Strong enhancement in a photoacoustic signal due to plasmonic absorption in Au nanostructures was measured using piezotransistive GaN microcantilevers. A pulsed 790 nm laser focused on the Au metallization of the piezotransistor resulted in a much larger photoacoustic signal compared to the non-metallized areas. Upon deposition of a 5 nm Au layer, the photoacoustic signal increased significantly for both previously metallized and non-metallized areas, while 2 nm Ni deposition decreased the photoacoustic signal, confirming the role of Au nanostructures in facilitating plasmonic absorption. Infrared microscopy images covering the boundary of Au metallized and non-metallized surfaces indicated a much larger rise in temperature of the former region with laser exposure, explaining the generation of photoacoustic signals through plasmonic absorption.

In recent years, photoacoustic spectroscopy has emerged as an attractive technique with strong potential applications in chemical and biological sensing with high sensitivity and selectivity.1–3 The technique is based on the detection of ultrasonic waves generated in a medium due to absorption of intensity modulated light energy (typically from a laser), causing a periodic change in temperature. Similar to Fourier Transform Infrared Spectroscopy (FTIR) (but different from it due to its zero background nature), this technique utilizes different infrared (IR) absorption (causing proportionally different amplitudes of the acoustic signal) by the chemical and biological analytes at different wavelengths, resulting in unique signatures that can be utilized to perform specific detection. Although the acoustic signal generated has been traditionally detected using microphones, recent studies have demonstrated that microcantilevers can offer orders of magnitude higher detection sensitivity in resonant mode due to quality factor enhancement.4,5 In addition, with the emergence of highly sensitive piezoresistive and piezotransistive microcantilevers, cumbersome optical means of detecting nanoscale and picoscale deflection can be avoided and the photoacoustic detection system can be dramatically scaled down.

Although Si based piezoresistive microcantilevers and other MEMS structures have been extensively studied over the years, advances in III-Nitride MEMS technology in the past decade utilizing epitaxial layers grown on Si (111) substrates have enabled realization of inexpensive piezotransistive microcantilevers with ultra-high sensitivity that are orders of magnitude higher than their Si piezoresistive counterparts.6,7 Deflections in these piezotransistive microcantilevers are transduced by highly sensitive AlGaN/GaN heterojunction field effect transistors (HFET) integrated at their base, which under deflection induced strain undergoes changes in the 2-dimensional electron gas (2DEG) and electron mobility, offering much higher sensitivity and gauge factor (GF) compared to the Si piezoresistors where only mobility change occurs in response to strain.8 Due to this, piezotransistive microcantilevers have demonstrated a gauge factor (GF) of ∼9000, which is almost 2 orders of magnitude higher than the best Si piezoresistors, and have been able to detect femtometer level acoustic excitations. Such devices have also been used to successfully detect the presence of nanogram level explosives with high specificity.9 

In recent years, plasmonic enhancement of visible and IR light absorption, typically utilizing metal nanoparticles, has led to significantly higher performance in many electronic and optoelectronic devices including solar cells10,11 and photodetectors.12 The valence electrons in metals are excited by the electric field of the photons, leading to their collective oscillations at a particular frequency13 that is dependent on the size and shape of the nanoparticles or nanostructures.14 Au nanoparticles are widely used for plasmonic signal enhancement in various chemical and biological sensing since their plasmonic frequency falls in the visible and near-infrared (NIR) region.15,16 Utilizing the plasmonic effect to enhance optical absorption and hence the photoacoustic signal is an attractive way to enhance the sensitivity of photoacoustic sensors. However, this has remained unexplored to date, as few studies are devoted to photoacoustics involving solid media, which can take advantage of the plasmonic effect.

In this work, we report on the observation of strong enhancement in a photoacoustic signal due to plasmonic absorption in Au nanoparticles and consequent localized heating. The photoacoustic signal was found to increase dramatically when a pulsed NIR laser was focused in areas covered with Au nanostructures. In contrast, the photoacoustic signal reduced after deposition of Ni nanoparticles but recovered upon further deposition of Au nanoparticles, reconfirming the plasmonic enhancement effects of the latter. A significant increase in temperature of Au nanoparticle covered areas was confirmed through infrared microscopy images.

Piezotransistive microcantilevers were fabricated using III-nitride epitaxial layers grown on a Si (111) substrate (from DOWA Semiconductor Akita Co., Ltd.). The details of this fabrication process have been discussed by Qazi (2011) and Talukdar (2015). The HFET was fabricated [at the Institute for Electronics and Nanotechnology (IEN), Georgia Institute of Technology] at the base of the cantilever, exploiting the maximum strain occurring there due to deflection. Source/drain ohmic contacts were formed by Ti (20 nm)/Al (100 nm)/Ti (45 nm)/Au (55 nm) metal stack deposition and rapid thermal annealing thereafter. Schottky gate contact was formed by Ni (25 nm)/Au (375 nm) deposition following the source-drain contacts. The microcantilevers used in these experiments have variable widths but a fixed length of 250 μm, with the embedded HFET having a mesa dimension of 17 × 29 μm and a gate length of 5 μm. To characterize the microcantilevers, we used two kinds of excitation to oscillate them: one using a piezo chip actuator and the other using photoacoustic excitation generated using a pulsed LASER. The piezo chip actuator of dimension 5 × 5 × 2 mm3 was purchased from Physik Instrumente, while the adjustable focus pulse laser module (790 nm, 25 mW, focal spot ∼30 μm) was purchased from World Star Tech. For our experiments, the HFET devices at the base of the microcantilevers were wire bonded to a custom designed chip carrier from Oshpark. For the piezo based excitation, the piezo chip was kept in contact with the top surface of the chip carrier, very close to the silicon base of the device assembly. For photoacoustic based excitation, the laser module was mounted on an xyz positioner and the focused laser beam was pointed at a desired location on the device on and around the base of the microcantilever. To obtain the resonance curve of the cantilevers, a variable frequency 5 V rms AC signal was applied from a SR850 Lock-In amplifier either to the piezo chip assembly or to the transistor-transistor logic (TTL) module of the LASER, for piezo and photoacoustic excitations, respectively. A constant drain-source current (IDS) was maintained, and the variation in drain-source voltage (VDS) was plotted as a function of frequency. Since the drain-source resistance (RDS) is proportional to strain and hence to cantilever deflection or oscillation amplitude, the plot of VDS vs. frequency is representative of resonance characteristics.8,17

In our previous work, a photoacoustic signal was generated by directing a pulsed laser beam near the base of the microcantilever, where the pulsed light was absorbed by the Si substrate (underneath the GaN epitaxial layer) to generate photoacoustic waves, which were measured by taking advantage of the strong piezoelectric properties of the AlGaN/GaN heterostructure, which gives rise to a two dimensional electron gas (2DEG) very close to the sutrface.8,9,18 In this work, to investigate the effects of plasmonic absorption, we directed the red laser beam (790 nm, 25 mW), pulsed at the resonant frequency of the sensing cantilever, on GaN overhang (realized through intentional lateral over etch of the Si substrate in the Bosch process) near the base of the cantilever. Since the photoacoustic technique is “zero background” (i.e., acoustic waves are generated only when there is an absorption of photons and consequent temperature rises), focusing a sub-bandgap light (1.57 eV) on GaN (bandgap 3.42 eV at room temperature19) ensures no light absorption and no acoustic wave generation. Indeed, we observed that careful positioning of the laser beam only on the GaN overhang produces no significant photoacoustic signal. Interestingly, however, when the laser was directed toward the base of the cantilever (on the GaN overhang), where Au metallization lines (Ti (20 nm)/Al (100 nm)/Ti (45 nm)/Au (55 nm) metal stack for ohmic and Ni (25 nm)/Au (375 nm) for Gate Schottky contacts)9 for source, drain, and gate contacts were present, a very strong photoacoustic signal was observed. Since the pulsed red laser is not likely to penetrate the 200 nm thick Au layer on the ohmic (source, drain) or Schottky (gate) contacts or be absorbed in the GaN layer even if it does so, we conclude that the Au layer absorbs the photonic energy from the red laser through generation of plasmons, which leads to a periodic temperature rise and thermal expansion of the gold layer and the GaN layer underneath as the plasmons lose their energy. The acoustic waves thus generated are detected by the cantilever which produces the resonance curve as the frequency of the pulsed laser is varied.

To further verify the plasmonic absorption, we deposited 5 nm Au on the top surface near a microcantilever (using a shadow mask) and recorded its resonance characteristics before and after Au deposition. Figure 1(a) shows the comparison between pre-Au and post Au deposition resonance amplitudes (ΔVDS). A dramatic increase in ΔVDS by ∼48 times as a result of the Au nanolayer deposition is observed, which is attributable to the much higher plasmonic absorption post Au deposition. The figure also shows a leftward shift in resonance frequency by ΔfR = 988 Hz after the Au deposition due to mass loading of the cantilever. To verify that the Au/GaN Schottky contact is not playing any role in the photonic absorption (by the electrons near the Schottky contact) and consequent acoustic wave generation, we recorded the cantilever responses with the laser focused at the base of the cantilever from both top and bottom. The responses are shown in Fig. 1(b) which shows that both the resonance curves in Fig. 1(b) have comparable values of ΔVDS although the one obtained by laser focused at the bottom seems to be somewhat asymmetric. We are still investigating this aspect. Nonetheless, if Schottky barrier related absorption were important, we would see a much higher cantilever response with a laser directed from the bottom side, which would let much higher intensity of light reach the Schottky contact, since there is reflection/attenuation by the Au metallization layer. Therefore, we rule out any significant role of electronic absorption at the Schottky barrier in photoacoustic generation.

FIG. 1.

(a) Resonance curves of a cantilever (250 × 50 × 1 μm) showing a change in resonance amplitude (ΔVDS), before and after Au deposition, in both linear and log scales. The laser was focused at the base of the cantilever, directly on the HFET. For both these curves, VDS was maintained at 0.8 V with no gate voltage applied. (b) Resonance curves of the same device with laser illumination from top and bottom at approximately the same position partially overlapping the HFET metallization at the far end of the HFET (away from the Si substrate). The inset shows the effect of initial Ni metallization (2 nm) and subsequent Au (2 nm) deposition on the resonance amplitude. For these measurements, a constant biasing point (IDS = 100 μA and VDS = 1.06 V) was maintained for both pre and post Au deposition experiments by varying the gate voltage slightly.

FIG. 1.

(a) Resonance curves of a cantilever (250 × 50 × 1 μm) showing a change in resonance amplitude (ΔVDS), before and after Au deposition, in both linear and log scales. The laser was focused at the base of the cantilever, directly on the HFET. For both these curves, VDS was maintained at 0.8 V with no gate voltage applied. (b) Resonance curves of the same device with laser illumination from top and bottom at approximately the same position partially overlapping the HFET metallization at the far end of the HFET (away from the Si substrate). The inset shows the effect of initial Ni metallization (2 nm) and subsequent Au (2 nm) deposition on the resonance amplitude. For these measurements, a constant biasing point (IDS = 100 μA and VDS = 1.06 V) was maintained for both pre and post Au deposition experiments by varying the gate voltage slightly.

Close modal

To probe further, we coated the cantilever chip with 2 nm Ni and measured the photoacoustic response with the laser focused on the HFET metallization. We found that the photoacoustic signal intensity actually reduces significantly compared to the pre-exposure signal. It is likely that Ni formed a more continuous film unlike Au (indeed some devices were shortened by 2 nm Ni deposition) which tends to form isolated islands (see discussion below) upon deposition. Such a structural difference can result in much reduced plasmonic absorption in the deposited Ni compared to Au.20 In addition, it can simply cut down the intensity of light reaching the Au metallization layer underneath, causing a reduced photoacoustic signal. Further coating with 2 nm Au, as expected,21,22 enhances the photoacoustic signal and makes it larger than the signal prior to Ni coating. This further underlines the impact of specifically Au nanoparticles in enhancing the photoacoustic signal.

Figure 2(a) shows the SEM image of the top surface of the cantilever after 5 nm Au deposition. The inset shows a further magnified section, which clearly shows isolated islands of Au nanoparticles on the GaN surface (average size, ∼2 μm). Due to the discontinuous nature of the Au film, there is no significant change in the ID-VD and ID-VG characteristics (inset) as shown in Fig. 2(b). There is a slight increase in the saturation drain current and a reduction in knee voltage as seen from the ID-VD plots (more prominent at higher gate biases), in addition to an improvement in transconductance as seen from the ID-VG plots. These improvements in device characteristics can be attributed to a reduction in surface trapping as a result of Au surface deposition, which fixes the surface barrier height.

FIG. 2.

(a) SEM image of an Au nanoparticle coated cantilever. The scale bar is 300 μm. The dark-colored clusters are islands of the Au nanoparticle. The inset shows a magnified SEM image of a section showing the dimensions of the Au nanoparticles. (b) I–V characteristics of the AlGaN/GaN HFET transducer at the base of a microcantilever before and after Au deposition. The inset shows the IDS–VGS plot for the same device (with VDS = 2.5 V), before and after Au deposition. (c) Resonance curves of the same device before and after Au deposition, for both Piezo chip and pulsed laser excitation. For these measurements, a constant biasing point (IDS = 100 μA, VDS = 1.08 V) was maintained for both pre and post Au deposition experiments by varying the gate voltage slightly.

FIG. 2.

(a) SEM image of an Au nanoparticle coated cantilever. The scale bar is 300 μm. The dark-colored clusters are islands of the Au nanoparticle. The inset shows a magnified SEM image of a section showing the dimensions of the Au nanoparticles. (b) I–V characteristics of the AlGaN/GaN HFET transducer at the base of a microcantilever before and after Au deposition. The inset shows the IDS–VGS plot for the same device (with VDS = 2.5 V), before and after Au deposition. (c) Resonance curves of the same device before and after Au deposition, for both Piezo chip and pulsed laser excitation. For these measurements, a constant biasing point (IDS = 100 μA, VDS = 1.08 V) was maintained for both pre and post Au deposition experiments by varying the gate voltage slightly.

Close modal

Figure 2(c) shows a comparison of the resonance curves obtained with laser and piezo chip based excitations on the same device, before and after Au deposition. Even when ΔVDS is observed to increase by ∼50 times for the laser excitation, ΔVDS resulting from piezo-excitation did not change much in magnitude. The magnitude of ΔVDS from piezo chip based excitation is typically significantly more than that from the photoacoustic excitation as seen from our previous work.18,23 In this work, the same holds true for the data recorded from before Au deposition on the cantilever [from Fig. 2(c)]. However, after Au deposition, the photoacoustic response was found to have increased many times while the piezo based response remained the same. The enhancement in the photoacoustic response is thus solely attributed to plasmonic absorption by the Au nanoparticles, ruling out any contribution from the variation in the cantilever mechanical properties due to the Au deposition.

The plasmonic absorption by Au nanoparticles was verified for several other microcantilevers for multiple locations around their base. Figure 3(a) shows an optical image of a microcantilever (250 × 70 × 1 μm) with spatial locations marked where plasmonic enhancements were recorded. Figure 3(b) shows the resonance characteristics of the microcantilever for some of the representative locations after 5 nm Au deposition. The pre-deposition amplitude for location 1 is also shown for comparison. Although the magnitudes vary widely, the peak position for the responses remains very similar after deposition for different locations and red-shifted compared to that measured before deposition due to the mass loading effect. The peak amplitudes before and after Au deposition are plotted for all the locations in Fig. 3(c). We find very large amplification in almost all locations varying from 1 to 7 and 9 to 14. Interestingly, for some locations, i.e., those on the GaN overhang [i.e., locations 4–7 and 14–16 in Fig. 3(a)] where there was no response prior to Au deposition, there was a small yet definite response recorded after Au deposition. Even for the locations [10–13 in Fig. 3(a)] where the GaN layer is over the silicon substrate, laterally farther away from the cantilever base, finite responses were recorded after deposition as opposed to no response from before Au deposition. The amplification factor corresponding to location 1 was found to be ∼18, while the highest amplification factor of ∼350 was recorded at location 6 [Fig. 3(c)].

FIG. 3.

(a) Optical image of a microcantilever (250 × 70 × 1 μm) with a map of the laser illumination points. Position 1 is directly on HFET. Positions 2–7 are on GaN overhang. Positions 8–13 are on GaN on the silicon substrate, and positions 14–15 are on the channel with only GaN layer (Si substrate etched away). (b) Resonance curves corresponding to selected positions after Au deposition compared with the response on HFET prior to deposition. (c) Resonance amplitudes (ΔVDS) before and after Au deposition plotted on a log scale for laser focus positions 1–16. For the measurements, a constant bias current of IDS = 100 μA and a voltage of VDS = 1.33 V were maintained.

FIG. 3.

(a) Optical image of a microcantilever (250 × 70 × 1 μm) with a map of the laser illumination points. Position 1 is directly on HFET. Positions 2–7 are on GaN overhang. Positions 8–13 are on GaN on the silicon substrate, and positions 14–15 are on the channel with only GaN layer (Si substrate etched away). (b) Resonance curves corresponding to selected positions after Au deposition compared with the response on HFET prior to deposition. (c) Resonance amplitudes (ΔVDS) before and after Au deposition plotted on a log scale for laser focus positions 1–16. For the measurements, a constant bias current of IDS = 100 μA and a voltage of VDS = 1.33 V were maintained.

Close modal

To verify that temperature indeed rises due to plasmonic absorption in the Au nanostructures, we took a bare AlGaN/GaN sample on the Si substrate and coated part of it with a 5 nm thick Au layer using a shadow mask. The same 790 nm laser, as used in the above experiments, was then used to illuminate across the junction of the metal covered and bare surfaces, and the steady state temperature profile was imaged using an infrared (IR) microscope (Thermacam SC100). The results are shown in Fig. 4(a). We find that the steady state temperature of the Au coated side is significantly higher than the uncoated side, which is expected due to the plasmonic absorption and the consequent thermal dissipation. An IR video (not presented here) of the region indicates that the steady state temperature profile is established over several tens of seconds. The temperature information from the image in Fig. 4(a) was extracted, and the temperature profile averaged over 6 line scans along the length of the AlGaN sample is shown in Fig. 4(b). The x-axis represents the data points on the lines scanned along the length of the sample, while the y-axis shows 6-fold averaged temperature at a particular point on the line. The temperature plot shows that the side coated with Au is about ∼6.3 °C hotter than the uncoated side. Although more accurate estimation of the temperature change needs careful calibration, nonetheless, the temperature rise clearly highlights the effect of plasmonic absorption.

FIG. 4.

(a) An IR image (728 × 368) of a partially Au coated (5 nm) AlGaN sample across the coating boundary, under laser (790 nm, 25 mW) illumination. The approximate boundary is shown by a dotted line, and the laser spot (defocused to expose a larger area) is shown by a dotted ellipse. The temperature scale bar shows the range of temperature variations. (b) Temperature profile along the length of the sample averaged over six lines [#170 to 175 marked in (a)] shows a variation of temperature by ∼6.3 °C across the boundary. The inset shows a schematic band diagram illustrating the plasmonic excitation of hot electrons facilitated by the Au nanostructure and subsequent generation of acoustic waves due to rising temperature because of hot carrier relaxation.

FIG. 4.

(a) An IR image (728 × 368) of a partially Au coated (5 nm) AlGaN sample across the coating boundary, under laser (790 nm, 25 mW) illumination. The approximate boundary is shown by a dotted line, and the laser spot (defocused to expose a larger area) is shown by a dotted ellipse. The temperature scale bar shows the range of temperature variations. (b) Temperature profile along the length of the sample averaged over six lines [#170 to 175 marked in (a)] shows a variation of temperature by ∼6.3 °C across the boundary. The inset shows a schematic band diagram illustrating the plasmonic excitation of hot electrons facilitated by the Au nanostructure and subsequent generation of acoustic waves due to rising temperature because of hot carrier relaxation.

Close modal

The inset of Fig. 4(b) shows the band diagram at the Au nanostructure/GaN interface, explaining the proposed plasmonic generation of photoacoustic waves. The pulsed 790 nm light incident on the Au/nanostructures excites plasmons, pushing electrons to a higher energy level (hot electrons).24 These electrons subsequently relax back to their original energy state by losing energy that gets converted into thermal energy, causing a temperature rise and hence local expansion of the material creating the acoustic waves. We note again here the difference with regular photonic absorption at a Schottky barrier; since the amplification is only seen for Au nanostructures and not Ni, although they both produce similar Schottky barrier heights with GaN, it is clear that plasmonic effects and not the regular absorption at a Schottky junction is the dominant cause for the photoacoustic amplification.

In conclusion, we have demonstrated strong enhancement in photoacoustic signals measured using piezotransistive GaN microcantilevers arising out of plasmonic absorption in Au metallization and nanostructured films. While nanoscale Au deposition facilitated signal enhancement by 350 fold, deposition of Ni resulted in no amplification but reduction in the photoacoustic signal, underlining the unique plasmonic absorption by Au nanostructures. IR imaging across the boundary on a partially Au coated sample clearly established that the temperature rise is related to plasmonic absorption in the Au nanostructures that resulted in photoacoustic signal enhancement.

Financial support for this work from the National Science Foundation (Grants Nos. ECCS 1512342, IIP-1602006, and CBET-1606882) is thankfully acknowledged.

1.
H.
Morkoc
,
S.
Strite
,
G. B.
Gao
,
M. E.
Lin
,
B.
Sverdlov
, and
M.
Burns
,
J. Appl. Phys.
76
,
1363
(
1994
).
2.
M. R.
Werner
and
W. R.
Fahrner
,
IEEE Trans. Ind. Electron.
48
,
249
(
2001
).
3.
T.
Hussain
,
M.
Micovic
,
T.
Tsen
,
M.
Delaney
,
D.
Chow
,
A.
Schmitz
,
P.
Hashimoto
,
D.
Wong
,
J. S.
Moon
, and
M.
Hu
,
Electron. Lett.
39
,
1708
(
2003
).
4.
M.
Stutzmann
,
G.
Steinhoff
,
M.
Eickhoff
,
O.
Ambacher
,
C. E.
Nebel
,
J.
Schalwig
,
R.
Neuberger
, and
G.
Müller
,
Diamond Relat. Mater.
11
,
886
(
2002
).
5.
M.
Razeghi
and
A.
Rogalski
,
J. Appl. Phys.
79
,
7433
(
1996
).
6.
B.
Jogai
,
J. Appl. Phys.
93
,
1631
(
2003
).
7.
K.
Naeli
and
O.
Brand
,
J. Appl. Phys.
105
,
014908
(
2009
).
8.
M.
Qazi
,
N.
DeRoller
,
A.
Talukdar
, and
G.
Koley
,
Appl. Phys. Lett.
99
,
193508
(
2011
).
9.
A.
Talukdar
,
M.
Faheem Khan
,
D.
Lee
,
S.
Kim
,
T.
Thundat
, and
G.
Koley
,
Nat. Commun.
6
,
1
(
2015
).
10.
S.
Pillai
,
K. R.
Catchpole
,
T.
Trupke
, and
M. A.
Green
,
J. Appl. Phys.
101
,
093105
(
2007
).
11.
K.
Nakayama
,
K.
Tanabe
, and
H. A.
Atwater
,
Appl. Phys. Lett.
93
,
121904
(
2008
).
12.
J.
Fang
,
D.
Wang
,
C.
DeVault
,
T.-F.
Chung
,
Y. P.
Chen
,
A.
Boltasseva
,
V. M.
Shalaev
, and
A. V.
Kildishev
,
Nano Letters.
17
,
57
(
2017
).
13.
J. M.
Pitarke
,
V. M.
Silkin
,
E. V.
Chulkov
, and
P. M.
Echenique
,
Rep. Prog. Phys.
70
,
1
(
2007
).
14.
S.
Link
and
M. A.
El-Sayed
,
J. Phys. Chem. B
103
,
4212
(
1999
).
15.
D. M.
Schaadt
,
B.
Feng
, and
E. T.
Yu
,
Appl. Phys. Lett.
86
,
063106
(
2005
).
16.
K.
Saha
,
S. S.
Agasti
,
C.
Kim
,
X.
Li
, and
V. M.
Rotello
,
Chem. Rev.
112
,
2739
(
2012
).
17.
O.
Ambacher
,
J.
Smart
,
J. R.
Shealy
,
N. G.
Weimann
,
K.
Chu
,
M.
Murphy
,
W. J.
Schaff
,
L. F.
Eastman
,
R.
Dimitrov
, and
L.
Wittmer
,
J. Appl. Phys.
85
,
3222
(
1999
).
18.
A.
Talukdar
,
M.
Qazi
, and
G.
Koley
,
Appl. Phys. Lett.
101
,
252102
(
2012
).
19.
O.
Madelung
,
Semiconductors: Group IV Elements and III-V Compounds
(
Springer Science & Business Media
,
2012
).
20.
B.
Abel
,
T. S.
Kabir
,
B.
Odukoya
,
M.
Mohammed
, and
K.
Aslan
,
Anal. Methods
7
,
1175
(
2015
).
21.
T.
Ung
,
L. M.
Liz-Marzán
, and
P.
Mulvaney
,
J. Phys. Chem. B
105
,
3441
(
2001
).
22.
S.
Link
,
M. B.
Mohamed
, and
M. A.
El-Sayed
,
J. Phys. Chem. B
103
,
3073
(
1999
).
23.
A.
Talukdar
and
G.
Koley
,
IEEE Electron Device Lett.
35
,
1299
(
2014
).
24.
R.
Jia
,
D.
Zhao
,
N.
Gao
, and
D.
Liu
,
Sci. Rep.
7
,
40483
(
2017
).