Porous Alumina Protective Coatings on Palladium Nanoparticles

Porous Alumina Protective Coatings on Palladium Nanoparticles by Self-Poisoned Atomic Layer Deposition

Junling Lu,?Bin Liu,?Je?rey P.Greeley,?Zhenxing Feng,⊥Joseph A.Libera,?Yu Lei,?Michael J.Bedzyk,⊥Peter C.Stair,§,#and Je?rey W.Elam*,?

?Energy Systems Division,?Center for Nanoscale Materials,and§Chemical Sciences&Engineering Division,Argonne National Laboratory,Argonne,Illinois60439,United States

⊥Department of Materials Science and Engineering and#Department of Chemistry,Northwestern University,Evanston,Illinois 60208,United States

*Supporting Information

transforms into Al(OH)3*species during the

support this TMA dissociation/hydration

which the Pd surface becomes poisoned by

complete monolayer of adsorbed Al species.

the net result is a porous Al2O3?lm.This

gases with access to the Pd surface sites,

stabilization,alumina,growth mechanism,

Atomic layer deposition(ALD)was developed in the late1970s to meet the requirements for the growth of large-area thin?lms for electroluminescent,?at-panel displays.1,2ALD relies on self-limiting sequential binary reactions between gaseous precursor molecules and a substrate to deposit?lms in an atomic layer-by-layer fashion.An inert gas purge period is introduced between each precursor pulse to prevent mixing of the chemicals,which would cause non self-limiting growth.2?4 Because of the unique feature of self-limiting growth in each deposition cycle,ALD can deposit uniform and conformal coatings regardless of whether the substrate is?at or possesses high aspect ratio features,high surface area,or high porosity.3,5?9As a consequence,ALD has attracted great attention and applications have extended far beyond micro-electronics into?elds such as catalysis,10?17photovoltaics,18?21 batteries,22fuel cells,16,23polymers,3,24,25and microdevices.26,27

applications,well-dispersed and uniform metal nanoparticles(NPs)have been successfully prepared using ALD with precise particle size control.28?31These ALD metal NPs often showed comparable or better catalytic performance than those synthesized by conventional methods,such as impregnation,ion-exchange,and deposition-precipita-tion.11,12,14,16Recently,there have been a number of attempts to stabilize supported metal NPs using ALD metal oxide coatings.13,17,32?34These studies were motivated by the atomically precise control over the thickness and composition of the protective layers a?orded by ALD in comparison with other methods such as chemical vapor deposition,grafting, microemulsion,and dendrimer encapsulation.35?41These less Received:January18,2012

Revised:April11,2012

Published:May8,2012

precise methods can yield overly thick protective shells that impede mass transport and reduce catalytic performance.

Al 2O 3ALD performed using TMA and water is one of the most successful ALD procedures,and has been extensively investigated.4,42?45The mechanism of Al 2O 3ALD on oxide surfaces is well understood:?rst,TMA reacts with hydroxyl groups on the starting surface forming Al(CH 3)x *(x =1?2,where the asterisk designates a surface species)and CH 4(Figure 1a);next,the Al(CH 3)x *terminated surface transforms

to an Al(OH)x *(x =1?2)terminated surface after the following H 2O exposure and again releases CH 4(Figure 1b).4,44,45In situ quartz crystal microbalance (QCM)measure-ments performed during Al 2O 3ALD demonstrated that the ratio of the total mass gain in one ALD cycle (Δm 0)to the mass gain after the TMA pulse (Δm 1)is Δm 0/Δm 1≈1.1,which implies x =1.6according to the mechanism of Figure 1.4,44,46Highly linear growth rates of 1.1?1.3?/cycle were measured on planar surfaces using spectroscopic ellipsometry,with the lower values occurring at higher temperatures where the surface hydroxyl coverage is reduced.4,42,43,47

Although the mechanism for Al 2O 3ALD on oxide surfaces requires surface hydroxyls,Al 2O 3ALD can also be grown on noble metal surfaces.For instance,Zhang et al.demonstrated that a subnanometer thick ALD Al 2O 3layer coated on silver ?lm-overnanosphere (AgFON)substrates can maintain and stabilize the activity of the underlying silver for surface-enhanced Raman spectroscopy (SERS).33,48Based on X-ray photoelectron spectroscopy and in situ QCM results,Whitney et al.suggested that the Al 2O 3ALD initiates when TMA decomposes on the Ag surface.49In a di ?erent study,Liang et al.prepared aluminum alkoxide (alucone)hydride ?lms on Pt surfaces using alternating exposures to TMA and ethylene glycol (EG)and thermally decomposing these ?lms to form highly porous Al 2O 3.34Pt/SiO 2catalysts coated with these porous,ultrathin alumina layers were very stable when calcined in air at 1073K.The 20-cycle alucone-coating (about 1.2nm)reduced the catalytic activity of the Pt NPs in CO oxidization reactions.This is likely due to the small pore size in the alumina layer.34

We recently utilized ALD Al 2O 3protective layers with precise thicknesses to inhibit the sintering of supported nanosized ALD Pd catalysts in the methanol decomposition reaction carried out at elevated temperatures.13Up to a certain thickness,the Al 2O 3protective layers preserved or even slightly enhanced the catalytic https://www.sodocs.net/doc/a210855329.html,ing CO as a probe molecule,we found that the ALD Al 2O 3overcoats preferentially nucleate at corners,steps,and edges of the Pd NPs while leaving the

Pd(111)facets accessible for methanol conversion,and this site preference became more pronounced after reaction testing.Thicker Al 2O 3overcoats (with thicknesses of ～8nm)were further tested on supported Pd catalysts and found to e ?ectively prevent the catalyst deactivation through either sintering or coking in excess of the Tammann temperature.17These remarkable improvements in catalytic performance were again suggested mainly due to the preferential blocking of the more reactive low-coordination Pd sites by the Al 2O 3overcoat.Although signi ?cant improvements in catalytic performance can be achieved by applying ALD Al 2O 3overcoats onto Pd,the mechanism for Al 2O 3ALD on Pd surfaces is unknown.In this work,we have investigated Al 2O 3ALD on Pd surfaces with in situ QCM,in situ quadrupole mass spectrometry (QMS),and transmission electron microscopy (TEM).In particular,we sought to understand how the catalytic activity of the Pd NPs was maintained even after many ALD Al 2O 3cycles that would be expected to bury the Pd surface.Density functional theory (DFT)calculations were also adopted to further elucidate the details of TMA adsorption and the subsequent hydration process on both Pd(111)and Pd(211)surfaces.

■

EXPERIMENTAL SECTION

ALD Reactor.ALD was performed in a viscous ?ow stainless steel tube reactor system.46Ultrahigh purity nitrogen (99.999%)carrier gas continuously passed through the tube reactor at a mass ?ow rate of 300sccm and a pressure of 1Torr.The ALD reactor was equipped with a quadrupole mass spectrometer (QMS,Stanford Research Systems RGA300)located downstream of the ?ow tube in a di ?erentially pumped chamber separated from the reactor tube by a 35μm ori ?ce and evacuated using a 50L/s turbomolecular pump.A quartz crystal microbalance (QCM)was mounted in a commercial QCM housing modi ?ed to allow a nitrogen purge that prevents growth on the back of the sensor.The QCM was installed in the middle of reaction tube for in situ monitoring of the ALD.

Al 2O 3and Pd ALD.Al 2O 3ALD was carried out by alternately dosing TMA (Sigma-Aldrich,97%)and deionized water at 473K.Pd ALD was performed by alternately dosing Pd(II)hexa ?uoroacetyla-cetonate (Pd(hfac)2,Sigma-Aldrich,>97%)and formalin at 473K.The Pd(hfac)2precursor was contained in a stainless steel bubbler heated to 333K to increase the vapor pressure.28,50Ultrahigh purity nitrogen with a ?ow rate of 50sccm passed through the bubbler and carried the Pd(hfac)2precursor to the reaction chamber.Formalin is a solution of 37%formaldehyde in water with 10?15%methanol added as a stabilizer.The precursor inlet lines were heated to 423K to prevent condensation of the ALD precursors.To de ?ne the ALD cycles,the ?rst precursor pulse time,the ?rst nitrogen purge time,the second precursor pulse time,and the second nitrogen purge time are expressed as t 1?t 2?t 3?t 4,in seconds (s).

In situ QCM and QMS Measurements.In situ QCM and QMS measurements were performed to study the Al 2O 3growth on Pd surfaces and to monitor the reaction products evolved during each half ALD cycle,respectively.Prior to these measurements,300ALD Al 2O 3cycles were performed using the timing sequence 1?5?1?5to deposit ～30nm Al 2O 3on the QCM sensor as well as on the inner surfaces of the ALD reactor that generate a majority of the species detected by the QMS.Next,150ALD Pd cycles were performed using the timing sequence 2?2?2?2to deposit a high density of Pd NPs.50After preparing the supported Pd NPs,50ALD Al 2O 3overcoating cycles were performed using the timing sequence 2?10?2?10and the process was monitored by QCM.For the in situ QMS studies,the TMA and water doses of the ALD Al 2O 3overcoating cycles were divided into 10×0.5s pulses to capture the temporal evolution of the gaseous products during each half ALD cycle.Nitrogen purge periods of 10s were used after each TMA and water pulse.The QMS signals for all of the gaseous products were calibrated using standard

gases

Figure 1.Schematic illustration of one Al 2O 3ALD cycle on a metal oxide surface.(a)TMA reacts with surface hydroxyl groups liberating methane;(b)water reacts with the Al(CH 3)x -terminated surface forming methane and regenerating the hydroxylated surface.

(1000ppm of hydrogen,methane and ethane in nitrogen,Air Liquide).

Morphology of ALD Al 2O 3Overcoated Pd NPs.To character-ize the ultrathin ALD overcoatings on the Pd NPs,high resolution transmission electron microscopy (TEM)was performed using spherical Al 2O 3powder (Al 2O 3NanoDur,99.5%,Alfa Aesar)as the starting support.About 600mg of Al 2O 3support was loaded into the ALD reactor.Four cycles of Pd ALD (300?300?300?300)were ?rst performed after a 10min ozone cleaning treatment and 20min stabilization in the ?ow of nitrogen at 473K.28Next,di ?erent numbers of ALD Al 2O 3overcoating cycles were performed with the timing sequence (60?180?120?180).13The ALD Al 2O 3overcoated Pd samples were characterized using a JEOL JEM-2100F fast TEM system (NUANCE facility,Northwestern University)operated at 200kV.DFT Calculations.The thermodynamics of the surface species were obtained by performing periodic DFT calculations using the Vienna Ab initio Simulation Package (VASP).51?54The ionic cores were treated with the projector augmented wave (PAW)method.55,56The PW91generalized gradient functional (GGA-PW91)functional was used to describe the electron exchange-correlation interac-tions.57,58

The Pd nanoparticle surfaces were modeled using the idealized (111)and (211)facets to represent the terrace and the stepped regions,respectively.The Pd(111)slab consisted of a 3-layer,p (3×3)unit cell.The Pd(211)slab consisted of a 3-layer unit cell with three atoms included along the step edge.The top layer was allowed to relax in each case.A vacuum equivalent to ?ve metal layers was used between successive metal slabs.The lattice constant was determined to be 3.95?,which compares well with both the experimental and theoretical bulk lattice values.59,60The surface Brillouin zone was sampled with 4×4×1and 3×3×1k-point based on the Monkhorst-Pack sampling scheme for Pd(111)and Pd(211)respectively.61The Kohn ?Sham valence states were expanded in the plane wave basis sets up to 25Ry (or 340eV).The self-consistent iterations were converged with a criterion of 1×10?6,and the ionic steps were converged to 0.02eV/?.The Methfessel-Paxton smearing scheme was used,62with the Fermi population of the Kohn ?Sham state being k B T =0.2eV.The total energies were extrapolated to 0eV.Dipole corrections were included in all cases.Zero-point energy corrections were not incorporated in this work.Gas phase energies were calculated in a box with dimensions of 17×17×18?,using only the gamma-point.The Gaussian smearing parameter was 0.02eV in this case.Spin polarization was included in both gas phase and surface calculations.Transition states (TSs)were calculated using the climbing-image nudged elastic band (CI-NEB)method.63,64The dimer method was also used to further re ?ne the determined transition states.65Each transition state was con ?rmed to have only one imaginary (negative)vibrational mode.The energy barriers were calculated using the lowest energy initial state con ?gurations.If reactions involve more than one reactant,the barriers are reported with respect to the most stable reactant states at in ?nite separation from one another.

Free energies were determined with entropy corrections estimated assuming loss or gain of translational entropies of gas phase species as these species adsorb on,or desorb from,the surfaces at 473K and standard pressure.The translational entropy correlations used for TMA,CH 4,C 2H 6,H 2O,and H 2in this study were 0.84,0.75,0.79,0.76,and 0.62eV,respectively.

■

RESULTS

The Al 2O 3ALD on high-density Pd NP-covered alumina surfaces was studied with in situ QCM and QMS,where 150Pd ALD cycles were performed ?rst to coat the entire ALD chamber and the quartz crystal surface with a high density of Pd NPs.Figure 2a shows the mass gain per cycle recorded using the QCM during the Al 2O 3ALD and exhibits four distinct regions of growth.In region I,the ?rst Al 2O 3ALD cycle yielded a mass gain of ～60ng/cm 2.In region II,the mass gain in the second cycle dramatically decreased to only 29ng/cm 2and was followed by a gradual increase to a maxium of 58ng/cm 2at about 13cycles.Next,the mass gain gradually decreased to a value of 37ng/cm 2(region III)and remained at this steady state value (region IV).Figure 2b shows the ratio of the total mass gain in one ALD cycle to the mass gain after the TMA pulse (Δm 0/Δm 1).In the ?rst Al 2O 3ALD cycle,this mass ratio was 1.6,signi ?cantly higher than the value of ～1.1in the subsequent cycles,which is the expected value for Al 2O 3ALD on oxide surfaces (also seen in Figure S1in the Supporting Information).44,66

For the in situ QMS studies,we divided the TMA and H 2O exposures into ten pulses in each Al 2O 3ALD cycle to investigate the detailed temporal evolution of the reaction products.Figure 3a shows an expanded view of the gaseous reaction products released during the ?rst ALD cycle on the Pd NP surfaces.In this ?gure,the QMS signals for each species have been scaled by their experimentally determined calibration factors so that the traces represent the relative partial pressures for each species.Surprisingly,a small amount of C 2H 6was observed (about 2.1%,m /e =30)along with the dominant CH 4product (m /e =16)during the ?rst TMA pulse.In the following nine TMA pulses,the C 2H 6signals dramatically decreased to the noise level,and the CH 4gradually decreased to a constant value consistent with the cracking pattern of the TMA.It was also surprising that during the H 2O pulses,a small amount of H 2was formed (m /e =2)along with the dominant CH 4product,which both decreased to the noise level after the ?rst and fourth H 2O pulses,respectively.Figure 3b shows a compressed view of the ?rst 9ALD Al 2O 3cycles on the Pd NP surfaces and demonstrates that C 2H 6was only observed during the ?rst cycle,whereas H 2was observed for about 7cycles

with

Figure 2.In situ QCM measurements of Al 2O 3ALD on a Pd NP-coated sensor.(a)Al 2O 3mass gain per cycle,where the growth can be divided into 4regions as described in the text;(b)the Δm 0/Δm 1step ratio for each cycle.

a gradually decreasing intensity.Meanwhile,the total CH 4product (CH 4_total)in each cycle increased with increasing number of ALD cycles.

Figure 4shows more clearly the evolution of the CH 4QMS signals during the TMA and H 2O exposures during the Al 2O 3ALD.The ratio of CH 4_total/CH 4_TMA (the total CH 4released divided by the amount released during the TMA pulse)in Figure 4a shows a steady state value of ～2after ca.5cycles that is consistent with previous measurements and implies that ～1/2of the CH 3ligands are released during TMA

adsorption on the ALD Al 2O 3surface (Figure 1).44,66In contrast,higher mass ratios are seen in the initial 5cycles,suggesting that more than 1/2of the CH 3ligands remain on the surface when TMA reacts on Pd.Figure 4b shows that the total amount of CH 4released during each cycle gradually increased to a maximum at ～10?15cycles and subsequently decreased.Since CH 4_total is proportional to the amount of Al 2O 3deposited,this observation is consistent with the QCM data in Figure 2a.

High-resolution TEM images provide direct evidence for the ALD Al 2O 3overcoats on the Pd NPs (Figure 5).The Pd NPs formed using 4ALD Pd cycles have a diameter of ～3nm.Prior to the ALD Al 2O 3overcoating,the NPs exhibit a sharp interface with the vacuum of the TEM environment (Figure 5a,inset).The thickness of the ALD Al 2O 3overcoats grew linearly with the number of ALD cycles at a growth rate of 0.16nm/cycle (Figure 5f).This value is consistent with the ALD Al 2O 3on BN NPs reported in the literature,5,67and the linear growth agrees well with our in situ QCM measurements (see Figure S2in the Supporting Information).The detailed structure of the amorphous Al 2O 3over layers,such as porosity,however,cannot be resolved by these TEM images.

The results of the DFT calculations are presented in Figure 6,which shows the free energy changes for TMA dissociative adsorption on Pd surfaces during the TMA pulse (Figure 6a),together with the hydroxylation of the Al during the subsequent water pulse (Figure 6b)calculated at 473K and standard pressure.Gas phase TMA,a clean Pd surface (either Pd(111)or Pd(211)),and the adsorbed atomic H (1/9ML H coverage)are used as the reference states.

On Pd(111),TMA adsorbs most strongly on the 3-fold (hcp)site (adsorption at fcc sites is only slightly less stable).Two of the Al ?C bonds become elongated from 1.97?in the gas phase to 2.34?upon adsorption.The dissociation to dimethylaluminum (DMA,Al(CH 3)2*)and CH 3*(on top sites),represented by eq 1,is 0.38eV exothermic,with an energy barrier of 0.16eV for the Al ?C scission.DMA also prefers the hcp site (again,fcc adsorption is only slightly less stable)in a tilted position with one of the Al ?C bonds elongated to 2.36?(see Figure 1S in the Supporting Information for the schematic geometries).DMA can further dissociate into AlCH 3*(methylaluminum,MA,which also has a slight preference for hcp adsorption)and CH 3*,which is 0.37eV exothermic (eq 2below).The energy barrier for the second Al ?C bond scission is also 0.16eV.We also note that a direct,concerted pathway for TMA conversion to MA,with a single transition state and a low barrier of 0.11eV,also exists,as indicated by the dashed line between states 2and 4.It is not favorable to further convert MA to Al *and CH 3*(eq 3)on terrace sites,a process which is endothermic by 0.85eV with an energy barrier of 1.37eV.

*+*→*+*

Al(CH )Al(CH )CH 33323(1)*+*→*+*Al(CH )Al(CH )CH 3233(2)*+*→*+*

Al(CH )Al CH 33(3)

Cleavage of Al ?C bonds in Al(CH 3)3*and Al(CH 3)2*is more thermodynamically favorable on the step site of a Pd(211)surface.Both Al(CH 3)3*and Al(CH 3)2*are very unstable on the steps;in fact,TMA directly dissociates into Al(CH 3)*at the step edge site via the path indicated by the dashed line in Figure 6a.Further,the predicted barrier to break an Al ?

C

Figure 3.In situ QMS measurements of Al 2O 3ALD performed after coating the inner surfaces of the ALD reactor with Pd NPs.The TMA and H 2O pulse sequences are indicated to correlate the QMS signals with the precursor exposures.(a)An expanded view of reaction products generated during the ?rst ALD cycle.The green ?lled areas under the dashed lines designate the CH 4background signals.(b)Reaction products formed during the ?rst nine ALD Al 2O 3

cycles.

Figure 4.Summary plots of the in situ QMS measurements of Al 2O 3ALD on Pd NP surfaces.(a)The ratio of the total CH 4formed to the CH 4formed during the TMA exposure in each ALD cycle;(b)the total amount of methane formed in each ALD cycle.The red curves are intended to guide the eye.

bond in DMA (eq 2)is also very low,resulting in stable Al(CH 3)*(adsorbed on the step edge)and CH 3*(also preferentially adsorbed on the step edge).In contrast to Pd(111),the relatively small free energy increase (～0.1eV)for reaction 3suggests that some Al(CH 3)*might further dissociate into Al *and CH 3*on step edges;the barrier for this process is 0.43eV.The free energy of Al(CH 3)*adsorption on the step edge site is lower than that on Pd(111)by approximately 0.81eV.The DFT results thus demonstrate that the step sites are much more thermodynami-cally favorable for TMA dissociative adsorption than are the Pd(111)terraces.These calculations are consistent with our previous CO chemisorption measurements showing that the ?rst cycle of ALD Al 2O 3preferentially nucleated at the low-coordinated Pd NP sites.13

The dissociated methyl groups,CH 3*,bind relatively strongly to Pd(111)and Pd(211),with binding energies of ?1.88and ?2.03eV,respectively,calculated relative to a gas phase methyl radical and a clean palladium surface.It is thus possible that these species will remain on the surface and become abundant under the conditions used in our study.Literature results suggest that additional decomposition of the CH 3*species is not expected.Adsorbed CH 3*on Pd(111)was found experimentally to be thermally stable to at least 440K,as shown in methanol decomposition studies.68,69Additionally,DFT calculations by Paul and Sautet on CH x *fragments on Pd(111)surfaces showed an endothermic path for the decompositions of CH 3*.70

The DFT results can be used to rationalize the gas phase products observed by in situ QMS.Methane released during the TMA exposures (Figure 3)results from CH 3*combining with preadsorbed H *(eq 4,see states 6and 8in Figure 6a).The ethane seen during the ?rst TMA pulse could result from the coupling of two adjacent CH 3*species according to eq 5(corresponding to state 7in Figure 6a).The formation of methane and ethane would be assisted by the entropy gained upon desorption.The relatively small ratio of ethane/methane observed by in situ QMS is likely due to the higher energy barriers for C ?C bond formation (1.58and 1.02eV

on

Figure 5.TEM images of spherical alumina supported Pd catalysts with di ?erent numbers of Al 2O 3ALD overcoating cycles (insets show higher magni ?cation images).(a)0cycle Al 2O 3;(b)5cycles Al 2O 3;(c)10cycles Al 2O 3;(d)15cycles Al 2O 3;(e)20cycles Al 2O 3;(f)thickness of Al 2O 3overcoats versus ALD

cycles.

Figure 6.Free energy diagrams of surface species formed during (a)exposure to TMA and subsequently to (b)water on Pd (111)(dark blue)and Pd(211)(magenta)surfaces calculated at 473K and 1atm.The gas phase TMA,clean surface,and a preadsorbed H atom are used as the reference state (black bar at 0eV).Energy barriers are not shown in the diagrams.Dashed lines in panel a indicate the formation of MA from TMA via two simultaneous Al ?C bond scissions.Schematic representations of the surface geometries for these species are shown in Figure S3in the Supporting Information.

Pd(111)and Pd(211),respectively,see Table1)compared to C?H bond formation(with energy barriers of0.67and0.63eV

on Pd(111)and Pd(211),respectively,in good agreement with literature values.71).Nevertheless,this small amount of ethane does support the existence of a large concentration of CH3*on the Pd surface from TMA dissociation.72?74Compared to CH3*,the coverages for both CH x*(x=0?2)fragments and H*should be low because no hydrogen was seen by in situ QMS during the TMA exposures,and because the decom-position of CH3*is suggested to be endothermic70(see also discussion above).The CH3*radicals that remain on the surface after the TMA exposure would act as a poison to block further TMA adsorption.

*+*→+*

CH H CH(g)2

34(4)

*+*→+*

CH CH C H(g)2

3326(5) Below,we brie?y summarize the DFT-determined thermo-dynamics of the subsequent reactions of Al(CH3)*and CH3* with water or water derivatives during the water exposure to the Pd(111)and Pd(211)surfaces(Figure6b).The adsorbed structures are schematically represented in Figure S3in the Supporting Information.The transformation of Al(CH3)*into Al(OH)*,accompanied by the formation of methane is represented by eq6.The free energy change of this reaction is highly exothermic,by～1.3eV,on both Pd(111)and Pd(211)surfaces.Further reactions with water to insert hydroxyls and form Al(OH)2*and Al(OH)3*(eqs7?10), are found to be thermodynamically favorable on both Pd(111) and Pd(211)surfaces.In fact,Al(OH)3*was found to be the most thermodynamically stable intermediate on the Pd surfaces (Figure6b).Reactions involving further O?H bond cleavage to form O x Al(OH)3?x*species(x=1?2)are endothermic.

*+→*+

AlCH H O(g)Al(OH)CH(g)

324(6)

*+→?*

Al(OH)H O(g)H O Al(OH)

22(7)?*+*→*+*

H O Al(OH)Al(OH)H

22(8) *+→?*

Al(OH)H O(g)H O Al(OH)

2222(9)

?*+*→*+*

H O Al(OH)Al(OH)H

223(10) *+*→+*

H H H(g)2

2(11) As shown in eqs8and10,atomic H*is produced during the hydration of the Al(OH)x*species(x=1,2).We note that this mechanism for cleaving H?OH bonds in water has a lower barrier than direct dissociation of water on palladium surfaces (see Table1for comparison of water dissociation on clean surface and in Al(OH)x*complexes).These H*could react with either CH3*species to release methane,(eq4),or combine with an additional H*to form molecular hydrogen (eq11).These results are consistent with the observation of methane and small amount of hydrogen during the water pulses during the in situ QMS experiments(Figure3).

The thermochemistry for the reaction of water with the Al(CH3)*and CH3*species on the planar(111)and stepped (211)are qualitatively similar in that the substitution and additional hydration reactions of Al(OH)*are all exothermic. However,these reactions are much more thermodynamically

favorable on the step sites of the Pd(211)surfaces.■DISCUSSION

In situ QCM and QMS are valuable tools for exploring the surface chemistry during ALD processes.These techniques are especially useful in our study for monitoring the initial cycles of Al2O3ALD where the TMA and water interact directly with the Pd surface.In our experiments,the150-cycle Pd ALD pretreatment yielded an average Pd coverage of～6ML (～14.7?).Previous scanning electron microscopy(SEM) studies demonstrated that this6ML coating is comprised of densely packed,discrete Pd NPs.50This morphology will increase the Pd surface roughness compared to a smooth?lm as illustrated by the schematic model in Figure7a.If the Al2O3 nucleated uniformly on this rough Pd surface,the QCM would show a high initial mass gain per cycle that gradually decreased to the expected steady-state value of37ng/cm2per cycle as the nanoscale roughness was eventually?lled by the ALD Al2O3?lm.However,Figure2a shows a much di?erent behavior.The Al2O3mass gain during the?rst cycle was60ng/cm2, signi?cantly higher than the steady-state value.Next,the mass gain increased gradually from an initial value of～29ng/cm2 (below the steady state value),reached a maximum at～13 cycles,and then decreased to the steady-state value.

The evolution in QCM mass changes can be explained by incomplete Al2O3nucleation and changes in substrate surface area.The behavior in regions II?IV of Figure2a resembles “type-2”substrate-inhibited ALD in which growth initiates at discrete sites forming islands that eventually coalesces to form a continuous?lm.4,75Here the much more pronounced maximum value at～13cycles compared to that in the“type-2”model is a consequence of the rough starting Pd surface in our case(Figure7a).This similarity indicates that the Al2O3 initiates nonuniformly on the Pd surface(Figure7b,c)in agreement with our previous DRIFTS studies of CO chemisorption that showed preferential nucleation in the?rst ALD Al2O3cycles only at the low-coordination Pd NP sites.13 The DFT calculations support this interpretation since the free energy for MA adsorption is～0.81eV stronger on the Pd(211) step surface than on the Pd(111)hcp sites.

With increasing ALD cycles,the Al2O3patches grow in3 dimensions,accompanied by the continuous Al(OH)3* nucleation(Figure7d).The continuous nucleation in this Region II will be discussed later.Consequently,the surface area increases causing the mass gain per cycle to grow(Region II in Figure2a)and reach a maximum as the Al2O3patches begin to coalesce.Afterward,the Al2O3growth changes to a layer-by layer growth mode.Meanwhile,the growth per cycle decreases as the alumina patches merge into a continuous?lm(Figure2a,

Table1.Activation Energy Barriers(in eV)for Methane, Ethane Formation,and Water Dissociation on Pd(111)and Pd(211)Surfaces(without zero-point energy corrections)a

Pd(111)Pd(211) CH3*+H*→CH4(g)0.670.48

CH3*+CH3*→C2H6(g) 1.58 1.02

Water Dissociation:H2O*→H*+OH*

Pd(111)Pd(211) on clean surface 1.05 1.19

In H2O?Al(OH)*0.580.52

In H2O?Al(OH)2*0.580.39

a Associative desorption barriers are referenced to the adsorbed reactants at in?nite separation from one another.

region III and Figure 7e),and eventually the roughness disappears so that the mass gain in each cycle reaches a constant value (Figure 2a,region IV,and Figure 7f).

Overall,the in situ QMS data in Figure 4support the island coalescence mechanism described above.Beyond ～5cycles,the CH 4product ratio of ～2(Figure 4a)is consistent with the conventional mechanism for Al 2O 3ALD where 1/2of the CH 3*species are lost during each half-reaction (Figure 1).Furthermore,the overall increase and then decrease in the total CH 4QMS signals (Figure 4b)match well with the QCM data (Figure 2a)after the ?rst ALD cycle.However,there is an apparent discrepancy in that the QCM mass increase during the ?rst ALD cycle on Pd is anomalously large while the corresponding CH 4QMS signal is nominal.Furthermore,the QCM mass ratio is anomalously high during the ?rst cycle (Figure 2b),and the QMS mass ratio remains above the steady state value for the ?rst ～5cycles.These apparent discrepancies,as well as the unusual gas phase products observed during the initial stages of Al 2O 3ALD on Pd,can be explained by the unique chemistry for the TMA and water reactions on the Pd surface.As suggested by the DFT calculations,TMA can nucleate on Pd surfaces through Al ?C bond scission (eqs 1-2)without the need for surface hydroxyl species.The observed ethane directly supports TMA dissociative adsorption on the Pd,since ethane can form through the coupling of two CH 3*at

higher coverages.72?74This is also consistent with the CH 4_total to CH 4_TMA ratio showing higher values in the ?rst ?ve cycles than the remaining cycles (Figure 4a).On the other hand,the DFT calculations showed that the energy barriers to form C 2H 6on the Pd(211)and Pd(111)surfaces are 1.02and 1.58eV,respectively,(Table 1),suggesting that the C 2H 6product observed in the ?rst cycle is likely generated through the coupling of two adjacent CH 3*species at the under-coordinated Pd sites rather than on the Pd(111)terraces.Therefore,it is not a surprise that we observed C 2H 6only in the ?rst cycle (Figure 3b),because the under-coordinated Pd sites become occupied by Al(OH)3*species after the ?rst ALD Al 2O 3cycle.

The signi ?cantly higher Δm 0/Δm 1mass ratio of 1.6in the ?rst cycle (Figure 2b)indicates that the surface species produced when water interacts with the TMA-treated Pd surface are di ?erent compared to TMA-treated oxide surfaces.As stated above,Al(OH)3*is the most stable intermediate species identi ?ed by DFT calculations.Therefore,the water exposure in the ?rst ALD cycle should convert the Al(CH 3)*into Al(OH)3*rather than Al(OH)*,according to eqs 6?10.The QCM step ratio resulting from this complete hydrox-ylation will depend on the number of CH 3*species remaining on the Pd after the TMA exposure:Δm 0/Δm 1=2.00,1.41,and 1.08for 1,2,and 3CH 3*species,respectively.Our in situ QCM results showed a ratio of 1.6,suggesting ～2CH 3*species remain on the Pd surface.Table 1lists the energy barriers for water dissociation on clean Pd as well as in the presence of Al(OH)*and Al(OH)2*complexes.The O ?H bond cleavage in water becomes signi ?cantly easier in the presence of under-coordinated Al species.These modeling results provide an explanation for why H 2was only observed during the initial pulses of the H 2O exposures (Figure 3).The H 2is produced by the recombinative desorption of H *species (eq 11).These H *species form through the dissociation of H 2O on under-coordinated Al species (eqs 7?10)which are most abundant near the beginning of the H 2O exposures.The inhibited Al 2O 3growth following the ?rst ALD cycle is likely caused by a large concentration of CH 3*species that block potential TMA adsorption sites.This site blocking is only temporary,since the CH 3*species are released as CH 4during the following water dose,so that the Pd surface becomes exposed again for reaction with TMA in the next cycle (Figure 7b and c).This self-poisoning and self-cleaning process continues for ～7cycles,as demonstrated by the persistent but gradually decreasing H 2production (Figure 3b).This mechanism will allow the continuous nucleation of new Al 2O 3islands on the exposed Pd sites as well as the growth of existing Al 2O 3patches in 3dimensions (Figure 7d).As a consequence,the Al 2O 3will grow as a discontinuous ?lm rather than a continuous,pinhole-free coating typical for ALD.Our recent CO chemisorption studies on ALD Al 2O 3-coated Pd showed that bare Pd sites remained even after 8cycles of ALD Al 2O 3overcoating.13In agreement with this ?nding,the H 2signals that persist for ～7cycles suggest that the Al 2O 3overcoat is su ?ciently porous to allow water to access the Pd NP surfaces (Figure 8).In contrast to our results on Pd,Al 2O 3ALD on silver uniformly blankets the metal surface.33,49This behavior is likely due to the weaker bonding of methyl species on Ag versus Pd,resulting in far fewer CH 3*inhibitors on the Ag.Recently,Weimer et al.synthesized aluminum alkoxide (alucone)hydride ?lms by molecular layer deposition using TMA and ethylene glycol (EG)over Pt/SiO 2catalysts.

Upon

Figure 7.Schematic illustration of Al 2O 3ALD on a surface coated with Pd NPs.The Roman numerals correspond to the labeled regimes in Figure 2a.(a)Initial surface is decorated with a high density of Pd NPs that increase the surface roughness;(b)Region I,nucleation through the TMA dissoicative adsorptioin on the Pd corner and edge sites,forming Al(CH 3)*and CH 3*species;(c)Region I,?rst H 2O exposure to the Al(CH 3)*and CH 3*terminated Pd surface,forming Al(OH)3*species;(d)Region II,island growth in three dimensions as well as additional nucleation;(e)Region III,layer-by layer growth occurs and develops after Al 2O 3patches coalesce to form a continuous ?lm on the rough Pd surface;(f)Region IV,layer by layer growth,where initial roughness generated by the Pd NPs has vanished so that the surface is smooth.

annealing to decompose the alucone,the resulting Al 2O 3?lms were found to be highly porous.34A direct comparison between the porosity of the ALD Al 2O 3overlayers produced in this study with those of Weimer et al.would require further investigation.

It is remarkable in our studies that the Pd NPs overcoated with up to 7?8cycles remain accessible to small reagent gases like water,CO,and methanol,even though TEM measure-ments show these particles to be embedded in a 1nm coating.Furthermore,porosity can be induced into much thicker ALD Al 2O 3overcoats on Pd NPs surface through temperature treatment or long-term reaction at elevated temperatures,which was indicated by the gain in CO chemisorption capacity on the used catalysts.13,17For example,we recently showed that when a Pd/Al 2O 3catalyst with 45cycles of ALD Al 2O 3overcoat was treated at 973K in an oxygen environment,the Pd surfaces became accessible to reagent gases through the development of microporosity (～2nm)inside the Al 2O 3layer.17Through overcoating and annealing,it might be possible to tune the activity and selectivity of Pd catalysts by controlling the porosity of the Al 2O 3(Figure 8).13,17

■

CONCLUSIONS

We have investigated the growth mechanism of Al 2O 3ALD using TMA and water on Pd NPs surfaces by combining in situ QCM,in situ QMS,and TEM experimental studies with DFT calculations.Both QCM and QMS presented a consistent picture that Al 2O 3only grows on certain portions of the Pd NPs rather than forming a continuous coating in the initial cycles.The high CH 4_total to CH 4_TMA ratio and the production of ethane suggest that the coverage of CH 3*should be high after the TMA pulse.The hydrogen product,observed during the water exposures in the ?rst ～7cycles,suggests that the ALD Al 2O 3layer is porous,consistent with our previous CO chemisorption studies.A mechanism is proposed for TMA dissociative adsorption and hydration of TMA intermediates based on DFT calculations.Thermodynamic analysis shows that the TMA dissociation is more favored on the stepped sites than the terrace sites,suggesting that low-coordination Pd sites may facilitate TMA nucleation in the ?rst ALD cycle.Moreover,DFT calculations also show that the hydration product Al(OH)3*is more thermodynamically favorable than Al(OH)*during the water pulses,explaining the anomalously

high Δm 0/Δm 1mass ratio observed during the ?rst cycle by in situ QCM.

■ASSOCIATED CONTENT

*

Supporting Information Additional ?gures (PDF).This material is available free of charge via the Internet at https://www.sodocs.net/doc/a210855329.html,.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail:jelam@https://www.sodocs.net/doc/a210855329.html,.

Notes

The authors declare no competing ?nancial interest.

■

ACKNOWLEDGMENTS

J.W.E.,J.P.G.,B.L.,and J.L.were supported as part of the Institute for Atom-e ?cient Chemical Transformations (IACT),an Energy Frontier Research Center funded by the U.S.Department of Energy (DOE),O ?ce of Science,O ?ce of Basic Energy Science.Z.F.and M.J.B.thank the support of the Institute for Catalysis and Energy Processes (U.S.Department of Energy Grant,DE-FG02-03ER15457).P.C.S.and Y.L.acknowledge support from the U.S.Department of Energy,BES-HFI,Chemical Sciences,under Contract https://www.sodocs.net/doc/a210855329.html,e of the Center for Nanoscale Materials was supported by the U.S.Department of Energy,O ?ce of Science,O ?ce of Basic Energy Science,under Contract DE-AC02-06CH11357.We also acknowledge grants of computer time from EMSL,a national scienti ?c user facility located at Paci ?c Northwest National Laboratory,and the Argonne Laboratory Computing Resource Center (LCRC),and resources of the National Energy Research Scienti ?c Computing Center (NERSC).

■

REFERENCES

(1)Ritala,M.Appl.Surf.Sci.1997,112,223.

(2)Suntola,T.;Hyvarinen,J.Annu.Rev.Mater.Sci.1985,15,177.(3)George,S.M.Chem.Rev.2010,110,111.

(4)Puurunen,R.L.J.Appl.Phys.2005,97,121301.

(5)Ferguson,J.D.;Weimer,A.W.;George,S.M.Thin Solid Films 2000,371,95.

(6)Libera,J.A.;Elam,J.W.;Pellin,M.J.Thin Solid Films 2008,516,6158.

(7)Lim,B.S.;Rahtu,A.;Gordon,R.G.Nat.Mater.2003,2,749.(8)Elam,J.W.;Xiong,G.;Han,C.Y.;Wang,H.H.;Birrell,J.P.;Welp,U.;Hryn,J.N.;Pellin,M.J.;Baumann,T.F.;Poco,J.F.;Satcher,J.H.J.Nanomater.2006,2006,1.

(9)Elam,J.In Atomic Layer Deposition of Nanostructured Materials ;Pinna,N.,Knez,M.,Eds.;Wiley-VCH:Weinheim,Germany,2012;p 227.

(10)Lei,Y.;Mehmood,F.;Lee,S.;Greeley,J.;Lee,B.;Seifert,S.;Winans,R.E.;Elam,J.W.;Meyer,R.J.;Redfern,P.C.;Teschner,D.;Schlogl,R.;Pellin,M.J.;Curtiss,L.A.;Vajda,S.Science 2010,328,224.

(11)Enterkin,J.A.;Setthapun,W.;Elam,J.W.;Christensen,S.T.;Rabu ?etti,F.A.;Marks,L.D.;Stair,P.C.;Poeppelmeier,K.R.;Marshall,C.L.ACS Catal.2011,1,629.

(12)Feng,H.;Elam,J.W.;Libera,J.A.;Setthapun,W.;Stair,P.C.Chem.Mater.2010,22,3133.

(13)Feng,H.;Lu,J.L.;Stair,P.C.;Elam,J.W.Catal.Lett.2011,141,512.

(14)Christensen,S.T.;Feng,H.;Libera,J.L.;Guo,N.;Miller,J.T.;Stair,P.C.;Elam,J.W.Nano Lett.2010,10,3047.

(15)Feng,H.;Elam,J.W.;Libera,J.A.;Pellin,M.J.;Stair,P.C.J.Catal.2010,269,

421.

Figure 8.Schematic illustration of porous ALD Al 2O 3overcoat on Pd NP.(a)Oxide-supported Pd catalyst;(b)dense Al 2O 3?lm on oxide support and porous Al 2O 3overcoat on Pd NP formed by ALD.

(16)King,J.S.;Wittstock,A.;Biener,J.;Kucheyev,S.O.;Wang,Y. M.;Baumann,T.F.;Giri,S.K.;Hamza,A.V.;Baeumer,M.;Bent,S.F. Nano Lett.2008,8,2405.

(17)Lu,J.L.;Fu,B.S.;Kung,M.C.;Xiao,G.M.;Elam,J.W.;Kung,

H.H.;Stair,P.C.Science2012,335,1205.

(18)Martinson,A.B.F.;Elam,J.W.;Pellin,M.J.Appl.Phys.Lett. 2009,94,123107.

(19)Libera,J.A.;Hryn,J.N.;Elam,J.W.Chem.Mater.2011,23, 2150.

(20)Nanu,M.;Schoonman,J.;Goossens,A.Adv.Mater.2004,16, 453.

(21)Chen,Y.W.;Prange,J.D.;Duhnen,S.;Park,Y.;Gunji,M.; Chidsey,C.E.D.;McIntyre,P.C.Nat.Mater.2011,10,539. (22)Scott,I.D.;Jung,Y.S.;Cavanagh,A.S.;An,Y.F.;Dillon,A.C.; George,S.M.;Lee,S.H.Nano Lett.2011,11,414.

(23)Chao,C.C.;Hsu,C.M.;Cui,Y.;Prinz,F.B.ACS Nano2011,5, 5692.

(24)Adarnczyk,N.M.;Dameron,A.A.;George,https://www.sodocs.net/doc/a210855329.html,ngmuir 2008,24,2081.

(25)Du,Y.;George,S.M.J.Phys.Chem.C2007,111,8509.

(26)Mayer,T.M.;Elam,J.W.;George,S.M.;Kotula,P.G.;Goeke, R.S.Appl.Phys.Lett.2003,82,2883.

(27)Davidson,B.D.;Seghete,D.;George,S.M.;Bright,V.M.Sens. Actuators,A2011,166,269.

(28)Lu,J.L.;Stair,https://www.sodocs.net/doc/a210855329.html,ngmuir2010,26,16486.

(29)Feng,H.;Elam,J.W.;Libera,J.A.;Stair,P.C.;Miller,J.T.ACS Catal.2011,1,665.

(30)Lu,J.L.;Stair,P.C.Angew.Chem.,Int.Ed.2010,49,2547.

(31)Christensen,S.T.;Elam,J.W.;Rabu?etti,F.A.;Ma,Q.; Weigand,S.J.;Lee,B.;Seifert,S.;Stair,P.C.;Poeppelmeier,K.R.; Hersam,M.C.;Bedzyk,M.J.Small2009,5,750.

(32)Ma,Z.;Brown,S.;Howe,J.Y.;Overbury,S.H.;Dai,S.J.Phys. Chem.C2008,112,9448.

(33)Zhang,X.Y.;Zhao,J.;Whitney,A.V.;Elam,J.W.;Van Duyne, R.P.J.Am.Chem.Soc.2006,128,10304.

(34)Liang,X.H.;Li,J.H.;Yu,M.;McMurray,C.N.;Falconer,J.L.; Weimer,A.W.ACS Catal.2011,1,1162.

(35)Joo,S.H.;Park,J.Y.;Tsung,C.K.;Yamada,Y.;Yang,P.D.; Somorjai,G.A.Nat.Mater.2009,8,126.

(36)Ott,L.S.;Finke,R.G.Coord.Chem.Rev.2007,251,1075.

(37)Seipenbusch,M.;Binder,A.J.Phys.Chem.C2009,113,20606.

(38)Park,J.N.;Forman,A.J.;Tang,W.;Cheng,J.H.;Hu,Y.S.;Lin,

H.F.;McFarland,E.W.Small2008,4,1694.

(39)Takenaka,S.;Matsumori,H.;Nakagawa,K.;Matsune,H.; Tanabe,E.;Kishida,M.J.Phys.Chem.C2007,111,15133. (40)Scott,R.W.J.;Sivadinarayana,C.;Wilson,O.M.;Yan,Z.; Goodman,D.W.;Crooks,R.M.J.Am.Chem.Soc.2005,127,1380.

(41)Arnal,P.M.;Comotti,M.;Schuth,F.Angew.Chem.,Int.Ed. 2006,45,8224.

(42)Ott,A.W.;Klaus,J.W.;Johnson,J.M.;George,S.M.Thin Solid Films1997,292,135.

(43)Dillon,A.C.;Ott,A.W.;Way,J.D.;George,S.M.Surf.Sci. 1995,322,230.

(44)Rahtu,A.;Alaranta,T.;Ritala,https://www.sodocs.net/doc/a210855329.html,ngmuir2001,17,6506.

(45)Juppo,M.;Rahtu,A.;Ritala,M.;Leskela,https://www.sodocs.net/doc/a210855329.html,ngmuir2000,16, 4034.

(46)Elam,J.W.;Groner,M.D.;George,S.M.Rev.Sci.Instrum. 2002,73,2981.

(47)Groner,M.D.;Fabreguette,F.H.;Elam,J.W.;George,S.M. Chem.Mater.2004,16,639.

(48)Whitney,A.V.;Elam,J.W.;Stair,P.C.;Van Duyne,R.P.J. Phys.Chem.C2007,111,16827.

(49)Whitney,A.V.;Elam,J.W.;Zou,S.L.;Zinovev,A.V.;Stair,P.

C.;Schatz,G.C.;Van Duyne,R.P.J.Phys.Chem.B2005,109,20522.

(50)Elam,J.W.;Zinovev,A.;Han,C.Y.;Wang,H.H.;Welp,U.; Hryn,J.N.;Pellin,M.J.Thin Solid Films2006,515,1664.

(51)Kresse,G.;Hafner,J.Phys.Rev.B1993,47,558.

(52)Kresse,G.;Hafner,J.Phys.Rev.B1994,49,14251.

(53)Kresse,G.;Furthmuller,J.Phys.Rev.B1996,54,11169.

(54)Kresse,G.;Furthmuller,https://www.sodocs.net/doc/a210855329.html,put.Mater.Sci.1996,6,15.

(55)Blochl,P.E.Phys.Rev.B1994,50,17953.

(56)Kresse,G.;Joubert,D.Phys.Rev.B1999,59,1758.

(57)Perdew,J.P.;Chevary,J.A.;Vosko,S.H.;Jackson,K.A.; Pederson,M.R.;Singh,D.J.;Fiolhais,C.Phys.Rev.B1992,46,6671.

(58)Perdew,J.P.;Chevary,J.A.;Vosko,S.H.;Jackson,K.A.; Pederson,M.R.;Singh,D.J.;Fiolhais,C.Phys.Rev.B1993,48,4978.

(59)Kittel,C.Introduction to Solid State Physics;7th ed.;John Wiley &Sons:New York,1996;p673.

(60)Lo?reda,D.;Simon,D.;Sautet,P.Chem.Phys.Lett.1998,291, 15.

(61)Monkhorst,H.J.;Pack,J.D.Phys.Rev.B1976,13,5188.

(62)Methfessel,M.;Paxton,A.T.Phys.Rev.B1989,40,3616.

(63)Henkelman,G.;Jonsson,H.J.Chem.Phys.2000,113,9978.

(64)Henkelman,G.;Johannesson,G.;Jonsson,H.;Kluwer Academic:New York,2000,p269.

(65)Henkelman,G.;Jonsson,H.J.Chem.Phys.1999,111,7010.

(66)Puurunen,R.L.J.Appl.Phys.2005,97,121301.

(67)McCormick,J.A.;Rice,K.P.;Paul,D.F.;Weimer,A.W.; George,S.M.Chem.Vap.Dep.2007,13,491.

(68)Kruse,N.;Rebholz,M.;Matolin,V.;Chuah,G.K.;Block,J.H. Surf.Sci.1990,238,L457.

(69)Levis,R.J.;Jiang,Z.C.;Winograd,N.J.Am.Chem.Soc.1989, 111,4605.

(70)Paul,J.F.;Sautet,P.J.Phys.Chem.B1998,102,1578.

(71)Liu,Z.P.;Hu,P.J.Am.Chem.Soc.2003,125,1958.

(72)Marsh,A.L.;Becraft,K.A.;Somorjai,G.A.J.Phys.Chem.B 2005,109,13619.

(73)Fairbrother,D.H.;Peng,X.D.;Viswanathan,R.;Stair,P.C.; Trenary,M.;Fan,J.Surf.Sci.1993,285,L455.

(74)Fuhrmann,T.;Kinne,M.;Trankenschuh,B.;Papp,C.;Zhu,J.

F.;Denecke,R.;Steinruck,H.P.New J.Phys.2005,7.

(75)Puurunen,R.L.;Vandervorst,W.J.Appl.Phys.2004,96,7686.