Thermal breakup of NH3 adsorbed on W(211)

SURFACE

SCIENCE 15 (1969) 37-76 0 North-Holland

THERMAL JOHN Department

BREAKUP

W. MAY*,

OF NH, ADSORBED

ROLAND

of Applied Physics,

Publishing Co., Amsterdam

J. SZOSTAK**

Cornell

and LESTER

University,

Received 19 June 1968; revised manuscript

ON W(211)t H. GERMER

Ithaca, New York

14850, U.S.A.

received 21 November

1968

Interaction of ammonia with a (211) tungsten surface is studied by flash desorption mass spectrometry, low energy electron diffraction, and measurements of work function change. The ammonia is adsorbed at room temperature. Upon heating, one-third of the hydrogen is evolved at 500°K leaving a residue having stoichiometry NH*. At about 800°K NHz groups are ordered without further decomposition into a centered rectangular array resting upon the substrate. The overlayer and substrate together form a C(4 x 2) structure. Onset of partial evaporation of the NH2 layer begins at about 900°K with the rectangular unit mesh of the NH2 array being continuously stretched in the direction parallel to the troughs of the substrate. The maximum stretching amounts to about 12x, with stoichiometry apparently remaining NHz. At 1050”K, the rectangular array becomes unstable and there is a drastic surface rearrangement after which LEED patterns are similar to patterns produced by pure nitrogen, though stoichiometry seems to be still NHa. Finally, heating to even higher temperatures causes destruction of this “pseudo-nitrogen” NH2 structure. Nitrogen and hydrogen evaporate together in a pressure burst near 1200°K and the surface is completely clean at about 1300°K. NH3 can be adsorbed as a second weakly held layer on top of an NH2 primary layer. Kelvin method measurements have shown that the work function is hardly changed by a layer of NHa, but NH3 reduces the work function by about one volt after adsorption either on a clean surface or on a thermally prepared NH% layer.

1. Introduction Adsorption and decomposition of ammonia on tungsten have been studied for over forty years, but basic understanding has not yet been reached. Polycrystalline surfaces have been used extensively l-13) and lack of knowledge of surface composition and of surface structure have been severe handicaps. t Research supported by National Aeronautics and Space Administration Contract NGR 33-01&029 and American Iron and Steel Institute Contract No. 148. We gratefully acknowledge the Advanced Research Projects Agency fat financial support of this project through the use of the Central Facilities of the Materials Science Center. * Now at Bartol Research Foundation of the Franklin Institute, Swarthmore, Pennsylvania 19081, U.S.A. ** Present Address: Institut fiir Automation, Miinchen 71, Katharinenstrasse 6, W. Germany. 37

38

J. W. MAY,

R. I. SZOSTAK

AND

L. H.GERMER

In the work reported here, we adsorb NH, at room temperature on a (211) tungsten surface and draw conclusions from flash desorption measurements (section 3), from LEED determinations of surface structures (section 4) and from work function changes (section 5). Well defined tungsten surfaces have been used in some other recent work and these results 14-18) can be profitably compared with those of the present paper. Our observations are in excellent agreement with measurements of ammonia adsorption on a tungsten (100) surface made by Estrup and Anderson 17~1s). They find that, after ammonia adsorption at room temperature, heating to a rather low temperature causes hydrogen evolution leaving NH, on the surface. This is decomposed at a much higher temperature with simultaneous evolution of nitrogen and hydrogen. “There are no indications that adsorbed NH or N are formed as intermediate products.” NH, which is held weakly to the (100) surface causes a reduction of the work function by about 1 eV but NH, produces only a very small change of work function. From our own measurements upon a tungsten (211) surface, we have independently drawn these same conclusions. Well ordered structures of adsorbed NH2 are found in both sets of experiments although, of course, the structures are quite different on the (100) and (211) surfaces. These conclusions are in irreconcilable disagreement with those which have been reached by Dawson and Hansen r6).

2. Experimental Three different tungsten crystals were used and gave identical results. Each was welded to heavy tungsten supports and mounted in a Varian Associates LEED system. Temperatures were measured by tungstenrhenium thermocouples. Gas composition was monitored by a Varian Associates quadrupole mass spectrometer. When ammonia was admitted large quantities of hydrogen were formed, probably by cracking of ammonia in the sputter-ion pump (Varian Associates, 140 I/set). Hot filaments of the LEED gun, mass spectrometer and ion gauge also decomposed ammonia, though decomposition from these sources could be eliminated almost completely by reducing power to the filaments. After this was done, the partial pressure of nitrogen was very small, but that of hydrogen still high but not harmful for reasons explained later. Experiments were necessarily conducted in mixtures of ammonia and hydrogen with some nitrogen. Relative amounts of these gases depended on the time interval after baking the system to 250 “C. Immediately after baking the ratio of hydrogen to ammonia was less than 0.1 and over a period

THERMAL

BREAKUP

OF NH3 ADSORBED

ON w

(211)

39

of days gradually increased until it might exceed unity, at which time the apparatus including the ion pump was baked again. Although hydrogen is initially adsorbed with the ammonia, its presence presented no serious problem because initially adsorbed hydrogen is gradually displaced by ammonia completely. The partial pressure of ammonia in the experimental mixtures was estimated from the combined readings of the mass spectrometer and the ion gauge, using for NH, an ionization cross-section ls) of 3.54 x lo-l6 cm’. Pressures measured this way are inherently inaccurate, and the real pressures of ammonia near the crystal and at the ion gauge and mass spectrometer were probably somewhat different. Calculated values of exposure to ammonia in these experiments, therefore, are liable to error and are only approximate. Admission of ammonia proved difficult because of adsorption on the apparatus walls. This has also been found by others lo, rs,aO). When the ammonia valve was opened, the partial pressure of ammonia at the mass spectrometer showed no significant increase for many minutes, and only after hours did the gas mixture of ammonia, nitrogen and hydrogen reach equilibrium at a steady flow. Sluggish behavior was found also when one closed the ammonia valve, several hours being necessary to pump down to lo-’ Torr. During this time, ammonia was desorbing from the walls which were a source of ammonia, rather than a sink as during admission. Because of the awkwardness of admitting and pumping out ammonia, experiments were usually conducted in a Ilow system in which theequilibrium partial pressure of ammonia was conveniently small. In most cases, the ammonia pressure was in the range lo-’ to lo-’ Torr, but in some experiments pressures in the 10e6 Torr range were used*. The times of flash heating were always kept so short that no significant amount of ammonia struck the crystal while it was above room temperature. Ammonia was adsorbed at about room temperature in all experiments described here. It seems to be necessary to point out that our short crystal, welded to heavy leads which do not get very hot, cools extremely rapidly and is quite different from the usual long filament for which cooling time is considerable. Isotopic ammonia (99 at. % 15N from Volk Radiochemical Co.) was used rather than ordinary ammonia. Decomposition of ammonia gave 30N2 so that contamination by 28C0 could be detected. Adsorbed CO was not present in appreciable amounts in any of the experiments, and in flashoff CO never exceeded 1% of the 30N2. Apparent pumping speeds for nitrogen and hydrogen were measured by admitting these gases to 10M6 Torr with the pump off, then measuring the * After high pressure exposures apparatus was pumped out before flashing.

40

J. W.MAY,

R. 3. SZOSTAK

AND

L.H.GERMER

time constant of evacuation when the pump was turned on. This method simulates to a good approximation a pressure burst to the pump during a flash-off. Apparent pumping speeds were somewhat lower than the rated values

for steady

state flow because

of the buffering

effects of the walls.

Fig. 1. Typical schematic mass spectrum after 15NH3 had been flowing through Vacionpumped apparatus for several days. Gas mixture: Hz - 2.5 x 10d7, 15NHa - 1.6 x 10m7, 30Na - 0.4 x 10-7 Torr. Note high partial pressure of Ha after this long flow. Peaks at 16, 17 and 18 are cracking pattern of ammoniaa’). Peaks at 15,28,29 and 30 are atomic and molecular nitrogen from decomposed ammonia. Partial pressure of CO< 5 x lo-lo Torr.

adsorption experiments, the only ambient gases detected in significant amounts were ammonia, nitrogen and hydrogen. In fig. 1 are drawn the relative peak heights of a mass spectrum of a typical gas mixture after long times of ammonia flow. The peaks at m/e= 18 and 17 are due to 15NH: and 15NHl which are expected in about equal proportions21). During

They are much larger than the *‘NH+

peak.

3. Flash desorption measurements We have stated in the introduction that one-third of the hydrogen from adsorbed ammonia is evolved upon heating to a rather low temperature (~500 “K), and that the residual NH, remains on the crystal surface until it finally comes off as nitrogen and hydrogen at temperatures above 900 “K. These conclusions are based upon measurements of the amounts of gas detected by the mass spectrometer upon heating the crystal. Because there

THERMAL

BREAKUP

OF NH3 ADSORBED

ON

w(211)

41

is always present some free hydrogen mixed with the NH, that strikes the crystal, and even also some nitrogen, interpretation of the flash off experiments requires great care. Various supplementary tests have been carried out that would have been unnecessary if these added gases had not been present. The reasons for these tests will be obvious. All flash-offs were made by putting a pre-set voltage across the crystal for a fixed time. The output of the mass spectrometer, giving the pressure rise of a particular gas - hydrogen or nitrogen - is traced out on a recorder or on the screen of a storage oscilloscope. Integrals of these curves are converted to numbers of molecules by using the published cross-sectionsz2), and the apparent pumping speeds determined as described above. We have already pointed out that absolute values obtained in this way may involve considerable uncertainty, but most deductions are based upon relative changes which were quite reproducible. Among the supplementary tests which were necessary were observations upon the adsorption of the individual gases hydrogen and nitrogen. 3.1. ADSORPTIONOF HYDROGENANDNITROGEN The supplementary hydrogen measurements were carried out by exposing the clean crystal surface to hydrogen for various lengths of time, following each exposure by flashing the crystal clean. We report a series of tests in which the exposures were respectively 0.6 x 10m6 Torr-see (0.6 L), 3L and 12 L*, all at the hydrogen pressure 5 x lo-* Torr. For the lowest exposure there was a single burst of hydrogen at about 600 “K representing an amount of hydrogen corresponding to a sticking probability of 0.3. (The agreement of this value of 0.3 with Armstrong’s estimate2s) gives some small support to our method of obtaining quantitative values.) For the larger exposures, there was in addition to the burst of gas at 600 “K a second peak in the flashoff curve at about 400 “K, only slightly developed after 3 L exposure but well developed after 12 L exposure. No other hydrogen evolution was found on flashing to 1400 “K. From a more extended series of measurements, the sizes of these two peaks measured by their integrals converted into numbers of Hz molecules per square centimeter, have been plotted in fig. 2 for a series of values of hydrogen exposure. One sees that the lower temperature peak appears strongly only after the higher temperature peak has reached its maximum. Each corresponds to about one monolayer of H, molecules (0.82 x 1015 cm-‘), very likely two monolayers of H atoms in the high temperature peak and one monolayer of H, molecules in a weakly held second layer. (Mimeault and Hansen24) have shown that on polycrystalline tungsten hydrogen is * This abbreviation

is used throughout

this paper.

J. W. MAY,

42

R. J. SZOSTAK

AND L. H. GERMER

adsorbed as atoms in a primary layer and as molecules in a set ond layer.) These hydrogen adsorption experiments yield two results that are of great importance for the interpretation of experiments of this paper: 1) Evolution of hydrogen at 600 “K is characteristic of hydrogen adsorbed from the gas phase. 2) When hydrogen is found to be evolved at high temperatures after NH, exposure in experiments not yet described, we know that it comes from decomposition of the original ammonia and ~~~~~~~2 ~~~~oge~ adsorbed as such on the crysfal, nor from the crystal leads, nor from the walls ofthe chamber.

Peak

at 600°K Peak at 400’K

H2 Exposure(t)

Fig. 2.

Maguitudes of the hydrogen flashoff peaks after various hydrogen exposures.

Extensive nitrogen adsorption experiments have not been required and were not made. In those tests that were carried out, isotopic 30N2 was used to check freedom from CO. (Volk Radiochemical, 99% “N.) Nitrogen adsorption measurements on this crystal surface have been reported by Changs5) also. Two observations only are of importance: 1) Nitrogen is adsorbed strongly although with low sticking probability at room temperaturez5). All of it is desorbed in a high temperature pressure burst with desorption most rapid at about 1200 “K. 2) If hydrogen is present, mixed with the nitrogen, it preempts available adsorption sites and further impedes nitrogen adsorption. This considerably simplifies matters since ammonia adsorption is always at room temperature. At elevated temperatures, however, (e.g. 850 “K) the sticking probability of nitrogen was greatly enhanced and interference from hydrogen was not observed. That the blocking effect of hydrogen may be important on other crystal planes also is suggested by observation of this effect on polycrystalline tungsten by both Rigby and Beeck 26). 3.2. ADSORPTION OF AMMONIA Although

ammonia,

hydrogen

and nitrogen

can individually

be adsorbed

THERMAL

BREAKUP

OF NH3 ADSORBED

ON

w(211)

43

upon a tungsten crystal surface at room temperature, the constant presence of nitrogen and hydrogen in our chamber does not seriously interfere with our ability to study ammonia adsorption by itself. The reasons for this are the facts that at room temperature the effective sticking probability of nitrogen in the presence of hydrogen is so much lower than that of either of the other two gases that it does not interfere detectably, and that hydrogen which is adsorbed initially is completely replaced by ammonia after a sufficiently long exposure. (See also Estrup and Anderson I”).) Also, nitrogen does not retard ammonia adsorption at room temperature on polycrystalline tungsten 4,5,g, 27). On W(211), this simplicity is probably true only at room temperature. At elevated temperatures, serious and confusing competition from strongly held nitrogen adsorbed from the gas phase is to be expected.

Fig. 3. Hydrogen evolution after coadsorption of ammonia and hydrogen. NH3 pressure 1.0 x 10-s Torr, Hz pressure 1.7 x 10e8 Torr. Exposure time 350 sec. Peaks A1 and B are characteristic of ammonia, peak AZ comes from hydrogen adsorbed directly (the 600°K peak of fig, 2). Initial heating rate 200” set-I.

This replacement of hydrogen by ammonia is illustrated by the flash-off curves of figs. 3 and 4. In fig. 3 is shown the hydrogen evolved as the crystal was heated rapidly to high temperature after an initial moderate exposure to an ammonia-hydrogen mixture. The hydrogen peak marked A, represents hydrogen adsorbed directly from the hydrogen-ammonia mixture; it is the 600 “K peak of fig. 2. In the presence of ammonia, we have never observed a hydrogen peak which can be identified with the weakly held (second layer) hydrogen of the 400 “K peak of fig. 2. The hydrogen peak marked A, is identified with adsorbed ammonia. From fig. 4, one sees that it grows with ammonia exposure and after an ammonia exposure of 12 L, it is inferred that hydrogen had been completely displaced. The hydrogen that desorbed at about 1200 “K came from the originally adsorbed ammonia, because it was observed (previous section) that there was no evolution at 1200 “K when hydrogen alone was adsorbed. Companion flash-off curves for nitrogen were obtained also. These are simpler than the curve of fig. 3; they show evolution of nitrogen in a single pressure burst only, at about 1200 “K.

44

J. W. MAY,

R. J. SZOSTAK

AND

L.H.

GERMER

The basic conclusion that the A, peak is produced by directly adsorbed hydrogen was established even more directly by flowing extra hydrogen into the chamber with a pre-set steady flow of ammonia. If one maintains

0.5

min

t AP m/e q2 1.0 min

2.0min 500°K

600°K

Fig. 4. Room temperature displacement of hydrogen by ammonia. Hydrogen desorption traces after ammonia exposures of 3,6 and 12 L with PNH~= 1.0 x 1O-7 and PHI = 7 x 1O-8 Torr. Initial heating rate 150” set-l. Figure illustrates decrease of peak AZ and growth of peak A1 with increasing exposure to the ammonia-hydrogen mixture. The B peak of hydrogen (not shown here) and a corresponding B peak of nitrogen grow also with ammonia exposure.

constant the total gas pressure, ammonia plus hydrogen, then this change to an increased percentage of hydrogen gave always an increase of the A, peak and partial suppression of other peaks. When, however, the ammonia exposure is kept constant for different hydrogen pressures, one obtains the results of fig. 5. In this figure are plotted flash-off measurements obtained after constant ammonia exposure of 3.5 L, with different amounts of added hydrogen to vary the ammonia hydrogen pressure ratio in the range between 0.001 and 4. Separate curves show the amounts of hydrogen in the A,, A, and B peaks, and also the amount of nitrogen flash-off in the B peak (at about 1200 “K also). One notes at once that the ratio of the hydrogen B peak to the nitrogen B peak, the two peaks at 1200 “K, is near to 2 under all conditions - as the total flash-off of these peaks varies by a factor of 5 and the PNHn/PH2ratio by a

THERMAL

BREAKW

OF NH3 ADSORBED

ON

45

w(211)

factor of 103. This means clearly that the residue left on the surface after all low temperature hydrogen has been removed is always just NH,. The constancy of this ratio has been confirmed in another set of experiments in which flash-offs were measured after exposures for widely different times to a constant mixture of NH, and H,. The ratio of the hydrogen and nitrogen B peaks was found to be completely constant, even down to very small exposures giving small surface coverages. This constancy of the ratio rules out disassociation of NH, to NH or N on the surface at lower coverages. The quantitative value of this ratio depends, however, upon the ratio of our calibrations of the mass spectrometer for nitrogen and hydrogen. This is subject to considerable uncertainty. Ratios of amounts of hydrogen do not involve this uncertainty and are, therefore, to be trusted more.

10-3

IO”’

10-Z P

N”3 ‘p

I

IO

tit

Fig. 5. Flash-off results after NHa exposures of 3.5 L for different amounts of added hydrogen. Peak A2 is not found for high ratios of ammonia to hydrogen. Note the coolant hydrogen/nitrogen ratio of peak B appropriate to NHa.

The hydrogen Al peak is properly stoichiometric with the B peaks of hydrogen and nitrogen only when there is no significant hydrogen adsorption, i.e., no A, peak. Only then is the hydrogen of the A, peak equal to half that of the B hydrogen peak as required to account for all the atoms* * One notes that Melton and Emmettas) observed minute quantities of evolved NH and NH2 radicals from surfaces of platinum and iron, but in amounts far too smalt to be detectable in our apparatus.

46

1.W.

MAY,

R. J. SZOSTAK

AND

L. H. GERMER

in NH3. At lower values of the ammonia/hydrogen gas pressure ratio, the A, peak is too large. This suggests that the A, peak “feeds” the A, peak, with some re-association of adsorbed hydrogen atoms with NH, leading to an evaporation route at 500 “K via NH, fast enough to complete with normal evaporation of ordinary hydrogen at the higher temperature of 600 “K. This feeding of A, by A, must be closely related to the ammoniadeuterium exchange mechanism, which is very efficient at these temperatures29). One notes that at the extreme right of fig. 5, there is no A, peak and the ratio of the B and A, hydrogen peaks closely approaches the theoretical value of 2. The collision rates of our isotopic ammonia and hydrogen molecules become equal when the ratio of their pressures is 3, and one notes that at this ratio, the retarding effect of hydrogen is rather small. There is fast replacement of adsorbed hydrogen by NH3, a high sticking probability for NH,, and lower sticking probability for hydrogen; avalue of 0.8 for ammonia is deduced below, in satisfactory agreement with 0.45 reported by Estrup and Anderson Is) on clean (100) tungsten, and the value of 0.3 for hydrogen which was reported above. For convenience, we represent the amounts of hydrogen in the flash-off peaks by the letters identifying the peaks. With this nomenclature, the fraction

Fraction

of H2

in Peak Ai not from NH3

lOI5 molecules

cme2 He-Peak

B

Fig. 6. Fraction of hydrogen flashed off in peak AI that is not from NH3, plotted against experimental values of peak B hydrogen. Data from fig. 5. When this fraction falls to zero, the value of peak B is 1.3 x 1Ol5 molecules cmm2 which corresponds to a stricking pobability of NH3 of 0.8 in the absence of hydrogen.

of “extra” hydrogen in the A, peak is (l-O.5 B/A,). In fig. 6, this fraction is plotted against the number of hydrogen molecules in the B peak. From this plot, the number of H, molecules in the B peak is about 1.3 x 1015 cmP2 for the exposure 3.5 L when there is no interference from adsorbed hydrogen. From kinetic theory, the number of molecules of NH, at 300 “K reaching the surface for an exposure Q measured in L units is 4.8 x 1Ol4 Q cm-*, or

THERMAL

BREAKUP

OF NH3 ADSORBED

1.7 x 1015 for exposure 3.5 L. The indicated is 0.8 in the absence of hydrogen. 3.3. COMPLETE NH,

ON

w(211)

sticking

47

probability

for NH,

COVERAGE

We have shown that complete coverage of NH,, that is without coadsorbed hydrogen, is attained by long exposure permitting NH, to replace initially adsorbed hydrogen. Maximum NH, coverage, however, is not attained by flashing such a saturated surface to about 800 “K. This is found out by observing that, after further exposing such a surface to NH,, a second flash to 800 “K evolves additional hydrogen in an A, peak, and, of course, an enhanced B peak upon complete cleaning. This is taken to mean that adsorbed NH, radicals are smaller than adsorbed NH, molecules.

t AP

Three

Doses

m/a=2

500 700 900 1100

1300 1400

1500

1600°K

Fig. 7. Hydrogen flashed off after a single saturating exposure to NH3 (90 L) at 1O-6 Torr, followed by pump-out before flashing-curves marked (1); and after further saturating exposures followed by pump-out with intermediate flashing to 800°K to remove the AI hydrogen-curves (2) and (3).

Maximum NH, coverage is attained by several exposures with intermediate heating to 800 “K. In figs. 7 and 8 are shown flash-off observations after I,2 and 3 exposures of 90 L each with intermediate heating to 800 “K. Relative values of numbers of molecules arriving at the mass spectrometer are marked on the peaks in the figures. The integrated numbers of molecules of hydrogen and of nitrogen in the different flash-off peaks are collected in table 1, and for comparison theoretical values calculated from the LEED

J.w.

48

MAY,

R.J.SZOSTAK

AND

L.H.GERMER

measurements of section 4. The agreement is good in view of uncertainties in our flash desorption measurements. The B hydrogen/C A,hydrogen ratio is inherently reliable, not involving calibrations as does the B hydrogen/B nitrogen ratio. The first ratio alone establishes the NH, stoichiometry, and then is well supported

by the second ratio also.

AP m/e=30

1 Fig. 8.

I

I

I

I

I

500 700 900 1100

I

1300 1400

1500

1600

Same as fig. 7 - for nitrogen. Note that there is never evolution low temperatures.

of nitrogen at

TABLE1 Surface coverage and composition lOI5 Molecules per cm2 in peak B Hydrogen molecules

Flash-off value

(fig. 7)

Calculated from LEED nattern isec. 4)

1.4 2.1

Flash-off value

3ne iose rhree loses

B hydrogen B nitrogen >

Nitrogen molecules

(fig. 7)

(figs. 7, 8)

(fig. 8)

Calculated from LEED pattern isec. 4)

_

0.66

_

1.9

2.1

1.84

I.0

0.92

2.0

2.1

3.4. DESORP~ION OF MOLECULESAND OF ATOMS The hydrogen atomization experiments of Brennan and Fletcherso) and of Hickmottsl) lead one to expect that atomic hydrogen might be desorbed when NH, is decomposed at 1200 “K. We have some strong circumstantial evidence this may be so, with H atoms combined into molecules on the walls of the chamber before they reach the mass spectrometer located in a

THERMAL

BREAKUP

OF NH3

ADSORBED

ON

w(211)

49

side arm. Our available evidence suggests, however, that the nitrogen from NH, leaves the surface as molecules, as does the hydrogen of the A, peak. Desorption curves for hydrogen and nitrogen, measured in the mass spectrometer as H, and as N,, are superposed in the lower half of fig. 9. The surface before the flash-off had been completely covered by NH, by the method of alternate exposure and low temperature flash as described above. Not only does the maximum in the hydrogen trace come after that of nitrogen, but also there is a long tail on the hydrogen trace which is lacking in the nitrogen trace. This is more obvious in the upper box of fig. 9, which shows the integrated number of hydrogen and nitrogen molecules sensed at the mass spectrometer. We suggest that hydrogen reaches the mass spectrometer later than nitrogen because it leaves the crystal as atoms which linger on the walls of the chamber until they are combined there into molecules. The hydrogen delay is not a pumping speed effect because hydrogen is pumped about three times faster than nitrogen by the Varian pump. An alternate explanation for this delay is absorption of hydrogen into the crystal fotlowed by desorption at a higher temperature as reported by Moore and Unterwaldss).

Spectrometer

Signal

500

700900

1100

Ix)0

1400

1500

1600°K

Fig. 9. Desorption from surface having a saturated NH2 layer. Superposed traces for nitrogen and hydrogen evolution are shown below (initial heating rate 200” set-1). Integrated amounts of hydrogen and nitrogen reaching the mass spectrometer are shown above.

We report another observation supporting pointed out above that almost all experiments

our speculation. We have were carried out in a stable

50

J. W. MAY,

R. J. SZOSTAK

AND

L. H.GERMER

steady flow of ammonia, with its accompanying (excepting only the “high pressure” experiments

hydrogen

and

nitrogen

leading to fig. 7). In for many hours with the walls of

general, steady flow had been maintained the chamber in equilibrium with this flowing gas mixture. When we have deviated from this situation and carried out a test after the chamber was

[Ioli1

BEAM

Fig. 10. Model of bee (211) surface. All LEED patterns are referred to this orientation of the crystal. Figure shows positioning of crystal with respect to beam. The [21 l] direction is pointing to toward the electron gun. Dimensions of unit mesh (outlined) are (3*/2) ao x 2*ao with ao = 3.16 A for tungsten. Fig. 11. LEED patterns at normal incidence, from clean W(211) at top, hydrogencovered (middle) and ammonia-covered (bottom). Left and right hand columns, respectively 55 and 90 eV. It is interesting to note that the intensities of the patterns are asymmetric about a horizontal line through the center, reflecting the assymmetry shown in the model of fig. 10. Illumination of crystal supports in the hydrogen patterns is by the mass spectrometer filament. The center spot in these, and other, photographs is caused by light from the primary filament seen around the crystal by the wide aperture camera lens.

THERMAL

BREAKUP

OF NH3 ADSORBED

ON

w(211)

51

Clean -+

t

t

55 eV

90 eV

52

J.W.

MAY,

R. J.SZOSTAK

AND

L.H.GF.RMER

freshly baked and its walls not in equilibrium with the gas, peak A, hydrogen and peak B nitrogen were found just as if the apparatus had been “aged” in flowing ammonia for several days - but peak B hydrogen was not detected at all. It could be found, although still small, in a second experiment after the lapse of one hour, and aging for more than one day was necessary for peak B hydrogen to become stoichiometric with peak B nitrogen. We interpret these results to mean that nitrogen and low temperature hydrogen desorb as molecules, but high temperature hydrogen desorbs as atoms; these H atoms stick to the walls if the walls are clean, but if the walls are in equilibrium with the flowing gas they soon recombine and leave as molecules. The inference is that strongly held NH, decomposes at high temperature to eject H atoms and leave mobile adsorbed N atoms; these immediately combine and are at once desorbed as molecules. 4. Diffraction measurements The orientation of the W(211) surface relative to the beam for all the LEED patterns in this section is shown in fig. 10. We need to specify this orientation because there is only one mirror plane perpendicular to the surface, as can be seen from the ball model. The most striking feature of the (211) surface, which is the third densest face of tungsten, is its furrowed nature. Troughs lie between close packed rows of W atoms and give the surface something of a one-dimensional character. This structural feature has a decided influence on the nature of the structures that are formed by thermal decomposition of adsorbed NH,. That these structures are formed of NH, radicals has been concluded from the flash-off experiments. 4.1. ADSORPTION AT ROOM TEMPERATURE In

fig. 11 are

ammonia There

adsorbed

compared at

are no fractional

LEED

patterns

room

temperature

order

beams,

and

produced and

not

by

hydrogen

subsequently

ail the patterns

and heated.

are (I x 1) with

Fig. 12. Patterns at 90 eV after saturation room temperature exposures to NH3 followed by heating. (a, b, c) each one dose and one heating, respectively to 750,800 and 850°K. (d and e) respectively two doses and three doses, heated to 875OK after each dose. Patterns observed at room temperature. Sequence a, b, c shows development of an ordered C(4 x 2)-NH2 structure, with degree of order dependent on flashing temperature. The photos of d and e illustrate intensification of the LEED patterns after redosing, and show also that relative intensities of extra spots are not significantly affected by redosing. (Light from mass spectrometer filament illuminates the crystal supports in a to d). Spots marked by arrows are primary beams of the overlayer.

THERMAL

BREAKUP

OF NH3 ADSORBED

ON

w(211)

53

54 low background.

J. W.

MAY,

Relative

R. J.SZOSTAK

beam

AND

intensities

L. H.GERMER

are not

seriously

affected

by

adsorbed hydrogen but are quite markedly changed by adsorbed ammonia. At the present state of the art, little can be deduced from the LEED intensities concerning the surface positions of adsorbed H atoms from hydrogen adsorption, or positions of adsorbed NH, molecules after ammonia adsorption. We can infer from the low background of fig. 11 that the unit mesh of an NH,-covered W(211) surface has the size shown in fig. 10, and contains an integral number of monolayers of NH, molecules (defining 1 monolayer as the number of W atoms in a (211) plane). Heating the hydrogen-covered surface causes no change from (1 x 1) symmetry. The pattern simply becomes a (1 x 1) clean surface pattern after all the hydrogen is off the surface at low temperature. But great changes can be produced by heating an ammonia covered surface. 4.2. THERMALLYORDEREDNH, After the low temperature hydrogen (peak A J is driven off by heating to about 700 “K (fig. 7), the background in the LEED patterns increases considerably, indicating a disordered arrangement of NH, radicals on the surface. The patterns are, however, still (1 x 1). Heating to higher temperatures causes the NH, radicals to become well ordered with no further gas evolution. The resulting C(4 x 2) structure is described below. Development of this C(4 x 2) structure is shown in fig. 12. The photographs of figs. 12a, b and c were obtained after single saturation doses of ammonia and then heating respectively to 750, 800 and 850 “K. Evidently degree of ordering after single doses depends on the flash temperature, and 850 “K is sufficiently hot to effect good ordering of the NH, groups. The patterns of figs. 12d and 12e were obtained after multiple saturation doses of NH, separated and followed by flashing at 875 “K, fig. 12d after two doses and fig. 12e after three. It is evident from figs. 12c, d and e that redosing causes virtually no changes in the relative intensities of diffraction beams from the C(4x 2) structure. This is taken to mean that redosing merely increases the covergae of C(4 x 2) and does not change the structure by forming, for example, a second layer of NH,. This is consistent with creation after 1 dose and low temperature flashing of regions where the NH, C(4 x 2) structure is well developed, as well as regions thought to be bare giving a (1 x 1) pattern. Formation of such “islands” means strong lateral attractive forces between NH, groups in the C(4x2) layer. Redosing experiments can be explained by thermal breakup of NH, in (1 x 1) regions only. Eventually, saturation coverage of NH, is attained and more redosing causes no further addition to the C(4 x 2) coverage. This is the conclusion that was reached in the flash-off experiments described earlier.

THERMAL

4.3. NH,

AT REDUCED

BREAKUP

SURFACE

OF NH3 ADSORBED

ON

w(211)

55

DENSITY

The C(4 x 2) structure contains the maximum NH, density. Structures of slightly lower density are produced by heating above 900 “K. This heating causes some of the NH, to evaporate as nitrogen and hydrogen, and when the temperature has reached 1050 “K there is about 89% of the original NH, still on the surface. Whether or not the hydrogen evaporating in this restricted temperature range is atomic we cannot say. We have already inferred from the flash-off data that the major fraction of the high temperature hydrogen is probably atomic, but this may not apply to decomposition below 1050 “K. In this temperature range 900-1050 “K, a remarkable series of structures is found when a C(4 x 2) surface (redosed or not) is flashed progressively hotter. Heating the C(4 x 2) structure to about 1000 “K causes a small gas evolution and a change of structure to C(10 x 2). This new structure gives LEED patterns shown in fig. 13. As in the case of C(4 x 2), the effect of redosing and flashing to the proper temperature is not to change relative intensities but only to increase the overall sharpness and brightness of the patterns (fig. 13b). In other words, the structure is determined solely by the maximum temperature to which the crystal has been flashed. If only one dose is given at room temperature and the crystal then heated to 1000 “K, some parts of the surface are then presumed covered with C(10 x 2) structure and other parts are presumed to be are, just as in the case of C(4 x 2). Redosing and reheating to 1000 “K increases the fraction of the surface covered by the C(10 x 2) structure, which is 7% less dense than the C(4 x 2) structure. (See below.)

Fig. 13. Diffraction patterns at 90 eV from ammonia saturated surface heated to 1OOO’K. Symmetry is C(10 x 2). (a) one dose, (b) three doses. Spots marked by arrows are primary beams of the overlayer.

56

J. W.

MAY.

R. J. SZOSTAK

AND L. H.GERMER

THERMAL

BREAKUP

OF NH3 ADSORBED

ON

w(211)

Fig. 14. LEED patterns at 90 eV from a surface first saturated with NHz by redosing, and then heated above 9OO’K. The sequence a to h were obtained after heating to progressively higher temperatures, as tabulated below.

Max. temp. (“K)

a

Calculated NH2 coverage (monolayers)

d e f

910 920 935 955 970 1000

0.738 0.727 0.719 0.714 0.707 0.700

2.21 2.18 2.16 2.14 2.12 2.10

:

1030 1040

0.687 0.675

2.06 2.03 _ .--

__~ a b C

Patterns

__~

of d and f are respectively

.- _P(7 x 2) and C(l0 x 2).

It is obvious by comparing figs. 12 and 13 that in the C( 10 x 2) structure, the surface spacings parallel to the [Iii] direction have changed from those in C(4x2), but that spacings parallel to the [Oli] are still the same as in C(4 x 2). In going from C(4 x 2) to C(I0 x 2), there has been an expansion of

the NH, structure in the [Iii] direction, along the furrows that are so prominent in the model of fig. 10. This expansion is not, however, simply a sudden change to C(l0 x 2) as the crystal is heated. Rather, it appears to be continuous, and a whole family of related structures is found in the temperature interval 900-1050 “K. The C(4 x 2) and C(10 x 2) structures are merely two members of this family. They give LEED patterns of high symmetry, but most of the structures in the family do not. In fig. 14 is shown a selection of patterns obtained after the C(4x2) structure had been heated to pro-

58

J. W.MAY,

R. J. SZOSTAK

AND

L.H.GERMER

gressively higher temperatures. Each heating to a higher temperature caused desorption of a small amount of hydrogen, and presumably nitrogen although this was not measured, and the corresponding expansion of the NH, structure in the [Iii] direction. The patterns of fig. 14 are generally very complicated. Nevertheless, we have succeeded in interpreting all of them in terms of a continuously expanding layer of NH, radicals lying above the (undisturbed) tungsten (211) surface, with this layer becoming progressively more dilute with the evaporation of more and more nitrogen and hydrogen. The layer is always a simple centered rectangular mesh, no matter what the extent of its expansion in the [iii] direction. Complexity of the LEED patterns is due to multiple diffraction of the beam between overlayer and substrate, and not to structures which are inherently more complicated. 4.4. STRUCTURESOFTHENH, LAYER To understand the diffraction patterns of fig. 14, we analyze first the simple C(4 x 2) pattern of fig. 12e. A schematic diagram of this pattern is given in fig. 15. Beams arising from multiple diffraction between the NH, overlayer and the substrate are shown as light crosses, while the primary overlayer beams and primary substrate beams are indicated respectively as crosses-in-circles and filled circles. Primary beams from the centered overlayer (crosses-in-circles) are at the 3 m, 3 y1positions (referred to the substrate pattern) with m+n even. The integers wzand IZare the Miller indices of the overlayer itself. For simplicity consider double scattering, once by the overlayer and once by the substrate, because double scattering alone can give in principle all the reflections indicated by the light crosses in fig. 15. The positions of these are given3s) by the Miller indices hk (referred to the substrate) with h =p + 1 m and k = q + 3 n where pq are indices of a substrate reflection. For example, the beam g indexed in fig. 15 could arise from a combination of the substrate 10 and overlayer ii beams (p= 1, q=O, m= - 1, n= - 1) or from the substrate 13 and il beams, or from several other low order combinations. Referring back to fig. 12e, one notes first of all the great intensity of the 2 ; beam (i.e., the ii beam -.. of the overlayer) and then the correspondingly bright beams at 10 and d i, in good agreement with our assigment of the primary beams. One also notes the marked asymmetry of the intensities in the [iii] direction and the presence of only one mirror plane in the pattern, which is proof that the substrate assymmetry (fig. 10) plays a major role in determining the relative intensity of multiple diffraction beams. The patterns of figs. 14a-h are produced also by multiple di~raction, but with the overlayer becoming more and more expanded in the [iii]direction,

THERMAL

BREAKUP

OF NH3

ADSORBED

ON

w(211)

59

as already described. Positions hk of multiple scattered beams in all of these patterns can be expressed by the formulae h=p + ctm and k =q +-it_n, with m+n even because the overlayer is always a simple centered mesh. The quantity a changes with expansion of the overIayer, and is the quotient of the separation between substrate W atoms in the [Iii]direction (i.e. 3+aO/2) and the separation between NH2 groups in the same direction. For most of

(34x2)

/*

*

QD

\ *

e

rg

x

Y

*

l

l\

l

IO

.I1

\

/ e /

. Substrate W (1121 Reflections QD Primary Reflections of Overlayer x Muitiple Scattering bX!IS

Fig. 15. Schematic diagram of C(4 x 2) pattern (compare with fig, 12e). Filled circles represent primary beams from the substrate, and crosses-in-circles primary beams of a centered NHa overlayer. Multiple scattering between overlayer and substrate gives rise to al1 the other reflections (light crosses). (The fi beam of the NH2 overlayer is given indices referred to the substrate -2 3 .)

the patterns of fig, 14, the quotient is not a simple fraction. When Q is a ratio of small integers, the overlayer and substrate are in registry every few spacings in the [iii]direction, but when CIis not such a ratio, the combined meshes are in registry only very infrequently. With CIa ratio of small integers, the LEED patterns are relatively simple and have high symmetry. On the basis of our model, only four simple patterns are expected within our range of expansion. They are C(4 x 2), P(7 x 2), C( 10 x 2) and P(3 x 2), with respective

I. W.

60

MAY,

R. J. SZOSTAK

AND

L.H.

GERMER

IYvalues of t,+, & and 5. No other patterns are expected, and none found, for which c1is a quotient of integers smaller than eleven. Patterns P(7 x 2) and C(10 x 2) are shown in figs. 14c and 14f. The structures C(4x 2), P(7 x 2), C(lOx 2) and P(3 x2)

inferred

are for from

the LEED patterns are shown in fig. 16. The open circles represent NH, groups in the overlayer and the filled circles W atoms in the top layer of the (211) surface. The effect of expansion parallel to [ lii]is very striking, and in each case we show a particular plausible placing of the overlayer upon the substrate, chosen from placements that differ by translations parallel to the surface. (The actual placement cannot be deduced here because LEED intensities, which alone are affected by the placement, cannot at present be analyzed on account of lack of theory.) The smallest repeating mesh of combined overlayer and substrate is a centered mesh when the denominator of the fraction CIis even, and is a primitive mesh when it is odd.

r-l d d 0

0

0

a=5/7

0

0

d= 7110

0

CJ

0

0.0

0

d= 3/4

0.0 0

0

4

C(4x2)

P(7x2)

C(lOx2

1

P(3x2)

Fig. 16. Sketches of the simplest composite structures, showing their unit meshes. Open circles represent NH2; filled circles, top layer W atoms of substrate. The proposed structures are a:ranged in order of increasing unit edge of the overlayer mesh from 3.65 8, to 4.10 8.

The structures of fig. 16 are compared in table 2. The centered rectangular unit mesh at any point of expansion of the NH, overlayer has the dimensions (1 Ia) (3%/2)

by

(2/3) (2%))

where a, = 3.16 8, is the cube edge of tungsten. The number 0 of monolayers of NH, is always 3a. A complete layer of the C(4 x 2) structure has a theoretical coverage of $ which, if thermally destroyed, would lead to evaporation of 0.92 x lOI molecules of nitrogen per cm’. Comparison of these numbers

THERMAL

with the experimental

BREAKUP

OF NH3 ADSORBED

values for a redosed

ON w

(21 t )

61

surface (table 1) shows an agree-

ment that is well within experimental error. This supports the model structures of fig. 16 and is a good indication that the assignment of the positions of the primary

over-layer beams in the diffraction TABLE

patterns

is correct.

2

Four simple overlayer structures

Pattern

a

NHz coverage (0) *

Overlayer unit mesh dimension in [iTi] (A)

C(4 x 2) P(7 x 2) C(10 x 2) P(3 x 2)

0.750 0.714 0.700 0.667

2.25 2.14 2.10 2.00

3.65 3.83 3.91 4.10

* 1 monolayer = number of W atoms in a (211) plane

Overlayer mesh dimension in [Ol T] (A)

Area per NH2

2.98 2.98 2.98 2.98

5.43 A2 5.70 5.82 6.11

= (2/3)*a0-~ = 8.19

x

10“’ cm-2.

The data of table 2 have been tabulated for cases in which CIis a ratio of integers smaller than eleven. We have, however, analyzed all the patterns of fig. 14 (as well as many others) and have determined corresponding values of tl as the temperature was raised and the NH1 coverage reduced. These values, given in the plot of fig. 17, were obtained from two different sets of measurements of the positions of diffraction spots in the photographs. The apparently continuous decrease in a with increasing temperature means that any stage in the expansion could probably be observed. The coverage 0 (equal to 3~) and the unit mesh edge of the overlayer parallel to the [Iii] direction (3*a,/211) are indicated by the scales of ordinates at the right. The experimental points of fig. 17 were obtained by starting with a repeatedly redosed surface having the C(4 x 2) structure. The crystal was then flashed for a second or two to the temperature of the abscissa, then allowed to cool and a photograph of the pattern obtained. The ammonia pressure was of the order of lo-’ Torr. After waiting for readsorption of more NH,, the crystal was flashed again to a different temperature, either higher or lower than the preceding flash, and another photo taken. Values of Mso obtained were independent of the temperature of the preceding flash and determined only by the temperature of the last flash. Thus it is possible with redosing to go from say a C (10 x 2) to a C(4 x 2) structure and back and forth as many times as desired. Evidently the state of expansion of the overlayer from an initial C(4 x 2) structure is a function of the last flash temperature only.

62

J. W.MAY,

R. J.SZOSTAK

AND

L.H.GERMER

The fact that a high temperature (850 OK) is required to order NH, groups into a uniform C(4 x 2) structure is proof that they are effectively immobile on the surface. This immobility, we explanation for the observation that heating less and less dense structures rather than structure. An important cause of immobility

believe, of NH2 islands of NH,

can give a satisfactory overlayers results in of compact C(4 x 2) in these structures is

0.66

0.72

850

900

950

1000

I050

1100°K

Fig. 17. Expansion of NH2 overlayers in the temperature range 9OC-1050°K. For each experimental point the surface was first saturated with NH2 by repeated dosing, then flashed to the temperature of the abscissa. Values of (xwere computed from measurements of the LEED pattern obtained after cooling. Crosses and circles represent values of u deduced from sets of measurements of two types. The photos of fig. 14 were used (and others). Three different scales of ordinates are marked on the figure; at left - (Y,at right coverage B and unit mesh edge of overlayer parallel to the troughs of the substrate (in angstrom units).

the assumed strong lateral attractive forces within the overlayer which effectively prevent diffusion of the NH, groups. That is, only by a cooperative motion of all NH, groups could a whole structural region shrink by evaporation and maintain a C(4x2) structure (providing that evaporation from island edges is ruled out). Such cooperative motion is highly unlikely, and one therefore infers that evaporation proceeds in such a way as to maintain a uniform distribution of NH, over regions of essentially unchanging area,

THERMAL

BREAKUP

OF NH3 ADSORBED

ON w

(2 11)

63

i.e., dilution of the overIayer as we have described. We also point out that development of islands of NH, after a single dose of ammonia and heating to 500 “K does not require long range mobility of NH,, but is explained by diffusion of NH3 and decomposition of the NH3 at the edges of disordered islands of NH,. The ordering of such islands is difficult owing to the strong forces between the NH, groups, and even at 750 “K ordering is not complete (fig. 12a). TRANSITION STRUCTURES

Cf4x2) I

I

I

I

P(7x2)

C(lOx2)

“PSEUDO-N” I

I

? I I I I 1 I

800

900

1000

llOO°K

18. Measurements of amounts of HZ flashed off in the temperature range 900 to 10.50°K, each point representing a separate experiment starting with a single saturation room temperature dose of NHa. The amounts are given as fractions of the total amount of HZ in the C(4 x 2) structure after one dose. Temperatures at which P(7 x 2) and C(10 x 2) are found are marked by arrows. The theoretical fraction 0.11 for formation of P(3 x 2) is marked also. At about this point, the surface a~angement is drastically changed into what we call the “pseudo-N” structure. Fig.

An alternative explanation of why heating NH, overIayers does not create islands of compact C(4 x 2) structure is that evaporation of nitrogen and hydrogen may deviate slightly from the ratio 1:2. This would make the surface stoichiometry a weak function of a, and the real structures on the surface would be more complicated than those already deduced. Careful checks of stoichiometry of the desorbing gases in the range of expansion from C(4 x 2) to P(3 x 2) have not been made, because of the small amounts of desorbate. For the present, we assume that NH2 is the only surface species in this range of coverage.

64

.I. W. MAY,

R. J.SZOSTAK

Although the amounts of hydrogen tions between the different structures

AND

L.H.GERMER

and nitrogen evolved during the transiare small and have not been measured

with precision, those measurements that have been made do agree within experimental error with values calculated from the structures. Measured values for hydrogen evolution are given in fig. 18, expressed as fractions of the total hydrogen contained in the original C(4 x 2) structure. In each case, before flashing, the surface had saturation coverage after a single dose of ammonia. The temperatures at which P(7 x 2) and C(l0 x 2) are formed are marked by arrows; for these the theoretical fractions are respectively l-(a,.,/a,.,)=0.048 and l-(cr,,.,/cr,.,)=O.O67 from table 2. The theoretical fraction for the P(3 x 2) which is never quite reached, is 0.11. In all three cases, the experimental values of fig. 18 are in excellent agreement. One might expect unusual stability for those surface structures for which the epitaxial relation with the substrate is simples4). That this is true is evident in the case of the C(4 x 2) structure which is stable over a considerable temperature range before suddenly changing at about 900 “K. Judging from the shape of the curve of fig. 17, this seems also to be true to a lesser extent for the C(10 x 2). 4.5. SUGGESTEDLOCATIONSOFIIYDROGENATOMS The proposed C(4 x 2) structure of fig. 16 is redrawn in fig. 19 with added derails. Here, N and H atoms are drawn to scale with their accepted covalent radii of 0.75 and 0.37 A. If this structure is correct, the H atoms are shared by the N atoms in a way analogous to hydrogen bonding, but with the N...H...N bonds stronger than normal and considerably shorter (2.35A) than the usual distanceas) of 3.1 A. The obtuse angle HNH in these structures is similar to the angle HNH = 103’ of the bent ground state of the amidogen radical (NH,) in the gas phasea6). In table 3 are collected the values that we infer for the N...H...N distances and the obtuse angles HNH for the range of structures from C(4 x 2) to P(3 x 2). There are other interesting attributes of this model. It is expected theoreti tally that each H atom should move in a single potential well centered on a point midway between nitrogen atoms if the N...H...N distance is small, but if the N.. .H.. .N distance is greater a double minimum in the potential is expected. This has very recently been calculated by Sabins’) for N...H.. .N bonding in a pyridine-pyridinium complex. His calculations show that a double minimum first appears for an N-N separation of 2.2 A. At an N-N separation of just 2.35 A (first entry in table 3), the calculations indicate a hump between minima of height only about 8 kcal/mole. Although the absolute values of the calculated barrier energies are not very reliableaT),

THERMAL

BREAKUP

OF NH3 ADSORBED

ON

W(211)

65

they may, perhaps, receive some support from the observation that the final instability of our model occurs at 2.53 A for which the theory predicts a barrier height of 20 kcal/mole, which seems a reasonable value for breakage of the bond. In the C(4 x 2) structure of fig. 19, the H atoms are localized halfway between N atoms, and as the overlayer expands to stretch the N...H...N bonds, the H atoms are assumed to oscillate back and forth across the potential hump between the N atoms. By the time the P(3 x 2) structure is reached, these oscillations are presumably so disruptive that the overlayer becomes unstable. It is then changed drastically into what we call

HO NO W.

(34x2) Fig. 19. Proposed model of the C(4 x 2)-NHz structure. The centers of substrate W atoms are shown as black dots. The overlayer atoms are drawn to scale as open circles (N diameter = 1.50 A, H diameter = 0.74 A). A C(4 x 2) combination mesh is outlined and dimensions within the overlayer are also shown. A plausible placing of the overlayer is given (differing from others only by translations of the overlayer parallel to the surface). Within the overlayer, H atoms are shared by N atoms with bonding analogous to hydrogen bonding. The N atoms are bonded to the substrate tungsten.

the “pseudo-N” structure. We believe that the NH, groups are then bonded to the substrate in essentially the same way as are N atoms after adsorption of pure nitrogen. Presumably, a hydrogen-bonding type of interaction plays little or no role in the bonding in the “pseudo-N” structure.

J. W.MAY,R, J. SZOSTAKANDL.H.GERMER TABLE3 Structural characteristics Structure

of suggested NH2 overlayer

N...H...N

distance 6%

._.. -..___~ ___^ C(4 x 2) P(7 x 2) C(10 x 2) P(3 x 2)

2.35 2.42 2.46 2.53

Obtuse angle HNH (deg)

_____._~._.

_ - _._-

101.4 104.2 105.4 108.1

._._~_ _.~...-...

4.6. AN ALTERNATE SURFACE STRUCTURE Consideration must be given to the possibility of accounting for the Cf4x2) diffraction pattern (fig. 15) by a structure differing from that sketched in fig. 16. C(4 x 2) patterns can be produced (by multipie difIraction) for any centered rectangular overlayer having unit mesh dimension parallel to [Iii] of (3*/2) a,/($-++ r ) an d unit mesh dimension parallel to [oli]of 2%,/@-l-s), wh ere Y and s are positive integers including zero. In the C(4 x 2) structure of fig. 16, T--S= 1, and the question we must examine is whether another choice of r and s would not be equally reasonable. We give arguments here to support the structures already deduced. One can conclude that Ymust be an odd integer, because if r is even the sequence of diffraction patterns for an expanding overlayer (and it must be expanding, not contracting) is not C(4 x 2) P(7 x 2), C(I0 x 2), P(3 x 2) but a completely different sequence. Because the field of intense diffraction spots is always dense, CIis restricted to small values. Possible values are r= 1 or Y= 3, because if P were 5 the overlayer would be compressed to a physically impossible degree. Next, we take up possible values of $. From considerations of coverage, it is readily conctuded that there are only two possible pairs of values for Yand s that could give the experimental flash-off results, Y=s= 1 described already, and r= 3, s=O which has not yet been considered. We shall show that this second structure is less plausible. For this second structure (not illustrated), NH, groups are packed exactly along the troughs of the substrate, with alternate rows staggered by half a spacing. This spacing aiong the rows wouid be (4/7) (3*/2) a, = 1.57 A for the C(4 x 2), expanding to (3/5 (3*/2) a, = 1.64 A for a final P(3 x 2). This spacing is always too short for an N-H-N hydrogen bond and too long for a single N-N bond. (The N-N bond length of hydrazine is 1.46 A.) The NH, coverage of this speculative structure is 3 monolayers for the C(4 x 2) structure, falling by only & of a monolayer to SQ monolayers for a P(3 x 2) structure. These numbers are in poorer agreement with resuhs

THERMAL

from flash-off

BREAKUP

measurements

OF NH3 ADSORBED

than

are those

ON

67

w(211)

obtained

from the structures

of fig. 16, although certainly not beyond possible experimental error. The most congent reasons for concluding that the overlayer structure is that for which r =s = 1 rather than r = 3, s=O come from qualitative comparisons of observed and expected intensities. Individual comparisons of this sort are subtle and fallible, but the weight of many such comparisons seems strong enough to carry conviction. The strong beams marked by arrows in the C(4 x 2) pattern of fig. 12e and in the C(10 x 2) pattern of fig. 13b are primary beams of the overlayer of the Y=s = 1 structure. They are not so readily accounted for by the r=3, s=O structure. For example, the $4 beam of fig. 12e is most simply interpreted for this structure as double diffraction of an overlayer i3 beam (referred to its own mesh) and the strong 10 beam of the substrate. Although this interpretation is perhaps barely possible, one would expect to find also z + probably equally strong as it would result from il of the overlayer and 10 of the substrate; this latter beam is, on the contrary, quite weak. A completely similar agreement applies to the very strong & 3 beam of the C(10 x 2) pattern of fig. 13b which here also is most simply interpreted as a primary beam of the overlayer of a r=s= 1 structure. A remarkable feature of all of these diffraction patterns is the fact that whenever they are well developed, they are everywhere dense; with almost no exceptions, all possible multiple diffraction spots are present and many are strong. This seems to us to mean that they do not require high scattering orders which are predicted to produce, in general, weak multiple diffracted beamss*). In the same way that we defined c1for r = 1, we can define ~1’for r=3.

One can show that a’= 1 +cc so that h=p’+(l TABLE

+c() m’ for a spot hk

4

Simplest double scattering combinations for the index h of some beams of the C(10 x 2) pattern* Structure of fig. 16 h

l/l0 2/10 3/10 4/10 S/l0 6110 7110 g/10 9110 * h==p+am=p’+(l+a)m’

m=m’

P

3

-2

Alt. Str. pl=p-nl

3

-4 -1

1 -1

2 5

-3 2 0

-2 1 4

-2 3

-3

and

ar=7/10.

-5 -8 4 -1 -6 6

68

produced

3.W.

by double

MAY,

diffraction

R. J. SZOS’TAK

(using

AND

L. W.GERMER

our previous

nomenclature).

Hence,

pn =FB’ and p’ =p - m, and fp’j = IpI + Jn/ll. Therefore, beams requiring p’ (structure for Y= 3, s=O) have orders greater than or equal to those of beams involving p (structure of fig. 19) which argues strongly against r=3 and favors r = 1, i.e., favors the structure of fig. 19. For example, we give in table 4 the lowest order combinations for values of h in the C(lOx2) pattern. The numbers p’ in the last column of the table are in some cases so Iarge that one believes the corresponding beams should not be observabless). Yet, all of these beams are reasonably strong in fig. 13b. In figs. 12, 13 and 14, the spots 2, cz and z, cxare very intense for all values of a, as are aiso their combinations with the substrate beam 10. This is again evidence that $,a and $,a are primary beams of the overlayer; i.e., the structure is that of T=.s= 1. Finally, we have the observation that weak second layer adsorption of ammonia can alter relative diffraction intensities severely (section 5). The alterations of intensity are consistent with the structure of fig. 16 rather than that for r=3, s=O. 4.7. “PSEUDO-N” STRUCTURE This structure that forms at 10.50 “K gives diffraction patterns very similar to those produced by adsorbed nitrogen that has been heated to the same temperature, reported by Chang2s). The similarity is the reason for giving this name to the structure. The “pseudo-N” structure begins to develop just at the temperature at which a P(3 x 2) structure would be expected on the basis of the curves of figs. 17 and 18. Actually, well-formed P(3 x 2) patterns are not observed and the transition to “pseudo-N” occurs just before the overlayer has expanded to the extent required by P(3 x 2). In all discussions so far, the P(3 x 2) structure has been invoked only for reason of heIping the arguments. Patterns sufficiently near to P(3 x 2) have been detected just before the “pseudo-N” to justify inclusion of P(3 x2). Such a pattern is shown in fig. 20a, along with a well developed “pseudo-N” pattern in fig. 20b. Examination of fig. 20a shows weak spots characteristic of a structure P(3 x 2) along with other incipient features characteristic of the “pseudo-N”. In this transition stage, the background is very high owing to surface disorder accompanying the transformation into “pseudo-N”. Some of this disorder may be connected with the evolution of considerable amounts of desorbate (see fig. 18). We estimate that NHp coverage in the “pseudo-N” is about 75 % of that in C(4 x 2), i.e., approximately 1.7 monolayers (table 2). Owing to the complexity of the “pseudo-N” patterns, we have not yet succeeded in deducing the surface structure and more experiments are

THERMAL

BREAKUP

OF NH3

ADSORBED

ON w

(2 t 1)

69

required. There is reason for believing that the “pseudo-N” surface is a faceted surface; that is, the original (211) plane has been rearranged into a hill-and-valley arrangement of other planes covered with adsorbed NH, groups. The similarity of the “pseudo-N” patterns and those from adsorbed nitrogen suggests that the facets are the same in the two cases. We propose to study these two patterns further.

Fig. 20. LEED patterns at 90 eV. (a) transition pattern with incipient ‘“pseudo-N” pattern. Note presence of beams attributable to P(3 x 2) not observable in other patterns published in this paper. High background indicates a disordered structure. (b) “Pseudo-N” pattern fully developed.

An important similarity between the structures produced by N atoms and by NH, is the fact that both are destroyed at the same temperature to give a clean surface (- 1200 “I(). This observation implies that the bonding modes as well as the surface structures, for N atoms and for NH, groups in “pseudo-N”, are essentially the same. This point is important in assessing the deductions drawn by Dawson and Hansen 16) from their field emission microscope study of ammonia on a tungsten tip. The present LEED results and those of Estrup and Anderson 1s) suggest strongly that the data of Dawson and Hansen require reinterpretation. 5. Second layer adsorption When any of the well ordered NH2 structures described above is further exposed to ammonia at room temperature, the diffraction pattern becomes clouded over by a general diffuse haze arising from physisorbed NH, randomly held in a second layer. Many sharp features of the original pattern are unchanged, but those diffraction beams produced only by high order

70

J.W.MAY.

R. J.fZOSTAK

AND L.H.GERMER

multiple scattering are largely eliminated. This is shown in fig. 21 in which are compared a pattern from a weil ordered P(7 x 2) structure of NH Z radicals without superincumbent NH, molecules and a pattern from the same surface after further exposure has produced a second layer of NH,. Weakening of beams produced by high order multiple scattering is observed as well as development of a general haze. These effects are to be expected as a

Fig. 21. LEED patterns at 90 eV. (a) Pattern from layer physisorbed NE-Is molecules. (b) After exposure NH3 molecules gives rise to high background and beams involving high orders. Beams involving

P(7 x 2) structure without second to NH3. Disordered second layer of suppression of multiple scattering only tow orders are still present.

result of general diffuse scattering from a second layer lacking good order. Similar effects have been observed when NH, has been adsorbed upon C(4 x 2), C(10 x 2) and other structures. That this effect is due primarily to physisorbed NH3 cannot be confirmed in direct manner by flash-off measurements because changes of NH3 pressure are buffered by the walls of the chamber as described in section 3. Measurements of changes of work function have, however, given desired information, and have yielded also data of interest for their own sake. These measurements were made with the help of 5. C. Tracy, Jr3”) using the Kelvin vibrating capacitor method. Particular care was taken to prevent thermal disturbance of the vibrating reference reed ftin-oxide costed stainless steel), and the reference reed was always moved away from the vicinity of the crystal whenever the crystal was to be flashed. Checks were made which showed that flashing the crystal below about 1800 “K had no appreciable effect on the work function of the reed, providing the reed was always moved away during flashing. Tests were also made for reproducibility of work function after the reed was moved and returned to its measuring

THERMAL

BREAKUP

OF NH3 ADSORBED

ON

w(21f)

71

position ; measurements were always reproducible to better than 100 mV. Experiments were conducted with a continuous flow of NH3 at low pressure in an equilibrium mixture containing hydrogen as described in section 3. From curve 1 of fig. 22, one concludes that NH 3 molecules adsorbed upon clean W(211) reduce the work function by about 0.8 V, and from curve 3 that NH3 upon a nearly full monolayer of NH, gives a reduction of about 0.9 V. Extrapolations to the instant of flashing of curves 2 and 3 indicate that adsorbed NH, produces no change of work function. The results are collected in table 5, together with observations from the literature of the work function changes of (211) tungsten resulting from the adsorption of hydrogen and nitrogen. 750’K

720’K

450°K

clean

surface

cl

- 0.5 Aip e.% -1.0

- 1.5

0

200

400

600

800

SECONDS

Fig. 22. Changes in work function with crystal exposed at room temperature to a gas mixture containing ammonia and hydrogen at respective partial pressures 3 and 2 x 10-a Torr. This is one continuing experiment with d+ measurements interrupted by flashes to the indicated temperatures at the times of the arrows. Dotted sections represent extrapolations covering the times required to return the vibrating reed to its measuring position. (Curve I) Obtained after cleaning crystal by flashing to 1450°K for 1 sec. Large decrease in work function is due to polar NH3 molecules adsorbed on the bare surface. (Curve 2) Crystal next flashed once to 720°K to create NH2 layer on surface plus extra sites. The newly adsorbed NH3 lies in part upon first layer NHa and, in part, upon clean surface. (Curve 3) Second layer adsorption of NH3 upon a nearly complete NH2 monolayer. Extrapolations of curves 2 and 3 show that A$ for NH2 covered surface is-approximately 0.0 eV.

The work function reduction of curve 1 is less than that of curve 2 because of the presence of some hydrogen on the surface after the first flash and very

72

.I. W. MAY,

R. J. SZOSTAK

AND

L. H.GERMER

TABLE 5

Work function changes on W (211) Ref.

Adsorbate .___NH3 NH2 NH3 upon NH2 HZ NZ

ctr$ -- 0.8 0.0 - 0.9 + 0.65 -t 0.43

This paper This paper This paper 23 40

little hydrogen after the second flash. This we might expect from the collected data of table 5. The slope of curve 1 is reduced also. We suppose that this is due to the retarding effect of hydrogen upon ammonia adsorption in the first layer, as reported earlier in section 3. It is believed that the influence of hydrogen upon ammonia adsorption in the second layer is not serious, although this has not yet been carefully appraised. The second layer is extremely sensitive to CO contamination which was about 1 x lo-“’ Torr in this experiment. The CO replaces the second layer of NH, with very high efficiency and causes a resulting gradual decay in the measured value of LIP. Our explanation of curve 1 of fig. 22 is that ammonia molecules are adsorbing undissociated on the bare W(211) surface with the lone pair of electrons pointing down into the surface. Evidently lowering of 5, is a simple consequence of the permanent dipole moment of the NH, molecule of 1.5 D. Heating causes development of NH2 structures and evolution of hydrogen (section 3) and 4 changes to a value expected from a covalent type of bonding. Molecules of NH3 in a second layer are then presumably held to the NH, layer by Van der Waals attraction, and again the permanent dipole moment of the NH, molecule explains the fall of 4 in curves 2 and 3 of fig. 22. Results obtained from a continuation of this experiment are recorded in fig. 23. The first three points of fig. 23 are taken from the curves of fig. 22, and other points from similar curves as the experiment progressed. The work function change produced by NH2 layers is essentially 0.0 eV, hence the changes in #J shown in fig. 23 are probably due entirely to NH, molecules. For example, flashes 4-9 show that a layer of NH, on a saturated sublayer of NH2 lowers 4 by about 0.9 eV. This is a smaller change than after flash 2 and the explanation may be depolarization within the second layer. Flashes lo-13 demonstrate that second layer NH, molecules are thermally quite labile and desorb readily at low temperatures, as expected for a Van der Waals type of binding. Note the general symmetry of the curve of fig. 23 developed by the final high temperature Bashes.

THERMAL

BREAKUP

5

OF NH3 ADSORBED

10 FLASH

Fig. 23.

continuation

Flash NO. 1 2

3 4 5 6 7 8 9 10 11

ON w(zi

15

73

1)

20

NUMBER

of the experiment of fig. 22 which shows the first three flashes. The temperature of each flash is given beiow.

Temp. (OK)

Rash No.

Temp. (“K)

1450 720 750 750 720 760 720 760 760 330 370

12

400 490 560 610 670 720 780 850 75ft 1130 1450

13 14 15 16 17 18 19 20 I21 22

d$ for NHa - covered surface without second layer NH3 is N 0.0 eV (see fig. 22). FIashes 4-9 show the change in + caused by a second layer of NH3 upon a saturated primary layer of NH2. The level for NH3 upon a full layer of NH:! (A# = - 0.9 eV) is lower than after Bash No. 2, possibly because of depolarization within the second layer. The same effect is seen when the NH2 layer is partially removed (flash 21). Flashes IO-13 demonstrate that second layer NH3 molecules are thermally desorbed very readily.

6. Summary and discussion

The chief conclusions of this paper can be summarized in a few sentences. Ammonia is adsorbed upon our tungsten surface with high sticking probability, resulting in a reduction of the work function of the surface by nearly I V. Upon heating an ammonia covered surface, decomposition occurs at

14

J. W.

MAY,

R. 1. SZOSTAK

AND

L.H. GERMER

about 500 “K with hydrogen evolution, leaving the surface and return of the work function

NH, radicals firmly held to to that of a clean surface.

Further, NH3 can be adsorbed in a weakly held second layer with again work function reduction by about 1 V. A large fraction of this second layer ammonia is evolved without decomposition just above room temperature (fig. 23). Upon heating to higher temperatures, the NH, radicals become well ordered in simple rectangular centered structures with N atoms bonded strongly to the surface and in a speculative model apparently held together by hydrogen bonds; heating to about 850 “K is necessary to obtain almost perfect ordering. When heated above this temperature, there is some evaporation from the surface with expansion of the rectangular NH, structure along the troughs of the tungsten surface, the maximum expansion being about 12% which is reached after heating to 1050 “K. Heating above this temperature results in an entirely different structure not yet investigated. Nitrogen and hydrogen are evolved together in a pressure burst at about 1200 “K, and the surface is clean at 1300 “K. Results entirely compatible with these have been reported by Estrup and Anderson 1’s 18) for the adsorption of ammonia upon a tungsten (100) surface. We believe that the deduction of Dawson and Hansen16) that all hydrogen is removed from the surface at low temperature is erroneous. This can be attributed in part to their experimentation without the aid of a mass spectrometer. TABLE

__._

_.-.

~_

Enthatpy changes

-..

~_.

._~_

Description

Reaction

Nz + 2N~as 2Nads+Nz

6

AH1 = D(N-N) AHz = Heat of desorp-

..-.Enthalpy change (kcal)

_~~ Ref. _

+ 225

+ 95

41

103 + 40

42-44

tion from evaporated W film AN3

-

~~~~~~~~~~~~~~~ -_-- ____

=

-

D(H-H)

AH4 = dHn(NH&as

-

A comment is in order upon the extreme complexity of the diffraction patterns produced by ordered NH2 layers upon our tungsten surface. The models we have suggested account for all of the enormous array of diffraction spots of these patterns and permit no other spots whatsoever. This complexity leads to a degree of certainty which we believe is unique. We point out that this certainty applies only to the multitudinous and often large unit cells of the models, not to the locations of atoms within them although a certain extremely simple arrangement is strongly suggested.

THERMAL

It is important

BREAKUP

OF NH3 ADSORBED

to have estimates

ON

of the magnitude

w(211)

of enthalpy

75

changes

as

ammonia is degraded. We have listed in table 6 the enthalpies of several reactions. The enthalpy LIH, refers to dilute coverage on an evaporated W film rather than to a (211) face, and the numbers we shall calculate are, therefore, only approximately correct. Our observation that adsorbed nitrogen and NH, groups (as “pseudo-N”) have similar diffraction patterns, and are removed from the surface at about the same temperature, makes calculations of the bond energy D, for N atoms adsorbed dissociatively upon the surface and the bond energy D, for isolated NH, groups upon the surface especially interesting. We can write D, = *AHI + $AH2

and

- D, + 20,

+ 3 AH, + AH, + AH, = 0,

where D, is the bond energy of N-H bonds. We obtain D, = 160 kcal/g atom, and assuming D, = 86 kcal/g atom which is the normal value for this bond45), D, = 156.5 kcal/g atom. The agreement is suprisingly good. At the present time, we believe it is too early to assess the new data presented here and in related work 16-18 ) in terms any stronger than at a level of hypothesis for the mechanism of ammonia decomposition on tungsten. The decomposition behavior is sensitive to the particular orientation of the surface, making generalization difficult. In fact, there is evidence that ammonia is not appreciably adsorbed on the W(110) surface46) and, therefore, the (110) surface is probably inactive for both decomposition and synthesis of NH,. There is not only sensitivity to orientation, but also the surface bonding is not simple for the single plane which we are studying here. We have seen that ammonia decomposes on a tungsten (211) surface in two main stages. The first stage gives adsorbed NH, radicals which are thermally destroyed in a second stage at elevated temperatures with evolution of both nitrogen and hydrogen. Nitrogen is never evolved unaccompanied by hydrogen. Tamaru has concluded la) that nitrogen desorption is rate controlling. This is apparently in conflict with the experimental fact7,s) that ND, decomposes more slowly than NH,. It seems that the rate controlling step is actually the break up of NH, (or ND,) and simultaneous evaporation of the nitrogen and hydrogen. Thus, the isotope effect is not incompatible with the conclusion of Tamaru. and there is no conflict. References 1) 2) 3) 4)

C. N. Hinshelwood and R. E. Burk, J. Chem. Sot. 127 (1925) 1105. G. M. Schwab, Z. Physik. Chem. 128 (1927) 161. C. H. Kunsman, J. Am. Chem. Sot. 50 (1928) 2100. C. H. Kunsman, E. S. Lamar and W. E. Deming, Phil. Mag. 10 (1930) 1015.

76

5) 6) 7) 8) 9) 10) 11) 12) 13)

14) IS) 16) 17) 18) 19) 20) 21) 22) 23) 24) 25) 26) 27) 28) 29) 30) 31) 32) 33) 34) 35) 36) 37) 38) 39)

401 41) 42) 43) 44) 45) 46)

J.W.MAY,

R. J. SZOSTAK

AND L. H. GERMER

H. R. Hailes, Trans. Faraday Sot. 27 (1931) 601. W. Frankenburger and A. Hodler, Trans. Faraday Sot. 28 (1932) 229. J. C. Jungers and H. S. Taylor, J. Am. Chem. Sot. 57 (1935) 679. R. M. Barrer, Trans. Faraday Sot. 32 (1936) 490. N. Nagasko and S. Miyazaki, Bull. Tokyo Inst. Technology 13 (1948) 124; (Chem. Abstr. 44 (1950) 10552). M. Wahba and C. Kemball, Trans. Faraday Sot. 49 (19.53) 1351. S. R. Logan and C. Kemball, Trans. Faraday Sot. 56 (1960) 144. K. Tamaru, Trans. Faraday Sot. 57 (1961) 1410. Review articles are in: Catalysis, Ed. P. H. Emmett (Reinhold, New York, 1954); G. C. Bond, Catalysis by Metals (Academic Press, New York, 1962) p. 371; K. Tamaru, Advan. Catalysis 15 (1964) 65. K. Azuma, Nature 190 (1961) 530. K. Azuma, J. Res. Inst. Catalysis Hokkaido Univ. 9 (1961) 55. P. T. Dawson and R. S. Hansen, J. Chem. Phys. 45 (1966) 3148; 48 (1968) 623. J. Anderson and P. J. Estrup, Surface Sci. 9 (1968) 463. P. 3. Estrup and 5. Anderson, J. Chem. Phys. 49 (1968) 523. F. W. Lamps, J. L. Franklin and F. H. Field, J. Am. Chem. Sot. 79 (1957) 6129. A. J. B. Robertson and E. M. A. Willhnft, Trans. Faraday Sot. 63 (1967) 476. Mass Spectral Data, Vol. 1, Serial Number 90, Am. Petroleum Res. Inst. Project 44, Chem. Thermo. Properties Center, A and M College of Texas, College Station, Texas. D. Rapp and P. Englander-Golden, J. Chem. Phys. 43 (1965) 1464. R. A. Armstrong, Can. J. Phys. 44 (1966) 1753. V. J. Mimeault and R. S. Hansen, J. Chem. Phys. 45 (1966) 2240. C. C. Chang, Ph. D. Thesis, Cornell University (1967). L. J. Rigby, Can. J. Phys. 43 (1965) 1020; 0. Beeck, Adv. Catalysis 2 (1950) 151 - see fig. 2 of this reference. E. Molinari, F, Cramarossa, M. Capitelli and A. Mercanti, Ricerca Sci. 36 (1966) 109. C. E. Melton and P. H. Emmett, J. Phys. Chem. 68 (1964) 3318. C. Kemball, Proc. Roy. Sot. (London) A 214 (1952) 413. D. Brennan and P. C. Fletcher, Proc. Roy. Sot. (London) A 250 (1959) 389. ‘I. W. Hickmott, .I. Chem. Phys. 32 (1960) 810. G. E. Moore and F. C. Unterwald, J. Chem. Phys. 40 (1964) 2639. E. Bauer, Surface Sci. 7 (1967) 351. N. H. Fletcher, J. Appl. Phys. 35 (1964) 234, 2276. G. C. Pimentel and A. L. McClellan, The Hydrogen Bond (Freeman and Co., San Francisco, 1960) p. 262. D. A. Ramsay, J. Chem. Phys. 25 (1956) 188. J. R. Sabin, Intern. J. Quantum Chem. 2 (1968) 23; also private comm~ication. P. W. Palmberg and T. N. Rhodin, J. Chem. Phys. 49 (1968) 147. The authors are much indebted to J. C. Tracy, Jr., for his valuable help in carrying out the work function measurements. Details of the apparatus are described in J. C. Tracy, Jr, Ph. D. Thesis, Cornell University (1968). A. A. Holscher, Dissertation, Leiden University, Netherlands (1967) p. 63. 0. Beeck, Advan. Catalysis 2 (1950) 151. M. Scwarc, Chem. Rev. 47 (1950) 75. J. A. Kerr, R. C. Sekhar and A. F. Trotman-Dickenson, J. Chem. SOC. (1963) 3217. J. A. Kerr, A. F. Trotman-Dickenson and M. Woiter, J. Chem. Sot. (1964) 3584. Handbook of Chemistry and Physics, 44th Ed. (The Chemical Rubber Publishing Co., Cleveland, 1962) p. 3519. J. W. May, unpublished results. These preliminary results are somewhat uncertain because CO contamination was quite high.