Appl. Phys. A 90, 219–223 (2008)
Applied Physics A
DOI: 10.1007/s00339-007-4321-1
Materials Science & Processing
n. misra1 y. pan2 n.w. cheung3 c.p. grigoropoulos1,u
Selective processing of semiconductor nanowires by polarized visible radiation 1
Department of Mechanical Engineering, University of California, Berkeley, CA 94720-1740, USA Nanosys, Inc., 2625 Hanover Street, Palo Alto, CA 94304, USA 3 Department of Electrical Engineering, University of California, Berkeley, CA 94720-1770, USA 2
Received: 26 June 2007/Accepted: 21 September 2007 Published online: 8 November 2007 • © Springer-Verlag 2007 ABSTRACT Semiconductor nanowires have attracted intense in-
terest due to potential applications in electronics, sensors and photonics. Introduction of dopants and their subsequent activation are essential for exploiting the electronic properties of semiconductor materials. In this work, we demonstrate pulsed laser annealing of silicon nanowires by visible radiation to be an efficient way for activating incorporated dopants and repairing implantation damage in a process that is compatible with sensitive flexible substrates. In situ electrical monitoring was used to study the laser annealing process. The absorption of laser light in SiNWs was shown to be strongly dependent on the light polarization and nanowire diameter based on finite difference time domain simulations. PACS 42.62.-b;
1
64.70.Nd; 78.67.n; 81.16.c
Introduction
The performance of semiconductor nanowires as active components for both nano-electronics devices [1] and large area macroelectronics [2] applications has been extensively investigated. Apart from the obvious advantage of scaling, the use of nanowires can have several additional benefits. For instance, recent reports indicate that the performance of nanowire devices fabricated on flexible substrates is not significantly altered when repeatedly bent [3]. Utilizing these attributes of silicon nanowires (SiNWs), one can employ thin film transistors (TFTs) based on nanowires [4] for flexible electronics instead of silicon (amorphous/polycrystalline) or organic semiconductors. Besides displays, such nanowire devices on lightweight and cheap flexible plastic substrates can have a plethora of applications in fields such as wearable electronics, mobile computing, information storage and biodetection. Other material systems based on nanowires, such as polymer-nanowire composites, may also have novel applications in a variety of fields. The challenge in implementing nanowires with polymers and other such temperaturesensitive materials is to be able to make the assembly and processing steps compatible with our choice of materials. Thermal processing of semiconductor nanowires [5, 6] has been u Fax: +1-510-642-6163, E-mail:
[email protected]
shown to be a highly beneficial step in the successful fabrication of functional devices based on nanowires. We recently demonstrated the processing of nanowires by UV radiation to be possible at lower fluence levels than bulk silicon [7]. We further investigate the laser processing of nanowires in the visible regime which promises easier integration with transparent plastic substrates. Additionally, the polarization anisotropy in the absorption of nanowires at wavelengths in the visible range is leveraged to give us an orientation selective processing scheme. As an example, electrical activation of implanted shallow dopants in SiNWs is investigated. In situ electrical monitoring is used to understand the effect of fluence levels and number of pulses on the processing of the SiNWs. The absorption of laser light in SiNWs is shown to be strongly dependent on the light polarization and nanowire diameter based on finite difference time domain (FDTD) simulations to corroborate our experiments. 2
Experimental
2.1
Nanowire devices
The SiNWs used in this study were grown via the vapour-liquid-solid (VLS) method, using gold colloids as catalysts. The wires were then thermally oxidized to form a ∼ 10 – 20 nm conformal oxide shell around the silicon core. The mean outside diameter of the core-shell structures was 60 – 70 nm. Quartz wafers were then coated with solutions of the SiNWs such that the wires were deposited on the wafer aligned along a predominant direction to within ±15◦ . The SiNW contact regions with metal electrodes were implanted and activated with rapid thermal annealing at 950 ◦ C for 10 s. This was done to ensure good contacts for electrical measurements of the silicon nanowires. A BF+ 2 beamline implant at a dose of 2 × 1015 cm−2 at 15 keV was used. Ti(80 nm)/Al(150 nm) contact electrodes 10 – 15 µm apart were then sputter-deposited after oxide removal and the remaining portion of the SiNWs were implanted with BF+ 2 to give boron-doped wires. The devices fabricated typically had about 1–5 nanowires with contacts to both electrodes. 2.2
Experimental setup
Nd:YAG laser annealing was used to electrically activate the dopants and repair the damage caused due to implantation in the SiNWs. The second harmonic output
220
Applied Physics A – Materials Science & Processing
of 532 nm from an Nd:YAG laser (New-wave Polaris) with a pulse width of 3 – 5 ns was used. Great care was taken to ensure the SiNWs were subject to uniform laser fluence. A micro-lens array homogenizer was used to achieve a topflat beam profile with a non-uniformity of less than 10% in the top-flat region. The generated flat-top size can be varied from a few microns to several millimeters by the appropriate choice of focal length of the lens used after the homogenizer assembly. A top-flat size of ∼ 7.5 mm was used for all the experiments. Multiple pulses were run at a repetition rate of 1 Hz in all experiments. The Nd:YAG laser beam used for processing is horizontally polarized as it exits the second harmonic generator. Consequently, by orienting our devices, we could control the angle between the incident polarization and the alignment direction of the nanowires. In situ two-point probe electrical measurements of the resistance of fabricated nanowire devices were used to monitor the YAG laser processing. All electrical measurements were performed with a semiconductor parameter analyzer (Hewlett Packard 4155A). 2.3
In situ electrical monitoring
To study the effect of fluence and the number of pulses on the laser processing of the nanowires, we specifically fabricated nanowire device structures with electrical contacts to nanowires that enabled us to monitor the nanowires. We have considered the activation of implanted dopants in the nanowires to demonstrate controllable selective thermal processing of the nanowires by visible radiation. The devices consisted of ion implanted nanowire regions to be processed by the laser, with thermally annealed low resistance contact regions to facilitate observation of the electrical behavior of the nanowires during laser processing. A scanning electron microscope (SEM) image of a fabricated SiNW device is shown in Fig. 1. The nanowire device structures were irra-
diated with laser pulses and the resistance of the nanowire devices was recorded after the desired number of pulses. The devices prepared by the process described above typically had resistance levels of the order of GΩ prior to laser annealing. The resistance of the devices after annealing with a specified number of laser pulses normalized to the initial resistance is plotted as a function of the number of pulses at different fluence levels. In situ electrical monitoring tests were run for two different polarization cases – electric field parallel to the predominant alignment direction of the SiNWs and electric field perpendicular to the predominant direction of alignment of the SiNWs (hereafter referred to as cases A and B, respectively). It is to be noted that our electrical data are based on a two-point probe measurement wherein contact resistance between the SiNWs and the metal electrodes can be substantially high to be neglected. To minimize the contact resistance for our purposes, a two-step process was employed to make contacts to the SiNWs and during the in situ experiments we only annealed the exposed regions of the SiNWs by the laser. The contact regions of the wire that have been implanted and thermally annealed to obtain ohmic contact should not be affected by the laser annealing process, since the metal contact acts as a reflecting photomask. Therefore, observed changes in resistance can be ascribed primarily to the effects of laser annealing in the sections of the nanowires subjected to laser annealing. 2.4
To further investigate the nano-structural evolution in the laser annealing process, separate structural characterization experiments were performed on the SiNWs. In these experiments, aligned SiNWs on a quartz substrate were subject to different fluence levels and pulse numbers. Transmission electron microscopy (TEM) was used to study the structure of these wires. Representative results from these studies are shown in Fig. 3. To gain better insight into the role of structural changes within the wires and corroborate the electrical data, we present the results of these experiments concurrently with the results of electrical monitoring experiments. 3
FIGURE 1 Scanning electron microscope image of a fabricated device showing a silicon nanowire between two metal electrodes
Structural characterization
Numerical
To estimate the absorption of radiation by the SiNWs during the laser processing, the FDTD technique was employed to compute the field distributions within the coreshell nanowires. For this purpose, the near-field simulation software “TEMPEST 6.0” developed at the Department of Electrical Engineering and Computer Science at UC Berkeley was employed. Simulated intensity distributions in the nanowires for the cases of different polarization of incident light are shown in Fig. 2. A total computational volume of 1000 × 1000 × 1 nm3 with a core-shell nanowire of outer diameter 60 nm was used for the cases shown in Fig. 2b. A grid size of 1 × 1 × 1 nm3 was implemented in the simulations. The silicon core radius was assumed to be 40 nm, surrounded by a 10 nm thick silica sheath. The nanowire was modeled as lying on top of a quartz substrate. The nanowire core and shell structure was constructed with cylindrical nodal elements of different radii. Light with a wavelength of 532 nm and
MISRA et al.
Selective processing of semiconductor nanowires by polarized visible radiation
221
FIGURE 3 Transmission electron microscope images of (a) laser processed nanowires with polarization along the nanowire axis, and (c) laser processed nanowires with polarization perpendicular to the nanowire axis for the indicated fluence and number of pulses. Higher fluence levels in cases (b) and (d) show evidence of melting and recrystallization within the oxide shell
FIGURE 2 Finite difference time domain simulation results for the interaction of visible radiation at λ = 532 nm with core-shell nanowires on a quartz substrate. (a) Predictions of absorption efficiency of the silicon nanowires as a function of the silicon core diameter. (b) Intensity distributions within the nanowires for a core-shell nanowire with a silicon core diameter of 40 nm. An oxide shell of 10 nm was assumed for all diameters in (a). The scale bars in (b) are 20 nm
specified polarization was incident normally on the nanowiresubstrate assembly. Periodic boundary conditions were employed on the sides of the simulation domain to simulate any effects of other wires in a parallel array configuration. At the top and bottom, 100 nm thick perfectly matched layers (PML) were used to limit the boundaries. Using the computed intensity distributions within the nanowire from the FDTD simulations, we were able to calculate the absorption cross-section of the nanowires, by integrating over the volume enclosed within the nanowire, according to the relation [8] σabs = kεr |E|2 dV , (1) V
where, k is the imaginary part of the refractive index, εr is the dielectric constant, |E| is the magnitude of the electric field predicted by FDTD simulations and V is the volume of the body under consideration. To verify the predictions of absorption using the FDTD simulations we compared the results to the analytical solution of Maxwell’s equations for infinitely long circular cylinders of different diameters. For the sake of comparison, the domain was assumed to contain a single suspended cylindrical silicon nanowire, bearing no thermally grown oxide. Good agreement was observed between the analytical and FDTD pre-
dictions of absorption. Interesting observations can be made from the computational results. The size of the SiNWs leads to different absorption in the nanowires. An absorption peak is apparent at core diameters of around 40 nm in the case of polarization parallel to the nanowires. A previous study has attributed this effect to resonance-enhanced absorption in silicon nanowires at visible wavelengths [9]. It has also been reported that enhanced photothermal effect can lead to strong absorption in thin SiNWs [10]. The cylindrical geometry of the nanowires gives rise to stronger absorption at smaller diameters for polarization parallel to the nanowires, which is absent in the case when polarization is perpendicular to the nanowires. Based on our computational results, it should be noted that a tight size distribution of SiNWs would be ideal for Nd:YAG laser annealing. The size dependence of absorption can be minimized by using different wavelengths such as UV. 4
Results and discussion
In case A, at laser fluence levels lower than 16 mJ/cm2 we observed improvement in the conductance when subjected to multiple pulse irradiation. On irradiating the SiNW devices at a higher fluence level of 21 mJ/cm2 , conductance increases of the order of ∼ 104 times that of non-annealed nanowires were observed. We believe that the substantial increase in conductance at larger fluences can be attributed to higher temperatures induced in the wires as compared to lower fluences. During the thermal annealing of bulk silicon, low temperature processing has been shown to help in the removal of crystalline defects [11, 12] whereas, higher temperatures are needed for activating higher fractions of implanted dopants (which have activation energies of the order of a few eV [13]). As a consequence, higher fluences are
222
Applied Physics A – Materials Science & Processing
needed to provide sufficient energy for the activation of the implanted dopants in the SiNWs. On further increasing the fluence to 28 mJ/cm2 , we observed the conductance initially increased, but subsequently dropped to lower values, eventually resulting in loss of signal. This indicates that damage threshold of the wires had been reached which was further confirmed by the examination of structure of the nanowires after pulsing. Ion implantation was seen to lead to the introduction of defects within the SiNWs that were grown as single crystalline wires. It was observed that the structure of the SiNWs in case A at 15 mJ/cm2 and 21 mJ/cm2 was crystalline. However, at 30 mJ/cm2 the wires showed damage. The damage was observed to be in the form of void formation in the silicon core, melting and recrystallization within the oxide shell and beading up of the wires. Such structural changes leading to the damage of the wires are the cause for the observed sharp decrease in the conductance of the wires near or above the melting threshold. The nucleation of defects in ion-implanted bulk silicon during non-melt laser annealing has been studied previously [14] and found to be dependent on the number of pulses and laser fluence levels. Multiple pulses at higher incident energy levels are more likely to cause defect formation in the nanowires and eventually lead to deterioration of the crystalline structure, which explains why the initial decrease in resistance is followed by an increase upon further pulsing. It is evident from the electrical monitoring data for case B, shown in Fig. 4b, that absorption is much stronger when polarization is along the axis of the nanowires rather than when polarization is perpendicular to the nanowires. At a fluence of ∼ 44 mJ/cm2 , which was much greater than the melting threshold of the SiNWs in case A, we do not see any signs of damage to the wires. As we go higher in fluence, we observe faster increases in conductance. At ∼ 71 mJ/cm2 , the conductance of the nanowires increases by ∼ 104 –105 . I –V curves acquired during the laser annealing process are shown in Fig. 5. Further increase in the fluence levels, eventually leads to melting and damage. TEM images of SiNWs annealed in case B are shown in Fig. 3c and d. The SiNWs showed damage at around ∼ 100 mJ/cm2 . Compared to case A, much larger fluences are required in case B to anneal or damage the SiNWs. In both cases, the nanowire damage threshold fluence was much lower than the reported values of melting threshold for silicon thin films and bulk that are typically hundreds of mJ/cm2 . This is attributed to the strong absorption efficiency of the nanowires, as well as small heat losses to the ambient and the thermal conductance to the substrate. The latter is responsible for prolonging the cooling cycle, thereby allowing sufficient residence of the nanowires at elevated temperatures and facilitating an efficient annealing process. A rough estimate of the temperature rise in the nanowires may be obtained using the lumped capacitance model for the thermal behaviour of the irradiated SiNWs [9, 10] that neglects the transverse thermal gradients through the nanowire cross section. Assuming bulk thermal properties, at a fluence level of 30 mJ/cm2 for a 30 nm wire with an absorption efficiency of 0.1, a temperature rise of about 790 K is predicted if all the absorbed energy were to heat up the nanowire. Such an estimate would predict a linear dependence on the fluence and a strong dependence on the diameter of the wires as ∼ σabs /r . Hence,
FIGURE 4 In situ electrical monitoring of the silicon nanowires during the Nd:YAG laser annealing process showing the variation of normalized resistance with multiple pulses and indicated fluence levels for (a) polarization along the nanowires, and (b) polarization normal to the nanowires. Data for two different sets of devices is shown for each experimental fluence level
diameters which exhibit absorption peaks would be heated to higher temperatures. In particular, the temperature rise in smaller wires would be accentuated due to the inverse proportionality to the diameter. However, such a simplified calculation should be used as qualitative guidance with caution since it does not take into account more intricate effects such as thermal losses from the wires, size and temperature dependent thermal properties of the wires as well as free carrier absorption and temperature dependent optical properties [15]. The exhibited polarization anisotropy in the absorption can have several benefits in the processing of SiNWs. The process window for compatibility with various substrates is broadened. The melting/damage to the wires occurs at different fluence levels depending upon the polarization. This ‘dual threshold’ of the SiNWs enables selective melting and possibly ablation of the SiNWs. Wires in unwanted direc-
MISRA et al.
Selective processing of semiconductor nanowires by polarized visible radiation
I–V curves of a nanowire device with a single nanowire during the Nd:YAG laser annealing process with polarization normal to the nanowires at a fluence level of 71 mJ/cm2 for the indicated number of pulses
FIGURE 5
tions could be removed with such a technique, while improved alignment and tighter size distribution of the wires should be possible. Different wires and dopants can be activated by control of the polarization and wire directions. In devices using crossed-nanowire layout, the process flow has to be more diligently designed, but the above described effects can be used advantageously to process the SiNWs. The selective nature of such processing also can be adapted to implement layer-by-layer nanowire based devices [16]. Alignment of the nanowires and/or the use of intermediate layers between layers of nanowires can be used for sequential processing which is not possible by conventional thermal techniques. Furthermore, this technique may be extended to induce temperature based reactions in the nanowires such as, silicidation [17] and controlled phase change [18]. If this effect of polarization is to be eliminated, i.e., the SiNWs need to be annealed irrespective of the orientation, laser annealing can be done below the lower melting threshold with circularly polarized light for larger number of pulses to ensure all the SiNWs are annealed to optimum levels. We expect that the YAG laser processing is compatible with plastic substrates. Although all the hereby reported experiments were performed on a quartz substrate, we have tested the damage threshold of PET and PEN substrates to evaluate the compatibility of the laser processing with flexible plastic substrates. Our experiments have revealed that the damage threshold of these substrates is much higher than the fluence levels required to anneal the SiNWs at 532 nm. Fabrication of SiNW devices on plastic substrates with laser processing is currently underway. 5
Conclusions
Laser processing of SiNWs has been carried out with a 532 nm Nd:YAG laser. In situ electrical monitoring and structural characterization of the SiNWs was used to demonstrate laser annealing of nanowires to be an efficient way
223
of activating implanted dopants as well as repairing the implantation damage in SiNWs. At optimum annealing conditions, increases in the conductance of the SiNWs on the order of ∼ 104 –105 were observed. Laser processing of nanowires was strongly polarization dependent due to the polarization anisotropy in the absorption of SiNWs. The absorption was also found to be size dependent. Furthermore, the Nd:YAG laser processing is a versatile tool for the selective processing of nanowires as there can be dual processing levels corresponding to device layout and polarization. Based on the structural characterization as well as the in situ electrical monitoring data, we believe that the melting threshold of SiNWs with diameters of outer diameters ∼ 60 nm for a wavelength of 532 nm is around ∼ 21 mJ/cm2 . When the polarization of the incident light is perpendicular to the wire axis the melting threshold is much higher ∼ 100 mJ/cm2 . The Nd:YAG laser processing technique developed in this work, displays the potential to be employed with flexible substrates such as plastics due to the low fluence levels needed to anneal the SiNWs. Substrates with much lower damage thresholds can also be employed by using polarization along the SiNWs. This processing technique can be a powerful tool in the fabrication of flexible displays, wearable electronics, large area chemical and biological sensors. Since the processing is thermal in nature, it may be adapted to other applications as well. ACKNOWLEDGEMENTS This was supported by DARPA under contract W31P4Q-05–CR193 through Nanosys, Inc. (PI Yaoling Pan). The authors would like to thank Dave Stumbo, Jeffrey Miller and David Zaziski at Nanosys, Inc. for their support. The authors are grateful to Erin Becker for the TEM studies.
REFERENCES 1 L.J. Lauhon, M.S. Gudiksen, D. Wang, C.M. Lieber, Nature 420, 57 (2002) 2 M.A. Alam, N. Pimparkar, S. Kumar, J. Murthy, MRS Bull. 31, 466 (2006) 3 M.C. McAlpine, R.S. Friedman, S. Jin, K.H. Lin, W.U. Wang, C.M. Lieber, Nano Lett. 3, 1531 (2003) 4 X. Duan, C. Niu, V. Sahi, J. Chen, J.W. Parce, S. Empedocles, J.L. Goldman, Nature 425, 274 (2003) 5 J. Niu, J. Sha, Y. Wang, X. Ma, D. Yang, Microelectronic Eng. 66, 65 (2003) 6 S.-W. Chung, J.-Y. Yu, J.R. Heath, Appl. Phys. Lett. 76, 2068 (2000) 7 N. Misra, L. Xu, Y. Pan, N. Cheung, C.P. Grigoropoulos, Appl. Phys. Lett. 90, 111 111 (2007) 8 A. Ishimaru, Wave Propagation and Scattering in Random Media (Academic Press, New York, 1978) 9 G.-H. Ding, C.T. Chan, Z.Q. Zhang, P. Sheng, Phys. Rev. B 71, 205 302 (2005) 10 N. Wang, B.D. Yao, Y.F. Chan, X.Y. Zhang, Nano Lett. 3, 475 (2003) 11 S. Baek, T. Jang, H. Hwang, Appl. Phys. Lett. 80, 2272 (2002) 12 K.K. Ong, K.L. Pey, P.S. Lee, A.T.S. Wee, X.C. Wang, C.H. Tung, L.J. Tang, Y.F. Chong, Appl. Phys. Lett. 89, 122 113 (2006) 13 A. Mokhberi, P.B. Griffin, J.D. Plummer, E. Paton, S. McCoy, K. Elliott, IEEE Trans. Electron. Dev. ED-49, 1183 (2002) 14 S. Earles, M. Law, R. Brindos, K. Jones, S. Talwar, S. Corcoran, IEEE Trans. Electron. Dev. ED-49, 1118 (2002) 15 N.G. Nilsson, K.G. Svantesson, J. Phys. D Appl. Phys. 13, 39 (1980) 16 A. Javey, S. Nam, R.S. Friendman, H. Yan, C.M. Lieber, Nano Lett. 7, 773 (2007) 17 Y. Wu, J. Xiang, C. Yang, W. Lu, C.M. Lieber, Nature 430, 61 (2004) 18 D. Yu, J. Wu, Q. Gu, H. Park, J. Am. Chem. Soc. 128, 8148 (2006)
