Guni, L. el d.

(Editors), New Frontiers in Catubsb

Proceedings of the 10th International Congress on Catalysis, 19-24 July, lW2, Budapest, Hungary
0 1993 Elsevier Science Publishers B.V. All rights reserved

ALKALI PROMOTER SYNERGISM IN SELECTIVE OXIDATION

M.M.Bhasin and C. D. Hendrix

Union Carbide Chemicals & Plastics Company, Inc., P.O.Box 8361, South Charleston, West
Virginia 25303, USA

Abstract
Individual alkali metals (as various salts) have been used as promoters to
improve selectivity in a variety of oxidation processes. Typically, selectivity and/or
activity response of an oxidation system to changes in alkali concentration is non-
linear and large amounts of alkali can act as a poison. To date mixtures of alkali
have not been explored in a systematic manner. It will be shown that thru the use
of well planned, statistically designed experiments, mixtures of alkali gave
synergistic selectivities higher than that obtainable from even the best of the
individual promoters. Mathematical models will be presented that describe such
beneficial synergistic selectivity (and activity) improvements alongwith their
experimental confirmation.

INTRODUCTION

Alkali metal promoters have played a key role in improving selectivity and
activity to desired oxidation products in a variety of commercial selective
oxidations (1). Few examples of commercial interest are: ethylene oxidation to
ethylene oxide, benzene and butane oxidation to maleic anhydride as well as
dehydrogenation of ethylbenzene to styrene. Recently, alkali metal promoters
have also contributed to the selective oxidation of methane to produce ethylene
and ethane. Extensive literature on alkali metal promoters for ethylene oxidation
to ethylene oxide has been reviewed in a recent patent (2) and publications (3).
To date, the literature in catalysis has searched and discovered optimum catalysts
for such reactions by laboriously varying one-variable at a time. When dealing
with mixtures of alkali metals, classical approach has been to hold one promoter
variable constant while attempting to vary the other and vice versa. Such an
approach assumes no interaction between the variables, i.e., the effects of each
promoter on catalytic performance is independent of the concentration of each
promoter. In this work, it will be shown that through the use of statistically
designed experiments in 3 or more alkali promoters and/or catalyst composition
variables, synergistic action between such promoters was discovered for selective
oxidation of ethylene to ethylene oxide. Typically, alkali metal promoters give
1432

non-linear response, i.e., at low concentration (below optimum) increasing alkali

promoter concentration increases efficiency until optimum performance is attained
while at concentrations beyond optimum, effect of increasing alkali promoter
concentration is to decrease efficiency. Synergism or synergistic action means
that the combined action of promoter A with promoter B (over an unpromoted
catalyst) results in an enhancement of efficiency and/or activity beyond that could
have been achieved by either promoter acting alone or in a merely additive
manner. Efficiency to any desired product is described as the rate to that product
divided by total rate of reaction. Examples will be given of synergistic promoter
action including examples whereby addition of second promoter, lithium to an
inactive, high concentration, cesium containing catalyst resulted in an excellent
efficiency and activity catalyst. Methodology of statistical designed experiments
that led to discovery of such synergism will also be highlighted.

EXPERIMENTAL

Catalysts were prepared using standard vacuum impregnation technique by

dissolving silver oxide in an ethylenediamine-water-oxalic acid solvent mixture
with appropriate amounts of promoter (2). Carrier was soaked with excess
solution under vacuum and the excess solution drained off. The impregnated
carrier was roasted in a monolayer over an endless spiral weave belt passing over
a hot zone at 50OOC for 2.5 minutes; thereby combusting the organics and
reducing Ag+ salts to Ago metal with appropriate amounts of promoters.
Catalyst testing was done in a differential rector of Berty design running at
1500 rpm such that catalyst bed was essentially isothermal (4). 80 ml of catalyst
was tested with 22.6SCFH with a GHSV of 8000 hr-1 at 275 psig. Temperature
was adjusted to produce 1.0% ethylene oxide using an inlet gas composition of
6.0% 02, 8% C2H4, 0.5% C2H6, 6.5% C02 and 7.5 ppm ethyl chloride (balance
Ng). Catalysts were stabilized for performance for 4-6 days and temperature,
efficiency and balances (carbon and oxygen) measured. Efficiency and selectivity
terms used in this paper are synonymous.

Typically, Central Composite design was used in three variables consisting of

wt. Yo of silver, alkali promoter A and alkali promoter B. Such a typical design in
two variables, X2 and XI, is shown in Figure 1. The dark circles in Figure 1
represent typical factorial design points while the open circles are the axial points
and a center point. The latter points, representing experiments, are critical to
modelling and defining the non-linear and interactive response of variables. In
addition, planned experiments were carried out in varying levels of each promoter
alone to assist in defining the concentration ranges of each variable. Efficiency
and temperature (a measure of activity) responses were modelled as a function of
linear, square and interaction terms of all three variables. These mathematical
models were obtained by fitting experimental efficiency and temperature data to
the controlled independent variables by the method of Least Squares (5). Such a
regression analysis can be readily done using a digital computer or any of the PCs
(Personal Computer). Typical promoters can be any binary or multicomponent
mixtures of alkali metals, Cs, Rb, K, Na and Li.
 1433

RESULTS AND DISCUSSION

16 catalysts of one typical Central Composite design set were evaluated for
ethylene epoxidation reaction to obtain Efficiency and Temperature responses for
producing -1Yo ethylene oxide. Mathematical relationships or Models for the
Efficiency and Temperature responses were developed as a function of the three
independent variables, the non-linear square terms and interaction terms
according to the method of least squares (5) using an available computer program
for Regression Analysis. For example, a typical Efficiency and Temperature Model
for one specific Ag, Cs and K synergistic system is:

Efficiency, YO= 77.19+0.196 (BG)+133.4 (BCs)+120.5 (BK)-63787.0 ( B C S ) ~

-47229.0 (BK)2+52.1 (BG x BC~)-60823.0(BCS x BK)

Temperature, "C = 259.44-1.279(BG)+1555.8(BCs)+794.1(BK)+0.3466(BG)2

+168544(BCs)2+147856.0(BK)2+132807.O(BCs X BK)

Where: BG = (wt. Yo Ag-11 .O), BCS = (wt. Yo CS-O.Ol),

BK = (wt. Yo K-0.003)

0
a a
0 0 0

Factorial points
a
0 Axial and center points

Variable, XI

Figure 1. Central Composite Design in Two Variables, XI and X2.

From these models, contour maps were generated representing lines of

constant efficiency and temperature. Thus catalysts of maximum efficiency or
activity or any other desired combination of activity and efficiency could be
predicted and made with very high degree of precision-several times lower
standard deviation of a single test.
1434

An excellent example of such synergism is shown in Figure 2.. Thus, for 7% Ag,
the optimum K-only catalyst has an efficiency of -74% while Cs-only catalyst has
75.5% while the best Cs+K catalyst has a significantly higher synergistic efficiency
of 76.7%. Also note that Cs promoter has a very sharp efficiency response while K
and K+Cs have very broad efficiency response to respective promoter
concentration. Contour maps for efficiency and temperature generated in the
experimental space are shown in Figure 3 for 7% Ag in the catalyst.

77

76

75

74

73

72
0 1 2 3 4 5
Total Alkali (GEW/Kg Catalyst x 103)

Figure 2. Cs-K Synergism at 7% silver

Figure 3 shows contour lines of constant efficiency and temperature in

potassium-cesium concentration space. The darker lines represent selectivity
contours with selectivity values being indicated for each line. The lighter, broken
lines represent temperature contours with temperatures indicated in degrees
centigrade. The maximum selectivity achievable for any combination of cesium
and potassium as a function of cesium concentration is indicated by the curve
labelled SelMax. This curve represents the maximum synergism achievable for
any given level of cesium. The synergism curve (designated "A" in Figure 3)
separates the region of synergism and regions of additive and antagonistic effects.
Additive effects occur when the efficiency obtained with a combination of cesium
and potassium is a weighted average of the efficiencies containing the respective
amounts of cesium and potassium. Antagonistic effects occur when the efficiency
obtainable with a mixture of cesium and potassium is less than that obtainable
from corresponding amounts of individual alkali. The area bound by ordinate, the
abscissa and the synergism curve (curve " A ) defines the area of synergism.

The region of synergism varies with the silver content of the catalyst as
described by the Efficiency Model. Figure 4 illustrates the region of synergism at
13% silver. The Selmaxcurve and synergism curve (curve " A ) of Figure 4 are the
same as defined with reference to Figure 3. Comparison of Figures 3 and 4 show
that the efficiency contour lines are simitar but are displaced with respect to the
 1435

0.020

0.015
�
� 0.010

0.005

0.005 0.010 0.015 0.020

Wt.% Cs
Figure 3. Contour maps of selectivity (%) and temperature (°C) for 7% silver
catalysts

cesium and potassium axes. For example, the 76% efficiency contour line in
Figure 3 envelopes an area which is smaller than the area covered by the
corresponding contour line in Figure 4. Further, the maximum value of selectivity
obtainable at 7 weight percent silver, i.e., the maximum along the Selmax line is
slightly above 76.6% {Figure 3), while the corresponding maximum for 13 weight
percent silver is slightly above 77.8% (Figure 4).

0.020

0.015
�
� 0.010

0.005

0.005 0.010 0.015 0.020

Wt. %Cs

Figure 4. Contour maps of selectivity (%) and temperature ( ° C) for 13% silver
catalysts
 1436

The Model predictions were checked out experimentally by preparing three

sets of catalysts at 7% Ag; one set containing optimum cesium concentration of
0.0906% (Catalyst l A , 1B and 1C), second set (catalysts l D , 1E and 1F)
containing 0.0906% Cs plus 0.004% K (synergistic combination) and third set
(Catalyst 1G) containing 0.004% K only (Table I).
The efficiency model predicts the difference between synergistic Cs + K
catalysts (1D, 1E,and 1F) and the Cs only catalysts (1A, 1B and 1C) to 0.6%. The
experimentally observed efficiency difference is 0.7% (76.2-75.5%). This is an
excellent agreement with the predictions of the Efficiency Model. Also note that
the synergistic combination catalyst is more efficient even though it is less active
by 5-6OC. In addition, the temperatures observed were also in excellent
agreement with the Temperature Model prediction (see Figure 3). Replicate
experiments here were necessary to reduce the standard deviation of efficiency.
The K only catalyst (1G) gives 72.8%--once again in very good agreement with
model predictions (see Figure 3).

TABLE I
Confirmation of Model Predictions

Cesium Only Catalysts:

7.0% Aa. U 9 0 6 ~ CS
0
1A 75.4 263.9
1B 75.7 261.4
1c 75.4 266.1
Average 75.5 f 0.17*

I !&sium Plus Potassium CataIvsts- 7 OO/ Aa- 0 009060/ Cs 0 .004% K

1D
1E
76.0
76.1
271.4
271.5
1F 76.4 270.9
Average 76.2 0.21'

1G P o t W m Onlv Catalyst.. 7.0o/o Aa. 0.0040% K.

1G 72.8 265.3
*Values reflect standard deviations.

Synergism between Cesium and lithium promoters was also studied. This
alkali mixture demonstrated some of the most striking examples of synergism
(Table 11). Synergistic catalyst combinations 2C, containing cesium and lithium
was about 10% more efficient than the corresponding lithium catalyst, 28 and
about 3.5% more efficient than cesium only catalyst, 2A.

Catalyst 2D demonstrates that when the lithium concentration of catalyst 2C

was increased about ten-fold, the resulting efficiency was reduced by more than
 1437

TABLE II
Catalysts Containing Synergistic Combinations of Silver, Cesium and Lithium

I Catalvst
2A
Wt. 70AQ

12.8
Wt. Yo CS

0.0052
Wt. Yo I i

0.000 72.6
WD.("C)

257
2B 13.1 0.000 0.0030 65.9 257
2c 12.8 0.00521 0.0030 76.0 257
2D 13.1 0.0053 0.0302 72.8 258
2E 13.0 0.0300 0.0300 78.6 258
2F 13 0.000 0.0300 65.4 263
I 2G 13.0 0.0300
*No measurable ethylene oxide was produced.
0.000 Inactive*

3%. However, a comparison of catalysts 2D and 2E discloses that the efficiency

obtained with catalyst 2D was increased by increasing the cesium concentration
from 0.0053% to 0.0300%. It should be noted that the cesium only catalyst, 2G
(corresponding to the cesium concentration of 2E was inactive; that is, at that
concentration of cesium, no ethylene oxide was produced. Yet, the combination of
this concentration of cesium with 0.0300% lithium resulted in a very active (and
efficient) catalyst (2E). Similarly, the lithium only catalyst of 2F (corresponding to
the lithium concentration of 2E) provided a relatively low efficiency of 65.4%.
Thus, the synergism achieved in catalyst 2E is underscored by the sharp contrast
of its efficiency relative to the poor results achieved with catalysts 2F and 2G, the
corresponding lithium only and cesium only catalysts.

Cesium-sodium mixtures also exhibited synergistic behavior. Table Ill below

demonstrates the synergism achieved with the catalyst of experiment 3C relative
to the corresponding cesium only catalyst of experiment 3A and the corresponding
sodium only catalyst of experiment 3B. Thus, the catalyst of 3C was 5% more
efficient than the cesium only catalyst of 3A and 7.2% more efficient than the
sodium only catalyst of 3B. Catalyst 3D discloses that increasing the sodium
concentration ten-fold relative to the concentration in catalyst 3C had the effect of
lokering the catalyst efficiency. Temperature of the synergistic catalyst was lower
than the corresponding cesium only and sodium only catalysts.

TABLE Ill
Catalysts Containing Synergistic Combinations of Silver, Cesium and Sodium

Catalvst kYcz&i uUd2.s !A!t&k %LEff& I Q m l l l

3A 12.8 0.005 0.000 72.6 257
38 13.0 0.000 0.010 71.4 259
3c 13.0 0.005 0.010 78.6 251

More complex mixtures of alkali were also studied and likewise they exhibited
synergistic behavior. Table IV below demonstrates such a synergism achieved
1438

with a catalyst containing silver and cesium in combination with potassium,

sodium and lithium as indicated in experiment 4A. The presence of chloride ion in
the catalyst concentrations of 4A, B and C indicates that at least a portion of the
alkali metal salts used in the catalyst preparation procedure were in the form of
chlorides. The efficiency achieved with the synergistic combination of experiment
4A was about 1.2% greater than the efficiency of the corresponding cesium only
catalyst of 4C and about 9% greater than the corresponding combination of
potassium, sodium and lithium of catalyst 48.

TABLE IV
Catalysts Containing Synergistic Combinations of Silver, Cesium, Potassium,
Sodium and Lithium
Wt. % Wt. Yo Wt. ?loWt. Yo Wt. Yo Wt. Yo Yo Temp.
C a t a l v s t A a C s K - Na - Li _CI Eff.eC2.l
4A 13.1 0.0132 0,001 0.002 0.006 0.0137 78.4' 262
48 13.0 0.000 0.001 0.002 0.006 0.0137 69.4 267
4C 13.1 0.0132 0.000 0.000 0.000 0.0035 77.2 265
*The efficiency represents the average of two experiments

CONCLUSIONS

It has been shown that via the use of well planned, designed experiments,
mixtures of "poor" and "good" alkali promoters gave synergistic selectivity
improvement that was even higher than the highest selectivity obtainable from the
"good" promoter alone. Mixtures of cesium and lithium gave one of the striking
examples of synergistic behavior where addition of lithium made an inactive
cesium catalyst very active and efficient. Moreover, the efficiency response to
alkali concentration was broad for the synergistic mixtures--resembling the broad
response of the "poor" alkali promoter. Higher optimum efficiency was observed
with higher Ag content catalysts in comparison with lower Ag content catalysts.
High precision mathematical models were developed from well planned
statistically designed sets of experiments. The high precision mathematical
models developed accurately predicted the observed performance of mixed alkali
catalysts.

REFERENCES

1 M. Sittig, "Handbook of Catalyst Manufacture", Oxidation Process Chapter,

pp 367-412, Noyes Data Corp. 1978
2 M.M. Bhasin, C.D. Hendrix and P. C. Ellgen, U S . 4,916,243, April 10, 1990
3a J.N. Cawse, J.P. Henry, M.W. Swartzlander and P.H. Wadia, Kirk Othmer:
Ency. of Chemical Technology, Vol. 9, 432 (1980), John Wiley & Sons, Inc.
3b J. M. Berty, Applied Industrial Catalysis, Vol. 1, 207 (1983), Academic Press
4 J. M. Berty, Chem. Eng. Prog., Vol. 70, No. 5. pp 78-84, 1974
5 O.L. Davis & P.L. Goldsmith, Chp. 8 Longman Group Ltd. London 4th Ed, 1976
 1439

DISCUSSION
Q: J. M. Berty (USA)
Industry by "Enlightened empirics", as Professor M. Boudart called it 30 years ago
has to solve a technical economic problem and to make money. Then it can support
academic research to explain the success in the next 30-40 years. Even empirical
experiments need good reactors and good experimental methodology.
A: M. M. Bhasin
I agree with your comment for the most part. Since, typically, there is no training
early on with the "Statistical Designed Experiment" approach, one-variable-at-a-time is
the approach overwhelmingly followed. Both approaches have their place in research and
development, although the Designed Experiment approach is more commonly practiced
in development. I have attempted to highlight its utility in the research arena as well.
Q: W. N. Delgass (USA)
I agree that the statical approach you describe is an excellent way to search for an
optimum in a complex variable approach. If you want to know why the optimum occurs,
i.e. the chemistry, I am inclined to think you must return to single-vanable experiments.
Do you agree ?
A: M. M. Bhasin
No, not entirely. In fact, for a more complex system, the Designed Experiment
approach is all the more important. Of course, the "one-variable" approach is needed to
scope out the region to study a variable. Following scooping experiments (which also
must be well planned), designed experiments offer the fastest way to explore complex,
interactive, multivariable space.
Q: Chyi-gang Huang (Taiwan)
Do you have any theory (hypothesis) speculation behind the synergism of alkali
metals in selective oxidation ?
A: M. M. Bhasin
I do not have a concrete theory to offer at this time. A possible explanation may be
that different ion size alkali have different propensity to reduce ethylene oxide burning
on the catalyst-thereby leading to synergism. In this context, I refer you to the works of
R. M. Lambert and C. T. Cambell.
Q: D. D. Suresh (USA)
Have you added I more alkali element to the two alkalis already present ? How
complex does the system get '?
A: M. M.Bhasin
Yes,we have. Synergism with one such multicomponent system is given in Table IV
of the paper. The complexity with a multicomponent system does increase, but once
again, Designed Experiment approach is invaluable in reducing the number of
experiments needed to model (map out) the space.