Improving Nanowire Sensing Capability by Electrical Field

Kuan-I Chen , Chien-Yuan Pan , Keng-Hui Li , Ying-Chih Huang , Chia-Wei Lu , Chuan-Yi Tang , Ya-Wen Su , Ling-Wei Tseng , Kun-Chang Tseng , Chi-Yun Li...
0 downloads 0 Views 326KB Size
Download PDF
Letter pubs.acs.org/NanoLett

Improving Nanowire Sensing Capability by Electrical Field Alignment of Surface Probing Molecules Chia-Jung Chu,†,∥ Chia-Sen Yeh,†,∥ Chun-Kai Liao,†,∥ Li-Chu Tsai,*,† Chun-Ming Huang,∥,‡ Hung-Yi Lin,*,‡ Jing-Jong Shyue,§ Yit-Tsong Chen,⊥ and Chii-Dong Chen*,∥ †

Institute of Organic and Polymeric Materials, National Taipei University of Technology, Taipei, 10608, Taiwan Department of Materials Science, National University of Tainan, Tainan 70005, Taiwan § Research Center for Applied Sciences, Academia Sinica, Taipei 11529, Taiwan ∥ Institute of Physics, Academia Sinica, Nankang, Taipei 11529, Taiwan ⊥ Department of Chemistry, National Taiwan University, Taipei 106, Taiwan ‡

S Supporting Information *

ABSTRACT: We argue that the structure ordering of selfassembled probing molecular monolayers is essential for the reliability and sensitivity of nanowire-based field-effect sensors because it can promote the efficiency for molecular interactions as well as strengthen the molecular dipole field experienced by the nanowires. In the case of monolayers, we showed that structure ordering could be improved by means of electrical field alignment. This technique was then employed to align multilayer complexes for nanowire sensing applications. The sensitivity we achieved for detection of hybridization between 15-base single-strand DNA molecules is 0.1 fM and for alcohol sensors is 0.5 ppm. The reliability was confirmed by repeated tests on chips that contain multiple nanowire sensors. KEYWORDS: Structure ordering, molecular dipole charge, silicon nanowire field-effect transistor, DNA sensor, alcohol sensor

S

be the same for both polar and nonpolar molecules. The electrical field alignment technique is thus applicable to all molecular monolayers or multilayer complexes. A similar mechanism is employed in liquid crystal5 as well as selfassembled fibers,6 but there the molecules are in the form of aggregated macroscopic clusters. It should have a significant impact for applications in which structure ordering is important, such as organic semiconductor transistors7 and opto-electronics.8,9 In this study, the effect of alignment is evaluated by using underneath silicon nanowire FETs (SiNWFETs). Using the SiNW-FETs for detection of molecular interactions, we show a prominent enhancement on the detection capability for DNA and alcohol sensors. The chips used here contained electron-beam lithographically made p-type SiNW-FETs, as shown in the inset of Figure 1a. The backsides of the silicon chips were coated with a metal layer, which served as a backgate electrode. The transistors were made on silicon-on-insulator (SOI) wafers and were electrically isolated by a 10 nm thick thermally oxidized SiO2 insulating layer. Details of the device fabrication4 are reported elsewhere, and device structure10−12 is described in the Supporting Information. There are eight SiNWs on each chip, and a homemade current amplifier enables simultaneous

emiconducting nanowire-based sensor is a promising candidate for accurate, label-free, real-time detection of biological and chemical molecules. An example of the sensors is silicon nanowires made by either bottom-up1 or top-down technologies.2 In these devices, electron transferring is blocked by an insulating layer covering the nanowires and conventional field-effect gating mechanism is employed. The needed gating field is provided by the probing molecules attached on the insulating surface. Upon interaction between the target and probing molecules, a change in the molecular dipole charges alters the number of conduction carriers in the nanowires, providing a change in the nanowire current and conductance. The response of the nanowire conduction upon molecular interaction is largely determined by the strength of the molecular dipole field generated. Structural ordering of the surface-modified monolayer3 is crucial for high sensitivity4 since a well-aligned molecular monolayer can dramatically enhance collective charge polarization brought about by molecular interactions. More specifically, a global dipole moment produced by well-aligned straight molecules perpendicular to the nanowire surface is desirable. Here, we propose and demonstrate a simple, unique, and noninvasive approach in which the molecular dipole is forced to align to an externally applied electrical field. Depending on the polarity, this dipole could be permanent for polar molecules or electric-fieldinduced for nonpolar molecules. Although the latter is much weak, the proposed electric field alignment mechanism would © XXXX American Chemical Society

Received: February 20, 2013 Revised: April 20, 2013

A

dx.doi.org/10.1021/nl400645j | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

Figure 1. Electric field alignment of APTES on SiNW FETs. (a) SEM image of a SiNW-FET. The width, length, and thickness of the wire are 200 nm, 3 μm, and 40 nm, respectively, and the wire is covered by a 10 nm thick thermally oxidized layer. The scale bar is 2 μm. (b, c) Schematic illustrations of the possible APTES molecular structures before and after alignment process.

In Figure 2a, we examine the onset field strength that is needed for the effect of alignment to become apparent and the

measurement of the current on all eight SiNW-FETs. This is important in terms of measurement reliability, as there are uncontrollable factors that may affect the quality of molecular modification on the SiNW surface. With this device design and measurement configuration, the reproducibility of the study presented in this paper is ensured. The surface of the SiNWs was modified with the APTES (3-aminopropyltriethoxysilane) monolayer. APTES is an intermediate self-assembled organic monolayer, which consists of a silane group at one end and amino group at the other end. It has been widely used in affinity-based biosensors13 because the silane group can bind tightly to silica or glass substrates while the amino group can form covalent bonds with carboxyl groups, which are commonly found in biomolecules.14 With only three carbons on number of molecular configurations15 are reported, representing random polarization orientations (see Figure 1b). Because of these complications, APTES is a good candidate for testing and illustrating the proposed molecule alignment technique. The electric field exerted on to the molecules was produced by applying a voltage VE on a metal plate about 1 mm above the chip surface while the backgate electrode was grounded. The field experienced by the molecules was uniform and could be precisely determined. To study the effect of alignment, the molecules were subjected to an electrical field for a short time (referred to as “alignment process”), and then the field was set to zero for a couple of seconds (called relaxation time) to allow for molecule internal stress to relax. After this alignment process, the degree of structural ordering was probed by measuring the nanowire current of the underneath SiNWs versus backgate voltage (IVg) characteristics; this is referred to as a “probing process”. Both alignment and probing processes were performed in the vacuum environment. The experiment started from attachment of APTES monolayers on the surface of SiNWs, and the monolayer was then subjected to an alignment process, as shown in Figure 1c. The effect of alignment was initially confirmed by improved attachment efficiency of silver nanoparticles on E-field-aligned APTES modified on SiO2 surface (detailed in the Supporting Information). After that, we carried out two independent tests to show that the alignment process would improve the detection capability for molecular interactions. In the first test, APTES was linked to glutaraldehyde, which was, in turn, linked to poly-T single-strand DNA (ssDNA). The modified SiNW-FETs were then employed for detection of hybridization with poly-A ssDNA. In the second test, the APTES on the wire surface was linked with iron porphyrin, which served as a sensing molecule for detection of ethanol and methanol.

Figure 2. Nanowire response to the alignment process. Measured SiNW current at fixed Vds and Vg after E-field alignment with varying VE. In (a) VE was ramped from 0 to 0.7 V with a step of 0.1 V; Vds and Vg values were +2.5 V and −3.0 V, respectively. The effect of alignment became clear for VE greater than 0.3 V and saturated at VE ≈ 0.6 V. The data were extracted from measured I−Vg curves at each stage displayed in the inset, in which the curves from left to right are shown for VE from 0 to 0.7 V with a step of 0.1 V. In (b) VE jumped directly from 0 to 0.5 V and then back to 0; Vds = +1.5 V and Vg = −4.5 V. The retaining time was examined by simply setting VE = 0, but all the alignment and probing processes remained the same. The I−Vg curves are displayed in the inset; the two overlapped curves are for VE = 0.5 and 0 (after alignment). All measurements were performed in a vacuum.

saturation field under which the structure is well aligned. It was found that the nanowire current at given bias voltage Vds and gate voltage Vg increases and then saturates after consecutive alignment/probing cycles. The field strength was increased step by step; each step consists of 20 cycles. In each cycle, the field was applied for 1 min and the relaxation time was set to 30 s. The current values shown in the plot were extracted from the current vs gate voltage (I−Vg) curves (see inset) obtained in the probing process. It appeared that a small field of 0.3 V/mm was enough to produce a clear alignment effect, and the effect B

dx.doi.org/10.1021/nl400645j | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

Figure 3. Shifts in the I−Vg curves after each attachment and alignment process. I−Vg curves for SiNW with unaligned and aligned molecules are displayed as dashed and solid curves, respectively. The curves were taken at Vds = +4 V in a vacuum.

saturates at a field of about 0.6 V/mm. In a subsequent experiment, we set a field of 0.5 V/mm directly but kept alignment and relaxation times unchanged. As shown in Figure 2b, after about 20 alignment processes, the current started to saturate. Nevertheless, we continued to perform several alignment processes and found the current did not increase further, suggesting that the molecular structure was stabilized at the alignment field and a global ordered state was achieved. We then turned the alignment field to zero but continued to probe the molecule for more than 30 times, corresponding to an elapsed time of ∼2 h, and found that the wire current remained unchanged. The I−Vg curves shown in the inset clearly display the stability of the molecules after alignment. This experiment reveals that once aligned, the molecule structural conformation can maintain for hours or longer. Within this retaining time, subsequent modification process or molecule sensing applications can be carried out. It is noted that the molecules at the sidewalls of the SiNWs would experience a tilted fringing electric field, and the efficiency of alignment for these molecules is lower. Nevertheless, this should not affect the applicability of this method. Detection of DNA hybridization was then used as an example for demonstrating the effect of E-field alignment. In this experiment, p-type SiNWs were employed. The first step was APTES modification, and this was followed by an alignment/probing process. Subsequently, glutaraldehyde molecules were linked to the APTES layer, and the resultant multilayer was subjected a second alignment/probing process. Finally, the APTES/glutaraldehyde layer was attached by poly15T ssDNA, and a third alignment/probing process was performed. The alignment process for APTES was the same as that described previously. The alignment fields for APTES/ glutaraldehyde and for APTES/glutaraldehyde/poly-15T ssDNA were −0.1 and +0.1 V/mm, respectively, and were applied for 1 min for both cases. Although the polarity and the magnitude of the needed alignment electric field can in principle be analyzed based on the molecular charge models, the analysis is complex due to the intricacy of the actual molecular configurations and of the experimental environments (e.g., air or various solutions). The polarity and magnitude of the alignment electric fields were thus figured out by trials. Even before alignment, the asattached molecular layer can shift the I−Vg curve toward positive or negative Vg, but the curve is not necessary parallel to the one before attachment of the molecules. As a rule of thumb, the alignment field polarity can be determined from the direction of this shift. For example, in the inset of Figure 2a, the as-modified APTES layer shifted the I−Vg of a p-type SiNW

toward positive Vg; the alignment field should then be positive. Conversely, if the as-processed molecular layer shifts the I−Vg curves toward negative Vg, negative alignment fields should be employed. For the case of n-type SiNWs, the same rule is applied, but the field polarity has to be reversed. The shifts in the threshold voltage after modification of subsequent molecule layers were attributed16 to the transferring of the effective dipole moment from the molecule chain toward the nanowire oxide surface. For example, a positive shift in the threshold is due to negatively charged oxygen atoms located right on top of silicon oxide surface (see Figure 1b,c) when APTES binds to SiO2. The magnitude of the alignment field is set to a smallest possible value provided that the shift is appreciable, and this shift should gradually decrease and diminish after several alignment processes. As shown in Figure 3, before and after attachment of each molecular layer the I−Vg curves shifted parallel; that is, for any nonzero currents, every point in the I− Vg curve move along the Vg axis by the same amount. The amount of shifts in Vg may be correlated to the change in the effective molecular dipole field induced by the attachment of the adding molecules. The same type of parallel shifts in I−Vg curves was also reported.16 In our devices, eight SiNWs were modified and measured at the same time and statistical analysis helped to identify appropriate alignment fields. After the above modification and alignment procedures, a PDMS pad containing microfluidic channel was bonded to the SiNW chip and detection of DNA hybridization reaction with poly-15A ssDNA was carried out. The effect of the alignment on the efficiency of detection can be seen from Figure 4. For ptype SiNWs, the hybridization causes I−Vg to shift toward positive Vg. The shift is highly reproducible, and in Figure 4a we show five typical sets of I−Vg curves that reflect the reliable detection of DNA hybridization. In a control experiment, the modification was carried out in the same beaker, but the molecular layers were not subjected to any alignment process. As displayed in Figure 4b, the I−Vg curves did not shift parallel. A parallel shift in I−Vg curves before and after modification of subsequent molecule layer means that the hybrid molecules present a constant dipole field, just like that of an offset in the backgate voltage. However, we note that the Coulomb interaction between the charge carriers in the nanowires and the molecular charges as well as the solution ions within the electric double layer (which has a thickness of Debye length) may result in a force acting on the surface molecules. Since the carrier concentration and distribution in the nanowires may be modulated by the backgate voltage, the net force acting on the surface molecules also changes with gate voltage. In the case of C

dx.doi.org/10.1021/nl400645j | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

Tris buffer (from Bioman Scientific Co.) and 0.01×PBS (Dulbecco’s Phosphate Buffered Saline, from Invitrogen). To ensure the accuracy of the DNA concentration, we diluted the DNA solution by one-tenth and performed hybridization detection tests at all concentrations. As shown in Figure 5, upon introduction of target DNA the nanowire current increased, indicating occurrence of hybridization reaction. In all measurements, control experiments were performed using the poly-15C ssDNA of the same concentration as well as buffer solution to exclude possibility of false detection. Depending on the buffer solution, the achieved sensitivity was different even for detection of the same DNA molecules. While for DNA in Tris buffer solution we obtained a detection sensitivity of 1 pM, for DNA in PBS buffer solution the achieved sensitivity was 0.1 fM. The high sensitivity in PBS solution may be attributed to long Debye length; The Debye length in Tris is estimated to be about 1 nm, whereas it is about 7 nm in PBS. Although the same detection limit was reported17 very recently for detection of hybridization between 24-base ssDNA using SiNW FETs, a shorter DNA that would require a higher detection capability18 is employed in this work. Note that the time interval between introduction of target DNA and saturation of Ids is considerably short in the 1 pM case as compared with that in the 0.1 fM case. The fast response in the 1 pM case is attributed to a high probability for the target molecules to find the probe molecules, in spite of a short Debye length. We further comment that the additional E-field alignment can improve not only the detection sensitivity but also the detection reliability. Detection of 0.1 fM DNA hybridization could be performed with high repeatability. Detection of methanol and ethanol gases was used as another example to demonstrate the application of the proposed electrical field alignment technique for improving sensitivity. Iron porphyrin was employed as the sensing molecule for these gases,19 and APTES served as the interfacial molecule linking porphyrin and the oxide surface of SiNWs. In the alignment process, an electric field of −0.5 V/mm was applied on the APTES/porphyrin molecules for 6 min. Figure 6 shows a comparison of the response in the nanowire current upon detection of the alcohol gases for as-attached and field-aligned molecular sensors. Since the alignment process lifts the

Figure 4. Comparison of I−Vg characteristics showing the effect of Efield alignment process on detection of DNA hybridization. The panels show I−Vg curves of five SiNWs (a) with and (b) without E-field alignment. Blue and red curves are I−Vg curves before and after hybridization of 15T ssDNA with 15A ssDNA measured in Tris buffer solution.

unaligned molecules, structural disorder makes it less resistant to the disturbance produced by the gate, and the molecule structure change results in nonparallel I−Vg curves. Conversely, the parallel shift in I−Vg curves before and after modification suggests that the molecule structure is more rigid. Structural ordering promotes efficiency of molecular reactions, strengthens the molecular dipole field, and can improve detection sensitivity. With the E-field alignment, we achieved a high sensitivity for real-time detection of hybridization between single-strand DNA (ssDNA) molecules of as short as 15 base pairs using sensors based on p-type SiNW field effect transistors. In this experiment, we prepared target poly15A ssDNA solutions in two types of buffers, namely 10 mM

Figure 5. Real-time detection of DNA hybridization. In (a) the target is 1 pM 15A ssDNA in Tris solution, while in (c) the target is 0.1 fM 15A ssDNA in PBS solution. Upon introduction of poly-15A ssDNA, a clear increase in the nanowire current was observed. In both (a) and (c) buffer was first added to ensure stability of the sensor. (b) and (d) show control experiments for (a) and (c), respectively. The bias conditions were Vds = +1.5 V, Vg = −1.0 V in (a, b) and were Vds = +1.6 V, Vg = −2.5 V in (c, d). D

dx.doi.org/10.1021/nl400645j | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

Figure 6. Nanowire current change upon introduction of (a) methanol and (b) ethanol as a function of gas concentration. Dashed and solid curves indicate responses before and after alignment process, respectively.

into a NaOH solution (1 M) containing iron-porphyrin (10 M) for 2 h.20 This was followed by a thorough rinse with NaOH solvent and sonication in NaOH bath for 5 min to remove physisorbed porphyrin molecules.

porphyrin molecules to promote the interaction, the enhanced response is consistent with our expectation. It is interesting to note that methanol always produces a larger nanowire current response than for ethanol, which suggests that absorption of methanol yields a greater negative equivalent charge. Conclusions. Taking APTES monolayer as an example, the SAM structure was shown to align to the direction of the applied electrical field. The structural ordering of APTES monolayers covalently bound on a silicon nanowire surface was examined by studying the response of underneath SiNW FETs. This field-alignment technique was shown to improve the binding efficiency of Ag nanoparticles on APTES molecules. For multilayer complexes, the alignment yielded parallel shifts of I−Vg curves at each stage of molecule modification. This parallel shift is a signature of high quality surface modification and is a key to the performance of SiNW-FET sensors. This alignment technique was shown to enhance detection capability and reliability for alcohol and DNA sensors. Detection of hybridization between 15-base ssDNA molecules with a high sensitivity of 0.1 fM was demonstrated. In addition, we also showed detection of methanol and ethanol down to 0.5 ppm using iron porphyrin as the probing molecule. Statistically, the alignment process can improve the sensitivity by up to 3 orders, but the actual effect depends on types of molecules as well as the modification procedures. This technique is important for many applications where interfacial molecule layer is employed. Experimental Section. Attachment of APTES Molecules. The procedures for the attachment of the APTES monolayer on the chips are as follows: The chips were first soaked in a 2% cholic acid in ethanol for 12 h to generate −OH on the SiO2 surface. This was followed by soaking in APTES (from SigmaAldrich, purity 99%) solution (2% in acetone) at room temperature for 1 h to form a monolayer on the surface. The devices were then cleaned with distilled water and blow-dried with nitrogen to remove the unbound molecules. Finally, the chips were bake-dried at 110 °C for 1 h. Attachment of Probe Single-Strand DNA. The attachment was carried out by immersing amino-terminated silanized nanowire FETs into 12.5% glutaraldehyde (from SigmaAldrich) in 0.1×PBS (Dulbecco’s Phosphate Buffered Saline, from Invitrogen, pH = 8) for 1 h. The chip was then rinsed with distilled water and blow-dried with nitrogen. This is then followed by immobilization of probe DNA, in which poly-15T ssDNA (10nM) was soaked in 0.1×PBS for 1 h. Finally, the SiNW FETs were rinsed thoroughly with distilled water and blow-dried with N2. Attachment of Iron Porphyrin. The attachment was carried out by immersing amino-terminated silanized nanowire FETs

■

ASSOCIATED CONTENT

S Supporting Information *

Experimental details; Figures S1 and S2. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected] (L.-C.T.); [email protected]. tw (H.-Y.L.); [email protected] (C.-D.C.). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This research was funded by the National Science Council Nos. NSC 101-2311-B-027-001- and NSC 101-2627-M-001-003and by National Taipei University of Technology. Prof. Dar-Bin Shieh is appreciated for supplying the Ag nanoparticles. We thank Tzu-Hui Hsu, Hsun-Chung Wu, Ming-Chou Lin, and Yann-Wen Lan for their assistance during the course of this work. Technical support from NanoCore, the Core Facilities for Nanoscience and Nanotechnology at Academia Sinica is acknowledged.

■

REFERENCES

(1) Patolsky, F.; Zheng, G.; Lieber, C. M. Anal. Chem. 2006, 78, 4260−4269. (2) Stern, E.; Klemic, J. F.; Routenberg, D. A.; Wyrembak, P. N.; Turner-Evans, D. B.; Hamilton, A. D.; LaVan, D. A.; Fahmy, T. M.; Reed, M. A. Nature 2007, 445, 519−522. (3) Ulman, A. J. Am. Chem. Soc. 2000, 122, 4047−4055. (4) Lin, M. C.; Chu, C. J.; Tsai, L. C.; Lin, H. Y.; Wu, C. S.; Wu, Y. P.; Wu, Y. N.; Shieh, D. B.; Su, Y. W.; Chen, C. D. Nano Lett. 2007, 7, 3656−3661. (5) Shah, A. A.; Kang, H.; Kohlstedt, K. L.; Ahn, K. H.; Glotzer, S. C.; Monroe, C. W.; Solomon, M. J. Small 2012, 8 (No. 10), 1551−1562. (6) Yoshio, M.; Shoji, Y.; Tochigi, Y.; Nishikawa, Y.; Kato, T. J. Am. Chem. Soc. 2009, 131, 6763−6767. (7) Hutchins, D. O.; Weidner, T.; Baio, J.; Polishak, B.; Acton, O.; Cernetic, N.; Ma, H.; Jen, A. K. Y. J. Mater. Chem. C 2013, 1, 101−113. (8) Fichou, D. J. Mater. Chem. 2000, 10, 571−588. (9) Yu, L. S.; Tseng, H. E.; Lu, H. H.; Chen, S. A. Appl. Phys. Lett. 2002, 81, 2014−2016. (10) Lin, S. P.; Pan, C. Y.; Tseng, K. C.; Lin, M. C.; Chen, C. D.; Tsaia, C. C.; Yua, S. H.; Sun, Y. C.; Lin, T. W.; Chen, Y. T. Nano Today 2009, 4, 235−243.

E

dx.doi.org/10.1021/nl400645j | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

(11) Chang, K. S.; Sun, C. J.; Chiang, P. L.; Chou, A. C.; Lin, M. J.; Liang, C .; Hung, H. H.; Yeh, Y. H.; Chen, C. D.; Pan, C. Y.; Chen, Y. T. Biosens. Bioelectron. 2012, 31, 137−143. (12) Lin, C. H.; Chu, C. J.; Teng, K. N.; Su, Y. J.; Chen, C. D.; Tsai, L. C.; Yang, Y. S. Jpn. J. Appl. Phys. 2012, 51, 02BL02. (13) Knopp, D.; Tang, D.; Niessner, R. Anal. Chim. Acta 2009, 647, 14−30. (14) Bierbaum, K.; Kinzler, M.; Woell, C.; Grunze, M.; Haehner, G.; Heid, S.; Effenberger, F. Langmuir 1995, 11, 512−518. (15) Golub, A. A.; Zubenko, A. I.; Zhmud, B. V. J. Colloid Interface Sci. 1996, 179, 482−487. (16) Paska, Y.; Haick, H. Appl. Phys. Lett. 2009, 95, 233103. (17) Gao, A.; Lu, N.; Wang, Y.; Dai, P.; Li, T.; Gao, X.; Wang, Y.; Fan, C. Nano Lett. 2012, 12, 5262−5268. (18) Kim, B.; Lee, J.; Namgung, S.; Kim, J.; Park, J. Y.; Lee, M. S.; Hong, S. Sens. Actuators, B 2012, 169, 182−187. (19) Tonezzer, M.; Quaranta, A.; Maggioni, G.; Carturan, S.; Mea, G. D. Sens. Actuators, B 2007, 122, 620−626. (20) Belyakova, L. A.; Besarab, L. N.; Roik, N. V.; Lyashenko, D. Y.; Vlasova, N. N.; Golovkova, L. P.; Chuiko, A. A. J. Colloid Interface Sci. 2006, 294, 11−20.

F

dx.doi.org/10.1021/nl400645j | Nano Lett. XXXX, XXX, XXX−XXX