Composition and structure of Pd nanoclusters in SiO x thin film
Annett Thøgersen, Jeyanthinath Mayandi, Lasse Vines, Martin F. Sunding, Arne Olsen, Spyros Diplas, Masanori Mitome, Yoshio Bando
CComposition and structure of Pd nanoclusters in SiO x thin film Annett Thøgersen, Jeyanthinath Mayandi, and Lasse Vines
Centre for Materials Science and Nanotechnology,University of Oslo, P.O.Box 1126 Blindern, N-0318 Oslo, Norway
Martin F. Sunding and Arne Olsen
Department of Physics, University of Oslo, P.O.Box 1048 Blindern, N-0316 Oslo, Norway
Spyros Diplas
SINTEF Materials and Chemistry, P.B 124 Blindern, N-0314 Oslo,Norway and Centre for Material Science and Nanotechnology, University of Oslo.
Masanori Mitome and Yoshio Bando
National Institute of Material Science, Namiki 1-1, Tsukuba, Ibaraki, 305-0044 Japan (Dated: October 2, 2012)The nucleation, distribution, composition and structure of Pd nanocrystals in SiO multilayerscontaining Ge, Si, and Pd are studied using High Resolution Transmission Electron Microscopy(HRTEM) and X-ray Photoelectron Spectroscopy (XPS), before and after heat treatment. The Pdnanocrystals in the as deposited sample seem to be capped by a layer of PdO x . A 1-2 eV shift inbinding energy was found for the Pd-3d XPS peak, due to initial state Pd to O charge transfer inthis layer. The heat treatment results in a decomposition of PdO and Pd into pure Pd nanocrystalsand SiO . PACS numbers:
I. INTRODUCTION
Material systems containing silicon and germaniumnanocrystals have attracted much attention due to theiroptical and electronic properties , as well as their po-tential applications in photo detectors , light emitters ,single electron transistors and non-volatile memories .Nanoclusters embedded in a SiO matrix is an attractiveoption towards nanocluster based device development .The most important factors influencing the optical prop-erties of the SiO -nanocluster devices are size, spatialdistribution, atomic and electronic structure as well asthe surface properties of the nanoclusters.Pd is an especially interesting material when used forcatalytic converters in auto-mobile technology for theelimination of NO X (Nitrogen Oxides) in the exhaustgases of gasoline engines . Considerable research has alsobeen conducted in the use of Pd catalysts for the combus-tion of methane. Particle morphology and oxidation statecan play an important role in defining the active sites onPd catalysts. Pd is also found in other applications suchas Granular Metal (GM) films, cermets or nano-cermets,where metal particles on MgO cubes . Transition metalparticles have interesting properties due to quantum sizeeffect, owing to the dramatic reduction of the numberof free electrons . The nanoparticle matrix may formedadvanced material system with new electronic, magnetic,optic, and thermal properties .In this work, samples containing both Pd nanocrys-tals and Ge clusters embedded in SiO layers supersat-urated with Si, were studied in detail, before and afterheat treatment. The formation, composition, distribu- tion, and the atomic and electronic structure of Ge andPd nanoclusters were studied by High Resolution Trans-mission Electron Microscopy (HRTEM) and X-ray Pho-toelectron Spectroscopy (XPS). II. EXPERIMENTAL
The samples were produced by growing a ∼ on a p-type Si substrate by Rapid Thermal Ox-idation (RTO) at 1000 ◦ C for 6 sec. Prior to growingthe RTO layer the wafers were cleaned using a standardRCA procedure (Radio Corporation of America, indus-try standard for removing contaminants from wafers) fol-lowed by immersion in a 10 % HF solution to remove thenative oxide. Then a ∼
10 nm layer of silicon rich ox-ide (46 at. %) was sputtered from a SiO :Si compositetarget onto the RTO-SiO and heat treated in a N atmo-sphere at 1000 ◦ C for 30 min, as described in our previousarticle . A ∼ layer containing 0.5 at. %Ge and 0.5 at. % Pd, was then sputtered. This sampleis reffered to as the as deposited sample (sample ASD )The heat treated sample (sample HT ) was then annealedagain at 900 ◦ C in a N atmosphere for 30 minutes tonucleate Ge and Pd nanocrystals. Cross-sectional TEMsamples were prepared by ion-milling using a Gatan pre-cision ion polishing system with a 5 kV gun voltage.The nanocrystals were studied with HRTEM using a300 keV JEOL 3100FEF TEM with an Omega imagingfilter. Additional HRTEM images were acquired usinga 200 keV JEOL 2010F TEM. XPS was performed ina KRATOS AXIS ULTRA DLD using monochromatic Al a r X i v : . [ c ond - m a t . m t r l - s c i ] S e p FIG. 1: HRTEM images of A) sample
ASD and B) HT . Thearrows show both crystalline (in the SiO -Ge layer) and amor-phous (in the SiO -Si-layer) nanoclusters of Pd. Fast FourierTransform patterns of the nanocrystals are presented as insetsand shows that the clusters are crystalline. K α radiation ( hν = 1486 . ◦ angle of emission (vertical) and charge compen-sation with ion energy electrons from a flood gun. TheX-ray source was operated at 10 mA and 15 kV. Theinelastic mean free path of the Ge d electrons in SiO is ∼ . Emission normal to the surface results in aphotoelectron escape depth of 11.8 nm , which allowedus to study the silicon nanoclusters located 11.8 nm be-low the surface of the oxide. Accordingly the inelasticmean free path of the Pd-3d electrons in SiO is 3.38nm, which allows us to study Pd nanocrystals within ananalysis depth of 10.1 nm. The spectra were peak fit-ted using the CasaXPS program after subtraction ofa Shirley type background. The spectra were calibratedby adjusting peak positions of the O-1s and Si-2p signalsfrom SiO at 533 eV and 103.6 eV respectively, and theSi-2p from the Si substrate at 99.5 eV . Compositiondepth profile was performed with Ar ion etching, on a3x3 mm spot area, at 2 kV, 100 µ A current and a cycletime of 20 s.
III. RESULTS AND DISCUSSION
Nanoclusters of different sizes in various layers were ob-served in both the as deposited (
ASD ) and heat treatedsample ( HT ), see Figure 1. Dark Ge and Pd nanoclustersare visible in the SiO -Ge-Pd layer, 10 nm from the outer-most surface (SiO /glue interface) and 13 nm from the Sisubstrate. Lattice fringes can be seen in the largest nan-oclusters, indicating that many of them are crystalline.The average nanocluster size for sample HT and ASD is 4nm and 2.5 nm, respectively . Smaller (1-2 nm) clustersare visible in the SiO -Si layer 6 nm from the Si substratein sample HT . No lattice fringes were observed in thesenanoclusters, indicating an amorphous structure. SinceSi has almost no contrast when embedded in SiO , thesenanoclusters contain either Pd or Ge.Figure 2 shows the high resolution Si p , Ge d , and Pd d XPS spectra acquired during depth profiling with Ar + sputtering of samples HT and ASD . The XPS spectra of sample HT show that Ge is distributed wider in theSiO -Ge-Pd layer, compared to sample ASD , where Geis located in a narrow band. In addition, Ge surface seg-regation seems to occur upon annealing in accordancewith previous work by Agan et al. and Marstein et al./citemarstein:1. Sample ASD looks very similar to sam-ple HT , apart from the location of the Si and Ge (seeFigure 2). The Pd distribution is almost the same inboth samples. As seen in Figure 1, small nanoclusterswere observed in the SiO -Si layer of samples HT . TheGe-3d and Pd-3d spectra of sample HT show very little orno increase in the Ge and Pd concentration in the SiO -Si layer. This indicates that the nanoclusters shown inFigure 1 that are present in the SiO -Si layer, have dif-fused to this area during electron beam exposure. Thereis a visible shift in the position of the Pd peaks withthe depth in sample ASD . This shift, not seen in sample HT , will be discussed in Section V B. Figure 4 shows theelemental distribution with depth for sample ASD and HT . The quantification is based upon peak fitting, beingshown in more in the following sections.The results of the as deposited samples are more com-plicated to interpret. We therefore present first the lesscomplicated results of the heat treated samples, followedby the as deposited ones. IV. RESULTS AND DISCUSSION: HEATTREATED SAMPLES
The composition of the heat treated samples measuredwith XPS, EELS, EFTEM and EDS is presented in Sec-tion IV A, while the atomic structure found by TEM ispresented in Section IV B.
A. Nanocluster composition
XPS was used to identify the chemical state of differentelements present in sample HT . Figure 3B and D showsthe fitted XPS spectrum of the Pd-3d and Ge-3d peak.The measured binding energies of the two Ge peaks were29.8 eV and 33 . ± . E = 3 . and 32.5 eV for GeO . The Ge-3denergy separation/chemical shift between Ge and GeO is reported to be 3.3 eV . The chemical shift is lesssusceptible to energy referencing when the same referenceis used and both peaks defining the chemical shift existin the same spectrum. Therefore the above data suggeststhat the peak at 29.8 corresponds to pure Ge, while thepeak at 33.1 ± . . ± . .The XPS spectrum in Figure 3B shows the Pd-3d peak.The spectrum is fitted with six components. The twopeaks located at binding energies of 348.9 eV and 345.9eV belong to Ge-LMM. The remaining four peaks maybe contributions from Pd d and Pd d . The measuredbinding energies are presented in Table II, and the chem- FIG. 2: XPS depth profiles of the Si p -, Ge d -, and Pd d -peak of sample HT and ASD . ical shifts in Table III. Pure Pd has a binding energy of340.4 eV (3d / ) and 335.1 eV (3d / ) , PdO of 341.6eV (3d / ) and 336.3 eV (3d / ), and PdO of 343 eV(3d / ) and 337.9 eV (3d / ) . The reference valuesshow chemical shifts for the Pd -Pd =1.2 eV, and forPd -Pd =2.6 eV.The two smallest peaks in Figure 3B at 341.6 eV and336.3 eV fit well with PdO, while the larger peaks at340.1 eV and 334.8 eV correspond to pure Pd. This isin agreement with the chemical shift of 1.5 between thePd -Pd of both the 3d / and 3d / peaks. The Pd peaks were fitted using slightly asymmetric components,with a FWHM of about 1 eV. These XPS results areshown and discussed in Section V B and V C.Figure 2 shows significant differences in the Si /Si intensity ratio in the SiO -Si layer between the two sam-ples. The increased Si /Si intensity ratio in the SiO -Sizone as well as in the Si substrate of the annealed sampleis attributed to oxidation upon annealing. This oxidationmay also be responsible for the small decrease in pure Geconcentration, as seen in Figure 4. B. Atomic structure of the nanocrystals
TABLE I: The observed d-values found from the FFT pat-terns compared to reference d values.Sample Observed Pd Pd O PdO Pd Si d-value (nm) (nm) (nm) (nm)(nm) HT ASD
ASD
The atomic structure of the nanocrystals in sample HT FIG. 3: XPS spectra of A) the Pd-3d peak of sample
ASD , B)the Pd-3d peak of sample HT , C) the Ge-3d peak of sample ASD , and D) the Ge-3d peak of sample HT . The detectiondepth is with respect to the sample surface. were studied by Fast Fourier Transform (FFT) patterns.FFT patterns of the Si substrate and two nanocrystalsare presented as insets in Figure 1. The FFT of the Sisubstrate is used as a reference, since the Si unit cell di-mensions are known ( d = 0 .
543 nm). The FFT of the twonanocrystals yielded a d-value of 0 . ± .
001 nm. TheHRTEM image shown in Figure 1 shows a squared pat-tern with the same lattice plane spacing of d = 0 .
198 nm.The measured d-values as well as references are presentedin Table I. Pure Pd has a Fm-3m space group, with lat-tice parameter 0.389 nm. The largest d-value is (200):0.195nm . This fits well with the measured d-values. FIG. 4: Elemental profile obtained from the XPS depth profilespectra of A) sample
ASD , and B) sample HT .FIG. 5: A) HRTEM image of a Pd nanocrystal in sample ASD , B) a sketch of the (100) zone axis of a Pd nanocrystal,and C) a higher magnified image of the atomic structure ofthe nanocrystal. FIG. 6: FFT patterns from A) the Si substrate, and B) andC) are from two Pd nanocrystals in sample
ASD . V. RESULTS AND DISCUSSION: ASDEPOSITED SAMPLES
The composition and atomic structure of sample
ASD is presented in the following sections. XPS measurementsof sample
ASD is presented in Section V B, and the dis-cussion of the results in in Section V C.
A. Nanocluster composition and atomic structure
A HRTEM image of a Pd nanocrystal in sample
ASD is presented in Figure 5. FFT patterns from the Sisubstrate and from two nanocrystals are presented inFigure 6. FFT patterns of three different nanocrystalareas in sample
ASD yielded the measured d-values of0 . ± .
002 nm, 0 . ± .
002 nm, and 0 . ± .
002 nm.The experimentally observed d values are then comparedto reference values (presented in Table I). The d-valuesmeasured on the three nanocrystal areas in sample
ASD fit with Pd O , PdSi, Pd Si, PdO, and pure Pd.Voogt et al. studied PdO particles with a metallicPd core. The surface tension of PdO is lower than thesurface tension of pure metallic Pd, and upon annealingPd will not dissolve in PdO . It is therefore reasonableto find particles containing, a metallic core and an oxideskin when heat treated at higher temperatures, becausethe migration of atoms is relatively easy. During heattreatment this core will grow linearly proportional to thesurface area, transforming to Pd and SiO x towards reach-ing thermodynamic equilibrium. This may explain whywe see more Pd oxide in the as deposited samples.It is therefore reasonable for the Pd nanocrystals inthe as deposited samples to have a few atomic layers ofPdO and/or Pd Si around them. During heat treatment,these compounds will react to form pure Pd and SiO .XPS results of these samples is shown and discussed inthe next section.The HRTEM and XPS of the as deposited and heattreated sample shows that Ge and Pd behave in an oppo-site manner. After annealing, Pd decomposes into purePd and SiO , whereas Ge oxidises. This may be influ-enced by residual oxygen in the annealing ambient. Thedifferences in oxidation may also be due to oxygen trans-fer from Pd to Ge (and Si), as a result of differences inthe enthalpy. B. Pd-3d binding energy shifts
TABLE II: The binding energy peak positions in sample
ASD and HT Peak Sample
ASD
Sample HT References ab E B (eV) E B (eV) E B (eV)Pd -3d / -3d / -3d / -3d / -3d / -3d / a Reference b Reference The Pd-3d and Ge-3d spectra from sample
ASD areshown in Figure 3A and C, while the Pd-3d spectra fromsample HT are shown in Figure 3B and D. The exper-imental binding energies, chemical shifts and referencevalues are shown in Table II and Table III. We observea shift in the Pd-3d binding energy with depth in theas deposited sample compared to the heat treated one(Figure 3). The chemical shift is defined as the bind-ing energy difference 3d / / - 3d / . HRTEM image ofsamples ASD shown in the previous section gave d-valueswhich match well with Pd O (Pd +4 and Pd +2 ), PdSi(Pd +4 ), Pd Si (Pd +2 ), PdO (Pd +2 ), and pure Pd (Pd ).The Pd-3d peak for sample ASD is visible at sampledepth range of 7-17 nm and 10-20 nm, due to a 10 nmphotoelectron escape depth for Pd in SiO . The depthwas determined using the Si-2p peak from the Si sub-strate and the photoelectron escape depth. The spectrumat 7-17 nm is from the top of the band of Pd nanocrystals,and the spectrum at 10-20 nm is from the bottom.The Pd-3d spectra of both samples at a depth of 7-17nm show four peaks in addition to the peaks belongingto Ge-LMM. Figure 2 shows the Si-2p peaks of sample ASD . Only Si from SiO (Si ) was detected at eitherdepths. This in combination with the Pd peaks meansthat Pd Si and PdSi are not present in the samples. Thesmaller peaks in Figure 3 A must therefore result fromPdO and/or Pd O .The chemical shift between the two 3d / and 3d / peak maxima in Figure 3 A at a depth of 7-17 nm are 2eV and 2.2 eV respectively for sample ASD . The chemicalshift values are higher than the expected one for Pd (1.2 eV) and lower than the one expected for Pd (2.6eV ). A reduced chemical shift could be attributedto the presence of Pd /Pd mixed valency. Pd O contains a mixed valency of two Pd and one Pd ions.The peaks at 343 eV and 337.9 eV most probably are dueto the presence of Pd O .The Pd-3d spectra from a depth of 10-20 nm in sample ASD contain six compounds in addition to the Ge-LMMpeaks. The two largest compounds were fitted with apure Gaussian peak. The chemical shift between the
TABLE III: The chemical shift of sample HT and ASD , andthe reference values.Sample Chem. shift: 3d / / ( P d / − P d ) (eV) (eV)( ± ± HT Pd +2 - Pd ASD Pd +2 - Pd ASD Pd +4 - Pd Pd -Pd +2 Pd -Pd +4 largest compounds and the Pd peaks is 0.9 eV for boththe 3d / - and 3d / - peak in sample ASD . These peakscan probably be assigned to a Pd peak (PdO or PdO x ).Figure 3 shows that pure Pd is mostly found in theupper part of the nanocluster band, while the (sub)oxideswere found in the lowest part of the nanocluster band,either as an oxide skin around the Pd nanocrystals and/ or as pure oxide nanocrystals. This inhomogenity hasmost probably occurred during sputtering deposition. Asmall shift to a higher binding energy is observed for bothPd (1 eV) and Pd peaks (0.5-0.7 eV) as comparedto the reference values . The chemical shift betweenPd and Pd is lower than the reference values and lowerthan what was found in sample HT . This will be discussedin the next section. C. The nature of the Pd-3d binding energy shift
In a previous paper, we studied the binding energyshifts of Er O nanoclusters in SiO , and the variousfactors influencing the binding energy were discussed indetail . In this work we performed a similar study onPd nanoclusters in SiO in order to evaluate the decreasein chemical shifts found in the as deposited sample ascompared to the heat treated sample. Shifts in bindingenergy can be expressed as∆ E B = K ∆ q + ∆ V + ∆ ϕ − ∆ R (1)In the above equation, K is a measure of the Coulombinteraction between the valence and core electrons, and∆ q expresses changes in the valence charge. K ∆ q reflectstherefore charge transfer effects. ∆ V is the contribu-tion of the changes in Madelung potential. ∆ ϕ containschanges in energy referencing, including variations of thesample work function and of the energy of charge com-pensating electrons, which may be important in the caseof insulators. The first two terms in Equation 1 refer toinitial state effects, while the third term expresses thedependence of ∆ E on energy referencing in the case ofinsulators. The fourth term is the contribution of therelaxation energy R , which is the kinetic energy gained(negative sign) when the electrons in the solid screen thephotohole produced by the photoemission process; thisis a final state effect.A 1-2 eV core level shift to higher E B in the Pd-3dpeak was observed in the work by Ichinohe et al. , whostudied Pd clusters in SiO . The shift was attributedto final state effects as a consequence of the decrease inparticle size, which is an initial state effect, and a subse-quent decrease in screening. This demonstrates how theinitial state influences the final state effects. In the caseof nanoclusters in an insulating matrix, the core hole re-laxation, screening could contain matrix contributions tosome extent. Since Pd is a metal, it is characterized by alarge screening efficiency. SiO , on the other hand, is aninsulator and has a low screening efficiency. Therefore,assuming an external screening contribution by SiO , thescreening in bulk Pd is higher than the screening in Pdnanocrystals embedded in SiO , due to the low screen-ing contribution from the oxide. Quantum confinementeffects and an increased band gap may also reduce screen-ing since the core hole screening by the conduction banddepend on the band gap. The larger the band gap isthe lower the screening efficiency becomes. A reductionin core hole screening appears as an increase in bindingenergy.The (sub)oxide in the as deposited samples has adecreased chemical shift compared to the heat treatedsamples. PdO has a higher dielectric constant thanits surrounding SiO matrix similarly to Er O inthe same matrix . In accordance with the previousargumentation , the screening contribution of the sur-rounding SiO on the Pd and PdO clusters is expected tobe small. Considering absence of energy referencing is-sues, initial state effects seem to have a dominant role in the increase of the binding energy of Pd nanocrystals andPdO x clusters. As for the Er O nanoclusters , chargetransfer from Pd to O can lead to the creation of positivesurface charges. The increased binding energy may there-fore be due to initial state effects, such as charge transferfrom Pd towards the interface. The smaller the nan-ocluster size, the higher the surface/volume ratio. There-fore interfacial transport phenomena are more enhanced.Variations in the effectiveness of charge neutralization onPd nanocrystals with and without a PdO x skin may alsoaccount for differences in peak shifts. VI. CONCLUSION
Multilayer samples containing Pd, Ge and Si weremade in order to study the nucleation, distribution, com-position as well as atomic and electronic structure of Geand Pd nanoclusters. The nanocrystals were observed byHRTEM, EDS and EFTEM imaging. Ge was observedin the form of small amorphous nanoclusters. The asdeposited samples contained not only pure Pd nanocrys-tals, but also Pd-oxides. A 1-2 eV shift in binding energyfound for the XPS Pd-3d peak of pure Pd and Pd wasattributed to initial state effects arising from an increasedcharge transfer from Pd to O in the nanocrystals and /or to electrostatic charging. According to the combinedTEM and XPS data, the Pd nanoclusters in the as de-posited samples consist of Pd and PdO x . S. Agan, A. Dana, and A. Aydinli, J. Phys.: Condens.Matter , 5037 (2006). R. Salh, L. Fitting, E. V. Kolesnikova, A. A. Sitnikova,M. V. Zamoryanskaya, B. Schmidt, and H.-J. Fitting,Semiconductors , 387 (2007). K. L. Wang, J. L. Liu, and G. Jin, J. Cryst. Growth , 1892 (2002). Y. Q. Wang, G. L. Kong, W. D. Chen, H. W. Diao, C. Y.Chen, S. B. Zhang, and X. B. Liao, Appl. Phys. Lett. ,4174 (2002). D. V. Averin and K. K. Likharev, J. Low-Temp. Phys. ,2394 (1986). M. Kanoun, , A. S. A. Baron, and F. Mazen, Appl. Phys.Lett. , 5079 (2004). Y. M. Yang, X. L. Wu, L. W. Yang, G. S. Huang, G. G.Siu, and P. K. Chu, J. Appl. Phys. , 064303 (2005). J. Kielhorn, C. Melber, D. Keller, and I. Mangelsdorf, In-ternational Journal of Hygiene and Environmental Health , 417 (2002). T. Ichinohe, S. Masaki, K. Uchida, S. Nozaki, andH. Morisaki, Thin solid films , 27 (2004). T. Teranishi and M. Miyake, Chem. Mater. 1998 , 594(1998). A. Thøgersen, J. Mayandi, J. S. Christensen, T. Finstad,M. Mitome, Y. Bando, and A. Olsen, J. Appl. Phys. ,094315 (2008). A. Thøgersen, S. Diplas, J. Mayandi, T. Finstad, A. Olsen,J. F. Watts, M. Mitome, and Y. Bando, J. Appl. Phys. , 024308 (2008). J. F. Moulde, W. F. Stickle, P. E. Sobol, and K. D.Bomben,
Handbook of X-Ray Photoelectron Spectroscopy (Perkin-Elmer Corporation, 1992), p. 201. NIST, http://srdata.nist.gov/xps/ (2009). X. Wu, M. Lu, and W. Yao, Surf. and Coat. Tech. , 92(2002). K. Kim, A. Gossmann, and N. Winograd, An. Chem. ,197 (1974). E. Owen and E. Yates, Phil. Mag. , 472 (1933). H. J. Meyer and H. Mueller-Buschbaum, Zeitschrift fuerNaturforschung, Teil B. Anorganische Chemie, OrganischeChemie (33,1978-41,1986) , 1661 (1979). A. G. Christy and S. M. Clark, Physical Review, Serie 3.B - Condensed Matter (18,1978-) , 9259 (1995). A. Nylund, Acta Chemica Scandinavica , 2381 (1966). E. Voogt, A. Mens, O. Gijzeman, and J. Geus, Surf. Sci. , 21 (1996). A. Thogersen, J. M. A. Finstad, A. Olsen, S. Diplas,M. Mitome, and Y. Bando, J. of Appl. Phys.106