Mechanism o f Activation at Catalytic Surfaces.

77
and XI exhibit the variation of u and v along Oy and 0 The feature of the
motion in this plane is the steadily increasing velocity tangential to the sphere,

J 2 3 4
V a l u e s of 7ya

for the magnitudes of v in diagram IX and in diagram X increase steadily

with the time for r — a. The diagrams VIII and IX should be compared with
those giving the radial and transverse components of velocity in the equatorial
plane in the case of a spherical source in a rotating liquid.

The Mechanism o f Activation at Catalytic Surfaces.

By H ugh S. T aylo r , Professor of Physical Chemistry, Princeton
University, N.J., U.S.A.

(Communicated by Dr. E. F. Armstrong, F.R.S.—Received July 6, 1926.)

A considerable body of experimental evidence* is now available with reference

to the properties of the catalytic surface, showing that such a surface is not
uniformly active and that chemical reactions occur over portions of the surface
which may vary in extent from a very small fraction of the surface to the whole
area. A theory which takes account of these facts was recently put forward
by the writer and has received a very favourable acceptance by workers in this
field.f For reactions in which only a relatively small fraction of the surface
is catalytically active, the activity is associated with occasional groups of atoms
fixed in metastable positions associated with high energy and chemical un­
saturation relative to the atoms in the regular lattice of the catalyst granule.
This concept, however, of the catalyst surface, satisfactory though it be, does
not answer what, in the last analysis, must be the central problem of the study
* 3rd and 4th Reports, Committee on Contact Catalysis, *J. Phys. Chem.,’ vol. 28, p. 898
(1924); vol. 30, p. 145 (1926); Annual Reports of the London Chemical Society, vol. 22,
p. 17 (1926); Constable, ‘ Roy. Soc. Proc.,’ A, vol. 110, p. 283 (1926).
t ‘ Roy. Soc. Proc.,’ A, vol. 108, p. 105 (1925), and developed in detail in 4th Report,
l oc. cit.
 78 H. S. Taylor.
of contact catalysis, namely, the nature of the activation induced in the
reactants, by association with these activating centres of the surface. It is this
problem which has called forth, in the recent literature, vague and ill-defined
possibilities which lack any experimental foundation. A series of recent
experimental investigations, carried out in the Princeton Laboratories, provide,
however, a very concrete picture as to the nature of the activation process in
several different examples. It is the object of the present communication
to summarise these researches and point out the conclusions to which they
logically lead.
Hinshelwood and Pritchard,* in considering this problem, believed that
adsorption of the reactant might increase as often as it decreased the stability
of the adsorbed molecule. They see in excessive momentum of parts of the
adsorbed molecule a possible cause of chemical change. Norrishf writes of a
change of configuration or distortion of the molecule consequent on an associa­
tion with a polar catalyst. His observation that such a change must occur
with the absorption of energy needs further qualification when it is remembered
that the absorption process is often accompanied by the liberation of large
amounts of heat energy. BodensteinJ refers to the deformation of adsorbed
molecules. Rideal§ uses the same expression in a recent series of lectures.
Constablej| is more specific. He speaks of a distortion of the OH group in an
adsorbed alcohol molecule and of a straining of the adsorbed molecule by the
electrostatic and electromagnetic fields of force of the atoms which constitute
the centres of activity of the surface. He believes that there must be a limit
to the distortion of the adsorbed molecule produced by the centres of activity
of a particular substance, though he is not specific as to what that limit may be.
The problem, as studied experimentally at Princeton, has been restricted
as far as possible to the simplest possible conditions, and attention has been
centred on the nature of the activation of simple diatomic molecules at catalytic
surfaces containing only one atomic species. Thus far, metallic catalysts and
hydrogen and nitrogen gases have been studied. I t is apparent that, with a
diatomic gas, there are two possible limiting states in the deformation postulated
by the above authors. The deformation may result in the production of the
atomic species or it may be still more drastic and result in the production of the *§

* ‘ J. Chem. Soe.,’ vol. 123, p. 2725 (1923).

t Ibid., vol. 123, p. 3006 (1923).
t “ Annalen,’ vol. 440, p. 177 (1924).
§ See ‘ Nature,’ vol. 117, p. 626 (1926).
|| ‘ Roy. Soc. Proc.,’ A, vol. 108, p. 355 (1925).
 Mechanism of Activation at Catalytic Surfaces. 79
ions of the atoms. This latter alternative has been frequently postulated,
especially in the case of hydrogen. The most recent attempts to interpret
catalytic hydrogenation in terms of the presence of hydrogen ions are due to
Bennewitz and Gunther* and to Schmidt.| The experimental evidence is
meagre. The theoretical evidence involves a consideration of the metal as a
medium of high dielectric constant and therefore capable of effecting the ionisa­
tion of the dissolved gas atom. The experimental evidence of the properties
of a contact catalyst indicate that reaction is due, not to dissolved gas, but to
gas adsorbed on the catalyst surface. The writer is of the opinion that the
application of the concept of a dielectric to an adsorbed atom involves an assump­
tion which requires careful consideration before it receives unqualified accept­
ance. In the meantime, the evidence in support of the existence of atoms at
the surface is at hand and increasing in quantity. The evidence available
concerning the state of adsorbed caesium on tungsten filaments indicates that,
even with atoms of such low ionisation potentials as caesium atoms, consider­
able fractions of the surface may be covered with atoms. With hydrogen,
whose ionisation potentials are as high as 13*4 volts for the atom and 16 volts
for the molecule, it seems tolerably certain that the great majority of the adsorbed
gas will be present in the non-ionised condition.
Gaugerf attempted an analysis of the state of hydrogen adsorbed on nickel
by a study of the critical potentials of the gas. He reported several such
potentials, which he associated with the first members of the Lyman series and
with the ionisation of atomic hydrogen. The method employed by Gauger
was modified by Wolfenden§ to give a somewhat less equivocal set of experi­
mental results with hydrogen gas on both nickel and copper, measuring
ionisation currents rather than radiation effects. Wolfenden concluded that
substantial quantities of atomic hydrogen are present at the metal surface even
with the plain metal as well as with a catalytic surface. His evidence for this
consisted in a marked “ break ” in the curves (obtained by plotting E /I against
the voltage of the ionising electrons) at a voltage of 13-4 volts. This break
he associated with the known ionisation potential of hydrogen atoms at this
voltage. The concordance between the ionisation potential of adsorbed
hydrogen atoms and of the free gaseous atomic species is a phenomenon not
yet sufficiently elucidated and requiring further experimental examination.*§

* ‘ Z. Phys. Chem.,’ vol. I l l , p. 257 (1924).

f Ibid.,vol. 118, p. 193 (1925).
X ‘ J. Am. Chem. Soc.,’ vol. 46, p. 674 (1924).
§ ‘ Roy. Soc. Proc.,’ A, vol. 110, p. 464 (1926).
 80 H. S. Taylor.
Wolfenden also found a break in bis curves at a potential of 11*4 volts, for the
existence of which he could offer no completely satisfactory explanation.
Some more recent data, now secured, enable us to offer such an explanation.
The experiments of Taylor and Marshall on the reactivity of hydrogen atoms
produced by collision of hydrogen molecules with excited mercury atoms*
served to show that activation at a surface producing atoms was entirely
adequate to secure rapid combination with such gases as ethylene, carbon
monoxide and oxygen. This evidence would serve to show that the extra
activation involved in the production of ions was unnecessary. This con­
clusion is further confirmed by Bonhoeffer’s results! with hydrogen atoms
produced by Wood’s method, the passage of a high-tension discharge through
molecular hydrogen. Both these sets of experiments are also concordant in
that they indicate that hydrogen atoms do not react with molecular nitrogen to
produce ammonia. The ease with which nitrogen and hydrogen react on an
iron catalyst suggested, therefore, that nitrogen must undergo some type of
activation at an iron surface. This activation has now been investigated, at
my request, by Dr. G. Kistiakowski in Princeton, using the method developed
by Wolfenden in the study of hydrogen on nickel and copper.
Kistiakowski has improved the technique of Wolfenden by an altered
arrangement of grid potentials enabling a sharper differentiation of ion effects
from those produced by radiation from the several electrode sources. He also
obtains a more accurate elimination of the several correction factors by calibra­
tion of the set-up, with argon as a reference gas. By this means he finds ionisa­
tion potentials in presence of catalyst and nitrogen at corrected voltages of
11*1, 12*9, 16-0 and 17*05 volts respectively. Those at 12*9 and 16-0 volts
are regarded as the potentials of adsorbed atomic and of molecular hydrogen,
from which, it is apparently not possible to free iron, even by out-gassing at
high temperatures. The value of 17*05 volts is clearly that of molecular
nitrogen. The value of 11 •1 volts is to be interpreted as the ionisation potential
of nitrgoen atoms adsorbed on iron. This break at 11*1 volts is quite sharp
and definite when iron is present either as a catalytic preparation or as ordinary
metal. It is quite absent when the target is not present in the system, or even
when a target poisoned with oxygen is employed. That it is definitely attribut­
able to nitrogen may be deduced from the following critical experiment of Kis­
tiakowski :—A target of pure plain iron, after out-gassing at 400° 0. for 10 hours
at a pressure less than 10~5 mm., showed voltage breaks at 11*1 and 13 volts
* ‘ J. Phys. Chem.,’ vol. 29, p. 1140 (1925).
t c Z. Elektrochem.,’ vol. 31, p. 521 (1925).
 Mechanism o f Activation at Catalytic Surfaces. 81
respectively, corresponding presumably to nitrogen and hydrogen atoms.
The target was then bombarded by high-speed electrons and heated thus to
incipient melting with continued evacuation. Examination then showed
no break at 11 •1 volts but a break still clearly defined at 13 volts. After 1 mm.
of highly purified nitrogen had been admitted to the system for five minutes and
the residual gas then immediately removed by evacuation, the break at 11*1
volts was again registered very much more sharply than initially. It is there­
fore evident that nitrogen can only be eliminated from the surface by incipient
fusion, and that, even under such conditions, hydrogen still remains present,
probably being released gradually from solution in the metal or by decomposi­
tion of the cement used in the glass joints. The evidence, however, is now
definite and unequivocal that nitrogen is present on iron surfaces in an activated
state, probably as atoms, and from the observed break in Wolfenden’s experi­
ments with nickel and copper at 11-4 volts the presence of nitrogen atoms
on his metal targets can also be deduced. This conclusion has been extended
by Kistiakowski to platinum surfaces.
An entirely independent set of investigations may now be considered in support
of the thesis that atomic hydrogen is present on the surface of an active hydro­
genation catalyst. The work of Beebe and Taylor* on the heat of adsorption
of hydrogen on nickel has now been extended by Frylingf to both promoted
and unpromoted nickel catalysts with varying amounts of adsorbed hydrogen
up to saturation capacity at one atmosphere gas pressure. The data of Russell
and Taylor J on the relative influence of promoters on the adsorptive capacity
and the reactivity of nickel indicated that the function of the promoter in this
case lay rather in a qualitative improvement of the nickel surface than in a
quantitative extension of the surface. The experiments of Fryling now indicate
the nature of this improvement of the surface. With unpromoted nickel
catalysts the heat of adsorption (Qv) per mol. of hydrogen adsorbed has a value
around 30 kg.-cals, for the first cubic centimetre of adsorbed gas and then
shows steadily diminishing values for successive adsorbed amounts up to satura­
tion capacity of the sample, at which point an adsorption of as much as 20 c.c.
may show an integral heat of adsorption of about 15 kg.-cals.per mol. of adsorbed
gas. With a nickel catalyst promoted with 13-5 per cent, thoria, on the other
hand, the initial heat of adsorption is low, increases with additional adsorp­
tion to a maximum, and then follows closely the falling values normal with
* ‘ J. Am. Chem. Soc.,’ vol. 46, p. 43 (1924).
t ‘ J. Phys. Chem.,’ vol. 30 (June, 1926).
J ‘ J. Phys. Chem.,’ vol. 29, p. 1235 (1925).
VOL. CXIII.— A. a
 82 H. S. Taylor.
unpromoted nickel. These facts are evident from the following sets of data
taken from Fryling’s work:—
V

Unpromoted nickel. Promoted nickel (13*5 per cent. Th02).

Cubic centimetres Integral per Cubic centimetres Integral Qv per

adsorbed. mol. adsorbed. adsorbed. mol. adsorbed.

1-32 25-6 0-73 13-5

2-57 21-4 1-42 20-3
3-81 19-5 213 24-4
6-00 17-7 2-81 25-8
8-66 16-3 4-83 26-1
26-5 13-6 5-69 25-4
7-70 21-9
16-7 16-3

The interpretation which the presence of the maximum in the results with
the promoted catalyst suggested to Fryling is that on the more active centres
of the nickel surface the adsorption of molecular hydrogen is accompanied
by dissociation into atoms ; this endothermic reaction, involving the dissocia­
tion of hydrogen, is superposed upon the normal and exothermic adsorption
of the molecular species. In the example cited, it is evident that, of the first
4*83 c.c. of hydrogen adsorbed, some fraction of the gas suffers such an endo­
thermic change. Beyond this maximum at 4-83 c.c., the curve is similar to
that with straight nickel. From the data of Russell and Taylor* it can be
computed, on the assumption that catalytic activity in their experiments is
measured by extent of “ hydrogen dissociating surface,” that there should be
a similar maximum on the straight nickel catalyst at about 0-55 c.c. adsorbed
gas. The. experimental method does not permit accurate measurement of
thermal magnitudes at such small adsorptions; but the validity of this con­
clusion is confirmed by Fryling’s observation that a 6 per cent. ThOa promoted
catalyst showed a similar maximum at 2-45 c.c. adsorbed gas, and a 4 per­
cent. Ce20 3 promoted catalyst at 1 -83 c.c. adsorbed gas.
The values obtained by extrapolation to zero adsorption of the two branches
of the curve obtained on plotting the differential heats of adsorption, i.e.,
the heat liberated per unit of gas adsorbed, are of interest, since they indicate,
with the approximation which is inevitable in extrapolations, the values for the
heat of adsorption of hydrogen molecules on the most active surface and the
net heat accompanying such adsorption when dissociation simultaneously
* Loc.
 Mechanism o f Activation at Catalytic Surfaces. 83
occurs. Fryling’s results for these are about 50 kg.-cals. for the former and a
value of less than 10 kg.-cals. for the process involving dissociation. From one
point of view this evidence may be taken as indicating that the change in the
hydrogen molecule is not more drastic than a dissociation into atoms. Polanyi*
has already pointed out that the heat of dissociation is lowered by adsorption.
The extrapolated value for the process assumed to involve dissociation is
of interest in that it indicates a value to be assigned to the heat of adsorption
of hydrogen atoms on the most active centres of the nickel. Since this extra­
polated value is less than 10 kg.-cals., it is evident that the heat of adsorption
of hydrogen atoms on such centres cannot be more than 10 kg.-cals. different
from the heat of dissociation of molecular hydrogen. Assumingf this latter
to be 97 kg.-cals., we may assess the heat of adsorption of atomic hydrogen on
active nickel centres at (97 + 10)/2, or approximately 53 kg.-cals. per gram
atom of hydrogen. Some unpublished data on heats of adsorption on platinum
indicate that a similar value holds for hydrogen atoms on platinum.
The utility which attaches to such data may be illustrated by reference to the
catalytic decomposition of hydrogen iodide and of ammonia at platinum surfaces.
These reactions have been studied by Hinshelwood and Burk,± and shown to
be unimolecular with respect to hydrogen iodide and to ammonia.
In the gas phase, the unimolecular decomposition of hydrogen iodide involves
the absorption of 68 kg.-cals. of energy per gram molecule of hydrogen iodide
decomposed, as is evident from the following data :—
1/2 Ho = H — 48-5 kg.-cals.
1/2 I 2 = I — 18 kg.-cals.
HI = 1/2 H 2 4- 1/212 — 1-5 kg.-cals.,
whence, by summation,
HI = H + 1 — 68 kg.-cals.
As the corresponding bimolecular decomposition of the iodide only involves
3 kg.-cals.,
2HI = H 2 4- I 2 — 3 kg.-cals.,
and its critical increment as measured from the temperature coefficient is only
44,000 calories per two molecules, it is not surprising that the gaseous thermal
decomposition is bimolecular.

* ‘ Z. Elektrochem.,’ vol. 27, p. 142 (1921).

t Wohl, ‘ Z. Elektrochem.,’ vol. 30, p. 49 (1924); Langmuir, ‘ Gen. Elec. Rev.,’ vol. 29,
p. 153 (1926).
t ‘ J- Chem. Soc.,’ vol. 127, p. 2896 (1925); vol. 127, p. 1105 (1925).
84 H. S. Taylor.
For purposes of comparison, we may now consider one possible mechanism
of hydrogen iodide decomposition at a platinum surface, concentrating our
attention on a particular platinum atom in the surface such that the heat of
adsorption of a hydrogen atom on that particular platinum atom is 53 kg.-cals,
per mol. of such atoms. We shall assume that the first step involved is an
adsorption of hydrogen iodide on the platinum and its decomposition to give
a free iodine atom and an adsorbed hydrogen. The net thermal change for
such a process is given by the equation
P t + HI = P t (H) + I — 15 kg.-cals. (a)
68 53
We shall next assume that the adsorbed hydrogen atom is struck by a hydrogen
iodide molecule, whereupon the following reactions occur :—
P t (H) + III = Pt(H 2) + I + x kg.-cals. (b)
P t (H2) = Pt + H* + y kg.-cals. (c)
The values of x and y are readily obtained when it is remembered that
I -f-1 = I 2 -f- 36 kg.-cals. (d)
and H 2 -f-12 = 2HI -)- 3 kg.-cals. (e)
For the summation of (a) -}- (b) -j- (c) -j- ( is equal to (e), whence
x+
or x-j- —
Now, y is the heat of desorption of hydrogen molecules from platinum and has
given measured values between —10 and —25 kg.-cals. It follows, therefore,
that the corresponding values of * must lie between —8 and + 7 kg.-cals. It is
very evident, therefore, that, by the mechanism postulated, the unimolecular
decomposition of hydrogen iodide has been changed from a single stage requiring
68 kg.-cals. to a series of stages no one of which requires more than 25 kg.-cals.,
so that the energy of activation has probably been correspondingly reduced
and the catalytic acceleration can very readily be understood. Additional
data as to heats of adsorption would permit the analysis of alternative possible
modes of decomposition at a platinum surface, which might, for example,
involve attachment to the platinum by iodine rather than by hydrogen.
The decomposition of ammonia at a platinum surface shows certain diver­
gencies from the case of hydrogen iodide just considered . The gaseous thermal
decomposition of ammonia has not been measured. It probably involves very
considerable energy increments, as is evident from the following thermal data
 Mechanism of Activation at Catalytic Surfaces. 85
deduced on the assumption that the dissociation of nitrogen into atoms requires
approximately 11*4 volts.*
N + N = N2 + 263-0 kg.-cals., (a)
6H = 3H2 + 292-5 kg.-cals., (6)
N2 + 3H2 — 2NH3 + 22-0 kg.-cals., (c)
whence
2N + 6H = 2NH3 + 577-5 kg.-cals., (d)
from which the heat of breaking of a N-H linkage in ammonia may be assessed
at (577 -5)/6 = 96-2 kg.-cals. By analogy with the hydrogen iodide dis­
sociation into atoms, it is probable that this reaction also never occurs uni-
molecularly in the gas phase, but probably bimolecularly by reversal of reaction
(c) above.
The unimolecular decomposition found by Hinshelwood and Burk at a
platinum surface may now be studied with the aid of the data already utilised
in the preceding case. At a platinum atom, on which the heat of adsorption
of a hydrogen atom is 53 kg.-cals., the breaking of a N-H linkage involves
P t + N-H = P t (H) + N — 43 kg.-cals. (e)
96 53
The critical increment has therefore presumably been approximately halved.
By this process the active platinum atoms become covered with adsorbed
hydrogen atoms. In the hydrogen iodide example just considered these atoms
were assumed to be removed by reaction with colliding hydrogen iodide mole­
cules, and the thermal data indicated this as quite possible. In the present
case, removal of hydrogen atoms with colliding ammonia is not possible. This
conclusion can be reached on the basis of experimental evidence, since Bon-
hoeffer and Boehmf have shown that no reaction occurs between hydrogen
atoms and ammonia. It is evident, therefore, that reaction between ammonia
and the Pt-TI grouping, involving, as this latter does, a heat of formation of
53 kg.-cals., will be still less likely to occur. This means that removal of hydrogen
atoms from platinum by evaporation or by reaction with adjacent atoms only
is possible. This gives a thermochemical reason for the observation of Hinshel­
wood and Burk that, in the ammonia decomposition, the surface was 'poisoned
by hydrogen atoms, and agrees with their conclusion that the temperature
coefficient of the reaction is partly determined by the freeing of the surface from

* Birge and Sponer, ‘ Proc. Am. Phys. Soc.’ ; ‘ Phys.’ Rev.,’ vol. 27 (1926).
f ‘ Z. physik. Chem.,’ vol. 119, p. 390 (1926).
86 Mechanism of Activation at Catalytic Surfaces.
hydrogen atoms as the temperature increases. On the assumed basis the
evaporation of the atoms requires an absorption of 53 kg.-cals.
This case of ammonia thus represents an example in which the apparent heat
of activation differs considerably from the true heat of activation. As Hinshel-
wood* has pointed out, these two magnitudes are related to one another by the
equation
Eapp< = ®true + ^ — \

where X' is the heat of desorption of the products and X is the heat of adsorp­
tion of the reactants. On the above basis Etrue cannot be less than the 43
kg.-cals. of equation (e) and the value of X' not less than 53 kg.-cals. assumed
for the heat of desorption of hydrogen atoms, values in accord with the high
value of 140 kg.-cals. found by Hinshelwood and Burk for the apparent heat of
activation.
It may also be pointed out that Hinshelwood’s belief that the X values will
in general be small as compared with the heat of activation, Etrue, is not sub­
stantiated by our experimental data. The values already cited in this article
show this, and they can be supplemented by data as yet unpublished, in which
the heats of adsorption of ammonia on several metals are greater than 10 kg.-cals.
and the heat of adsorption of carbon monoxide on copper may be as high as
30 kg.-cals.
Summary.
A review has been given of recent experimental work which indicates that
the activation of diatomic molecules at metal catalyst surfaces consists in the
production of adsorbed atoms. The evidence is concerned with both ionisation
potentials and heats of adsorption.
The data on heats of adsorption have been utilised to present a satisfactory
explanation of the decomposition of hydrogen iodide and of ammonia at
platinum surfaces and the reasons for the acceleration of velocity observed.

* ‘ Kinetics of Chemical Change,’ pp. 176-178, Oxford University Press (1926).