NANO LETTERS
Direct Quantification of Gold along a Single Si Nanowire
2008 Vol. 8, No. 11 3709-3714
A. Bailly,† O. Renault,*,† N. Barrett,‡ L. F. Zagonel,‡ P. Gentile,§ N. Pauc,§ F. Dhalluin,| T. Baron,| A. Chabli,† J. C. Cezar,⊥ and N. B. Brookes⊥ CEA-LETI, MINATEC, 17 rue des Martyrs, 38054 Grenoble Cedex 9, France, CEA-DSM/IRAMIS/SPCSI, CEA Saclay, 91191 Gif-sur-YVette, France, CEA-DSM/INAC/SP2M/SiNaPS, 17 rue des Martyrs, 38054 Grenoble cedex 9, France, CNRS-LTM, UMR 5129, 17 rue des Martyrs, 38054 Grenoble cedex 9, France, and ESRF, 6 rue Jules Horowitz, BP220, 38043 Grenoble cedex 09, France Received July 3, 2008; Revised Manuscript Received September 22, 2008
ABSTRACT The presence of gold on the sidewall of a tapered, single silicon nanowire is directly quantified from core-level nanospectra using energyfiltered photoelectron emission microscopy. The uniform island-type partial coverage of gold determined as 0.42 ( 0.06 (∼1.8 ML) is in quantitative agreement with the diameter reduction of the gold catalyst observed by scanning electron microscopy and is confirmed by a splitting of the photothresholds collected from the sidewall, from which characteristic local work functions are extracted using a model of the full secondary electron distributions.
Semiconductor nanowires are currently inspiring intense research efforts due to unprecedented applications in highperformance nanoscale electronic, photonic, and thermoelectric devices.1 Silicon nanowires (Si NWs) can be prepared with single-crystalline structures and yield enhanced performances in field-effect transistors.2 Most are grown using the vapor-liquid-solid (VLS) mechanism3 from liquid alloyed catalyst droplets during silicon chemical vapor deposition. These structures were long considered as perfect crystals with straight and clean sidewalls. However, recent work suggests structural imperfections such as sidewall faceting upon growth in an ultrahigh vacuum (UHV),4 which could have considerable impact on the NW final properties. Chemical imperfections have also been reported,5 providing evidence that the catalyst droplet evolves during growth. Hannon et al. observed consumption of the gold catalyst droplet during growth determining an upper limit to the NW length and causing sidewall tapering6 and gave convincing arguments for gold migration from the catalyst on the NW sidewall. However, the direct identification of the presence of gold on the sidewall was beyond the scope of their work. Recently, a qualitative, but very local, identification was reported for technologically relevant Si NWs grown under controlled conditions.7,8 However, a more global, nondestructive me* Corresponding author. E-mail address:
[email protected]. † CEA-LETI, MINATEC. ‡ CEA-DSM/IRAMIS/SPCSI. § CEA-DSM/INAC/SP2M/SiNaPS. | CNRS-LTM. ⊥ ESRF. 10.1021/nl801952a CCC: $40.75 Published on Web 10/25/2008
2008 American Chemical Society
soscopic analysis (i.e, all along the sidewall of a wire), which would include a reliable quantification of the sidewall surface chemistry, is still lacking. Such quantified information has important implications on the investigations of defective NWs and the understanding of growth mechanisms. In this letter, we show how gold present on the sidewall of a single, tapered Si NW, due to diffusion from the catalyst, can be directly identified and quantified with high reliability using a surface-sensitive, spectro-microscopic technique, energyfiltered X-ray photoelectron emission microscopy (XPEEM). The Si NWs were grown by the VLS method using lowpressure chemical vapor deposition with a silane gas precursor at 0.1 mbar of partial pressure and at a temperature of 650 °C. After growth, the silicon nanowires were dispersed on a silicon dioxide substrate covered with 60 µm-size gold square patterns in order to facilitate their location and also to provide an energy reference for the XPEEM gold signal collected from the wires. The wire of interest was located using scanning electron microscopy (SEM). Prior to XPEEM experiments, heating at 700 °C in an ultra-high vacuum (1 × 10-10 mbar) for 2 min removed most of the adventitious carbon. An X-ray photoelectron spectroscopy survey spectrum recorded less than 1% residual surface contamination, low enough not to have a significant influence on the spectroscopic measurements. The quantitative results we provide in the following strongly support that the surface preparation adopted here did not induce any interference between the gold substrate and the gold located on the nanowire. SEM analysis performed after the XPEEM experi-
Figure 1. Scanning electron micrographs of the Si NW taken with (a) secondary and (b) backscattered electrons, showing the gold catalyst and the tapered shape.
ments on the same NW did not show significant changes regarding the tapered shape. XPEEM is one of the most promising full-field imaging methods to have emerged recently,9-11 providing nondestructive spectromicroscopic analysis. Aberration-corrected energy filtering, using either time-of-flight techniques12 or hemispherical analyzers,13,14 is important for surface nanoanalysis of elemental, chemical, and electronic states. The microscope (NanoESCA, Omicron NanoTechnology) features a fully electrostatic PEEM column together with an aberrationcorrected energy filter consisting of two hemispherical analyzers coupled by a transfer lens.13 The imaging mode works at high energy resolution and with photoelectron kinetic energies up to 1.6 keV without deteriorating the spatial resolution. Soft X-ray radiation from the ID08 beamline at the European Synchrotron Radiation Facility (ESRF) in Grenoble operated in the multibunch mode was used. The photon flux density on the sample, in a typical field of view of 25 µm, was ∼1.4 × 1010 ph s-1 0.1% BW-1 at 700 eV. Core-level XPEEM imaging was performed at 500 eV for both Si 2p and Au 4f7/2 core levels, with an overall energy resolution of 0.9 eV. For threshold spectromicroscopy, the photon energy was 500 eV, and the overall energy resolution was 0.48 eV. The final XPEEM data, that is, the laterally and energetically resolved photoemitted intensity Ij(x, y, Ej), where E is the electron energy referenced to the Fermi level of the sample (Ef, see below), are collected as a three-dimensional data set formed by single-energy images recorded across the selected spectral feature in energy steps of 0.5 eV for core levels and 0.1 eV for secondary electrons. The data sets are easily exploited off-line by choosing any region of interest and extracting the corresponding I(Ej) curve. In this way, one can obtain nanospectra from selected zones along the imaged nanowire. The size of the contrast aperture had no effect on the threshold position. The residual energy dispersion of the analyzer along the vertical coordi3710
nate of the image was found to be less than 30 meV, negligible with respect to the energy shifts in the spectra extracted from the image series. The Si NW is shown on the SEM images (Figure 1) taken after the XPEEM experiments. The gold catalyst sits at the upper extremity (Figure 1a) and is identified by the contrast with the nanowire (Figure 1b). The diameter varies from 305 to 250 nm along the wire length (8.1 µm), evidencing that tapering occurred during the growth. We first present quantitative data regarding the local work function before addressing the results of the partial gold coverage along the NW using the core levels. Figure 2a-c shows the same nanowire observed with energy-filtered XPEEM at the photoemission threshold using secondary electrons of 4.2, 4.5, and 5.1 eV of kinetic energy. The overall shape of the wire, including the gold catalyst at the top, is distinctly observed at 4.5 eV in Figure 2b. The length is 8.7 µm, consistent with the SEM results given the typical 10% uncertainty for the field of view of the instrument. The image taken at 4.2 eV exhibits a shorter wire length, 8.3 µm, due to the lower intensity from the catalyst at this energy with respect to the wire itself, reflecting the difference in the work function of the catalyst and the wire sidewall: the higher work function is expected in the case of a gold-rich catalyst. At 5.1 eV, close to the gold work function, there is a contrast inversion between the wire and the substrate. The wire appears darker than the surrounding gold (Figure 2c). We present in Figure 2d the threshold nanospectra extracted from selected regions depicted in the images (b-c), for the secondary electron energy distribution over an energy range of 3-15 eV. Regions I and II cover the wire sidewall; region III covers the catalyst, and region IV is on the gold substrate. The results are independent of the absolute area of the region selected, as checked by considering smaller sampling regions for the catalyst, sidewall, and substrate. The inset of Figure 2d is a close-up of Nano Lett., Vol. 8, No. 11, 2008
Figure 2. (a-c) XPEEM images of the Si NW at the photoemission threshold, taken at different photoelectron energies relative to the Fermi level of the sample. Contrast aperture, 70 µm; extractor voltage, 12 kV; acquisition time, 2s. (d) XPEEM nanospectra of the secondary electron distribution from the regions of interest (200 × 200 nm2) depicted in b and c: on the wire sidewall (regions I and II), on the catalyst (region III), and outside the wire on the gold substrate (region IV). Inset: zoom of the threshold region. (e) Energy level diagram illustrating the determination of the local work function.
the threshold region from 3 to 7 eV. The energy scale on the abscissa axis is referred to the Fermi level, EF, of the sample surface (EF ) 0), as illustrated in Figure 2e. If EK denotes the kinetic energy of the photoelectrons measured at the entrance of the imaging analyzer, and E denotes the final-state energy relative to EF, then E ) EK + eVS + ΦA, where ΦA is the work function of the analyzer and VS the bias voltage applied to the sample surface. An electron having an initial state energy Ei just below EF, excited with photons of energy hν, will have a measured kinetic energy EK given by EK ) (Ei + hν) - eVS - ΦA. Thus, the threshold kinetic energy E0K is given by Φ - eVS - ΦA. The secondary electron energy distributions, S(E), presented in Figure 2d are thus characterized by a sharp threshold corresponding to the local work function Φ of the emitting region under consideration, provided ΦA is known. The substrate emission (region IV), is similar to the energy distribution of gold reported by Henke et al.,15 who showed moreover that the shape of the secondary electron distribution is independent of the incident photon energy in the 0.1-10 keV range. The local work function of the substrate is obtained from the fit (Figure 3a) to the secondary electron distribution S(E) Nano Lett., Vol. 8, No. 11, 2008
described by Henke et al.16 After correction for the Schottky effect (due to the high surface potential created by the extractor17), we obtain Φ ) 5.01 ( 0.01 eV, in good agreement with the literature.18 For the sidewall and catalyst regions of interest (I, II, and III), the threshold energy distribution has a double structure. The structure was also observed before in situ UHV annealing; consequently, it cannot be attributed to the preparation conditions. A similar threshold shape reported for macroscopically averaged measurements of the Cu(111) work function19 was attributed to the photoemission from step edges having a lower work function than the (111) planes. In our case, the intensity maximum is at the same energy as for the gold substrate, and there is a shoulder 1 eV lower in energy. The double structure suggests two chemically distinct surface regions present within the 200 × 200 nm2 area of the sidewall, with a work function difference of 1 eV. Similar behavior is observed for region III generated on the catalyst; however, the position of the low-energy threshold is about 4.3 eV, intermediate between that of the gold substrate and the sidewall. The shift of the low-energy threshold between the catalyst and the sidewall regions is 1 order of magnitude 3711
Figure 3. Henke’s fit of the nanospectra of Figure 2b, (a) on the gold substrate (region IV) and (b) on the sidewall surface of the Si NW (region II).
larger than the vertical-coordinate-induced dispersion of the energy filter and therefore is not attributable to an instrumental artifact. For the extraction of the local work function, one can normally either model the whole secondary electron energy distribution with Henke’s model or perform a fit limited to the threshold region itself using a complementary error function.17 However, for the particular case of the double threshold structure involving two convoluted secondary electron distributions, the secondary electron tails need to be taken into account. Therefore, we have extended Henke’s model,15,16 to a double surface termination of work function Φ1,2, in the region of interest. The secondary electron distribution can be written as S1,2(E) ) S1(E) + S2(E), with
{
Ai(E - Φi)
c for Ei+1 g E g Eci Si(E) ) (E - Φi + Bi)4 for E e Eci 0
(1)
where A1,2 are scaling factors, B1,2 are the fit parameters, and Ec1,2 are the initial energies of the corresponding Henke distribution for a single threshold. In Figure 3b, we present the fit for the sidewall (region II) using eq 1; the values of Φ1 and Φ2 for all regions are reported in Table 1. The uncertainty in the value of the work function varies between 4 and 20 meV. The extracted values for Φ2 are very close to that determined for the gold substrate. Thus, elemental gold is present not only on the NW sidewall but also in the residual catalyst after growth, as expected from the gold-silicon equilibrium phase diagram below the 3712
Table 1. Local Work Function Determined from the Fitting of the XPEEM Nanospectra at the Photoemission Threshold Generated over the 200 × 200 nm2 Regions Using Henke’s Function local work function (eV) region
Φ1
Φ2
NW (I) NW (II) NW (III) substrate (IV)
4.04 ( 0.01 4.06 ( 0.01 4.31 ( 0.02
5.05 ( 0.01 4.98 ( 0.01 4.99 ( 0.01 5.01 ( 0.01
eutectic temperature. On the sidewall, Φ1 is between 4.0 and 4.1 eV, much smaller than the work function of pure, intrinsic bulk silicon (4.55 eV). An average work function of 3.6 eV for 20-nm-diameter, randomly aligned Si NWs prepared by laser ablation was reported by field-emission measurements,20 and 4.6 eV was found for vertically aligned Si NWs after hydrogen plasma etching.21 Our value of Φ1 therefore falls in the middle of the range of the reported work functions for Si NW systems. On the catalyst (region III), Φ1 is 4.3 eV, in between the local work function related to the pure silicon NW phase and that of gold on the sidewall. Complete phase separation between Au and Si is expected below the bulk Au-Si eutectic temperature (363 °C); therefore, a 4.3 eV work function cannot be assigned to a bulk, gold-silicon alloy. From the reported morphological observations of solidified Au-Si droplets using SEM, the formation of two chemical terminations, Au and Si, is clearly shown.22,23 The formation of Au/Si surface alloys is notorious on Si surfaces with gold coverage of few monolayers.24,25 Therefore, it is Nano Lett., Vol. 8, No. 11, 2008
Figure 4. Core-level XPEEM images of the Si NW, with (a) Au 4f7/2 electrons (EK ) 416 eV) and (b) Si 2p electrons (EK ) 396 eV). Contrast aperture, 70 µm; extractor voltage, 12 kV; acquisition time, 15 min. The contrast inversion in the Si 2p image is due to the strong contribution of the gold loss spectrum from the substrate. (c) XPEEM Au 4 f7/2 nanospectra, recorded with an energy step of 0.5 eV, along the Si NW and on the substrate. (d) Profile of the gold coverage as calculated from the Si 2p and Au 4f7/2 intensities; the dotted line indicates the average value of θAu.
not excluded that such an alloy forms in the topmost surface layers of the silicon phase upon cooling of the catalyst. Because of the extreme surface sensitivity of our analysis, only this Au/Si surface alloy is probed in the pure Si region, with a work function in between that of pure Si and pure gold.26 Figure 4a,b presents the background-subtracted XPEEM Au 4f7/2 and Si 2p core-level images. The corresponding nanospectra for gold extracted are shown in Figure 4d. As expected, there is a strong gold substrate signal at 84 eV of binding energy with a full-width at half-maximum of 1 eV, consistent with the overall energy resolution of the analysis. The Au 4f7/2 intensity in the XPEEM image on the wire is much weaker. This high apparent contrast is due to the respective signal-to-background ratios of the wire and the substrate. However, by extracting nanospectra along the wire (Figure 4c), we find a significant Au 4f7/2 signal. The nanospectra are fitted with the same parameters as for the gold substrate. The Au 4f7/2 intensities of regions I-III are half that of the substrate. We generated additional Au 4f7/2 nanospectra every 500 nm and found that the intensity relative to the substrate, ranging between 0.4 and 0.5, was position-independent over the entire length of the sidewall. This supports the conclusion that elemental gold is evenly Nano Lett., Vol. 8, No. 11, 2008
but discretely distributed along the sidewall. The Si 2p XPEEM image recorded at 396 eV of kinetic energy displays only a weak contrast between the substrate and the Si NW. Due to the proximity of the Si 2p and Au 4f transitions, there is a background contribution to the Si 2p signal from the Au 4f energy losses at lower kinetic energies. Despite this, we were able to extract, from the same regions of interest as considered for the Au 4f7/2 transition, Si 2p core-level nanospectra with characteristic binding energies between 103 and 104 eV, as expected for oxidized Si. This is supported by the O 1s XPEEM image (not presented), showing oxygen mainly located on the sidewall. From the integral intensities IAu and ISi of the core-level nanospectra, we can estimate the gold sidewall coverage as θAu )
IAu ⁄ σAu (ISi ⁄ σSi) + (IAu ⁄ σAu)
where σAu and σSi are the photoionization cross-sections of the Au 4f7/2 and Si 2p transitions. Taking Yeh and Lindau’s cross-sections,27 we obtain in Figure 4d θAu along the sidewall length. Given the uncertainties in the values of the cross-sections, we estimate the error in the coverage determination to be 0.1. The mean coverage over the 11 regions of interest is 0.42 with a standard deviation of 0.06. It is 3713
nearly constant along the sidewall, as expected for a diffusion process in a permanent regime. The hemispherical catalyst diameter decreases from 305 to 250 nm during growth (SEM image, Figure 1), liberating ∼2 × 108 Au atoms for migration along the sidewall. An independent STEM analysis,8 on NWs grown under similar conditions to those considered here, demonstrated that Au is present as 3-nm-diameter droplets on the sidewall, whose total area is therefore ∼2 × 106 nm2. The coverage of 0.42 determined from our XPEEM results would therefore require ∼4 × 105 droplets, corresponding to ∼1.9 × 108 Au atoms: this is in excellent agreement with the gold loss from the catalyst derived from the SEM image of Figure 1. We cannot totally exclude a contribution of the substrate to the surface amount of gold determined from our measurements; however, from this quantitative comparison, it appears that a possible substrate contamination does not significantly affect the results. Therefore, the mesoscopicscale, quantitative analysis by XPEEM shows that the gold present on the sidewall surface is distributed homogeneously along the entire length of the wire, corresponding to ∼1.8 ML of effective coverage with respect to the sidewall surface area. This figure agrees with the estimate from Hannon et al.6 In conclusion, we have performed quantitative energyfiltered XPEEM imaging with soft X-ray excitation of the surface chemistry of a single Si NW grown by the VLS process, and we show how the technique can be used for the nondestructive, quantitative surface analysis of single nanowires on the mesoscopic scale. The presence of elemental gold on the tapered Si NW sidewall is directly measured and quantified using XPEEM. Both threshold and core-level spectra provide a consistent picture of the sidewall surface. A partial gold coverage is derived from the XPEEM measurements along the whole length of the wire sidewall and is determined to be 0.42 ( 0.06 (∼1.8 ML). It is shown, by a quantitative comparison with morphological results using SEM and STEM, that gold analyzed by XPEEM comes mostly from the surface diffusion of the catalyst on the sidewall. Given the spatial resolution and sensitivities with core-level electrons, the method can be extended to much smaller nanowires and nanoscale heterostructures of varying chemical natures, providing hitherto unavailable physical parameters and contributing to our understanding of how growth conditions impact the final NW properties. Acknowledgment. This work was supported by the French ANR “XPEEM” project nr 05-NANO-065. Support from the Basic Technological Research Program (RTB) at CEA-LETI is also gratefully acknowledged. Special thanks go to the
3714
technical staff of the ESRF ID08 beamline for their help during the experiments. References (1) Lu, W.; Lieber, C. M. J. Phys. D: Appl. Phys. 2006, 39, R387–R406. (2) Cui, Y.; Zhong, Z.; Wang, D.; Wang, W. U.; Lieber, C. M. Nano Lett. 2003, 3 (2), 149–152. (3) Wagner, R. S. Whiskers Technology; Wiley: New York, 1970. (4) Ross, F. M.; Tersoff, J.; Reuter, M. C. Phys. ReV. Lett. 2005, 95 (14), 146104-1146104-4. (5) Pan, L.; Lew, K.-K.; Redwing, J. M.; Dickey, E. C. J. Cryst. Growth 2005, 277 (1-4), 428–436. (6) Hannon, J. B.; Kodamka, S.; Ross, F. M.; Tromp, R. M. Nature 2006, 440, 69–71. (7) Dhalluin, F.; Desre, P. J.; den Hertog, M. I.; Rouviere, J.-L.; Ferret, P.; Gentile, P.; Baron, T. J. Appl. Phys. 2007, 102 (9), 094906–5. (8) den Hertog, M. I.; Rouviere, J.-L.; Dhalluin, F.; Desre, X. J. P.; Gentile, P.; Ferret, P.; Oehler, F.; Baron, T. Nano Lett. 2008, 8 (5), 1544– 1550. (9) Rempfer, G. F.; Skoczylas, W. P.; Hayes Griffith, O. Ultramicroscopy 1991, 36 (1-3), 196–221. (10) Swiech, W.; Fecher, G. H.; Ziethen, C.; Schmidt, O.; Schonhense, G.; Grzelakowski, K. M.; Schneider, C.; Fromter, R.; Oepen, H. P.; Kirschner, J. J. Electron Spectrosc. Relat. Phenom. 1997, 84 (1-3), 171–188. (11) Bauer, E.; Koziol, C.; Lilienkamp, G.; Schmidt, T. J. Electron Spectrosc. Relat. Phenom. 1997, 84 (1-3), 201–209. (12) Schonhense, G.; Oelsner, A.; Schmidt, O.; Fecher, G. H.; Mergel, V.; Jagutzki, O.; Schmidt-Bocking, H. Surf. Sci. 2001, 480 (3), 180–187. (13) Escher, M.; Weber, N.; Merkel, M.; Ziethen, C.; Bernhard, P.; Scho¨nhense, G.; Schmidt, S.; Forster, F.; Reinert, F.; Kro¨mker, B.; Funnemann, D. J. Phys.: Condens. Matter 2005, 17 (16), S1329– S1338. (14) Renault, O.; Barrett, N.; Bailly, A.; Zagonel, L. F.; Mariolle, D.; Cezar, J. C.; Brookes, N. B.; Winkler, K.; Kromker, B.; Funnemann, D. Surf. Sci. 2007, 601 (20), 4727–4732. (15) Henke, B. L.; Smith, J. A.; Attwood, D. T. J. Appl. Phys. 1977, 48 (5), 1852–1866. (16) Henke, B. L.; Liesegang, J.; Smith, S. D. Phys. ReV. B: Condens. Matter Mater. Phys. 1979, 19 (6), 3004. (17) Renault, O.; Brochier, R.; Roule, A.; Haumesser, P. H.; Kro¨mker, B.; Funnemann, D. Surf. Interface Anal. 2006, 38, 375–377. (18) Eastman, D. E. Phys. ReV. B: Condens. Matter Mater. Phys. 1970, 2 (1), 1. (19) Takeuchi, K.; Suda, A.; Ushioda, S. Surf. Sci. 2001, 489 (1-3), 100– 106. (20) Au, F. C. K.; Wong, K. W.; Tang, Y. H.; Zhang, Y. F.; Bello, I.; Lee, S. T. Appl. Phys. Lett. 1999, 75 (12), 1700–1702. (21) Cheng, T. C.; Shieh, J.; Huang, W. J.; Yang, M. C.; Cheng, M. H.; Lin, H. M.; Chang, M. N. Appl. Phys. Lett. 2006, 88 (26), 263118. (22) Gentile, P. T. D.; Dhalluin, F.; Buttard, D.; Pauc, N.; Den Hertog, M.; Ferret, P.; Baron., T. Nanotechnology 2008, 19 (12), 125608. (23) Ressel, B.; Prince, K. C.; Heun, S.; Homma, Y. J. Appl. Phys. 2003, 93 (7), 3886–3892. (24) Calliari, L.; Sancrotti, M.; Braicovich, L. Phys. ReV. B: Condens. Matter Mater. Phys. 1984, 30 (8), 4885. (25) Yeh, J. J.; Hwang, J.; Bertness, K.; Friedman, D. J.; Cao, R.; Lindau, I. Phys. ReV. Lett. 1993, 70 (24), 3768–3771. (26) Wong, K.; Vongehr, S.; Kresin, V. V. Phys. ReV. B: Condens. Matter Mater. Phys. 2003, 67 (3), 035406. (27) Yeh, J. J.; Lindau, I. At. Data Nucl. Data Tables 1985, 32 (1), 1–155.
NL801952A
Nano Lett., Vol. 8, No. 11, 2008
