NANO LETTERS
Au-Free Epitaxial Growth of InAs Nanowires
2006 Vol. 6, No. 8 1817-1821
Bernhard Mandl,†,‡ Julian Stangl,‡ Thomas Ma˚rtensson,† Anders Mikkelsen,§ Jessica Eriksson,§ Lisa S. Karlsson,| Gu1 nther Bauer,‡ Lars Samuelson,† and Werner Seifert*,† Solid State Physics, Lund UniVersity, Box 118, S-221 00 Lund, Sweden, Institut fu¨r Halbleiter- und Festko¨rperphysik, Johannes Kepler UniVersita¨t Linz, Altenbergerstr. 69, A-4040 Linz, Austria, Synchrotron Radiation Research, Lund UniVersity, Box 118, S-221 00 Lund, Sweden, and National Centre for High-Resolution Electron Microscopy (nCHREM), Lund UniVersity, Box 124, S-221 00 Lund, Sweden Received February 26, 2006; Revised Manuscript Received June 7, 2006
ABSTRACT III−V nanowires have been fabricated by metal−organic vapor-phase epitaxy without using Au or other metal particles as a catalyst. Instead, prior to growth, a thin SiOx layer is deposited on the substrates. Wires form on various III−V substrates as well as on Si. They are nontapered in thickness and exhibit a hexagonal cross-section. From high-resolution X-ray diffraction, the epitaxial relation between wires and substrates is demonstrated and their crystal structure is determined.
Semiconductor nanowires as one-dimensional structures and building blocks for nanodevices have received increased attention in recent years. Controlling the one-dimensional growth on a nanometer scale offers unique opportunities for combining materials, manipulating properties, and designing novel devices. For most cases, nanowire growth has been performed by using “catalysts” in the so-called vaporliquid-solid (VLS)1 or vapor-solid-solid (VSS) mechanisms.2,3 A very common catalyst metal is Au, which is, however, known to produce unwanted deep levels, especially in silicon.4 Alternatively, nanowires can be grown by socalled self-catalytic processes. Here one constituent of the wire material forms the catalytic droplet, enabling VLS growth on top of the wire, for example, in In(liquid)/InP.5 Also, a few examples of catalyst-free III-V nanowire growth have been reported so far; the detailed mechanisms for these cases are still under debate.6,7 In some cases, traces of other materials, for instance, oxygen, have been found to be involved in growth (oxygen-assisted growth).8 In this letter, we present a general method for producing epitaxial nanowires of III/V materials, without using Au* Corresponding author. E-mail:
[email protected]; tel: +46 46 2227671. † Solid State Physics, Lund University. ‡ Institut fu ¨ r Halbleiter- und Festko¨rperphysik, Johannes Kepler Universita¨t Linz. § Synchrotron Radiation Research, Lund University. | National Centre for High-Resolution Electron Microscopy (nCHREM), Lund University. 10.1021/nl060452v CCC: $33.50 Published on Web 07/13/2006
© 2006 American Chemical Society
particles as a catalyst. We demonstrate this for InAs, which has one of the highest electron mobilities within the family of III-V compounds and is therefore predestined for electronic applications of nanowires. It has been shown that InAs nanowires can be contacted and gated easily.9 With inbuilt barriers (e.g., InP), the functionality of such structures in single electron transistors10 and resonant tunneling devices11 has been demonstrated. Moreover, InAs has a high potential to be used in combination with Si for high-speed electronics applications. For this purpose, however, nanowires grown Au-assisted impose severe restrictions because of the introduction of deep-level defects into Si. We show in this letter that InAs nanowires can be grown epitaxially on various substrates without any metal catalyst by covering the substrates with a thin layer of SiOx (x ≈ 1) prior to InAs growth. X-ray diffraction measurements indicate that the wires form in part in the wurtzite modification and grow spontaneously in the c direction [000.1h], equivalent to the cubic [111] direction (in this letter, we use the [hkk′.l] notation for hexagonal structures, and the [hkl] notation for cubic structures). For wire-growth, we used low-pressure metal-organic vapor-phase epitaxy (LP-MOVPE) at a pressure of 10 kPa, with trimethylindium (TMI), arsine (AsH3), and phosphine (PH3) as precursor materials, transported in a flow of 6000 mL/min of H2 as carrier gas. For the precursors, typical molar fractions of 2 × 10-6 for TMI and 2 × 10-4 for AsH3 were used. For TMI, higher molar fractions were also tested but
Figure 2. Length of InAs wires grown on InP(111)B during 60 s as a function of growth temperature. For 680 °C too few wires are found to establish an error bar. The error bars represent the standard deviation. Figure 1. SEM images (45° tilt) of InAs wires grown on InP(111)B at different temperatures; the growth time was 60 s.
had no significant effect on the growth rate. The molar fraction for PH3 was varied in the range (3.5-15) × 10-3. As substrates, we used epitaxy-ready III/V wafers and Si wafers. In the case of Si substrates, the native oxide was removed by an HF dip. Before loading the substrates into the growth chamber, a thin SiOx layer was sublimated onto the surface. The substrates were then heated to the growth temperature between 520 °C and 680 °C in a H2 atmosphere. As soon as the growth temperature was reached, the precursors were switched on simultaneously. The growth was stopped by switching off the TMI source, and the samples were cooled under AsH3 flow, or for InAsP deposition, under additional PH3 flow. To characterize the wires, we employed scanning electron microscopy (SEM) and X-ray diffraction (XRD). SEM investigations were performed using a JEOL 6400 and a LEO Supra 35 microscope. From SEM we obtain the length, orientation, shape, and density of the wires. XRD experiments have been performed at beamline ID10B (Troı¨ka II) at the ESRF in Grenoble to get information on the orientation of the wires relative to the substrate and relative to each other, as well as to determine the lattice constant and the crystalline structure of the wires. SEM images of InAs wires grown at different temperatures on InP(111)B surfaces, covered by 1.3 nm SiOx prior to growth, are presented in Figure 1. We observe the following trends: (i) The wires grow spontaneously on the InP(111)surface and appear epitaxially oriented, that is, they are standing vertically on the surface and grow in continuation of the substrate [111]B orientation. Remarkably, the wires are homogeneously thick, that is, in contrast to typical MOVPE wire growth12,13 we cannot determine any measurable tapering or any thickening at the wire foot/substrate connection. (ii) The length of the wires is a function of growth temperature. Figure 2 presents the wire length for a growth time of 1 min. We see a broad maximum with a peak at around 600 °C. In general, the growth temperatures are much higher than those for Au-assisted growth of InAs 1818
nanowires in MOVPE, for which growth suddenly ceases when the temperature exceeds 500 °C.14 We should mention that the wire length is not a linear function of time: in a separate investigation, not presented here, we have seen that the length growth rate starts from a high value and decreases over time according to a power law R ∝ tn (n ≈ -0.5). This fact can be seen as a hint on the growth mechanism. (iii) The wire density is also a function of growth temperature. In general, the density decreases with increasing temperature. This indicates that there are relations to the general laws that govern the nucleation kinetics of clusters on surfaces, where the density, F, of critical nuclei follows the general proportionality F ∝ R/D(T), with R being the deposition rate and D(T) being the temperature-dependent coefficient of surface diffusion.15 This result indicates that clusters, or cluster-related nucleation sites, may be involved in the prestages of wire growth. At higher temperatures, these clusters grow anisotropically and form one-dimensional wires. At lower temperatures, only a part of the clusters adopts this growth mechanism, others grow by isotropic expansion instead. In fact, we find that at low temperatures the wires compete with InAs clusters nucleating on the surface (see the 540 °C sample in Figure 1). (iv) In parallel to the decreasing density, the aspect ratio length/width of the wires decreases with increasing temperature: At higher temperatures, the radial growth on the side facets gets more and more important. At the highest growth temperature of 680 °C, the aspect ratio dropped down to below 1 and we obtained epitaxially grown hexagonal platelets. (v) The morphology of the nanowires is rod-shaped with a regular hexagonal cross-section. The top of the thicker wires is flat, visible at least in the case of the wires grown at higher temperatures in Figure 1. This is also in agreement with the transmission electron micrograph (TEM) image shown in Figure 3. From this TEM image, it is also evident that no metal catalyst exists at the nanowire tip. In Figure 4 we show that in this growth mode InAs wires can also be grown in regular arrays by prepatterning the SiOx layer prior to growth. The pattern was prepared by spinning PMMA onto the InP(111)B substrates and writing a regular Nano Lett., Vol. 6, No. 8, 2006
Figure 5. SEM images (45° tilt with respect to the (001) normal) of InAs wires grown on an InAs(001) substrate. The bottom part with mainly InAs clusters shows the substrate surface, which was covered by 20 nm SiOx prior to growth. The upper part shows [1-11] wires grown at the (1-10) cleavage plane, where the oxide was thinner. The growth time was 8 min at 560 °C. Figure 3. High-resolution TEM image of a wire tip grown under SiOx-assisted growth conditions at 620 °C. The arrow shows the growth direction. No metal catalyst is found at the end of the nanowire.
Figure 4. SEM images (45° tilt) of InAs wires grown on InP(111)B using a prepatterned SiOx layer. Growth was initiated during heating at 500 °C and stopped after 240 s at 580 °C.
array of dots using electron-beam lithography. After developing the resist, the SiOx layer was deposited, and the remaining resist was removed in a lift-off process. To grow wires on such patterned SiOx films, the sources had to be activated already during the heating-up period at 510-520 °C. Activating the sources later, that is, at higher growth temperatures, leads to a loss of the pattern. This most probably has to do with the thermal stability of SiOx, which becomes mobile at temperatures above 500 °C. Scanning tunneling microscopy (STM) investigations have shown that a uniform 1.3-nm-thick SiOx layer on InAs(111) becomes very mobile (forming larger clusters or evaporating) by annealing at 500 °C in time scales comparable to our growth times.16 It should be noted that in our case the wires were not grown out of openings in the SiOx layer, but at the site where, after e-beam lithography and lift-off, islands of SiOx remain. This is in agreement with the STM measurements, which show that after heating SiOx covered InAs(111) or InP(111) at 500 °C the majority of the surface on both InAs and InP are crystalline and clean with dispersed clusters of disordered material.16 This nucleation and growth process is not yet understood in detail. Obviously, the SiOx islands Nano Lett., Vol. 6, No. 8, 2006
act as “catalysts” similar to Au particles in case of patterned growth in the InP(Au)/InP(111)B system.17 The best results, that is, wires with high homogeneity in size and growing vertically, were obtained for InAs on InP(111)B surfaces, covered by 1.3 nm SiOx. The process is, however, not limited to this material combination alone. A very interesting result was obtained when we used InAs(001) substrates. Covering the substrate surface by an about 20-nm-thick oxide layer, the SiOx was actually acting as a mask, preVenting wire growth on the InAs(001) surface. We found in this case, however, a high density of wires growing at the {110} cleavage planes, see Figure 5. These cleavage planes, although not in the perpendicular direction of the SiOx evaporation beam, obviously got covered by a sufficiently thick SiOx layer to enable wire growth. It is worth mentioning that a lattice mismatch does not prevent epitaxial wire growth in the investigated growth mode. InAs wires were also grown on other III-V substrates, as, for instance, GaAs (mismatch ≈ 7.2%) and GaP (mismatch ≈ 11%). The versatility of the process is demonstrated in two further examples: First, with our method InAs wires can also be grown epitaxially on Si substrates (mismatch ≈ 12%), making the method attractive for applications with an integration of such wires with Si microelectronics technologies. We could demonstrate this for InAs growth on Si(001), where we clearly see that the wires grow randomly in the four 〈111〉 directions available on a Si(001) surface, as shown in Figure 6. Second, we have grown alloyed InAsxP1-x nanowires by adding PH3 to the other precursors into the reactor cell, which also opens the field of heterostructure growth for this growth method. Details will be published elsewhere. The crystalline structure of the nanowires was obtained from XRD experiments. To reduce the scattering contribution from the substrate, we used grazing incidence diffraction geometry, with incident and exit angles close to the critical angle of external reflection for the substrate. From reciprocal space maps (RSMs), we determined the lattice plane spacings parallel to the sample surface. Additional RSMs around the 1819
Figure 6. SEM images (top view) of InAs wires grown on Si(001) at 620 °C; the growth time was 60 s.
(111) Bragg peak of the substrate also give the lattice plane spacing along the growth direction. The InAs peaks are found to be very close to where they are expected for fully relaxed InAs, proving that the InAs wires grow epitaxially on InP(111)B. Streaks along the 〈110〉 directions are observed in the RSMs, consistent with the hexagonal wire shape with {110} side facets observed in SEM. In InAs wires grown by chemical beam epitaxy (CBE) at lower temperatures, around 390 °C, wurtzite (WZ) rather than zinc blende (ZB) crystal structure was observed, with consequences for the electronic properties.18 Hence, we investigated whether the wires grown by our method at higher temperatures also contain wurtzite parts. From transmission electron micrograph (TEM) images, a rather large density of stacking faults in the nanowires was detected, where the stacking sequence changes from (ABCABC) along the [111] growth direction to an (ABAB) sequence. Figure 7 depicts this situation: panel a sketches the (ABC) sequence of a cubic fcc lattice. Only one sublattice of zinc blende InAs is shown; the other sublattice is shifted by a quarter of the space diagonal along [111]. The cube edge in a is an “A” atom, lighter and darker gray spheres and planes indicate the “B” and “C” atoms and planes, respectively. The corresponding (AB) stacking sequence is shown in Figure 7b, where the ”C” layer is shifted to become an “A” layer. Without any other shift of atoms than within (111) planes, the resulting crystal structure is hexagonal close packed, that is, including the second sublattice, the structure is wurtzite. In XRD, the two alternative lattice structures can be distinguished easily: In reciprocal space, the ZB [111] direction is equivalent to the WZ [000.1] direction and the
Figure 7. Sketch of the cubic (a) and hexagonal (b) close-packed structure, corresponding to one sublattice of the zinc blende and wurtzite lattices, respectively. Base vectors and cubic directions are indicated. 1820
Figure 8. RSMs measured in GID geometry around indicated reflections of InP and InAs. The axes are scaled to show the same distance between InAs and InP reflections.
ZB [112h] direction is equivalent to the WZ [1h01.0] direction. Considering the complete lattice structure and the corresponding selection rules for Bragg reflections in reciprocal space, the lowest order reflection in growth direction is the (1h1h1h) Bragg peak for ZB, which corresponds to the (000.2h) peak of WZ. Perpendicular to the growth direction, the ZB (224h) peak is equivalent to the WZ (3h03.0) peak. However, in WZ the (1h01.0) and (2h02.0) peaks are also allowed, whereas in ZB no corresponding reflections exist. This gives us a straightforward method to investigate whether the InAs nanowires grow in the ZB or in the WZ crystal structure. Using GID, we recorded the intensity distribution around the WZ (1h01.0), (2h02.0), and (3h03.0) peaks. The corresponding RSMs of a sample with InAs wires grown for 60 s on InP(111)B at a temperature of 620 °C are shown in Figure 8. The peaks due to InP and InAs are indicated. Streaks in the direction perpendicular to wire side facets, indicated by the dotted lines, are visible, which is proof that the InAs peak is due to the wires. For the ZB-forbidden reflections, the InP substrate peaks vanish almost completely, while the (3h03.0), which is equivalent to the ZB (224h), shows a strong InP peak. Contrary, for InAs, which constitutes the nanowires, all reflections are similarly strong. Hence, it is clear that at least a considerable part of the InAs wires has WZ structure. To quantify this amount, we compared the total integrated intensity of the three reflections. If all InAs were wurtzite, then these intensities should follow the structure factor of WZ. If some material grows in the zinc blende lattice, then the (3h03.0) will contain additional intensity from the cubic part of the wires. The analysis reveals exactly this behavior, yielding about 20% of InAs in the WZ modification. From the lattice plane spacings, we obtain lattice parameters of the wurtzite InAs of a ) 4.274 Å and c ) 6.985 Å, respectively. Hence, we obtain a c/a ratio of 1.640, slightly larger than the ideal value of 1.633 for the hexagonal closepacked structure. As parts of the wires grow in the ZB structure, which has a lower lattice plane spacing in the [111] direction, the c value might be slightly underestimated. Within the (111) plane, we obtain exactly the same lattice spacings for both the WZ and the ZB + WZ reflections. The tendency of III-V materials to adopt the thermodynamically metastable WZ instead of the stable ZB modificaNano Lett., Vol. 6, No. 8, 2006
tion is characteristic for several III-V materials when growing as wire structures. This has been reported, for instance, for InAs wires grown Au-assisted at 450 °C by Takahashi and Moriizumi.19 These wires were reported to exhibit lattice parameters of a ) 4.27 Å and c ) 7.02 Å. The c/a ratio of 1.644 also shows a positive deviation from the ideal value for the WZ modification, like in our case. This is characteristic of materials whose stable modification is ZB20 and in fact, their wires converted to the stable ZB modification after annealing for 5 min at 800 °C. InAs grown Au-assisted in CBE at a temperature of 390 °C is also reported to crystallize in WZ.18 Similar observations, that is, crystallization of WZ segments, were made for GaAs, grown by CBE21 as well as InP grown by MOVPE.22 It seems that this tendency to adopt the metastable modification depends on two factors: (1) it is characteristic for growth in the c direction [000.1] or [111], respectively: InAs as well as InP, grown under similar conditions in the [001] direction, develop in a perfect ZB-structure;13,23,24 (2) it is characteristic for growth under high supersaturations: parts of wires growing under lower supersaturation, as, for instance, the typical “neck” parts below Au particles in Auassisted GaAs wire growth, develop in perfect ZB structure.21 This indicates that the formation of metastable modifications in wire growth is related to the specifics of the growing interface between the supersaturated mobile phase and the solid phase; that is, it is most probably the nucleation phenomenon well known as “Ostwald’s rule”:25 “The phase that nucleates needs not to be the stable phase, but the one that is closest in free energy to the parent phase”. In this sense, the observation of the metastable phase in the growth of nanowires can be attributed to well-known phenomena in crystal growth. In conclusion, we present a method for growing InAs wires in MOVPE without using Au or another metal as a “catalyst”. The “catalyst” function in our case is taken over by a thin film of SiOx, evaporated onto the substrate surface prior to growth. Wires of InAs grow randomly or in patterns epitaxially oriented on substrate surfaces. The wires are straight and homogeneous in thickness over the length, with no measurable tapering. The growth temperatures are generally higher than those for Au-assisted growth, which could be an advantage concerning the better CH4-elimination process, leading to lower C incorporation. The structure of the wires is in part wurtzite, which is a thermodynamically metastable modification of InAs. Because this method avoids the use of metallic particles, such as Au, as “catalysts”, it is a promising way to add high-mobility materials, such as InAs, to Si-based microelectronics.
Nano Lett., Vol. 6, No. 8, 2006
Acknowledgment. This work was carried out within the Nanometer Structure Consortium in Lund and was supported by grants from the Swedish Research Council (VR), the Swedish Foundation for Strategic Research (SSF), the FWF Vienna, the EC (“SANDiE” NMP4 CT-2004-500101), EU Education Program Sokrates/Erasmus and “Kepler Internationalisierungsprogramm KIP”, and the ESRF, Grenoble. We would like to thank the staff of ID10B for beamline setup and assistance, and Mariusz Graczyk for SiOx depositions. References (1) Wagner, R. S.; Ellis, W. C. Appl. Phys. Lett. 1964, 4, 89. (2) Kamins, T. I.; Williams, R. S.; Basile, D. P.; Hesjedal, T.; Harris, J. S. J. Appl. Phys. 2001, 89, 1008. (3) Persson, A. I.; Larsson, M. W.; Stenstro¨m, S.; Ohlsson, B. J.; Samuelson, L.; Wallenberg, L. R. Nat. Mater. 2004, 3, 677. (4) Pantelides, S. T. Deep Centers in Semiconductors; Gordon and Breach: New York, 1986. (5) Novotny, C. J.; Yu, P. K. L. Appl. Phys. Lett. 2005, 87, 203111. (6) Poole, P. J.; Lefebvre, J.; Fraser, Appl. Phys. Lett. 2003, 83, 2055. (7) Motohisa, J.; Noborisaka, J.; Takeda, J.; Inari, M.; Fukui, T. J. Cryst. Growth 2004, 272, 180. (8) Lee, S. T.; Wang, N.; Lee, C. S. Mater. Sci. Eng. 2000, A286, 16. (9) Bryllert, T.; Samuelson, L.; Jensen, L. E.; Wernersson, L. DeVice Research Conference Digest 2005, p 157. (10) Thelander, C.; Mårtensson, T.; Bjo¨rk, M. T.; Ohlsson, B. J.; Larsson, M. W.; Wallenberg, L. R.; Samuelson, L. Appl. Phys. Lett. 2003, 83, 2052. (11) Bjo¨rk, M. T.; Ohlsson, B. J.; Thelander, C.; Persson, A. I.; Deppert, K.; Wallenberg, L. R.; Samuelson, L. Appl. Phys. Lett. 2002, 81, 4458. (12) Hiruma, K.; Yazawa, M.; Katsuyama, T.; Ogawa, K.; Haraguchi, K.; Koguchi, M.; Kakibayashi, H. J. Appl. Phys. 1995, 77, 447. (13) Seifert, W.; Borgstro¨m, M.; Deppert, K.; Dick, K. A.; Johansson, J.; Larsson, M. W.; Mårtensson, T.; Sko¨ld, N.; Svensson, C. P. T.; Wacaser, B. A.; Wallenberg, L. R.; Samuelson, L. J. Cryst. Growth 2004, 272, 211. (14) Dick, K.; Deppert, K.; Mårtensson, T.; Mandl, B.; Samuelson, L.; Seifert, W. Nano Lett. 2005, 5, 761. (15) Venables, J. A. Surf. Sci. 1994, 299/300, 798. (16) Mikkelsen, A. et. al, to be submitted for publication, 2006. (17) Mårtensson, T.; Borgstro¨m, M.; Seifert, W.; Ohlsson, B. J.; Samuelson, L. Nanotechnology 2003, 14, 1255. (18) Bjo¨rk, M. T.; Fuhrer, A.; Hansen, A. E.; Larsson, M. W.; Fro¨berg, L. E.; Samuelson, L. Phys. ReV. B 2005, 72, 201307(R). (19) Takahashi, K.; Moriizumi, T. Jpn. J. Appl. Phys. 1966, 5, 657. (20) Lawaetz, P. Phys. ReV. B 1972, 5, 4039-4045. (21) Ohlsson, B. J.; Bjo¨rk, M. T.; Magnusson, M. H.; Deppert, K.; Samuelson, L.; Wallenberg, L. R. Appl. Phys. Lett. 2001, 79, 3335. (22) Mohan, P.; Motohisa, J.; Fukui, T. Nanotechnology 2005, 16, 2903. (23) Krishnamachari, U.; Borgstro¨m, M.; Ohlsson, B.; Panev, N.; Samuelson, L.; Seifert, W.; Larsson, M.; Wallenberg, L. Appl. Phys. Lett. 2004, 85, 2077. (24) Bjo¨rk, M. T.; Ohlsson, B. J.; Sass, T.; Persson, A. I.; Thelander, C.; Magnusson, M. H.; Deppert, K.; Wallenberg, L. R.; Samuelson, L. Appl. Phys. Lett. 2002, 80, 1058. (25) Ostwald, W. Z. Phys. Chem. 1897, 22, 289.
NL060452V
1821