Tysoe56 the First Publication Trimethyl Aluminium Ftir Reaction on a Surface Al2o3

8/12/2019 Tysoe56 the First Publication Trimethyl Aluminium Ftir Reaction on a Surface Al2o3

1/7

Chem . Mater. 1994,6, 1705-1711 1705

Infrared Spectroscopy of Trimethylaluminum andDimethylaluminum chloride Adsorbed on Alumina

C. Soto, R. Wu, D. W. Bennett, and W. T. Tysoe*Department of Chem istry and Laboratory for Surface Studies University of

Wisconsin-Milwaukee Milwaukee Wisco nsin5321

Received Ma rch 15 1994. Revised Manuscript Received July12 1994@

The species formed by adsorption of trim ethyl alum inum and dimethylaluminum chlorideon y-alumina that has been dehydroxylated at 400 K are identified using infraredspectroscopy. Gas-phase trime thyl alumi num dimer is proposed to adsorb via reaction ofth e bridging methyl groups forming adsorbed dimethylaluminum. A small amou nt of dimeris also found on the surface immediately following exposure to t rimet hyla lumin um, but thisis removed by prolonged evacuation at room temp eratu re. Adsorbed mono- and dimethyl-aluminum species are distinguished by their corresponding peak profiles in the methylbending region where, in the lat ter case, a splitting is found due to coupling between adjacentmethyl species. Relative dimethyl- and monomethylalu minum coverages can be monitoredas a function of anne aling tem per atu re where it is found that dimethylaluminum convertsto monomethylaluminum by reaction with surface hydroxyl groups an d evolving methane.The remaining methyl group in th e monomethylaluminum surface species reacts in a similarman ner w ith surface hydroxyl species to form methane bu t is found to be significantly lesslabile than th e dimethylaluminum species. A similar reaction path way is also found fordimethyl aluminum chloride. The surface species formed from dimethy laluminum chloride,however, react more rapidly t ha n those formed by trimeth ylalu minu m adsorption.

Introduction

Metal alkyls provide an important class of volatileprecursors for the growth of inorganic thin films usinga method generally known a s organometallic chemicalvapor deposition (OMCVD).l-12 These can react withnon-metal hydrides t o form the inorganic layer and t oeliminate methane and a re most often used for growing

films of 111-V compounds. In addition, aluminum oxidecan be grown by a variant of this chemistry by thereaction of trimethylaluminum and water.13 This reac-tion has the advantage, from the point of view offundamental studies, that it is sufficiently fast foralumina layers to be grown by this method both inultrahigh vacuum as well as a s a t higher pressures -1Torr) on high-surface-area alumina substrates. It hasbeen previously demonstrated that the chemistry in

*To whom correspondence should be addressed.@ Abstract published inAdvance AC S AbstractsAugust 15, 1994.

1) Manasevit, H. M.; Simpson,W. I. J . Electrochem. SOC .196911 6 . 1725.

I- .

2) Duchemin, J.-P.; Bonnet,M.; Koelsch, F.; Huyg he, D.J . C y s t .

3) Leys, M. R.; Veenvliet, H.J . Cryst. Growth 1981 5 145.4) Hersee, S. D.; Duchemin,J.-P. Ann u. Rev. Mater. Sc i .1982,12,

Growth 1978, 5 181.

65.5) Dapkus , P. D.; Manasevit, H. M.; Hess. K. L.; Low, T. S.;

Sti l lman , E. G.J . C y s t . G ro w th 1981, 5 10.6) Glew, R. W. . Phys . 1982, C5 281.7) Stringfellow,G. . . Cryst . Growth 1984, 8 111.8) Den Baars,S. P.; Ma a, B. Y.; Dapk us, D.P.; Danner, A. D.; Lee,

9) Larson, C. .; Buchanan, N. I.; Stringfellow, G. B. App l. Phys.

10) Suzuki, H.; Mori, K.; Kawasaki, M.; Sato, H.J . Appl . Phys .

11) Luckerath, R.; Tommack, P.; Hertling, A.; Koss, H. J. ; Balk,

12) Stnngfeellow,G. B rganometallic Vapor-Phase Epitox y: T h e o y

13) oto, C.; ysoe, W. T. J . Vac. Sci . Technol.1991 9 2686.

H . C. J . C y s t . G ro w th 1986 7, 188.

Lett . 1988, 2 480.

1988 4 371.

P. . C y s t . G ro w th 1988 3, 51.

and Practice; Academic Pres Inc.: New York,1989.

both regimes appears to be identical and can, therefore,be followed in ultrahigh vacuum on a planar aluminasubstrate o r alternatively using high-surface-area pow-ders to identify surface species using infrared spectros-copy.l3 Alumina films can be grown in a controlledmanner using this strategy by sequentially exposing thesurface to trimethylaluminum (TMA) and water; eachcycle forming an alumina layer approximately 1

thick.13 The stoichiometry of the reaction during eachcycle is consistent with the initia l formation of adsorbeddimethylaluminum due t o the adsorption of trimethyl-aluminum on hydroxylated alumina. However, th eadsorption of TMA on silica leads to the formation ofmonomethylaluminum as well as the deposition ofmethyl groups on silica This chemistry canbe followed on silica since distinct methyl stretchingmodes due to both Al-CH3 and Si-CH3 can be distin-guished. Such direct experimental evidence is notpossible on alumina surfaces since an analogous reac-tion on alumina leads to two identical Al-CH3 species.To obtain direct experimental evidence for the existenceof either mono- o r dimethylaluminum species followingtrimethylaluminum adsorption on alumina and t o in-

~~ ~

14) Murray, J . ; S harp, M.J.; Hockey, J. A. J . Catal. 1970,18, 2.15) Yates, D. J.; Deinbinski, G.W.; Kroll, W. R.; Elliot, J. J . J

16) eglar, R. J . ; Murray, J. ; Hambleton, F. H.; Sharp, M. J . ;

17) eglar, R. J.; Hambleton,F. H.; Hockey, J. A. J . Catal . 1971,

18) Kunawicz, J.; Jones, P.; Hockey, J. A.Trans . Faraday SOC .

19) ow, M. J. D.; Se verdia, A. G.; Chan.J. J . Catal. 1981, 9

20) Kinney, J B.; taley, R. H. J Phys . Chem. 1983 7, 735.21) Morrow, B. A.; McFa rlane, A. J. J . Non-Cys t . So l ids 1990

22) ertram, M. E.; Michalske, T. A.; Rogers,J. W . Jr. J . Phys .

Phys . Chem. 1969 3, 11.

Parker, A. .; Hockey,J. A. J . C h e m . S O C . 1970, 170.

20 309.

1971, 7 848.

384.

120 61.

C h e n . 1991,95,4453.

0897-4756/94/2806-1705 04.50/0 994 American Chemical Society


2/7

1706

vestigate their relative reactivities, we take advantageof the splitting in the methyl bending modes of adimethyl species compared to monomethyl group resu lt-ing from coupling between these modes. This is a well-known effect in the infrared analysis of organic com-pounds where a splitt ing of -10 cm-l in th e bendingregion of the spectrum is found for terminal methylgroups in 1,l-dimethylalkanes compared to n-

This effect is used t o identify the natu reof the surface species formed on hydroxylated aluminafollowing trimethylaluminum adsorption where high-quality, high-resolution infrared spectra can be collectedon the high-surface-area substrate.

Chem. Mater. , Vol. 6 No. 10 994 Soto et al

Experimental Section

The experimental apparatus used t o obtained infraredspectra has been described in detail in a previous p~b1ication.l~Briefly, a pressed y-alum ina pellet is mounted to t he end of atransfer rod which is enclosed in an evacuable cell whichoperates at a base pressure of -1 x Torr. They-alumina is degassed in vacuo for 90 min to remove molecularwater . The magnetically coupled sample mounting rod canbe moved so th at the alumina pellet is located eit her betweentwo sodium chloride windows to allow infra red spectra t o be

collected o r t o be retracted into a tube furnace where thesample can be heated up t o -1200 K in vacuo. The infraredcell is also connected to a gas-handling line to allow reac tant sto be introduced into the cell. Infrar ed spectr a were collectedusing a Midac Fourier transform infrared spectrometer gener-ally using a resolution of 0.5 cm-l for 1000 scans, and thespectrum due to the background was subst rated from thesignal due to the pellet. Initial infrared spectra showed abroad feature due to surface hydroxyl species along with smallpeaks between 1300 and 1700 cm-l which may be due to eithera small amount of surface carbonate o r residual water. Thespectrum of the gas-phases species could be measured usingthe s ame apparatu s merely by retracting t he sample from thepath of the infrared beam.

Both trimethylalumin um (Aldrich Chemicals; 97 ) anddimethylalu minum chloride (Aldrich Chemical, 97 ) weretransferred into glass vials inside a nitrogen-filled glovebagusing a syringe. Both were purified by repeated freeze/pump/thaw cycles and their purities determined using infraredspectroscopy. The major contam inant in both compounds wasmethane presumably arising from reaction with atmosphericwater.

Results

Figure 2 displays the infrared spectra of gas-phasetrimethylaluminum (TIM) and dimethylaluminum chlo-ride DMA) showing the methyl stretching region (be-tween 2790 and 3000 cm-l) and the methyl bendingregion (between 1150 and 1300 cm-l). Ball-and-stickdepictions of dimeric trimethylaluminum and dimethyl-aluminum chloride are shown in Figure 1. The generalappearance of the methyl stretching regions is es-sentially identical, showing three peaks due to thesymmetric and asymmetric methyl stretches and aFermi resonance at lower frequencies. The asymmetricstretch for TMA is a t 2944 cm-l, whereas for dimethyl-aluminum chloride it is a t a somewhat higher frequency

23) heppard, N.; Simpson,D. M. . Reu. 1953 , 19.24) Rao, C. N. Chemical Applications of Infrared Sp ectroscopy;

Academic Press Inc.: New York,1963.25) ilverstein, R. N.; Bassler,G. C.; Morril, T. C. Spectroscopic

Identification of Organic Com poun ds 5th ed.; John Wiley and Sons:New York, 1991.

26) Suther land,G. . B. M; Simpson,D. M. J . Chem. Phys . 1947,15 153.

Figure 1. Ball-and-stick depictions of (a) rimethylaluminumdimer and (b) dimethylaluminum chloride dimer.

Gas-phase T M A an d \

D M A , in f r a r e d s p e c t r a/

M e t h y l b e n d s 1r

\ TM AL.

L 1- --

1250 1200/7

\

1 M e h y \

s t r e t c h e s

/

\ / '\ --./- . T M A.-

1 I I

roo0 2 9 2800Figure 2. Infrared spectra of gas-phase trimethylaluminum(TMA) and dimethylaluminum chloride DMA) showing themethyl bending region between 1150 and 1300 cm-l and themethyl s tretching region between 2790 and 3000 cm-l.

of 2960 cm-l. The symmetric stretch is a t 2901 cm-'in the case of trimethylaluminum and is at a slightlyhigher value of 2907 cm-l for dimethylaluminum chlo-ride. The Fermi resonances appear a s a broad peak at-2835 cm-1.27-29 Shown also in this figure are th ebending regions for both TMA and DMA and, in theformer case, the spectrum exhibits peaks at -1254 cm-land a feature at -1207 cm-l which consists of two peaksseparated by -5 cm-l. This will be discussed in greaterdetai l below. The corresponding region for DMA shows

27) Herzberg, G. olecular Spe ctra and Molecular Structure; VanNostrand R einhold: New York,1989; Vol 11.

28) Bennett , W. H.; eyer, C. F. Phys. Rev. 1928,3 2, 88.29) Adel, A.; Barker,E. F. J . Chem. Phys . 1964,2, 27.


3/7

M e d l and M eAlCl Adsorbed on Alumina

I I I

Chem. Mater., Vol. 6, No. 10 1994 1707

I I

~

G a s - p h a s e T M A ,i n f r a r e d s p e c t r u mTe m p e r a t u r e d e p e n d e n c e

An n ea l i n g

\ Temperature

3 7 3 K/-_

---____1 \----/-+----- -,

4 7 3 K- ---- -_ 1 1

5 2 3 K

/ \_

1250 1200Frequency cm-1

Figure 3 Infrared spectra of the bending region of gas-phasetrimethylaluminum TMA) displayed between 1160 and 1280cm-l as a function of tempera ture. The sample temp erat ureis displayed adjacent to the corresponding spectrum.

only a single feature a t a frequency of -1209 cm-l; the1254 cm-l feature is completely absent. Since TMAexists as a dimer in t he gas phase a t room temperature,it contains two chemically distinct methyl species, theterminal methyl groups and the bridging methyl groups.Chlorine is the bridging group in DMA so tha t the peakat 1254 cm-l is assigned to the bending modes of thebridging methyl group and the 1200 cm-l modes to theterminal groups. This serves to point out the relativesensitivity of these bending modes to changes in th enature of the methyl groups. This assignment is further

confirmed since the integrated intensity under the 1254cm-l peak is 0.51 0.03 of th e integrated intensity ofthe 1207 cm-l stat e and therefore scales well with theirrelative abundances. The differences in methyl stretch-ing frequencies for TMA and DMA (Figure 2) alsosuggest tha t th e terminal methyl groups are modifiedby the presence of a bridging chlorine species. Thebending modes appear to be less sensitive to this effect.

Features due to the terminal methyl groups a t -1200cm-l, in fact, consist of two states a t 1207 and 1203 cm-(spli t by -5 cm-l). It is proposed tha t this splitting isdue t o coupling between adjacent methyl groups con-sistent with the existence of aluminum dimethyl species.To ensure th at thi s splitting is not due to the combina-tion of low-frequency skeletal modes with t he bendingmodes (i.e., hot bands), the gas-phase TMA spectrumwas recorded, using exactly identical spectrometer set-tings, as a function of temperature and the results aredisplayed in Figure 3. Here, both th e 1254 and 1207cm-l peaks decrease in intensity as the temperatureincreases but the spectral profile remains essentiallyconstant, with perhaps some loss in fine structure.These result s indicate th at th e peak splitting is not dueto hot bands and clearly indicates the existence of twostates.

Figure 4 shows the methyl stretching region (between2700 and 3300 cm-l) for TMA adsorbed on alumina th atha s been pretreated a t 400 K showing th e characteristic

T M A IA 1 20 3, i n f r a r e d s p e c t r u mn

2 9 3 K

3 7 3 K


4/7


5/7

Medl and MefilCl Adsorbed on Alumina

adsorbed monomeric species. This could thereforeeither be Al(CH3)3, Al(CH3)2, or Al(CH3). The trimethyl-aluminum monomer (Al(CH3)3) does exist in t he gasphase but only at high temperatures and is thereforeunlikely to be present on the alumina a t room tempera-ture. In addition, significant methane evolution is notedat room temperature whereas none would be expectedfor a surface consisting entirely of Al(CH313. The surfacespecies therefore could either be mono- o r dimethyl-aluminum groups. The methyl stretching feature shifkfrom 1208 to 1218 cam-l on heating the surface fromroom temperature to -398 K and becomes significantlynarrower as well as decreasing substantially in inten-sity. Since the final removal of all methyl groups fromth e surface is immediately preceded by t he formationof a Al(CH3) species, the narrow peak centered a t 1218cm-l (Figure 5) is assigned to this monomethyl species.This is consistent with assignments made el ~ ew h er e. ~ la ~The broader feature centered a t 1208 cm-l found afteradsorbing TMA at room temperature (Figure 5) istherefore assigned to an aluminum dimethyl species. Asshown by the gas-phase TMA and DMA data (Figure2), the presence of two methyl groups bonded t o thesame aluminum causes a splitting of -5 cm-l betweenthe two modes. This splitting is not well resolved inthe spectra for TMA adsorbed on alumina (Figure 5) andmerely results in a broadening of the peak. Neverthe-less, the observation of a broad asymmetric feature isconsistent with this assignment. The spectra of Figure5 therefore indicate th e initial formation of a dimethylsurface species which converts to a monomethyl specieson heating. This process will be addressed in greaterdetail below. Other explanations for this shift, such asadsorbate-adsorbate interactions, are precluded sincesolvent shifts for even very polar groups like C=N areonly -10 cm-l between the gas and liquid phases.33

It is proposed that the TMA dimer initially adsorbson hydroxylated alumina preferentially by reaction with

the bridging methyl groups (evolving methane) to yieldtwo adjacent (CH3)2Al= species. The aluminum coor-dination environment in this species is currently beingexamined using magic-angle spinning nuclear magneticresonance spectroscopy. One of the methyl speciesreacts further with surface hydroxyl groups eliminatingfurther methane t o yield a monomethyl species. It isclear fiom the da ta of Figure 5 that the dimethyl speciesis more reactive than the monomethyl species sincemonomethylaluminum remains on the surface at higherannealing temperatures. A similar difference betweenthe reactivity of the fi rst and second methyl group hasbeen observed previously in studies of the reaction ofthe surface dimethyl species with water.13 The forma-

tion of monomethyl species from the dimethyl groupcould occur in two possible ways. Fir st, one of themethyl species could react directly with an adjacenthydroxyl group and evolve methane to leave the mono-methylaluminum on the surface. Alternatively, it couldreact in a manner analogous to th at seen on silica wherea dimethyl group rapidly decomposes yielding bothAl(CH3) and Si(CH3) surface species. To distinguishbetween these possibilities, the spectra in Figure 5 arefit to the profiles due to a dimethyl and a monomethylspecies. The aluminum dimethyl fingerprint is taken

Chem. Mater., Vol. 6, No. 10 1994 1709

33) Thomas, B. H.; Orville-Thomas,W. J. J. Mol. Struct. 1969 3191.

TMA I A l 2 0 3 F i e d s p e c t r a

Exp e r m e n a l

1160 1180 1200 1220 1240F r e q u e n c y c m - l

Figure 7 Fits of infrared profiles of C H ~ ) Z A ~ ~ ~ ~ )nd CH3)-a ) to the spectra of the bending regions shown in Figure 5.

to be the profile shown at the top of Figure 5 (for asample that has been evacuated for 10 min followingTMA exposure), and for the monomethyl species, theprofile shown in Figure 5 for a surface heated t o 398 K.The integrated area of the monomethyl profile is takento be 50 of the dimethyl species, reflecting the relativemethyl stoichiometries. The relative contributions ofthe component species were determined using multi-component linear regre~sion.~~ n this technique, theexperimental spectrum is modeled as a linear combina-tion of the component spectra and correlation coef-ficients in excess 0.99 were obtained for th e fits in allcases. Typical fits to th e spectral profiles are displayedin Figure 7 and th e resulting relative coverages 0) reshown plotted in Figure 8 as a function of the annealingtemperature. Note that the 293 K point i n this curveis for the spectrum taken following TMA adsorption andevacuation for 30 min (Figure 5; second spectrum) andindicates some Al(CH3) formation even a t room tem-perature. If a single surface dimethyl species decom-poses to yield two surface monomethyl species, thecoverage of monomethyl groups should substantiallyexceed that of the original dimethyl groups as the

sample temperature is raised. Alternatively, if thedimethyl species react with surface hydroxyl groupsforming monomethyl species and evolving methane, themonomethyl coverage should never exeed that of theinitial dimethyl species. The data in Figure 8 indicatethat the relative coverage of the monomethyl speciesnever exceeds unity and th e total relative coverage, infact, remains constant at -1.0 up to an annealingtemperature of -400 K. This indicates that the di-methyl species react with surface hydroxyl groups to

34) Osten, D. W. In Computerized Quantitative Infrared Analysis;McClure, G . L., Ed.; American Societyfor Testing and Materials:Phildadelphia,1987; p 6.


6/7

1710 Chem. Mater., Vol. 6 No. 10 994 Soto et al.

C o v e r a g e of m o n o - a n dd i m e t h y l a l u m i n u m , T M A / A l 2 0 3

Te m p e r a t u r e/K

Figure 8. Plots of the coverages of (CHS)fl(ads) nd (CH3)-A(&) n alumina tha t has been pretreated at 400 K followingtrimethylaluminum TMA) adsorption a t 300 K as a functionof annealing temperat ure.

form the monomethyl group. The monomethyl groupis significantly less labile than dimethyl since it remainson the surface up to -550 K.

A similar plot of relative surface coverages is dis-played in Figure 9 for dimethylaluminum chlorideadsorbed on alumina (Figure 6). An analogous effectis observed in this case since the total coverage doesnot exceed unity. The major difference is that both thedimethyl- and monomethylaluminum species react morerapidly when formed from dimethylaluminum chloridethan when formed from trimethylaluminum.

Note that no significant change is observed in theshape of the spectrum due t o the methyl stretchingmodes (Figure 41, in particular, for spectra taken a t 300K where the surface is predominantly covered withdimethylaluminum and at -400 K where the surfaceis covered predominantly with monomethyl species(Figure 8). The plot in Figure 10 indicates th at the datafor the methyl stretching and bending regions areconsistent. This figure displays the relative integratedarea of the methyl stretching modes versus sampleannealing temperatur e m) as well as the correspondingcoverages calculated from the data shown in Figure 8

0 ) . These show an almost linear decrease in totalmethyl coverage as a function of sample temperaturederived from both the methyl stretching and bendingmodes. The agreement between the two sets of dat asuggests tha t the relative mono- and dimethyl coveragesplotted in Figure 8 are reasonable.

Finally, shown in Figure 11 are the proposed struc-tures for both th e mono- and dimethylaluminum surfacespecies. This proposed alumin um dimethyl structureprovides an indication why it is more reactive than t hemonomethyl species. Clearly, the dimethyl species ismore flexible and a methyl group in this species can

Co v erag e of mo n o - an dd i m e t h y l a lu m i nu m , D M A I A l 2 0 3

0 360 440 520Te m p e ra t u re I K

Figure 9. Plots of the coverages of (cH~)&l(~d~) nd (CH3)-A( ) n alumina that has been pretreated at 400 K followingdimethylaluminum chloride DMA) adsorption a t 300 K as afunction of annealing temperature.

To t a l -CH3 co v e rag e , TM AIA1 2 0 3

f r o m m e t h y l s t r e t c h e s0 28 AL CH312)+B Al CH3 1112

f r o m m e t h y l b e n d s

0' 0.82

0 6

Wu

al

U

I I 1n.n? S O 300 350 400 450 500 550

Te m p e r a t u r eI KFigure 10. Comparison of th e tota l methyl coverage followingadsorption of trimethyl aluminum on alumina that has beenpretreated at 400 K as a function of annealing temperature,W ) obtained from the a rea unde r the methyl stretching modes,0 ) btained from the data of Figure 8 from (2@((CH3)&l(ads)

access adjacent surface hydroxyl groups relatively eas-ily. On the other hand, th e monomethyl species issignificantly more rigid and also furthe r away from thesurface so that reaction with an adjacent surface hy-droxyl group is more difficult.

There may be two possible explanations for thegreater reactivity of the surface species derived from

-I- @((CH3)&ads))/2.


7/7

Me and Me AlC l Adsorbed on Alumina

Figure 11. Schematic depictions of the adsorbed dimethyl-aluminum and monomethylaluminum surface species.

DMA th an TMA. Fir st, th e reaction of trimethyl-aluminum to form the surface dimethylaluminum spe-cies involves reaction of the bridging methyl group withsurface hydroxyl groups to evolve methane. This im-plies that the local surface hydroxyl concentration is

Chem. Mater., Vol. 6 No. 10 1994 1711

lower in this case tha n for surface species derived fromDMA. In the latter case, the bridging species arechlorines which will react with surface Lewis sites andtherefore not deplete the surface of surface hydroxyls.Alternatively, the reactivity may be affected due toelectronic interactions because of adjacent chemisorbedchlorine.

Conclusions

Trimethylaluminum reacts with alumina that hasbeen dehydroxylated at 400 K t o yield an adsorbeddimethylaluminum species in which the aluminum is4-fold coordinated immediately following adsorption.This suggests tha t the TMA dimer reacts preferentiallyvia the bridging methyl species and evolves methane.A small amount of adsorbed TMA dimer is found on thesurface immediately after adsorption, although this isremoved by prolonged evacuation a t room temperature.The surface dimethylaluminum species further reactrather rapidly with surface hydroxyls forming mono-methylaluminums. These also react with surface hy-droxyls so th at all surface methyl species are eventuallyremoved, although significantly more slowly th an doesdimethylaluminum. This behavior can be rationalizedon the basis of their corresponding surface structures.

Documents

Tysoe56 the First Publication Trimethyl Aluminium Ftir Reaction on a Surface Al2o3