Surface Science 59 (1976) 638-642 Q North-Holland Publishing Company

A LOW TEMPERATURE PRECURSOR TO CHEMISORBED OXYGEN ON

TUNGSTEN (110)

Received 25 May X976; manuscript received in final form 6 July 1976

Oxygen adsorption on tungsten is generally considered to be dissociative in con- trast to, for instance, CO which is adsorbed nondissociatively at T 5 300 K. We find that for T < 40 K oxygen is adsorbed in a weakly bound state, possibly molecular, from which the more strongly bound, atomic chemisorption state seen at high tem- perature is derived.

The apparatus is shown in fig. 1.02 from a -300 K effusion source impinged on

an electrically heatable tungsten (110) surface in a liquid H2 cooled chamber. Thermal desorption was measured with a Veeco GA4 magnetic sector mass spec- trometer which could also be used to analyze neutrals and ions produced in elec- tron impact desorption, using 150 eV electrons. Work function differences could be measured with a retractable vibrating condenser, using a polycrystalline tungsten ribbon as reference electrode.

Oxygen adsorbed at 300 K is desorbed thermally between 1700 and 2200 K

Fig. 1. Schematic diagram of apparatus. Major components are labelled. Liquid Hz dewar and most solenoid coils are not shown. Interior of glass cell conductivized by stannic oxide coating.

638

C. Leung, R. Gomer / Low temperature precursor to chemisorbed oxygen on W 639

ELECTRON IMPACT DESORPTION

CHANGE -

THERMAL OESORPTION _

27 TEMPERATURE (K)

Fig. 2. Correlation of EID products, work function increase, and thermal desorption spectrum

between 20 and 70 K: initial exposure = 1200 sec. The relative EID intensities of neutral 02

and O+, each collected in a different way, do not represent the total yield of each of the EID

products in the correct ratio. Electron current density 1.5 ~A/cm’.

predominantly as oxygen atoms. A layer deposited at 20 K by 300 set exposure [l] to the source yields the same thermal desorption spectrum with no observable desorption at lower temperatures. With progressively longer initial exposure a de- sorption peak is first seen at 45 K (fig. 2); this peak reaches maximum intensity after 600 set exposure. For even longer exposure a desorption peak occurs at 27 K and in turn reaches maximum intensity after 900 set total exposure. For X00 set exposure a third peak at 25 K occurs, whose height’increases indefinitely with ad-

ditional exposure, up to 3600 set, the longest time used. This peak evidently cor- responds to desorption of higher layers of physisorbed oxygen beyond the first

physisorbed layer, which corresponds to the 27 K peak. EID and work function measurements show that oxygen in the layers desorbing at 25 and 27 K has nearly identical properties while oxygen corresponding to the 45 K peak is not physi- sorbed.

Work function increases monotonically with adsorption at 20 K. A~#I reaches a maximum of 0.25 eV after -600 set exposure and then remains constant with the additional exposure required for the observation of the thermal desorption peaks at 25 and 27 K. On heating the work function remains unchanged until desorption at 40-50 K occurs, where the work function jumps abruptly by 0.35 eV to A@I = 0.60 eV (fig. 2). A rise in @ on heating to 45 K is also observed for smaller doses. For ex- posures of 300 set the full A@ is observed; for 100 set exposure A@ rises from 0.08 to 0.20 eV at 45 K. No other discontinuous changes in the work function are then

640 C. Leung, R. Gomer / Low temperature precursor to chemisorbed oxygen on W

observed with further increase in temperature up to the final desorption between 1700 and 2200 K. Redosing at 20 K after heating to >45 K leaves A# unchanged at 0.6 eV. However, a desorption peak is seen at 40-45 K.

EID of a 20 K layer formed by <600 set exposure or by heating to 27 5 T < 40 K for any exposure yields no detectable ion current initially (fig. 3). With continued EID, O+ is produced, increases in intensity, reaches a maximum after 21 mC/cm*, and then decays with an initial cross-section of 2 X lo-l7 cm* (fig. 3). Heating after electron impact desorption to the point of maximum O+ production shows the dissappearance of the 45 K thermal desorption peak. These results suggest that electron impact causes part of the 20 K layer to desorb as neutral O,, or possibly as O+ or 0; coming off at angles which preclude its detection in the mass spectrometer with conversion of the remainder to an O+ yielding state. The absence of a neutral 0, signal in the mass spectrometer does not preclude 0, desorption since, except for the very high cross section cases observed in physisorption neutral EID products are generally below the limits of detection of the mass spectrometer.

EID after an exposure >600 set immediately yields neutral 0, and O+ ions (figs. 2, 3). First neutral 0, and then O+ decay with large cross sections, (00~ g 7 X lo-l5 cm*, oO+= IO-l5 cm2). 0, disappears irreversibly. The O+ signal goes through a minimum and then rises to a maximum, whose intensity corresponds to that seen with dosing times less than 600 set, or with heating to 27 < T < 45 K. This maximum is only l/SO of the O+ signal seen initially with 900 set exposure. These results indicate that 0, and the initial large O+ signal correspond to physi- sorbed 0,. The rise of O+ after the depletion of physisorbed oxygen by EID, and its

‘~~~ 200 400 600 800 1000 I200

EID TIME (5)

Fig. 3. EID at 30 PA/cm’ for three temperature ranges: initial exposure = 800 sec.

C. Leung, R. Gomer /Low temperature precursor to chemisorbed oxygen on W 641

subsequent slow decay correspond to the desorption/conversion of the chemisorp- tion precursor. The desorption of neutral 0, with very high cross section is signifi- cant. A similar phenomenon is also seen with physisorbed CO,. The explanation is

undoubtedly that EID leads to dissociation (neutral and ionic) of physisorbed 0, with essentially gas phase cross sections. The dissociation fragments have sufficient kinetic energy to kick off several additional physisorbed 0, molecules per EID event and thus explain the very high total desorption cross sections. It is possible that the more rapid decay of 0, relative to O+ is connected with the variation in binding energy or orientation of 0, in the first physisorbed layer, relative to higher layers.

The absence of 0; for physisorbed 0, parallels the absence of CO+ from physi-

sorbed CO2 and is explained by the fact that such molecule-ions do not find them- selves on a repulsive part of their potential curve, relative to the surface.

If EID is carried out after heating a 20 K layer to >45 K, an O+ signal is seen without any induction period, and decays with an initial cross section of 2 X lo-l7 cm2. Thus thermal desorption at 45 K is accompanied by irreversible conver- sion of an equal amount to that desorbed (for the full 20 K monolayer) to a more tightly bound, presumably atomic form, with a concomitant rise in work function. For initial amounts corresponding to exposures of 5300 set conversion at 4.5 K still occurs, but without desorption. This is shown (a) by the abrupt rise in work func- tion at 45 K and (b) by the fact that EID of such a layer at 20 K produces an O+ signal which rises initially, goes through a maximum and then decays, as already discussed.

Finally, it is noteworthy that the “saturated” T > 45 K layer is capable of ad- sorbing l/3 of a virgin layer of CO (relative to the clean surface), which behaves in thermal and electron impact desorption like virgin CO, yielding CO+ and converting to an O+ yielding beta-precursor (as seen with C12018 experiments) and eventually

beta CO. Thusonly a fraction of the surface is in fact oxygen covered. This behavior also finds its parallel in CO adsorption, in that beta-precursor coverage is 40% of the low temperature virgin layer. It is possible that with oxygen further increases in atomic adsorption from molecular 0, are prevented by the requirement that both atoms be firmly bound in order to make the process energetically or at least kineti-

cally feasible, and that this becomes geometrically impossible, at least at low tem- perature.

We conclude that a uniform precursor state is formed at 20 K and converted at 45 K to a less densely populated, more tightly bound chemisorbed state, probably similar to that observed at >300 K. A similar sequence of changes has been reported for Co on W (110) [3]. In that case, a low temperature undissociated state, virgin CO, is converted to a less densely packed, more strongly bonded /.I precursor state at 375 K, accompanied by partial thermal desorption, a work function decrease, and a change from CO+ to O+ as the principal ionic EID product. Aside from the fact that 05 is not observed at 27 <_ T< 45 K, the changes for 0, on W (110) are very similar to those for CO although they occur at much lower temperatures, cor-

642 C. Leung, R. Gamer / Low temperature precursor to chemisor~gd oxygen on W

responding to the much weaker molecular and much stronger dissociative adsorp- tion of oxygen. Evidence for a weaMy bound molecular species at low temperatures has also been reported by Ehrlich [4] for hydrogen on tungsten (110). It thus ap- pears that the phenomenon of molecular precursors to atomic adsorption is fairly widespread. Whether and at what temperature such states convert to atomic adsorp- tion depends of course on the specific binding and activation energies of a given system.

This work was supported in part by NSF Grant MPS 74-12059-AOl. We have

also benefitted by support from the Advanced Projects Research Agency of the Department of Defence under Contract ONR NOOO~4-75-C-1107 administered by the Office of Naval Research.

C. LEUNG and R. GOMER

The James Franck Institute, The University of Chicago, Chicago, Illinois 60637, USA

References

[l] 300 SW of exposure in this arrangement is equivalent to I Langmuir (10m6 Torr see) of oxygen at 300 K. Since beam intensity, source to target distance and crystal orientation are fixed throughout the experiments, other exposure times scafe.

[a] C. Leung and R. Gomer, to be published. [3 J C. Leung, M. Vass and R. Gomer, J. Vacuum Sci. Technol. 13 (1976) 286;

R. Gamer, Japan. J. App. Phys. Suppl. 2 part 2 (1974) 213. [4] G. Erhlich, to be published.