilibrium constant for equation (21) is shown in equation
(22) .
y
x

a [er] a(Si0 2)

(22)

a(Cr xOy) a [Si]


If we assume that (1) the activity coefficients of chromium and
silicon in the metal are essentially constant over the range of
concentrations studied in Ref .13, that (2) the activity coefficient
of chr omium oxide is also constant, and that (3) log activity of
silica varies linearly with the slag basicity, then an expression
can be written for the weight per cent chromium in the slag which
has the following form :
B

log (%C)slag =A + - + C log [%Cr]metal


T

+ D log [<y,Si] metal


+ E log V

(23)

wher e A, B, C, D and E are constants, T is the absolute temperature,


and V is the slag basicity.
The slag- metal equilibrium studies needed to develop the constants
in equation (23) were performed in the rotating furnace shown in
Fig . 3 whi ch prevented slag-refractory reactions. The metal chemistries
were l imited to different randomly selected combinations of chromium
(11 to 18%) and silicon (0.10 to 0.80%) at a fixed low carbon content
f 0.06%. The s l ags were of the ternary CaO-Si0 2-15% Al 2o3 system
wi th CaO/Si0 2 rat i os from 0.3 to 0. 9. The slags and metals were
< r1u i librated at temperatures from 2780 to 3140 F.
Typical concent,ruUon-tim curv s for heats approaching equilibrium from both sides
1u1 1-lhown in l' g. 4 wh ich is only ill ust r ative of the time and
1 on < nLr1LLio11 h m11{ s inv lv d since te mperat ur es and compositions
1t r 1

<111' 1'1 11 nL.

C.W.McCOY, A.F.KOIEK AND F.C.LANGENBERG


...--.---,...----.---........--....-----.1.20
1.00

W20.o.

rf

i 1.5.0 f~x...
. . ~----..
X~ x_xJL... .\

Ill

J
.5

10.0

r_l,,.--1-

'-t---t-

--'CrJ Heat

v
7B

0.80

-~ 55

-_."le,...Cr,__-

0.40~

5.0v

o0

0.20

30

60

90

120

150

180

210

240

Time , Minutes

Fig.4.

~I

Typical Slag-Metal Equilibriation Runs

The experimental data shown in Table VI were treated by


statistical methods and the resultant equation for the equilibrium
log (%Cr) in the slag was:
log (%Cr)

8866
4. 887 - - - + O. 340 log [%Cr J
T

- 0, 178 log [%Si] - 1. 721 log V

(24)

where
T = absolute temperature in degrees Kelvin
[rt.Cr] = equilibrium weight ,per cent chromium in the metal
[%Si] = equilibrium weight per cent silicon in the metal
V = slag basicity expressed as wt %Ca0/wt o/J)i0 2.

'
The agreement between the experimental
and the estimated values
of chromium in the slag predicted from equation (24) i~ shown in
Fig.5.

:~on

TABLE ' VI
Summary of Chemical Analyses, Temperatures and Calculated Final Slag Basicities
Chromium Slag-Metal Equilibrium Study

Final
[er]
Heat
in
Metal
(%)

~-=

....::

5S
5V
5W
5Z
6A
60
6E
6G
61
6J
6K
6L
6Q
6U
6P

11. 95
11. 05
11.14
11. 70
11. 74
11. 71
11. 66
11.30
11.81
11. 73
10.97
16.24
17.91
16.65
17.32

Final
[si]
in
Metal
(%)

[cao]
in
Slag
(%)

0.39
0.09
0.11
0.52
0.52
0.27
0. 32
0.12
o. 29
0.28
0.09
0. 27
0.80
0.16
0.32

37
30
35
30
35
30
40
30
50
50
30
35
30
30
30

[A1203J
in
Slag
(%)

15
15
15
15
15
12
12
12
12
12
12
12
12
12

(CaF 2)
in
Slag
(%)

3
3
3
3
3
3
3
3
3

(Si0 2 )
Originally
in
Slag
(%)

Final
(Si0 2)
in
Slag
(%)

Final
(Cr)
in
Slag
(%)

Temperature
(OK)

63
55
50
55
50
55
45
55
35
35
55
50
55
55
55

50.3
42.8
44.9
47.7
46.0
46.8
45.6
45.2
43.6
43.6
42.4
37.8
49.3
47.1
44.1

10.0
14.9
10. 2
7.6
5.4
10.0
5.1
13. 0
4. 1
4.1
14.0
15.6
8.5
19.8
9.8

1839
1872
1839
1855
1805
1900
1878
1864
1866
1865
1914
1972
1814
1814
1889

Calculation,
Final
Basicity

>

1-t

a
~
1-t

0. 71
0.564
o. 63
0.582
0.715
0.56
0. 75
0.539
0.89
0.89
0.595
0.745
0.524
0.37
0.637

i
~

==

1-t

~
~>

C.W.McCOY, A. F.KOIEK AND F. C.LANGENBERG


25.0 ...----.---.-----r---....--....,..------,

c
G>

:. 20.0
CJI

iii 15.0
.5

i 10.0

Cr = 11.0 - 18.0%
Si = 0.10 - 0 .80%
Temp = 1800-2000 K

J:!

5.0

u"

= 0.4- 0.9

<t

5.0

10.0

15.0

20.0

25.0

Estimated Chromium in Slog , Percent

Fig.5.

Actual versus Estimated Chromium Contents in Slags.

The experimental slag basicities were low, varying from 0.4 to


0.9 as compared with the usual greater-than-one values in actual
stainless steel production. To check the applicability of the
estimating equation to higher basicities in actual slag systems of
a more complicated nature, data for the finishing slags on 80-ton,
Type 430 stainless steel heats were used . Here the slags contained
MgO , MnO, CaF 2, and FeO in addition to the Cao , Si0 2 and Al 2o3 .
One would expect at this stage, because of the longer slag-metal
contact time and the intimate slag-metal mixing action of the
stirring and tapping operations , the closest approach to chemical
equilibrium found in actual stainless steel production. Using the
final slag basicities expressed as (%Cao + %Mg0)/%Si0 2, the final
metal compositicm.s, and the tapping temperatures shown in Table VII,
the slag chromium levels were calculated from equation (24) and
compared with the actual valueey shown. There is fairly close
agreement at these high basicity levels, with the actual value somewhat higher than the estimated one . This is to be expected.in view
of the time needed tb achieve equilibrium, as indicated in Fig.4 .

...

The next step is the possible use of the equilibrium estimating


equation in measuring the amount of deviation of a flush slag in
an actual practice from the equilibrium value and the evaluation of
the economic incentive for increased efforts to improve chromium
recovery at this stage of . ~roduction .

.on

APPLICATION OF THERMOCHEMICAL DATA


For a specific set of temperature and slag-metal compositions, it
is possible to calculate the minimum chromium level in the slag that
can be expected. This can be compared with actual practice values
for consideration of practice changes. As an illustration of such
a study, data equivalent to that taken in the laboratory study were
obtained at slag-off on twelve special-practice, 80-ton, electric
furnace , stainless steel heats and analyzed by the same statistical
methods . A similar estimating equation was derived as seen below:
log (%Cr)

949

1.118 + -

o. 550 log [3cr]

- 0.154 log [o/oSi] - 0.508 log V

(25)

The amount of uncertainty for this equation, because of the


scatter in the data, is much greater than that for the laboratory
controlled equilibrium study, as expected. There are also
apparent contradictions on the direction of the effect of the
temperature and the chromium content of the metal for this limited
study. However, as in the equilibrium study, the dominant effect
of the slag basicity is real, and the overall expression will serve
well for illustrative purposes.
For the average assumed conditions of 15. 00% chromium and 0.40%
silicon in the metal and a temperature of 3000 F, Fig,6 shows a
plot of the per cent chromi.um in the slag versus basici ty for the
equilibrium values [equation (24)] and the operating values
[equation (25)], together with the statistically expected deviation
from the unique value line as shown by the shaded areas ( 2s). It
is seen that, despite the scatter, the equilibrium concentrations
of chromium in the slag 0 are substantially below all indicated operating values. The difference here, and more particularly, between
t he equilibrium values and the known behavior for a particular shop
practice, shows the economic justification for increased expenditures
i n the form of flux, time and power usage to recover more of the
chr-omium left in the slag by a closer approach to chemical
quilibri um.

;!,11 l

C.W.McCOY, A.F.KOLEK AND F.C.LANGENBERG


16.00

~ --I

14.00
c

~12 . 00

&.

.;io.oo

en
.~

8.00

~ 6 .00

.:

u 4 .0 0 W J 1 1 J t t 1 1 . i

2.00 ~

o ~~~~~~~~~~~~~~~~~~~

1.00

1.20

1.40

1.60

1.80

2 .00

. .
( %Cao + %MgO)
%Si0
Slag Bas1c1ty,
2

Fig. 6. Comparison of Operating and Equilibrium Chromium Contents in Slags

Si

Cr = 15.0%
Temp= 1825 K (2826 Fl

=0.40%

810.00
~

a..

,!;;

5.00

E
~

.:

0.50

1.00

1.50

2 .00

.
( %Cao + %MgO)
Slag Basic1ty,
%S i 0
2

'

Fig. 7. Estimated Equilibrium Chromium Content in Slag versus Basicity


The equilibrium estimating equation can also be used to determine
the value of changing some of the conditions (slag basicity, metal
concentrations, temperature) to achieve a lower chromium level in
the slag, for a given relatively close approach to equilibrium or a
constant deviation from it. Figur 7 illustrnt 1:1 I'! v rnl points thnt
could h 00111'11<1 rod f r 111i 11H1-1 111n1 d ohrom1 um o nL1 11L (l 5% r) 1i11tl

APPLICATION OF THERMOCHEMICAL DATA


temperature (1825K). It shows the equilibrium effect of increased
basicity at a given metal silicon concentration or, conversely, the
effect of different silicon levels at a given slag basicity.
SULFUR AND NITROGEN
Other chemical reactions of interest in the finishing period are
the removal of sulfur from the bath and the possible addition of
nitrogen. Sulfur specifications in stainless steels are usually
low with a few exceptions such as the free-machining stainless grades.
Therefore, sulphur removal is important and it is recognized that
desulfurization is favored by basic slags with high fluidity, a
condition normal to stainless steel finishing. If an extra low
sulfur specification must be met, it is sometimes necessary to use
several finishing slags.
Equilibrium sulfur distribution data have not been obtained
directly on the slag-metal systems found in the finishing period
of stainless steel heats. However, the equilibrium can be approximated by using available data for the CaO-Si0 2-Al 2o3 system and
the appropriate interaction parameters in the metal phase. Of
particular importance is the effect of chromium on the activity of
sulfur. It has been determined 14 that chromium has a substantial
negative effect on the activity coefficient of sulfur which up to
about 20 per cent chromium can be expressed by log t 8Cr = - 0.019
[%Cr). Thus, the activity of sulfur in a heat containing 16% Cr
is approximately one-half the activity of the same concentration of
sulfur in a chromium-free alloy.
Gas (nitrogen)-metal equilibrium data are of little practical
value for those stainless steels where nitrogen is not intentionally
added as an alloying element. The final nitrogen content is well
below the solubility limit and is affected by operating variables.
However, for those stainless steels where nitrogen is intentionally
added as an alloying element (nitrogen is a strong austenite former
ll.nd 'stabilizer ), the solubility data 15 16 17 are of great practical
i mportance. A knowledge of the solubility limit as a function of
mp ition and Lomperature is useful in obtaining optimum recoveries
or n tr K n J'rorn 111 tr f( n-b aring ferroalloys or revert scrap.
NI t.rnr 11 T< cov1 1 I <H ur< low< r as th st l approaches its saturation
llmlt. iLll<I It. hlLI 111<11 11how11 l. lmL 11lLrng111 roov r,1
are related to
:.l'/1

C.W.McCOY, A.F.KOIEK AND F.C.LANGENBERG


the residual nitrogen in the bath before the nitrogen addition, the
amount of nitrogen added, and the nitrogen saturation solubility
limit 18 1 9. The solubility data are also of use in designing new
nitrogen bearing stainless steels.
LINEAR PROGRAMMING OF STAINLESS STEEL MELTING
As an example of the use of some of this and other thermochemical
data, let us review briefly the recent attempts at perfecting linear
programming mode1s in melting and refining of stainless steels. The
basic idea in linear programming is a method of choosing, from many
possible solutions of a set of linear equations, the solution which
minimizes a single additional linear equation. For stainless steel
melting and refining, the set of linear equations represents the
chemical and physical balances occurring during the process and the
equation which is minimized is the cost of the raw materials involved.
The independent variables appearing in these equations consist chiefly
of the various possibl e feed materials.
Linear progrll!:liiling of stainless steelmaking was first described
in the published literature by Hilty in 195920. The heart of the
model consisted of an energy balance during oxygen decarburization
developed by earlier work of Healy and Hilty21 and Healy22. Their
simplified general energy balance equation was:

(26)

The left side of the equation represents the energy input into the
furnace. A, B and C are the respective heats evolved when one per
cent of the element in solui ion per ton of charge is oxidized, and
concentration changes in the brackets are expressed as weight per
cent. E is the electrical energy input during the oxygen blow. The
factor 70 represents the heat 1 of fusion of a ton of steel and m is
the fraction of the bath stiil solid at the start of the oxidiz ing.
period.
The right side of the equation represents the energy output of
I
the system. Cp is the 'apparent heat capacity' of the molten bath
plus the furnace on a per-ton basis. The temperature change during
oxidation is given by the differ ence between the final t emp rature
272

APPLICATION OF THERMOCHEMICAL DATA


T2 and the initial temperature T1; g 0 is the heat loss rate of the
furnace to cooling water and surroundings per unit time (assumed to
be steady state); t is the total time from the start of the oxygen
blow. D is the heat content of the nitrogen accompanying a unit of
infiltrated air and Va is the rate of air infiltration per unit
t i me per ton of charge in the furnace.
The temperature T2 at the end of the oxygen blow is calculated
from a linearized formulation of the metal-carbon-temperature
relationships previously discussed. Assumptions concerning initial
chromium, manganese, nickel, carbon and s i l i con in the bath are made
for the appropriate final carbon and nickel levels desired. The
levels are subsequently corrected during successive approximations
with successive solutions until an overall acceptable solution i s
reached relative to the overall material and energy balances. No
diff i culty has been experienced in converging to a solution, even
though very large errors in the initial assumptions were made.
~

UJ
UJ

....----....-----.------,.------....----....----~

1-

Ul

1-

......

U)

1Ul

(.)
~

ii:
UJ

!q

:::E
~
UJ
U)

~ 0

II:
(.)

UJ

0.06

0.08

0.10

0.12

CARBON IN BATH AFTER BLOW (%)

Fi g.8.

Effect of Carbon after Blow on Materials Cost

, The Cr ucible linear programming model consists of twenty-four


quations of the chemical and physical balances, and about sixty
v1iri1tbl s , approximately ha l f of wh i ch are available raw materials

C.W.McCOY, A.F.KOLEK AND F.C.LANGENBERG


and the remainder such things as slag weight, slag composition, bath
weight, etc. Extensive testing of the program and comparison with
actual 80-ton electric furnace data have indicated that the
mathematical solutions are reliable. Its use in evaluation of new
materials strictly on an economic basis and in assessing the effect
of possible practice changes has been tested and adopted as standard
practice in the 302, 304, 430, 434 and 201 grades with tangible cost
savings. An example of these is shown in Fig.8 which shows, for
certain specific practical charge composition limitations, the
material cost savings than can be realized by stopping the oxygen
blow at higher-than-conventional carbon levels.
DEOXIDATION
The final and a very important area for the application of
thermochemical data to stainless steelmaking is deoxidation with
its subsequent effect on product quality. Numerous papers have
been written on deoxidation practices for stainless steels and the
relationship of these practices to quality and properties. However,
the study which relates most to the theme of this paper is that of
Ramachandran, Walsh and Fulton2 3 24 who developed a thermodynamic
model for calculating the oxygen content of molten stainless steels.
Their work is an excellent example of how fundamental data can be put
to practical use.
The investigators assumed that the controlling deoxidation
reaction is:
(27)
where <sio 2> denotes impure silica with variable activity in a
complex inclusion. Implicit in this assumption is that the
equilibrium oxygen in the finishing stages of a stainl'ess steel heat
is controlled by the metal ch~mistry, temperature and inclusion _.
composition: the function of the finishing slags is to protect the
steel from the furnace atmosphere and to permit the equilibrium
described in equation (27) to be reached.
For K = - - = - - - -- - - -

(28)

APPLICATION OF THERMCX::HEMICAL DATA


TABLE VII

Estimated Versus Actual Chromium in Tap Slags


for 80-Ton Electric-Furnace Heats

Estimated Cr
Temperature Metal Composition (%) Slag Composition (%) in Slag.from
Equation (24)
v
Cr
F'
Cr
Si
K
(%)
1824
1834
1830
1806
1806
1812
1829
1817

2824
2842
2834
2792
2792
2803
2833
2812

17.62
16.50
17.65
17.50
17. 15
17.59
18.60
18. 62

1. 73
1. 81
1. 90

0.24
0.39
0. 18
0.38
0.42
0.45
0.42
o. 50

1. 68
1. 78
1. 59
1. 59
1. 48

1. 89
1. 90

2.30
2. 98
2.16
3. 00
2.70
2.30

1. 41

1. 25
1.82
1. 23
1. 08
1.35
1. 64
1. 59

0.025.....----.--~--~--~-~

OBS.Q=.OOIOt.9082 C.ALC.Q
2cr=.0016

01

lZ 0.020
l&J

a:

l&J

11..0.015
~

::c
~

l&J

31: 0.010
0
l&J

&:

~0.005
m

0~--~--~----'-----'----'

J11 (

.9.

Act,nnl v rsus

O.OO!S
0.010
0.015
0.020 0.025
CALCULATED WEtGHT PERCENT Q

ul ulnt d Oxyg n Contents in Production Heats


( ll.i f

l" 11 H

'.l'il1

23 , 24 )

C.W.McCOY, A. F. KOLEK AND

F.C.LANGENBER~

the authors determine a 8i from the composition of the steel utilizing


the necessary interaction parameters.* The activity of silica is
determined by adding 'equivalent' silica activities which are a
function of the aM 0 /A.!' aA 1/a 8 i and al'.i./a~!. ratios in the metal.
Thus, a 0 can be calculated. Since f 0 can be determined from the

interaction parameters and metal composition, the oxygen content


can be calculated in weight per cent. The calculations are somewhat
tedius . So the procedure has been programmed for computer applica tion. Fig.923 demonstrates the use of the model on routine
production heats of stainless steel where analytically determined
oxygen contents are plotted against calculated oxygen contents.
Practical insights gained by utilizing the model are discussed in
the original papers.
SUMMARY
In 1964, over 1, 500,000 tons of stainless steel ingots will be
cast in the United States. This material will contain .anywhere
from 12 per cent to 30 per cent alloying elements costing from twelve
cents to several dollars per pound. The role of the industrial research and development metallurgical thermodynamicist in this
industry is principally the application of fundamental data to
improve the efficiency and subsequently lower the cost of the
melting processes. In this endeavor, he must be able to recognize
the basic chemical reactions and exercise some judgment to simplify
them to some manageable and useful form. This discussion has been
a review of some of these efforts. Some of the simple oxidation
reactions have been investigated. A beginning has been made toward
understanding interaction effects during this reaction period but
much more needs to be done. For example, more complex systems, where
several metallic elements are being oxidized concurrently to form a
homogeneous oxide slag, must be investigated. Here the analysis is
complicated by both sl~g and meta'I. interactions.
Simple slag-metal equilibrium studies for both the major alloying
elements and the residual elements are lacking, let alone gas-slagmetal studies to account for the effect of furnace atmospheres.
Refer to the original papers for the assumptions and approximations
used.
;,7()

APPLICATION OF THERMOCHEMICAL DATA


Some elementary thermochemical models of the stainless steel
melting process have been constructed which have proved of economic
value, but these rely principally on empirical constants rather than
a fundamental understanding of the chemical equilibrium, chemical
kinetics, mass transfer, or heat transfer involved. Only when the
major gaps in our knowledge of these areas are filled in the basic
research laboratories will the applied research man achieve understanding and, therefore, control over these and other industrial
processes.
ACKNOWLEDGEMENT
The helpful advice and criticism of Professor George W.Healy of
Pennsylvania State University in preparing this manuscript are
gratefully acknowledged.

'I'/ '/

C.W.McCO~,

A.F.KOIEK AND F.C.LANGENBERG


REFERENCES

1.

D.C.Hilty, H.P.Rassbach and W.Crafts, Observations of Stainless


Steel Melting Practice, J.Iron Steel Inst. 180, 116-128 (1955).

- 2.

D.C.Hilty, Relation Between Chromium and Carbon in Chromium


Steel Refining, Trans.A.I.M.E. 185, 91-94 (1949).

3.

D.C.Hilty, G.W:Healy and W.C.Crafts, Metallic Oxidation in


Chromium Steel Refining, Trans.A.I.M.E. 197, 649-653 (1953).

4.

F.D.Richardson, The Thermodynamics of Metallurgical Carbides


and of Carbon in Iron, J.Iron Steel Inst. 175, 33-51 (1953).

5.

T.Fuwa and J.Chipman, Activity of Carbon in Liquid-Iron Alloys,


Trans.A.I.M.E. 215, 708 (1959).

6.

A.Simkovich and C.W.McCoy, The Effect of Nickel on the Chromium


and Carbon Relationship in Stainless Steel Melting, Trans.
A.I.M.E. 221, 416 (1961) .

. 7.

D.C.Hilty, W.D.Forgeng and R.L.Folkman, Oxygen Solubility and


Oxide Phases in the Fe-Cr-0 System, Trans.A.I.M.E. 29, 253
(1955).

8.

B.V.Linchevskii and A.M.Samarin, Solubility of OXygen in


Manganese-Containing Liquid Iron, Izvestiya Akademii Nauk
S.S.S.R. otdel.Techn.Nauk, No.2, 9-18 (February 1957).
(Brutcher Trans.No.4030) .

9.

C.R. Taylor, Some Observations on the Metallurgy of Electric


Furnace Melting, Electric Furnace Steel Proceedings, A.I.M.E.
8, 91-~7 (1950).

10.

W.Crafts and H.P.Rassbach, Melting Low-Carbon Stainless Steels,


Electric Furnace Steel Proceedings, A.I.M.E. 9, 95-104 (1951).

11.

H.P.Rassbach and E.R.Saunders, Reducing Period in Stainless


Steel Melting, Tr.ans.A.I.M.E.. 197, 1009-1016 (1953).

12.

...

'

A.J.Texter, Discussion, Electri c Furnace

A. f .M.F.. 11, 73 -77 (1953)~

','/II

l Proceedings ,

APPLICATION OF THERMOCHEMICAL DATA


13.

C.W.McCoy and F.C.Langenberg, Slag-metal Equilibrium During


Stainless Steel Melting, J.Metals 16, 421-424 (1964).

14.

N.R.Griffing and G.W.Healy, The Effect of Chromium on the


Activity of Sulfur in Liquid Iron, Trans.A.I.M.E. 218, 849-854
(1960).

15.

J.C.Humbert and J.F.Elliott, Solubility of Nitrogen in Liquid


Fe-Cr-Ni Alloys, Trans.A.I.M.E. 218, 1076-1087 (1960).

16.

R.D.Pehlke and J.F.Elliott, Solubility of Nitrogen in Liquid


Iron Alloys, Trans .A. I.M.E. 218, 1088-1101 (1960).

17.

H.Schenck, M. Frohberg and H.Graf, Effect of Additions on the


Equilibrium of Nitrogen with Liquid Iron, Arch.Eisenhuttenwessen,
29, 673-676 (1958) and 30, 533-537 (1959).

18.

F.C.Langenberg and M.J.Day, Application of Nitrogen Solubility


Data to Alloy Steelmaking, Electric Furnace Steel Proceedings,
A.I.M.E. 15, 7-15 (1957).

19.

H.P.Rassbach, E.R.Saunders and W.L.Harbrecht, Nitrogen in


Stainless Steel, Electric Furnace Steel Proceedings, A.I.M.E.
11, 244-256 (1953).

20.

D.C.Hilty, R.W.Taylor and R.H.Gillespie, Predicting Minimum


Materials Cost for Stainless Steels, Journal of Metals 11,
458-464 (July 1959).

21 .

G.W.Healy and D.C.Hilty, Effect of Oxygen Input Rates in the


Decarburization of Chromium Steel, Trans.A.I.M.E. 209, 695-707
(1957).

22 .

G.W. Healy,, Simplified Calculation of Effects of Oxygen


Blowing Rates, Electric Furrace Conference Proc.A.I.M.E. 16,
252 ( 1958).

23.

R.A.Walsh, S.Ral!Rchandran and J.C:Fulton, A Thermodynamic


Mod 1 for th Deoxidation of Alloy Steels, Trans.A.I.M.E. 227,
1380-1:193 ( 106::1} .

...,'

C.W.McCOY, A.F.KOlEK AND F.C.LANGENBERG


24. S.Ramachandran, R.A.Walsh and J.C.Fulton, Calculation of Oxygen
Contents in Molten Stainless Steels, Electric Furnace Steel
Proceedings, A.I.M.E. 19, 128-140 (1961).

'

THERMODYNAMICS OF MATERIALS AT HIGH TEMPERATURE


AND LOW PRESSURE
by
Paul W. Gilles
Department of Chemistry
University of Kansas
Lawrence, Kansas

These are topics about which I should like to talk this morning:
pressure -- composition diagrams , (2) the factors that influence
the mode of vaporization and determine the vapor pressure of a
substance, (3) the steps i n a vaporization study and (4) finally, a
few suggestions of practical uses.
(1 )

All substances vaporize and each does so at a characteristic rate


dependent upon its temperature and its environment. The rates of
vaporizat i on of different substances vary enormously; yet at
suffic i ently high temperatures, every substance will vaporize at an
appreciable and destructive rate.
It is now being recognized that the limiting faqtor in many high
t mperature systems will be the rate of loss of matter by vaporization.
Former ly, melting temperatures were of utmost importance in the
d sign and use of high temperature systems. As more highly refractory
Hllbf)tances become available and as our temperature limits increase,
vuporization becomes the most important mode of deterioration of
rnut ria ls . The purpose of this paper is to discuss some of the
l>rinciples of vaporization phenomena and to relate them to the
Lh r modynamic properties of substances. Pitzer and Brewer 1 have
discussed some of these problems in an unexcelled way.
Vaporizati on phenomena are best elucidated through an investigation
l111own as a vaporization study. Such a study frequently has purposes
h< yond t hat of s i mply characterizing the vaporization of a chemical
11.YRt m. Better than any other kind of study, a vaporization
I t~V _Atigation can yi eld t he thermodynamic properties of simple
1o11u1unus molecules . In some instances such properties can be obLt ln d in no other way . These pr operties are of immense and critical
lumorLo.n
in t ting models and theories of chemical binding. In
r 01111 11yHt1 utH, u vnpor 1zat ion st udy may yield even the thermodynamic
t11' ol11rt.I!11 of' t.111 co11rh.1111 d pho. s. A vaporization st udy also
''.Ill

P. W.GILLES
establishes the nature of high temperature reactions. It provides
data on the vapor pressure which establishes the ultimate limit on
the use of a material in a high temperature system. Finally, a
vaporization study may be useful in the preparation of new compounds
or in the preparation of single crystals. Professor Alcock 2
mentioned yesterday that there is no better way to obtain the partial
molal free energies of components in solutions then by carrying out
a vaporization study, and with that I surely agree. Some industrial
processes are interrelated with vaporization reactions. Among these
can be named metal preparations, vapor plating and metal purification.
Some reasons for vaporization studies are summarized in Table I.
TABLE I
Purposes of Vaporization Studies
1. To obtain for simple gaseous substances thermodynamic data
essential for chemical binding theory.
2. To obtain thermodynamic data for refractory phases.
I

3. To investigate the stoichiometry of high temperature reactions.


4. To measure vapor pressures of pure substances.
5. To measure partial molal free energies and enthalpies of
components in solutions.
6. To investigate the kinetics and mechanisms of high temperature
reactions.
7. To prepare new compounds with elements in unusual oxid,ation
states.
8. To produce single crystals.

'

THE PRESSURE -- COMPOSITION DIAGRAM

A pressure -- composition diagram is the most convenient way to


show the results of a vaporization study. In such a diagram for a
two-component system the total pressure is plotted as a function of
'fl2

MATERIALS AT HIGH TEMPERATURE AND LOW PRESSURE


the composition for a fixed temperature. Simple diagrams of t his
type are found in elementary physical chemistry texts.

VAPOR

Fig.1 .

Ideal Solution

Figure 1 shows the behavior of a system of two components, ~ and


containing ideal gaseous solutions and ideal solid solutions.
whenever the pressure is very low, only the gas phase is present;
whenever the pressure is high, only the solid phase is present. The
total pressure varies linearly with the composition of the solid
vhase, as shown by the upper line, and varies along the lower curve
with the composition of the gas phase. At intermediate pressures
lutd compositions as in the lenticular area, two phases are present;
lhe compositions of these two phases are given by the intersections
nf a tie-line with the curves.
~.

Now, if a system does not show ideal behavior, it may exhibit the
h havior shown in Figure 2. In such a system there is no intermediate
ompound, but partial solubility of the components occurs. Again the
total pressure is plotted as a function of the solid composition for
t.h upper curves and of the gaseous composition for the lower curves.
AL , low pressures only vapor is present. At temperatures and pressures
Cul ling within the area at the upper left which is labeled~ (solution),
th Rystem contains a variable composition solid solution phase of ~
111 P!. Similarly at the upper right, mutatis mutandis. In the upper
c:cntrul 1)1\r'L Lw Holid phas , each saturated with the minor component,
1xlr1t. In 10111ll hrlu m. 'l'h cq111l j br 1um phases in the lenticular areas
111 1

1 11 01111 10 1111. 11111 11rnl n 1-11L1H

ow 1

1ml111.lcm with compoH itions shown

P.W.GILLES

A (8-SAT'D SOLN)
0

A(SOLN l

8 (SOL Nl
8 ( A-SAT'D SQ' N)

VAPOR

Fig.2
at the ends of tie-line. Along the horizontal line three phases, ~
saturated with ~. vapor of composition f and ~ saturated with A. exist
in equilibrium.
In Figure 3 the same areas are shown, but both solid solutions
yield vapors relatively richer in component ~ Fractional distillation of the system represented in Figure 3 produces pure ~ in the
distillate and pure ~ in the residue. Regardless of the initial
composition, partial vaporization always prod.uces a residue richer
in ~ If the residue is a single phase, its equilibrium pressure is
less than that of the initial sample.
Fractional distillation of the system shown in Figure 2 causes the
composition of the residue to change toward ~ if the original solid
composition lay to the left of the point ~. and toward ~ if to the
right. Again if the residue is a single phase, its equilibrium
pressure is smaller than that of the initial sample.
0

Now, if a binary ;;ystem contains an intermediate phase, which . a


chemist is inclined to call a compound, the diagram might have the
behavior shown in Figure 4. At the upper left and upper right are
the component solid solutions. In the upper center is the variable
composition solid solution sometimes called a compound, which is
shown here as AB. Two-phase regions lie as shown -between these

2fJ4

MATERIALS AT HIGH TEMPERATURE AND LOW PRESSURE

A (8- SAT D SOLN)

B(SOLN)

A(SOLN)

B (A-SAT'D SOLN)

VAPOR

Fig.3

BW!-5A

A(AB-SAT D)

+
A(SOLN)

AB (SOl.N) AB (B-SA

AB (A-SAT[j

VAPOR

Fig . 4
tt111gle-phase regions. The vapor phase is at the bottom; the lenti'Ular ar as at the sides have the same significance as before, and
t.hoH in th
nter r pr sent equilibrium between the solid
11o l11L.Lon ' ompcmnd' und vapor solution with compositions given at
I. ht

1 111111

ul'

I\

LI 1 - 1 Im

P.W.GILLES
The intermediate phase is shown as a congruently vaporizing one.
At the minima in the curves in the center of the diagram, the vapor
and solid have the same composition, and vaporization of the phase
at this composition occurs at constant pressure without a cJmposition
change. Moreover, samples of the 'compound' with different
compositions produce a vapor relatively richer in the component which
is in excess compared to that at the minimum, thus causing the
residue to approach the composition of the minim1.p11. Again the
horizontal lines join the compositions of the three phases in
equilibrium with one another.

B
(AB-SATD B (SOCN)

A
A

(AB-SATO)

AB

(SOLNl

(SOLNl

AB
(A-SATO)

AB
(B-SATOl

AB

+V

VAPOR

Ne

Fig.5
Figure 5 shows the incongruent vaporization of an intermediate
phase. The left and right sides and the bottom of 'the diagram
have the same significance as previously. The vapor in equilibrium
with the 'compound' is always relatively richer in component ~ than
the condensed phase. Accordingly, as vapor is removed, the residue
must grow richer in component ~ Vaporization of 'compound' ultimately produces the~~ solid SO'lution. Fractional distillation ~of .
any sample ultimately produces pure solid ~ and pure gaseous .
Figure 6 shows the results of an extensive experimental investigation on the titanium-oxygen system carried out in our laboratories
by Drs.Q.D.Wheatley, P.G_.Wahlbeck, H.F.Franzen, K.D.Carlson and
286

MATERIALS AT HIGH TEMPERATURE AND LOW PRESSURE

a-Ti

a+/3

TiO

Ti 20
+

VAPOR

Q2

0.4

0.6

0.8

Ti 20 3 rTi30,
+
+
Ti3 0 5

fr;,p.,

Ti 3 0 5

TiO
+
Ti 2 0 3

TiO

m,

Ti 20 3

Ti20

f3

------

1.0

1.2

1.4

1.6

_./

1.8

2.0

o/r;

Fig . 6
Mr.G. D.Stone. In the titanium-oxygen system numerous condensed
phases h~ve been identified, and some are shown in the figure . The
most important of these are Tio, Ti 2o 3, Ti 3o 5, and Ti0 2. Between
Ti 30 5 and Ti0 2 are many phases discovered by Magneli and coworkers 3 4 5 whose compositions are described as Tino 2n-l These
nre only roughly indicated in the figure. The vaporization studies
have revealed that Tio and Ti 2o3 vaporize incongruently, always
producing a titlll'ium-rich vapor and a residuP containing the next
higher oxide. It is well known that Ti0 2 loses oxygen, and so also
do most of the complex phases. The congruently vaporizing compoHition has been shown to lie in the Ti 3o 5 solid solution region.
Th se f acts are summarized in the figure.
This kind of diagram does not of course reveal the nature of the
aseous species . It is only the composition of the gas phase which
s. indicated. In the titanium-oxygen system, the principal species
ov r the TiO phase are Ti(g) and TiO(g) ; over the Ti 2o3 phase, the
vr, 11 iriul gas is TiO(g); over the Ti0 2 phase, the principal gas is
o:x.yg 11, but Ti0 2(g) is appreciable. The principal species over the
m1 11 gr111 utly vo.p ril'.jng T o solid solution appear to be TiO and
3 5

llXYK\ n

~Lom11.

'!II'/

P.W.GILLES
SCHEMATIC

PRESSURE - COMPOSITION

PHASE

DIAGRAM

li

s~

l i,. S TV

VAPOR

03
SI Ti

Fig . 7.

ATOMIC

RATIO

Schematic Pressure-Composition Phase Diagram

The results of a related study on the titanium-sulfur performed


by Dr.Hugo F.Franzen 6 are shown in Figure 7. The higher sulfides
lose sulfur preferentially and Ti 2S loses metal preferentially.
Accordingly, TiS is the congruently vaporizing phase in this
particular system.

(1)

us 2 (a) +

U\

US

(1)

us, .. '9> + <1l

U(i)

u(.Q l +us,.,.<1

us 2

U.f'
(sl

u2.s3 (1)

"':::>a:
"'"'a:"'

(1)

u3s5 <> +

a..

u (i)t v

_ ___......-- ._ u s Cal + v
2 3

..J

..:

....
....0

us,..(S) + v

VAPOR

1.0

1.5

1.67

s I u

Fig.fl
'1.11 11

2.0

3.0

MATERIALS AT HIGH TEMPERATURE AND LOW PRESSURE


The uranium-oxygen system has been studied and discussed extensively and pressure-composition diagrams for it have been
published7,8,9. Figure 8 results from the work of Dr.E.D.Caterl0 , 11
who studied the uranium-sulfur system. The congruently vaporizing
phase in this system is the uranium monosulfide . Similar diagrams
have been presented for several additional two-component oxide
systems and boride systemsl2,13.
FACTORS INFLUENCING VAPORIZATION PHENOMA
Turn now to some of the various factors that influence the mode of
vaporization and the vapor pressure of a refractory substance.
The words mode of vaporization refer to the matters of the congruence of the processes and the variance of the system undergoing
vaporization. It has been pointed out previously that meaningful
pressures can be obtained only under certain circumstances 1 14 .
For two-component systems either two condensed phases and one gas
phase ~ust exist in equilibrium or a single condensed phase and a
gas phase of the same composition must co-exist .
TABLE II
Some Factors Influencing Equilibrium Vaporization Phenomena
1. Thermodynamic stabilities of the gaseous molecules.
2. Thermodynamic stabilities of the condensed phases.
3. Compositions of the gaseous compounds.
4. Compositions of the condensed phases.
5. Free energies of sublimation of the elements .
6. Curvature in the free energy-composition diagram.
7. The nat ure of t he experiment.
1'n.bl

gj v

1:1

in approximate decr easing order of importance, the

flL :LorH w)l l IJ 111J' l u n


Quilibrium vaporization phenoma. The
11L1Lhl l lt. l111of'1u 11on 1111 rnnlu 11 11r-1 ur
f nlmost.overwhel ming i mportance

P.W.GILLES
in establishing the mode of vaporization and the vapor pressure. If
a gaseous molecule of great stability exists, the condensed phase
of that same composition will tend to vaporize congruently with a
high pressure. The stabilities of the condensed phases are less
important, but to illustrate the influence of this factor, consider
the hypothetical case of a solid compound whose stability can be
altered. As the stability of the solid compound is increased
without a change in the stability of a gas phase, the vapor pressure
falls. Inasmuch as congruent vaporization occurs at a locally
minimum pressure, the tendency of the condensed phase to vaporize
congruently increases. Also its tendency to vaporize by decomposition
into the elements instead of into a gaseous compound increases .
Not only does the existence of a stable gaseous compound of a
particular composition tend to cause a condensed phase of that same
composition to vaporize congruently, but also its existence causes
other solid phases to tend to vaporize incongruently. Brewer 15
aites several such cases among the metal halides, ~ or example, the
titanium trihalide disproportionation. An interesting illustration
is provided by the work of Hildenbrand 16 on the lithium-aluminumfluoride system; Li 3AlF 6 (s) vaporizes incongruently primarily
because of the appreciable stability of the gaseous molecule LiAlF 4
which has no counterpart in the condensed phase. This discussion
also shows that the composition of the condensed phase, more
particularly its r elationship to the composition of gaseous molecules,
influences the vaporization mode and pressure.
In the absence of gaseous molecules, vaporization proceeds by
decomposition to the elements, and the heats of sublimation of these
elements become the most important factor in establishing the mode
of vaporization and the vapo~ pressure. In a binary system with
greatly different heats of vaporization for the components the usual
mode of vaporization is by preferential vaporization of the component
with the higher volatility. In extreme cases , the ultimate residue
is the other componept . A goocf example is the tantalum-boron
sys t em17 in which boron is preferentially vaporized from all the
phases. Another example is provided by the chromium-boron system
in which chromium is preferentially vaporized from all phases. If
the elements have comparable volatility, then a congruently
vaporizing compound frequently exists such as has been shown for
zircon!~ diboride1 8. The fact ors are t empered of course by t he
?.90

MATERIALS AT HIGH TEMPERATURE AND LOW PRESSURE


stability of the condensed phase; for as its stability increases its
tendency to vaporize congruently by decomposition to the components
increases.
Of perhaps some importance in some systems is the nature of the
curvature in the free energy-composition diagram near the minimum
but the author knows of no case in which this effect is important.
Finally, in some systems in which stabilities of condensed phases
ar e very close to one another, different modes of vaporization might
be expected for static experiments and for Knudsen experiments owing
to the existence in the latter of the molecular weight dependence.
In many instances the mode of vaporization and the transport of
matter in a vapor pressure experiment are not governed by these
equilibrium properties. Brewer 19 has pointed out numerous times
t hat very low rates of vaporization are expected if the molecular
configuration in the vapor differs radically from that in the condensed phase. High energy barriers contributing to slow rates of
reaction may completely suppress the equilibrium properties. Notable
among these cases are phosphorus 20 , As 2o321 , and several metallic
nitrides2 2 , 23 . In some cases, particularly those in which incongruent
vaporization occurs, surface impoverishment arises because of slow
rates of diffusion, and vaporization rates are much lower than
equilibrium rates.
THE COURSE OF A VAPORIZATION STUDY
Ordinarily a vaporization study proceeds through several distinct
parts as shown in Table III.
TABLE III
Stages in a Vaporization Study
-1. General

charact~ristics

of the chemical system.

2. Ide nt ification of the vapor species .


3. WrH. i ur, oC th
11 ,

Mt

ILi 111 ' 1 1111 111.

I)!'

11

t reaction.

1.1 11

Vlq)OI' lll' l llH111'0.

P.W.GILLES
5. Calculation of thermodynamic properties.
6. Deduction of molecular structures and constants.
7. Measurement of the kinetics of vaporization .
8. Establishment of the mechanism of vaporization .
These are listed in the order in which they might most reasonably
be accomplished, but rarely is this order followed in actual practice .
In the initial characterization of the vaporization process, one
first investigates possible reactions between the sample and other
parts of the system, which are numerous. A crucible is frequently
present; residual gas is present in a vacuum system; gases in a
transpiration experiment are not always pure; shields surround the
vaporizing s ample; and sometimes a suspension is present in the
apparatus. The apparatus should be designed and operated in such a
fashion so as to reduce or eliminate reactions between the sample
and the other ingredients. This initial characterization also is
used to establish which phases are congruently vaporizing and what
is the variance of the system. Simple, easy experiments can be used
to accomplish these investigations. The crucible material may be
changed; the apparatus may be altered; x-ray diffraction studies may
be employed; a mass balance can be attempted; a series of step -wise
vaporization experiments can be performed; chemical analyses can be
performed; and the closed crucible technique may be employed. The
results of the initial characterization of the chemical system are
usually presented in the form of a pressure-composition diagram
similar to those described previously.
After this study into the ,properties of the system has been accomplished, the gaseous specie ~ should be identified. The mass
spectrometer has been spectacularly successful in deducing the
vaporizing species 24 but other techniques may be used .

...

'

At this stage meaningful vapor pressure measurements may be made.


The ,methods will depend primarily on the magnitude of the pressure
to be measured. At high pressures , a static or transpiration
apparatus can be used. The Knudsen method, the Langmuir method and
the tors~on effusion meth_od are appr opriat e for measur ements a t low
press ures. These met hods are adequately described
elsewhere25, 26 , 27 28, 29. . Tw kinds of apparA.tu1::1 m:1od i n ur
20

MATERIALS AT HIGH TEMPERATURE AND LOW PRESSURE

r: :;

MAGNETICALLY

ACTUATED

EJECTOR ROD

{-

~?~f-

--

, ,., CE- ACETONE DEWA

.,

COPPER TO PYRE)( HOUSEKEEPER SEAL

JOINT

TARGET MP<:;AZINE

TARGET RECEIVER

rnc~ACT::Eo=;;~
~
SHVHER i~ 1~~ ~fO; WATE::~~NG OUTLET
~

INDUCTION COtL
CRUC18LE ASSEMBLY

I
.

QUARTZ TABLE
PYREX WATER JACKET
PYREX

(I)
'

WATER

TO

COOLI N G INLET_ 1

OIFF'USION

.f

CON OE NSER

JOINT

B&S

~;

p~""-'---~ I

-1

--f- --1 ,---, ~ ' '"'"T


0

GAUGE

~1' ~r ~
I ~1

'
'

~~
'""'" , :

O~:AL

WINDOW

J1!NT

r i-- a1..Ass wouNTJNG nPs

Po

,'"'"'
f
f

'"'"T
'"'"T

~ltf'
) -.~ : ----~ ~:@+-.=.

PRISM

MAGNETIC SHUTTER

024t
W.WJ

,:c:~~

:.

Fig.9.

Vacuum Effusion Apparatus

laboratory are shown in Figures 9 and 10. In Figure 9, a crucible


containing the sample is placed in the center of a glass apparatus
and is heated inductively by 450kc rf power. Gases arising from
the vaporization processes escape through a hole in the lid of the
cruc_ible anti travel upward in the apparatus. The weight loss of
the crucible can be measured, or the vapor escaping from the crucible
can be collected on a target. After a target has been exposed to
the vapor stream, it can be ejected and allowed to fall into a
rocreiver. Meanwhile, the next target falls into the receiving
p ition. The collected matter can be assayed by one or more of
s v ral techniques. Modification of this technique employs an
inv rtod tub i n plac of the target collection attachment so that
Lhc o l lt L d 1111mpl
nn b ' ch mi cally dissolved and appropriately
lllllLl .Y~!

d.
,f)

:1

P.W. GILLES
0

L..L..J
SCA LE
IN CM.

~ f ..IOINT

SILA STI C

GA SKETS

WATER COOLED
SUPPOflT FLANGE

PRE SSUR E

GAUG E .
TO DIFFU S ION Pl.MP

LIQU ID NITROGEN

COLO

CURRENT

CO NCENT~(ITOR
VACUU M

Fig.10.

SY STEM

TRAP

APPARAT US

14

Current Concentrator Apparatus - Vacuum System #14

In Figure 10, an eddy current concentrator is shown. The sample


is placed inside the concentrator and heated either directly, if it
is a conducting sample, or by radiation from a susceptor. With this
apparatus the weight loss of the sample is measured, although it is
possible under certain circU!Ilfltances to collect the vapor for
subsequent assay .
Vapor pressure data obtained 1 from the measurements can be
appropriately treated by well-known methods to deduce the thermo dynamic properties for the reactions that occur. Both second-law
and third-law evaluations should be attempted.
Most vaporization studies terminate at thi s point, but one r ec ent
development illustrating t he use of vapor i zation studies for
' 291

MATERIALS AT HIGH TEMPERATURE AND LOW PRESSURE


establishing molecular constants was accomplished independently by
Cubicciotti and Eding 30 and by Hildenbrand 31 . They deduced
values for vibrational frequencies by comparing second-law and
third-law results on dihalide molecules.
Studies of the kinetics and mechanism of vaporization are underway
in several laboratories. It is to be expected that activity in this
field will increase rapidly in the next few years.
PRACTICAL USES
The most important uses of high temperature vaporization studies
from the point of view of a chemist are those directed mainly toward
establishing the principles of chemical binding; obtaining vapor
pressures and thermodynamic properties of substances; and preparing
new compounds. But in addition, certain practical aspects of
vaporization studies can be cited. An entire industrial endeavor
is associated with vapor plating. Yesterday's newspaper 32 contained
an article concerning a new process by which aluminum is to be plated
on steel. Another interesting development which has occurred as a
result of high temperature investigations was presented in the paper
by Brewer and Rosenblatt 33 on the principles of the purification of
metals by evaporation and by vacuum fusion.
Still another development not closely related to vaporization
processes but which came from Professor Brewer and which therefore
must be mentioned is his theoretical paper on the deduction of phase
diagrams 34 . In this report he predicts about two billion phase
diagrams.
ACKNOWLEDGEMENTS
The author is pleased to acknowledge the support of the United
States Atomic Energy Commission through its Contract At(ll-1)-1140
with the University of Kansas.

:,!)

P.W.GILLES
REFERENCES
1.

G.N.Lewis, M.Randall, K.S.Pitzer and L.Brewer, Thermodynamics,


Second Edition, Chapter 33, McGraw-Hill: New York (1961) .

2.

C.B.Alcock, Panel I of this Conference, 39-50.

3.

S.Anderson, B.Collen, U.Kuylenstierna and A.Magneli, Acta.


Chem.Scand. 11, 1641 (1957).

4.

S.Andersson and A.Magneli, Naturwiss. 43, 495 (1956).

5.

S.Andersson, Acta Chem.Scand. 14, 1161 (1960).

6.

H.F.Franzen and P.W.Gilles, J.Chem.Phys. 42, 1033-8 (1965).

7.

R.J.Ackermann, E.G.Rauh and R.J.Thorn, J.Chem.Phys. 37,


2693-8 (1962).

8.

P.W.Gilles, Chapter in Thermodynamics of Nuclear Materials,


401-15, International Atomic Energy Agency, Vienna (1962).

9.

International Atomic Energy Agency, Thermodynamic and Transport


Properties of Uranium Dioxide and Related Phases. Technical
,Report Series No.39. International Atomic Energy Agency,
Vienna (1965).

10.

E.D.Cater, P.W.Gilles and R.J.Thorn, J.Chem.Phys. 35, 608-18


(1961).

11J

E.D.Cater, E.G.Rauh and R.J.Thorn, ibid. 619-24.

12.

R.J.Ackermann and E.G.Rauh, J.Phys.Chem. 67, 2596 (1963).

13.

P. K.Smith, Ph.D. Thesis, l}piversity of Kansas (1964) .

14.

P.W.Gilles, Pure and Applied Chemistry 5, 543-56 (1962).

15.

...

L.Brewer, Chapter in Chemistry and Metallurgy of Miscellaneous


edited by L.L.Quill, 193-275, McGraw-Hill (1950).

Mat~rials,

200

MATERIALS AT HIGH TEMPERATURE AND LOW PRESSURE


16.

D.L.Hildenbrand, Aeronutronic Division, Philco Corporation,


Newport Beach, California, Report No.U-2403, (Sept . 15 1963) .

17.

J.M.Leitnaker, M.G.Bowman and P.W.Gilles, J.Electrochem.Soc.


109, 441-3 (1962).

18.

J.M.Leitnaker, M.G.Bowman and P.W.Gilles, J.Chem.Phys. 36,


350-8 ( 1962).

19.

L.Brewer, Proceedings of the Robert A.Welch Fowuiation


Conferences in Chemical Research 6, 47-92, especially 82
(1962).

20.

L.Brewer and J.S.Kane, J.Phys.Chem. 59, 105 (1955).

21.

K.A.Becker, K.Plieth and I.N.Stranski, Progress in Inorganic


Chemistry 4, 1-72 (1962).

22.

D.L.Hildenbrand and W.F.Hall, J.Phys.Chem. 67, 888 (1963).

23.

L.H.Dreger, V.V.Dadape and J.L.Margrave, J.Phys.Chem. 66,


1556-9 (1962).

24.

M.G.Ingrhram and J.Drowart, Proceedings of an International


Symposium on High Temperature Technology held at Asilomar,
California, October 6-9, 1959, 219-240, McGraw-Hill: New York
(1960).

25.

C.B.Alcock and G. W.Hooper, 'Measurement of Vapor Pressures at


High Temperatures by the Transpiration Method', Phys.Chem. of
Process Metallurgy, Pt.1, 325-40, G. R.St.Pierre, Editor,
Interscience Publishers: New York (1961).

26.

Robert D.Freeman, Molecular Flow and the Effusion Process in


the Measurement of Vapor Pressures, Technical Documentary
Report No.ASD - TDR-63-754 (August 1963) Wright-Patterson Air
Force Base, Ohio.

27.

J.L.Mar grav. a chapter in J. OM. Bockris, J.L.White and


J.D . Ma k nz , Rds., Physicoche mica l Measurements at High
'1'1111wrnt1u1 s , 22 5-46, A ad mic Press, Inc.:
New York, New
York (llhn) .
:~ 11'1

P.W.GILLES
28.

U.Merten and W.E.Bell, 'The Transpiration Method', GA-2144,


Rev., General Atomic Division of General Dynamics, San Diego,
California (September 29 1961).

29.

A.N. Nesmeyanov, Vapor Pressure of the Chemical Elements, Edited


by Robert Gary, NBS, Washington, D.C . translated by Scripta
Technica , Washington, D.C.; Elsevier Publishing Co.: New York
(1963).

30.

D.D . Cubicciotti and H.Eding, J.Chem.Phys. 40, 978 - 82 (1964) .

31.

D.L.Hildenbrand, J.Chem.Phys. 40, 3438 - 9 (1964).

32.

Pittsburgh Press, article entitled 'U.S. Steel to Construct


Coating Mill' (Nov.30 1964).

33.

L. Brewer and G.Rosenblatt , Trans.Met.Soc.A.I .M.E. 224, 1268-71


(1962) .

34.

L. Brewer, Prediction of High Temperature Metallic Phase


Diagrams, University of California, Lawrence Radiation
Laboratory, Berkeley, California, Report No.UCRL-10701, Rev.
(January 1964) .

'

'l 111

A SUMMARY OF EXPERIENCES IN APPLYING


THERMODYNAMICS TO METALLURGICAL PROCESSES
by

G.R.Fitterer
Distinguished Professor, Metallurgical Engineering and
Director of the Thermodynamics Center, University of Pittsburgh

I would like to start

imr

talk with two questions;

First: Why is it that with all of the available thermodynamic


data which we have at our disposal today, are so many new metallurgical processes being developed through large scale, empirically
conceived projects such as, for example, the vacuum degassing. and
the basic oxygen steel processes?
Secondly: Is it not now time for us to start with basic thermodynamic data and apply these directly to the development of new
processes?
This Conference has been conceived for the purpose of directing
discussion towards the transition of the pure and abstract science
of thermodynamics into applications to industrial processes.
In the arrangement of the four panels of the conference, an
at tempt has been made to review the current status of the science
with respect to the amount and types of data which are available,
together with some of the steps which are needed to apply them to
commercial processes.
The first panel discussed the current status of available data
ana suggested some methods by which these might be expressed in
usuable form.
The second panel discussed some of the applications of thermodynamic principles to non-ferrous metallurgical processes. Some of
these problems and methods of solution were perhaps unfamiliar to
i:rorne of those in this audience, but they were included purposely
n th hope of stimulating new ideas concerning their applications
1n var iou industries.
P rJmpH Lh 1H J,llLl)Ol' is h t dir cted towards the engineer and the
1ipull1rl 11l1 11t. t11L, um! 111 doin H it Ahould be emphasized 'that there

ion

G.R.FITTERER
has been a tremendous number of laboratory experiments which have
produced considerable reliable thermodynamic data. However, there
are a surprisingly few engineers who are currently using this
information to develop new materials or new processes.
In other fields of science this procedure has become known as
the 'spin-off' of scientific discoveries into useful commercial
products. Fortunately, there is no need today for one to wait for
new data in order to initiate many of these developments. Much
useful information has been waiting for its commercial application
for many years.
The scientist has already done a good portion of his job and has
furnished s with more information than we can use for some time.
One should not expect the scientist to visualize our needs. Instead,
the engineers must anticipate the needs of commerce and direct the
'spin-off', accordingly.
Forty-one years ago, the publication of the text on Thermodynamics
by Lewis and Randall(l) marked the beginning of the applications of
this science. However, at that time, the amount at available data
was meagre and it was often necessary to perform costly pilot-plant
or full-scale commercial experiments in order to improve a commercial
process or to develop a new one. Today, much of the groundwork can
be done by careful thermodynamic calculations.
At this point, I would like to relate just a few experiences of
one who has in the past resorted to full-scale, commercial experiments for answers. However, thermodynamic principles and data were
kept in mind and used whenever possible.
This work began in 1926, when as a student at Carnegie Tech, I
was assigned a thesis on the Deoxidation of Steel with Silicon'
under the direction of Dr.C.H.Her}y, Jr., at the U.S. Bureau of
Mines <2 ). This was a p'a rt of a large program under his direction on
the 'Physical Chemistry of Steel Making' in co-operation with some
70 industrial firms.
In these investigations , the 'test tubes' were open hearth furnaceR
ranging in siz fro m 30 to 250 tons capacity .. Literally hundreds of
mm r 1ul hottLH f liquid Ht, 1 w.r st udi d tn rd r to d termine
t.h

l11in , t,1 111t.lr:11 rrr Lh 1111vorL1l11L 1

:lllfl

LllmH.

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


It must be recalled that in the 20's there were but few laboratory tools such as the high frequency induction furnace and, thus ,
the vacuum fusion analysis as we know it today did not exist. In
f act, practically none of the current experimental procedures were
available at the time. At the start of the program, there was
considerable dispute as to which reactions occurred during the
r efining of steel.
ACQUISITION OF THERMODYNAMIC DATA FROM COMMERCIAL PROCESSES
The system of chemical thermodynamics (as we know it today)
was established(!) only three years before the start of this
program on the 'Physical Chemistry of Steel Making' . As stated,
the available data, the research techniques and the equipment were
meagre at that time to say the least. Nevertheless , one did the
best with what was available and many who have read the publications by
Berty , et.al. ( 2 ) still regard this work as marking the transition
fr om the 'art' of steelmaking to the science. Only three examples
of these and other attempts to reduce thermodynamics to commercial
us e on this scale will be presented here .
The Carbon-Oxygen Reaction in Liquid Steel
Inasmuch as this is the most important refining reaction in
liquid steel , its characteristics were the first concern of the
Herty group. A large number of other workers have also studied
the equi librium of the reaction*
Q+Q

= co

and i t is now quite well established that the equilibrium relation


can be defined by the expression
m = (%Q) (%Q)

= 0.0022

This constant does not change much with temperature but because
f the CO gas evolution, the value of m is pressure sensitive .

Untl n r I lw 11 11 ,y111ho l 11 l 11cl l<mL<

nit

t.1 rlrtlli 111 H

:101

lut _ion i n 11Quid ir n.

G.R.FITTERER
At one atmosphere pressure, m, expresses the minimum amounts of
oxygen which one can expect to find in steel at the respective
carbon contents. Thus, at equilibrium with 1% carbon, a steel would
contain about 0.0022% oxygen. Similarly, a 0.10% carbon steel would
contain about 0.022% oxygen. Actually it is doubtful if this
equilibrium is reached for any length of time by any steel process.
Instead, much more oxygen is found in commercial steels for a given
carbon content.

TABLE I
Equilibrium in the Carbon - Oxygen Reaction in Liquid Steel

Year

Investigator

m = o/iQ x %Q

t 0c

o. 0022

1600
1600

1927

Le Chatelier< 3 >
Styri( 4)
Pierarct< 5)

0.0019

1600

1928

Field( 6 )

0.0023

1600

1931

Vacher and Hamilton< 7 >


Chipman< 8)
Kalling and Rudberg< 9>

0.0025

1650

0.0019

1600

0.0025 +

Variable

Mccutcheon and Chipman(lO)


Fetters and Chipman< 11 )

0.0022

1500

0.0033 +

Variable

0.0028

1600

1957

Marshall and Chipman<1 2)


Fitterer 0 3)

0.0022

1600

1963

Chipman and Elliott<l4)

0. 00186 to
0.00218
0.00245

1500 to
1600
1900 ~ . -

1912
1919

1934
1937
1938
1940
1941

0.002

....

,I

Table I lists some of the determinations of the value of m. by


different investigators over the years. These values were
determine.ct in many differ~nt ways and it should be noted that the
. grand old man of French science, i.e., H.Le Chatellier<3) had
suggested 0.002 for th value otm as early aA 1912.
:HI:.'.

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


Chipman and his students contributed many of the most important
laboratory studies of this reaction and as late as 1963 Chipman and
Elliott< 14 ) suggested that the value of m varies slightly with
temperature from 0.00186 to 0.00245 at 1500 and 1900c respectively.
0.03

0.3
OBSERVED
REBOILED
m 0.00214

500# FeSi
0.2

"'k
a

0.02

%Q

% Si

0.01

0.1

10
0
TIME FROM SILICON

Fig.1.

20
30
ADDITION, MINUTES

Carbon, Oxygen and Silicon Relations in a 200 Ton Basic


Open Hearth Heat of Steel Reboiled with Silicon

In compiling the Herty memorial volume< 2 > for A.I.M.E. in 1956,


it became my task to include the log sheets of some 14 basic open
hearth heats, which were studied in 1930. These heats had been
T~boiled' with silicon - a practice which was in vogue in those
days. As shown in Figure 1, a large ferro-silicon addition was
normally made towards the end of the refining stages of these heats
so as to delay further carbon elimination, while the carbon analysis
was being determined. Rapid analytical methods for carbon had not
yet been developed and it was necessary to restrain the carbon
reaction by means of the silicon deoxidation until the carbon
content of the heat could be reported back to the melter. With this
information he could then calculate the need for further additions
and proceed to tap the heat.
, 9enerally, about 15 to 20 minutes after the silicon addition, the
diffuSion of oxygen from the slag to the metal supplied sufficient
oxygen to convert essentially all of the silicon to Si0 2. This then
asc nded through the metal bath (because of its low density) to
b com n onstitu nt part of the slag.

:10:1

G.R.FITTERER
At this time in the heat, a carbon 'reboil' (i.e. a renewal of
the CO evolution) was observed. This point, as indicated in
Figure 1, was taken to indicate that the reaction had just exceeded
the Q + Q = (CO) equilibrium level. Thus, an analysis of carbon
and oxygen in these steels at this time was assumed to indicate the
equilibrium level.
Although the values were determined at different temperatures,
the average of these 14 (50 - 250 ton) commercial basic open hearth
heats indicated <3 ) that m = (o/<&) (o/i:Q) = 0. 0022. The point to be
made here is that selected data from commercial open hearth heats
can supply a fairly accurate indication of the equilibrium level in
high temperature refining reactions under controlled c.o nditions.
In fact, these commercial heats of 1930 have (more or less)
confirmed the later laboratory results as well as Le Chatelier's
1912 value,
In summarizing all of the work which has been done, we should now
consider that the equilibrium relations in the carbon and oxygen
reaction in liquid steels has been established for all time. We
are, however, not finished with this reaction, but will return to it
shortly.
The Silicon-Oxygen Reaction in Liquid Steel
All medium and high carbon steels are deoxidized with some element
such as silicon so as to prevent the Q-Q reaction from occurring
during solidification. This addition produces a steel which is
solid or free from 'blow-holes', which are caused by CO evolution.
A~.stated previously, a study of the Si-0 reaction in liquid
steel was my first assignment by Herty< 15Y.- The total silicon
content of a steel deoxidized with silicon is the sum of the silicon
which has combined with the oxygen to form Si0 2 and that which is
dissolved but uncomb ined with ol ygen in the iron. Obviously in 'the
study of the reaction

~i +

2Q

= Si0 2 (glass)

one must .analyze the stee~ for both the c ombined and the unc ombined
. silicon. Thus, it was necessary t o deve l op a special analytical
3CM

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


method for Si0 2. The electrolytic method(lG,l7) was used to
determine the values for combined silicon which when subtracted
from the total silicon gave values for the uncombined silicon.
This is the value which must be used in establishing the
equilibrium constant between silicon and oxygen in steel.
The amount of uncombined (or 'free') silicon was determined in
a number of small ingots from a heat of low carbon-plain carbon
steel. The ingots had been deoxidized with varying amounts of
silicon, and the uncombined or 'free' oxygen was determined by the
Ledebur method< lB).
An average value was reported from these experiments for the
deoxidation constant with silicon, as
K0

= (o/.:Si) (%0) 2 = 7.43 x 10- 6

Unfortunately, it was assumed, at the time of these studies, that


the corresponding temperature for this value of &0 was 1600C which
was the temperature at the time the silicon was added.
This was the only value of the constant available at the time and
apparently when Chipman< 8> presented his first paper on deoxidation
with silicon in 1934 he used this value to establish the free energy
equation for .this highly temperature-sensitive reaction.
I t was later obvious that our value should have been reported at
t he freezing point of iron because the silicon reaction continued
t o occur in the liquid steel until it solidified. This correction
was suggested in the Berty memorial lecture< 2 >. Meanwhile, Gokcen
and ' Chipman< 29 ) had apparently discovered this error and Hilty and
Cr afts< 20 > investigated the reaction by further experiments. A very
i mpor tant correction factor was later added by Chipman< 21 ) to his
previous thermodynamic calculations . This involved an increase of
up to 7000 calories in the value of the heat of formation of Si0 2
over older values . others added evidence to verify this
oonc lusion< 22 23 24 ). This correction is obviously included in the
Chipman and Ell i ott< 14 ) 1963 values . The constant may now be
n id r d as being established as far as practical applications
1tl'O

on rn d .

:w

G.R.FITTERER

-6.0

-5.5

I.
2.
3.
4.
5.
6.
7.
8.
9.
10.

HERTY 8 FITTERER, 1928


HERTY, FITTERER 8 CHRISTOPHER, 1931
KORBER 8 OELSEN, 1933
CHIPMAN, 1934
SCHENCK 8 BRUGGEMANN, 1935
SCHENCK, 1945
HILTY a CRAFTS, 1950
GOKCEN 8 CHIPMAN, 1952
CHIPMAN a GOKCEN, 1953
CHIPMAN 8 ELLIOTT, 1963

04
LOG Kd-5.0

-4.5

04
-4.0
1500

Fig.2.

1600
1550
TEMPERATURE, C

1700

The Silicon-Oxygen Equilibrium in Liquid Steel

It might be mentioned in passing that if the Chipman and Elliott


values< 14 ) for the deoxidation of steel with silicon are extrapolated to the melting point of iron they correspond exactly with
the Berty and Fitterer results< 15 ) reported in 1928 and with data
from open hearth heats which Berty, Fitterer and Christopher< 25 )
reported in 1931. All of these values are shown in Figure 2.
Table II records these and oth~r values< 26 27 28 29 ) for this
constant.
In a discussion at the meetinp in 1928 when these results were
reported, W. B. Batfie Hf' stated that in England a different value for
the equilibrium constant had been found. Berty who was always quick
in repartee stated that 'this was the first time that he had ever
heard of the equilibrium constant changing with geographical
location'. At this late date, I might state that I am also pleased
to learn that the constant - has not changed w?.th time. It imparts
durability to the definition of t~e word 'constant '.
:11H1

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


TABLE II

Determination of K0 = (o/~i) (o/i:Q) 2

Investigators

Year

Method

K0x10 5 at Various Temp.


1536 1550 1600 1650

1.

Herty & Fitterer< 15 )

1928

250# heats

2.

1931

open hearth 0.813

3.

Herty, Fitterer &


Christopher <25 )
Korber & Oelsen< 26 )

1933

laboratory

4.

Chipman< 8)

1934 thermo.
calc*

o. 743

5.

1935

open hearth

1. 76

6.

Schenck &
Bruggeman< 27 )
Schenck< 28 )

1945

open hearth

1. 24

7.

Hilty & Crafts<20)

1950

laboratory

8.

Gokcen & Chipman<19)

1952

laboratory

9.

Chipman & Gokcen<29)

1953

laboratory

10. Chipman&. Elliott<14)

1963

thermo.
calc

0.743

---

--

--

--

1. 34 3.60 10.6

1.60 3.00

5.30

2.8
1. 05 2.80

7.00

0. 731 1.00 2.80

7.60

It is obvious from these examples that it is possible for some


plant scale experiments to 'fix' the thermodynamic 'level' for
reactions. Once this has been done , a great deal of the process
becomes better understood and lends itself to control .
A few weeks ago, I asked Dr.Kubaschewski if he would calculate
't he standard free energy equation for the silicon reaction using
the ery latest thermodynamic data. His results coincided almost
exactly with those of Chipman and Elliott< 14 ) as follows:
6G 0 = 148, 100 - 37 . 25T + 9. 15TlogN 0 + 4. 574TlogN 81
.6 G0 = 0 at equilibrium and
'" Us ing K0x10 5 O. 743 at 1600c from Ref.15.
:107

where

G.R.FITTERER
(riBi) (%Q) 2 = 0.72 ,x 10 5 at the melting point of iron according
to Kubaschewski
ilF 0 = 142,000 - 55T

-4. 574(1809)log(Si) (0) 2 or

(%Si) (%Q) 2 = 0.729 x 10 5 according to Chipman and Elliott.


Two sets of values 0 5 , 25 ) obtained from both laboratory and plant
heats back in the 'twenties' were 0. 743 x 10 5 and 0. 813 x 10 5 respectively.
From the practical standpoint, the equilibrium levels for this
reaction have also been essentially established for all time.
However, there is much to be said later about the application of
this information to the deoxidation of steel with silicon.
Metallurgical Reactions with Silicate Slags
A third metallurgical process which has been thoroughly explained
through consideration of thermodynamic principles is the behavior of
FeO in commercial silicate slags.
At the University of Pittsburgh we have been engaged for some
twenty years in co-operative research with a group of companies
which is incorporated as the Ac id Open Hearth Research Association.
In the course of this work it was found (30, 31 32 ) that the
refining of steel by this process could be controlled.completely
through consideration of the Fe0-Si0 2 phase diagram. These steps
have been described in detail elsewhere, but a brief summary of the
principles involved is essential here.
Figure 3 illustrates the currently accepted Fe0-Si0 2 phase diagram
by Bowen and SchairerC 33 ). It was found in the work of the AOHRA
that no matter what oxidizing additions are made to the acid furnace,
or at what temperatur~. the meta.'l and slag always tend to approach equilibrium with the furnace bottom which is composed of Si0 2. In
other words, it was observed that at any temperature, the slag
composition changed by increasing in Si0 2 content so as to approach
line A-C of the figure. This, of course, is the line indicating the
saturation of the liquid s).ag with Si0 2. To the right of this line,
solid Si0 2 and liquid slag co-exist.
:'! Otl

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES

Solid 510
0

(Cristobalite)
2
and Liquid

..:

Solid 510

(Tridymite)
2
and Liquid

f..
2000
'13stite and
Fayalite

Tridymite and
Fayalite

Fayalite
Fe25i04

15
0

100%
FeO

Fig.3.

20

40

6o

Si0 , Per Cent


2

80

100%
5102

System FeO-Si0 2 and Its Role in Acid Steel Refining

One might think that this equilibrium is attained merely by


erosion of the furnace bottom, but fortunately this is a slow
process. Instead the equilibrium is approached by the transfer of
oxygen from the slag to the liquid steel. This loss of FeO reduces
the slag volume and increases the Si0 2 concentration. The important
point which should be emphasized here, however, is that as the slag
analysis approaches line A-C, the carbon reaction subsides and
finally stops completely. In other words, this is an illustration
of a process which involves many tons of liquid steel but which
essentially attains slag-metal equilibrium. At this point, the
metal may be held for hours, without ~ppreciable loss of carbon
content or change in alloy composition. In fact, heats are often
i nt entionally held in the furnace this way for various practical
r as ons .
th

If or 1A 1tdd ci t a heat which has attained this equilibrium,


a.rbon rm t Lo11 w1Jl i,itart ttgain but will shortly return to
:10 I

G.R.FITI'ERER
the same equilibrium analysis providing the temperature has not
changed. In using the oxygen lance*, the carbon reaction subsides
very quickly after the lance has been removed, for the same reason.
The practical usefulness of this discovery will be described
shortly but at this point, the thermodynamic significance of this
phenomenon must be discussed.
From the phase diagram, it is obvious that the activity of Si0 2
in the slag is unity along line A-C because it is in equilibrium
with solid Si0 2 , which is thus in its standard state .
llol Fraction, 510
O.l

0.2

l.O

0.3
Activity of 510

0.4
2

l5J6c

' '

0.6

'

'

o.6

"
.\l

o.4

Normal range of
canposition of acid
open hearth slags
0.2

0
i.o

Fig.4.

0,9

o.6
0.7
Mol Fraction, FeO

,,

o.6

O.S ~ao

Apparent Variation of Activities of FeO and Si0 2 in Acid


Open Hearth Slags with Temperature and Composition

Because of the fact that the carbon reaction ceases as. the slag
composition approaches this saturation line it was concluded several
A tube in which oxygen is
slag le vel.

for~ e d

into the liquid steel below the

:1 10

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


years ago that the activity of FeO in the slag behaves as if it were
zero in a silica saturated slag as shown in Figure 4. This figure
is drawn so that at any temperature, as the N810 approached the
saturation value according to the phase diagram , 2the a 810 = 1.
Meanwhile, at this same point the FeO in the slag becomes 2inactive
ins ofar as the carbon reaction is concerned.
However, this is contrary to previous investigators< 34 35 ) who
as sumed that FeO essentially behaved in such silicate slags
acc ording to Raoults' law (i.e ., aFeO = NFeo)* . This indicates
that instead of zero, the activity of FeO is approximately 0.45
when the activity of Si0 2 is unity.

0.9

0.8

0 .7
0

ol

0.6

0.5

.................

0.4

o.3L.....-......__ _.__ __.__ __._ _


0

10

15

20

~-~-~-~-~-~

25

WEIGHT "

Fig.5.

., Wh n

30

35

40

40

So<\

Variation-of aFeO w ~th Percent Sio 2 <36 )

N 1H Lhn m 1

fni t 1 n.
:111
-

'

G. R.FITTERER
Schuhmann and Ensio< 36 ) studied high FeO-silicate slags with the
iron activity fixed at unity since the slags were melted in iron
crucibles in the gamma iron temperature range.
Their results are shown in Figure 5 where it may be seen that
the activity tends to follow the ideal or Raoult's law at high FeO
concent rations but beyond 20 weight per cent of Sio 2, the trend is
markedly below the ideal line.

SCHUHMANN

AND ENSIO, 1258-1374 C

I. N (SAT SI02l = 0.4484 @ 1340C


2. N (SAT SI 02) = 0 .476
@ 1500C

3. N (SAT SI02l = 0.494


4. N (SAT Si02) = 0.528
5. N (SAT Si02) = 0.52 6

1.0

@ 1535C
@ 1600C
@ 1650C

0.9
o.8
0.7
0.6

a=N

Fe0 0.5

.... ......

.....

0.4
0.3
0.2
0.1
0
0

Fig. 6.

0.1

0.412345

0.2

Measured vs . Predicted Activities for FeO


in Ferrous Silicate Slags

When their resuJ ts are plotted on a graph depicting zero ~ .


activity of FeO at the Si0 2 saturation points, they coincide quite
we 11 with _the line (Figure 6) showing saturation at 1340 which was
the average temperature of their determinations. In fact, the
points coincide exactly with this line at concentrations greater
than N8102 = 0.20.

:11 2

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES

1.0

SCHUHMANN AND ENSIO


'FITTERER

............
,

... a=N... ... ...


o, ......
... ......

0...

0.5

"" "" O(l600C) .,. .,.


O(l340C) ' , ' ,
.,. .,. "'

' ...

... ...

'----'~~~-'-~-""-~---~.L-~'--~~_.__...
~

Nsio 2 --+-

.,,. 2-

...

' LIQUIDS'
SOLID Si02
CRYSTOBALI TE

Nsi02 ( 1600)
SATURATED

t C

1.0

LIQUID

1500

SOLID Si02
TRIDYMITE

LIQUID

1000

Fig.7.

0.2

0.8

1.0

The 'Apparent' Activity of FeO in Acid Steel Slags


as Related to the FeO-Si0 2 Phase Diagram

The relationship of such lines with the phase diagram are shown
i n Figure 7 and it can be seen that the saturation mol fracti on for
i 0 2 has affected the activity of FeO in these slags. Also the
Ht epness of the saturation line or liquidus in the phase diagram
i ii the r egion of 1340C explains the COQlment by Schuhmann and
,naio that a variation of temperature within a range of 116 C
clo 1::1 not 1::1 m to affect the activity of FeO materially. Neither
cl A Lt 1Lff t th saturation concentration of Si0 2 very much.

:1 1:1

G.R.FITTERER
These three examples illustrate large scale thermodynamic
experiments which have established three equilibria and which have
been verified by labora~ ory experiments. However, excessive
concentration on the E!'Iact' level of equilibrium can produce a
sterile science.
'THE FORGO'ITEN EQUATION''

Attention should be given now to some of t he more applied or


engineering aspects of thermodynamics. The same three examples
will be used.
It is obvious from the previous discussion that one of the major
goals of those working in this science is to determine the equi librium conditions for the reaction. The establishment of the
values for the equilibrium constant ar various temperatures is
thereforP. mo~t important .
The commonly used relation for t h is purpose is

wherein, Ka is the equilibrium constant expr essed in activities of


the various constituents and 6F 0 is the standard free energy.
According to Glasstone< 37 ), this expression gives 'the increase
in free energy when the reactants, all in their standard states of
unit activity, are converted into the products, in their standard
states'.
To illustrate the signi(icance of this expression, it may be
suggested that two crucibles containing liquid iron could be heated
to the same temperature. The iron in one of these crucibles would
be intentionally saturated w;it h carbon and the other with oxy~en !
The oxygen and carbon are thus at unit activity. Now, 'if the two
liquids were mixed, a violent reaction would occur but it would
eventually attain equilibrium. Carbon monoxide would, of course, be
given off in great quantities at one atmosphere pressure . This
illustrates the significance of the standard free energy ( 6 F0 ).

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


Certainly this not the way steel is made. The standard free
energy cannot describe the usual refining reactions in liquid steel.
It does, however, define the equilibrium constant referred to the
standard states. The expression 6F 0 = -RTZnKa, is somewhat like
the civil engineer's benchmark, with which he can locate any
arbitrary geographic position with reference to this base.
The nearest approach to the standard free energy in steel
r efining is probably attained in the case of the basic oxygen furnace. Here the surface of a high carbon iron is impinged upon by a
steady stream of essentially pure oxygen with the formation of
carbon monoxide.
However, the 'standard' free energy does not describe the
reaction tendencies in very many metallurgical refining processes.
Instead, the expression which might be labelled the 'Forgotten
Equation' should be used .
Both of these equations are forms of the 'react ion isotherm'
pr oposed by van't Hoff in 1886, but the following expression seldom
appears in the literature:

or
6 FR

= RTZnJ - RTZnK

The standard free energy is thus only a part of a larger and for
.some purposes, a more important equation. 6FR indicates the free
energy of a reaction which starts with the reactants at any activity
t o generate products at any activity with the reaction being
re f erred to the standard free energy. If the initial and final
st ates of the reactants and products respectively are unity, then
J = 1 and 6 FR = 6F 0
INDUSTRIAL APPLICATIONS
My upp al to t he ngineer or applied scientist is that principles
t,h nu ti n fr
nor gy, 6 FR, be given more consideration
wl11 11 <lnr 111111 wI Lil comm r e: II 1 pro <HR s.
Wit h i ts use, thermodynamics
lnv o lv i n~t

:1 1r1

G.R.FITTERER
becomes a more vitally important science, It describes the
characteristics of re '!.jtions which can be made to occur under a wide
variety of conditions. The standard free energy, 6F 0 , describes
one set of conditions which can be changed only by selecting
different standard states or temperatures .
Vacuum Degassing of Liquid Steel
Returning to the three examples used in this paper, Figure 8
illustrates the normal trend of the Q + Q reaction in liquid steel
in forming CO. The reaction for a 0.25% C steel with the usual
degree of oxidation at point J will tend to follow line J-K which
has the slope of

16

12

= 1. 33.

This suggests a reaction free energy value of 6FR


1
calories.

*EQUILIBRIUM

AT

ATMOSPHERE
0

A.

AFR = AF 2
2

+ RTlnJ

B.

AFR =AF~
1

+ RTlnJ

-4,150 \

PRESSURE

= -12,715 cal.

0 .04

= -4,150 cal.

1-

lf

0.03

w
Q.

z
~ 0 .02
>x
0

0.01

0.10

0.20

0.30
CARBON,

Fig.8.

0.40

AF 1 = RTlnK1
., _
= -22,764 cal.

0.50

0.60

PERCENT

The Carbon-Oxyge n Equilibr i um i n.Iron at


One and 0.'1 At mosJ;>h .l'O Pr Hsur
:1 J(J

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


If, however , the same liquid steel is poured into a vacuum
chamber at 0.1 atmosphere as in the tap-degassing process the
'reaction' free energy, 6 FR = -12, 715 calories.
2

At present, there is some difference in op1n1on as to the best


method of deoxidizing steels by the reaction of carbon with oxygen
i n a partial vacuum on a commercial scale . On the one hand, it is
s uggested that the oxygen content for a given carbon should be
reduced to a minimum prior to tapping in the vacuum. Thus the best
possible vacuum or lowest pressure attainable should be used for
the reduction .
However, it is also known that in the commercial process the
most efficient reaction occurs when the tapping stream is 'exploded'
into thousands of droplets. Generally , it has been found that the
~ acuum' must be reduced below about 50 mm. of mercury for the CO
evolution to attain the necessary violence to explode' the tapping
stream . This factor alone apparently cannot always be relied upon to
accomplish the desired degree of deoxidation by this method and
var i able results are obtained.
It might be suggested that a clue to consistent results might be
fo und i n a study of the 'reaction' free energy l1FR expressed
graphically by the line J -K2 in Figure 8. It is po~sible that 6 FR
2
must exceed a certain minimum level so as to overcome an activation
onergy barrier to the reaction in such a rapidly falling stream of
meta l under such a high ferro-static head.
Perhaps the oxygen content prior to tapping should always be
uffic iently high to ensure a minimum 'reaction' free energy (or
.J/K - va lue) in order to 'explode ' the stream and thus permit an
fficie nt elimination of oxygen down to the K2 level shown in
1''igure 8.
neoxidation of St eel with Sil icon After Vacuum Degassing
With an addit i on of 0.2 5% s ilicon to a heat with a normal oxygen
c:o11L nt during a vigorous carbon boil at point J (as in the previous
1x.rn11pl ), th' :-illl on-oxygen reaction wil l pr oceed along line J to K
2Q
32
!11 l'l1111ro II. 'l'h l H I l ru lnLH Lh HlOJK of
= 1.14
Sj
28
:11'/

G.R.FITTERER
0.05

.J

*C+ 0

AT ATMOSPHERE

PRESSURE

0.04
I-

Si+2Q AT

LIJ

<.> 0.03
a=
LIJ

a..

LIJ

(!)

>-

0.02

ti.FR= LlF(sn + RTlnJ(c)

= -16,600

cal.

0.01

Fig.9.

The Deoxidation of Previously Evacuated Steel with Silicon

The 6F 0 valu_e is -42,500 calories at this temperature, i.e., the


melting point of iron.
The amount of Si0 2 inclusions formed in proceeding from J to K
would involve 0.04% silicon and produce about 0.086 weight per cent
of Si0 2. However, non-metallic inclusions should be considered on
a volume percentage basis because this is on the basis of microand supersonic inspection methods and is the basis on which these
impurities affect physical properties. Because of the difference
in density between silica and steel (i.e., 2.3 to 7. 8), 0.292 volume
per cent of non-metallics wou'fd be formed in the above example "wh"ich
only involves 0.04% of silicon.
It has long been known that in making clean steel, the oxygen
content prior to deoxidation should be as low as possible. In other
words, J should be as c~ose as possible to K or below, prior to
deoxidation with silicon.
:J iii

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


In recent years, vacuum processes have been resorted to in order
to reduce the non-metallics to a minimum. If the undeoxidized steel
i s tapped into a partial vacuum of 0.1 atmosphere, the oxygen
content may be lowered to a point below which no silicon will react
even at the solidification point as shown in Figure 9. It can be
seen in Figure 9 that the steel which has been vacuum degassed at
0.1 atmosphere contains an oxygen content which is less than the
equilibrium value for the silicon-oxygen reaction. Hence no
ox idation of silicon will occur and the non-metallic impurities will
have been reduced to a minimum.

Deoxidation of Acid Steels through Slag Control


In the third example which involved the refining of liquid steel
in an acid lined hearth, the concept of non-equilibrium or reaction
fre e energy ( tiFR) is also useful.
With the phase equilibrium and the FeO activity concepts described
previously, one can proceed with confidence to change the operating
procedures for the acid steel processes. One need not hesitate to
throw such a system off equilibrium to his advantage when one knows
where equilibrium is and how it can be approached .
With this information several remarkable changes have been made
in acid open hearth practice. In consideration of Figure 3, furnace
practices have been designed so that with prescribed charging procedure for a given furnace with its special fuel characteristics,
the re is no longer any need to make oxidizing additions such as ore.
The oxidation can be supplied merely by increasing the fuel input so
as to raise the temperature rapidly as from point A to B of the
fig ure. This produces a non-equilibrium-oxidizing condition and the
heat times have been reduced several hours by this manipulation.
Point B of Figure 3 may be considered as a function of J in the
Reaction Free Energy equation and C as K. The line B-C thus is
~8e ntially the driving force for the reaction.
Carbon is eliminated
a~ t he slag composition passes from B to C where the reaction stops
b cause of the 'apparent' zero activity of FeO.
n ra.1Hi11g th t mp rat ur e from the 1340 degree level of
He l111h111n 1111 1i11d T:nH 1 Lo tho l 600 d gr
leve 1 in stee 1 processes, the
:1 trl

G.R.FITTERER
saturation percentage of Si0 2 increases according to the phase
diagram and c<ft'r esponds to a decrease in FeO content in the slag of
some 12 weight per cent.
With a 50 ton heat and a six pe r cent slag weight this amounts
to an addition of approximately 720 pounds of FeO. This, together
with the action it creates , can account for the elimination of about
0. 30% carbon in the 50 ton heat .
In addition, steels in the acid hearth can be partially deoxidized
merely by chilling the slag from point C to D in F'igure 3 thus
precipitating Si0 2 from the slag . This can be accomplished by
reducing the fuel input.
With an excess of Si0 2 in the slag blanket as indicated at point
D and with excess oxygen in the metal, the oxygen will diffuse from
the metal upwards into the slag in an attempt to approach
equilibrium by passing from point D to A.
In this reaction, D is a function of J and A a function of K in
the re.action free energy equation. This procedure will reduce the
oxygen content to near the ~ - Q equilibrium. All of these
manipulations have been made successfully on conmercial heats .
In pr esenting this paper it is hoped that others may be
encouraged to seek thermochemical data from industrial processes
when they are not otherwise available . Some plant work is always
desirable in any case so as to properly define the important
problems and to identify the controlling reactions. Such
applications are not easy but it is rewarding to see them in
operation on such a large scale.
In the work of the new Thermodynamic Center at the University
of Pittsburgh it is planned to explore such applications whenever
possible but also t ~ determine fundamental data where vacancies exist. The bulk of our work will be at higher temperatures than
those which are usually reported and will involve a variety of
metallic and some refractory compound systems.

'

:120

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


REFERENCES
1.

G.N. Lewis and M. Randall, Ther modynamics, McGraw-Hill (1923).

2.

G.R.Fitterer, Deoxidation of Steel, A Memorial to Dr.C.H.Herty,


Jr., A.I.M.E. (1957).

3.

H.Le Chatelier, Revue de Metallurgie 9, 513 (1912).

4.

H.Styri, Chemistry and Metallurgy 20, 478 (1919).

5.

M.Pierard, Rev. de Met. 24, 41 (1927).

6.

A.L.Field, Trans.A.I.M.E. 80, 114 (1928).

7.

H.C.Vacher and E.H.Hamilton, Trans.A.I.M.E. 95, 124 (1931).

8.

J.Chipman, Trans.A.S.M. 22, 385 (1934).

9.

B.O.Kalling and N.Rudberg, Jernkontorets Ann. 121, 93 (1937).

10.

K.C.McCutcheon and J.Chipman, Trans.A.I.M.E. 131, 206 (1938),...

11.

K.Fetters and J.ChiprrRn, A.I.M.E.Open Hearth Proc. 140, 208.

12.

S.Marshall and J.ChiprrRn, Trans.A.I.M.E. 722 (1942).

13.

G.R.Fitterer, A. I.M.E. Open Hearth Proc. 40, 281 (1957).

14.

J.Chipman and J.F.Elliott, Electric Furnace Steel Making 2,


Ch.16, Interscience Publishers (Wiley) (1963).

15.

C.H.Herty, Jr. and G.R.Fitterer, Bulletin # 36, Co-operation


Metallurgical Investigations, U.S.Bureau of Mines and Carnegie
Institute of Technology.

16.' G.R.Fitterer et.al., U.S.Bureau of Mines, Report of


Investigations 3205 (May 1933).
17.

.Jr. T. r.ukashew itsch -Duvanova, Schlackene inschlusse in Eisen


1mcl SLuhJ , Vb V rla.g T cknik, Berlin (1955).

G.R.FITTERER
18.

A.Ledebur, Stah[ und Eisen 2, 193 (1882).

19.

N.A.Gokcen and J.Chipman, J.Metals, 111 (Feb.1952).

20.

D.C.Hilty and N.Crafts, Trans.A.I.M.E. 188, 425 (1950).

21.

J.Chipman, J.Am.Chem.Soc. 83, 1762 (1961).

22.

C.N.Foster and L. M.Foster, J.Physical Chemistry 66, 380 (1962).

23.

W.D.Good, J.Physical Chemistry 66, 381 (1962).

24 .

S.S.Wise, J.L.Margrave, H. M.Feder and W.N.Hubbard, J.Physical


Chemistry 66, 381 (1962).

25.

C.H.Herty, Jr., G.R.Fitter and C.F.Christopher, U.S.Bureau of


Mines Tech. Paper 492 (1931).

26.

F.Korber and W.Oelsen, Mitt Kaiser Wilhelm, Ins.Eisenforsch.


Dusseldorf 15, 280 (1933).

27.

ff.Schenck and E.O.Bruggeman, Arch v.d.Eisenhuttenw. 9, 543


(1935).

28.

ff.Schenck, The Physical Chemistry of Steel Making, British


Iron and Steel Research Institute (1945).

29.

J.Chipman and N.A.Gokcen, Trans.A.I.M.E. 197, 1017 (1953).

30.

G.R.Fitter, Campbell

31.

G.R.Fitterer, A.I.M.E., J.Metals, 1634 (Dec.1953).

32.

G.R.Fitterer et.al., Acid, 10pen Hearth Research Association


Bullet ins 1 and 2, University of Pittsburgh.

33.

N.L.Bowen and J.F.Schairer, Am.J.Science 24, 177 (1932).

34.

F.D.Richardson, Proc.Faraday Soc. Physical Chemistry of


Process Metallurgy, 244 (1948).

~emorial,

Trans.A.S.M. 34, 1 (1945) .

...

APPLYING THERMODYNAMICS TO METALLURGICAL PROCESSES


35.

M.Rey, Proc.Faraday Soc., Physical Chemistry of Process


Metallurgy, 257 (1948).

36.

R.Schuhmann and P.J.Ensio, Trans.A.I.M.E. 191, 411 (1951).

37.

S.Glasstone, Thermodynamics for Chemists, D.Van Nostrant, 283


(1947).

PANEL III
DISCUSSION
Dr.Berger - We will now open the session for the intra-panel
discussion. Dr.Fitterer, it seems to me that in this panel we have
thoroughly discussed many interesting applications of thermodynamic
processes. Dr.Gilles' discussion of some of the low pressure systems
and some of the effects we get by going to low pressures was very
interesting and yet it seems that much of the discussion this after noon was relative to vacuum degassing. Also, some of the reactions
of the metal-metal oxide systems will be bette~ understood as a
r esult of the experiment which he described. We will start with
Dr . St.Pierre.
Dr. St . Pierre - I would like to bring up a point with Dr.Langenberg
about the large error in the temperature difference which can be
produced by a rather small error in the interaction coefficient. We
we re just making some calculations and a 5% uncertainty in the interaction coefficient produces a 20 degree difference between the
ca lculated temperature and the observed temperature for the chromiumcarbon equilibrium. Even a two per cent error would represent about
a s ix degree uncertainty in the temperature.
Or.Langenberg - I think the point that Dr.St . Pierre makes is
exce l lent . The fact that a five per cent error in determining an
interaction parameter can cause a temperature deviation of 20 degrees
at 2, 000K is very important . Also, the errors in analysis for carbon at very low percentages is a problem which can lead to errors,
especially when you get down to less than 0. 03% carbon which is not
typica l of stainless steel blowing conditions .
llJ.Fitterer - Well , along that line, Dr . Langenberg, I was wondering

if any of your slags were studied for probable formation of some


spi nels? We tried to apply Hilty's equation to some low chro~ium
st els in t he ac i d open heart h, but we found that there is a very
Htrong for mation of an FeO Cr 2o3 spinel. This threw us completely
off. In fact , we proved t h is by studying thin sections of slags.
W( co uld s e th spinels in every one of t hem.
Ur. l.111t J.1 1111 lH' g - No Prof'c HHOr l"ULor r , w did not
11 Lur l lnt1 illl 11 11.v of' 1. 111 11 1nKH.
W1 11HH11m ci Lhat Lho

_:I!? '

make petrogr aphic


act vity of t he

PANEL III

metal oxide was unity in the oxygen impingement zone. If the slag
is not the pure oxide or is not saturated with the oxide, the activity
is less than unity. There was some German work done in 1961 where
the slags contained quite a bit of lime and fluorspar and the slags
were on the metal before blowing started; in this case, there was
not agreement with the other investigators who studied this system.
We feel this difference occurred because the activity of the
chromium oxide was not unity. The temperature difference is what
one would expect assuming that the oxide activity was less unity.
We have not studied our slags for spinel formation.
Dr.Fitterer - I think und~r the conditions that you and also Dr.Hilty
worked, this is a very important contribution, and I think it's
probably been verified. The only thing is that in more oxidizing
slags, you're liable to be affected by spine! formation. Your slags
were not too far from a neutral condition as far as acid and bases
are concerned.
Dr. Langenberg - They were essentially pure chromium oxide plus the
iron oxide which was formed concurrently. The slags would be
saturated with the chrome oxide, so the activity would still be
unity..
Dr.Fitterer - I'd like to ask Dr.Gilles what he's done on metallic
systems by his method.
Dr.Gilles - None of my students has worked on a metal since 1955 or
so when two of them measured the vapor pressure of cerium. That
work has not yet been completely analyzed because the free energy
function for gaseous cerium has not yet been available. Some of the
borides with which students have been working have many metallic
characteristics.
Dr. Fitterer - I watched your work with interest on metallic systems. .
with oxygen, and this is what I mean by being very important
relative to the vacuum degassing work.
~

Dr.Langenberg - I'd like to comment although it's probably a littl~


premature in view of the _d iscussion coming up this afternoon on
vacuum degassing. There is an indication from quit e a f ew

DISCUSSION
investigations that in the commercial vacuum degassing processes
t here is a tendency to approach an equilibrium with a carbon monoxide
pressure of 100 millimeters. In other words, it' s very difficult to
get carbon-oxygen products which are below this pseudo-equilibrium
val ue equal to a CO pressure of 100 millimeters. If you plot carbonoxygen results for the various commercial degassing processes they
tend to level off at this value. Now, there probably are ways, and
we think we know some, that wil~ permit us to go below these values.
Ther e might be some disagreement on this because it is a highly
commercial field with very strong sales pressures depending on the
degassing process you have.
Dr.Ber ger - Dr.Langenberg, I'd like to ask you if you've carried
your work down to the very low carbon contents i n some of the
stainless materials. I know some of your curves stop around 0.06%
carbon, or might that be a proprietary question?
Dr. Langenberg - No, the data which you have in the paper is all we
have developed in these experiments.
Dr.Berger - Thank you. Now, I would like to distribute the
ques tions from the audience to the various participants, and then
we' 11 discuss them one at a time. I think we might start with
Dr.Gi l l es . He has one that's rather interesting,
llr. Gilles - This is not a question. It is an assertion . It comes
from Pr ofessor Chipman. I quote, 'Headline in Pravda. Distinguished
~cien t i st states - U.S. vaporizes congruently.'
(Laughter)
llr . Fitterer - I have three questions from the audience, all of which
are essentia lly the same. These are relative to my suggestion that
the activity of FeO in a silica saturated slag tends towards zero.
One of t he discussers states that aFeO cannot be equal to zero in such
slags. Let ' s put it this way. It is probable that the time element
i.e., t he rat e of diffusion in a very viscous slag is a controlling
fact or. Never t hel es s , heats ~n the acid furnace may be held upwards
t,o an hour, for example, wit hout a carbon reaction under a silica
Ho.t.urnt d slag. This is tr ue even with the silicon content under
n fuct, t h silicon reaction will r everse under these
0. 01%.
con cl It, I onH HO nl'l Lo nor n1:10 th silicon content . In all of the
1x1>1. l'l 1111 11 l.11 ll<1<l'lh1cl In Lh lri t)l\IJ 1 l, h 1.-1 11 n
nt nt of the metal

PANEL III
was under 0.01%. So, let's say that FeO behaves as if its activity
is zero under these conditions .
Dr.Berger - Does that answer the question satisfactorily to the
benefit of those who participated? Are there any further
discussions? Dr . st.Pierre .
Dr . St . Pierre - Two questions have been referred to me from th r!
audience. The first one is , 'Can the 6F of the reaction be w~ed in
general to calculate rates of re actions quantitatively? For example,
by the methods of thermodynamics of irreversable processes?' I
have to say no . In order to apply irreversable thermodynamics, it is
necessary to have all of the coefficients on the intensity gradients
within the system, and, of course, these are unknown for most
systems. However, the magnitude of the overall 6F {s some indication
of how fast we might expect a process to procede, so it is a
qualitative indication.
The second question is , 'How could knowledge of the free energy
balance be used to change processing procedures?' There is some
possibility that this can be done quantitatively. If a proposed
minor change in a process results in an increased dissipation of
free energy, all other things beirig relatively constant, the process
should procede a little faster. When the kinetics are not changing
drastically, that is , there are no changes in the kinetic processes,
an increase in the dissipated free energy should represent an
increased reaction rate.
Dr.Langenberg - The first question I have is from Mr.Burr, United
states Steel. 'Have we made any production scale studies on the
effect Of bath temperatur e at s! ag- off on chromium recovery?' From
equilibrium considerations, we would expect that the higher the
temperature, the poorer would be the chrome recovery. But. as you
know from your operating experie~qe, the higher the temperature, t ~e
more fluid the slag beco mes. This then becomes a kinetic considera~
tion which masks and overwhelmes any of the equilibrium conditions .
In fact, if you looked at the two equations, one representing
equilibrium, and the other which we developed from th~ 12 production_
heats you noticed that the sign in front of the temperature term was
different. In practice, the "higher the t emperatur e during the
r eduction. step, the bet t er is your .chrome recovery . However, you
f'
have to watch t he r fractories.
:!',II

DISCUSSION
Dr.Langenberg - The second question refers to the slides showing the
difference between the measured and calculated temperatures for the
chromium-carbon-nickel system. It asks 'what is the meaning between
the two degree difference at zero per cent nickel?' We are
referencing our work against the Hilty equation within two or three
degrees. Instead of listing our own constant which is only slightly
different, we went along with the Hilty values without concern for
the two degree error in temperature measurement. There was no use
developing a new equation for that slight difference since Hilty's
equation is already established in the literature and probably built
into several linear programs by steel companies that are using it.

Another question - 'in the carbon-manganese - nickel system, do


you have a loss of manganese due to volatilization during oxygen
inj ection?' Yes, and we have ignored it. Manganese volatilization
would lower the final temperatures. This has been reported by
Samar in and others.
The final question - 'How does the manganese in a reduced slag
compare to the per cent chrome?' We have not carried out equilibrium
studies which include manganese in the system. To get a better
rec overy of manganese you should operate at a little higher basicity
in your slag. You should operate at a V-ratio of about 1.7 compared
to a V-ratio of about 1.4 to 1.5 for the straight chromium grades.
Dr.Gilles - Dr.C.H.Lee from Grumman Aircraft asks two questions .
'Do you have a pressure composition diagram for tungsten alloys?'
The tungsten-oxygen system has been studied by Ackerman and Thorm in
the Argonne National Laboratory, and the results were published
recently, I believe, in the Journal of Physical Chemistry. So far as
I know , that is the only system involving tungsten in which a
pressure composition diagram has been studied. I'm confident I can
draw it properly for the tungsten -boron system even though we have
made no extensive measurements .

, His second question is, 'Can you predict the composition changes
during iso-thermal evaporation of binary alloys? If yes, how?'
Well, _yes, s ure. Brewer and Rosenblatt have as a matter of fact
don pr cis ly that for metal phases containing oxygen. The data
Umt nrn r oq n1r cl in or,d r to mak an appropriate prediction are
l.hoHi

Ind 1<:1 t< <I I 11 'l'1tb l 11 I

of my pn.p r.

PANEL III
Professor Hultgren has been asked, 'In measuring equilibrium
vapor pressures, equilibrium is established with the surface of the
sample. For non-stoichiometric samples the surface becomes more
depleted of the more volati'ie component. The result depends on the
relative rate of diffusion and evaporation. Can you comment on
this?' H~ recognizes the difficulty that all of us in vaporization
work recognize, and we do our best to ascertain first of all whether
surface impoverishment is causing difficulty. If the surface is
being depleted so as to cause departure from equilibrium, we take
steps to rectify the difficulty. The means that we use to establish
whether depletion of the surface is causing problems are several.
Most simply, one can observe the vaporization rate as a function of
the extent to which vaporization has already occurred. If the rate
is diminishing with the temperature, we know we're in trouble. One
sensitive means of indicating whether the rate is following is to
observe the vaporization in the mass spectrometer. Now, if there
are problems, several remedies can be used. First , one can stir the
sample occasionally. Second, he can change the proportions of the
two phases if there are two condensed phases present. Third, he can
change the orifice size in the crucible in order to estimate
appropriate corrections to be made for the effect with both orifices.
Finally, he can work with a sufficiently small orifice in some
circumstances so as to reduce the surface impoverishment to a
minimum.
Mr.Alex Simkovish from Latrobe Steel Company asks the following
question. _ 'From your pressure-composition diagram for the titaniumoxygen system, what was the oxygen pressure at which Ti0 2 began to
decompose?' The diagram that I drew was for a constant temperature.
It was not a quantitative diagram. The pressure at which Ti0 2
loses oxygen depends, of course, on the temperature. The feeling
that. I usually have is that i f one heats Ti0 2 in a vacuum to a
temperature in the neighborhood of 1200 to 1400 degrees centigrade,
appreciable oxygen losses
... occur. ,1
Dr.Berger - Are there any additional questions from the audience?
If not, I'd like to briefly summarize the program this morning by
saying that Dr.St.Pierre told us something about the theoretical
calculations of the Free ~nergy Balances and illustrated those v ry
effectively with laboratory applications.

PANEL III

Dr.Langenberg illustrated how thermodynamic data and modifications


can fit modern stainless steel manufacturing processes by describing
the evolution process from the chromium stainless steels of 400 series
to the austenitic chromium nickel steels.
Dr.Gilles gave us an interesting description involving the
applications of thermodynamics to vaporization studies at low
pressures.
Dr.Fitterer described some of his experiments related to the applications of deoxidation processes for plain carbon steels. These
were illustrated with the carbon - oxygen reaction followed by the
deoxidation constant for the silicon - oxygen reaction. He described
the equilibrium constants for the Si-oxygen reaction and tabulated the
changes from the year 1928 to 1963. His brief description of some
of his work with the Acid Open Hearth Research group gave some
insights as to what we may expect concerning vacuum degassing.
I think these authors are to be congratulated for their excellent
papers. I'd like to thank the audience for their kind participation
and interest that has been shown.
I have just been informed that Mrs.Kubaschewski is in the audience.
I am pleased to present Mrs.Kubaschewski. Will she please rise? We
would like you to meet the other half of the Kubaschewski team.
Mrs.Kubaschewski is well-known as a thermochemist herself because of
her many calorimetric experiments. Some of her publications were as
Miss O.Von Goldbeck.

PANEL IV
NEEDS FOR FUTURE APPLICATIONS OF THERMODYNAMICS
Chairman

Dr.J.F.Elliott, Professor of Metallurgy,


Massachusetts Institute of Technology, Cambridge, Massachusetts

1.

Requirements for Space Materials


by Dr.Charles E.May, Head, High Temperature Chemistry
Section, Lewis Research Center, National Aeronautics
and Space Administration, Cleveland, Ohio

2.

The Aluminum Industry


by Dr.P.T.Stroup, Assistant Director of Research,
Aluminum Company of America, New Kensington, Pa.

3.

Problems in the Non-Ferrous Industries and a Note


on Estimation of Heat Capacity
by Professor H.H.Kellogg, Extractive Metallurgy,
Columbia University, New York City

4.

Unfinished Business in the Thermodynamics of the


Basic Oxygen Steel Making Processes
by Professor William O.Philbrook, Metallurgical
Engineering Department, Carnegie Institute of
Technology, Pittsburgh, Pa.

5,

Vacuum Degassing of Liquid Stee 1


by Mr.J.E.Fogarty, Assistant to the Superintendent,
Smith Division, Republic Steel Co., Canton, Ohio

6.

The Problems in Continuous Casting


by Dr. T. B. Winkler, Assistant General r11anager,
Homer Research Laboratories, Bethlehem Steel Company,
Rethlehem, Pa.

7.

D1A ttHt-d on

PANEL IV
Dr.Elliott
Ladies and Gentlemen, this is the final session of the Conference,
and as you can see by your program, it is entitled 'Needs for Future
Application of Thermodynamics' . We have a panel of six people who
are eminently suited and able to point up what these needs are in
several specific areas. We hope to cover more than just those areas
as we get into the discussion area of the session.
On our Panel we have several college faculty members and several
people from the industrial world who are face to face with some
real problems of making things go even though they don't have all
the information necessary to make a complete and final analysis.
As one looks into the availability of information on phases and
reactions of interest to process metallurgy, it is clear that
we still do not have facts with regard to the thermodynamic
properties of many of them. In addition, it seems that any kinetic
study that we undertake is severely handicapped at the start for
lack of some important thermodynamic data. It is our responsibility
this afternoon to point out some of these areas that are critical to
the technology.

Our speakers today will mention directly and obliquely the areas
where these numbers are yet unknown. We would like the audience to
bring up issues where these numbers are known.
Dr.Charles E.May will speak on Requirements of Space Materials and
I think he has primary reference to thermodynamic information.
Dr.May.

REQUIREMENTS FOR SPACE MATER IALS


by

Dr.Charles E.May
Head, High Temperature Chemi st ry Section
Lewis Research Center , NASA

During the first part of this conference, we have been discussing


how thermodynamics can be applied to various problems . This part
of the program, however, is intended to emphasise the nature of the
particular problem rather than the exact met hod of solution .
As we all know, the requirements for space materials are often
far more stringent than those for other applications. Depending on
the specific use, materials may either have to be refractory or have
to withstand stresses caused by subjection to liquid hydrogen
temperature. A refractory must inevitably contact another refractory
and thus be compatible with it . In many applications extremes in
thermal cycling are involved . Moreover, these same materials
should be light in weight while having a high rupture and notch
strength, and yet possess sufficient ductility to allow easy fabrication. Over and above this, materials must be compatible with the
environmental conditions of the upper atmosphere and outer space,
i ncluding activated molecules, ions, micrometeorites, vacuum and
radiation. Obviously, such materials must not only include metals
but ceramics, plastics, elastomers and lubricants. No single
material possesses all the desired properties; a separate selection
of the proper material must be made for each individual application.
This talk will concentrate on three general aspects of the space
materials problem, In each of these aspects, thermodynamics can be
an invaluable tool for the solution of the probiems involved. The
topics to be discussed are: the stability of refractories, the
compatibility of solids at high temperature and the effect of
environmental gases on materials.
,When talking about the stability of refractories we immediately
th i nk of tungsten and so does the designer of a piece of space hardwar e . Occasionally, molybdenum may be chosen due to ease of
fa brication. Ther e ar e , however, quite a few refractories which
m ~ h huv b n ho n. Figur e 1 is a partial list of such materials
1u11l Hhowr-i how Lho1r 1111 1ti n f{ poi nt s compar e with that of tungsten. Of
Hl'111 t,111 1.-1 111.1 1111t., 111 1111i p11, lri liaf n1um arb 1d wh ich has the highest

CHARLES E.MAY
MELTING POINTS OF SOME REFRACTORY MATERIALS
M. P.,
OF
c

Ta
Mo
Nb

6600*
6120
5430
4710
4530

Th02
Zr02
Cr203

5970
4920
4410

M. P . ,
OF

M.P . ,
OF
4 TaC : l ZrC
Hf'C
Tac
ZrC
NbC
T1C

we

~~c

Mo2C
S1C*

7110
7030
7020
6380
6330
5710
5190
5170
5120
4870
4710

ThB4

5880
5500
5430
5400
5250
5180
4530

Hf'N
TaN
ZrN
T1N

5990
5600
5400
5310

HfB2
ZrB2
TaB2
T1B2
NbB2
WB

*SUBLIMES

Fig.1
melting poi nt of any known compound . Some of the other carbides
should also be considered . These materials have the properties of
metals with room temperature electrical conductivities greater than
some of the common metals . These refractories are not chosen for
space applications at present because many of their properties are
unknown . Their stability limits at high temperature are not
adequately known and the nature of their vapor species is for the
most part speculative . In short , there is a great need for a concerted investigation into the thermodynamics of these materials;
the results of such an investigation would allow these materials to
be considered for space applications .

A few additional comments should be made concerning these refractories . First, these materials have high heats of formation.
For this and other reasons, they should definitely be considered to
be compounds. However, for sqme of these materials, in particular
tantalum carbide, the optimum properties apparently occur at other
than the expected stochiometric composition. Figure 2 illustrates
this; tantalum carbide has a minimum magnetic susceptibility at
Tac 0. 8. Other proper.;t.ies are a'l'so a maximum or minimum at this same
composition. Thermodynamics might also give an insight into this
unusual situation .
Looking a bit further , one should note that the thermodynamics of
the pseud9-binary alloys . ~f the refractories would al s o be of much
interest. It has been reported that haf nium carbi de .and t antal um
carbide form an alloy with a high"er me l ti ng p i nt than hafni um

REQUIREMENTS FOR SPACE MATERIALS


MAGNETIC SUSCEPTIBILITY OF TANT ALUM CARBIDE
sx10-&
5.4x10-~
I

4
VOLUME
SUSCEPTIBILITY,

CGS

.8
.9
MOLAR RA TIO, C/T a

1.0

Fig.2
carbide, itself. Other examples of this type may also exist among
the refractories.
The second general topic to be discussed is the compatibility of
solids. The important item to be discussed here is the various
problems where the question of compatibility arises. One example
where two phases contact each other is in the strengthening of
materials; the well known method of precipitation hardening dates
back many years. However, there are other methods of strengthening
involving more than a single phase; the potentialities of some of
these are being investigated at our laboratory. One of these involves
TRANSVERSE SECTION
TUNGSTEN REINFORCED COPPER COMPOSITE COMPOSED OF 48.3 - 5 MIL WIRES

UNETCHE'D MAG:

Fig.3

:1:w

SOX

CHARLES E.MAY
the embedding of whiskers or fibres in a matrix in such a way that
the fibres will retain their high tensile strength. The systems of
most practical interest are, of course ; those for which strong
whiskers are available. The preliminary study at our laboratory is
being carried out for fine tungsten wire in a copper matrix . Figure
3 shows a photomicrograph of a test specimen.
Another strengthening method involves dispersion similar to the
SAP process for aluminium. The first metal investigated was nickel.
The strict SAP process would require submicron particles of nickel
oxide to be dispersed in submicron particles of nickel metal. Since
the oxide is soluble in the metal at elevated temperatures, a judicial
selection of another oxide had to be made; in this case, the choice
was alumina. Figure 4 shows some results; the successfulness of the
method is easily noted. Application of this method would require
the finding of a compatible oxide or a substitute for each metal of
interest.
STRESS FOR RUPTURE VS RECIPROCAL OF MEASURED
INTERPARTICLE SPACING IN Nl-Al 2 0, ALLOYS
10 HRS AT 1500 F

14

10

STRESS,

IOOO PSI

PURE NICKEL

.I

.2

.3

.4

.5

RECIPROCAL OF MEASURED INTERPARTlCLE SPACING,

10

.6

P. ' MICRONS

TlP. ', MICR ON S

Fig.4
Another example of"'two phases' in contact is in the containment
of a nuclear fuel in a matrix metal. Our selection of materials is
quite restricted in this case. The matrix must be a refractory and
the dispersed material is limited to the few compounds that act
as a nuclear fuel.
The final two phase problem is associated with coat i ngs . . It is.
true that, many times , th prob l m i nvolved is.J;ho Atab ility of t he
:MO

REQUIREMENTS FOR SPACE MATERIALS


coating; however, the compatibility with the base metal must also be
considered. A typical example of this is the use of a disilicide
coating for the protection of tungsten from oxygen. Another example
is a coating to protect chromium alloys from nitrogen and its
embrittling effects.
This leads to the final general topic to be covered, that of the
effect of gases on materials. This problem is present even though
coatings are used since the gas may react with the coating, itself,
or the gas may still diffuse slowly through the coating to the substrate. It is not always gross reactions such as oxidation that we
may be concerned with. Many times slight solubility of gases in
materials has detrimental effects such as embrittlement of tantalum
by hydrogen and of chromium by nitrogen. Such problems involving
solubility can be treated using thermodynamics.
Unfortunately, many of the gas-solid reactions involve kinetic
problems and not thermodynamics directly. Fortunately, however, for
a few systems, in particular the oxidation of platinum metals, the
thermodynamic values can be related to the activation energy and rate
of reaction in much the same way as we do in the case of
vaporization. The necessary condition is that the system must involve the formation of an unstable volatile component. Thus,
occasionally thermodynamics can help solve a kinetic problem.
In the afore discussion, I have commented on only a few of the
general problems associated with materials used in space and the
outer atmosphere. At present, problems are circumvented by designing
around the deficiencies inherent in the presently available materials.
Development and research on materials with desirable properties will
permit many simplifications in the design of spacecraft of the
future.

THE ALUMINUM INDUSTRY


by

P.T.Stroup
Assistant Director of Research, Aluminum Company of America

From a devout thermodynamic viewpoint, there should not be any


aluminum industry because aluminum is unstable towards its environment and proceeds toward the lowest energy state as the oxide.
However, the obstacle of reaction rate so frequently met in thermodynamic processes is effectively present as a surface oxide film
which completely stops the reaction in a manner obnoxious to kinetics
idealists.
In the Hall electrolytic process of aluminum ~melting, the
electrolyte consists essentially of molten cryolite in which is
dissolved aluminum oxide. Despite the many years of commercial use
during which this electrolyte has been employed and investigated,
the ionic constitution of the bath and mechanism of electrolysis
are not completely understood. Laboratory experiments involving
molten cryolite are conducted under difficulties caused by the
elevated temperature attack of equipment not only by the molten
cryolite but also by molten aluminum and fumes from volatilization
and hydrolysis of cryolite.
Of the several schemes proposed for dissociation of cryolite, the
present preference favors the formation of sodium tetrafluoraluminate.

This mechanism has been found to agree with calculations of the


liquidus curve by Grjotheim 1, the degree of dissociation calculated
by Frank and Foster 2 from the density values reported by Edwards et
al 3 and identification by Howard 4 of sodium tetrafluoraluminate
NaAlF 4 in the quenched vapor from cryolite.
'The nature of the solution of alumina in molten cryolite is not
as well understood. At present it is generally accepted that a
r action occurs between alumina and cryolite to form a new ionic
p c s rather than a simple physical type of solution.
'l'ILhl
h1iVf

l 11111. 11 Llw mn.uy ell rfor nt kind1:1 of ionic species that


1>1111 l!l'OIHH1d l, o 1xnl1L111 I.Im Ho lut n of alumina in cryolite
:14:1

P.T.SI'ROUP
and the subsequent behavior in the electrolytic production of
aluminum. The list starts with an oxygen ion and progresses
through stages of aluminate, oxyfluoride and aluminofluoride to
the fluoride ion.
TABLE I

Postulated Alumina Cryolite Anions


o-2

Alumina Formula

Allmand

1924

Alo 2

Aluminate Ion

Pointelli

1940

Al204 -2
Alo 3- 3

Alumina Solubility

Frejacques

1949

AlumiIJ,ate Ion

Fedotier

1923

C0 2 Solubility

For land

1953

Co-ordination of 4

1950

Cryolite Eutectic

Boner
Treadwell

1940

Solid Solution

Griinert

1942

Al302Fl4-9
AlOF 5 - 4

Alumina Solubility

For land

1953

Solid Solution

Grunert

1942

A1F 4-

Co-ordination
Phase Diagram

Boner
Frejacques

1950
1949

Dissociation of AlF 6
Dissociation of AlF 6

Fi:irland

1953

For land

1953

Fedotier

1932

Drossbach

1934

Al0 2F 2- 3
AlOF 2
Al20# 4-2
AlOF3 - 2

Al 2F9 - 3
AlF - 2
5

Al2F11.5
AlF 6 3
F

Ionization
Dissociation

From this wide choice Foster and Frank 5 have selected Alo 2 and
AlOF 2- to be the most accepta~\e from a thermodynamic approach ~~~g
Temkins 6 formulation"for the activity of a component in an ionic
melt from cryoscopic data. The log of activity of cryolite from
several ionization patterns in the cryolite-alumina system was
plotted as a function of temperature and the species Alo 2 and AlOF 2showed the closest similarity to the slope for the cryoscopic heat
of fusion. The presently. favored mechanism for solution of alumina
in molten cryolite. is:

THE ALUMINUM INDUSTRY

Frank and Foster 7 identified the ions that carry the current in
the electrolysis of alumina in cryolite by transport experiments
employing the radioactive tracers sodium 24, fluorine 18 and
aluminum 26, and following migration across semipe rmeable membranes
of highly sintered alumina or hot pressed boron ni t r ide. Substantially all the current was carried by the sodium ion from anode to
cathode. About 1% of the c'urrent was carried from cathode to anode
by an anion in which the fluorine to aluminum ratio was 2 to 1. The
oxyfluoride ion, AlOF2, was suggested as that anion .
The simplest explanation for the production of metallic aluminum
at the cathode requires a reaction between the discharged sodium ion
and some aluminum containing species , such as AlOF 2 -. A more positive
i dentification of this and other ionic species may some day be accompl ished by Raman effect studies of molten cryolite if a suitable
container can be developed for such experiments.
A,lmost nothing is known concerning the electrode reaction at the
an ode in the Hall process so much study is needed. The predominant
product is carbon dioxide mixed with varying amounts of carbon
monoxide. It is thought by most investigators that carbon dioxide
is the primary product and any carbon monoxide is formed by reaction
of carbon dioxide with carbon, aluminum or sodium in the form of a
vapor, mist, fog or suspension. This behavior is quite important
commercially because such a reaction means a loss of aluminum and
a lower current efficiency.
Another area of interest in the aluminum industry is concerned
with carbothermic smelting recently reviewed by Stroup 8. Thermodynamic calculations and experience have taught that all the major
oxide s in bauxite except zirconia are reduced by carbothermic
smelting before alumina is reduced, as shown in Figure 1 where the
free ~nergy of formation of these oxides is shown as a function of
temperature and indicates the temperature at which metal formation
might be expected from carbon reduction. In practice, the oxides
rlo not behave as simply as predicted. This is partly caused by the
r rnution of carbides , i nt rrnediate compounds and volatile sub0011111mnldA. A ording t . this dntn, al nmin um oxide should be reduced

P.T.STROUP
- >O

-40

_,o

~
~

i
~

-60

-70

-80

- 90

-100

t!)

~w

-110

~ -120

1000
TEMPERATURE,

Fig.1.

2000

3000

DEGREES K.

Free Energy-Temperature Relations in Carbothermic Reduction


of Oxides found in Bauxite

by carbon to give aluminum intimately mixed with carbon monoxide


with a pressure of 1 atm at a temperature around 2100c.
It is now realized that this simple picture of alumina reduction
is not possible. Nevertheless, it is the thinking that has guided
investigators on this problem for many years in the past and incited
a period of optimistic activity, which still exists, on carbothermic
reduction of alumina with the hope of superseding the electrolytic
process. Many of these past efforts are now considered of doubtful
value because of a lack of realization of the actual reactions and
products involved. ExtremeJy small charges were used in the experiments and inadequate or inaccurate analyses were made of the
products after completion of the reactions. The reactions of
interest were assumed to be:

'

4Al + 3C

...

THE ALUMINUM INDUSTRY


The absence of metal in the products from experiments based on
these reactions was blamed on carbide formation and the back-reaction
of aluminum vapor with carbon monoxide.
A new concept of the reaction between alumina and carbon was
i nt r oduced by the identification of the products at various stages
of the reaction by Foster, Long and Hunter 9 utilizing photomicrographic, X-ray, and chemical methods of analysis. The presence of
intermediate compounds and reactions had been suggested previously,
but identification of these compounds was not accomplished until
Foster demonstrated the existence of the oxycarbides, Al 4o4c and
Al 2oc, formed by the reactions:

The latter reaction was identified in photomicrographs of melt


s amples taken during the reaction. A background matrix of aluminum
tet r aoxycarbide was shown to be reacting with plates of aluminium
carbide to form monoxycarbide. The existence of these compounds
has now been confirmed by others and has opened the way for proposing
many new equations to explain a series of reactions occurring between
al umina and carbon.
Fi lonenkolO followed the course of the solid-solid reaction between
al umina and carbon by microscopically observing the appearance and
tr ansformation of phases in charges containing 4, 15 and 26 pct
petroleum coke. The initial product corresponded to a spinal Al 3o4
or AlO Al 2o3 formed from aluminum suboxide and alumina at a temperature as low as 1615C by the reaction:

The spinel transformed into the tetraoxycarbide Al 4o4c' in


increasing amounts as the temperature increased in the range 1680
to 1860bc when the carbon content was 15 pct and into the monoxycarbide Al 2oc when the carbon content was 26 pct. These results
onfirmed Foster' s announcement of the two oxycarbides, which were
ohta d by crystallization from molten cond i tions.

P.T.STROUP
The formation of suboxides can occur by several proposed types
of reactions:

Brewer and Searcy 11 concluded that Al 2o is formed when alumina


is heated with reducing agents, while AlO and atomic oxygen are the
products when alumina is heated alone.
Cochran 12 checked the conclusion of Brewer and Searcy by showing
that the spectrum from vapors of the alumina-aluminum reaction
exhibited the presence of Al 2o and absence of AlO.
The aluminum-producing reaction proposed by Miller, Foster and
Baker.1 3 is the one between aluminum tetraoxycar bide and aluminum
pa rb i de

The reaction between tetraoxycarbide and carbide is theoretically


capabl e of producing either the monoxycarbide Al 2oc or metallic
aluminum by .the equations

The Al 2oc may dissociate to form metallic aluminum and CO under .


the conditions impo~d in the ~recess, but there is no evidence ~ at
present to indicate such behavior .
The composition and temperature range found best for metal
formation, according to Foster 14 , is shown in Figure 2.
The, weight ratio of alumina t.o carbon and the react i on t.emperature
wer e fowid t o be i mportant factors which gov.wned th amount and ki nd
:11111

THE ALUMINUM INDUSTRY

"'
~"'
0
<I

a:

"'u

0:

.,."'

~----.----------!

Al20C

+ LIO.

I
I

J. _____,

O'
I

I I

I
I

...,..~ Al404C

__L_

"'

4 l

a:

,/
I

,\

-'I
+ I
N

LIO. /

~I
1900

I
I

"'
"'
"'0

a:

"'
"'::>

ci>

I
I

2000

ALUMINUM
PRODUCING
REGION

2100

+ LIQ .

I-

1800

'

20

40

60

80

MOLE PER CENT Al 4 C3

Fig,2.

Approximate Location of Aluminum Producing Region in the


Al 20g-Al 4C3 Phase Diagram

of product obtained. The effect of different ratios is shown in


Table II where the different products are respectively spinel, suboxide, tetraoxycarbide, monoxide, monoxycarbide, metallic aluminum
and carbide, all of which have been separated and identified.
TABLE II

Effect of Alumina Carbon Ratio on Reaction Products


Alumina Carbon Ratio

Theoretical
Mols C/Mols Al 2o3
1/3

Product
Al 3o4

96:4
90: 10

AlO

85: 15
81:19

1 1/2
2

Al 4o4c

74:26
74:26

3
3

65:35

4 1/2

Al 20
Al 20C
Al
Al 4C3

P.T.STROUP
400

300

._,

<.>

200

a:

...

"
:!j

100

:':

......

Al20:s
Af20

e:

Al 2 0,

+ 3C

2AI

+ 3CO

+ 2C

2CO

"'

"

'

- io~o~o---.,.1.,.i
oo~o---,-'000---,--'ooo
TEMPERATURE,

Fig.3.

Free

Ener~y-Temperature

DEGREES K.

Relations for Carbothermic Reduction


of Alumina

In Figure 3 is shown .6F as a function of temperature for four


competing reactions:

Al 2o3 + 3C

2Al + 3CO

At temperatures around 2000Pc the lower three curves to produce


carbide, suboxide and metal are quite similar in ~ values so that
the prevailing one will depend on factors affecting the reaction
rates or on some ~termediate reaction. Formation of carbide ~ may get
a head start because a solid-solid reaction proceeds to form carbide
at lower temperatures. The formation of metallic aluminum by a
solid-solid reaction has not1been reported and little or no data
exists for suboxide formation under these conditions .

...

TIIE ALUMINUM INDUSTRY


Another area of thermodynamic interest is the process

in which alumina is reacted with carbon and nitrogen to produce


aluminum nitride and carbon monoxide. In practice, the production
of aluminum nitride falls off rapidly at temperatures above l80o 0 c,
so it has been proposed that aluminum carbide becomes the stable
phase at these higher temperatures. TheL'iF-temperature curves for
the two reactions are shown in Figure 4. Extrapolation of the
two lines indicates the formation of carbide to be favored at still
higher temperatures. Considerable disagreement exists in the
literature concerning the equilibria in the reaction to form
aluminum nitride.

lOO

z
~
w
..."'
0

200

..,
"'w
~

00

w
w

:::

TEMPERATURE,

Fig.4.

DEGREES K.

Free Energy-Temperature Relation for Carbothermic Formation


of Aluminum Nitride

None of these considerations has recognized the formation of


al uminum monocyanide, AlCN, described by Pechiney 15 from the
reaction between aluminum nitride and carbon.
AlN + C

= AlCN

A ordlnl( to P bi n y; t h .formo.tion

:1 I

f th

volatile monocyanide

P.T.STROUP
increases with temperature above 17oo 0 c, and this explains the
apparent decrease in nitride formation. The presence of carbide
was reported in nitride made at elevated temperatures and explained
by the formation and disproportionation of monocyanide. Aluminum
monocyanide vapor can react with aluminum to form carbide and
nitride:
3AlCN

4Al

= Al 4C3

3AlN

It is suggested that some of these reactions involving carbides,


oxycarbides, nitrides , monocyanides and suboxides may occur in
molten steel when aluminum is added.
Another area of thermodynamic interest is the subhalide chemistry
of aluminum~ Willmore 16 and Klem and Voss 17 demonstrated that
aluminum trifluoride reacts with metallic aluminum to form a volatile
monofluoride which disproportionates at a lower temperature to
condense metallic aluminum.

2Al

AlF3

1200c
_ _ _ 3AlF

soo 0 c
100

""

1600

c
Cl
a:
C>
i= 1400

""
""""a:
""c

PER CENT AICI I~ VAPOR


80
60
40
20

CHARGE
50% FERROALLOY
90% AICI + 10% AICl3 -

1200

U)

C>

1000

..;
a:

::>

I-

800

Cl

a:

""
""

IL

;I:

600 .

...

Rt:TURN LINE

I-

Fig,5,

Temperature Cycle for Aluminum Monochloride Distillation


From 50 per cent Ferroalloy at 1500 Centi.grade and
Atmospheric Pressure

THE ALUMINUM INDUSTRY


A similar process utilizing the trichloride was described later
by Gross1 8. Figure 5 shows the process in which the A1Cl 3 vapor is
passed through a bed of impure crushed alloy above 1000c and only
aluminum is extracted and condensed by disproportionation at a
lower temperature around 7oo 0 c according to the reaction
1000C

2Al

3A1Cl

7oo 0 c
The ,reformed AlC1 3 is not condensed but is recirculated through
the reaction furnace.
~
ii:

;;;;;::-.;;;;;;;:::',....;;;;;::r-::::::J

0.40 rM:.c.A::..:X:.c.IMc:.UM"'-=-EF'-'F-'"'IC'-"IEo;.;Nc:.CV;__r---==:::::==:r---<;;;;::----==::::::r:-......

1sooc

1sooc

~
:!'

0.32

"3
"...
0

0.24

=>
0
a.
a:
~ 0.16

":!'=>
"=>_,
<l

...0

0.08

<f>
0

OL_----~~.::r::~::;;::;;;;::;;c::=::t::==,,,,,J,,======::::::::.L_..:::::i::::==::::::~::r==:::L-_j
10
100
50
TOTAL PRESSURE IN MILLIMETERS OF MERCURY

Fig.6.

500

760

Isotherms for the Formation of Aluminium Monochloride from


50 per cent Ferroalloy and Aluminum Trichloride

Russell, Martin and Cochran 19 determined the equilibrium and rate


of monochloride formation at several temperatures. Figure 6 shows
that the efficiency of extraction increases as the temperature is
increased and the pressure is decreased.
Attempts have been made to combine the carbothermic-reduction
and monochloride-extraction steps into a short-cut process:

P.T.STROUP
A1203 + 3C + AlC1 3 _, 3A1Cl + 3CO
l

2Al + AlC1 3
Oross 20 calculated that the above reaction would occur at 17oo 0 c
r t. ut..mospheric pressure, but warned that the reaction products must
In Hh c k-cooled or the aluminum quickly absorbed from the vapor to
1111 v nt r e versal to alumina and carbon. This is the same procedure
t.1 I 1 d f r carbothermic reduction of magnesia using hydrogen for
r hoc; lt
oling, which was not successful on a commercial scale. Free1111 rgy values for this reaction are shown in Figure 7 and confirm
llm 1H' ti calculations.

lOO

_j

;:

.."'

200

Alz03
3AICI

0:

3C

3CO

+ A ICl 3

"0

>-

<!>

0:

100

"'w
w

e:

- 100 ~--~----~---~
1000
lOO
2000
lOOO

TEMPERATURE, DEGREES K.

l'IK . 7.

l'r

111t.11'1HL

!t:norgy-Temperature Relation for Carbothermic,Formation


... of Alumin Monochloride
upplications of thermodynamics in the aluminum industry

hrw I 11u UH
nHidorably as more re liable values have appeared for
1.111 t.111 rmooh1 mion.l properties of cryolite, aluminum fluoride,
r l.11111l1111m cai rhl<l< 1uHI nlum1num. n1tr1d . Additional confirmations are

11111 d1 d for, Liu 111 comDtiUndH tLn<l tLLt 11tion Rh uld now also. b directed
lnw111d I.he 1111111ov1il111L ti1imLHllllHl11, o:xy ' urb1dlH, i111ho ld fl rind othr
I 1 111 1 1 h1fi1w11 111111uo1111d 11 011 re 111111, 11 O!iKt1 I I, lo11 ,

:11,11

THE ALUMINUM INDUSTRY


REFERENCES
1.

K.Grjotheim, Kgl.Norske Videnskab Selskabs Skrifter No.5 (1956).

2.

W.B.Frank and L.M.Foster, Jour.Phys.Chem. 64, 95 (1960).

3.

J.D.Edwards, C.S.Taylor, L.A.Cosgrove and A. S.Russell,

J.E lectrochem.Soc. 100, 508 (1953) .


4.

E.H.Howard, J.Am.Chem.Soc. 76, 2041 (1954).

5.

P.A.Foster, Jr. and W.B.Frank, J.Electrochem.Soc. 107, 998


(1960).

6.

M.Temkin, Acta.Physicochim.U.R.S.S. 20, 411 (1945).

7.

W.B.Frank and L.M.Foster, Jour.Phys.Chem. 61, 1531 (1957).

8.

P.T.Stroup, Trans.Met.Soc.of A.I.M.E. 230, 356 (1964).

9.

L.M.Foster, G.Long and M.S.Hunter, J.Am.Ceram.Soc. 39, 1


( 1956).

10.

N.E.Filonenko, I.V.Lavrov and S.V.Andreeva, Dokl.Akad.Nauk.


S.S.S.R. 124, 155 (1959).

11.

L.Brewer and A.W.Searcy, J.Am.Chem.Soc. 73, 5308 (1951).

12.

C.N.Cochran, J.Am.Chem.Soc. 77, 2190 (1955).

13.

M.A.Miller, L.M.Foster and C.D.Baker, U.S.Patent 2829961


(1958).

14.

L.M.Foster, unpublished results, Alcoa Research Laboratories.

15.

Pechiney Companie, British Patent 842726 (i960); Canadian


Patents 605599 (1960); 623160 (1961); u.s.Patent 3032398 (1962).

16.

C,B.Willmore, U.S.Patent 2184705 (1939).

17 .

W. K-l mm and E.Vos s, Z Anorg.A llge m.Chem. 251, 223 (1943).


:1

P.T.STROUP
18.

P.Gtoss, British Patent 582579 (1944); U.S. Patent 2470306


(1949).

19.

A.S.Russell, K.E.Martin and C. N.Chochran, J.Am.Chem.Soc.


73, 1466 (1951).

20.

P.Gross, Congress International de L'Aluminium, Paris, June


14-19, 1, 167-71 (l954).

'
.;

...
:i

(I

PROBLEMS IN THE NON-FERROUS INDUSTRIES, AND


A NOTE ON ESTIMATION OF HEAT CAPACITY
by
Herbert ff.Kellogg
Henry Krumb School of Mines
Columbia University, New York 27, N.Y.

Abstract

The future role of thermochemistry as applied to the non-ferrous


metal industries is discussed in the light of the wide scope of the
chemistry involved, the relatively small size of the industries and
the competitive problems facing these industries. There is need
for a wide variety of thermochemical data particularly on solution
systems.
An improved meth~d for estimation of the heat capacity of solid
compounds at 298K is presented. The method is analogous to
Latimer's method for estimation of entropy at 298K, and has an
average accuracy of 0.12 cal/C/atom in the compound for normal
valence.compounds.

HERBERT ff.KELLOGG
Sixty-nine of the ninety-two natural elements qualify as nonferrous metals, and this fact has some important consequences for
the subject under discussion. About half of these metals are
recovered and used on an industrial scale. These are: Li, Na, Be,
Mg, Ti, Zr, Hf, V, Nb, Ta, Cr, Mo, W, Mn, Co, Ni, Pt, Cu, Ag, Au, Zn,
Cd, Hg, Al, Sn, Pb, Sb, Bi, U. New additions to the list of
'industrial' non-ferrous metals have been frequent because of the
special requirements of the materials engineer in the space age.
Ferrous metallurgy involves the chemistry of iron and some ten
to fifteen associated elements. Other papers on this program have
illustrated the complexities of ferrous thermochemistry, and the
many gaps in knowledge which stand between us and fuller understanding
of the subject.
The principles of thermochemistry remain unaltered when we shift
our attention from ferrous to non-ferrous metallurgy. Any one nonferrous system is much like ferrous metallurgy in the degree of
complexity which must be unraveled by thermodynamic methods to
achieve understanding. But there are thirty or more industrially
important non-ferrous systems to be understood and only one ferrous
metallurgy. This fact alone explains much of the apparent backwardness of the non-ferrous industries in application of
thermochemistry to their problems. Activity-concentration relations,
for example, are relatively well known for molten pig iron and steel,
and this knowledge contributes greatly to our understanding of iron
and steel making. By comparison we have only rudimentary knowledge
of activity concentration relations in blister copper, black copper,
crude nickel, tin, aluminium, magnesium, lead bullion, etc.
Not only is the scope of non-ferrous thermochemistry many times
the .scope of ferrous thermochemistry, but the industries which must
support research in this area are small compared with the ferrous
industry. All of the non-ferroqs industries, lumped together, don't
equal the iron and st'eel industry in dollar volume of business. ~ The
largest of the non-ferrous industries (aluminium and copper) are
smaller than the ferrous industry by a factor of ten or more.
The enormity of the task and the limited resources of numerous
small industries represent the legitimate reasons for the slower
growth of' thermochemical knowledg e in non "' ferrous m tallurgy. One

....

:!!ill

PROBLEMS IN THE NON-FERROUS INDUSTRIES


can not help but conclude that another, less defensible, reason also
contributes to the slow pace of research in the older non-ferrous
industries (particularly copper, lead, zinc and tin), This is the
attitude, engendered by the long history of these industries, which
tends to belittle the value of modern science for solution of
industrial problems, Such an attitude has been possible in the ruidtwentieth century only because these industries inherited a vast
storehouse of empirical knowledge from previous centuries and have
made effective use of it. Their inherited capital is now mostly
consumed, however, and significant improvements in processes will
require much closer alliance with modern science and engineering
than has beeri the custom in the past.
I am firmly of the opinion that the next decade will show a
marked increase in the level of research activity in the non-ferrous
industries. Thermochemical research will represent a significant
part of this new activity. I base my conclusion on the belief that
the inexorable problems which these industries now face can only be
solved through modern research methods. How to extract copper of
ever greater purity, from ever lower-grade ores, without increase
in price, is not a problem which will be solved by reading De Re
Metallica, or asking father how he did it. If it can be solved at
all, it willbe solved by application of all pertinent aspects of
modern scientific knowledge,
To a process metallurgist in the United States it comes as a
disquieting revelation that most of the significant process developments since World War II must be attributed either to foreign
countries err the chemical industry. Basic-oxygen steelmaking, the
zinc-lead blast furnace, flash-smelting of copper and nickel sulfides,
and electro-winning of nickel from sulfide anodes are all European
or Canadian developments. The fluid-bed originated from the petroleum
industry. The big chemical industries - duPont, Dow, Union Carbide,
Mallinckrodt, Kaiser, Olin-Mathieson, etc. have assumed an active
ro le in production of the newer metals - uranium, titanium, nobium,
t antalium, aluminum, magnesium - and have contributed new process
ideas and a new, research-oriented outlook on metallurgical problems.
Is this the shape of things to come? Will non-ferrous extractive
m ta! urgy, as we know it, slowly become the province of the chemical
nclui-1Lry r f rr r pl'ogr 1:1siv f r ign compan ies? Present trends,

HERBERT H.KELLOGG
extrapolated into the future, seem to answer these questions in the
affirmative . I take the optimistic view, however, that a resurgence
of research and development in the traditional non-ferrous industries
will soon reverse these trends and help to re-establish their
competitive position.
In summary, then, the needs of the non-ferrous industries are
for a more active and modernized approach to metallurgical research
and process development. For this purpose , a vast amount of presently
unknown thermochemical data will be required. Other papers on this
program have already indicated many of our needs for thermochemical
data, but I will summarize those most urgent for the non-ferrous
industries .
In hydrometallurgy, there exists a deplorable lack of data on
aqueous reactions at elevated temperature and pressure. The development engineer , concerned with pressure-leaching and
pressure-precipitation, lacks data on solubilities, phase relations
and equilibrium constants at temperatures in excess of 60c.
Aqueous equilibria involving ions often exhibit abnormally large
val ues of ,t",cp the change in heat capacity, with the result that
1 g K vs 1/T is a highly curved relation (often showing a maximum
r minimum). It follows that estimation of equilibrium constants
nt. 150-200c by calculation from known values of ,t",ff 0 298 and,i",s 0 298 1
on t.h ass umption that ,t",cp equals zero, can lead to entirely
1 1..-c n ous resul t s . The need for extensive new data in this area is
11 rl{t nL.
d-1:10.lt chemistry, which holds so much promise as the basis
l e r - winn ing of metals, for refining of spent atomic fuel, and
rnl' 111 t ornplishi ng many difficult separations of one metal from
n111ll.h1 r, tn n knowledge of solubilities, vapor pressures and activities
I 1ml't, I 11lnrly meager.
It will require many new data and conlih rnhlc d v lopment of solution t~eory before we can adequately
1111 cl let. b ho.vi r i n these.... fascinating : solution systems.
111 l' m-1

r111 1

In tlY romotnllur gy

- still the major means of metal production


1111 d fol' do.ta is so gr eat that it is difficult to select the
111wll. 11 rg 11t it ms. Po.nel I, of this conference, has emphasized the
11111 1 r1n l\ Llv Hy data on soitcl and li Qui d s olutions .
I echo these
11 11 t. l11i1111.11, bnL wc'l ul<l urg t udy f syst m.,.typ s mor clo ely
11 111

PROBLEMS IN THE NON-FERROUS INDUSTRIES


related to non-ferrous metallurgy. The systems M-0, M.::-S, MC, M-N
and the ternary systems M-0-C , M-0-S, etc. need careful study for
many of the non-ferrous metals. We should not overlook the fact
that there are still serious gaps in our knowledge of the thermodynamic properties of pure substances. Thus, we lack data for many
silicates, basic sulfates, sulfides, carbides and other similar pure
compounds.
Clearly, a huge task of data production faces the non-ferrous
metal field . It seems unlikely, that in the lifetime of most of us,
we will be able to solve real industrial problems in non-ferrous
metallurgy by application of measured equilibrium data. No single
real process is so simple that we can hope to describe it entirely
by measured thermodynamic data. Estimation of unknown data now is,
and will continue to be, an essential part of the application of
equilibrium data to process design and control. It follows that
training and experience in data estimation are necessary for the
research and development engineer.
ESTIMATION OF HEAT CAPACITY OF SOLIDS AT 298K
In order that this paper may represent more than a collect i on of
generalities, I offer below a small contribution in the form of an
improved method for estimation of the heat capacity of solid
compounds at 298K.
Thi s past sunmer, while re~writing lecture notes on estimation
met hods, it occurred to me that a scheme, similar to that used by
Lat imer for estimation of entropy, might prove more accurate ' for
estimation of cp 298 than existing methods . You will recall that
Latimer estjmates the entropy of solid compounds by adding together
cont ribut i ons from the catonic and anionic groups in the compound.
He accounts for the variation of entropy contributions with the
valence of t he compound by arbitrarily blaming all this variation
on th~ ~nion and by listing the anionic entropy contributions as a
f unction of the cation charge. His method has proven itself most
us ful, rapid, and r elatively accurate .
Working from th

m a1:1w d h at capacitios of some 550 compounds,

1md J'o l l pwlnl( 1111 Ltl mrn-1L on Uri ly lllJ> ri e.l Ll))pr n.ch , I hav d vis d
ILll ( 11L,1 1111 t.111
11101:1 Cllll'I l'or c,,~Oll wht h l'( ti( mblo11 Ii" Lh J.1Lt.1nu 1 fl
:11 I

HERBERT H.KELLOGG
entropy method and a more detailed form of the Kopp-Neumann rule.
All the values needed to use the method are included in Tables I and
II. I can best summarize the value of the method by stating that the
Kopp-Neumann rule has an average error of 0.5 cal per 0 c per atom
in the compound when tested against a representative list of known
heat capacity values; my method reproduces the data with an average
error of .12 cal per 0 c per atom, when tested against normal valence
compounds, and .23 cal per 0 c per atom when tested against nonvalence intermetallics, carbides, silicides, borides, etc. The
increase in accuracy is noteable when estimating CP for a complex
compound such as M2(S0 4) 3, containing 17 atoms. The Kopp-Neumann
accuracy would be 8.5 cal/C/gram mole compared to 2.0 cal/C/gram
mole for the new method.
After compiling the tables, I tested the method by using it to
estimate CP for twelve compounds for which newly measured heat
capacities were reported in the current chemical literature. The
results are shown in Table Ill, and the accuracy of these estimates
(.1 cal/C/atom) is even a little better than I found for the data
on which the tables are based. This experience has given me some
confidence in the value of the method for estimating unknown data.
I am anxious for you to try the method and I would welcome your
suggestions for improvement. Following are a few typical examples
which illustrate the rules to be followed in using the data of TableH
I and II.
1) Estimate CP 298 for Al 2(S0 4) 3: In Table I find Al
valence compounds. In Table II find

so4=

3.5 for

= 17.9 in compounds .with tri-valent cation.

The measured value ..-is 61. 8

'

2) Estimate CP 298 for Ca 3N2: In Table I find Ca


compounds. In Table Ill find
N3 - = 5.3 in compounds with divalent cation.

The measured val ue is 27.0

5.6 in valenc

PROBIEMS IN THE NON-FERROUS INDUS'IRIES

TABLE I
Atomic Contributions to CP 298
pt

6.3

Hf 6.3

Rb

6.3

Hg 6.3

Rh 6.2

s [5. o]

Ag 6.1

Al 3.5 (5.6)
As 5.1
Au 6.3

1. 5b

[6. 8]

In 5.6
Ir 6.3
K 6.2
La 6.5
Lanthanides 6.3

Sb 6.1

Li 4.9

Sr 6.1

Ca 5.6 (6.0)

Mg 4. 4 (5. 7)

Ta 6.3

Cd 6.0
Cl [6. O]

Mn 6.0
Mo 6.1

Te ?
Th 6.7

Co 6.0

N 4.5 (4.5)

Ti 5.7

Cr 6.0

Na 6.0
Nb 6.2
Ni 6.0

Tl 6.1

u 6.9

[4. o]

6.1
Ga 4.8

3.6

Ge 5.2

Pd 6.1

1. 3 (2. 6)

Ba

6.5

Be 1.8 (2.8)

Bi 6.5
Br [6. 6]
C

[2.0] (2.5) 8

Cs 6.4
Cu 6.0
F [5 . 3]
Fe

Pb 6.5

Sc 4.8

Se [6. 5]
Si 3.0(4.8)
Sn 6.1

v 6.0
6.5
5.6

Zn 5.4 (6.0)
Zr 6.1

Use these ( ) values for intermetallics, carbides, silicides,


borides and non-valence nitrides. Use these [ ] values to estimate
values not given in Table II.
a. For salt-like carbides, such as Cac 2, Al 4C3, use C
b.
For H in acids, acid salts and hydrides
.
~

Abe tu l Vo. 111 s :

Nll 4+ 13.7; H20 9.8 in hydrates.


:10:1

= 4.0

HERBERT H.KELLOGG
TABLE . I I
Anion Contributions to Cp 298
Cation Charge
+1

+2

+3

+4

+5

+6

F.

5.5

5. 3

5.1

5.0

5.1

(5. 4)

c1

6.1

6.0

5.8

5.7

Br

6. 6

6.6

6.6

6.3

6.8

6.8

(7 . 0)

(6 . 8)

o=
s=

4.2

4.6

4.1

3.8

7.0

5.7

5.5

5.0

se=
Te=
N3

(7 . 5)

(6 . 5)
(8 . 8)

(6.)

(3. 8)

5.3

4.0

4.0

ow

(7. 7)

7. 1

SW

(12 . 6)

SeH.

( 11. 2)

co3=

14.5

13 . 8

N0 3so =
3
Cl03.

16.5

15.0

(16.7)

(16.3)

18.1

( 17)

Bro 3

(18. 9)

(18)

I03so 4=
Clo 4
crot
Cr204 =

(19)

BH -

4
CN-

(11)

3.6

4.1

(4 . 7)

(14.)

18. 2'

18.9

(5.7)

17. 9

(17.4)

19.2
22.2

...

25 . 7

'

15.5
(9)

(9)

(8)

These values ( ) are based on scant evidence. For other aniom1,


such as N0 2-. c10 2, use at omic . values listed in Table I.
:104

..

PROBLEMS IN THE NON-FERROUS INDUSTRIES


3) Estimate CP 298 for Mg 2Si0 4: For complex oxides (silicates,
titanates, ferrites, borates, aluminates, phosphates, molybdates,
tungstates, etc.) estimate CP as the sum of CP for the constituent
oxides. Thus we write 2Mgosio 2, and estimate MgO and Sio 2
separately:
Mg+ o = 4.4 + 4.6 = 9.0 =

cP

for MgO

Si + 2(0) = 3.0 + 2(3.8) = 10.6 = CP for Si0 2


2x9.0 + 10.6 = 28.6 = CP for Mg 2Si0 4
The measured value is 28.2
4) For hydrates add 9.8 for each molecule of hydration water to
CP for the anhydrous compound. Thus for Al 2(S0 4) 3 6H 2o we have

The measured value is 117.8


5) Estimate cp 298 for Fe 4N: This is a nitride for which simple
valence rules do not apply. Therefore we use the nitrogen value
from Table I
4(6.1) + 4.5 = 28.9 = CP for Fe 4N
The measured value is 29.3
6) Estimate CP 298 for Mg 3Sb 2: In Table I find Mg= 5.7 for
intermetallics and Sb= 6.1 for all compounds. Thus
3(5.7) + 2(6.1)
Th

= 29.3

= CP for Mg 3Sb 2

measured value is 29.8

:10

HERBERT H.KEILOGG
TABLE III
Estimation of CP 298 for
Measured*

Compound

Soli~

Compounds
Estimated

6/atom

11. 53

11. 3

. 08

ZrC

9.13

8.6

. 27

BN

4.69

5.3

. 30

Ga 2Te 3

35.99

36.0

.oo

V Si
3

21. 84

22.8

. 24

Mo 3 Si

22.22

23.1

. 22

SrC1 2

17.98

18.1

.04

156.9

.08

ZrB 2

K203Al 203 6Sio 2 2H 20

153.6

BeS0 4

20.65

20.0

.11

CaMg(C0 3 ) 2

37 . 65

37.6

.01

NbBl. 963

11. 42

11. 3

.04

UN

11. 31

10. 9

.20

...

'

Based on measurements reported in the current chemical literature


(Chem , Abstr. ~9 (1963)).
300

..

UNFINISHED BUSINESS IN THE THERMODYNAMICS


OF THE BASIC OXYGEN STEELMAKING PROCESSES
by
W.O.Philbrook
Metallurgical Engineering Department
Carnegie Institute of Technology

Abstract
Owing to their speed and inaccessibility to direct observation
during blowing, oxygen steelmaking processes require computerized
control methods based on mathematical models in order to achieve
their potential capability and versatility. For these autogenous
processes a heat balance is essential even to the most elementary
control of tapping temperature through proper proportioning of the
charge. Heat contents of hot metal and liquid slags are not known
reliably and new de terminations are needed. The scope of other
thermochemical and activity data for slag-metal reactions needs to
be extended to cover a broader range of temperature and composition.

W.O.PHILBRUOK
In its essential physical chemistry, oxygen steelmaking is not
intrinsically different from the older methods of steel refining in
basic-lined furnaces. The primary objective of all of these
processes is the selective oxidation from liquid iron of carbon,
silicon and other solute elements down to the required low levels
while at the same time minimizing the loss of iron and other valuable
elements. Basic, rather than siliceous, furnace refractories are
now used almost universally, both for compatibility with the basic
slags that are necessary to remove phosphorus and sulfur and for
their superior ability to withstand high operating temperatures.
Some interesting applications of thermodynamics to steel refining
and deoxidation have already been presented at this symposium by
Dr.Fitterer and Dr.Langenberg and his coauthors.
The significant feature of oxygen steelmaking for the present
discussion is that it combines some of the characteristics of older
processes in a way that accentuates the desirability - indeed , the
necessity - fo r a much more comprehensive background of thermochemical
data than now exists. Furthermore, this comes at a time when our
ability to make use of this information is greatly enhanced by the
availability of high-speed computers and by ~mproved analytical
methods and sensing devices that are being developed.
One important attribute of oxygen steelmaking is its speed , which
compresses the time scale by about an order of magnitude in comparison with its natural competitor, the basic open-hearth process.
The open-hearth furnace of 15 years ago took 8 to 10 hours or more
to produce a heat of several hundred tons of steel, depending on
furnace size. The refining period lasted a couple of hours, and the
furnace was readily accessible for observation and periodic sampling
for chemical analysis and temperature to follow the progress of the
heat. Physical chemistry was ' a nice plaything for the metallurgists,
but a good melter usually did pretty well without it .
In contrast, the bJ sic oxyge~ furnace processes a heat of lOO ~ tO:
300 tons in 45 minutes to an hour, of which only 20 to 25 minutes is
the actual blowing time wherein all of the melting and refining are
accomplished. The equivalent of the refining period in the open
hearth has been compressed to less than 15 minutes and the oxygen .
converter is virtually inacessible without halting the blow t o
permit turning down the vessel. It mi ght seem logic~l to t hink t hat,

:rnn

...

BASIC OXYGEN STEELMAKING PROCESSES


since the process has been speeded up so tremendously, one could
afford to interrupt it for a few minutes to check temperature and
analysis and obtain closer control. However, this runs counter to
human nature and the psychology of production men. A delay of 10
minutes becomes an intolerable 20 per cent of the potentially
achievable heat time. Instead, control methods to match the speed
of the process are sought.
Another cogent consideration is that the oxygen steelmaking
process is autogeneous. In the absence of an independently
adjustable fuel or energy supply, it is crucial that the scrap
portion of the charge and other thermal demands be properly balanced
against the heat availability, as determined by the temperature and
composition of the hot metal. Modern steelmaking standards call for
control of tapping temperatures with little more than 10 C degrees
spread about a specified value in the vicinity of 1600c where the
accuracy of the measurement itself is no better than 5 C degrees.
Without going into detailed explanations it can be said that, if
brought under suitable control, the basic oxygen steelmaking process
has economic and metallurgical attributes that give it a strong
competitive position in the production of the grades of steel that
account for the major share of the tonnage. A great deal of effort
is being expended to perfect mathematical models to simulate the
process in order to predict and control its performance with the aid
of computers, so that its full potential may be realized. The
const~uction of such models(l) involves many complex problems of
reac~ion kinetics, fluid dynamics, sensors for continuous monitoring
of process variables, etc. These problems are probably more severe
than the thermodynamic ones, but are outside the scope of this
conference. There is obviously a need for complete information about
equilibrium limitations of the process and for thermo~hemical data
having a high degree of internal precision, if not absolute accuracy,
to place oxygen steelinaking under the level of control that is
desired for a dependable, versatile, general-purpose process.
These needs are not solely restricted to the basic oxygen process.
Cheap oxygen, computers, and other technological advances are
available as well to the older steelmaking processes, and competitive
pr ssur has force d changes throughout the industry. Physical
h miHtry huH fino.lly come of age in the steel business and is being

:inn

W.O.PHILBROOK
called upon to do a man's job. The present situation is almost like
expecting a skilled mechanic to build an automobile engine with hand
tools. Even though the need for and ability to use physical chemistry
has suddenly jumped by an order of magnitude, the purchase or rental
of computers is not by itself going to produce the capital stock of
process data that still needs to be developed.
THERMODYNAMIC NEEDS
Most physical chemists prefer to work with nice, tractable
stoichiometric substances or binary solutions. The relatively small
number who have chosen to study the complex, multicomponent systems
that the metallurgist is forced to cope with, have properly concentrated on the problems that were most important at the time. It is
to their credit that the amount of information that has been developed
on steelrnaking systems is quite impressive as may be judged from
recent compilations. <23 > The emphasis thus far has been on
essentially isothermal studies of activity coefficients and equilibrium
constants of importance to terminal slag-metal reactions and
deoxidation. The available experimental data have been extended to
the utmost through the assumption of quasi-regular solution behavior
and other approximations. Not only do we need information on
additional systems, but we need also the enthalpy and heat capacity
data for extending calculations over the temperature span of
approximately 1600 to 2000K (2420 to 3140F) which is of active
interest in iron and steelmaking.
ENTHALPIES AND HEAT CAPACITIES
Because of the importance that heat balances have assumed in
oxygen steelmaking it is conv~nient to subdivide the discussion of
incomplete thermodynamic information. Enthalpy and heat capacity
data needed for heat balances and temperature coefficients will be
considered first, followed by ~~ee-energy and activity values for _
isothermal equilibritniJ calculations. The theoretical and experimental
interrelations among these quantities are recognized but are not of
primary concern in this treatment.
Metal Phases
The largest single contribution to the heat balance (298K
reference temperature) of the basic oxygen process is t h h at
content of the hot metal, wh ich a countA or ""ough y lml f t h
<l70

BASIC OXYGEN STEELMAKING PROCESSES


input to a standard, top-blown converter. The accuracy with wh ich
this important bit of information can be estimated leaves much to
be desired.
The only direct calorimetric measurements of the heat content of
molten pig iron that the writer has found were reported by
S.Umino<4,5) in 1926 and 1927. His results on two samples were in
good agreement and have been widely used; for example, in Fig.19-2 of
Ref.2. However, there is reason to question their accuracy on
several grounds.
To begin with, the reference state for Umino's results is ill
defined. In his procedure the hot specimens were dropped into a
water calorimeter at 20 to 25c. Hence, his reported enthalpy
values are relative to a chilled iron at room temperature rather
than to the elements. The carbon was probably present as Fe 3c in
the final state. Neglecting the low concentrations of phosphorus
and sulfur, other elements presumably were in solid solution in
iron. For silicon, at least, the heat of solution at 298K is
significant, the Korber and Qelsen<7) data being the basis for the
reported( 3) heat of formation of 'Fe 3si: It is as good a guess as
any that the reference state at calorimeter temperature for Umino's
enthalpy values was approximately: Fe 3c + 'Fe 3Si' + remaining
elements to account for the reported composition of the specimens.
The second question concerns the numerical accuracy of the
reported enthalpies. With correction for carbon content, Umino's
later< 6 ) (1935) values for the heat content of iron are in good
agreement with the presently accepted values< 8 ), but his 1926
results< 4) for low-carbon iron were about 4-5 per cent higher in the
range from 1300 to 1570C. One may suspect, without being certain,
that his contemporaneous work on cast iron suffered from the same
sources of error. Furthermore, 'umino's 1926 value< 4) for the
specific heat of liquid iron, 0.222, was 18 per cent higher than
the presently accepted one of 0.188 for pure iron. His specific
heat of 0.215 for molten pig iron was probably too high in absolute
value but not as much as would be expected compared with his value
f or low-carbon iron, since the dominant solutes have higher specific
heat than iron. It is difficult to assess the reliability of
Umino's data for cast-iron compositions.

~71

W.O.PHILBROOK
400

~'Y-

375

./'.3V

/,

350

Y/
.' :/-I
_,/~
11V,~ /
'2P

.v

325

~ 300

'">

275

"

>-

~ 250

I
I
I
I

'-'

3 225

I
I

:i:

200

..

175

PIG IRON
o------o Experimental, Umino, (1926)

- {l)Colculoted relative to elements


ot 298 K.

- {2}Colc ulo led relative to ~


"c hilled iron"

I/

150
1000

1100

1200

1300

1400

1500

1600

1700

TEMPERATURE, C.

Fig.1.

Heat Content of Blast-furnace Iron


4.22%C, 1.48%Si, 0.73%Mn, 0.22% Cu

An alternative approach to estimating the heat content of hot


metal is to begin with reported enthalpies of elements<B,3) and
heats of solution(2, 3 ) for binary alloys with iron at 1600C and to
extrapolate to lower temperatures by assuming that the Kopp-Neumann
rule of additivity of heat capacities is valid in the absence of
experimental data. Such a calculation has been made for an iron of
the same composition as one of Umino's samples, and the results are
compared with his experimental points (slightly corrected for
temperature scale and reference temperature) in Fig.1. Two calculated curves are shown, one relative to the free elements at 298K
and one for the 'chilled iron' reference state described two paragraphs above. The phenomenally close agreement between the calculat d
line for 'chilled ir'<ln' and the' extrapolation of Umino' s curve is
regarded as sheer coincidence rather than a confirmation of either
the calculated or the experimental values.
The calculated enthalpy vs temperature relation for hot metal haH
to be used with caution because it depends too heavily on avproxi- .
mations and assumptions with onlr meager experimental information n
:17'

..

BASIC OXYGEN S'IBELMAKING PROCESSES


partial molar enthalpies of solutes in binary iron alloys. As a
matter of fact, the only heats of solution that have been directly
measured in systems of interest to steelmaking are the old calorimetric values of Korber and Oelsen (7) for Fe-Si alloys. Turkdogan,
Grieveson and Beisler< 9> have recently proposed different limiting
values of ~Hsi and ~ItFe for dilute solutions of Si in Fe and of Fe
in Si, respectively, than were deduced by Chipman and
associates< 100203 > from the Korber and Oelsen<7) data. For reasons
that need not be detailed here, there has been feverish action during
the past year or two to revise the thermodynamics of the Fe-Si
system, and activity diagrams have heen proliferating more rapidly
than plans for steel mills in under-developed countries. The range
of disagreement among the rival camps is rapidly narrowing, and the
Fe-Si system should soon be the best understood of the liquid
binaries of iron and be deserving of inclusion in the next edition
of Hultgren et al< 11).
For other iron binaries the partial molar enthalpies are on an
even more tenuous basis. Rist and Chipman< 12 2 3 ) stretched the
available data to derive activities in the Fe-C system as a function
of temperature and to deduce the heat of solution of graphite, on
the assumption that the latter is independent of temperature together
with other approximatio~s. The heats of mixing that are reported <3 )
for other liquid iron binaries were apparently derived from activity
data at one temperature and the assumption of ideal entropy of
mixing. Hultgren, Orr, Anderson and Kelley(ll) did not consider
the temperature coefficients reliable enough to warrant tabulation
of heats and entropies of mixing for any of the liquid iron binaries
in their book. There is no basis yet for computing heats of mixing
in ternary or multicomponent solutions. This situation can be
expected to improve in the next few years with continuing research
on interaction coefficients and their temperature as well as
composition dependence.
Summing up, we do have some information on heats of solutions of
elements in liquid iron at about 160o 0 c. These values have been
derived from a minimum of experimental data and their accuracy is
not wel l established, but they are better than the information
available on heat s of mixing at 2981<.. In the absence of knowledge
aboul!_ partial molar heat capacities, the Kopp-Neumann rule is the
on ly o.vail o.bl bu.sis for stimating the heat capacity for molten

W.O.PHILBROOK
alloys. This seems like a reasonable approximation for solutes for
which the heat capacity in the liquid state can be used but is less
palatable for graphite. At their normal low concentration levels,
phosphorus and sulfur are best ignored in heat balance culculations.
The specific heat derived by the additive rule for the liquid pig
iron of Fig.1 was 0.200 compared with 0.215 reported by Umino. For
heat balance purposes, the writer prefers the calcuJation of the
heat content of the hot metal relative to elements at 298K from
high-temperature data by the methods outlined, instead of extrapolating Umino's experimental values of uncertain accuracy and
reference stdte. With the limitations of present thermodynamic data
there is a possible error of several per cent in this important entry
in the heat balance.
The heat content and heat capacity of the liquid metal at various
stages of processing to steel are obviously obtainable by the same
type of calculation, but the more closely the composition approaches
that of pure liquid iron, the less is the probable error.
Slag Phase
Slags are complex ionic melts of eight to a dozen or more componentR.
For heat balance purposes it is not necessary to understand their stru
ture or the reactions by which they are formed. It must be possible,
however, to compute the net enthalpy change for forming a melt of final
composition and temperature from components in a defined initial state.
Even disregarding considerations of accuracy, a formally correct
calculation of this kind is not now possible for steelmaking slags.
Heats of reaction and fusion are the first problem. It is customary
and convenient to define the components of slags as oxides, sulfur
compounds appropriate to the system, and fluorspar; phosphorus is
best handled as a phosphate to avoid complications of gaseous SP,ecies.
The heats of formation of these components from elements are well
known <3 13 > at 298K Q.nd in most' cases at steelmaking temperature's
as well . However, aside from molten wiistite and fluorspar, the major
components thus defined are solids at 1600C, and their heats of
fusion and reaction to form the melt are not known with any degree of
accuracy if at all. While it is interesting to know t hat nearly
ideal mixing occurs< 14 > between homologous molten silicates having.
cations of like valence and similar size, t his infor mation has limit d
usefulness when the
. heat s of formation a11d f usion of tho individ ual

BASIC OXYGEN STEELMAKING PROCESSES


silicates
silicates
of mixing
fusion of

are unavailable. Little is known about the mixing of


with aluminates, phosphates, etc. Presumably the heats
are small in comparison with the heats of formation and
such melts .

With the present paucity of data, the calculation scheme that


seems to involve the least uncertainty is to assume the formation
at 298K of mineral species that would be expected to constitute
the crystalline slag. To do this accurately would require multicomponent phase diagrams to establish the primary mineral species
and their solid solubility limits in the crystalline slag, together
with the heats of formation of these equilibrium phases from the
oxide components 'at 298K. In practice, the best that can be done
is to use such heats of formation as are available for the more
likely compounds as judged from phase diagrams< 15 23 ) of simpler
pr ototype systems, ignoring heats of solid solution and formation
of more complex compounds for which information is lacking.
There is no sound basis for computing the enthalpy increment
r equired to transform a commixture of oxides and their compounds
at 298K to a molten slag at steelmaking temperatures . Enthalpy
data up to moderately high temperatures in the solid state are
available for a number of the primary compounds of simple slag
systems, but heats of fusion and heat capacities in the liquid state
are very scarce . K.K.Kelley(B) lists6H(fusion) and C (!)for less
than a dozen high-melting compounds that are even remofe!y related
to steelmaking slags. Data for these compounds, together with
calcium borates, sodium ferrite and sodium silicate , which might
have structural similarity to slags, are presented in Table I.
(Heats of fusion, mostly estimated, for a few more such substances
have been tabulated< 3 .i 3 ,l6,l7) . ) For the compounds listed in
Table ~. the entropy of fusion per gram formula weight divided by
t he number of atoms in the formula, ranges from 1.4 to 2.6 e . u.
per g. atom. There is no consistent trend with type of compound or
melting point, so any generalization on the entropy or heat of fusion
is liable to low precision . Similarly, although the heat capacity
of' the liquids averages 8.40 cal/(g.atom) (deg.K), the range is from
7.1 to 9.5 wi th no obvious trend . Hence, the limited amount of data
f or stoichi ometr i c compounds is not very helpful for predicting
. HT - H298 for molten slags .

:w

W.O.PHILBROOK
TABLE I
Heat and Entropy of Fusion and HeatCapacity
of Molten Substances of Possible Relevance
to Steelmaking Slags

Tabulated Values< 8 >


Compound

6u

Tm(oK)

CP(t)

Calculated per g. atom


No.of
atoms

6sm/at.

Cp(.e,)/at.

ca 2Fe 2o5

36, 110

1750

74.20

2.29

8.25

CaFe 2o4

25,870

1510

54.90

2.45

7.84

Ca 2P2o7

24,100

1626

96.80

11

1.35

8.80

CaTiSi0 5

29,590

1670

66.80

2.22

8.35

CaF 2

7, 100

1691

23.90

1. 40

7. 97

Fe.9470
FeS

7,490
7,730

1650

16. 30

1. 95

2.33

8.36

1468

17.00

2.63

8.50

Fe 2Si0 4

22,030

1490

57.50

2.11

8.21

FeTi0 3

21, 670

1640

47. 60

2.64

9.52

MnS

6,240

1803

16.00

1. 73

8.00

V204

27, 210

1818

51. 00

2. 50

8.50

Ca3B206
Ca 2B2o5

35,490

1760

94.00

.. 11

1. 83

8.55

24,090

1585

68.20

1. 69

7.58

CaB 2o4

17,670

1435

1. 76

8.81

CaB 4o7

27,060

1260

91. 70
106 . 30

7
12

1. 79

8.86

NaFe0 2

11, 760

1620

37.30

1. 81

9.32

Na 2Si0 3

12,470

) 361

42.'80

1. 53

7. 14

Average

8 ..40

...

BASIC OXYGEN STEEUfAKING PROCESSES


Direct calorimetric measurements of the heat content of basic
oxidizing slags are limited both in scope and utility. Umino<lS)
reported data for three open-hearth slag samples, one up to 155o 0 c
and two to 15oo 0 c; this work was the basis for Fig.19-1 of Basic
Open Hearth Steelmaking< 2 ) Naeser0 9) gave six sets of values up to
1400c representing eight basic open-hearth slag samples that were
similar to Umino's but a little higher in FeO and MnO and lower in
MgO content. Determined by dropping calorimeter, all of these heat
contents were relative to a poorly defined reference state of
quenched slag at about 20c instead of specified crystalline compounds. There seem to be no better data available. Recent Soviet
investigations on the thermochemical and other properties of slags,
which are reported in the abstract literature beginning about 1960,
have thus far been confined to temperature levels below 1300C and
have dealt mostly with slags of interest to lead and zinc smelting
but not directly relevant to steel refining.

I
50 0

BASIC OXIDIZ ING SLAGS

- - Calculated for Umino No. 4


.,.... - _,.. E xperimental, Umino No. 4

o-- ---o Experimental, Neeser No. 3,4

.-;

Expected trend of extrapo lation

~ 40 0

/I

30

"'>

."'

1-

"'
1-

.I

20 0

'

I-

;:\
:i:

1/

p '

"'

.' i

10 0

_.,..Y
/

0./
0
200

400

600

800

1000 1200 1400 1600 1800

TEMPERATURE , C.

Fig.2.

Heat Content of Basic Open Hearth Slags

In Fig.2 are reproduced Umino's (lS) data for his sample 4, which
was carried to the highest temperature, and Naesers< 19 ) composite
resul~s for his samples 3 and 4, which had about the same 18% Si0
2

W.O.PHILBROOK
content as Umino's slags. The analyses reported for these slags,
weight per cent, were as follows :
p

Umino, No.4
Naeser, No.3
Naeser, No.4

18.20 13.45 2.40 5.00 43.63 9.14 6.97 0.33 0.54


18.2 18.0
7.2 35.7 4.9 13.9
17.8

18.3

3.1

41 . 9 5.4 13.7

The work of these two experimenters was in excellent agreement up to


1100-1200c. At 1300-1400c Naeser' s results were systematically
higher than Umino's, but there were too few runs to establish
whether this was a result of slag composition and melting range or
of experimental technique. The most disconcerting feature of these
data is that Umino's results showed the specific heat of the slag to
increase continuously with temperature even up to 1550C, which was
reportedly well into the fully liquid region. Extrapolation of the
heat content curve with continuously rising slope leads to values at
temperatures of 1600c and above that are probably too high. It
seems more likely that the specific heat should finally level off
and the extrapolation should follow a course approximately like the
dotted line in Fig.2. (This~ is justified by very limited data on
blast-furnace slags above 1500C.) The slope of the extrapolated
heat content curve above 1600c corresponds to a specific heat of
about O. 32 cal/(gram) (C) or 8. 4 cal/(g.at.) (C) which is the
average from Table II .
The full line drawn on Fig.2 for temperatures up to 1100c was
calculated for the composition of Umino's slag No.4 by the KoppNeumann rule using Kelleys ~ 8 ) selected values for the constituent
oxides. It is seen that within the accuracy of the data, the KoppNeumann rule predicts very well the heat content of these basic
slags up to 1000c. Lack of data for the molten oxides prevents
any attempt to appl.J the rule. 'above the solidus temperature. Using
Kelleys< 8 > selected values of HT-H 298 for compounds and oxides,
respectively, the additive rule holds within 2 per cent at 800K,
4 per cent at 1200K and 5 per cent at 1600K for the following
compounds similar to the primary phases of some slag systems:
2 CaOFe 2o 3, 3 cao si02,_ 2 Ca0 Si0 2, CaOSi0 2 and
2 CaOAl 203si0 2. On the other hand, Esin and collabor ator s <2o. 21 )

..

BASIC OXYGEN STEELMAKING PROCESSES


have cited small deviations from additivity of enthalpies as evidence
of interactions in some binary and ternary melts of the Na 20-Pb0-Si0 2
system. Although it is often used and in many instances is the only
means of estimation available, the Kopp-Neumann rule can hardly be
expected to apply precisely to silicate and phosphate slags.
Incredible as it may seem, it is still necessary to use devious
and uncertain methods for what should be a routine calculation of
the contribution of slag formation to the heat balance of a steelmaking process. Not even a reasonable estimate of acc~acy is
possible. Yet an error of 10 per cent in the heat content of molten
slag estimated from Fig.2 is equivalent to about 1 per cent of scrap
in the charge for a top-blown oxygen process. There is need for
some applied research combining the use of high-temperature and
solution calorimetry to measure the total enthalpies of real slags
relative to simple oxides at 298K for the range of compositions and
temperature of importance to steelmaking. Properly planned and
interpreted, such work would have theoretical value as well as
practical utility.
EQUILIBRIUM DATA
Knowledge of the equilibrium conditions for steelmaking reactions
is needed to define the thermodynamic driving force available to
sustain high production rates and the limits to which refining can
be .carried. Refining consists of selective oxidation reactions that
may be represented by the generalized equation:

where Mstand for any oxidizable element and 0 for oxygen dissolved.
in the liquid iron. Exceptions are carbon, with its gaseous oxides
that escape from the system, and sulfur, which enters the slag
predominantly in sulfide form.
From a purely thermodynamic viewpoint, the data needed comprise
activities or activity coefficients of elements in the liquid metal
and of oxides in the slag, together with equilibrium constants for
the correspond i ng reactions. Just as the components of molten slags
ur " ommonly d fin d. in terms of oxide species , so also are pure
Jd o UHULlly ~~ lo t1 d uH Htnndard stat s for o.ctivities of slag
:1'{ I

W.O.PHILBROOK
constituents. Ionic models have had only limited success for thermodynamic applications thus far. The standard free energies of oxides
from elements at steelmaking temperatures are in general well known.
This provides the basis for a thermodynamic treatment that is formally
correct and utilitarian but is independent of any particular model of
slag structure and gives little insight into reaction mechanisms.
The underlying principles and the considerable body of avai : able
equilibrium data have been capably summarized(2,3).
Multicomponent solutions with liquid iron as the solvent have
been studied rather extensively, and activities and interaction
parameters for solute elements have been measured for both dilute
and carbon-saturated alloys. The data up to 1963 have been compiled
by Elliott and colleagues< 2 3 > subject to a systematic
correction<2 2023 > dependent on molecular weight. As noted in connection with heats of solution, most of the tabulated activities
are based on experiments in a restricted temperature range and their
variation with temperature has been assumed to follow regular
solution, behavior. More recent work in several quarters is directed
toward obtaining molal enthalpies and entropies, so continuing
improvement of the activity data for liquid steel is anticipated.
The slag phase is inherently more complex and less well understood than the metal. For the oxidizing slags of steel refining
the experimental problem is complicated by the reactivity of liquid
iron oxide toward all available crucible materials. Hence, the
activity data are by no means as complete as the elegant treatment
that has been worked out< 24 > for slags of the CaO-MgO-Si0 2-Al 2o3
system under reducing conditions. Laboratory studies of slag-metal
equllibria are usually restricted to comparatively simple melts, and
various empirical correlations have been contrived in attempts to
extend the results to mor.e complex real slags. To be sure, the
equilibrium relations for s.ulfur and phosphorus elimination are more
involved than for the metallic impurities, but one needs only to
scan the 14 pages devot~d to sulfllr and phosphorus distribution in
Reference 3 to recognize that satisfactory, generalized thermodynamic
models for slags have not yet been achieved.
It might be argued that a need for more comprehensive thermodynamic
data is not the bottleneck in the construction of process models at
the present moment. Dynamic models require a knowl dgo of kinetics,

...

:1110

BASIC OXYGEN STEELMAKING PROCESSES


and the rates of steel refining reactions are controlled primarily
by transport steps rather than chemical ones. Hence the rates are
strongly dependent on fluid mechanics, reaction paths and the
particular interface at which a given reaction occurs. In the real
process the apparent departure from equilibrium varies with the
individual reaction, the stage of the heat , lance manipulation , and
other features that are more physical than chemical in nature and
are outside the scope of this conference. Such an argument is
pointless because thermodynamic limits are prerequisite to
quantitative description and understanding of kinetics.
CONCLUSION
The speed and autogenous mode of operation of oxygen steelmaking
call for control methods based on mathematical models that simulate
the processes. Although a great deal is known about the physical
chemistry of steelmaking, the quantitative basis for such models is
far from complete. Also, the information that is available has been
extended to the utmost by theoretical assumptions or empirical
correlat i ons in efforts to supply estimates where needed data are
lacking. There is a risk that by repeated use an estimated value
may acquire the aura of experimental respectability.
Most of the research effort prior to the advent of oxygen steelmaking tended to be concentrated on deoxidation practice and on
conditions near the end of the refining stage of open-hearth and
electric-furnace heats. The temperature level was usually restricted
to the normal tapping-temperature range in the vicinity of 1soo 0 c.
Such information is fairly adequate for simple, over-all thermochemical models based on initial and final states of heats blown
consistently to the same low-carbon end point. Charge proportioning
to achieve temperature control is the primary objective of such
models . Statistical methods can be used if necessary as a supplement
to thermodynamics to obtain and update prediction equations for
residual phosphorus, sulfur or manganese contents at the full-blown
end point .
More elegant models are needed,
att ain the f ull potential of basic
ste.e ls and terminat i ng the blow at
nr b ing developed for monitoring
'.11l1

however, and are being sought to


oxygen processes for making alloy
any desired carbon level. Methods
t he t emperature and the carbon

W.O.PHILBROOK
and oxygen balances throughout most of the heat. As these are perfected there will be increasing demand for thermodynamic activity
values over a wider range of slag and metal compositions and temperatures than are covered by present data. Research to broaden the
scope of fundamental information on steelmaking systems should be
fostered now in anticipation of the need.
The most serious and glaring deficiency for both current and
future needs is the absence of reliable heat content data for molten
pig iron and slags. A precise heat balance is the keystone of even
the most rudimentary mathematical model of oxygen steelmaking, Yet
it is not possible to make with confidence a heat balance for this
or any other steel refining or iron smelting process that involves
molten charge or products. Filling this gap in the thermochemical
data should be the first order of business, and it will probably be
accomplished most quickly by utilitarian calorimetric work on real
or synthetic slags and metals covering composition ranges of
industrial importance.
After supplying the immediate need for some numbers for practical
heat balances, a systematic calorimetric study of heats of mixing in
binary and multicomponent solutions in liquid iron and in molten
oxide systems would contribute immensely to the objectives of extending the accuracy and range of usefulness of activity data for
steelmaking. Such work would also be important basic research in
high-temperature physical chemistry.

'

:111 2

BASIC OXYGEN STEELMAKING PROCESSES


REFERENCES
1.

W.O.Philbrook, 'The Problem of Mathematical Simulation of


Basic Oxygen Steelmaking' in John F.Elliott and T.R.Meadowcroft,
Eds: Steelmaking: The Chipman Conference 223-236, MIT Press;
(1965).

2.

Physical Chemistry of Steelma.king Committee, G.Derge, Ed.


Basic Open Hearth Steelmaking, 3rd Ed. A.I.M.E. (1964).

3.

John F.Elliott, Molly Gleiser and V.Ramakrishna, Thermochemistry


for Steelmaking, 2, Addison-Wesley: Reading, Mass. (1963); cf.
ibid, 1 (1960).

4.

S.Um'ino, Sci.Reports Tohoku Imper.Univ. 15, 597-617 (1926).

5.

S.Umino, idem 16, 775-798 (1927).

6.

S.Umino, idem 23, 665 (1935).

7.

F.K6rber and W.Oelsen, Mitt.K.-Wilhelm-Inst.Eisenforsch.


18, 109-130 (1936).

8.

K.K.Kelley, 'Contributions to the Data on Theoretical


Metallurgy, XIII', Bur.Mines Bul.584 (1960).

9.

E.T.Turkdogan, P.Grieveson and J.F.Beisler, Trans.Met.Soc.


A.I.M.E. 227, 1258-1264 (1963).

10.

J.Chipman, J.C.Fulton, N.Gokcen and G.R.Caskey, Jr., Acta.Met.


2, 439-450 (1954).

11.

R.Hultgren, R.L : orr, P.D.Anderson and K.K.Kelley, Selected


Values of Thermodynamic Properties of Metals and Alloys,
John Wiley: New York (1963).

12.

A.Rist and J.Chipma.n, 'Activity of Carbon in Liquid Iron-Carbon


Solutions' in J.F.Elliott, Ed: The Physical Chemistry of
Steelmaking 3-12, M.I . T.Technology Press and John Wiley:
New York (1958) .

:m:i

W.O.PHILBROOK
13.

O.Kubaschewski and E.Ll.Evans, Metallurgical Thermochemistry,


3rd Ed., John Wiley: New York (1958).

14.

F.D.Richardson, 'Activities in Ternary Silicate Melts' in The


Physical Chemistry of Steelmaking 68-75, M.I.T. Technology Press
and John Wiley: New York (1958).

15.

E.M.Levin, Carl R.Robbins and H.F.McMurdie, Phase Diagrams


for Ceramists, American Ceramic Society: Columbus, Ohio
(1964).

16.

M.F.Koehler, R.Barany and K.K.Kelly, 'Heats and Free Energies


of Formation of Ferrites and Aluminates of Calcium, Magnesium,
Sodium and Lithium', Bur.Mines R.I.5711. (1961).

17.

K.K.Kelly, 'Heats and Free Energies of Formation of Anhydrous


Silicates', Bur.Mines R.I.5901. (1962).

18.

S.Umino, Sci.Reports Tohoku Imp.Univ. 17, 985 (1928).

19.

G.Naeser, Mitt.K.-Wilhelm-Inst.Eisenforsch 12, 7-12 (1930).

20.

S.G.Bratchikov and A.E.Esin, Isvest.Vysshikh Ucheb.Zavedenii


Tsetnaya Met. 3, No.4, 39,44 (1960).

21.

G.A.Roporishchev, A.E.Esin and S.G.Bratchikov, idem 4, No.3,


37-43 (1961); 5, No.1, 50-58 (1962).

22.

H.Schenck, M.G.Frohberg and E.Steinmetz, Arch.eisenhuttenw.


31, 671 (1960).

23.

C.H.P.Lupis and J.F.Elliott, Trans.Met.Soc.A.I.M.E. 233, 257-8


(1965).

24.

R.H.Rein and Jollb Chipman,' Trans.Met.Soc.A.I.M.E. 233, 415-425


(1965).

:JIM

VACUUM DEGASSING OF LIQUID STEEL


by

John E.Fogarty
Superintendent
South Division
Republic Steel Corporation
The subject of this panel session is the 'Needs for Future
Applications of Thermodynamics' . In illustration of such needs as
related to the vacuum degassing of liquid steel, some of the problem
areas associated with Republic Steel Corporation's induction stirred
ladle degassing facility in Canton, Ohio will be reviewed.
Consider that in order to achieve a reasonable profit, operating
management must produce a high quality product at the lowest practi cal cost. From a technical viewpoint, therefore, the best possible
systems analysis must be made.
Probably, the first thermodynamic aspect the degassing engineer
encounters is defined by the K

Pco
% c x fc x % o x f 0

equation.

Immediately problems arise.


First, there is apparently no generally suitable system for
measuring the oxygen content of a 'low-oxygen' steel sample . Even
the American Society for Testing Materials make no recommendation
concerning an analytical system. Furthermore, the National Bureau
of Standards has no standard samples in the low oxygen (below 30 ppm)
range. This inadequacy is reflected in the results of a recent oxygen
analysis 'round-robin' of properly selected steel samples (1) . This
'round-robin' was conducted by Republic Steel with leading industrial,
research and government laboratories. The results are shown in
Figure 1. The lack of uniformity of results is obvious. Although
reproducibility within each laboratbry facility was good, the general
state-of-the-art requires more sophistication.
Could not a properly applied thermodynamic effort result in a
fas t er , more dependable, simple system for oxygen analysis that
could be used not only in the laboratory but also used as a control
tool for t he steelmaki ng process ? Cons ider the extremely broad
pot ntJul of tiu h o. syst m.

:m

JOHN E.FOGARTY

E 3 50

)//

~
~

/V

3 25

30

.."'

27

...>

12~

....
z

~ 2 50

....

....
:z:

PIG IRON

o-----o Experimental, Umino (19261

2 25 t---t--~,-1 '

00

,)
1---17'--t

(l)Colculoted relative to elemenh

at 298K.

- - (21Colculoled relative to "chilled iron"

75/

5,l6'-oo--110'--o--12....
oo-1~30_0_1---+40_0_1~50-o-1s~o-o-11""'=
!l!f
TEMPERATURE, c.

Fig.1
Referring again to the equilibrium equations , it is found that
when they are used in an attempt to estimate the oxygen content of
a degassed heat that a tremendous difference exists between the
approximate oxygen c.o ntent of the steel and the theoretical
estimation. For example, a 1% C low alloy steel that is degassed
in the fully 'open' condition, at a pressure less than 100 microns
(Hg), normally ranges over 100 times higher in oxygen content than
the thermodynamic potential.

Reference to the tables of interaction coefficients of the alloy~


ing elements hardly account for all of this. Incidentally, the
degassing engineer finds such data incomplete for the one element
(manganese) that is pr_esent in V!rtually all commercial steels. He .
attempts to establish this point (for his system) himself with
several 'open' 100 ton carbon steel heats making the manganese
addition before degassing on some heats and after degassing (but
still under vacuum) on other heats.
in the f or.m of low oxyge n electrolytic manganese

:mo

VACUUM DEGASSING OF LIQUID STEEL


The influence of molybdenum, chromium and nickel are also
determined by making AISI 4040, 4140 and 4340 heats. The results
are interesting when compared with the available published interaction coefficients. Nickel it is found has a beneficial effect on
the C - 0 reaction approximately double published values. Molybdenum exhibits a significantly more detrimental effect than expected.
Chromium is found to have a range reasonably close to published
values, and manganese is found to have no apparent influence. These
effects not only manifest themselves in the oxygen analyses of the
heats, such as they are, but also in internal steel cleanliness.
The engineer asks, 'Does exposure to a vacuum in effect modify
the interaction coefficients of some of the elements in an open
steel bath?' Or, are these observations in fact due to some other
aspect of the C - 0 reaction, such as effecting the minimum refractory pore size for bubble neucleation? If one considers the vast
potential of vacuum carbon deoxidation in the manufacture of steel,
answers to such questions are necessary. Furthermore, it is likely
that the effects of the individual alloying elements on the efficiency
of the C - 0 reaction will have a significant influence on the
chemical composition of the steels of tomorrow. Again, should not
dependable, specific data be developed which can guide alloy
development?
Next, the engineer considers the role of the refractory portion
of his system. Little has beP.n published concerning the effect of
ladle refractories in an evacuated system. A paper by Elliott and
Meadowcraft does offer a real insite into the problem. The engineer
supplements this work by experimenting with bricks of commercial
refractories.
Five pound controlled chemistry melts of AISI
1040 containing none of the strong deoxidizers
(Al, Zr, Si, etc.) were made in a zirconia-lined
induction furnace. Sections of commercial bricks
were immersed in the bath and the chamber evacuated
to 100.. The valve to evacuating system was then
closed. The rate of pressure rise to 10,000. was
determined and subtracted from the rate of pressure
rise to 10,000 on a similar trial which did not

:1117

JOHN E.FOGARTY
include the immersion of the brick sample. This
difference provided a measure of the rate of
deterioration (stability) of the brick .
220

.200

l \

.180

.160
(

'\

._\

Gas Evo111t1oti l4o


(.Briclc-Netal
Reaction)
.120
CuFt/Sq.Ft/Min . lOO

IYG

~s

2D

30

.032

.09

.030

.OS

~'

70

!
6

.<Yf

.06

(J

...,..
,,,,-

.
~

'!!

....
"!l

.05

.022. f<

.04

""" " ~
0

....
.,..0

. 021\

. 020

.03

.018

.()12

.016

. 01

...

:ll
8
u
::::

<ii

(~

X Gas ivol.Rate
Carbon IDss
Gl 51 Ucon "hop"
1<

.10

"Ev'o lltion Rate

;)

.020

. 03~

. 026

.060

.o4o

. 11

.0'2@

'

.080

.036

80

Fig.2
This experiment was conducted on samples of several commercial
Al 2o3-Si0 2 bricks of varying composition. Gas evolution rates were
calc ulated from the pressure rise rates . The results showing
increased stability for bricks with higher Al 2o3 content, are
plotted in Figure 2.
The difference in the rates of evolution for the various brickl?
w uld be related to the effect of carbon in the bath on the products
o , for example, the Si0 2 = Sio(g) + FeO reaction, as follows:
2

+ ( iO (g) + FeO) 2CO_(g) + Si + Fe

or, Lill 1J'f'1 I, of Ol rh n ln Lh1 hnL h 1111'

:11111

Lly

ou Liu H1 2 r fro. tory

VACUUM DEGASSING OF LIQUID STEEL


(Si0 2 + 2C = 200(g) +Si). Assuming either of these reactions
essentially describes the overall reaction, the bath C should be
depleted to a greater degree with less stable refractories because
they would provide more oxygen for the carbon of the bath to unite
with and leave the system as CO gas. Similarly, the silicon content
of the bath should increase.
To substantiate this, a second series of similar experiments was
conducted but, in this case, the brick sections were immersed for 15
minutes while the bath was exposed at 100. As expected, the results
did reveal both an increased silicon pick-up and a greater carbon
loss with the higher Si0 2 brick sections. The results are also
plotted in Figure 2.
It is obvious, therefore, that as the per cent of alumina in the
brick increases, the stability of the Al 2o3-Si0 2 brick increases.
And, conversely, as the per cent of Si0 2 increases, the rate of
deterioration of the brick increases as measured by the gas evolution
rate, the carbon drop rate, and the rate of silicon pick-up in the
bath. Concurrently, of course, the rate of oxygen transfer to the
steel is also greater with low stability brick.
Therefore, the refractory with the greatest chemical stability
(consistent with its cost, pl;lysical properties, etc.) must be utilized
if optimum efficiency is to be achieved by carbon deoxidation.
However, the best refractories for the system in terms of chemical
stability are unsatisfactory in terms of their physical stability
and cost. Could not programs be initiated whereby thermodynamic data
could be developed which could then be applied to the manufacture of
both chemically and physically stable refractories produced within
the framework of acceptable economic limits?
In conclusion, the application of available technology by the
steel industry is occurring at an ever increasing rate. It must in
order for that industry to remain competitive in the marketplace.
~t is therefore essential that more research be directed at existing
practical problems in order that further advances can be made quickly
and economically. The old trial and error method is too expensive
and time consuming .

:11111

JOHN E.FOGARTY
(1) The samples were centerless ground 1/4" rd
rod sections of AISI 52100 from the product
of a 25" x 27" x 12,000 pound ingot. A 500
pound length was sectioned and analyzed for
oxygen at 12" intervals to insure uniformity.
Included samples (3" long) were sent out for
the 'round-robin' analyses.

'

390

THE PROBLEMS IN CONTINUOUS CASTING


by

T.B.Winkler
Assistant to Director of Research,
Bethlehem Steel Co.

Continuous casting of .steel is already a commercially successful


process, especially in those instances where the tonnage output of
the steelmaking complex is relatively small and where the grade,
quality and application of product is relatively narrow. The
progress that has been made to reach even this degree of commercial
acceptance represents a tribute to those who have been working
diligently in the field. However, if continuous casting is to gain
broader commercial acceptance, particularly for tonnage steel
production, three aims must first be achieved:
1. Develop
portion
quality
casting

continuous casting practice to cover a substantial


of the broad range of steel grades and product
now made to established commercial standards by ingot
and primary-mill rolling.

2. Increase the continuous casting rate to .the point where it is


more nearly compatible with existing steelma.king and
steel-finishing rates in tonnage operations.
3. Improve the efficiency of section-size change to match
commercial requirements of size and gage, particularly in
flat-rolled product.
The attractive potential of continuous casting for both lower cost
and improved quality of product is an incentive for the speedy
achievement of these aims. To this end, many steel producers have
begun programs either of experimental investigation or of commercial
application .
.In keeping with the general theme of the conference, we shall
dtscuss these aims with reference to the application of fundamental

T.B.WINKLER
principles of thermodynamics, heat transfer, thermochemistry and
other di sciplines.
CHANGE OF SECTION SIZE
The last of the three aims, that of improving the efficiency of
section-size change, will be discussed only briefly, becaus~
solution of the problem is one of engineering design, hence not a
proper subject for extensive discussion at this conference.
Commercial requirements of size and gage for flat-rolled products
made on finishing mills of current design dictate that semifinished
slabs be made in a wide variety of sizes. This variety is readily
achieved with existing primary-mill facilities. If continuous
casting is to fit easily into an existing plant as a replacement
for a primary mill, then it must be capable of producing the same
variety of slab sizes. However, the current designs of continuous
casting machines do not provide for section-size changes to be made
easily or inexpensively .
At least three ways of compensating for this deficiency can be
advanced:
1. Finishing mills might be modified so that fewer slab sizes
would be needed for the same mix of final sheet or plate sizes.
2. The casting machine could handle a selected few sizes
representing a substantial portion of the tonnage and the primary
mi ll could continue to roll the balance.
3. The number of slab sizes could be reduced if re-examination
f the finishing mill operation' showed that not as many slab sizes

wr

actually needed.

A11y on or all of these or other modifications may apply in


I 11<1Iv ldunl cases . Nevertheless, the fact remains that some variety

...

u!' ttlub 6i7.e is needed in plants making flat-rolled pr oduct, and


onL!nuous casting's well-established competitor, the primary mill,
111 < u.1mbl
of easily producing this variety. The same condition
1 11. lt1LA for billets or r ounds , a1though the pr oblem is not as severe'
111 1 ' Lh r o these cases because the variety of sizes is not as
l 1u' ic1.
A oordingJy, if oont1nnou6 .custing is to be wid ly accepted

PROBLEMS IN CONTINUOUS CASTING


for flat-rolled product, a method must be developed that permits
quick change of section size and the solution of this important
problem lies in the field of engineering design.
CASTING RATE
In current practice, the casting rate for billets allows a
throughput of about 15 to 20 tons per hour per strand. For slabs,
the throughput ranges from about 40 tons per hour per strand for a
6 x 24" slab to about 125 tons per hour per strand for a 10 x 60"
slab. It is generally accepted that the casting operation for a
single ladleful of steel can proceed satisfactorily for periods up
to one hour. Caating periods shorter than one hour are desirable;
longer ones are possible through high ladle preheat, ladle covers,
ladle burners or any other procedure or equipment devised to minimize
heat loss from the ladle of steel.
On the basis of both the throughput rate and the duration of the
casting operation, it is clear that continuous casting is better
suited to steelmaking furnaces of small size, such as electric arc
furnaces and small basic oxygen furnaces. On the other hand, for
large-size furnaces a multiplicity of casting strands is needed. As
the number of strands increases, the capital and operating costs
increase and the operation of the facility becomes more complex. If
the casting rate could be increased without sacrifice of quality, the
number of strands could be decreased and continuous casting would
become more attractive to the tonnage producer who operates steelmaking furnaces of large size. An obvious alternative is to
substitute smaller furnaces. This alternative is a distinct possibility where new or replacement steelmaking capacity is needed and
there is evidence that at least one tonnage producer plans this.
Casting speed is controlled by the rate at which the solidifying
, skin grows in thickness as the solidification front progresses from
the surface to the center of the bar. The rate of growth of skin
thickness is in turn controlled by the rate of heat extraction from
the solidifying bar.
In actual practice, casting speed is adjusted so that, when the
bar m r..g A from t he bottom of the water-cooled mold and enters the
ttprny lmmh ~ r. , Liu HOlidified skin is thick enough and strong enough

:10:1

T.B.WINKLER
to withstand both the ferrostatic pressure of the liquid steel
within the bar and the tensile stress imposed by the stripping action
of the oscillating mold. If the casting speed is too fast and the
skin too thin and weak, the skin will rupture, a breakout will occur
and the casting operation will be forced to stop. In some instances,
a breakout results in damage to the machine.
If casting speed is to be increased, the rate of tieat transfer
from the solidifying bar to the water-cooled mold must be increased
to promote quicker formation of a strong skin. During casting, the
surface of the solidifying bar is in contact with the mold for a
distance of only a few inches below the molten metal surface. In
this area , the rates of heat transfer and of solidification are at a
maximum. Below this point, the solidifying bar shrinks away from
the mold and the air gap formed by the shrinkage lowers heat transfer , thus causing the rate of solidification to slow down. Ther efore
to accomplish higher casting speeds, mold design must be modified to
counteract the shrinkage of the solidifying bar so that the sur face
of the bar is in direct contact with the water-cooled mold wall for .
a greater distance , if not for the full length of the mold. In one
case, a mold of unique design tried on a German casting machine
permitted a casting rate nearly double the normal .
In the portion of the casting machine below the mold, casting
speed does not appear to be limited by the rate of heat transfer
from the surface of the bar. Optimum rates of heat transfer appear
to be possible in the spray chamber where heat is extracted by water
sprays. Below the spray chamber , where heat is dissipated mostly
by radiation to the surroundings, heat transfer rates are less than.
optimum but still adequate. In both sections the limiting heat .
transfer rate is nt that from the surface, but rather that from the
interior to the surface of the solidifying bar.
Further application of the principles of heat transfer, supported
by carefuH y designecf experiments and measurements both '0n casting
machines and in laboratory simulations, can lead to significant
improvements in casting speed, hence to broader commercial
application of continuous casting.

::194

PROBIEMS IN CONrINUOUS CAST ING


STEEL GRADES AND QUALITY
To broaden commercial application of continuous casting, l h Ir
process must be _expanded to cover a wide range of grades , quul!ty
and applications. At present, the great majority of steel b ing
cast commercially is silicon-killed carbon and low-alloy gr ad s 1111cl
austenitic stainless grades. It has been reported that ferr itic
stainless grades are also being cast but that the phase change wh ich
the ferritic grades undergo during solidification requires more
precise control of the casting operation.
ALUMINIUM-KILLED STEEL
Considerable difficulty is encountered when attempts are made to
cast aluminium-killed fine-grain carbon or alloy steels either for
billet and bar products or for low-carbon deep-drawing sheet. When
the residual aluminum concentration of the steel exceeds about
0.010 to 0.015%, two problems arise :
1. Aluminum oxide deposits on the bore of the tundish nozzle and
gradually restricts the stream. This restriction can lead to
complete blockage of the nozzle, thereby causing the casting
operation to be discontinued.
2. Gobs of aluminum oxide accumulate on the surface of the
molten metal in the mold and some .become trapped in the solidifying
skin of the bar, thus causing danger of breakout and/or poor
surface.
To avoid these problems other deoxidizing and grain refining
alloys, such as vanadium and titanium, have been used, but their
performance as grain refiners is not as reliable as aluminum ~ The
cast i ng of aluminum-killed steels is an area in which the fundamental
pr i nciples of thermochemistry might profitably be applied to provide
a s olution to problems both of nozzle restriction and of the
ent r apment of aluminum oxide in the solidifying skin.
RI MMI NG STEEL
For sheet product application, the most commonly used grades are
rbon rimming or mechani cally capped steels. Repeated efforts
V hrn II mndf L continuously cast rimmi ng st eel in slab form but

fow ~ 1

lu

:Hlfi

T.B.WINKLER
only with irregular success. These erratic results can be attributed
to our lack of adequate control of deoxidation that is needed to
develop and maintain, during solidification, the precise rate of gas
evolution required to produce both a sound skin on the bar and deeply
seated blow holes. In m.v opinion adequate control of deoxidation
would be possible if we had in-plant methods for rapid, accurate
analysis of both carbon and oxygen concentration of the steel bath
at or near tap of the steelmaking furnace. Until such methods
become available, the steelmaker will be forced to depend on approxi
mate indications and control will continue to be erratic. Much work
is being done to develop rapid carbon and oxygen analysis. Here
again application of fundamental principles of thermochemistry may
produce fruitful results.
VACUUM DEOXIDATION
Another approach to the production of slabs for application to
low-carbon sheet is to substitute vacuum-deoxidized steel for
rimming steel. Several companies have dnnounced their intention
to explore this route and it is reported that one or two are already
doing some work. The work that has already been done on the vacuum
deoxidation of ingot steel should be applicable to continuous
casting. It now seems clear that none of the commercially available
methods for vacuum treatment is capable of lowering the oxygen
concentration in low-carbon steel to a sufficiently low level to
fully kill the steel. If no metallic deoxidizers are added after
vacuum deoxidation, the ingots solidify as semikilled steel. The
assumption is that a steel fully killed by vacuum deoxidation would
produce superior quality from the standpoint both of internal
cleanliness and of freedo~ from surface defects. Application of
fundamental principles may lead to more successful vacuum deoxidation
on the commercial ~cale and thus make steel of superior cleanliness
available for continuous casting.

'

INTERNAL SOUNDNESS
Continuously cast billets are well suited to the production of
merchant bar products such as reinforcing bar, flats, small angles
and small channels . . These .products make no serious demands on
either internal soundness or on surface quality. Cast billets are
also used for bar product of more demanding application such as

....

:urn

PROBLEMS IN-CONTINUOUS CASTING

cold-heading stock, sucker rods, grinding balls, and seamless tubing.


It is not yet clear whether continuously cast material will meet the
even more critical requirements of applications such as automotive
and aircraft forgings, stock for cold extrusion, bearing materials,
gear stock and the like.
Requirements for these critical applications are:
1. Nearly all these applications call for _fine grain. The
problems associated with this have already been mentioned.
2. These products also call for minimum segregation. In this
continuously cast material has a distinct advantage since
the rapid rate of solidification practically eliminates chemical
segregation.

~espect,

3. Most of these critical applications require good cleanliness.


Here again, continuously cast material has an advantage. If the
liquid steel being delivered to the casting-machine mold is relatively
clean, the minor amount of non-metallic material that does exist will
be well distributed in the quickly solidified bar.
4. Critical applications require freedom from center unsoundness.
Continuously cast material has a distinct tendency tCJtVard the forma. t ion of center porosity as well as other internal voids and subsurface cracks. These defects are usually controlled by suitable
adjustment of casting speed and cooling rate in the secondary cooling
section. Accumulation of fundamental data on the effect of casting
r ate and cooling rate on internal soundness will be useful for
exercising good control in commercial operation. Also electromechanical stirring or vibration of the solidifying bar might provide
a method for improving center soundness.
SURFACE

Surface quality is also a requisite for critical applications .


Since the mold of a continuous casting machine is smooth, the
potential for producing superior surface on continpously cast material
is high. Re f inement in mechanical oscillation of the mold may be
n eded. Pr otection of exposed metal in the pouring stream and the
Lundish Should aid in reducing t he amount of slag trapped in the
3

T.B.WINKLER
solidifying skin of the bar. Proper design of tilndishes and nozzles
or other delivery devices should be helpful for controlling splash
of metal entering the mold. Adequate adjustment of casting speed
and of cooling rates in the spray chamber and oelow can result in
crackfree product. Any or all of these variables present
opportunities for the application of fundamental principles.
To sum up, the use of fundamental principles of thermodynamics
and related disciplines will be valuable for studies aimed at
increasing the continuous casting rate and at expanding the process
to include a broader range of steel grades, quality, and applications.

'

PANEL IV

DISCUSSION
Dr.Elliott Thank you Ted. The members of the panel have been very
well behaved. They've completed their task in a reasonable amount
of time, and now we will have a short period of discussion among
them. First, let's give the panel a chance. Who wants to raise the
first question? Dr.Stroup.
Dr.Stroup - I don't believe Dr . Kellogg ever told us why he had two
values for aluminium.
Professor Kellogg - There was quite a bit I forgot to t ell, really.
There is a group of 'compounds' for which ordinary rules of valence
don't apply. For example, it would be hard to decide what the
valence of boron is in zirconium boride, or the valence of silicon
in vanadium silicide. I found that I had to treat intermetallic
compounds, borides, carbides, silicides and some of the nitrides as
what I call non-valence compounds, and to assign different values
for the atomic contribution to heat capacity than for valence compounds. For aluminium I found a value of 3. 5 in valence compounds,
like Al 2o3 and 5. 6 for intermetallic compounds of aluminium, aluminium
boride, etc. In the same way, the value for boron is 1.3 in valence
compounds like B2o3, but 2.6 in borides. In all, my table I shows
special values for non-valence compounds of nine of the low atomicweight elements. For elements of high atomic weight I could find
no difference for the atomic contribution to heat capacity between
the non-valence type of compound and the valence type. Now what
all this means I will not attempt to explain, but these special
values are necessary for empirical correlation of the data.
Dr. Elliott Are there other questions ' on this point among the panel?
I have two quick ones. The first is, can we use these heat capacity
estimates for aqueous solutions and secondly, what is the prospect
of making a correction to these heat capacities somewhat in terms
of, say a Debye temperature so that your tables could be used at
higher temperatures, where as you point out yourself, we do need
add i t i onal information?
Profe~sor Kel l ogg - These tables are intended only for solid compoundH nt room t mperature, j ust as Lat i mer' s t ables for entropy

:100

DISCUSSION
estimation apply only for solids. There are other methods for
estimating entropy of ions and dissolved substances in aqueous
solution but they are considerably more complicated. It would
probably be possible to make such estimates of heat capacity in
aqueous solutions, but I have not tried to do so. YQur second
question, regarding the Debye temperature, contains a most
interesting idea. I have not yet progressed far with estimation
methods for heat capacity at elevated temperature, and I will
certainly investigate a Debye temperature correlation.
Dr.Elliott - Thank you.

Other questions now?

Dr. May - Do you attribute the change with temperature of the heat
cape.city for ammonia solution to a change in hydration with
temperature or some other phenomenon?
Professor Kellogg - There are remarkably few data available on the
heat capacity of ions in solution, but what little data we do have
suggests that all charged species in solution have abnormally low
heat capacities compared with uncharged species. so, if you write
a reaction, such as I did, involving two un-ionized species, such
as ammonia and water, going to two ions like the ammonium ion and
hydroxyl ion, you will get a large negative heat ca~city for that
reaction. But if you write a reaction in which the same number of
charges appear on either side of the equation, the abnormally low
heat capacity of the ions tends to cancel each other and you no
longer get this very large change in heat capacity for the reaction.
For example, writing the ammonia-ammonium equilibrium in the form,
aqueous ammonia plus hydrogen ion goes to ammonium ion, the heat
capacity change for that reaction is quite small. I'm not sure this
answers your question, but pethaps you can restate it and I can try
again.
Dr. Elliott - Are there other c<?.vunents from the panel?

...

Professor Philbrook - I'd like to ask Dr.Stroup if he would like to


elaborate on a hint he made of the possible participation of these
oxy-carbides or other sub-oxides of aluminum in steel. I presume
you had reference to de-oxidation.

...
'100

PANEL IV
Dr. Stroup - Yes, it looks like a natural thought. The high t mlH 11
ture chemistry of all these monovalent compounds of aluminium is i n
the range at which you add aluminum to molten steel, and I would
be suspicious that the aluminum combines with the carbon and nitrogen,
and that they interact with each other. I guess steel people
recognize aluminum-nitride and oxide and I would suspect aluminum
carbide can also be present. We are just becoming aware that some
of these reactions could occur in steel.
Dr.Elliott - I have a question for Dr.May. I noticed that you did
not say anything about sulfides in your tabulation or in your discussion. Some of them such as cerium sulfide are quite stable at
high temperatures. You also alluded to the problem associated with
non-stoichiometry of many of these compounds. Do you think that our
efforts in handling these defect lattice structures that we are
encountering here are sufficient, or do we need more attention to be
paid to them?
Dr. May - I would certainly consider the sulfides as reasonable high
temperature materials. But, the point o! reference is to look at
the compounds which have higher melting points, i.e., higher than
tungsten. If we look at these, we are limited roughly to hafnium
carbide, tantalum carbide, the alloy which exists between hafnium
carbide and tantalum carbide, and the alloy which exists between
zirconium carbide and tantalum carbide. These illustrate extreme
high in temperature , therefore, I wish to emphasize these. Obviously
for lower temperature, you would have silicides , sulfides and
phosphides as additional pertinent materials to investigate.
Dr. Elliott - The second question dealt with non-stoichiometry and the
associated problems.
Dr. May - Well , that is difficult to answer because you're asking a
sort of indirect question. One would have to consider the crystal
structure that is present in tantalum carbide and the other refractory
carbides and nitrides. It is the face centered cubic structure
composed of interpenetrating lattices of the metal and the carbon.
When there is a deficiency of what we may call anions (although they
are probably not present as anions) such as carbon or nitrogen,
obv iQ._us ly ther e are vacant s ites in these positions . As I alluded
to befor e , tantalum carbide is very difficult to make as a
401

DISCUSSION
stoichiometric compound. It always seems to lack carbon. At one
time we thought that the usual 3 per cent atomic deficiency in
carbon was perhaps related to an ordering of the carbon vacancies;
thus, we attempted to detect it by x-ray diffraction. Of course,
the scattering power of the super-lattice thus generated would be
about 1 millionth of the strength of that of the main lattice, and
therefore, even though we meticulously used counting techniques
where we thought the lattice lines would be, we found no ordering.
Because of the melting point data (I must reaffirm) we believe
that the composition like Tac is that which would correspond to
compound. However, for some reason, such compounds apparently do
not vaporize congruently: when you heat these compounds to a high
temperature, they seem to lose carbon. It may be due to the
presence of oxygen by which the carbon is removed through an oxidation process yielding C0 2. Thus, it is difficult at present to
be sure if such compounds do actually vaporize incongruently,
Dr.Elliott - can I paraphrase your answer by saying that
non-stoichiometry is a problem?
Dr.May - It's a very great problem.
Dr.Elliott - I mean we have a great deal of work to do even with
one material because of the influence of structure of properties.
Even when we consider just the few that you have highlighted.
I have a question for the other end of the row down there , i.e. ,
all three of those steel people. Vacuum pouring followed by continuous casting is being considered as a way of avoiding many of the
problems associated with the conventional means of deoxidizing
steels by additions of deoxidizing elements. It is possible in the
cyclic use of ladles, etc., that on exposure to the air, the
refractories will absorb appre~ lable quantities of iron oxide. This
iron oxide may be a ~~jor source of oxygen in subsequent use of the
ladle. Has this problem been grappled with?
Dr.Winkler - We've attempted to 1 determine the influence of used
ladles compared with new ladles, and we have found that there is a
greater pickup of non-metallics from used ladles versus very well
dried and very well cleaned newly: relined ladles. Disturbingly
'102

....

PANEL IV

enough, when we examined some of the non-metallics wi th t h mlc 10


probe, we found that many contained substantial quantities f
calcium, in the neighborhood of 30 to 50%. These steels had not
had any calcium used in deoxidation practice, so our obvious conclusion was that the source of the calcium was slag. It could come
from furnace slag and/or from slag adhering to the surface of the
ladle from the previous heat. To this extent at least, we have had
some experience to show that the material on the surface of a used
ladle can be troublesome, especially when one is looking for extreme
cleanliness.
Mr.Fogarty - We also have had a look at this, and we realize that it
is very likely a serious problem. We've tried to make some Judgment
comparing used ladles which have had several heats on them with
brand new ladles. However, the newly lined may not be completely
devoid of moisture to the extent that ladles that had heats in them.
This complicates the observation. It may be a factor, but I really
hav~ not been able to make a quantitative judgment.
Dr.Elliott - Professor Philbrook, what's your reaction?
Professor Philbrook - We don't use many ladles in our operation. I
don't have any authoritative experience that I think will contribute.
Dr. Elliott - Thank you. Do we have other questions from the panel
now, or shall we go on to questions from the audience? I propose
that we start back into and clear out the questions starting with
Dr.Winkler.
Dr.Winkler - I have been asked a question from the audience which
reads, 'Please state your opinion of the advantages and disadvantages
of the following continuous casting processes. First the curvedmold radial machine with no bending, second, the straight-mold
vertical machine and third, the curved-mold machine .with progressive
bending. Please give the advantages and disadvantages in terms of
qual,it.y, casting speed, changes of section size, secondary cooling,
etc. A proper answer to this question would require another
seminar . Since this question does not .pertain to thermodynamics I
woul d suggest that the gentleman who raised the question contact me
after the meet i ng.
"

DISCUSSION
Dr.Elliott Thank you Dr.Winkler. There is a great deal of
technology in this with a number of over-tones with regard to
thermodynamics. Perhaps you could meet with the questionner at the
Electric Furnace Conference tomorrow. Mr.Fogarty, may we have your
questions?
Mr.Fogarty - The first one is from Dr .C.B.Alcock of the Imperial
College and he asks, 'Don't you think the low pressure degassing
would be less misleadir.lg than your term vacuum degassing?' This
really is not my term. Probably you are correc~ but this is a
general term within the industry, and I'm sure it would be rather
difficult to change it at this time.
The second question is from someone at Crucible Steel, 'Did you
notice to pick-up of zirconia through the reaction with the zirconia
crucible in your five pound melts of ASI - 1040, and if so, how
much?' If we picked any up it was less than 0.01% and we really
don't have much confidence in the results because they were so low.
We probably did, but I'm afraid I don't have a quantitative value
to give you. The next question is what vacuum is required to degas
steel? What is the measure used to determine the end of degassing?
Well, I would expect that the lower the pressure, the better.
Normally on a production basis, we say what is the measure used to
determine the end of the degassing, it's when the boil subsides,
i.e., it ultimately goes to a steady state boil, indicating a
probable reaction with the refractory.
The other question is, 'Does degassing give any problems with
the ladle refractories?' And I'd like to turn this around and say
that ladle refractories give us trouble with degassing.
The next question is, 'Would you comment on the relative merits
of the exploding stream degassin~ versus your technique?' Well, by
exploding stream, I a~sume that ' you mean stream degassing. I would
call it ladle-to-ladle stream degassing or ladle-to-mold stream
degassing because of a very limited .number of molds. If you are
dealing with very large ingots where ladle-to-mold degassing is
practical or if you are interested in the flaking problem, and wish
to eliminate hydrogen, then I would say that the stream degassing
system is particularly satisfactory. However, if you have to teem

PANEL IV
numerous ingots ladle-to-ladle stream degassing becomes a necessity.
Here, you have to heat two ladles with your steel necessitating a
much higher tapping temperature. Looking at the economic side of
it, ladle refractory cost is doubled and the furnace refractory
costs are significantly increased.
The next question. is what is your justification in attempting to
double chec.4 .v:alties of interaction coefficients determined in systems
containing' high purity raw materials in refractories by an industrial
process where it is likely that small quantities of extraneous but
influential elements are present? Actually, I don t attempt to
justify this, because I have not attempted to do it. I merely stated
that the effect of nickel, for example, is observed to be twice as
great as could be accounted for by its interaction coefficient. You
will note that I did not give any specific values about over-all
interaction effects nor coefficients because they very likely would
not be of any value to any other system or perhaps the same system
in another plant. They are, nevertheless, real problems and some
general measurements have allowed us to at least cope with our
problems. In addition, you might find it convenient to put all the
blame on the interaction coefficients but I did state in the paper
that there are questions regarding the effect of ladle refractories,
kinetics , etc., including small quantities of extraneous but
influential elements.
Dr.Elliott - Now, there is a question asked jointly of Professors
Philbrook and Kellogg, and I don't know if they're going to sing it
in duo or in tandem or what, but we'll leave it to them.
Professor Kellogg - The question is from Dr.Moles from National
Lead Company and is, 'How do we stand on estimating high temperature
, heat capacities for solids either on extrapolating from room
temperature or making use of very sparse high temperature data?' I
don't think we stand too well. Most of you are probably familiar
with Dr.Kubaschewski's book in which he discusses estimation of heat
capacities at elevated temperature. He gives one method there which
I think has proved fairly successful; and that is taking the
temperature of the first transition which occurs in the substances,
whether it be a melting point or a crystal transformation and
assum,tng that the contribution to the heat capacity at the temperature
is about the same for each of the atoms involved. I think

DISCUSSION
Dr.K.K.Kelley of the U.S. Bureau of Mines recommends a contribution
per atom of 7 calories per gm.atom, and Dr.Kubaschewski recommends
7.25 calories per gm.atom. I have attempted to improve the accuracy
of this method for estimation of heat capacity at the first transit ion temperature by developing tables of empirical values similar
to those I have developed for room temperature. Relatively few
data are available from which to calculate the empirical values,
however, and the uncertainty of the estimation method is rather
large - about O. 5 calorie per 0 c per atom ln the formula ..
Professor Philbrook - I'd make one quick addition. For substances
of the nature of blast furnace slags, for example, it seems that
the heat content in the range where it rises virtually linearly
with temperature, before the u17Nard inflection on incipient fusion,
is fairly close to that calculated from the sum of the contributions
of the oxides making up the slag. This may be good up to about
1000c or thereabouts.
Dr.Elliott - Dr.Stroup, do you have a question?
Dr.Stroup - The question which was submitted is, 'Can you say up to
what temperature aluminum carbide is stable?' It starts getting
pretty restless around 2200 centigrade, if it's held there for a
long time, the aluminum evaporates and leaves behind some beaut'iful
graphite crystals.
Dr.Elliott - Is there a reference on this?
Dr.Stroup - Yes, you will find some reference in my recent lecture
for the AIME on carbo-thermic smelting of aluminum. There's quite
a few references given there.
Dr.Elliott - . Professor C.B.Alcock has a question for Dr.May as
follows: 'There already exists a satisfactory amount of thermodynamic data for the r&fractory metal
' silicides. Have you any
particular reason for eliminating these compounds from consideration
for space application?'
Dr.May - This question was essentially answered before when I gave
the reason for not considering sulfides. The diborides, nitrides and

400

PANEL IV

carbides have the highest known melting points and are , th r J'o ,
the compounds I emphasize.
Dr.Elliott - We have a question for Dr.May from Dr.J.B.Bennett, of
the General Electric Company. 'Please comment on expected
advantages or disadvantages of using complex carbides being
synthesized I guess beforehand as dispersion hardening components.
How, do you expect their stability at elevated temperatures to
compare with thorium oxide in tungsten or in nickel?'
Dr.May - I can give you one example that occurred in our laboratory
a week or two ago. Somebody was trying to roll or draw columbium
carbide and put a sheet of tungsten around it before processing at
high temperatures. When he was finished he found that he had tungsten carbide on the outside and columbium in the middle of his
sample. Therefore, I think, at least this particular system is not
a good one. As for a nickel system, I would expect you would find
a little of the same sort of difficulty.
Dr.Kubaschewski (from the audience) - Please note, Dr.May, that the
systems indicated by the refractory compounds in your Table I are
also the most difficult to investigate quantitatively by experimental
thermochemical methods. In fact, in some cases, we do not know of
any reliable or sufficiently accurate methods for the measurements of
~H or.~ G0 .
Dr.Elliott - Dr.Kubaschewski, this points up one of the areas needing
attention in the next ten years.
Gentlemen, I wish to thank the panel for a very interesting
afte!'noon. If I may speak for the audience, I wish to say we all
have enjoyed the papers and discussion a great deal. There are many
other things we could discuss but time has run out and I would like
to turn the meeting back to Dr.Fitterer for adjournment .
. Dr.Fitterer - Thank you Dr.Elliott. In closing this conference, I
would like to say that just five months ago when we had this idea
of this Conference we had absolutely no idea that it would turn out
as well as it has. In fact, we were quite discouraged, however, as
I mentloned at the beginning of the meeting. We feel that a great
d al of th credit for the success of this Conference can be given to
1107

DISCUSSION
Dr.Kubaschewski. We would like to thank all of you who have come to
the Conference, and thank the participants particularly for their
e~cellent thought-provoking papers.
Now, in closing, I would like
to say one more thing. I would like to thank our committees particularly I would like to thank Miss Penny Brown, David Blewitt and
Carl Cassler and the projectionists and Mellon Institute for the use
of this excellent auditorium. Now, one more item, and this is that
within a year and a half you should see the following sign in your
mail, we may be having another conference. Thank you very much for
coming!
~s

...

'

~Cp

dT

PARTICIPANTS
THERMODYNAMICS CONFERENCE
NOVEMBER 29 DECEMBER 1, 196 4
1.

Dr.C.B.Alcock
University of London
London, England .

11. Dr.Irving Johnson


Argonne National Laboratory
Argonne, Illinois

2.

Dr . J.A.Berger
University of Pittsburgh
Pittsburgh, Pa. 15213

12. Dr . O.Kubaschewski
National Physical Laboratory
Teddington, Middlesex,
England

3.

Dr.John Chipman
Massachusetts Institute of Tech. 13. Mr.A.F. Kolek
cambridge, Mass.
Crucible Steel Co.of America
Pittsburgh, Pa..
Dr.J.F.Elliott
Mass. Inst. of Tech.
14 . Prof . H.H.Kellogg
Cambridge, Mass.
Columbia. University
New York, New York
Dr.G.R.Fitterer
University of Pittsburgh
15. Dr.F. C. Langenberg
Crucible Steel Co . of
Pittsburgh, Pa . 15213
America.
Pittsburgh, Pa.
J.E.Fogarty
Republic Steel Co.
16. Dr.B.Meddings
Canton, Ohio
Sherritt-Gordon Mines
Alberta, Canada
Dr.P.W .Gilles
University of Kansas
Lawrence , Kansas
17. Dr . V.N.Mackiw
Sherritt-Gordon Mines
Dr .R.Hultgren
Alberta, Canada.
University of California
18. Dr.G.W.Mellors
B! rkeley, Calif.
Union Carbide Co.
br. T. R. Ingraham
Cleveland, Ohio
Canadian Mine.s Branch
19. Dr.C . E. May
otta.wa, Canada
NASA
Cleveland, Ohio
D~ . Everett J ohns on
N1lt.i no. l Bureau of standards

4.

5.

6.

7.

8.

9. '

10.

W1 1lll111l l, ou, D.C.


!IOR

PARTICIPANTS
20. Dr.C.W.McCoy

25. Dr. P.T.stroup

Crucible Steel Co. of America


Pittsburgh , Pa.
21. Dr . J.H.Oxley

Battelle Memorial Institute


Columbus, Ohio
22. Prof . W. O.Philbrook

ALCOA
New Kensington, Pa.
26. Dr.S.Senderoff
Union Carbide Co.
Cleveland, Ohio
27. Dr.J.J.Ward

Carnegie Inst . bf Tech .


Pittsburgh, Pa. 15213

Battelle Memorial Inst.


Columbus, Ohio

23. Dr . G.R.St.Pierre

28. Dr.W.E . Wallace

Ohio State U.
Columbus, Ohio

University of Pittsburgh
Pittsburgh, Pa. 15213

24. Dr.F.D.Rossini

29. Dr.T.B.Winkler

Notre Dame University


South Bend, Indiana

Bethlehem Steel Co.


Bethlehem, Pa.

Registrants
1.

Dr . Allen M.Alpern
Corning Glass Works
Corning, New York

6.

R.P. Abendroth
Owens Illinois Glass Co.
Toledo, Ohio

2.

Victor P. Ardito
Allegheny Ludlum Steel Corp.
Brackenridge, Pa .

7.

Lufti Amra
General Electric Co.
Cleveland, Ohio

3.

F. H. Allison
Blaw Knox Company
Pittsburgh , Pa.

8.

J.Berkowitz
Arthur D.Little, Inc.
Cambridge, Mass.

4.

Paul M.Audette
ottawa , Ontario, Canada

9.

Alfred Buchler
Arthur D.Little, Inc.
Cambridge, Mass.

5.

J.F.Anderson
Western Gear Cor p.
Pittsburgh, Pa .

10. P.E . Blac kbur n

Art hur D.L i ttle, Inc.


'o.mbr- id , Mo.sR.
111() .

THERMODYNAMICS CONFERENCE
11. H.L.Bishop
Allegheny Ludlum Steel
Brackenridge, Pa.

28. D. L. Ball
General Motors Corp.
Warron, Michigan

12. J.G.Bassett
U.S. Steel Corp.
Monroeville, Pa.

29. V.C.Boucek
Rim-blox Co.
McMurray, Pa.

13. W.F.Brizes
Westinghouse Astronuclear
Large, Pa.

30. W.H.Burr
U.S.Steel Corp.
Duquesne, Pa.

14. E.L.Bedell
320 Brunswick Rd.
Louisville, Ky.

31. C.R.Bieling
N.J.Zinc Co.
Palmerton, Pa.

15. Dr.J.G.Bennett
General Electric Co.
Cleveland, Ohio

32. Janes Biggers


Penn State U.
University Park, Pa.

23. R.E.Boni
Armco Steel Corp.
Middletown, Ohio

33. C.C.Brown
General Electric Co.
Cincinnati, Ohio

24. Eugene Bicknese


Ford Motor Co.
Dearoorn, Michigan

34. D.H.Baker
Reno Met. Res. Center
Reno, Nevada

25. Thomas M.Barnes


Interlake Iron Corp.
Toledo, Ohio

35. L.R.Bidwell
Aero Space Res. Lab.
Wright-Patterson Air Force Base
Ohio

26. Calvin H.Baloun


Ohio University
Athens, Ohio

36. D.E.Babcock
Republic Steel
Cleveland, Ohio

27. E.F.Bauer
Union Carbide
Niagara Falls, New York

37. D.W.Blewitt
University of Pittsburgh
Pittsburgh, Pa. 15213

lj 11

PARTICIPANTS
38. W.D.Bennett
Falconbridge Nickel Mines Ltd.
Thornhill, Ontario, Canada

48. D.L.Creazzi
Crucible Steel
Pittsburgh, Pa. 15330

39. J.Bindels
Bell Telephone Lab.
Allentown, Pa.

49. Dr.R.S.Cremisio
Universal Cyclops Steel Corp.
Bridgeville, Pa.

40. E.Aukrust
J & L Steel Corp.
Pittsburgh, Pa.

50. G.W.Cunningham
Battelle Memorial Inst.
Columbus, Ohio

41. F. W.Amery
Colarado Fuel & Iron Corp.
Pueblo, Colorado

51. L. Cooper
Heppenstall Co.
Pittsburgh, Pa.

42. Dr.D.Crimmins
U.S.Army Materials Res.
Agency,
Watertown, Mass .

52. Dr.J.Cuthill
National Bureau of Standards
Washington, D. C.
53. R.B.Corbett
Corbett Associates Inc.
Mars, Pa.

43. W.R.Chilcott
U.S.Steel Corp.
Monroeville, Pa.

54. L.P.Connor
University of Pittsburgh
Pittsburgh, Pa. 15213

44. D.Camp
Carnegie Tech
Pittsburgh, Pa. 15213

55. J.A.Clumpner
Olin Matheison Chemical Corp.
New Haven, Conn.

45. C.K.Chu
Westinghouse Electric
Youngwood, Pa.
46. J.M.Cigan
St.Joseph Lead Co.
Monaca, Pa.
47. L.Cogan
Linde Div-u.c.c.
Newark, New J er sey

56. D.C.Cusanelli
U.S. Metals Refining Co.
Carteret, New Jersey
57. Dr.G.N.Campbell
University of California
Los Alamos , New Mexico

THERMODYNAMICS CONFERENCE
58. E.J.Chapin
U.S.Naval Res. Lab.
Washington, D.C.

68. H.M.Davis
Army Res. Off ico
Durham, N.C.

59. H. P.Croft
Copper Range Co.
New York, New York

69. E.W.Dewing
Aluminum Labs., Ltd.
Arvida, Quebec, Canada

60. W,L.Davis
U.S.Steel Corp.
Monroeville, Pa.

70. J.R.Dietz
National Steel Corp.
Weirton, W. Va.

61. G.Derge
Carnegie Inst. of Tech.
Pittsburgh, Pa. 15213

71. T.F.Dudash
Allegheny Ludlum
Brackenridge, Pa.

62. F.J.Dewez
U.S. Steel
Monroeville, Pa.

72. E.A.Dancy
Carnegie Inst. of Tech.
Pittsburgh, Pa. 15213

63. J.B.Donihee
American University
College Park, Md.

73. L. S. Darken
U.S.Steel Corp.
Munroeville, Pa.

64. P.Duby
University of Pennsylvania
Philadelphia, Pa.

74. J.Feinman
u.s.steel
Universal, Pa.

65. J. M.Dugan
Ohio Steel Foundry Co.
Lima, Ohio

75. R.J.Fekete
Ford Motor Co.
Dearborn, Michigan

66. J.H.Downing
Union Carbide
Niagara Falls, N.Y.

76. H.A.Fisch
General Electric Co.
Cleveland. Ohio

67. W.S.Dritt
Union Carbide
Oak Ridge, Tenn.

77. W.B.Frank
ALCOA Research
New Kensington, Pa.

PARTICIPANTS
78. J.W.Frederickson
Dow Chemical Co.
Midland, Michigan

88. B.Grupen
Bell Telephone Lab.
Allentown, Pa.

79. H.V.Fairbanks

89. W.H.Goodnow
J & L Res.

West Virginia University


Morgantown, W.Va.

Pittsburgh, Pa.
90. J.G.Goodwin

80. T.P.Floridis
V.P. I.

University of Pittsburgh
Pittsburgh, Pa. 15213

Blacksburg, Va.
81. P.A. Foster

91. S.Gilby

Ohiq State U.
Columbus, Ohio

ALCOA Res. Labs.


New Kensington, Pa.

92. G.H.Geiger

82. J.F.Foley,

University of Michigan
Madison 5, Wisconsin.

Babcock-Wilcox
Beaver Falls, Pa.

93. J. Godshall

83. Tasuku Fuwa

Westinghouse Electric
Pittsburgh, Pa.

Tohoku University
Sendai, Japan

94. K.Grjotheim
Technical University of
Norway
State College, Pa.

84. J.C.Fulton

Allegheny Ludlum Steel Corp.


Brackenridge, Pa.
85. A.Filippi

95. W. Hogue

University of Pittsburgh
Pittsburgh, Pa. 15213
86. E.A.Gulbransen

United Engr & Foundry Co.


New Kensington, Pa.

...

96. D.Halliday

Westinghouse Res. Labs .


Pittsburgh, Pa.

University of Pittsburgh
Pittsburgh, Pa. 15213

87. J.Gordon
Blaw Knox Co.
Pi tts burgh, Pa .

97. P.J . Hanley

!NCO
uf1' rn , N w York

I I I

"

THERMODYNAMICS CONFERENCE
98.

Dr.J.P.Hansen
University, Alabama

108. E.J.Hohman
J & L Steel
Pittsburgh, Pa.

D.C.Hilty
Union Carbide
Niagara Falls, New York

109. P.Jackson
Ford Motor
Dearborn, Michigan

MTE
99.

100. R.L.Holmer
Union Carbide
Newark, N.J.

110. H.A.Johansen
Westinghouse Res.
Pittsburgh, Pa.

101. J . B.Huvos
Bureau of Mines
Greenbet, Md.

111. E.W.Johnson
Westinghouse Electric
Pittsburgh, Pa.

102. D.J.Hansen
Union Carbide
Niagara Falls, N.Y.

112. L.K.Johnson

103. C.C.Herrick
University of California
Los Alamos, N.M.

113. F.G.Jones

104 . H.P.Hopkins
Carnegie Inst. of Tech.
Pittsburgh, Pa. 15213

114. C.A.Julien

105. E.Hodges

115. J.M.Henderson
American Smelting&Refining
South Plainfield, N.J.

Crucible Steel
Pittsburgh, Pa.

University of Michigan,
Ann Arbor, Michigan
Crucible Stee 1
Pittsburgh, Pa.

Babcock & Wilcox


Beaver Falls, Pa.
106. S.C.Hayes
Armco Steel
Baltimore, Md.

116. P.Grieveson

107. T.S.Howard
Chase Brass and Copper Co.
E1:1olid . Ohio

117. J.Jasper

U.S. Steel
Monroeville, Pa.
Ohio State U.
Columbus, Ohio

11

PARTICIPANTS
118. S. G. Johnson
Standard Steel Co.
Lewistown, Pa.

128. T.M.Krebs
Babcock & Wilcox
Beaver Falls, Pa.

119. E.F.Jordan
General Motors
Bristol, Conn.

129. R.Caplan
Carnegie Inst. of Tech.
Pittsburgh, Pa. 15213

120. T.S.Jones
Allegheny Ludlum Steel
Brackenridge, Pa.

130. W.L.Kerlie
U.S.Steel
Monroeville, Pa.

121. R. Jackson
J & L Steel
Pittsburgh, Pa.

131. 0. J. Kleppa
University of Chicago
Chicago, Illinois

122. R.E.Johnson
Penn state U.
University Park, Pa.

132. G.H.Kessler
National Steel Corp.
Weirton, W. Va.

123. S.M.Kaufman
Ford Motor Co.
Dearborn, Mich.

133. T.F.Kaveney
Union Carbide
Niagara Falls, N.Y.

124. P.J.Koros
J & L Steel
Pittsburgh, Pa.

134. D.J.Knight
Atlas Steel Co.
Welland, Ontario, Canada

125. J.Krawczykiewicz
Westinghouse Electric
Youngwood, Pa.

135. A.ff.King
Arthur D.Little, Inc.
Cambridge, Mass.

126. W.Kernner
Wheeling Steel
Steubenville, Ohie

136. Dr.J.G.Kreible
General Electric Co.
Cleve land, Ohio

127. S. Klinvex
Bureau of Mines
Pittsburgh, Pa.

'

137. R.J.King
University of Pittsburgh
Pittsburgh, Pa. 15213

THERMODYNAMICS CONFERENCE
148. Dr.A.H.Larson
Colorado School of M n OH
Golden, Colo.

138. C. W.Kovach
University of Pittsburgh
Pittsburgh, Pa. 15213
139.

149. H. Lipsey

~.Kubit

Republic Steel Corp.


Cleveland, Ohio

University of Pittsburgh
Pittsburgh, Pa. 15213
140. Dr.W.A.Klemm
Kaiser Aluminum & Chem.
Corp.
Permanente, California

150. Dr.A.Liberman
Zenith Radio Corp.
Chicago, Illinois
151. S.S.Lin
University of Kansas
Lawrence, Kansas

141. W.Klawitter
Carpenter Steel Co.
Reading, Pa.

152. R.Littlewood
Steel Co. of Canada
Hamilton, Ontario, Canada

142. Dr.C.H.Li
Grumman Aircraft
Bethpage, New York

153. W.Mahan
Bureau of Mines
Pittsburgh, Pa.

143. J.W.Linhart

Ohio Steel Foundry


Lima, Ohio

154. J . .Morris
Bureau of Mines
Pittsburgh, Pa.

144. G.Long

Northern Illinois Gas Co.


Aurora, Illinois

155. W.H.McCluskey
Whee ling Stee 1
Mingo Junction, Ohio

145. T.S.Lundy
Union Carbide Nuclear Co.
Oak Ridge, Tenn.

156. R.C.Masse
Allegheny Ludlum
Brackenridge, Pa.

146. H. F.Leonard
Coppers Inc .
. Pittsburgh, Pa.

157. F.E.McBride
General Electric Co.
Cleveland, Ohio

147. D.C.Ludwigson
University of Pittsburgh
~ittsburgh, Pa. 15213

'11 '1

PARTICIPANTS
158. F.X.McCawley
Bureau of Mines
College Park, Md.

168. O. W.Moles
National Lead Co.
South Amboy, N.J.

159. R.T.MacMillan
Bureau of Mines
College Park, Md.

169. O.T.Marzke
U.S. Stee 1 Corp.
Pittsburgh, Pa.

160. J.J.Millen
General Electric Co.
Cleveland, Ohio

170. 0.0.Miller
Met. Engr.
New York, N. Y.

161. W. F. Morse
Columbia Gas System
Columbus, Ohio

171. L.R.Matricardi
Union Carbide Corp.
Niagara Falls, N. Y.

162. G.T.Motock
Met. Engr.
Hamden, Conn.

172. S.Mellgren
U.S.Metals Refining Co.
Carteret, N.J.

163. A.Muan
Penn State University
University Park, Pa.

173. Dr.L.A.McClaine
Arthur D.Little, Inc.
Cambridge, Mass.

164. A.E.Morris
Penn State University
University Park, Pa.

174. R.B.McLellan
Rice University
Houston, Texas

165. L.J.McGeaay
Lafayette College
Easton, Pa.

175. C.K.Mader
Swindell-Dressler Co.
Pittsburgh, Pa.

166. R.E.Musgrove
Climax Molybdenum Co.
New York, N.Y.
....

176. J.N.Marshall
Babcock & Wilcox
Beaver Falls, Pa.

'

167. R. S. Mateer

177. F.J.Millero
Carnegie Inst. of Tech.
Pittsburgh, Pa. 15213

University of Kentucky
Lexington, Ky.

'11 IJ

178. J.Megusar

188 . IJL ll, N111 hl

University of Pittsburgh
Pittsburgh, Pa. 15213

1111111

La n l lt 11 1,1 11 1 111
Chi cago , 1 1 l.

179. F.Muller
Inland Steel Co.
Hammond, Ind.

189. 0.0.0matete

180. H.D.Myers

190 . Dr .B.F. Oliver


U.S. Steel Corp.
Pittsburgh, Pa.

Princeton University
Princeton, N. J.

Dow Chemical Co.


Midland, Mich.
181. J. Morral

191. K.H. Osthagen


International Harvester Co.
San Diego, Calif.

Ohio State University


Columbus , Ohio
182. C.B.Moesta
University of Pittsburgh
Pittsburgh, Pa . 15213

192. E.F.Pearson
Colorado Fuel & Iron
Pueblo, Colo .

183 . W.McKewan

193. D.Puotinen
R.C.A.
Harrison, N. J.

U.S. Steel Corp.


Monroeville, Pa.
184 . B. M.Mihara
Republic Steel Corp.
Canton, Ohio

194 . R.E.Paine

185. Dr.Morse

195 . M. F. Parkman
Aerojet General
San Ramon, Calif.

Consultant
Whittier, Calif .

A.F.M.L.
Yellow Springs , Ohio
186. A.E. Nehrenberg
Simonds Saw & Steel Co.
Lockport , N. Y.

196. K.E. Pinnow

187. F. G.Norris
Wheeling Steel Co .
Steubenville, Ohio

197. R.E . Pocock


Bailey Meter Co .
Wicklif fe , Ohio

Crucible Stee 1
Pittsburgh, Pa .

119

PARTICIPANTS
208. N.Parlee
Stanford University
Stanford, Calif.

198. W.F.Portez
J & L

Pittsburgh, Pa.
199. J.Preston
Allegheny Ludlum
Brackenridge, Pa.

209. J.L.Peterson
La Salle Steel Co.
Chicago, Ill.

200. A.Prokopovitsh
U.S. Dept. of Interior
Washington, D.C.

210. W.F . Pikus


General Electric Co.
Cleveland, Ohio

201. W.A . Peifer


Allegheny Ludlum
Brackenridge, Pa.

211 . D.I.Reichmayr
Consultant
Pittsburgh, Pa.

202. Dr.A.Pehler
Westinghouse Electric
Pittsburgh, Pa.

212. D.L.Raymer
Latrobe Steel Co
Latrobe, Pa.

203. R.D.Pehlke
University of Michigan
Ann Arbor, Mich.

213. F.E.Rizzo
Aerospace Res. Lab.
Wright Patterson Air Fore
Base, Ohio

204 . G.W.Perbix
Republic Steel Corp.
Cleveland, Ohio

214. J.N.Reding
Dow Metal Products
Midland, Mich.

205. R.M . Powers


Assoc. Chemist
60 Shore Rd.
Cold Spring Harbor, N.Y.

215. E.A.Reid
Columbia Gas Co .
Columbus, Ohio.

206. D. S.Parsons
Sylvania Electric
Towanda, Pa.

216. W.P.Roe
American Smelting & R f
South Plainf i e ld, N.J.

207. C.Prasky
U. S. Bureau of Mi nes
Minn apol s. Minn.

217. R.J . Reyn i k


f T h .
Dr x 1 lnR
Ph 1 htd l Vh in, Pn.

I ' 0 -

11 111

THERMODYNAMICS CONFERENCE
218. E.Rosen
Penn State U.
University Park, Pa.

228. E.Salkovitz
University of Pittsburgh,
Pittsburgh, Pa. 15213.

219. R.Reese
Ohio State U.
Columbus, Ohio.

229. H.S.Spacil
G.E.
Schenectady, N. Y.

220. T.W.Ratcliffe
Babcock & Wilcox Co.
Beaver Falls, Pa.

230. Dr.L.L.Seigle
State Univ. of New York,
Stony Brook, L.I., N.Y.

221. R.Ricksecker
Chase Brass & Copper
Cleveland, Ohio

231 . W.F.Schilling
University of Pittsburgh
Pittsburgh, Pa. 15213

222. S.T.Ross
Brooks and Perkins Inc .
Detroit, Mich.

232. Dr.A.W.Schlechten
Colorado School of Mines
Golden, Color.

223. H. J.Ritz
Climax Molybdenum Co.
Langeloth, Pa.

233. E. J.Smith
National Steel Corp.
Weirton, w. Va .

224. P.Larosa
u.s.steel
Pittsburgh, Pa.

239 . J.P. Surls


Dow Chemical Co.
Pittsburg, Calif.

225. D.Ryan
National Forge
warren, Pa.

240. M.A.Scheil
A. 0. Smith Corp.
Milwaukee, Wisc .

226. G.Szilva
Climax Molybdenum
coldwater , Mich.

241. G.C.Sinke
Dow Chemical Co.
Midland, Mich .

227. C.E.Shoemaker
Bet hlehem St ee l Co.
B thl hem, Pa .

242. C. W.Spicer
U.S. Stee 1 Corp.
Duquesne, Pa .

ii' I

PARTICIPANTS
243. R.J.Slusar
Olin Matheison
New Haven, Conn.

253. G.Smith
J & L Steel
Pittsburgh, Pa.

244. L.F.Sander
Penn State U.
University Park, Pa.

254. W.Spear
International Nickel Co.
Pittsburgh, Pa.

245. A.Simkovich
Latrobe Steel Co.
Latrobe, Pa .

255. F. Tartaron
Bureau of Mines
Pittsburgh, Pa.

246. R.H.Spitzer
Carnegie Inst. of Tech.
Pittsburgh, Pa. 15213

256. E.S.Tankins
Naval Air Center
Philadelphia, Pa.

247. W.F.Stowasser
Air Products & Chemical
Inc.
Allentown, Pa.

257. M.S.Tubino
University of Pittsburgh
Pittsburgh, Pa. 15213
258. R. Tougas
Ecole Polytechnique
Montreal, Canada

248. P.E.Snyder
Westinghouse Res. Lab.
Pittsburgh, Pa.

259. E.T.Turkdogen
U.S.Steel Corp.
Monroeville, Pa.

249. R.B.Shaw
Allegheny Ludlum Steel
Brackenridge, Pa.
250. Dr.H.W.Schnaible
U.S. Steel Res.
Monroeville, Pa.
251. P.Schneider
Dravo Corp.
Pittsburgh, Pa.
252. A.J.Skalican
U.S.N.
Annapolis, Md.

260. S.Tager
Case Inst. Of Tech.
Cleveland, Ohio

'

261. G. H. Turner
Consolidated Mining &
Smelting Co.
Trail, Canada
262. M. C.Udy
Udy Processes Inc.
Niagara Falls, N' Y.

THERMODYNAMICS CONFERrnNcm
263. K.J.Vogel
3648 Forest Ave.
Munhall, Pa. 15121
264 .

o. Valla
Penn state U.
College Park, Pa.

273. L. A. Will y
ALCOA Rea . LR h.
New Kensingt n, I .
274. R. W. Wilson
Pittsburgh Met . Div. Alt
Cleveland, Ohio

265. Dr.G.Waddington
National Academy of Science
Washington, D.C.

275. W. Wilson
A.Fink! & Sons
Chicago, Ill.

266. P.N.Walsh
Union Carbide
Tarrytown, N. Y.

276. W.G.Wilson
Molybdenum Corp.
Pittsburgh, Pa .

267. J.S.Warner
Columbia University
Tappan, N. Y.

277. D.W.G.White
Depart. of Mines & Tech.
Surveys
ottawa, Ontario, Canada

268. I. H. Warren
University of B.C.
Vancouver, B. C.
269. J.R . Weeks
Brookhaven National Lab.
Upton, L.I., N.Y.
270. P. L.Weston
U.S.B.M.
.Minneapolis, Minn.
271. T. C. Wilder
Kennecott Copper Co.
Lexington, Mass.
272. D. II. W lkins on
N. r. 7. lnci Co.
I I111! l'l,011, Pt

278. F.H.Woehlbier
Battelle Mem. Inst.
Columbus, Ohio
279. Dr.E.F.Wondris
,National Steel Corp.
Weirton, W. Va.
280. G.W.Worth
Republic Steel
Cleveland, Ohio
281. P. Woolf
Bureau of Mines
Pittsburgh, Pa.
282. J.E.Werner
Bethlehem Steel
Bethlehem, Pa.

....

11

PARTICIPANTS
283. F.M.Wiberg
Swindell-Dressler Co.
Pittsburgh, Pa.

293. J.M.Gaines
Union Carbide Co.
New York, N. Y.

284. Dr.C.H.Wilkins
University of Alabama
Tuscaloosa, Ala.

294.

285. H. A.Wriedt
U.S. Steel
Monroeville, Pa.

295. K.Schwerotfeger
U.S. Steel
Pittsburgh, Pa.

286. Dr.C.A.Wulff
Carnegie Inst. of Tech.
Pittsburgh, Pa. 15213

296 . L.M.Diran
International Nickel Co.
Pittsburgh, Pa.

287. N.Yoshida
W.S .Atkins & Assoc.
Toronto, Ontario, Canada

297. R.L.Mickelson
Interlake Iron Corp.
Wellcmick, Ohio

288. C.V.Zabielski
1685 Parkline Dr.
Pittsburgh 27, Pa .

298. C.R.Lillie
Arnsted Industries
Bensenville, Illinois

289. Dr.C.A.Zapffe
earl A.Zapffe Firm
Baltimore, Md.

299 . F.Ernick
Westinghouse Electric
Greensburg, Pa.

290. G.R.Zellars
Aerospace Science
Cleveland, Ohio

300. W.M.Hyams
Crucible Steel
Midla~d, Pa.

291. R.B.Emerick
Allegheny Ludlum
Brackenridge, Pa.

301. H.K. Bruner


Interlake Iron Corp.
Wellowick, Ohio

o. M.Katz
Westinghouse Electric
Pittsburgh, Pa.

'

292. B.Hoffacker
St. Joseph Lead Co.
Monaca, Pa.

302. W.A.Roberson
Rutherford Co.
Cleveland, Ohio

421