NANO LETTERS
Polarized Surface-Enhanced Raman Spectroscopy from Molecules Adsorbed in Nano-Gaps Produced by Electromigration in Silver Nanowires
2009 Vol. 9, No. 2 672-676
Jeong Min Baik, Seung Joon Lee, and Martin Moskovits* Department of Chemistry and Biochemistry, UniVersity of California, Santa Barbara, California 93106 Received October 16, 2008; Revised Manuscript Received December 10, 2008
ABSTRACT Uniform nanogaps were produced in silver nanowires previously fabricated by AC electrodeposition in porous anodic alumina templates and released by dissolving the oxide matrix. Configuring the nanowires with electrodes and passing current through them created nanogaps by electromigration. Uniform nanogaps approximately perpendicular to the axes of the nanowires were produced with widths in the range ∼17 to ∼40 nm. Gaps as small as 5 nm but with less uniform geometries could also be produced in this manner. A heuristic model, in which electromigrative and diffusional effects are included, is used to relate the size of the nanogap produced to the local resistance of the region in the nanowire where the gap ultimately forms. Intense surface-enhanced Raman spectra were observed from Rhodamine 6G adsorbed on the nanowire in the gap. For gaps that uniformly divided the nanowire across its width, the surface enhanced Raman spectroscopy intensity was maximum when the electric vector of the exciting light was oriented across the gap, as predicted by electromagnetic enhancement calculations.
The development of reliable and reproducible fabrication techniques for making nanogaps in nanowires is an actively pursued field primarily because of their applicability to molecular electronics in which rather small molecules must bridge metallic junctions.1-3 The ability to measure the electrical and transport characteristics through one or more molecules bridging a nanogap while simultaneously identifying the molecules and perhaps determining their conformations by, for example, using surface enhanced Raman spectroscopy (SERS), would certainly advance both fields. A small nanogap in a SERS-active material such as silver or gold would be an almost ideal structure toward such a goal since we now know that the space between two closely located nanostructures is a SERS “hot spot” when illuminated with light of the appropriate wavelength and polarization.4-6 Nanogaps have been successfully fabricated using a number of techniques including forming mechanical break junctions7 and templated electroplating.8 Electromigration is a particularly simple method that yields fairly consistent nanoscale gaps when one starts with a nano-object such as a nanowire,9 or a nanostructure produced by optical or e-beam lithograph.10,11 Electromigration, the electric-fieldassist migration of metal ions counterpropagating against the * To whom correspondence should be addressed. E-mail: mmoskovits@ ltsc.ucsb.edu. 10.1021/nl803145d CCC: $40.75 Published on Web 01/21/2009
2009 American Chemical Society
flow of electron in a metal structure under bias, is a phenomenon studied for decades in the context of the degradation of contacts in microelectronics.12 For systems like nanowires in which the surface-to-volume ratio is large, surface diffusion, that is, a process that is not field-assisted but purely thermal in which material diffuses out of hightemperature regions of the nanowire and deposited at lowertemperature locations, might also play a significant role in thinning out high-resistance points in a nanowire eventually creating a nanogap.13 A number of studies have been carried out in an attempt to effect and control electromigration in nanowires, including one that reports preliminary simultaneous Raman and molecular conduction measurments.14 However, there still remains much to study regarding the electromigration process, especially in very small structures.1-3 In this article, we report the formation of nanogaps by electromigration (and possibly also by surface diffusion) in silver nanowires fabricated in porous anodic aluminum oxide, followed by polarized SERS measurements on Rhodamine 6G (R6G) molecules infused into the gap. Highly ordered porous aluminum oxide (PAO) templates were prepared using high purity aluminum (Alfa Aesar, Al foil, 2.5 mm/0.10 in. thick, Plutaronic 99.997% metal basis, 100 mm × 500 mm) that was annealed at 500 °C and 5 × 10-5 Torr for 4 h and electropolished in CH3CH2OH/HClO4 (0.8:0.2 v/v) at ca. 0 °C and 45 V DC for 45 s. PAO was
Figure 1. (a) Representative current-voltage (I-V) curve carried out at a rate of 0.012 V/sec under vacuum (5 × 10-6 Torr) at room temperature. (Inset) SEM images of Ag nanowires grown in porous anodic alumina with the alumina matrix etched to release the nanowires. (b) SEM image of single nanowire spanning gold contacts separated by about 15 µm. (Inset) Higher magnification image of a ∼19 nm nanogap.
produced using two-step anodization.15 The Al was first anodized in 0.3 M oxalic acid at 15 °C and 40 V DC for 4 h, the alumina was removed in aqueous H3PO4/H2CrO4 (6:1.8 w/w) at 65 °C by etching for 2 h, then rinsed in water and reanodized for 8 h resulting ∼40 µm long pores. The remaining aluminum was removed in an HCl and CuCl2 solution and a Ti/Au (30/800nm) electrode was thermally evaporated onto one side of the PAO template to provide a conductive contact. Ag electroplating was then performed in a three-electrode electrochemical cell. In the cell, the PAO template, an Ag/AgCl electrode, and a Pt plate were used as the working electrode, the reference electrode, and the counter electrode, respectively. The Ag nanowire arrays were prepared in commercial Ag plating solution (Alfa Aesar, 28.7 g/L (Ag)) without additives at a deposition potential of -0.8 V, a value determined to be optimal by cyclic voltammetry prior to deposition.16 After deposition, the samples were rinsed several times in water and dried in air at room temperature. All the chemicals otherwise specified were reagent grade, and Millipore-filtered water of resistivity greater than 18.0 MΩ · cm was used in making aqueous solutions. Nanowires were released by dissolving the alumina template in 1 M NaOH, rinsed several times in water (see the inset of Figure 1a), and suspended in pure ethanol by brief sonication (1 ∼ 2 s). To make electrical contacts to individual nanowires, an aqueous suspension of nanowires was dried on an oxidized silicon substrate, after which Ti and Au was evaporated in sequence through a shadow mask to make the electrodes. Figure 1a shows a representative I-V curve carried out at room temperature at a rate of 0.01V/sec under vacuum (5 × 10-6 Torr). At low values of the applied voltage, the current increased linearly with increasing voltage with a slope corresponding to a conductance value of 7.5 × 10-3 Ω-1. That is, the nanowire behaved as an ohmic resistor (130 Ω) with conductance corresponding approximately to that of metallic Ag. The conductance value decreased slowly to 2.8 × 10-3 Ω-1 as the voltage increased. Above ∼0.3 V, the I-V shape of the conductance curve became steeper indicating an even lower conductance. Eventually a voltage Nano Lett., Vol. 9, No. 2, 2009
is reached at which the current suddenly drops to zero. The details of the changes observed in the current-voltage behavior of the nanowire with increasing voltage are more easily appreciated in a plot of the conductance (di/dV) versus bias voltage (Figure 1a). At voltage values below ∼0.2 V the conductance is almost constant. The small decrease in conductance with increasing voltage observed in this region we believe is due to a small degree of resistive heating of the nanowire. This conclusion is supported by the fact that if the voltage ramp is halted before ∼0.2 V is reached and the voltage restored to 0 V, the initial conductance value is regained after a few minutes. Above ∼0.3 V, the conductance drops approximately linearly with increasing voltage. This, we believe, signals the onset of electromigration, since the change in conductance is not reversed if the voltage ramp is halted before gap formation occurs and the voltage restored to 0 and the measurements repeated. At approximately 0.58 V (for the particular nanowire illustrated), the conductance drops suddenly and irreversibly to 0 indicating the formation of a nanogap. Figure 1b shows a scanning electron microscopy (SEM) image of a nanowire connecting gold contacts separated by ∼15 µm after the gap-forming voltage was reached. The ensuing ∼19 nm gap is shown at higher magnification in the inset to Figure 1b. The above behavior can be semiquantitatively understood using a simple model based on existing theories of electromigration. We will assume that the system consists of a single, high-resistance nanojunction where the temperature increases to a higher value than elsewhere in the nanowire. Electromigration therefore occurs faster at that point which eventually becomes the locus of the nanogap. This will likely be a region of the nanowire of a somewhat smaller local diameter. This nanojunction will be modeled as a resistor (Rj) whose resistance increases as the electromigration process proceeds. The nanojunction is in series with a single series resistor (Ro), assumed constant in time and independent of bias voltage, that represents all other resistances in the circuit including the resistance due to the rest of the nanowire plus all of the contact resistances plus any resistors placed deliberately in the circuit to control the current. We assume that the series resistor will not change in value as the nanogap forms. (This may not be strictly true for the kinds of contact resistances that are formed between a nanowire and macroscopic contact pads, but those effects will be ignored in this heuristic analysis.) The voltage drop across the junction, Vj, is therefore given by Vj ) VRj ⁄ (Ro + Rj)
(1)
We will assume that the nanojunction temperature will be determined by resistive heating which will be primarily offset by conductive losses to the rest of the nanowire and to the underlying substrate. Assuming the voltage ramp rate to be slow enough for thermal equilibrium to be instantaneously achieved and assuming that the conductive heat losses can be written in the form K(T - To), where T is the temperature at the nanojunction and To that of the ambient, then writing the power balance equation and rearranging the terms to solve explicitly for T, one obtains 673
T ) To + Vj2 ⁄ (KRj) ) To +
V 2Rj K(Ro + Rj)2
(2)
Combining eq 2 with the expression for the total current through the circuit, I ) V/(Ro + Rj) gives the expression T ) To + (iV - i2Ro) ⁄ K
(2′)
The junction resistance, Rj, will vary with time and the ensuing increasing current through the junction on account of electromigration, as well as due to surface diffusion. The mass transport due to both of these processes has been considered in numerous publications.12,17 We will use the semiempirical expression17 that includes electromigrative and diffusional effects except that we will not include a threshold current, since in principle surface diffusion can occur at all temperatures, but being an activated process, its rate will be imperceptibly small at low temperatures. Hence the mass transport rate out of the junction region will be assumed17 to be of the form rate ∝
dRj Ai exp(-Ea ⁄ kBT) ) T dt
(3)
In eq 3, A is a collection of constants including the preexponential factor of the diffusion constant, i is the current through the junction, and Ea is the activation energy to atom migration. Since the resistance at the junction where the gap will ultimately form will be proportional to the crosssectional area of the nanowire at that point, the time rate of increase in Rj will be proportional to the rate at which material electromigrates out of the junction area. We will assume that the constant, A, is chosen so as to transform the proportionality in 3, into eq. 3. Using the relationship V ) i(Ro + Rj) one obtains the following expression relating the time rates of change of V, i, and Rj dRj dV di ) (Ro + Rj) + i dt dt dt dRj di dV (R + Rj) + i ) dV dt o dt
(4)
The current-voltage measurements were carried out by ramping the voltage, V, at a constant rate tthat we define as dV/dt t b. Substituting this identity into eq 3, recognizing that di/dV defines the conductance, G, and substituting for dRj/dt using eq 3 one obtains the following semiempirical expression for the conductance: G(V) )
(
)
i Ai2 -Ea⁄kBT e 1V bT
(5)
In which the temperature, T, is given by eq 2′. Equation 5 is semiempirical since one needs to substitute measured values for V and i to define the temperature. The result is a value for G that does not rely on taking a derivative of the measured i(V) values and therefore contains new mechanistic information, despite the fact that measured values if i(V) are used to define T. We show below how this approach can yield valuable predictions and mechanistic insight. Equation 5 was fit to the measured conductance values shown in Figure 1a by adjusting the two parameters, A and K. The value for Ea is assumed to be 0.1 eV as in ref 17 and Ro is taken to be the measured (low voltage) resistance of the nanowire (130 Ω). The results (Figure 2) show excellent 674
Figure 2. Calculated conductance (G(V) ) di/dV) curves as a function of bias voltage for two values of the voltage ramp rate, b, (0.01 or 0.005 V/s), calculated using eq 5 and compared with the measured values of the conductance using a ramp rate of 0.01 V/s (Parameters: K ) 0.000004 W/Kelvin; Ea ) 0.1 eV; A ) 3 600 000 Kelvin s-1 Ω ampere-1; Ro ) 130 Ω).
Figure 3. (a) Slope of the calculated conductance (dG/dV ) d2i/ dV2) for various values of the series resistance Ro, calculated in the linear G vs V region between 0.4 and 0.5 V, that is, in the voltage range where electromigration appears to begin to be significant just prior to the formation of the nanogap. (b) Measured d2i/dV2 values as a function of the measured width of the resultant nanogap. (Inset) SEM images of two nanogaps with 17 and 42 nm widths.
concordance between the measured values and those recalculated using eq 5. Figure 2 also shows that once a good fit is achieved for a given set of experimental values one can successfully predict the behavior for other values of the system’s parameters. For example, the expected conductance when the voltage ramp rate is reduced from 0.01 to 0.005 V/s is shown in Figure 2. Since the values of i(V) also change when the ramp rate is altered one needs to calculate a new function i(V). This was done by beginning with a linear (i.e., strictly ohmic) trial current corresponding to resistance Ro with which a trial conductance function is calculated and numerically integrated to yield an improved estimate of i(V). Then the process is repeated to self-consistency. This approach was also used to predict the effect on the conductance in the electromigration region for various initial values of Ro. The results are shown in Figure 3a. The slope of G(V) in the range 0.4 to 0.5 V range (just prior to gap formation) was numerically calculated from these curves and plotted as a function of the assumed values of Ro. This curve is reminiscent of Figure 3b, which shows a similar functional relationship between the slope of G(V) just before the gap Nano Lett., Vol. 9, No. 2, 2009
Figure 4. (a) Polarization-angle dependence of the SERS spectra obtained from R6G adsorbed onto a nanogap. (b) SERS intensity of the 1647, 1360, and 771 cm-1 lines of R6G as a function of polarization angle.
forms and the ultimate size of the gap. This implies that the initial value of the resistance of the nanowire at its most resistive point will determine the size of the gap (all other parameters being equal). That is, the smaller the initial nanowire resistance (i.e., the larger the initial nanowire diameter) at its “weakest spot”, the smaller the eventual gap size will be. SERS experiments were carried out on a number of nanogaps produced in the manner described above. One of the goals in the above nanogap-making program was to create gaps with relatively flat walls that transect the nanowire perpendicularly so as to be able to define the orientation of the polarization of light with respect to the metal surfaces in the gap region. Sample were incubated overnight in a 1 × 10-5 M aqueous R6G solution, washed with water, and dried prior to the Raman measurements. SERS spectra were excited with 514.5 nm Ar-ion laser light that was focused to a spot size ∼1 µm. A wavelength of 514.5 nm is an appropriate choice of wavelength for gaps ∼20 nm. (For very small gaps, excitation with a wavelength further toward the red would produce more intense SERS signals.) Representative SERS spectra are shown in Figure 4a as a function of the polarization direction of the illuminating and collected light. Figure 4b shows the polarization-angle dependence of the SERS intensities of the bands at 1647, 1360, and 771 cm-1. The observed strong and medium-strong bands at 1647, 1570, 1502, 1360, and 1180 cm-1 are assigned to the totally symmetric modes of in-plane C-C stretching vibrations. Their frequencies agree well with literature values although, not unexpectedly, their relative intensities in the SERS spectrum varies from those reported for ordinary Raman.18 Polarization-dependent SERS spectra were measured by rotating a half-wave plate in 30° increments, changing the angle (θ) between the electric vector of the incident light and the long axis of the nanowire. The (baseline-corrected) SERS intensities of the 1647, 1360, and 771 cm-1 lines of R6G are maximum when the polarization is parallel to the nanowire’s long axis (θ ) 0°) (i.e., when the light is polarized across the nanogap) and falls to a minimum value ∼3 times smaller when the polarization is perpendicular to the nanowire’s axis (θ ) 90°). Nano Lett., Vol. 9, No. 2, 2009
The dependence of the SERS intensity in polarization and its magnitude are as expected on the basis of near-field electromagnetic field enhancement calculations, which have primarily been carried out from two-particle systems.6,19,20 Such systems can show significantly larger enhancements when the gap size is small. For gaps ∼1 nm, the polarization contrast can be 4-5 orders of magnitude for light polarized across the gap and along the gap. However, with gaps ∼20 nm the observed intensity contrast on illuminating the gap with light polarized along and across the gap is approximately what is predicted. In the absence of a gap the SERS intensity would become maximum when the nanowire is illuminated with light polarized across the nanowire’s long axis, as was previously observed.21 Not all of the nanogaps we produced were as uniform and as perpendicular to the nanowire’s axis as the ones reported in this study. Occasionally, gaps with significant concavities and high curvature were obtained. It was gratifying to note that the polarization dependence of the SERS signals obtained with these always accorded with what one qualitatively expect for such structures as deduced from their SEM images, recorded subsequent to the SERS measurements. Likewise, the effect of the radius of curvature of the nanowire in the nanogap region (again not unexpectedly) was found to affect the SERS intensity and its polarization dependence. Occasionally gaps were formed which consisted in a rather sharp tip pointing to a concave surface on the other surface of the gap. Such structures produce more intense SERS signals and polarization effects consistent with their geometries. Summarizing, SERS measurements were carried out on R6G adsorbed in nanogaps produced in single Ag nanowire by electromigration. Uniform gaps that were approximately perpendicular to the long axes of the nanowires could be produced with gap widths in the range ∼17 to ∼40 nm. Gaps as small as 5 nm but with less uniform geometries could be produced in this manner. For gaps that divide the nanowire uniformly across its width, the SERS intensity was maximum when the electric vector was oriented parallel to the long axis of the nanowire (i.e., across the gap) as predicted by electromagnetic enhancement calculations. Acknowledgment. This work was supported by the Institute for Collaborative Biotechnologies through Grant DAAD19-03-D- 0004 from the U.S. Army Research Office and made extensive use of the MRL Central Facilities at UCSB by the National Science Foundation under Award nos. DMR-0080034 and DMR-0216466 for the HRTEM/STEM microscopy. References (1) Lambert, M. F.; Goffman, M. F.; Bourgoin1, J. P.; Hesto, P. Nanotechnology 2003, 14, 772–777. (2) Taychatanapat, T.; Bolotin, K. I.; Kuemmeth, F.; Ralph, D. C. Nano Lett. 2007, 7, 652–656. (3) Hadeed, F. O.; Durkan, C. Appl. Phys. Lett. 2007, 91, 123120. (4) Moskovits, M. ReV. Mod. Phys. 1985, 57, 783–826. (5) Kneipp, K.; Wang, Y.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Phys. ReV. Lett. 1996, 76, 2444–2447. (6) Nie, S.; Emory, S. R. Science 1997, 275, 1102–1106. (7) Park, H.; Lim, A. K. L.; Alivisatos, A. P. Appl. Phys. Lett. 1999, 75, 301–303. 675
(8) Reed, M. A.; Zhou, C.; Muller, C. J.; Burgin, T. P.; Tour, J. M. Science 1997, 278, 252–254. (9) Kervennic, Y. V.; Van der Zant, H. S. J.; Morpurgo, A. F.; Gurevich, L.; Kouwenhoven, L. P. Appl. Phys. Lett. 2002, 80, 321–323. (10) Fischbein, M. D.; Drndic´, M. Appl. Phys. Lett. 2006, 88, 063116. (11) Strachan, D. R.; Smith, D. E.; Johnston, D. E.; Park, T. H.; Therien, M. J.; Bonnell, D. A.; Johnson, A. T. Appl. Phys. Lett. 2005, 86, 043109. (12) Ho, P. S.; Kwok, T. Rep. Prog. Phys. 1989, 52, 301–348. (13) Douketis, C.; Haslett, T. L.; Wang, Z.; Moskovits, M.; Lannotta, S. J. Chem. Phys. 2000, 113, 11315–11323. (14) Ward, D. R.; Grady, N. K.; Levin, C. S.; Halas, N. J.; Wu, Y.; Nordlander, P.; Natelson, D. Nano Lett. 2007, 7, 1396–1400. (15) Masuda, H.; Fukuda, K. Science 1995, 268, 1466–1468.
676
(16) Gardner, C. E.; Ghanem, M. A.; Wilson, J. W.; Smith, D. C. Anal. Chem. 2006, 78, 951–955. (17) Trouwborst, M. L.; van der Molen, S. J.; van Wees, B. J. J. Appl. Phys. 2006, 99, 114316. (18) Jeong, D. H.; Zhang, Y. X.; Moskovits, M. J. Phys. Chem. B 2004, 108, 12724–12728. (19) Moskovits, M. J. Raman Spectrosc. 2005, 36, 485–496. (20) Sawai, Y.; Takimoto, B.; Nabika, H.; Ajito, K.; Murakoshi, K. J. Am. Chem. Soc. 2007, 129, 1658–1662. (21) Mohanty, P.; Yoon, I.; Kang, T.; Seo, K.; Varadwaj, K. S. K.; Choi, W.; Park, Q.-H.; Ahn, J. P.; Suh, Y. D.; Ihee, H.; Kim, B. J. Am. Chem. Soc. 2007, 129, 9576–9577.
NL803145D
Nano Lett., Vol. 9, No. 2, 2009
