arXiv:cond-mat/0603746v1 [cond-mat.mtrl-sci] 28 Mar 2006 ∗ h pia pcrm aigi trciefrotc ap- optics for attractive in it lies bandgap making Its spectrum, material. optical sensor the a as application its to tung- of substrate. formation the the at in- through carbide chemically case sten by this nanowire in enhanced that strain, time significantly duced first be the can for show growth results Our mechanism a 1950s. Fisher following by growth lead- tung- whisker oxide described surface of enhanced the to formation strains surface ing the the that at carbide propose sten We substrate. the ro ln.Ceia hrceiainsosta the WO that or crystalline shows are methane, characterization com- hydrogen, wires Chemical involves magnitude that alone. of process argon orders a two to of about pared yield by the increases nanowires methane the then surfaces forma- and tungsten hydrogen oxidized nanowire of first for exposure to that forces find driving We gives the tion. that into process tung- insight high-yield grown that new a have clear We with nanowires is form. unclear oxide nanowires it is sten the it While VLS, why to and 10]. related how not hy- 9, is argon, mechanism 8, growth of [7, the mixtures oxygen various and tungsten in drogen, when heated form groups are can Several whiskers substrates oxide particles. mechanism that catalyst different shown involve the have completely not on a does nanowires, by that oxide diame- grow Tungsten the hand, determines nanowire. other particle the 6]. catalyst of [5, the ter particle nanowire of catalyst gold the size liquified The which small, a in from grows mechanism, (VLS) liquid-solid [4]. sensors efficient nanowires make of surface/volume also high types can The new ratio [3]. allowing formed be dislocations, to bulk heterojuctions of creation through than be the rather can relaxation, mismatch radial lattice via accommodated devices. from arising planar strain over example, [2]. advantages For vari- certain sensors wide offer and demonstrated, Nanowires a photodiodes, nanoelec- been date, photodetectors, have in To devices including potential nanowire-based their of [1]. ety of optoelectronics and because tronics part, in grow, lcrncades [email protected] address: Electronic ugtnoiehsrcie osdrbeatnindue attention considerable received has oxide Tungsten h aoiyo aoie r rw sn h vapor- the using grown are nanowires of majority The to continues nanowires semiconducting in Interest hita Klinke, Christian efidta xoueo xdzdtnse lst yrgnan hydrogen to films tungsten oxidized of exposure that find We h omto fadnearyo yial 0n imtrnan diameter nm 10 WO crystalline typically are of wires array the that dense shows a analysis of formation the aoiegot.Ti ih eagnrlcnet applicabl concept, general a f be the might with This associated growth. strain nanowire interfacial which in mechanism B .J asnRsac etr 11KthwnRa,Yorkt Road, Kitchawan 1101 Center, Research Watson J. T. IBM ehv netgtdtefraino ugtnoienanowi oxide tungsten of formation the investigated have We ugtnoiennwr rwhb hmclyidcdstra chemically-induced by growth nanowire oxide Tungsten tal. et 3 n htW a omdon formed has WC that and 1]adFak 1]i the in [12] Franks and [11] ∗ ae .Hno,LneGga,Ktle etr n heo Av Phaedon and Reuter, Kathleen Gignac, Lynne Hannon, B. James 3 epooeaceial-rvnwikrgrowth whisker chemically-driven a propose We . tae n aoie rw t900 at grown nanowires and strate, 1: FIG. a eu o40 to up be can olwdb rwhpae n hnso oln to cooling slow then and phase, phase growth heating a a of involved by pressure process quartz base followed three-step a a Our to on Torr. evacuated furnace 1 was the furnace into The on introduced holder. tungsten were nm (100 – tungsten films either silicon) thin – evaporated substrates or foils, diameter The wires, cm used. 10 was standard furnace a first, tube the In growth. nanowire [15]. windows” ranging “smart applications to in [14], material sensors tung- promising from make a properties light, same oxide heat, The to sten potential. exposure op- applied by the an varied is, or be That can [13]. studied properties properties tical widely ‘chromogenic’ been its also to has due oxide Tungsten plications. H 2 eue w ieetCDssest investigate to systems CVD different two used We n )CH d) and , raino ugtncriestimulates carbide tungsten of ormation E irgah fa h rsietnse sub- tungsten pristine the a) of micrographs SEM loi te aoiesystems. nanowire other in also e e ne ieetCDconditions. CVD different under res wrs tutrladchemical and Structural owires. 4 µ w egt,N 09,USA 10598, NY Heights, own ehn t900 at methane d H + long. m 2 nteCDsse.e h nanowires The e) system. CVD the in ◦ ◦ ed to leads C ihb r )A + Ar c) Ar, b) with C in ouris 2

FIG. 3: Evolution of the morphology of the tungsten oxide nanowire film grown in the vacuum system with increasing temperature (from left to right). The comparison of pyrom- eter measurements along the heated filament with the corre- sponding SEM images allows to estimate the temperature in ◦ this series to be between 800 and 950 C.

perature range from 600◦C to 1000◦C. However, only at 900◦C did we observe a high yield, with wires longer than 1 µm. Above 900◦C tungsten oxide starts to sublimate in vacuum [13]. Surprisingly, extending the duration of the growth phase did not increase either the areal density FIG. 2: TEM micrographs showing the lattice fringes and or length of the wires. Furthermore, the yield was essen- the diffraction pattern (insets) of individual tungsten oxide tially unchanged when the total pressure of the growth nanowires. The long axis of the wires points in <001> direc- phase mixture was varied from 10 to 500 Torr. Varying tion of WO3. the methane/hydrogen ratio over a wide range also had no significant influence on the wire length or diameter. To summarize, high yield requires hydrogen pretreatment room temperature. In the heating phase, the tempera- followed by exposure to methane. ture is ramped to 900◦C at a rate of 50◦/min in a gas In order to determine the chemical composition and flow of 500 sccm Ar and 200 sccm H2 (total pressure of structure of individual nanowires we grew them directly 100 Torr). In the growth phase the temperature is held at on transmission electron microscopy (TEM) grids made 900◦C in a mixture of 200 sccm hydrogen and 500 sccm of tungsten. TEM images show that the wires are crys- methane for 30 min at a pressure of 100 Torr. Finally, talline, with fringes visible both perpendicular and paral- the samples were cooled to room temperature in a gas lel to the wire axis (Fig. 2). The corresponding lattice dis- flow of 500 sccm Ar. tances are 3.3 A˚ and 3.7 A,˚ consistent with the structure In the first experiments, nanowires were grown by heat- of WO3. The corresponding diffraction image (Fig. 2 in- ing W films (Fig. 1a) in an Ar flow in the tube furnace. set) suggests that the wires are monoclinic WO3, with the A scanning electron microscopy (SEM) image of a film <001> crystal axis oriented parallel to the long axis of exposed to 500 sccm Ar at 900◦C is shown in Fig. 1b. Un- the wire. The streaking in the direction orthogonal to the der these growth conditions the film partially dewets and long axis indicates significant twinning. Energy disper- roughens. The surface features are rounded and smooth sive X-ray (EDX) measurements and electron energy loss (i.e. no facets), and very few nanowires are observed. spectroscopy (EELS) performed directly on individual However, we found that by adding H2 to the gas flow (e.g. nanowires confirmed the presence of only tungsten and 500 sccm Ar + 200 sccm H2), we obtained a more uniform oxygen in the wires. X-ray photoelectron spectroscopy surface consisting of crystallites that were clearly faceted (XPS) was used to investigate the chemical composition (Fig. 1c), while the nanowire yield remained relatively of the nanowires and substrates after growth. The chemi- low. A dramatic change in the growth occurs when the cal shifts of the W 4f levels are compatible with a mixture growth stage flow includes methane (Fig. 1d). The yield of WC and WO3. Samples measured before growth show of nanowires for the mixed flow is about two orders of no carbide peak. As we describe below, tungsten car- magnitude higher than when either hydrogen or methane bide is formed on the W substrate during the exposure are used separately. The nanowires are often several mi- to methane. crons long, with a typical diameter of about 10 nm. Oc- In order to determine the source of the oxygen in the casionally, we observed extremely long nanowires of up nanowires, we grew them in a vacuum system with a base to 40 µm in length (Fig. 1e). pressure of about 10−7 Torr. The substrates consisted of We systematically varied the process conditions to resistively-heated W filaments. Nanowires were grown investigate the influence of temperature, gas pressure, by first backfilling the chamber to 200 mTorr with H2. growth time, and gas mixture on the yield and length of The filament was then heated for 30 min at 1000◦C (at the nanowires. We found that nanowires grew in the tem- the hottest point of the wire). Subsequently, the cham- 3

−7 ber was pumped to 10 Torr again and CH4 was intro- duced until the pressure reached 5 Torr. The filament was heated for 30 min in the methane atmosphere, and then cooled to room temperature. This process resulted in nanowire growth that was qualitatively similar to that achieved in the tube furnace: a high density of nanowires with a typical diameter of about 10 nm (Fig. 3a). How- ever, if the filament was flashed to 2000◦C just prior to processing, no nanowires were observed. We propose that the oxygen required for nanowire formation comes from the native oxide on the W substrates. The fact that the nanowire length can not be extended by extending the growth time supports this conclusion. In the vacuum system the filament was clamped to thick Cu feedthroughs, resulting in a significant temper- ature gradient along the length of filament. The gra- dient allowed us to systematically investigate the role of temperature on the nanowire growth under otherwise identical process conditions. A sequence of SEM images along the filament length (i.e. with increasing tempera- ture) is shown in Fig. 3. As the temperature is increased, FIG. 4: Proposed growth mechanism: The tungsten substrate the density and length of the nanowires decreases, while (1) gets oxidized in air (2). In the CVD system at elevated their diameter increases. As we describe below, this be- temperatures the tungsten oxide surface roughens and forms havior is consistent with conventional whisker growth. sharp faceted crystallites with the help of hydrogen (3). Intro- Whiskers are non-equilibrium structures that grow as a duced CH4 carburizes the underlaying tungsten. Carbide for- result of bulk dislocation motion (i.e. climb). In one mation in turn induces strain at the WC/WO3 interface (4). scenario described by Frank [16], the intersection of a The strain is relieved by formation of stress-free nanowires dislocation with the surface traces out a periodic loop. (5). Each time the loop is traversed, the area of the surface enclosed by the loop will be displaced by a Burger’s vec- tor. This mechanims gives rise to whiskers with irreg- ular cross-sections but smooth side walls, as shown in We propose that the strain driving the whisker forma- in Fig. 3c(inset). The shape cross section is determined tion is enhanced by carbide formation at the W/WO3 by the loop that the dislocation core traces at the sur- interface. In this picture carbide formation increases the face (the path is ultimately determined by irregularities interfacial strain (e.g. the W-W separation in WC in- in surface morphology). Whisker growth is thought to creases by %4), enhancing nanowire growth. A similar be governed by bulk vacancy motion (in this case, pre- enhancement has been observed in Sn films subjected sumably the tungsten oxide given the low mobility of to external loading [11, 12]. The difference is that in vacancies in W at 1000◦C). In simple geometries [16], the present case the strain is induced chemically, via car- the whisker growth rate is (approximately) proportional bide formation at the interface. The process is illustrated to the the inverse of the wire diameter. It is clear from schematically in Fig. 4. the data shown in Fig. 3 that nanowires with larger di- In order to clarify the role of the carbide at the in- ameters are shorter than those with smaller diameters, terface, we attempted to re-grow nanowires on the same consistent with that prediction. We conclude that the substrate. That is, we first grew nanowires in the vac- nanowires grow via dislocation climb in the oxide film. uum system on a pristine filament using the procedure While the growth mechanism is consistent with outlined above. The filament was then sonicated to re- whisker growth, the driving force for whisker formation move the nanowires, and left in air for several days. It is less clear. In many metal whisker systems (e.g. Sn) was then reintroduced into the vacuum system and sub- the oxidation of the whisker surface plays an impor- jected to the nanowire growth process for a second time. tant role in stabilizing the growth of elongated structures However, no nanowires grew during the second process- with large surface area to volume ratios. In the present ing. The filament was irreversibly changed in the first case the surface is already oxidized and the amount of processing, presumably because the carbide we measure oxide appears to be conserved. One possible driving using XPS is formed at the surface. Note that TEM force for whisker growth in the present case is interfacial measurements show no evidence for carbide formation in strain: whisker growth will convert strained oxide at the the nanowires. Subsequent processing does not produce W/WO3 interface to unstrained oxide in the nanowire. nanowires because either, (a) no oxide forms on the car- The relief of strain may be one important factor driving bide surface, or (b) further carbide formation does not whisker formation. change the strain state of the native oxide. We conclude 4 that carbide formation at the W/WO3 interface is as- that the nanowires are crystalline WO3 and that WC is sociated with the high nanowire yield. Hydrogen pre- formed at the surface of the tungsten substrates. The treatment is also required for high yield. As the Fig. 1c cross sectional shape and variation in nanowire length shows, hydrogen pretreatment significantly changes the with diameter are consistent with conventional whisker surface morphology, presumably by enhancing oxide dif- growth. We conclude that interfacial strain drives the fusion. H2 is known to enhance oxygen diffusion in semi- nanowire formation in a process analogous to that ob- conductors [17]. In whisker growth, the whisker diameter served in mechanically strained Sn films. This wire is determined by surface morphology. Hydrogen pretreat- growth mechanism is expected to be general, and may ment may help in forming the crystalline grains required be of use in enhancing the yield in other nanowire sys- for classical whisker growth. tems. In conclusion, our results show that exposure to both hydrogen and methane strongly enhances the formation We thank Bruce Ek for technical support. Chris- of nanowires on (natively) oxidized W films during CVD tian Klinke acknowledges financial support from the processing. Structural and chemical analysis indicates Alexander-von-Humboldt Foundation.

[1] Wang, Z. L. Adv. Mater. 2003, 15, 432. [2] Vayssieres, L. Adv. Mater. 2003, 15, 464. [3] Bj¨ork, M. T.; Ohlsson, B. J.; Sass, T.; Persson, A. I.; Thelander, C.; Magnusson, M. H.; Deppert, K.; Wallen- berg, L. R.; Samuelson, L. Appl. Phys. Lett. 2002, 80, 1058. [4] Wan, Q.; Li, Q. H.; Chen, Y. J.; Wang, T. H.; He, X. L.; Li, J. P.; Lin, C. L. Appl. Phys. Lett. 2004, 84, 3654. [5] Whisker Technology, ed. A.P. Levitt, Wiley-Interscience, New York, 1970. [6] Law, M.; Goldberger, J.; Yang, P. Annu. Rev. Mat. Res. 2004, 34, 83. [7] Okuyama, M. Japan. J. Appl. Phys. 1970, 9, 409. [8] Gu, G.; Zheng, B.; Han, W. Q.; Roth, S.; Liu, J. Nano Lett. 2002, 2, 849. [9] Li, X. L.; Liu, J. F.; Li, Y. D. Inorg. Chem. 2003, 42, 921. [10] Zhang, H.; Feng, M.; Liu, F.; Liu, L.; Chen, H.; Gao, H.; Li, J. Chem. Phys. Lett. 2004, 389, 337. [11] Fisher, R. M.; Darken, L. S.; Carroll, K. G. Acta Metall. 1954, 2, 368. [12] Franks, J. Acta Metall. 1958, 6, 103. [13] Bange, K. Sol. Energy Mater. Sol. Cells 1999, 58, 1. [14] Cantalini, C.; Sun, H. T.; Faccio, M.; Pelino, M.; San- tucci, S.; Lozzi, L.; Passacantando, M. Sens. Actu. B 1996, 31, 81. [15] Cui, H. N.; Costa, M. F.; Teixeira, V.; Porqueras, I.; Bertran, E. Surf. Sci. 2003, 532, 1127. [16] Frank, F. C. Philos. Mag. 1953, 44, 854. [17] Jones, R. Phil. Trans. Roy. Soc. A 1995, 350, 189.