Letter pubs.acs.org/NanoLett
Tensile Strained Germanium Nanowires Measured by Photocurrent Spectroscopy and X‑ray Microdiffraction Kevin Guilloy,*,†,‡ Nicolas Pauc,†,‡ Alban Gassenq,†,‡ Pascal Gentile,†,‡ Samuel Tardif,§,∥ François Rieutord,§,∥ and Vincent Calvo†,‡ †
Université Grenoble Alpes, INAC-SP2M, SINAPS, F-38000 Grenoble, France CEA, INAC-SP2M, SINAPS, F-38000 Grenoble, France § Université Grenoble Alpes, INAC-SP2M, NRS, F-38000 Grenoble, France ∥ CEA, INAC-SP2M, NRS, F-38000 Grenoble, France ‡
S Supporting Information *
ABSTRACT: Applying tensile strain in a single germanium crystal is a very promising way to tune its bandstructure and turn it into a direct band gap semiconductor. In this work, we stress vapor− liquid−solid grown germanium nanowires along their [111] axis thanks to the strain tranfer from a silicon nitride thin film by a microfabrication process. We measure the Γ-LH direct band gap transition by photocurrent spectrometry and quantify associated strain by X-ray Laue microdiffraction on beamline BM32 at the European Synchrotron Radiation Facility. Nanowires exhibit up to 1.48% strain and an absorption threshold down to 0.73 eV, which is in good agreement with theoretical computations for the Γ-LH transition, showing that the nanowire geometry is an efficient way of applying tensile uniaxial stress along the [111] axis of a germanium crystal. KEYWORDS: Germanium, nanowire, tensile strain, photocurrent, XRD, Laue
W
substrates13 has also been published, although requiring a custom sample holder to maintain the strain state in the material. However, while most attempts use germanium grown on (100) substrates and most theoretical works have been done for biaxial (100) and uniaxial [100] directions,14−16 the physical properties of uniaxially stressed germanium along [111] is still controversial. Indeed, while deformation potentials from the model-solid theory tend to dismiss the possibility of a fundamental direct bandgap for Ge stressed in this crystallographic direction,17 other works indicate the existence of a direct bandgap regime above a certain strain threshold, ranging from 1.0518 to 4.2%.19 Nanowires are an advantageous solution to apply very high level of uniaxial strains because more than 10% strain has already been described, showing that they had no defect compromising their mechanical properties.20 Biaxial stressing through a sputtered silicon nitride shell on ⟨111⟩ nanowires has been reported21 as well as uniaxial tensile stress using the flexion of the silicon substrate with a dynamic micromechanical 3-point stress module.22 In this work, we study vapor−liquid−solid (VLS) grown nanowires, which provide a potential way of growing germanium nanostructures along a [111] direction.23 We
ith the rapid increase of the need for bandwidth and computational capabilities, the integration of photonics devices on electronic chips is currently highly investigated. Passive devices and modulators are readily available but the lack of on-chip light source for silicon photonics is among the main obstacles for the realization of such devices. Besides the integration of III−V semiconductors on silicon, germanium is being intensely explored as a light source material due to its complementary metal oxide semiconductor (CMOS) compatibility. Germanium is known to be a poor light emitter due to the indirect nature of its bandgap. As its direct bandgap is only 0.14 eV above its fundamental gap, both optically and electrically pumped pulsed laser operation have been reported in strained (0.25% (100) biaxial) doped germanium waveguides. This was made possible since the application of a tensile strain reduces the difference between the direct and the fundamental bandgaps, and strong n-type doping inhibits the thermalization process toward the indirect fundamental bandgap.1,2 The next challenge is therefore the application of a higher strain level in order to reduce the required amount of doping to achieve laser operation. Different straining strategies have been demonstrated. Microstructuring of germanium epilayers on silicon have shown strain accumulation in tiny bridges of the thermal residual strain appearing at the cooling step of the crystal growth,3−6 as well as the use of external stressors, principally silicon nitride,7−11 which is known to be a convenient stressor for other materials such as silicon.12 The use of flexible organic © 2015 American Chemical Society
Received: December 16, 2014 Revised: February 18, 2015 Published: March 11, 2015 2429
DOI: 10.1021/nl5048219 Nano Lett. 2015, 15, 2429−2433
Letter
Nano Letters describe a microfabrication process taking advantage of the stress transfer from a silicon nitride thin film to germanium nanowires. We further investigate the influence of [111] uniaxial strain on the band structure of germanium nanowires by photocurrent spectroscopy and X-ray diffraction analysis. These two techniques correlates the optical absorption threshold to a direct strain measurement at the nanoscale level. Germanium nanowires were grown in a low-pressure chemical vapor deposition (LPCVD) chamber at 370 °C from GeH4 as a precursor. HCl was introduced during the growth in order to inhibit side facet growth and reduce tapering.24 A commercial colloidal solution was employed to dropcast 250 nm gold microspheres as a catalyst for the vapor− liquid−solid growth. B2H6 and PH3 were used as precursors for the p- and n-type dopants, their fluxes were alternatively switched on and off during the growth in order to obtain axial p-i-n junctions. Resulting nanowires, oriented along a [111] axis, were 20 μm long and 280 nm in diameter. The as-grown sample was dipped in a 10% diluted H2O2 solution for 20 s in order to remove the n-type doped surface layer, which causes a short-circuit of the p-i-n structure. We measured the diameter of the nanowires before and after etching by scanning electron microscopy (SEM). As-grown nanowires had a significant tapering in the p-type region (Figure 1a), which is reduced after etching (Figure 1b). Etched germanium nanowires have a diameter of 240 nm at the tip below the gold catalyst and a diameter of 150 nm in the central intrinsic region, while the base of the p-type and tapered region has a diameter of 320 nm (see Figure 1c for a schematic diagram). The reversed tapered geometry in the n-type region may be explained by a differential etch rate for different doping levels. In both situations, before and after etching, nanowires had smooth surfaces with rugosities below the 10 nm imaging resolution of the SEM (Figure 1f,g). The electrical properties of these nanowires were investigated after sonicating the growth sample in an isopropanol solution and dropcasting them onto an insulating thermal silicon oxide on a silicon sample. Metal contacts were patterned by deep UV lithography and 40 nm of chromium and 110 nm of gold were deposited in an electron beam evaporator. After the lift-off of the metal layer, the current−voltage characteristics of single nanowires were measured using a Keithley 4200 analyzer. The Figure 1d presents the current−voltage characteristic of an as-grown and an H2O2 surface etched nanowire. The asgrown wire exhibits a resistive behavior with a typical measured resistance of 30 kΩ. Etching drastically modifies the I−V characteristic because it switched from an ohmic to a strong rectifying behavior with 5 orders of magnitude change between forward and reverse biais at 1 V (Figure 1.e). A linear fit in the logarithmic-linear I−V relation for the forward biased nanowire below 0.35 V gives a saturation current of 96 mA/cm2 and an ideality factor of 1.29. The H2O2 etching step is therefore required to obtain longitudinal built-in fields to achieve photocarrier separation. The straining of the nanowires was performed by a clean room microfabrication process. Silicon nitride was employed as a stressor material in order to apply uniaxial tensile stress along the nanowire axis. For this purpose, we realized a microdevice comprising an electrically connected germanium wire anchored onto two Si3N4 pulling arms. First, a 20 nm layer of HfO2 was deposited by thermal atomic layer deposition (ALD) onto the growth sample (Figure 2a). Then, the nanowires were
Figure 1. (a,b) SEM images of the nanowire growth sample, before and after etching. (c) Schematic diagram of the geometry of the etched nanowires. (d,e) Current−voltage characteristics of germanium nanowires before etching (in dashed blue) and after etching (in solid orange) on a linear (d) and logarithmic (e) scale. (f,g) SEM images of the tip of an as grown nanowire and an etched nanowire, exhibiting smooth surfaces before and after etching.
dropcasted onto a stressed LPCVD silicon nitride layer on a silicon substrate (Figure 2b). Bow measurement on the bare Si3N4/Si substrate carried out by profilometry revealed a 1.3 ± 0.1 GPa biaxial tensile stress in the nitride layer, showing the high stress state of the silicon nitride thin film. An 110 nm layer of PECVD silicon oxide was then deposited on top of the nanowires in view of anchoring them on the silicon nitride thin film (Figure 2c). Metal contacts were obtained by the lift-off of a chromium and gold layer after the etching of the silicon oxide and HfO2 layers by Ar ion beam etching (Figure 2d). Then, the shape of two arms were patterned on each side of the nanowire by optical lithography. The silicon oxide and silicon nitride were further etched by inductively coupled plasma reactive ion etching (ICP RIE) with CF4 and O2 plasma, the HfO2 acting as a mask to prevent germanium etching (Figure 2e). The width of the arms was set to 5 μm for processing convenience and the 2430
DOI: 10.1021/nl5048219 Nano Lett. 2015, 15, 2429−2433
Letter
Nano Letters
length of the silicon nitride arm Larm, and the properties of the nitride layer ϵwire =
ΔLgap Lgap
=−
2ΔLarm σ L = −2 sin arm Lgap Esin Lgap
It is then possible to apply different amounts of elongation for a given substrate by choosing different arm lengths (Larm in Figure 2g). The electronic band structures of the stressed nanowire were explored by photocurrent spectroscopy using a homemade setup with a Xe arc lamp white light source and a grating monochromator. The light beam was focused onto the nanowire using an optical microscope with a Cassegrain objective also used for sample alignment, the spot size being significantly larger than the junction in the nanowire (about 20 μm). The light was polarized in the axis of the nanowire (which we will call TM polarization) in order to maximize absorption. The electrical contacts on the nanowires were connected through tungsten probes to a lock-in amplifier triggered by an optical chopper. The resulting spectrum in the 1200−1800 nm range was normalized by the spectrum of the source obtained by a reference extended InGaAs photodiode. The latter spectrum was finally normalized by the theoretical response of a core−shell Ge/HfO2 nanowire in air, with a diameter equals to the one in the intrinsic region, in order to eliminate the effect of Mie optical resonances.25−27 Because the absorption is proportional to the extinction coefficient in the small absorption regime,28 the absorption edge was measured by fitting the squared measured spectrum by a linear regression, plotted in Figure 3 (see Supporting Informations for details). We performed photocurrent spectroscopy measurements on unstressed germanium nanowires (Larm = 0 μm in Figure 3) and observed an absorption spectrum with a threshold which follows a relation α = K(E − Eg)1/2, that is, the usual relation between absorption and photon energy for the direct bandgap of a semiconductor.29 In this case, we measured a value of 1560 ± 20 nm (or 0.80 ± 0.01 eV), which is in agreement with literature data.30 Photocurrent spectra of germanium nanowires stressed by our microfabrication process exhibit a red shift as compared to unstressed germanium, as shown in Figure 3. We further notice that an increase in the silicon nitride arm length causes an increase in the redshift of the absorption edge, as the nanowires have an absorption threshold of 1610 ± 32 nm (respectively 1633 ± 25 and 1700 ± 20 nm) for an arm length of 3 μm (respectively 6 and 15 μm). Because the nanowires exhibit a strong rectifying behavior, the measured photocurrent is collected through a junction in the semiconductor material. The possibility of the carrier collection being obtained through Schottky junction can be ruled out because its depletion region would have been located close to the metal contacts in the unsuspended section and therefore unstrained areas. Any photocurrent generated by the metal−semiconductor junctions would have caused a parasitic contribution with an absorption edge at the unstrained level, which we did not observe. The central p-i-n junction is therefore the junction probed by our experiment. While photocurrent spectroscopy probes the influence of the stress on the electronic direct bandgap of germanium nanowires, the stress and strain components can be directly measured using X-ray Laue diffraction. Such measurements were performed on the X-ray microdiffraction setup at
Figure 2. (a−f) Illustration of the microfabrication process used to stress the germanium nanowires, displayed in green. The stressed silicon nitride layer is displayed in purple, the silicon oxide anchoring layer in pink, the metal contacts in yellow, and the silicon substrate in gray. (g) Scanning electron microscopy image of a stressed nanowire.
gap between the two arms was 6 μm, so that the 3 μm long intrinsic zone was entirely stretched. Finally, the silicon substrate was isotropically etched by SF6/ Ar RIE, releasing the suspended structure formed by the silicon nitride arms and the nanowires (Figure 2f). Hence, the arms retracted, applying a stretching force along the axis of the nanowire. Figure 2g shows an SEM image of a nanowire stressed using this process. The leading purpose of this process is to tune the amount of stretching displacement in the nanowire by adapting the geometry of the silicon nitride arms. Because the nanowire has a very small cross-section as compared to the arms (with an area ratio in the order of 100), the mechanical equilibrium of the device is only dictated by the shape of the arms, which can be modeled as a beam with one clamped and one free end. According to this hypothesis, its elongation can therefore be expressed as a function of its stress state σSiN and its Young’s modulus ESiN by applying Hooke’s law ΔLarm σ = ϵsin = sin Larm Esin
As a result, the free-standing part of the nanowire undergoes the fixed elongation of the two arms and its strain is expressed only as a function of the length of the suspended part Lgap, the 2431
DOI: 10.1021/nl5048219 Nano Lett. 2015, 15, 2429−2433
Letter
Nano Letters
Figure 3. Normalized photocurrent spectrum of an unstrained germanium nanowire, showing an absorption threshold at 1560 nm (blue circles); photocurrent spectra of stressed nanowires with increasing silicon nitride arm length L showing a red shift up to 1700 nm (orange squares, green triangles, and red diamonds). The linear fits of the absorption thresholds are shown as dashed lines.
Figure 4. (a,b) X-ray diffraction image of a strained [111] germanium nanowire. Red squares highlight the germanium diffraction peaks while blue rings highlight the diffraction pattern of the (100) silicon substrate. (c) TM photocurrent spectroscopy measurements (squares) as a function of longitudinal strain along [111], measured by XRD. Experimental data are compared to literature theoretical work for the Γ-LH transition (solid line)33 and show good agreement.
beamline BM32 at the European Synchrotron Radiation Facility.31 The diffraction pattern of a 300 nm wide white Xray beam was collected on a 2D detector at 90° from the incident beam, the sample being at 45°, and the position of the beam relative to the nanowire was monitored using an additional fluorescence detector set to the Ge K edge. Figure 4a,b shows the captured diffraction patterns of a nanowire. The crystal structure of the nanowire was refined from the reflections positions using the LaueTools software suite.32 Being insensitive to the hydrostatic strain, Laue diffraction can only yield the deviatoric part of the strain tensor. The full strain tensor could be obtained assuming no normal stress on the surfaces perpendicular to the axis of the nanowires (i.e., on the free surfaces). Data analysis on nanowires, for instance, the most stressed one (for L = 15 μm), show a deviatoric strain tensor with negligible diagonal coefficients (below 0.02%) and nondiagonal coefficients of 0.561 ± 0.005% in the cubic basis of the crystal. These values are compatible with the stress model (stretching along [111]) and lead to a value of 1.48 ± 0.01% for the uniaxial component of the strain tensor along the axis of the wire. Nanowires with intermediate silicon nitride arm length exhibit intermediate strains: 0.69 ± 0.01 and 0.89 ± 0.01% for 3 and 6 μm, respectively. Figure 4 shows the absorption threshold measured by photocurrent spectroscopy as a function of longitudinal strain, measured by XRD. We do not know the exact experimental value of the Young modulus of the silicon nitride used in our experiments. This does not give us an absolute value of the strain appearing upon
arm relaxation, but a linear increase of ϵwire is expected with increasing Larm. Experiment shows that ϵwire monotonously increases with Larm, thereby supporting in part the previous assumption, while exhibiting a non-null extrapolated strain at zero arm length (this relation is plotted in Supporting Informations). This can be explained by taking into account the residual elongation due to the undercut at the base of the suspended structure. To further understand the relation between measured light absorption shift and strain in the germanium crystal, the selection rules of the optical transitions have to be taken into consideration. Indeed, for a semiconductor with a diamond crystal structure such as germanium, uniaxial strain along a zaxis implies that only the Γ-LH transition couples to polarizations with electric fields along the z-axis because the momentum matrix element for the Γ-HH transition is zero along z34 (see Supporting Informations). This implies that the absorption threshold of photocurrent spectra measured with TM polarization correspond to the Γ-LH transition. We compare our measurements with a theoretical model for germanium stressed along a [111] axis. Deformation potentials provide a behavior for the shift of conduction and valence band edges under stress33 (see Supporting Informations for discussion about the values of the deformation potentials). Figure 4 compares the measured bandgap shift to theoretical EgΓ−LH energy difference, exhibiting a good agreement between theory and experimental data. The nanowire geometry is therefore an efficient way to apply tensile uniaxial stress along a 2432
DOI: 10.1021/nl5048219 Nano Lett. 2015, 15, 2429−2433
Letter
Nano Letters
(8) Jinendra, J. R.; Hryciw, A.; Baer, T. M.; B, D. A. B. M.; Brongersma, M. L.; Howe, R. T. Nat. Photonics 2012, 6, 398−405. (9) de Kersauson, M.; Kurdi, M. E.; David, S.; Checoury, X.; Fishman, G.; Sauvage, S.; Jakomin, R.; Beaudoin, G.; Sagnes, I.; Boucaud, P. Opt. Express 2011, 19, 17925−17934. (10) Ghrib, A.; El Kurdi, M.; de Kersauson, M.; Prost, M.; Sauvage, S.; Checoury, X.; Beaudoin, G.; Sagnes, I.; Boucaud, P. Appl. Phys. Lett. 2013, 102. (11) Capellini, G.; Reich, C.; Guha, S.; Yamamoto, Y.; Lisker, M.; Virgilio, M.; Ghrib, A.; Kurdi, M. E.; Boucaud, P.; Tillack, B.; Schroeder, T. Opt. Express 2014, 22, 399−410. (12) Passi, V.; Bhaskar, U.; Pardoen, T.; Sodervall, U.; Nilsson, B.; Petersson, G.; Hagberg, M.; Raskin, J.-P. J. Microelectromech. Syst. 2012, 21, 822−829. (13) Sanchez-Perez, J. R.; Boztug, C.; Chen, F.; Sudradjat, F. F.; Paskiewicz, D. M.; Jacobson, R.; Lagally, M. G.; Paiella, R. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 18893−18898. (14) Virgilio, M.; Manganelli, C. L.; Grosso, G.; Schroeder, T.; Capellini, G. J. Appl. Phys. 2013, 114. (15) Guan-Yu, W.; He-Ming, Z.; Xiang, G.; Bin, W.; Chun-Yu, Z. Chin. Phys. B 2012, 21, 057103. (16) Fan, W. J. J. Appl. Phys.. 2013114. (17) Chang, G.-E.; Cheng, H. H. J. Phys. D: Appl. Phys. 2013, 46, 065103. (18) Tahini, H.; Chroneos, A.; Grimes, R. W.; Schwingenschloegl, U.; Dimoulas, A. J. Phys.: Condens. Matter 2012, 24, 195802. (19) Zhang, F.; Crespi, V. H.; Zhang, P. Phys. Rev. Lett. 2009, 102, 156401. (20) Ngo, L. T.; Almecija, D.; Sader, J. E.; Daly, B.; Petkov, N.; Holmes, J. D.; Erts, D.; Boland, J. J. Nano Lett. 2006, 6, 2964−2968. (21) Dupre, L.; Buttard, D.; Gentile, P.; Benoit a la Guillaume, Q.; Gorisse, T.; Renevier, H. Phys. Status Solidi RRL 2014, 8, 317−320. (22) Greil, J.; Lugstein, A.; Zeiner, C.; Strasser, G.; Bertagnolli, E. Nano Lett. 2012, 12, 6230−6234. (23) Wagner, R. S.; Ellis, W. C. Appl. Phys. Lett. 1964, 4, 89−90. (24) Gentile, P.; Solanki, A.; Pauc, N.; Oehler, F.; Salem, B.; Rosaz, G.; Baron, T.; Hertog, M. D.; Calvo, V. Nanotechnology 2012, 23, 215702. (25) Kaliteevski, M. A.; Abram, R. A.; Nikolaev, V. V.; Sokolovski, G. S. J. Mod. Opt. 1999, 46, 875−890. (26) Solanki, A.; Gentile, P.; Calvo, V.; Rosaz, G.; Salem, B.; Aimez, V.; Drouin, D.; Pauc, N. Nano Energy 2012, 1, 714 − 722. (27) Solanki, A.; Gentile, P.; Boutami, S.; Calvo, V.; Pauc, N. Adv. Opt. Mater. 2014, 3, 120−128. (28) Hulst, H.; van de Hulst, H. Light Scattering by Small Particles; Dover Books on Physics Series; Dover Publications: Mineola, NY, 1957; p 323. (29) Singh, J. Semiconductor Devices: Basic Principles; John Wiley: New York, 2009. (30) Vorobyev, L. E. In Handbook Series on Semiconductor Parameters; Levinshtein, M., Rumyantsev, S., Shur, M., Eds.; World Scientific: Singapore, 1996; Vol. 1. (31) Ulrich, O.; Biquard, X.; Bleuet, P.; Geaymond, O.; Gergaud, P.; Micha, J. S.; Robach, O.; Rieutord, F. Rev. Sci. Instrum. 2011, 82, 033908. (32) LaueTools: Laue X-ray Microdiffraction Analysis Software, version 5, 2010; http://www.esrf.eu/home/UsersAndScience/ Experiments/CRG/BM32/lauetools-laue-x-ray-microdiffractionanalysis-software.html. (33) van de Walle, C. G. Phys. Rev. B 1989, 39, 1871−1883. (34) Signorello, G.; Karg, S.; Bjork, M. T.; Gotsmann, B.; Riel, H. Nano Lett. 2013, 13, 917−924.
[111] direction in order to probe the influence of strain on the band structure. In conclusion, we have been able to apply uniaxial stress along the [111] axis of germanium nanowires up to a longitudinal strain of 1.48 ± 0.01% thanks to the strain transfer from a stressed silicon nitride layer in a microfabrication process. In particular, this shows that the combination of synchrotron XRD measurements with photocurrent spectroscopy can provide an efficient way to probe the influence of strain on the band structure of germanium. Photocurrent measurements led to a direct bandgap measurement shifted from 0.80 to 0.73 eV, consistent with the deformation potential theory for the Γ-LH transition and the longitudinal strain amplitudes measured by X-ray Laue microdiffraction.
■
ASSOCIATED CONTENT
S Supporting Information *
Core−shell nanowire optical response calculation; strain−stress relations for [111] microbeam and deviatoric tensor; and redshift-strain relation by deformation potential for [111] strain. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work is supported by the CEA projects ‘Phare DSM-DRT: Photonics”, “Phare DSM-DRT: Operando” and the French national program “Programme d’investissement d’Avenir, IRT Nanoelec”. This work has been performed with the help of the “Plateforme technologique amont” in Grenoble with the financial support of the “Nanosciences aux limites de la Nanoélectronique” Foundation and the CNRS Renatech network. The author thanks Jean-Sébastien Micha and Odile Robach for their assistance in the XRD experiments, as well as Laurent Cagnon for performing ALD HfO2 depositions and Pierre Noé for providing stressed silicon nitride on silicon substrates.
■
REFERENCES
(1) Liu, J.; Sun, X.; Camacho-Aguilera, R.; Kimerling, L. C.; Michel, J. Opt. Lett. 2010, 35, 679−681. (2) Camacho-Aguilera, R. E.; Cai, Y.; Patel, N.; Bessette, J. T.; Romagnoli, M.; Kimerling, L. C.; Michel, J. Opt. Express 2012, 20, 11316−11320. (3) Nam, D.; Sukhdeo, D. S.; Kang, J.-H.; Petykiewicz, J.; Lee, J. H.; Jung, W. S.; Vuckovic, J.; Brongersma, M. L.; Saraswat, K. C. Nano Lett. 2013, 13, 3118−3123. (4) Suess, M. J.; Geiger, R.; Minamisawa, A. R.; Schiefler, G.; Frigerio, J.; Chrastina, D.; Isella, G.; Spolenak, R.; Faist, J.; Sigg, H. Nat. Photonics 2013, 7, 466−472. (5) Boztug, C.; Sanchez-Perez, J. R.; Cavallo, F.; Lagally, M. G.; Paiella, R. ACS Nano 2014, 8, 3136−3151. (6) Boucaud, P.; Kurdi, M. E.; Ghrib, A.; Prost, M.; de Kersauson, M.; Sauvage, S.; Aniel, F.; Checoury, X.; Beaudoin, G.; Largeau, L.; Sagnes, I.; Ndong, G.; Chaigneau, M.; Ossikovski, R. Photon. Res. 2013, 1, 102−109. (7) Kuroyanagi, R.; Nguyen, L.; Tsuchizawa, T.; Ishikawa, Y.; Yamada, K.; Wada, K. Opt. Express 2013, 21, 18553−18557. 2433
DOI: 10.1021/nl5048219 Nano Lett. 2015, 15, 2429−2433
