NANO LETTERS
Plasmon-Assisted Chemical Vapor Deposition
2006 Vol. 6, No. 11 2592-2597
David A. Boyd,*,† Leslie Greengard,‡ Mark Brongersma,§ Mohamed Y. El-Naggar,† and David G. Goodwin† DiVision of Engineering and Applied Science, California Institute of Technology, Pasadena, California 91125, Courant Institute of Mathematical Sciences, New York UniVersity, New York, New York 10012, and Department of Materials Science and Engineering, Stanford UniVersity, Stanford, California 94305-4045 Received August 31, 2006; Revised Manuscript Received September 18, 2006
ABSTRACT We introduce a new chemical vapor deposition (CVD) process that can be used to selectively deposit materials of many different types. The technique makes use of the plasmon resonance in nanoscale metal structures to produce the local heating necessary to initiate deposition when illuminated by a focused low-power laser. We demonstrate the technique, which we refer to as plasmon-assisted CVD (PACVD), by patterning the spatial deposition of PbO and TiO2 on glass substrates coated with a dispersion of 23 nm gold particles. The morphology of both oxide deposits is consistent with local laser-induced heating of the gold particles by more than 150 °C. We show that temperature changes of this magnitude are consistent with our analysis of the heat-loss mechanisms. The technique is general and can be used to spatially control the deposition of virtually any material for which a CVD process exists.
Chemical vapor deposition (CVD) is used widely in many different fields to deposit thin films and coatings of many different materials. For example, it is used in the manufacture of microelectronic and micromechanical devices to lay down the metal, dielectric, and semiconductor layers that, when patterned and etched by photolithography, form the micrometer- or submicrometer-scale features that make up the device. Although this approach can be used to fabricate many types of microdevices, there are also many types of devices one can imagine for which the multilayer deposit/pattern/ etch approach is not ideal. For example, truly threedimensional micromechanical devices are difficult to fabricate in this way. Another problem arises due to the relatively high temperatures needed by many CVD processes because all previously deposited structures must be able to withstand the process temperature of every subsequent deposition step. All of these challenges become greater as features become smaller; for fabrication of devices with feature sizes on the order of 10 nm, it is clear that other fabrication methods will be required. An alternative to blanket deposition followed by patterning and etching is to directly write the device features. Stuke et al. have demonstrated that quite complex three-dimensional structures can be formed by using a focused laser beam to * Corresponding author. E-mail:
[email protected]. † California Institute of Technology. ‡ New York University. § Stanford University. 10.1021/nl062061m CCC: $33.50 Published on Web 10/11/2006
© 2006 American Chemical Society
locally heat the spot where deposition is desired.1,2 This method is known as laser-assisted CVD (LACVD). Using computer control of the lateral beam position as well as its focus, it is possible to create structures that would not be possible with conventional methods, for example, freestanding 3D microstructures.2 LACVD, however, requires relatively high laser power for most CVD precursors, and is difficult to use for deposition onto transparent substrates (e.g., glass or silica), which do not absorb enough energy from the laser to be heated significantly. In addition, the size of the deposit produced by LACVD is limited by the laser spot size, and this can be used to create micrometer-scale, but not nanoscale, devices. In this paper, we demonstrate a technique similar to LACVD that we call plasmon-assisted CVD (PACVD). This has advantages over LACVD that much lower laser powers are needed, deposition on transparent substrates is simple, and the size of the deposit can, in principle, be much smaller than the laser spot size, making it possible to fabricate nanostructures with this technique. Here we use PACVD to deposit two different oxides, both at relatively low temperature and with low laser power. Photon-electron interactions in nanoscale structures, especially metals, have been the subject of many studies.3 Although the basic physics underlying these interactions is complex, resonant phenomena at these length scales have already given rise to a number of useful technologies, from
new waveguides, filters, and light sources to new spectroscopic and microscopic imaging devices.4,5 In nanoparticles of metals such as gold and silver, excitation of surface plasmons (a free electron oscillation coupled to an electromagnetic wave) by an incident electromagnetic field leads to strong, resonant coupling to the field when the plasmon wavelength and particle size satisfy certain shape-dependent resonance conditions. For a single particle, the precise location of the resonant frequency and the breadth of the resonance peak are functions not only of particle size but of particle shape as well.6 Furthermore, arrayed particles interact with each other so that the geometry of the particle distribution can cause shifts in both the resonant frequency and width.7 Finally, the presence of multiple dielectric materials (such as the ambient medium, the substrate, and the metal) make the calculation of the spectrum more complex. The excitation of surface plasmons tends to be lossy, however, because of coupling of the electron wave to lattice vibrations on a picosecond time scale. As a result, energy absorbed at or near the plasmon resonant frequency is quickly converted to heat. In metal optics, this is a well-known and undesired side-effect. In other application areas, it has been suggested as a useful tool, including the far-field optical detection of large molecules,8 the denaturation of proteins,9 and thermal treatment for cancer using gold nanoshells.10 The ability to easily heat nanoparticles of gold, silver, or other metals with a visible-wavelength laser suggests that this could allow very localized heating for selective chemical vapor deposition, if a dispersion of these particles coated a substrate and were illuminated with a laser near the plasmon resonance frequency. If many nanoscale metal particles are illuminated by the micrometer-scale focused laser beam, then this would result in a micrometer-scale deposit, similar to LACVD. However, as will be discussed in more depth below, it may be possible to create nanoscale structures much smaller than the laser wavelength or spot size. The CVD process we propose, which we call plasmonassisted chemical Vapor deposition (PACVD), is illustrated in Figure 1. A template of nanoscale gold lines or dots is first laid down on a substrate by any suitable technique (ebeam lithography, or block copolymer lithography, for example). The substrate is then exposed to a gaseous environment containing the CVD precursor in a carrier gas. However, unlike a conventional CVD process to deposit a blanket film, in which the entire substrate is held for considerable time at typical CVD process temperatures (400-1000 °C), the substrate is kept at a temperature low enough that chemistry does not occur. To initiate CVD growth on the gold, a low power visiblewavelength laser is focused onto the substrate, with a wavelength chosen to coincide with the plasmon resonance frequency in the gold nanostructure. The laser resonantly excites surface plasmons in the gold nanostructures, which then transfer energy to lattice vibrations on a picosecond timescale. As a result, substantial, very rapid heating of the gold is possible. At the same time, heat loss from the gold via conduction to the surrounding substrate is inhibited Nano Lett., Vol. 6, No. 11, 2006
Figure 1. Schematic of PACVD process. A template of nanoscale gold lines or dots is first laid down on a substrate (top). The substrate is then exposed to a gaseous environment containing the CVD precursor in a carrier gas, and a laser (green) is focused on the surface heating the nanoparticles (middle). Growth is initiated only on the heated nanoparticles (bottom).
because of the poor thermal coupling of the gold particle whose size is much smaller than the phonon mean free path in the substrate. This allows temperature rises of several hundred degrees to be achieved with only a low-power laser and minimizes heating of the surrounding substrate, even where it is directly illuminated by the laser. When the temperature of the plasmon-heated gold structure reaches the range needed to initiate decomposition of the CVD precursor on the gold particle surface, deposition begins on the gold, and involves essentially conventional CVD chemistry. The novel aspect is that the heating can be localized in space and time, allowing use of high-temperature CVD processes without excessive heating of the substrate. To explore the characteristics of PACVD, two different metal oxide systems were investigated. In the first, the metal (Pb) forms a low-temperature eutectic with gold, whereas in the second, the metal (Ti) does not. Gold nanoparticles were deposited on a glass microscope cover slip using the block copolymer lithography (BCPL) method.11 Shown in Figures 2, 4, and 5 are SEM images of deposits created by PACVD using the lead precursor. Because of size, only the thicker lines and larger dots were able to be identified by micro-Raman spectroscopy as orthorhombic PbO (massicot).12 Depositions occurred with and without 2593
Figure 2. SEM image of a series of lines deposited by PACVD at scan rates of 1, 5, 3, 3, and 2 µm/s, bottom to top, respectively. The dark area in the right of the of the image running across the lines is the scratch as noted in the text.
Figure 4. SEM image of a deposit created by PACVD using a short exposure at half of the full laser power.
Figure 5. PbO nanowires created by PACVD. Figure 3. SEM image of a line deposited by PACVD at scan rate of 1 µm/s.
oxygen, and the addition of oxygen increased the deposition rates. Figure 2 shows a series of lines deposited at scan rates of 1, 5, 3, 3, and 2 µm/s (bottom to top, respectively). As can be seen, the amount of material deposited along the line is inversely proportional to the scan speed. Figure 3 is a higher magnification image of a line deposited at a rate of 1 µm/s exhibiting a nanowhisker morphology. Figure 4 is a deposit that formed at half the maximum laser power and with an exposure that stopped just as the reflected signal began to diminish, limiting the deposition to the individual nanoparticles. The diameter of the spot is approximately that of the focused laser beam, which is 1.5 ( 0.1 µm. Note that no deposition occurred on areas of the sample where Au particles were not present, nor did deposition occur on gold particles not illuminated by the laser. Visible in Figure 2 is a scratch that extends along the lines. The scratch occurred during or after the deposition of the gold nanoparticles and before the PACVD deposition. There are no gold particles along the scratch, and it can be seen that deposition did not occur as the laser beam moved across it. 2594
In the absence of an oxygen flow, dots consisting of PbO nanowires were also produced, Figure 5. The nanowire lengths ranged from tens of nanometers to several micrometers, indicating growth rates on the order of several hundred nanometers per minute. Because this morphology is different than what is reported in the few other studies of PbO deposition by CVD12,13 and in particular, a study of films produced by photolytic LACVD,14 it strongly suggests that these nanowires grew by the vapor-liquid-solid (VLS) mechanism, in which material from the vapor is dissolved in a liquid at the seed particle, and the solid condenses from the liquid and emerges as a long nanowire.15 This is possible here because gold and lead form a eutectic at 215 °C. In pure VLS growth, the wire diameter equals the seed particle diameter. Here the diameter is several times greater than the seed diameter, indicating some degree of radial growth as well. Nevertheless, the aspect ratio of the wires indicates that by far the majority of mass addition occurred at the seed particle. For the nanowires, there was no oxygen flow added during this experiment, and so the oxygen in the deposit must have come from that in the precursor molecule itself. A compariNano Lett., Vol. 6, No. 11, 2006
Figure 6. TiO2 polycrystalline dot deposited by PACVD. MicroRaman spectra show that it is the anatase form.
son of Raman spectra taken from these deposits and a commercially available PbO powder shows that the spectral features of the PbO nanowires are red-shifted and broadened.16 This is consistent with studies of Si and ZnO nanowires and is believed to result from phonon confinement and or stress effects.17,18 The diameters of many of dots are ∼10 µm (not shown), which is larger than the beam spot. Similar results are seen for the lines. This could be due to coupling of light/heat from the nanoparticles in the illuminated spot. Alternatively, the precursors may decompose on particle surfaces in the beam, and then reactive fragments may diffuse outward along the surface or through the gas, initiating growth nearby. A similar procedure was used to deposit titania, but the results are very different. In these experiments, each spot on the substrate illuminated by the laser grew polycrystalline TiO2. The growth rate was high, with some deposits up to 40 µm in height. Shown in Figure 6 is an SEM image of a typical polycrystalline deposit. Micro-Raman spectroscopy measurements identified the deposit to be the anatase form of TiO2. The eutectic temperature for Ti-Au is 1063 °C, and we would not expect VLS growth to occur in this system given the available laser power. However, deposition can be nucleated on the laser-heated nanoparticles at lower temperatures, T. The thermal conductivity of semiconductors and insulators scales as 1/T, and once a deposition nucleates, the heating will be localized in the deposit, which will continue to grow.19,20 The new growth is not catalyzed on the surface of the nanoparticle as in VLS growth; deposition occurs on the outermost surface as in conventional CVD. This effect has been well-demonstrated by Stuke et al. in LACVD fabrication of micrometer-scale structures composed of Al2O3 rods.1,2 The results for PbO and TiO2 clearly show that significant local heating occurred in the laser-illuminated regions and that this heating was sufficient to allow CVD growth. With the current experimental arrangement, it is difficult to measure the temperature of the gold particles or CVD Nano Lett., Vol. 6, No. 11, 2006
deposits directly, but estimates can be made by comparing to the Pb-Au binary phase diagram and the known stability limits of the precursors. The eutectic between Pb and Au occurs at a temperature of 215 °C, and so if in fact the PbO nanowires grew by the VLS mechanism, then the temperature must have reached at least this value. However, the lead precursor is generally stable to at least 250 °C, and does not fully decompose until 300 °C, suggesting that a temperature higher than the Pb/ Au eutectic temperature was reached. The stability of the Ti precursor is similar to that of Pb. There is a report of MOCVD growth of polycrystalline titania (non-rutile) films on sapphire as low as 400 °C.21 A few simple estimates show that in fact heating of this magnitude is to be expected. For a CW laser, the steadystate temperature is determined by the balance between the incident power density and the heat transfer from the nanoparticles to the substrate. Following the approach of Pustovalov,22 the temperature of a spherical particle due to a power density I0 at steady state can be shown to be T0 ) T ∞ +
I0Kabs r0 4k∞
(1)
where Kabs is the absorption efficiency factor for a particle of radius r0, which can be calculated from Mie scattering theory, and k∞ is the coefficient of thermal conductivity of the surrounding medium at the macroscopic equilibrium temperature T∞. This equation assumes that the particle is fully embedded in the surrounding medium and neglects radiative cooling effects. The above calculations, however, are based on classical Fourier diffusion theory. In solids, it is believed that as the particle size becomes comparable to the phonon mean-free path, Λ, of the surrounding medium the heat transfer is reduced compared to what would be predicted from Fourier diffusion because on this length scale the phonon transport is transitional between the ballistic and diffusive regimes.23 A qualitative comparison of T0 calculated from eq 1 for a 25-nm-diameter nanoparticle on an amorphous SiO2 substrate in the diffusion and quasi-diffusive transport regimes is shown in Figure 7. The dashed vertical line is the maximum power density used in this experiment, 1.2 × 1010 W/m2. The temperature is offset to the base sample cell temperature used in the experiments, 100 °C. The value of k∞ used for the diffusion regime calculation is 1.5 W/m‚K, which is the known bulk value for silica. In the quasi-diffusive transport regime we simply reduce k∞ by a factor of 10, (to 0.15 W/m‚ K) and in both cases, we set Kabs ) 1.5. A reduction in the heat transfer is expected because the average radius of the nanoparticles is comparable to literature values of Λ for amorphous SiO2 at 300 K (0.8-8) nm24,25 and further because the nanoparticles are resting on the surface and are not fully embedded in the substrate. These numbers are all approximate and affected by a number of additional factors but are presented here only to illustrate that temperature rises of the order required to explain our deposition results are quite likely. 2595
Figure 7. Comparison of T0 calculated from eq 1 for a 25 nm nanoparticle in the Fourier diffusion and quasi-diffusive transport regimes (lower and upper lines (bold and red, respectively). The vertical line is the maximum power density used in this experiment.
The laser heating of the substrate in the absence of the nanoparticles must also be considered. For the present experiments, we examine the case of a weakly absorbing substrate, that is, SiO2. The rise in the substrate temperature in the center of the beam, which is assumed to be Gaussian, in the absence of radiation losses can be shown to be T0c ≈ T∞ + 0.89
Iaw0 R* 2 C ln k∞ xπ R* 2
(
)
(2)
where w0 is the beam waist, R* ) Rabsw0, Rabs is the optical absorption coefficient of the substrate (1/cm), Ia ) (1 - R)I0, I0 is the incident power density (W/m2), R denotes the reflectivity, and C ) 0.577 is Euler’s constant.20 For the case of SiO2, the values of both R and Rabs are small, and T0c is negligible for the power densities of interest. Although we have demonstrated PACVD only for two oxides, the principles are general and should enable selective growth of virtually any material that can be deposited by a CVD process, including metals, semiconductors, and dielectrics of many types. For systems where contamination or doping by gold is an issue, other metals with a plasmon resonance could be substituted or the particle encased in a transparent inert matrix. The ability to limit deposition to micrometer-scale or even nanometer-scale areas without a mask and without etching may well lead to lower-cost fabrication technologies for advanced micro- and nanodevices. A requirement for the method is the prior fabrication of a template with micro/nanostructures that are efficient absorbers of electromagnetic energy in specific frequency ranges. There are a number of technologies that can be used for template creation, including electron-beam lithography, nanoimprinting, and block copolymer lithography, as used here. Because the plasmon resonance frequency depends on the size and shape of the resonant nanostructure, it may even be possible to construct a template with multiple resonance frequencies, allowing selective heating of some portions of 2596
the template by tuning the laser. This could be useful, for example, in fabricating devices composed of multiple materials. In conclusion, a new technique for selective CVD has been proposed and demonstrated for two different metal oxide systems. The technique makes use of the plasmon resonances in nanometer-scale structures as localized sources of heat. The three principal advantages of our approach over existing LACVD methods are (1) that the length scale of the deposition in the CVD reactor is determined by the dimensions of the template, (2) that the substrate can be kept at relatively low temperatures, and (3) that the power required is low. In addition, although we have only considered CW heating here, very rapid heating and cooling should be possible with this technique, allowing one to initiate and terminate chemistry on picosecond time scales. Much remains to be done to understand the extent of spatial control that can be achieved, the effect of field coupling between nanoparticles, the effects of nanoscale heat transfer, and how the resonant absorption is affected by the CVD deposit. All of these topics are subjects of current work and will be reported at a later date. Acknowledgment. This work has been supported by ARO through MURI grant number DAAD19-01-1-0517, by DARPA through ONR award number N00014-06-1-0454, and a gift from MainMan Ltd. We also thank George Rossman for the use of equipment and the Raman microscope and Elizabeth Miura Boyd for technical assistance. Supporting Information Available: Experimental method, PACVD reactor system, deposition experiments, and microRaman analysis. This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Lehmann, O; Stuke, M. Science 1995, 270, 1644-1646. (2) Wanke, M. C.; Lehmann, O.; Mu¨ller, K.; Wen, Q.; Stuke, M. Science 1997, 275, 1284-1286. (3) Kreibig, U. K.; Vollmer, M. Optical Properties of Metal Clusters; Springer-Verlag: New York, 1995. (4) Zayats, A. V.; Smolyaninov, I. I. J. Opt. A: Pure Appl. Opt. 2003, 5, S16-S50. (5) Maier, S. A.; Brongersma, M. L.; Kik, P. G.; Meltzer, S.; Requicha, A. G.; Atwater, H. A. AdV. Mater. 2001, 13, 1501-1505. (6) Mock, J. J.; Barbic, M.; Smith, D. R.; Schultz, D. A.; Schultz, S. J. Chem. Phys. 2002, 116, 6755-6759. (7) Zhao, L. L.; Kelly, K. L.; Schatz, G. C. J. Phys. Chem. B 2003, 107, 7343-7350. (8) Boyer, D.; Tamarat, P.; Maali, A.; Lounis, B.; Orrit, M. Science 2002, 297, 1160-1163. (9) Hauttmann, G.; Birngruber, R. IEEE J. Sel. Top. Quantum Electron. 1999, 5, 954-962. (10) Hirsch, L. R.; Stafford, R. J.; Bankson, J. A.; Sershen, S. R.; Rivera, B.; Price, R. E.; Hazle, J. D.; Halas, N. J.; West, J. L. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 13549-13554. (11) Jaramillo, T. F.; Baeck, S.-H.; Cuenya, B. R.; McFarland, E. W. J. Am. Chem. Soc. 2003, 125, 7148-7149. (12) Madsen, L. D. J. Am. Ceram. Soc. 1998, 81, 988-996. (13) Hedinger, R.; Kradolfer, T.; Hegetschweiler, K.; Wrle, M.; Dahmen, K.-H. Chem. Vap. Deposition 1999, 5, 29-35. (14) Watanabe, A.; Tsuchiya, T.; Imai, Y. Thin Solid Films 2002, 419, 76-81. (15) Hu, J.; Wang, T.; Lieber, C. M. Acc. Chem. Res. 1999, 32, 435445. (16) Raman spectra of PbO is provided in the Supporting Information.
Nano Lett., Vol. 6, No. 11, 2006
(17) Hofmann, S.; Ducati, C.; Neill, R. J.; Piscanec, S.; Ferrari, A. C.; Geng, J.; Dunin-Borkowski, R. E.; Robertson, J. J. Appl. Phys. 2003, 94, 6005-6011. (18) Geng, C.; Jiang, Y.; Yao, Y.; Meng, X.; Zapien, J. A.; Lee, C. S.; Lifshitz, Y.; Lee, S. T. AdV. Funct. Mater. 2004, 14, 589-594. (19) Ba¨urele, D. Chemical Processing with Lasers, volume 1 of Springer Series in Materials Science; Springer-Verlag: New York, 1986. (20) Ba¨urele, D. Laser Processing and Chemistry, 2nd ed.; SpringerVerlag: New York, 1996.
Nano Lett., Vol. 6, No. 11, 2006
(21) Chang, H. L. M.; You, H.; Gao, Y.; Guo, J.; Foster, C. M.; Chiarello, R. P.; Zhang, T. J.; Lam, D. J. J. Mater. Res. 1992, 7, 2495. (22) Pustovalov, V. K. Chem. Phys. 2004, 308, 103-108. (23) Chen, G. J. Nanopart. Res. 2000, 2, 199-204. (24) Kittel, C. Phys ReV 1949, 75, 972-974. (25) Allen, W. P., Jr.; Bromwell, D.; Doyle, T.; Devlin, L.; Snider, R.; Watson, P.; Dixon, G. Phys. ReV. B 1994, 49, 265.
NL062061M
2597