Chemical Engineering Science, 1968, Vol. 23, pp. 981-990.

Pergamon Press.

Printed in Great Britain.

The kinetics of the oxidation of hydrogen chloride over molten salt catalysts
D. M. RUTHVENt and C. N. KENNEY Department of Chemical Engineering, Pembroke Street, Cambridge
(First received

13 December

1967; in revisedform

20 February

1968)

Abstract-This paper describes a study of the kinetics of oxidation of hydrogen chloride over unsupported melts containing copper, potassium and lanthanum chlorides. It is shown that kinetic data may be interpreted by assuming that the equilibrium: CuCl, = CUCI+ Cl, is maintained at all times between catalyst melt and gas and that the overall rate of hydrogen chloride oxidation is determined by the rate at which the melt absorbs oxygen. Rate constants calculated from the overall kinetic data agree weII with values obtained previously from a study of the kinetics of oxygen absorption. The activity of the catalyst is promoted by the addition of lanthanum chloride and it is shown that this effect is due to catalysis of the oxygen absorption step. The complex effect of temperature on the rate of the overall reaction may also be accounted for satisfactorily in terms of this mechanism. For melts containing equimolar amounts of potassium and copper chlorides it is shown that the apparent activation energy is.approximately 28 kcal per mole of hydrogen chloride oxidised. The conclusions from this studv should be applicable to supported Deacon catalysts which . have the same active components. INTRODUCTION

the Deacon process for the production of chlorine by the catalytic oxidation of hydrogen chloride has been revived in recent years with the growth of the demand for chlorinated organic chemicals which in their manufacture give large amounts of by-product hydrogen chloride. The process has been reviewed elsewhere [ l-31 and the use of a fluidised bed has been described. In spite of its increasing importance, little kinetic data has been published on the Deacon reaction. With a chromium based catalyst, Jones et al.141 observed complex kinetics and Soloshenko and Vilnyanski [5] found the reaction to be first order in oxygen. The purposes of this paper are two-fold: kinetic studies of the Deacon reaction are described using a catalyst based on that used by Deacon, containing copper chloride, to which however, potassium chloride and lanthanum chloride have been added. Such a catalyst is closely related to the doubly promoted ones listed in the recent patent literature which contain both alkali metal and rare earth halides in
INTEREST in

addition to copper chloride[6, lo] these catalyst systems overcome the principal disadvantages of the original Deacon catalyst. They are active at temperatures below 400C and so the equilbtjum conversion to chlorine is much greater than at the higher temperature which is necessary with less active unpromoted copper chloride catalysts. In addition, there is only a small loss of catalyst through volatilisation. It is argued here that the catalytic oxidation occurs by the following oxidation-reduction mechanism in which the reoxidation of cuprous chloride to cupric chloride is rate determining. 4Cu++ + 4Cl4cu+ +o, = 4Cu+ -I-2C12 Fast equilibrium (1) ---, 4Cu++ + 20-- Slow rate (2) determining step + 2H,O+4ClFast reaction (3) (4) catalyst for

20--+4HCl

4HCl+ 0, = 2H,O + 2C12 Lanthanum chloride is an effective the reoxidation step.

tPresent address: Department of Chemical Engineering, University of New Brunswick, Fredericton, N.B., Canada.

981

D. M. RUTHVEN

and C. N. KENNEY required proportions from cylinder gases and the exact composition was determined by analysis. The development of a suitable method of gas analysis proved difficult and time consuming. Gas chromatography ultimately proved satisfactory and the instrument, which has been described elsewhere[7], was used to analyse inlet and outlet gases for nitrogen plus oxygen, chlorine and hydrogen chloride. From the sample valve the gas stream was passed through a tube packed with caustic soda and molecular sieve pellets in order to absorb water, chlorine and hydrogen chloride and the proportion of oxygen to nitrogen was then determined using a paramagnetic oxygen meter. No attempt was made to determine the amount of water vapour directly and the partial pressure was assumed to be the same as that of chlorine. It was not possible to determine hydrogen chloride as accurately as other components and values were generally obtained by difference. In evaluating reaction rates, a mass balance over the reactor for each component was used to check the consistency of the analyses: agreement to within + 5 per cent was usually obtained. In the first series of experiments relatively large amounts of melt (50-1OOg) were used and the effect of varying the stirring speed, the surface area and the depth of the melt was investigated. The reaction rate was found to be proportional to the melt surface area and independent of the total volume of melt present. Stirring had little effect on the reaction rate provided that the rate of stirring was not fast enough to disturb the surface seriously. Having established this it was possible to use much smaller amounts of melt (- 10 g) and so to reduce the time required to reach the steady state which varied from a few hours at low gas flowrates to a few minutes at high flowrates. Even an extremely thin layer (< 1 mm) of melt proved to be an active catalyst suggesting that a surface reaction is involved, and this view is supported if not proven by the observation that the reaction rate is proportional to the surface area of the melt. Specific reaction rates were based on the geometrical area of the melt surface calculated from the diameter of the reactor tube. Some difficulty was encountered through the tendency of the melt to adhere to the tube walls thus increasing the effective area and although great care was taken in filling the reactor tube, the uncertainty in area is

The second aspect of this catalyst system which is of interest is that, at the reaction temperature, it is an homogeneous ionic melt. A kinetic study of the catalysis of sulphur dioxide oxidation by vanadium ions in molten pyrosulphates has recently been described[ 11,121 but otherwise little is known about the properties of such catalysts which may, as a result of the high mobility of charged species, show unusual properties. That some of these catalyst systems may be molten when distributed on a porous inert sup port has been recognised[6], but in the present work an unsupported pool of melt was used to obtain kinetic data. Such a system has considerable advantages for laboratory studies since the melt surface area is fairly accurately known, pore diffusion is eliminated, temperature control is simple and the analysis of catalyst composition is easier than with conventional supported catalysts.
EXPERIMENTAL The essential details of the apparatus are shown in Fig. 1. Gas mixtures containing oxygen, nitrogen, and hydrogen chloride were passed at known Rowrates over a heated sample of melt. A large glass syringe (cu. six litres) was used to prepare the inlet gas mixtures and to obtain accurately known flowrates. The syringe was fitted with a P.T.F.E. piston, and driven by a synchronous motor through a variable speed gear box, giving a range of flowrates from 3 to 300cm3/min. Over this range of flowrates the reactor behaved as a well mixed vessel. The piston was sealed by a rubber 0 ring covered by a small amount of a heavy liquid parat%. Gas mixtures were made up in approximately

Reoctionwssel he&d in ebctric fumoce Mixhre of 0, N,6 HCL Chmnotogmph sompb MIVO (in fnezing Monomabr

Oxm

Gas ayrirge with voriobk~apeed drive

Fig. 1. Diagram of apparatus for studying the overall kinetics of hydrogen chloride oxidation.

982

The kinetics of the oxidation of hydrogen chloride probably a major cause of experimental error. A wide range of reaction rates was encountered but it was not possible to measure with any accuracy rates lower than about 10-r g moles/cmg sec. The melts were analysed at the end of each series of experiments by methods which have been described previously[8]. The proportion of cupric copper was always high and the amount of oxide present always low. A typical analysis together with the range of experimental variables is given in Table 1. Table 1. Range of experimental variables investigated in the oxidation of hydrogen chloride Variable Range

sures over copper chloride melts 191it was shown that the reaction: CuCl, = CuCl + #Zl, (0 (9) (0 (5)

Total pressure atmospheric Temperature 35OC-450C Gas Composition 0.1-0.65 atm 0, 0.18-0.65 atm HCl O+Ol-0.20atm Cl, 3-300 cm3/min Gas flowrate Catalyst composition l-0 throughout KCL/(CuCl+ CuCl,) 0- 16 per cent mole LaCl, Typical catalyst analysis % mole (after use) K+ 44.5 La3+ 11.0 cu++ 39.0 cu+ 5.5 O-cl.8

RESULTS

AND

DISCUSSION

An extensive series of experiments was carried out to determine the approximate form of the rate equation. Oxygen and hydrogen chloride partial pressures were varied independently, and by changing the flow rate, different conversions and therefore different chlorine partial pressures were obtained. Results of one such series of experiments are shown in Fig. 2 in which reaction rates are plotted against chlorine partial pressure for various oxygen and hydrogen chloride partial pressures. The reaction rate depends strongly on oxygen and chlorine partial pressures but appears to be almost independent of the partial pressure of hydrogen chloride. The inverse dependence on chlorine partial pressure is particularly noticeable at very low conversions and is thus unlikely to be due to the reverse of the reaction given in Eq. (4). In our study of the equilibrium chlorine pres-

is fast compared with the overall rates of hydrogen chloride oxidation encountered in the present work. It is therefore probable that, when such melts are used as catalysts for hydrogen chloride oxidation, this equilibrium is always established between gas and melt once the steady state has been reached. The observation that the reaction is zero order in hydrogen chloride together with the low oxide content of the molten catalysts both suggest that the reaction of hydrogen chloride with oxygen in the melt is rapid. It is therefore possible that the overall reaction rate is controlled by the rate at which the catalyst melt absorbs oxygen, a conclusion which is supported by the strong dependence of the rate on oxygen partial pressure. These considerations suggest that it may be possible to formulate the mechanism of the Deacon reaction in terms of the steps shown in Eq. (l)-(3), and a rate expression can be simply obtained using these assumptions. We have shown in an earlier paper[8] that the rate of oxygen absorption by a melt containing cuprous chloride is given by a kinetic equation of the form: r = k,[Cu+]2Po* (6) where: [Cu] = mole fraction of cuprous copper; p& = partial pressure of oxygen, atm
k3 = third

order rate constant calculated from oxygen absorption studies, g moles/cm2 set atm.

If this process controls the overall reaction rate, the rate of hydrogen chloride oxidation as given by Eq. (4) is: 4kA [Cu+12Po, g moleslsec where
k = rate

(7) absorption,

constant for oxygen g moles/cm2 set atm

983

D. M. R U T H V E N and C. N. K E N N E Y

A = superficial area of catalyst, cm 2 [Cu +] = mole fraction of cuprous copper in catalyst. Considering equilibrium [5] we have:

j K -- [--C--a-~] . N/Pc==
where: K = equilibrium Constant j = activity coefficient ratio = yII/yI.

[Cu +]

(8)

Let the mole fraction of total copper in the melt be a = [Cu +] + [Cu++]. Eliminating the mole fractions of cuprous and cupric copper from (7) and (8), we obtain the expression:
4kj2 K2 a 2

R = (jK+X/(ecl,))z
where:

.Po,

(9)

R = Specific rate of hydrogen chloride oxidation (g moles/cm 2 sec). This may be written:

4_~=

1 1 "~/Pcl=. 2aN/'k -~ 2ajK'k/f~

-SO --

(10)

If the activity coefficient ratio, j, is independent of chlorine partial pressure, a plot of ~/(Po2/R) against ~/Po, will be linear, with slope 1/2ajKV'k and intercept - - j K on the ~/Pct, axis. It should thus be possible to determine ttoth j K and k from such a plot, although this requires accurate data over a wide range of chlorine pressures. The kinetic scheme given above may however be tested by using the values of the chlorine equilibrium function, j K , derived from our study of chlorine equilibrium pressures over copper chloride melts [9] to fix the intercept on the X/Pct~ axis and then the apparent rate constant, k, for oxygen absorption can be derived from the best straight line through the experimental points. This value may then be compared with the value of k3 obtained directly from oxygen absorption measurements. A linearised plot of the data in Fig. 2 is given in Fig. 3 which strongly supports the proposed kinetic model. This correlation could not be tested over the complete range of chlorine pressures, since at low chlorine pressures the kinetic measurements become inaccurate and on the high pressure side the rate is limited by the overall equilibrium. It does however appear to fit the data quite well over the range of chlorine pressures

Rob of HCL=oxklolton x I07 (g moles/cm%ec)

o~

Melt c ~ : K C L : 4 5 %

mole

\
~. o~
_

\
N.. ~ ~

c.cr+c.c~,~"~,~
Temperature420"C : Inlet gaj_

02756

v ~71
o 55 47

3,o

,.or\
/ "F
I ~ I1%0~

"
" ~

I0

15 % Clzin outlet go=

~!0

Fig. 2. Variation of rate of H CI oxidation with gas c o m p o s i t i o n - S e r i e s L2.

984

The kinetics of the oxidation of hydrogen chloride

Melt composition

KCL:45%molo CuCl t CuCL,:45% mole

Temperature

:42O*C

x o v
A

0 01 o.2 o3 @4

15 27 29 41 53

85 36 71 59 47
05

Fig. 3. Rate of HCI oxidation- linearised plot for Series L2.

from O-02 to 0.20 atm. Values of k were obtained from the slopes of plots of this type. The activity of these melts as catalysts for hydrogen chloride oxidation is strongly dependent on the proportion of lanthanum chloride present, in accordance with our earlier observations of the strong catalytic effect of lanthanum on the rate of oxygen absorptionE81. A com-

parison of the rate constants for a number of melts is given in Table 2 and these values are plotted in Fig. 4 together with the values of k3 determined from oxygen absorption studies. Although there is considerable scatter, the values of rate constants calculated from the overall kinetic data and the chlorine pressure data agree quite well with those measured directly.

,ooo

Rote conetont x UY
(gmole/cm2sec atm)
0 0

8oo _
600 400 500 200 -

Wues

of k for

HCL oxidation. x Values of k for oxygen otsorption

Melts containing e@rr&r orna~nts of potassium 8 coppar chlorides

0 mole %

IO

12

I4

16

LOCI,

Fig. 4. Variation of rate constants with lanthanum content of melt.

985

The kinetics of the oxidation of hydrogen chloride

lcco 800 700 600 so0 400 -

MBltsconloii~0fpolprimacopprc~ whm dkcakdl6fJ~rbpr

otlinrari66d pm.

300

A i-15 85Kmd6

LaCl,

200

Ll4 IDBY. Ll3 14.4%

co
I.3

I
I.4

I
I.6

Fig. 5. H Cl oxidation; variation of rate constant k with temperature.

Fig. 3 a single mean value ofjk was chosen since the variation in jk over the range of chlorine pressure used is cu. 5 per cent. The temperature dependence of the equilibrium function is given by alnjK -=r3T

AH
RT2

(11)

where k = 4ka2 (jK)* is the pseudo rate constant for a given melt at a given temperature.? It follows from this expression that the apparent activation energy will be approximately 286 kcal; twice the heat of reaction (5) since:
Ink = ln(4ka3 + 2 1njK.

where LW (- 14.3 kcals/mole HCl) refers to reaction (5). Provided that the partial molar enthalpies of the cupric and cuprous species are independent of temperature, the customary straight line plot of 1ogjK against l/T should be obtained. Because of the variation of the activity coefficients with melt composition however, this is a complex function and the linear relation holds only at low temperatures or high chlorine pressures. Fortunately, in spite of these complications it is possible to simplify the rate expression somewhat since the chlorine equilibrium constant is strongly temperature dependent falling to approximately O-01 at 350C from 0.055 at 450C. Consequently, at temperatures below about 400C and chlorine pressures greater than 0.05 atm, the value of this function is small in comparison with the square root of the chlorine pressure and the rate Eq. (8) simplifies to the approximate expression:

(13)

The variation of k with temperature is small and the temperature dependence of jK is given by: JlnjK -=-=-A/Y
R

14300
R

from

(11).

a+ 0

For a melt containing KCl: 45.75 per cent mole, CuCl+ CuCl,:45.75 per cent mole, LaCl, :85 per cent mole, the pseudo rate constant k may thus be written:
k = 445 exp (- 28,6OO/RT) g moles/cm2 sec.

The Arrhenius plot of the experimental values of k for a melt of this composition is shown in Fig. 6. For melts containing less than equimolar proportions of potassium to copper chloride, it was shown previously that there is an appreci-

987

D. M. RUTHVEN and C. N. KENNEY

Appmnt actktion cna(ly =28.4 k Cuts/mole HCL

\
@6 0.5

I
I,3 I.4 lo

I
I.5

I
I.6

TOK

Fig. 6. Temperature variation of pseudo rate constant for HCl oxidation.

able activation energy for oxygen absorption[8] and hence the apparent activation energy of the pseudo rate constant k will be greater than 28.6 kcalslg mole HCl. This value may be compared with the figure of 18.3 k&s/g mole HCl quoted for the activation energy over solid chromium based catalysts [ 51.
CONCLUSIONS

The kinetics of the oxidation of hydrogen chloride over copper chloride melts can be accounted for by assuming that the rate of reaction is controlled by the rate at which the catalyst can absorb oxygen, together with the assumption that reaction (5) is always at equilibrium. Varying the lanthanum chloride content of the melt and the temperature have been shown to affect both the overall kinetics and the rate of oxygen absorption into a copper chloride melt in similar ways, and there seems little doubt that the proposed mechanism is essentially correct. It would however be desirable to extend

the scope of the experiments to melts containing different proportions of potassium and copper chlorides to confirm this. From the datagiven in a recent patent specification[6] it appears that the rate constant per unit area for the supported catalyst is very much lower than for the unsupported melts, suggesting that the effective area is very much smaller than the B.E.T. area of the support. With such supports it is probable that some of the smaller pores may actually become blocked with the melt severely restricting the available area. In spite of the high specific activity of the unsupported molten catalyst it does not seem likely that a bubbler or other gas-liquid contactor would have a sufficiently large interfacial area to give overall reaction rates per unit volume of catalyst which would be as great as those employing supported catalysts. Nonetheless ionic melts appear to be unusual catalyst systems and it would be of great interest to know whether the Deacon reaction occurs on the surface or

988

The kinetics of the oxidation of hydrogen chloride

within the bulk of the melt, the nature of the reactive species, and the mechanism by which lanthanum can exert such a strong catalytic effect.
Acknowledgment-This work was carried out during the tenure of an S.R.C. studentship and the junior author (D.M.R.) wishes to acknowledge this financial support. NOTATION

A
k

[Cu+]+ [Cu], mole fraction of total copper in melt. surface area of melt, cm2 rate constant for oxygen absorption calculated from overall kinetic data for HCl oxidation, g moles cm-2 set-l atm-

k3 rate constant for oxygen absorption calculated from kinetic data for oxygen absorption, g moles cmm2set-l atm-l K equilibrium constant for reaction (5) atm05 partial pressure of oxygen,,atm partial pressure of chlorine, atm R specific rate of hydrogen chloride oxidation, g moles/cm2 set r specific rate of oxygen absorption g moles/cm2 set Ku1 mole fraction cuprous species in melt [Cu++] mole fraction cupric species in melt YI activity coefficient of Cu+ YII activity coefficient of Cu++ .i yZZ/-yZ= activity coefficient ratio

REFERENCES [l] JOHNSTONE H. F., Chem. Engng Prog. 1948 44 657. [2] ENGEL W. F. and WAALE M. J., Chemy Ind. 1962 76. [3] UANT J. Th., Bull. Ass. fr. Techns. Perrole 1963 157 8 1. [4] JONES A. M., BLISS H. and WALKER C. A., A ./.Ch.E.JI 1966 12 260. [5] SOLOSHENKO A. A. and VILNYANSKI Y. E., Chem. Abstr. 1964 62 86990. [6] Shell International. British Patent 908,022 (1962). r71 RUTHVEN D. M. and KENNEY C. N., The Analyst 1966 91 1086. c8] RUTHVEN D. M. and KENNEY C. N., Chem. Engng Sci. 1967 22 156 1. 191 RUTHVEN D. M. and KENNEY C. N .. J. inorn. nucl. Chem. 1968 Xl 93 1. [iOj SLAMA I., KODEJS Z. and PACAK P., Chem~~kyprumysi 1967 17142 179. [ll] GLUECK A. R., Ph.D. Thesis, Cambridge 1966. [ 121 GLUECK A. R. and KENNEY C. N., Chem. Engng Sci. In press. R&rmC-Cet article decrit une etude entreprise sur la cinetique doxidation dacide chlorhydrique sur des des fusions non soutenues contenant des chlorures de cuivre, de potassium et de lanthane. II est montre que Ion peut interpreter les don&es cinetiques en supposant que lequilibre: CUCI, = CuCl + Cl, est maintenu a tout moment entre la fusion et le gaz catalyseur et que le taux general doxidation dacide chlorhydrique est dCterminC par le taux auquel la fusion absorbe loxygene. Les constantes de taux calculees a partir des dot&es g&&ales de cinetique saccordent bien avec les valeurs obtenues par une etude anterieure sur la cinttique dabsorption doxygene. Le catalyseur est active par laddition de chlorure de lanthane et on observe que cet effet est dfi a la catalyse au moment de labsorption doxygene. Leffet complexe de la temperature sur le taux de la reaction gCnCrale peut aussi &tre considere dune maniere satisfaisante dans le contexte de ce mecanisme. 11est montre que, pour les fusions contenant des quantites CquimolCculaires de chlorure de potassium et de chlorure de cuivre, lenergie apparente dactivation est denviron de 28 Kcal par molecule dacide chlorhydrique oxydt. Les conclusions de cette ttude devraient supporter les catalyseurs de Deacon qui ont les mEmes composants actifs. Zusammenfassung-In diesem Artikel wird die Kinetik der Oxydation von Chlorwasserstoff iiber nicht untersttitzten Schmelzen, die Kupfer-, Kalium- und Lanthanchloride enthalten, untersucht. Es wird gezeigt, dass die kinetischen Daten unter der Annahme erkliirt werden kiinnen. dass das Gleichgewicht: CUCI, = CuCI+
Cl,

CESVol.

23 No.9-C

D. M. RUTHVEN

and C. N. KENNEY

jederzeit zwischen der Katalysatorschmelze und dem Gas vorhanden ist und dass die Gesamtgeschwindigkeit der Chlorwasserstoffoxydation durch die Geschwindigkeit mit welcher die Schmelze Sauerstoff absorbiert, bestimmt wird. Die aus den kinetischen Gesamtdaten berechneten Geschwindigkeitskonstanten stimmen gut mit den bereits in einer friiheren Untersuchung der Kinetik der Sauerstoffabsorption erhaltenen Werten tiberein. Die Aktivitit des Katalysators wird durch den Zusatz von Lanthanchlorid gesteigert und es wird gezeigt, dass dieser Einfluss auf die Katalyse der Sauerstoffabsorption zuriickzuftihren ist. Der komplexe Einfluss der Temperatur auf die Geschwindigkeit der Gesamtreaktion kann ebenfalls auf befriedigende Weise im Sinne dieses Mechanismus ausgedeutet werden. Es wird gezeigt, dass ftir Schmelzen, die 3quimolare Mengen von Kalium- und Kupferchloriden enthalten, die scheinbare Aktivierungsenergie cu. 28 Kcal pro Mel oxydierten Chlorwasserstoff ist. Die Ergebnisse dieser Untersuchung sollten auf unterstbtzte DeaconKatalysatoren, die gleichen aktiven Bestandteile haben, anwendbar sein.

990