Adsorption of oxygen on the {111}, {100} and {110} surfaces of clean nickel

SURFACE

SCIENCE 1 (1964) 319-348; 0 North-Holland

ADSORPTION

Publishing Co., Amsterdam

OF OXYGEN ON THE (1111, (100)

AND (110) SURFACES OF CLEAN NICKEL

A. U. MAC RAE Bell Telephone Laboratories,

Murray

Hill, New Jersey

Received 4 April 1964 A study of the adsorption of oxygen on the clean (11X), (001) and (110) surfaces of nickel, using low energy electron diffraction techniques, reveals that several structures form on each of these surfaces before a monolayer oxygen coverage is attained. These structures are composed of both oxygen and nickel atoms in the topmost layer. Each structure is characterized by a sticking probability, work function change, oxygen coverage, and degradation temperature. The oxygen is adsorbed via a different mechanism on each of these surfaces. An order-disorder type of transition is observed when the temperature of a (111) oriented crystal, containing a quarter of a monolayer of oxygen, is increased to 160 “C. The close similarity of the structures to three-dimensional nickel oxide, and the magnitude of the work function changes, indicates that the surface oxygen-nickel bond is quite ionic in nature.

1. Introduction The chemisorption of gases on solids has been studied extensively for many yearsl). Much of the work in this field has been stimulated by an attempt to understand the practical aspects of catalysis, corrosion, oxidation and recently the miniaturization of electronic components. Yet many questions about the basic nature of the process of chemisorption remain unanswered. Chief among these are the nature of the bond that exists between the adsorbed species and the surface material and the actual mechanism that leads to the adsorption process. The adsorption of oxygen on nickel is an example of a widely studied system that is not thoroughly understood. This lack of understanding in many instances has been complicated by experimental difficulties. Out of necessity many of the classical surface experiments have been performed on evaporated films and finely divided powders having large surface areas. These large surface areas are often needed to attain the degree of sensitivity that is required for the surface measurements, but the lack of ultrahigh vacuum techniques and the innumerable crystallographic surfaces of these samples has made the interpretation of the results extremely difficult in many instances. Many of the investigations of this system have been concerned with the 319

320

A.U.MAC

RAE

determination of the nature of the oxygen-nickel bond. Even though the compound NiO is highly ionic, it is nevertheless thought by many authors that the surface bond due to the adsorbed oxygen is predominantly covalents-4). This conclusion has been reached mainly on the basis of a comparison between the calculated and observed values of the change in the work function of nickei due to the adsorption of oxygen. As shown in table 1, the work function increases during the adsorption process. This increase is consistent with the formation of a dipole layer that involves the transfer of electrons from the nickel to the adsorbed oxygen atom@). Since the observed work function change is approximately 30% of the calculated vatue based on a completely ionic bond, it is often assumed that this NiO surface bond is 30% ionic. The results of the present investigation indicate that the model used in making these calculations is not correct and that fair agreement between the calculated and observed work function change can be obtained by assuming an ionic bond. TABLET

Measured values, listed in the literature, for the increase in the work function and the heat of adsorption due to the adsorption of oxygen on nickel Change in Work Function in VoIts -____--__ 1.6 (a) 1.4 (b) 0.55 (c)

Heat of Adsorption in k&/mole -.--_.... 107 (d) 125 (e) 100 (f) 150 (g)

(a) J. C. P. Mignolet, Discuss. Faraday Sot. 8 (1950) 105. (b) (c) (d) (e) (f) (g)

R. C. L. Bosworth, Trans. Faraday Sot. 35 (1939) 397. I. Ogawa, T. Doke and I. Nakada, J. Appl. Phys. (Japan) 21 (1952) 223: D. Brennan, D. 0. Hayward, B. M. W. Trapnell, Proc. Roy. Sot. A256 (1960) 81. 0. Beeck, Advances in Catalysis 2 (1950) 151. R. M. Dell, D. F. Klemperer and F. S. Stone, J. Phys. Chem. 60 (1956) 1586. D. F. Klemperer and F. S. Stone, Proc. Roy. Sot. A243 (1959)

Calorimetrically measured heats of adsorption, which are listed in table 1 also are not in agreement with the calculated values436). This is not surprising since these calculations rely on the estimated ionicity of the bond, which is apparently too low. The heat of formation of nickel oxide’) is 114 kcal/mole of O,, which is very close to the average of the measured values of the heat of adsorption given in table I. The similarity of these two values appears to be additional evidence to support the hypothesis that the adsorbed oxygennickel bond is similar to the ionic bond in NiO.

ADSORPTION

There is the possibility

321

OF OXYGEN

that the heat released during the adsorption

process

increases the temperature of the thin film or powder samples to the point where the formation of NiO is enhanced. Under these circumstances, it would be the properties of NiO that were being measured. Changes in the surface area of nickel films due to sintering during the adsorption process appear to bear this outs). Changes in the K X-ray absorption edge when oxygen is adsorbed on 30 8, nickel particles also indicate that the nickel-oxygen bond is ionicg). The significance of a series of publications on the change in the magnetization of similar nickel particles is open to question at the present time, since both increases in the magnetizationlo), which were interpreted as being due to the transfer of electrons from the metal to the adsorbed oxygen producing an 11, is), which were interpreted ionic bond, and decreases in the magnetization as being due to a covalent bond, have been observedls). Increases in the resistance of evaporated thin films due to adsorbed oxygen have also been observed and were interpreted in terms of a transfer of electrons from nickel to oxygen atoms14* 15). The usefulness of single crystals for surface studies has been recognized for many yearsis). However, with the exception of some recent field electron emission 1791s) and electron diffraction work 19-25), polycrystalline materials have been used in the investigation of the chemisorption of oxygen on nickel. The single crystal low energy electron diffraction studies have been confined mainly to the (100) and (110) surfaces. These results indicate that the adsorption process is not simple, i.e. the oxygen atoms are not adsorbed on top of the surface atoms, but form a surface layer that consists of both oxygen and nickel atoms. In addition, these results have indicated that several stable arrangements of oxygen atoms are possible on each of these surfaces. The purpose of this investigation is to use low energy electron diffraction, in conjunction with ultrahigh vacuum techniques, as a tool to study the adsorption of oxygen on the (11 l), (001) and (110) surfaces of nickel. These surfaces correspond to the three most densely packed planes of nickel. Most of the observations and conclusions concerning the (001) and (1 IO) surfaces are in agreement with previous low energy electron diffraction studies. Results were obtained in this investigation that provide additional detail on the mechanism of the chemisorption process and the structures formed on the surfaces. The results on the (111) surface are predominantly new. 2. Experimental The diffraction

tube and the ultrahigh

procedure vacuum

system were used in previous

322

A. U. MAC

RAE

investigations of nickel surfaces 22ps5). The tube is of the postacceleration display type and has been described in detail elsewhere25~26). Following conventional ultrahigh vacuum processing techniques, pressures of 2 x 10-r’ Torr were obtained. Oxygen, obtained from either a silver leak tube or a commercially available high purity oxygen bottle, was admitted to the system through a bakeable metal valve and allowed to interact with the nickel at an initial pressure of N 3 x lo-’ Torr. The formation of CO, by the interaction of the oxygen with carbon on the commonly used tungsten iongauge filament 27), was minimized by using low temperature (700 to 9OOC) filaments in the ionization gauge. The samples were cut from high purity single crystals and the surface was oriented to within two to four minutes of arc to the desired crystallographic plane. Following the completion of a mechanical polish with successively finer abrasives, the crystals were electropolished with a solution containing 3 % glacial acetic and 1 % perchloric acid. Sufficient metal was removed from the surface to eliminate the damaged layers that were produced by the mechanical polish. The temperature of the crystals was measured with Pt-Pt, Rh thermocouples welded to the back of the crystals. Carbon free tungsten leads were used to mount the crystals to their supports. The size of the leads was such that the crystal and the leads near the crystal attained a uniform temperature during the high temperature heating. The intensity of individual diffraction spots was measured with a photometer having a to field of view. It was not possible to clean the surface of any of the nickel crystals by heating them close to the melting point (1453°C) in an ultrahigh vacuum. This heating always resulted in the appearance of diffraction patterns that were attributed to carbon on the surface ss,s5). The presence of carbon on a heated nickel surface has also been observed in field emission studiesl7). The actual clean surfaces, identified by the diffraction patterns, were obtained by one of two techniques. The first involved heating the crystal in oxygen and hydrogen229 2s) and the second, which was more reliable, consisted of ion bombardmentsg) with 150 volt argon ions, followed by an anneal at about 700°C. In some instances it was possible to remove the adsorbed oxygen from the surface by heating the crystals to N 800°C for 30 set in ultrahigh vacuum. This removal was due to the diffusion of oxygen into the bulk of the crystal and not to the evaporation of the adsorbed oxygen. After several such treatments, the solubility of oxygen in the crystal was presumably attained, and it was no longer possible to remove the oxygen by heating alone. This oxygen could be leached out of the crystal by cycling the temperature between 200” and 1000°C in the presence of hydrogen. Oxygen adsorbed on

ADSORPTIONOF OXYGEN

323

the crystal subsequent to this procedure couId then be removed by heating until the solubility limit was again reached. 3. Diffraction patterns from clean surfaces The three most tightly packed surfaces of nickel, the (11 I), the (001) and the (IlO), represent an interesting system of increasing complexity. The surface atom density decreases in this order (see table 2) and atoms on these surfaces have, respectively, 9, 8 and 7 nearest neighbors, instead of the 12 for atoms in the bulk of a face centered metal. Photographs of marble models of these surfaces are shown in fig. 1, with the unit mesh of each of the surfaces outlined. These unit meshes, a 120” rhombus, a square and a rectangle, provide bases for the indexing of the diffraction patterns with two-dimensional Miller indices30). Diffraction patterns from clean crystals yielded only those TABLE 2 Nickel

surface characteristics.

surface atoms/cm2

I I /

No. of

(111)

1.8 x lOI

(OO1)

1.6x lOI5

(110)

1.1 x 1015

h and

k are the two-dimensional surface rows

Row spacings,

Miller

Unit Mesh

d&k

(~)*~0~(~0 + hk + kz)' no/(2h0 + 2k9" ao/(2h2 + kg)*

indices of the

dimensions

'

ao/2/2,120"

a0/2/2

X

aoid2

X aold2

a0/2/2X a0

spots that one would expect from these meshes. The position of each of the spots in the patterns satisfied the plane grating formula (A.= drtksin y) with the values of dhk that are listed in table 2. This indicates that the nickel atoms in the topmost Iayer of these three clean surfaces have exactly the same arrangement as atoms in similar planes in the bulk of the crystal. The elemental semiconductors, germanium and silicon, in contrast, have surface atom arrangements that are quite unlike those found in their bulks1732). New spots appear in the diffraction patterns when oxygen is adsorbed on these surfaces of nickel. These spots can be labeled by fractional Miller indices and are due to the formation of surface structures containing the adsorbed oxygen. These structures have dimensions that are integral multiples of the dimensions of the substrate unit meshes shown in fig. I. The new structures are designated by their dimensions relative to those of the unit meshes, which are listed in table 2. For example a Ni (110) (2 x 1) structure has dimensions 2(a,/,/2) in the [lo] direction and 1(go) in the [Ol] direction. Long exposures of the crystal to oxygen cause new spots to

324

A. U. MAC

RAE

appear in the patterns that cannot be labeled by simple fractional indices and are due to the formation of NiO on the surface. This oxide is not the subject of this paper since it does not come under the classification of adsorbed oxygen in the sense used here. Additional references to the topic of the oxidation of nickel can be found elsewherels,si-25333). 4. Surface coverage Proper interpretation of the results of many surface experiments requires an accurate determination of the density of the adsorbed species. Low energy electron diffraction is useful in this respect since a solution of the surface structure reveals the surface coverage. In many cases the surface structure may be so complicated that it is not possible to determine the density of the adsorbed atoms, however. It is possible to determine the sticking probability, or at least its minimum

Fig. 1. Photographs of marble models of the (a) (11 l), (b) (001) and (c) (110) surfaces of a face centered cubic metal. The unit mesh of each of these surfaces is outlined in black. The white marbles in these and subsequent photographs of surface models represent nickel atoms.

ADSORFTTON

value,

from the diffraction

results.

325

OF OXYGEN

The number

of incident

atoms

that are

necessary to complete a structure can be determined from the exposures4). The completion of a structure is usually signified by some feature of the diffraction pattern, such as the attainment of a maximum in the intensity of a particular fractional order beam. If the adsorbed atom density can be determined from a solution of the surface structure, the sticking probability N is the surface is given by (1.4 x 10-21) NE/(&G), f or oxygen adsorption. density of the substrate atoms, which is listed in table 2 for the three principal planes of nickel, &t is the exposure in units of Torr x set and E is the fractional density of adsorbed surface atoms. It is also assumed that the incident oxygen molecules dissociate into atoms. For the purposes of this discussion the coverage is defined in terms of the fractional surface coverage. Thus an adsorbed atom density of $ N is referred to as a quarter of a monolayer in spite of the fact that the maximum density of adsorbed atoms may be less than N. Some authors define the coverage in terms of this maximum density, but this definition is not convenient since a redefinition of a monolayer coverage may be required in certain temperature intervals due to the temperature dependence of the type of surface structure formed by the adsorbed gas. 5. A Ni(ll1)

surface

5.1. STRUCWRE Admission of oxygen, at a pressure of 5 x 10e9 Torr, to a vacuum system containing a Ni (111) crystal held at room temperature, caused new spots to appear at the h/2 k/2 position in the diffraction pattern, with h, k = 0, 1, with increasing oxygen ex2, 3, . , . . These spots increased in intensity, posure, until an exposure of approximately 1 x 10m6 Torr x set was attained. At this exposure the intensity of these fractional order spots reached a maximum. A photograph of a diffraction pattern that was formed at room temperature is shown in fig. 2a. The appearance of new spots in the diffraction patterns, at these positions, implies that the chemisorbed oxygen atoms are arranged in a structure that has dimensions that are twice those of the substrate unit mesh in both the [lo] and [Ol] directions 25*35). For this reason, this structure is referred to as the Ni (111) (2 x 2) structure. A determination of the exposure necessary to complete the formation of this structure reveals that the sticking probability of oxygen on a clean Ni (111) surface varies from 0.4 to unity, the average being 0.7. This figure represents an average over a dozen similar experiments using both ion bombardment-anneal and chemical cleaning techniques on several crystals.

326

This value is obtained

A. LT. MAC

RAE

under the assumption

that this structure

accomodates

one oxygen atom per four substrate nickel atoms, i.e. a coverage of 0.25. The location of the new spots in the diffraction pattern reveals only the dimensions of the new mesh formed by the adsorbed atoms; it does not reveal the position of the atoms in the structure. This is done by analyzing the dependence of the intensity of the diffraction beams on the voltage of the incident beam 22, 31*36). Such a curve, for the 3 $ beam, is shown in fig. 3. The maxima in the intensity of the beams due to this structure do not coincide with the maxima of the integral order beams from the substrate crystal,

Fig. 2. (a) Diffraction pattern at 90 volts and (b) a photograph of a marble model of a Ni (111) surface containing a (2 x 2) structure. The 00, or specularly reflected spot is at the center of the photograph and is partly obscured by the crystal. The dark marbles in this and subsequent photographs represent lattice positions that are not occupied by nickel atoms.

suggesting that curves, such as the one shown in fig. 3 are not due to double scattering but are due to coherence between the scattered beams from the surface atoms. No effects due to the double scattering of a diffraction beam from substrate to surface oxgyen atoms have been observed at the relatively high voltage used in this analysis. The strong intensity of these fractional order beams is surprising. Oxygen (Z = 8) is expected to have a much smaller scattering cross section than nickel (Z = 28), especially at the relatively high voltages indicated in fig. 3. The maxima of the fractional order beams have an intensity that is on the average 0.2 times the intensity of the strongest beams from the substrate. This indicates that the main contributions to the intensity of the new spots is not from oxygen atoms, but from nickel atoms. This is the same interpretation that was used in the analysis of the diffraction features from the Ni (110) surface s2); namely that the pattern must be due to a surface structure that is

ADSORPTION

composed

of both oxygen and nickel atoms.

The structure

at by an analysis of the intensity vs voltage positions of the intensity maxima, assuming

that was arrived

curves is shown in fig. 2b. The this structure, for the + + beam

are indicated by the arrows in fig. 3. Maxima calculated using the relationship /I

327

OF OXYGEN

for simple structures

can be

2+ + l-4

max = (n + j.4)”+ (Z/d&

where dhkis the inter row spacing given in table 2, I is the spacing between the appropriate layers, IZ= 0, 1, 2, 3, . . . and ,U is a simple fraction whose value is dependent on the structuress). p = 3 for the 3 + beam and I = a,/J3, the undistorted spacing between layers on a (111) substrate. The positions of these arrows is based on the assumption that all the scattering is due to the

I

0

I

I 100

1

I 200

,

I 300

I

I 400

VOLTS

Fig. 3. Intensity vs voltage for the + 4 beam from the Ni (111) (2 x 2) structure, with the incident beam normal to the surface. The positions of the calculated maxima are marked by arrows.

exposed nickel atoms and that no inner potential corrections are necessary. The agreement between the observed and calculated curves for both the + 0 and 0 _t beams is not as good as for the 3 + beam. The dark balls shown in fig. 2b and subsequent figures do not represent atoms; they only signify the absence of nickel atoms. Since the contribution to the diffraction data from the oxygen atoms is negligible at these relatively high voltages, it is not possible to locate their positions with any degree of certainty. It is felt that the oxygen atoms occupy these vacant positions, however. The separation between the oxygen atoms and the first complete nickel layer is not specified by the diffraction data.

A. U. MAC

328

RAE

Following the completion of the pattern due to the Ni (111) (2 x 2) structure, the intensity of the h/2 k/2 spots gradually decreased with continuing oxygen exposure, and the intensity of the background increased. The background did not have a uniform intensity distribution, however. Certain portions of the diffraction pattern contained widely distributed diffuse intensity maxima. Coincident with the decrease in the intensity of the h/2 k/2 beams and the increase in the intensity of the background, new spots gradually appeared at the h + 3 k ) 3 positions in the diffraction pattern. A photograph of such a diffraction pattern is shown in fig. 4a. When these new beams attained their maximum intensity, which was much less than that of the beams from the (2 x 2) structure, the background intensity was usually quite high and occasionally there was evidence of very weak spots at the h/2 k/2 positions in the diffraction pattern. Apparently this structure never completely covered the surface.

Fig. 4.

(a) Diffraction

pattern at 90 volts and a (b) marble model of a Ni (111) (2/3 x 43) R 30” structure.

The presence of the spots at the h * + k + + positions in the diffraction pattern indicates that a new structure formed on the surface. This structure has unit mesh dimensions that are equal to those of the substrate unit mesh in both the [lo] and [Ol] d irections but are three times those of the substrate unit mesh in the [ 11] d’erection. Such a mesh is shown in fig. 4b. Since both the a and b dimensions of this mesh are J3 times those of the corresponding sides of the substrate unit mesh and this mesh is rotated 30” with respect to the orientation of the substrate unit mesh, this structure is referred to as the Ni (111) (43 x J3)R(30”) structure. The beams due to the (J3 x J3) R (30”) structure on the surface exhibit intensity maxima and minima as the voltage on the crystal is increased. The

AD~R~ONOFOXYGEN

329

voltages at which the 3 3 beam have intensity maxima can be calculated from eq. (1) with p = 3 and I = a,/J3. This is under the assumption that nickel atoms are absent from the topmost layer in the manner shown in fig. 4b. The calculation yields intensity maxima at 26, 62 and 116 volts. This compares with the observed positions of 22,63 and 102 volts. It appears that this structure is also composed of both nickel and oxygen atoms in the topmost layer with the oxygen atoms located on the vacant nickel positions. This structure represents an oxygen coverage of 3; that is, the density of oxygen atoms in this structure is f the density of substrate nickel atoms on the (111) surface. An additional oxygen exposure of 20 x lo-” Torr x set is needed to cause the transition from the (2 x 2) to the (J3 x ,/3)R(30”) structure. Increased oxygen exposure did not result in the formation of any new diffraction patterns characteristic of chemisorbed oxygen. This exposure resulted in a decrease in the intensity of al1 the spots and an increase in the intensity of the background. Finally, after a long exposure, no diffraction beams were discernible. Heating the crystal to approximately 100°C resulted in the restoration of the diffraction pattern. Presumably the decrease in the intensity of the spots is due to the adsorption of a second layer of oxygen in a molecular form. This oxygen is adsorbed in a random array on the surface and is an example of what is usually referred to in the literature as physisorption. After the temperature of the crystal was increased to approximately 100°C to remove the amorphously arranged oxygen, the diffraction pattern normally contained spots due to the (2 x 2) structure, weak spots due to NiO and a weak diffuse background in certain portions of the pattern, Heating to approximately 400°C usually removed everything but the spots due to the (2 x 2) structure. This temperature was, however, dependent on the state of development of the NiO. Higher temperatures were needed to eliminate NiO if these spots were strong initially. As mentioned in section 2, it was sometimes possible to obtain a pattern characteristic of a clean surface by heating the crystal to 800°C. When the bulk of the crystal was saturated with oxygen, it was not possible to eliminate the (2 x 2) pattern by heating alone, even to the melting point of nickel. 5.2.

ADSORPTION

CHARACTERISTICS

The most evident feature of the diffraction pattern during the formation of the Ni (111) (2 x 2) structure was the extremely rapid increase in the intensity of the h/2 k/2 spots just prior to the completion of this structure. The dependence of the intensity of the 0 3 beam on the oxygen exposure is shown in fig. 5. If all the incident oxygen molecules were incorporated into this

A. Il. MAC

330

structure

RAE

as atoms,

the integrated intensity would then be proportional to the square of (&t), i.e. the square of the number of adsorbed atoms. (The integrated intensity of a diffraction beam from a mosaic crystal is proportional to the square of the number of scattering centers when all the centers scatter in-phase.) The line with slope 2 in fig. 5 is just the limiting value of the intensity for a given value of (x~t). It is thus not possible for the integrated intensity to be greater than the values indicated by this line. Clearly the intensity does not follow a square law, implying that the incident atoms are not incorporated into a (2 x 2) structure surrounding a single nucleation center. It does not follow a (&z)” law either. This would occur, with a low value of the resultant sticking probability, if the sticking probability at any instant were proportional to the collision cross section or diameter of a patch of the (2 x 2) structure around a nucleation center. The final rapid increase in intensity with a value of the sticking probability close to unity (i.e. 0.75) indicates that most of the incident molecules are adsorbed on the surface, presumably as atoms, but are not adsorbed in a single structure within the coherence width of the incident beam (a few hundred Angstroms). Instead, they are adsorbed either individually or in small patches consisting of (2 x 2) structures centered at sites on the surface that are restricted to mesh points. No contribution to the intensity of the half-order beams from such a random arrangement would occur since there are as many sites that scatter out-of-phase as there are sites that scatter in-phase for the (2 x 2) structure. When the oxygencoverage becomes appreciable, adsorption at random sites is no longer possible since the adsorbed atoms are so close to each other that there is an interaction between them. Under these conditions, some atoms that had been bound at random sites will change sites to satisfy the restriction that there be a separation of two subtrate mesh sites between the oxygen atoms and that the coverage be maximized. This movement produces the final rapid increase in intensity. Since this adsorption process appears to be characterized by the absence of the growth of large patches of the (2 x 2) structure, it may be possible that the distance that an atom is capable of migrating is much shorter than the distance between the (2 x 2) patches, which are presumably centered on nucleation sites. This would then lead to adsorption at random mesh points in the initial adsorption stage. From measurements that will be described in another paragraph, it is known that surface migration is appreciable at 120°C. The dependence of the intensity of the 0 + beam at 63 volts on the oxygen exposure was then determined with the crystal at a temperature of 100°C. These results, also shown in fig. 5, are almost superimposed on the room temperature results. This coincidence of the experimental results at these two temperatures indicates that even though migration can occur, the

ADSORPTION

short

range interaction

between

331

OF OXYGEN

oxygen

atoms

is not sufficiently

strong

to

result in the formation of a single (2 x 2) structure when the migrating atom approaches another oxygen atom or a small patch. A calculation of the dependence of the intensity on the number of incident atoms was made assuming that the initial adsorption occurred at random sites on a linear chain. This was done using a table of random numbers, with STICKING

1.c

PROBABILITY

1.0

0.6

0.65

8

108-6

0.6

0.4

0.08

0.06

0.04

0.02

l--

II 13-7

&POWRE

6 (TOUR-SEC)

Fig. 5. Intensity vs oxygen exposure for the 0 4 beam from the Ni (111) (2 x 2) structure at 63 volts. Room temperature results are marked by circles and the 100 “C results by dots. The dashed curve is calculated. The line of slope two intersects unit intensity at the exposure appropriate to unit sticking probability.

each random number in sequence representing adsorption on a particular numbered site with the restriction that no two consecutive sites be occupied. When the surface coverage became appreciable it was-necessary to change the the occupancy of some of the sites to meet this restriction. The results of this calculation are also shown in fig. 5. Even though the model is probably

A. U. MACRAE

332

oversimplified, there is some semblance of agreement between the calculated and experimental results. Thus it appears that migration to nucleation centers is not an important process in the adsorption of oxygen on the (111) surface of nickel. 5.3. TEMPERATURE EFFECTS It was observed that the pattern characteristic of the (2 x 2) pattern did not appear when oxygen was adsorbed on the crystal while it was held at an elevated temperature. The pattern appeared suddenly at 160°C when the temperature of the crystal was subsequently reduced, however. Thus the oxygen was adsorbed on this Ni (111) surface while the crystal was at the elevated temperature but was not bound to the surface in a well ordered

‘_.O

.o.o -

.o i.0 -

.o -

.o-

8.0 -

1.0 -

I

50

I

100

I 150 TEMPERATURE

I 200

I 250

I

300

3

D

(“C)

Fig. 6. Intensity vs temperature for the 0 1 and 0 g beams from a Ni (111) surface containing the (2 x 2) structure at 63 V. The solid curve indicates the temperature dependence of the intensity of the 0 4 beam, corrected for the Debye-Wailer factor.

ADSORPTION

OF OXYGEN

333

structure. Reducing the temperature to below 160°C caused ordering to occur and the pattern characteristic of the (2 x 2) was then observed. The dependence of the intensity on the temperature of the crystal was determined for many of the fractional order beams over a wide range in voltage. A typical result is shown in fig. 6. The intensity vs voltage for both the 0 3 and 0 1 beams at 63 volts is shown in this figure. Identical curves were obtained when the temperature of the crystal was either increased or decreased. The gradual decrease in intensity with increasing temperature for the 0 1 beam is due to the thermal vibration of the atoms. The slope of this curve can be characterized by a Debye temperature of 320°K. This effective Debye temperature is lower than the accepted value of 390”K37) and has previously been attributed to the increased vibrational amplitude of surface atoms 38). The rapid decrease in the intensity of the 0 $ beam is due to a disruption of the (2 x 2) structure. Above 160°C the ordering forces are so weak that the oxygen does not exist on the surface in a definite structure. Presumably, above this critical temperature there is an appreciable probability for the occupation of oxygen sites by nickel atoms and nickel sites by oxygen atoms. Under these conditions the (2 x 2) structure does not exist and no diffraction beams due to this structure are observed. In some respects this behavior is similar to the order-disorder type of phase changes that are observed in alloys. In this case, however, we are dealing with a two-dimensional phenomenon, which in principle should be easier to treat mathematically than the three-dimensional problem. Even so, formidable problems are encountered in the application of thermodynamics to systems that exhibit phase changes, and approximations involving the cooperative aspects of the assembly of atoms are usually introduced into an analysisas). A Bragg-Williams type of analysis40) of the Ni (111) (2 x 2) structure yields results that are very similar to those obtained for the AB, or Cu,Au type of three-dimensional structure. If it is assumed that the interaction energy between the surface oxygen-substrate nickel and surface nickel-substrate nickel atoms is temperature independent, these terms can be neglected and the problem simplifies to one that involves the counting of nearest neighbor interactions in the plane of the surface. Each surface atom in the (2 x 2) structure has six nearest surface neighbors. An oxygen atom is surrounded by six surface nickel atoms and a nickel atom is surrounded by two oxygen atoms and four nickel atoms. In the AB, structure each atom has twelve nearest neighbors, with each A atom surrounded by twelve B atoms and each B atom surrounded by four A atoms and eight B atoms. It is thus possible to analyze the results obtained from the Ni (111) (2 x 2) structure by applying a scale factor of two to the AB, analysis.

334

A. U. MAC

RAE

The Bragg-Williams treatment yields the expression v/kT = ($s) log (1 + 16s/3(1 --s)~) w h ere v is the ordering energy and s is the long range ordering coefficient, which is defined as equal to 1 - w(O)/F(Ni). w(O) is the fractional occupancy of oxygen sites by nickel atoms and F(Ni) is the fractional number of nickel sites. Complete order corresponds to s = 1 and complete disorder to s = 0. The above expression predicts a discontinuity in s at a critical temperature. The experimentally determined intensities can be converted to values of s. The scattering amplitude of the 0 3 beam is proportional to the number of oxygen atoms on oxygen sites minus two thirds the number of oxygen atoms on nickel sites. In terms of s, the intensity is thus proportional to (5s - 1)2. If the temperature factor is also included the intensity becomes I = k exp (- 2M)(5s - 1)’ where k is a proportionality factor and A4 is the Debye factor. The intensity of the 0 + beam was corrected for the temperature factor by using the temperature dependence of the 0 1 beam given in fig. 6. The corrected curve is shown as a solid curve in this figure. This is also a measure of the temperature dependence of s2, since 5s > 1 for most of this curve. No discontinuity occurs in the intensity, as predicted by the Bragg-Williams treatment. No difference in the temperature dependence of the intensity was observed when the temperature of the crystal was changed very slowly. This indicates that the arrangement of the atoms was at equilibrium for the measurements shown in fig. 6 and a long relaxation time was not responsible for the lack of a discontinuity in the intensity. No discontinuity in s is predicted by a B-W treatment of the AB type 01 alloy. The decrease in intensity with increasing temperature is more rapid for the Ni (111) (2 x 2) structure than is predicted for the AB alloy, however. Calculated curves of this type are reproduced in ref. 41. The critical temperature for the Ni (11 I) (2 x 2) structure should occur at 0.822 v/k, in the B-W treatment, where v is the ordering energy. This figure is arrived at from the original B-W treatment by considering the fact that each atom in the (2 x 2) structure has six nearest neighbors instead ,of the twelve which is characteristic of the AB, structure. The ordering energy of this system is then 0.045 eV for the critical temperature of 160°C. Detailed analyses of the AB, structuresQ+43), which include nearest neighbor interactions, also predict a discontinuity in s at a critical temperature. Even though the Ni (I 11) (2 x 2) structure is two-dimensional in nature, it does not fall in the class of structures that can be treated exactly. Surface imperfections may be responsible for the failure to observe a discontinuity in the intensity vs temperature curves for the half order beams. Actually, it is surprising that such an abrupt change in the intensity was even observed. Normally, the results of the analyses of two-dimensional cooperative phe-

.4DSORl?TlON

OF OXYGEN

335

nomena are extremely sensitive to small disturbances such as imperfections or impurities at a surface44). 6. A Ni(OO1) surface 6.1. STRUCTURE When oxygen, at a pressure of 5 x IO- 9 Torr, is adsorbed on a clean (001) surface of nickel, the first change in the diffraction pattern that is observed is the appearance of streaks in the zones that pass through the h + 3 k + 3 positions. These streaks, which are an indication of disorder, do not appear in the zones that pass through integral order spots, h k, and are parallel to the [ho]* nad [Ok]* directions in reciprocal space. A photograph of the diffraction pattern in this stage of development is shown in fig. 7a. Increasing the oxygen exposure results in the coalescing of these streaks into spots that are located at the h/2 k/2 positions in the diffraction pattern.

Fig. 7. (a) Diffraction pattern obtained from a Ni (001) surface with low oxygen exposure at 100 volts.(b) Diffraction pattern at 67 volts and (c) marble model of the Ni (001) (2 x 2) structure.

336

A. U. MAC

RAE

Such a diffraction pattern is shown in fig. 7b. Spots located at these positions imply that the chemisorbed oxygen is bound to the surface in a structure that has dimensions that are twice those of the substrate unit mesh in both the [lo] and [Ol] directions. In this respect, there is a similarity in the first structures that are observed to form on both the Ni (111) and Ni (001) surfaces. This structure is designated as the Ni (001) p (2 x 2) structure. The diffraction pattern is the same one that Farnsworth, et al.rg) refer to as the “double spaced, face centered” or simply the “ds-fc” pattern and Germer and Hartman2i) refer to as being due to the “four-structure”. Analysis of the intensity-voltage curves for the + 0 and 3 + beams, indicates that this structure is also composed of both oxygen and nickel atoms in the topmost layer. An intensity vs voltage curve for the + 3 beam is shown in

(a)

i/Z

1/2

EiEAM,(2x2)

t

ii

0

50

100

150

200

VOLTS

1 beam from the Ni (001) (a) (2 x 2) structure Fig. 8. Intensity vs voltage for the z1 ‘5 (b) c (2 x ) structure. Calculated positions of maxima are marked by arrows.

fig. 8a. If it is assumed that the oxygen atoms have the arrangement shown in fig. 7c, intensity maxima for the fractional order beams will occur at the wavelengths given by eq. (1) with I = ao, and ,LI= +. The maxima calculated for the 3 3 beam are marked in fig. Sa by arrows. The agreement between the positions of the calculated and observed maxima indicates that the oxygen atoms probably have the arrangement shown in fig. 7c. The fact that nickel atoms are incorporated into this (2 x 2) structure was not discussed by the previous investigators of this system is,si), possibly because they did not base their structure analyses on intensity vs voltage curves. Additional exposure of this crystal to oxygen results in the gradual decrease in the intensity of the h + $ k and h k + + beams until the only

ADSORPTION

fractional indices

OF OXYGEN

337

order spots that can be seen on the screen are those with Miller h + + k + 3. A photograph

of this final diffraction

pattern

is shown

in fig. 9a. From this pattern one infers that the structure on the surface has inter-row repeat distances in the [lo] and [01] directions that are identical to those of the substrate mesh but that the repeat distance in the [ll] direction is doubled. This is characteristic of a centered two by two structure or simply c (2 x 2). It is the same pattern that Farnsworth, et a119) have called the and Germer and “single-spaced, simple square”, or “s-s, s-s” pattern Hartmanzl) have referred to as being due to the “two structure”. The dependence of the intensity of the 3 3 beam on the voltage of the incident electrons is shown in fig. 8b. The shape of this curve is quite similar to that

Fig. 9.

(a) Diffraction pattern at 110volts and (b) model of a Ni (001) c (2 x 2) structure.

of the + + beam from the preceding Ni (001) (2 x 2) surface structure. This type of agreement is obtained if the oxygen atoms have the arrangement on the surface that is shown in fig. 9b. Both the Ni (001) p (2 x 2) and Ni (001) c (2 x 2) structures have the same arrangement of atoms in the [ 1 l] direction aside from the fact that there are twice as many oxygen atoms per row in the c (2 x 2) structure. Both of these structures are also composed of nickel and oxygen atoms in the topmost layer; the p (2 x 2) structure containing a quarter of a monolayer of oxygen and the c (2 x 2) structure a half a monolayer of oxygen. Farnsworth, et a119) arrived at a similar structure for the half a monolayer of oxygen on the basis of photoelectric data.

6.2. ADSORPTION CHARACTERISTICS No attempt was made to obtain the dependence of the intensity of any of the half order spots on the oxygen exposure because of the excessive streaking

338

A. U. MAC

RAE

that accompanied the formation of the p (2 x 2) structure. The sticking probability was determined from the total oxygen exposure needed to form the (2 x 2) structure and it varied from 0.04 to almost unity, depending on the temperature of the crystal. The lower value of the sticking probability was obtained when the crystal was held at room temperature, while the value close to unity was obtained when the cyrstal was kept at 150 to 250°C during the adsorption process. The 0.04 figure is a lower limit and is indicative of the amount of oxygen incorporated into the p (2 x 2) structure. It is possible that the actual sticking probability is higher due to the adsorption of oxygen in an amorphous arrangement, which was not included in the (2 x 2) arrangement. The intensity of the patterns that were formed when the crystal was held at the elevated temperatures was always greater than the intensity of the patterns formed at room temperature. The fact that a unit sticking probability was obtained is an indication that only a quarter of a monolayer of oxygen adsorbed on the surface. The increase in the sticking probability observed when the temperature was increased may be due to an increase in the surface migration distance of an oxygen atom or molecule with increasing temperature. If the migration or diffusion distance of an atom were too low, the incident atoms would not be able to reach a nucleation center before evaporating from the surface. At the higher temperature, all the incident atoms could be adsorbed due to an increase in the migration distance, and a unit sticking probability would be obtained. A similar effect has been observed for the adsorption of oxygen on the (110) surface of tungsten 45). In the case of the Ni (001) c (2 x 2) structure, the sticking probability appears to be independent of the temperature of the crystal and is approximately 0.02, in reasonable agreement with previous results193 2r). 6.3, TE~PERA~RE EFFECTS It was possible to degrade the Ni (001) c (2 x 2) structure to the Ni (001) p (2 x 2) structure by heating the crystal to approximately 700°C. No abrupt decrease in the intensity of any of the fractional order spots at a particular temperature was observed when the temperature of crystal was increased up to this value. This lack of an intensity decrease indicates that the ordering energy of these structures is greater than that of the Ni (111) (2 x 2) structure. ‘7. A Ni(ll0) surface 7.2. STRUCTURE The adsorption of oxygen at room temperature on a clean Ni (110) surface

ADSORPTION

resulted

in the appearance

of streaks

OF OXYGEN

in the diffraction

339

pattern

that were

parallel to the [ho]* direction in reciprocal space. These streaks were located between the adjacent h k and h f 1 spots. As more oxygen was adsorbed on the surface, these streaks rapidly coalesced into spots that were located at the h f 3 k positions in the diffraction pattern. These spots then became more intense as the oxygen exposure was increased. Photographs of both a streaked and a final diffraction pattern are shown in figs. 10a and lob.

C

Fig. 10. (a) Diffraction pattern at 128 volts from a Ni (110) surface at low oxygen exposure. (b) Diffraction pattern at 70 volts and (c) model of a Ni (110) (2 x 1) structure.

The streaks that appeared in the diffraction patterns initially can be attributed to structures having large spacings in the [lo] direction. A distribution of spacings will produce many spots which will appear in the diffraction pattern as continuous streaks 22). These diffraction patterns indicate that the first structure that forms on the (110) surface has a repeat distance that is twice that of the substrate unit mesh in the [lo] direction, and equal to that of substrate unit mesh in the [01]

A. il. MAC

340

direction

as). This is referred

RAE

to as the Ni (110) (2 x 1) structure.

In previous

publications this was the (1 x 2) structure, but the designation has been changed to be consistent with accepted crystallographic nomenclature. This is the convention that the sides of the unit mesh are defined with a < b. The considerable intensity of the h + + k spots and the location of the maxima in the intensity vs voltage curves require that this structure be composed of both oxygen and nickel atoms in the topmost layer. Since this point has been discussed in detail along with the pertinent calculations and experimental results in a previous publication (ref. 22, 1962), it will not be repeated here. A photograph of the marble model of the Ni (110) (2 x 1) structure is shown in fig. 10~. Similar conclusions regarding the presence of both oxygen and nickel in the topmost layer of the Ni (110) surface have been reached on the basis of field emission studiesls,*6}. Following the completion of the (2 x 1) structure, several other structures form with additional oxygen exposure. Streaking of the fractional order spots portends the transition to a structure of higher oxygen content. Models of these structures, the (5 x 2)‘, (3 x 1) and the (5 x l), are shown in fig. 11. They represent oxygen coverage of 0.40, 0.67 and 0.80 monolayers,

b

a

Fig. 11.

Models of (a) Ni (110) (5 x 2)‘, (b) Ni (1 IO) (3 x 1) and (c) Ni (I 10) (5 x 1) structures.

ADSORPTION

respectively.

The formation

341

OF OXYGEN

of these structures

has also been discussed

in

detail (ref. 22, 1962). 7.2. ADSORPTION CHARACTERISTICS The dependence of the intensity on the oxygen exposure for the 3 0 beam from the Ni (110) (2 x 1) structure is shown in fig. 12. The maximum in the intensity of this beam is attained at an oxygen exposure that corresponds to a unit sticking probability. The intensity of this emerging beam follows a square law, which indicates that every incident oxygen atom is bound to the surface in a (2 x 1) structure and makes a contribution to the intensity of the 1.0 0.6 0.6 -

0.4-

0.2 -

: si z p z -

o.i0.08

-

o.oe-

0.04

-

0.02

-

0.01

I

10-a

I

z

I 4

III 6 EXPOSURE

Fig. 12.

Intensity

10-7 (TOW?-

I

I

2

4

Ill 6

10-e

SEC‘)

vs oxygen exposure for the S 0 beam from the Ni (110) (2 x 1) structure at 73 volts.

spot. Thus it is apparent that nucleation is probably an important mechanism in the formation of the Ni (110) (2 x 1) structure. The oxygen atoms migrate along the troughs on this surface, as evidenced by the streaking, until they finally reach a nucleation center and then are bound in this structure. Surface imperfections such as steps, dislocations or impurities are possible centers for such a process. The sticking probability decreases in a stepwise fashion as each of these structures is formed. A decrease in the sticking probability from unity to form the (2 x 1) structure, to approximately 0.2, 0.05 and 0.01 to form the (5 x 2)‘, (3 x 1) and (5 x 1) structures respectively is observed

342

A. U. MAC

RAE

In the sequence of structure changes, the (2 x 1) and the (3 x 1) structures always form, but the “five structures” are sometimes not found. This lack of reproducibility seems to indicate that the perfection of the surface has an important role in the adsorption process. These structures must result from the migration of atoms in the troughs formed by the nickel substrate. It seems probably that a very small number of impurity atoms or imperfections in the troughs can greatly affect this movement. The variation in the sticking probabilities and the occasional failure of the “five structures” to form, seem to bear this out. 7.3. TEMPERATURE EFFECTS The intensity vs temperature for the 3 0 beam from the Ni (110) (2 x 1) structure does not show a sudden sharp decrease as do corresponding curves for the fractional order beams from the Ni (111) structure (fig. 6). One draws the conclusion that the ordering energy of the structure that is initially formed on the Ni (1 IO) surface is much higher than the energy for the corresponding structure that forms on the Ni (111) surface, and that sudden “melting” of the structure on the (110) surface does not occur within the observed temperature range. Each one of these structures can in turn be degraded to the previously formed structure by heating the crystal. An increase in the temperature to appro~mately 100°C results in the removal of the molecular oxygen and the appearance of the pattern due to the (5 x 1) structure. Temperatures of approximately 250”, 400” and 800°C result in the formation of the (3 x l), (5 x 2) and (2 x 1) structures respectively. Temperatures close to the melting point of nickel (1450°C) do not result in the removal of the (2 x 1) structure from the surface unless the oxygen can diffuse into the bulk of the crystal. 8. Work function changes It is possible to measure the change in the work function of the specimen due to the adsorption of a gas by a technique that utilizes the diffraction equipmentsi). This technique is based on the fact that the angular position of a diffracted beam is very sensitive to the energy of the incident electrons when v,, the angle between the incident and diffracted beams, is approximately 90”. If the voltage between the crystal and the low work function filament is adjusted to a value such that a diffracted beam occurs at p = 90”, an increase in the work function of the crystal due to adsorption of a gas will cause the diffracted beam to disappear since there is no longer a solution of the plane grating formula. The diffracted beam will reappear if the voltage between the filament and the crystal is increas.ed. This increase

343

ADSORPTIONOFOXYGEN

in voltage is equal to the value of the change in the work function due to the adsorption process. It is necessary to rotate the crystal with respect to the incident beam to obtain a diffraction beam at the low voltages because the relatively low dhk values of nickel do not permit allowed solutions of the grating formula at low voltages for normal incidence. It was demonstrated that any contribution to the measured work function change due to the adsorption of oxygen on the electron gun filament was negligible. Good agreement was obtained between the work function measurements that were determined by either adsorbing oxygen on the surface or removing the oxygen from the surface by diffusing it into the bulk. The changes in the work function that were measured for the adsorption of oxygen on the Ni (1 lo), (001) and (110) surfaces are summarized in table 3. An increase in the work function, which corresponds to the formation of a negative charge outward, occurs on each of these surfaces. Except for the results on the (001) surface, the total changes in the work function are similar in magnitude to the changes in the work function that were measured for the adsorption of oxygen on thin films of nickel (table 1). While the changes in the work function due to the adsorption of half a monolayer of oxygen on the (001) and (110) surfaces are approximately equal, they are about one half the value obtained for only a quarter of a monolayer of oxygen on the (111) TABLE 3

The characteristics of the structures formed on nickel due to the adsorption of oxygen. The work function increases shown in the parentheses represent the total work function change due to the structure, starting from a clean surface.

Structure

Ni(lll)clean Ni(ll1) (2 x 2) Ni (111) (~‘3 x ~‘3) R(30”) Ni (001) clean Ni (001) p (2 x 2) Ni(OOl)c(2 x 2) Ni (110) clean Ni (110) (2 x Ni (110) (5 x Ni (110) (3 x Ni(110)(5 x

Fractional )xy.Covarege

Approx.Temp To Degrade to Str. Of Lower Oxygen Content

I

7-1

I

‘6

A.01

1.2 volts

i ~

400°C

.*

0.02

4

i

> m.p.* ‘.

I 0.04- 1

.“. * 9

. 1) 2)’ 1) 1)

Approx. Work Fnc. Increase Due to Oxy.

Approx Oxy. Sticking Prob

4 B

I

0.25 0.2 (45)

> m.p.* 700”

1 0.2

0.6

0.05 0.01

0.6 (I .2)

> m.p.* 800 350” 250”

..

i * This temperature

is reduced when the crystal is not saturated with oxygen.

344

A. U. MAC

RAE

surface. Thus it is apparent that a knowledge of the number of adsorbed atoms alone is not sufficient to predict a work function change. It is commonly assumed that changes in work function can be interpreted in terms of the number of adsorbed atoms but this assumption need not necessarily be true, as evidenced by the present results. Numerous caI~uIations of changes in work function due to the adsorption of different species on surfaces exist in the fiterature. These calculations rely on the Helmholtz formula; A V = 2m,u/c, where A V is the work function change, n is the number of dipoles per cm2 having moment ,!Iperpendicular to the surface and ): is the dielectric constant 5). In certain circumstances the coefficient 271is replaced by 47~47). If we assume that this surface layer has a dielectric constant of five, that there is a 45” angle between a first layer oxygen and second layer nickel atom and that 85% of the total charge transfer occurs in the topmost layer, agreement can be obtained between the calculated and observed work function change due to the adsorption of half a monolayer of doubly ionized oxygen on the Ni (001) surface. One is tempted to apply the same type of calculation to all the structures observed to form, but such calculations are highly artificial. For instance, it is extremely difficult to justify the choice of a particuIar dietectric constant for these structures. While these surface structures have many points in common with NiO, the surface structures would certaimy not be characterized by the dieIectric constant of NiO, which has been estimated to be ten@). Perhaps the most important drawback to the calculation of such surface properties is the lack of knowledge of the distortion of the charge distribution at the surface before and after the adsorption process. Understanding of chemical binding has not advanced to the point where it is possible to go beyond an ad hoc type of explanation of this charge distribution. The significant aspect of these work function vs structure measurements is that charge transfer between oxygen and nickel atoms in the topmost layer may be appreciable. Charge transfer of this type will yield a work function change that is lower than the value that would be obtained if only oxygen atoms composed the topmost layer. Thus it is not necessary to assume an unrealistic value for the covalency of the surface nickel-oxygen bond, just because there is disagreement between the work function change that is measured and the one that is calculated on the basis of an over-simplified model. The large differences in the work function changes that were observed between the (111) and the (001) and (110) surfaces may be due to the vertical positions of the oxygen atoms. The oxygen and nickel atoms that form the surface structures on the (001) and (110) surfaces may be close to being

ADSORPTION

coplanar,

345

OF OXYGEN

as they are in the corresponding

faces of NiO.

They

are

not

coplanar in the (111) face of NiO, however. If the oxygen atoms in the Ni (111) (2 x 2) structure are displaced slightly outward from the positions of the dark balls shown in fig. 2b, a relatively high change in the work function would be obtained. This is due to the effective separation of the negative charge in a plane through the oxygen atoms and the positive charge in a plane through the topmost nickel atoms. Since no direct diffraction effects were obtained from the oxygen atoms, it was not possible to ascertain the vertical positions of these atoms precisely. A slight vertical displacement of the oxygen atoms may account for the low critical temperature of the Ni (111) (2 x 2) structure. One might expect that moderate heating would introduce disorder into this structure since the nickel atoms should be able to move quite easily into the partial “hole” that exists directly under the oxygen atoms. The lack of a discontinuity in the ordering parameter may also be due to this particular type of disordering.

9. Discussion The adsorption of oxygen on the (11 l), (001) and (110) surfaces of clean nickel results in the formation of several surface structures that apparently are composed of both oxygen and nickel atoms in the topmost layer. Each of these structures has a characteristic mode of adsorption, sticking probability for oxygen, degree of coverage and degradation temperature associated with it. These points are summarized in table 3. The mode of adsorption on these clean surfaces is especially orientation dependent. While every oxygen atom that strikes the (1 IO) surface at room temperature migrates to some nucleation center to form the initial structure, this process occurs on the (001) surface only at an elevated temperature and does not occur on the (111) surface. The initial structure on the (111) surface forms appreciably only when the coverage is such that it is necessary to form this structure to accommodate additional oxygen on the surface. The initial or (2 x 2) structure on the Ni (111) surface is very interesting since it displays cooperative phenomena that are similar to the order-disorder effects that are observed in alloys of the type Cu,Au. In many aspects, oxygen adsorbed on nickel has the characteristics of a three-dimensional compound rather than a two-dimensional one that is commonly assumed for adsorbed gases. The presence of both oxygen and nickel atoms in the topmost layer suggests a comparison of these structures with the three-dimensional compound nickel oxide. The arrangement of oxygen and nickel atoms in the Ni (001) c (2 x 2) and Ni (110) (2 x 1)2s) structures is the same as that found in the (001) and (110) planes of NiO,

346

respectively.

A. U. MAC

RAE

NiO has the rock salt structure.

The corresponding

dimensions

of the surface structures are 20% larger than those of NiO, however. It has been suggested that these surface structures are due to the formation of a Ni,O type of structure on the surface35). This explanation is limited to the Ni (111) (2 x 2), Ni (001) c (2 x 2) and Ni(ll0) (2 x 1) structures and does not include the other structures, however. In addition, the actual structures appear to be confined to only one layer on the surface and not to several layers as a comparison with a three-dimensional structure would imply. That the structures due to the adsorbed oxygen are confined to the topmost layer is indicated by the structure analyses and the fact that unit sticking probabilities are observed for the formation of these structures. The unit sticking probabilities imply that all the incident oxygen atoms are accounted for by the structures shown in figures 2b, 7c, and 10~. The changes in work function that were observed for each of these structures indicate that the oxygen exists in a highly ionic state on the surface. Additional evidence from the structures seems to support this conclusion. The (J3 x ,/3) R (30”) and the c (2 x 2), which are the final structures that form on the (111) and (001) surfaces, are those structures that can accommodate the greatest amount of oxygen at lattice sites if it is assumed that the oxygen has the radius of the oxygen ion, 1.40 A. There is just not enough space on the surface to accommodate additional oxygen of this size at lattice sites. Thus these results tend to indicate that the surface nickel-oxygen bond probably has a degree if ionicity somewhere between that predicted for the Ni-0 molecule on the basis of electronegativity, 51%4g), and that of the compound NiO. Acknowledgment The author would like to thank L. H. Germer for his suggestions for improving the manuscript, J. R. Arthur for several discussions on chemisorption and C. Herring for a discussion on work function changes.

References 1) Many references to the subject of chemisorption may be found in: G. C. Bond, Catalysis by Metals (Academic Press, New York, 1962). W. E. Gardner (Editor), Chemisorprion (Butterworths, London, 1957). B. M. W. Trapnell, Chemisorption (Butterworths, London, 1955). 2) F. C. Tompkins, Le Vide 17 (1962) 72. 3) B. M. W. Trapnell, Chemisorption p. 151. 4) I. Higuchi, T. Ree, and H. Eyring, J. Am. Chem. Sot. 79 (1957) 1330. 5) I. Langmuir, J. Am. Chem. Sot. 54 (1932) 2798. 6) M. B. W. Trapnell, Chemisovption p. 150. 7) B. J. Boyle, E. G. King and K. C. Conway, J. Am. Chem. Sot. 76 (1954) 3835.

ADSORPTION

OF OXYGEN

347

8) R. M. Dell, D. F. Klemperer and F. S. Stone, J. Phys. Chem. 60 (1956) 1586. 9) P. W. Lewis, J. Phys. Chem. 64 (1960) 1103. 10) P. W. Selwood, S. Adler and T. R. Phillips, J. Am. Chem. Sot. 77 (1955) 1462. L. E. Moore and P. W. Selwood, J. Am. Chem. Sot. 78 (1956) 697. 11) J. J. Broeder, L. L. van Reijen, W. M. H. Sachtler and G. C. A. Schuit, Z. Elektrothem. 60 (1956) 838. J. J. Broeder, L. L. van Reijen and A. R. Kornswagen, J. Chem. Phys. 54 (1957) 37. 12) R. J. Leak and P. W. Selwood, J. Phys. Chem. 64 (1960) 1114. 13) P. W. Selwood, Adsorption and Collective Paramagnetism (Academic Press, New York, 1962) p. 175. 14) D. Brennan, D. 0. Hayward, B. M. W. Trapnell, Proc. Roy. Sot. A256 (1960) 81. 15) R. Suhrmann and K. Schultz, Z. Physik. Chem. (Frankfurt) 1 (1954) 69. 16) A. T. Gwathmey and A. F. Benton, J. Chem. Phys. 8 (1940) 431 and 569. A. T. Gwathmey and R. E. Cunningham, Advances in Catalysis 10 (1958) 57. 17) R. Gomer, J. Chem. Phys. 21 (1953) 293. 18) G. Ehrlich, Annals N.Y. Acad. Sci. 101 (1963) 722. 19) R. E. Schlier and H. E. Farnsworth, Advances in Catalysis 9 (1957) 434. H. E. Farnsworth and J. Tuul, J. Phys. Chem. Solids 9 (1959) 48. H. E. Farnsworth and H. H. Madden, Jr., J. Appl. Phys. 32 (1961) 1933. 20) R. L. Park and H. E. Farnsworth, Appl. Phys. Letters 3 (1963) 167. 21) L. H. Germer and C. D. Hartman, J. Appl. Phys. 31 (1960) 2085. 22) L. H. Germer, A. U. Mac Rae and C. D. Hartman, J. Appl. Phys. 32 (1961) 2432. L. H. Germer and A. U. Mac Rae, J. Appl. Phys. 33 (1962) 2923. L. H. Germer and A. U. Mac Rae, Robert A. Welch Foundation Research Bulletin 11 (1961) S-26. 23) L. H. Germer, E. J. Scheibner and C. D. Hartman, Phil. Mag. 5 (1960) 222. 24) A. U. Mac Rae, Appl. Phys. Letters 2 (1963) 88. 25) A. U. Mac Rae, Science 139 (1963) 379. 26) W. Ehrenberg, Phil. Mag. 18 (1934) 878. E. J. Scheibner, L. H. Germer and C. D. Hartman, Rev. Sci. Instr. 31(1960) 112. L. H. Germer and C. D. Hartman, Rev. Sci. Instr. 31 (1960) 784. J. J. Lander, J. Morrison and F. G. Unterwald, Rev. Sci. Instr. 33 (1962) 782. 27) J. A. Becker, E. J. Becker and R. G. Brandes, J. Appl. Phys. 32 (1961) 411. 28) L. H. Germer and A. U. Mac Rae, J. Chem. Phys. 37 (1962) 1382. 29) H. E. Farnsworth, T. H. George, and R. M. Burger, J. Appl. Phys. 26 (1955) 252. 30) E. A. Wood, J. Appl. Phys. (in press, 1964). 31) J. J. Lander and J. Morrison, J. Appl. Phys. 34 (1963) 1403. 32) R. E. Schlier and H. E. Farnsworth, J. Chem. Phys. 30 (1959) 917. J. J. Lander and J. Morrison, ibid 33 (1962) 729. 33) M. Otter, Zeit. fur Naturforsch. 14 (1959) 355. 34) Formulae for this quantity are given in: S. Dushman, Scientific Foundations of Vacuum Techniques Second Edition (John Wiley & Sons, Inc., New York, 1962) p. 14. 35) H. E. Farnsworth, Appl. Phys. Letters 2 (1963) 199. 36) J. J. Lander and J. Morrison, J. Chem. Phys. 37 (1962) 729. 37) J. A. Rayne and W. G. Kemp, Phil. Mag. 1 (1956) 918. 38) A. U. Mac Rae and L. H. Germer, Phys. Rev. Letters 8 (1962) 489 39) Recent surveys of cooperative phenomena in crystals are given by: C. Domb, Advances in Physics 9 (1960) 149, 245. T. Muto and Y. Tagaki, Solid State Physics 1 (1955) 193. L. Guttman, Solid State Physics 3 (1956) 145. 40) W. L. Bragg and E. J. Williams, Proc. Roy. Sot. (London) 145 (1934) 699. 41) F. C. Nix and W. Shockley, Rev. Mod. Phys. 10 (1938) 1. 42) R. Peierls, Proc. Roy. Sot. Al54 (1936) 207.

348

A. U. MAC

RAE

43) H. Bethe, Proc. Roy. Sot. A150 (1935) 552.

44) 45) 46) 47) 48)

F. H. Stillinger (private communication). L. H. Germer and R. M. Stern (private communication). W. Ambs (private communication). R. V. Culver and F. C. Tompkins, Advances in Catalysis 11 (1959) 67. F. J. Morin, Semiconductors edited by N. B. Hannay, (Reinhold Publishing Corporation, New York, 1959) p. 620. 49) L. Pauling, The Nature ofthe Chemical Bond Third Edition (Cornell Univerisity Press, Ithaca, New York, 1960) p. 99.