Electrically Excited Plasmonic Nanoruler for Biomolecule Detection

Aug 22, 2016 - Support. Get Help · For Advertisers · Institutional Sales; Live Chat. Partners. Atypon; CHORUS; COPE; COUNTER; CrossRef; CrossCheck ...
0 downloads 0 Views 1MB Size
Subscriber access provided by United Arab Emirates University | Libraries Deanship

Communication

Electrically excited Plasmonic Nanoruler for Biomolecule Detection André Dathe, Mario Ziegler, Uwe Huebner, Wolfgang Fritzsche, and Ondrej Stranik Nano Lett., Just Accepted Manuscript • DOI: 10.1021/acs.nanolett.6b02414 • Publication Date (Web): 22 Aug 2016 Downloaded from http://pubs.acs.org on August 26, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Nano Letters is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Electrically excited Plasmonic Nanoruler for Biomolecule Detection André Dathe(1), Mario Ziegler(2), Uwe Hübner(2), Wolfgang Fritzsche(1) and Ondrej Stranik*(1) (1) Department of Nanobiophotonics, Leibniz Institute of Photonic Technology (IPHT), Jena (2) Department of Quantum Detection, Leibniz Institute of Photonic Technology (IPHT), Jena Keywords: plasmonic nanoruler, electrical excitation, inelastic tunneling, nanoparticle, biomolecule detection Plasmon based sensors are excellent tools for a label free detection of small biomolecules. An interesting group of such sensors are plasmonic nanorulers which rely on the plasmon hybridization upon modification of their morphology in order to sense nanoscale distances. Sensor geometries based on interaction of plasmons in a flat metallic layer together with metal nanoparticles inherit unique advantages, but need a special optical excitation configuration not easy to miniaturize. Herein we introduce the concept of nanoruler excitation by direct, electrically induced generation of surface plasmons based on the quantum shot noise of tunneling currents. An electron tunneling junction consisting of a metaldielectric-semiconductor hetero-structure is directly incorporated into the nanorulers basic geometry. By applying voltage on this modified nanoruler, the plasmon modes are directly excited without any additional optical component as a light source. We demonstrate by several experiments that this electrically driven nanoruler possesses similar properties as an optically exited one and confirm its sensing capabilities by the detection of the binding of small biomolecules such as antibodies. This new sensing principle could open the way to a new platform of highly miniaturized, integrated plasmonic sensors compatible to monolithic integrated circuits.

Plasmonics became a very progressive field of research in the last decades, because of its promise to control and monitor light at nanometer and femtosecond scale1. The essence of plasmonics is a collective excitation of conduction electrons in metals2. Because the screening length of most metals is in the nanometer range, plasmons can be tuned by the shape, size and dimensionality of the supporting nanostructure, and the energy of the plasmons can be in the range of visible light3. Exploiting these properties plasmons advance their way into the application for nanoscale terahertz and optical devices4, optical antennas5,6, fabrication of metamaterials7–9, and enhanced Raman spectroscopy10,11. Recently, plasmonics expands into quantum and non-linear optics by offering applications in frequency conversion, switching of optical signals9,12 and single photon sources13. Among these, the most prominent and pushed ahead application is biomolecule detection14–20. The underlying principle of most of these applications is a spectroscopic characterization of the plasmon on the nanostructures during adsorption of biomolecules. Plasmons on a metal surface can be classified21 into propagating surface plasmons (pSP) generated on two dimensional metal surfaces, and localized surface plasmons (lSP) excited on 3D spatially confined nanostructures. As the terminology indicates, the pSP can propagate along the boundary, and lSP are localized on the nanostructure with a stronger spatial confinement of the optical energy. In complex nanostructures, plasmons can interact with the neighboring plasmons, and thereby

change their energies and spatial confinement. In this way, for example, the interaction of pSP in metallic layers leads to the generation of long/short -range plasmons22,23 and the interaction of lSP in nanostructures creates Fanoresonance type plasmons24–26, which can be - in analogy to the model of electron levels in atoms - described as ‘plasmon hybridization’27. Hybridization of lSP of metallic nanoparticles is the basis for a label free detection of molecules, e.g. in colorimetric bioassays28. The target molecules mixed with the nanoparticles induce aggregation of the nanoparticles; the resulting energy shift due to the plasmon hybridization is manifested by a change of the color of scattered or absorbed light. This principle allows detection of even a single DNA strand by measurement of plasmon hybridization of two nanoparticles (called nanoruler)29. The hybridization of DNA shortens the distance between the nanoparticles, and therefore changes the resonance condition of the plasmon. In general, the presence of the analyte molecule results in a changed distance between the two components of the nanoruler system, and this nanoscale change translates into significant changes in the spectroscopic behavior of the system, which are utilized as sensing signal. In our case, we are interested in the hybridization of propagating and localized surface plasmons. This happens, when a metallic nanoparticle (NP) is in close proximity to a metal film30. In general, the hybridization leads to several plasmon hybridization energies, but for small NP-substrate distances and a thick metal layer, the sys-

ACS Paragon Plus Environment

1

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

tem is in analogy to two particles. The plasmon resonance of the system strongly depends on the separation (few nanometer scale), and the energy of the plasmon is shifted to a lower energy31,32. This arrangement of a NP nearby a metal film has a very well defined NP-substrate separation (direction is perpendicular to the substrate), and therefore, it has a potential for a very sensitive detector based on the nanoruler concept described above. The challenges with the implementation of these effects, known as gap-modes, are their efficient excitation and detection. Currently, the spectroscopy of gap-modes on single nanostructures is usually done by microspectroscopy with dark-field illumination. Since the metallic layer inhibits detection in transmission, the freespace illumination has to be in reflection mode. An alternative illumination can be realized by evanescent field excitation of pSP on the metallic layer32. However, this can only be doneefficiently for a narrow energy range due to the necessary exploitation of resonant coupling conditions. The general disadvantage of these optical excitations is their demand for additional optical elements and the resulting difficulty of the integration. Our solution for the excitation of the gap-modes is an electrically induced generation of spectrally broad bandedpSPon the metallic substrates instead of the optical excitation. The main idea is the extension of the metallic substrate into a thin-film stack of metal-insulatorsemiconductor layers. Such a structure forms an electrical tunneling junction, which - after applying a voltage across - generates surface plasmons on the metallic layer based on the quantum fluctuation of the tunneling current33. While these light emitting tunnel junctions (LETJ) were initially intended for far-field emission34, it was noticed that the efficiency of the far-field light emission is lower than that of competing emission processes. One of the reasons was the generation of surface plasmons35. However, due to the formation of pSP, LETJ are excellently suited for the integration in the nanoruler sensor concept based on gap-modes. Here we demonstrate as proof-of-concept a novel type of bio-molecular sensor, employing ultra-compact electrical excitation of the plasmon modes combined with the highly sensitive distance dependent energy shift of the plasmon gap-modes in the NP-substrate system. We confirmed that the spectroscopic properties of the optically induced as well as electrically induced gap-modes are similar, and were able to demonstrate the electrically induced gap-mode even from single NP on the substrate. Such findings open new approaches for the development of compact ultrasensitive bio-sensors based on the spectroscopy of gap-mode plasmons. Currently, surface plasmons are usually optically excited. However, electron excitation was initially utilized to study surface plasmons36, when a beam of high energy electrons was directed towards the metallic substrate. In the modified forms of electron energy loss spectroscopy37 or cathode-luminescence spectroscopy38, a high spatial resolution regarding the characterization of plasmons on nanostructures in vacuum is achieved. Another approach utilizes the tip of a scanning tunneling microscope as

Page 2 of 11

electron source. The tip is placed only a few nm above a nanoscale metallic structure of interest, so that the inelastically tunneling electrons excite the plasmons in the nanostructure39–41. Recently, electron tunneling between lithographically produced electrodes was used to excite plasmons on particles located in between these biased films42,43. In our device, we utilize inelastic electron tunneling in a planar metal-insulator-semiconductor heterostructure44 (compatible to standard Si-processing technology) to excite pSP. This method is based on the discovery by Lambe45 (originally in a metal-insulator-metal system) that such biased structures (LETJ) emit light in the visible regime. The effect was fully theoretically derived by Uehara33,44 . Fig. 1(I) represents the energy diagram of the metal/oxide/n-doped semiconductor tunneling junction with the metal being positively biased against the semiconductor (the opposite bias is also possible, but less efficient). The applied voltage shifts the Fermi-energy level, and the electrons starts to tunnel through the upper, triangularly shaped barrier created in the oxide layer from semiconductor to the metal. For an oxide layer thickness of only a few nanometers, the structure should be in the regime of Fowler-Nordheim tunneling (j = aE²exp(b/E), where j is the current density, E is the applied voltage and a, b are constants). Herein, the height of the triangular barrier is determined by the energy difference between the electron affinity of the oxide, the work function of the metal and the semiconductor doping. The width of the barrier is governed by the thickness of the oxide layer and the magnitude of the applied voltage. Such tunneling processes exhibit fluctuations of the current, which have their origin in quantum mechanics, and their power spectrum reach up to the optical frequencies (expressions for the current characteristic and its power spectrum in dependence on the barrier parameter are in Fig. S1). The current fluctuation is equivalent to the sum of oscillating current sources over a broad range of frequencies and different amplitudes embedded in the oxide layer, which radiate the electromagnetic energy. The finite element method (FEM) simulation in Fig. 1A shows a current source that is oscillating at the optical frequency and placed in the middle of the oxide layer. The false color map represents the x-component of the electrical field in order to visualize the different paths of the light emission. As described in the theory, it is visible that the major portion of the emitted light goes to the silicon substrate and a small portion is emitted into the free space above the junction in the form of spherical waves, which was observed by Lambe45 and other authors46–50. A third portion is propagating along the metal boundary in the form of surface pSP, which do not radiate into the free space due to the large k-vector. In order to increase the light emission efficiency of the LETJ structures, several methods such as implementation of artificial surface-roughness by interlayers46,51–53,gratings54–56, multiple-tunneling barriers57, prism-coupling35,58 and annealing of the surface to form nano-agglomerates59– 61 were studied, which should enhance the conversion of pSPinto free space radiation.

ACS Paragon Plus Environment

2

Page 3 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Figure 1: (A)Principle of the light emission process in a LETJ structure. As shown in the energy diagram of the LETJ chip, if a bias Vg across the thin film junction is applied, tunneling processes (e.g. Fowler-Nordheim tunneling) dominate the conduction mechanism. Accompanying quantum shot noise leads to light emission, which originates from the oxide layer. The light emission consists of direct emission and pSP propagation along the metal boundary.(B) Taking a planar gold-air interface as a basis, the surface plasmon polariton is evanescent as shown exemplary in the depiction of the pSP magnetic field component HzpSP. If a nano antenna (e.g. gold particle d = 80 nm) is added in the vicinity, the pSP is perturbed and scattered into free-space radiation. The scattered portion Hzsca is displayed. (C) The scattering spectra were simulated for different distances between the metal surface and particle dFP. (D) The spectral response of the gap-mode was obtained by determination of the peak centroid as function of dFP. (Inset) Gap-mode coupling is represented by the time-averaged total electric field Eavg in the insets for three representative distances (a= 5 nm, b= 15 nm, c= 30 nm).

Our device design exploits NPs as the scatterers of the generated pSP in the LETJ and relies on spectral changes of the plasmon gap-modes on the distance variation between metal substrate and metal NP31. A FEM simulation of the present pSP that is scattered due to a metallic NP is shown in Fig. 1B (the color map represents the instant distribution of the z-magnetic field). The upper simulation displays the non-perturbedpSP, which exponentially decays with increasing distance from the surface. The simulation below demonstrates the effect of a NP for the conversion of the pSP into free space radiation. We carried out simulations for different NP-substrate distances over a broad spectral range and integrated the total scattered power into free space. The graph in Fig. 1C shows the spectral dependence of the scattered power for different NP-metal distancesdFP. Each spectrum exhibits a distinctive peak, which changes its position and intensity. The peak position as function of the separationdFP is given in Fig. 1D.For very small separations, the pSP is coupling with the gap-mode(strong field localization between the NP and substrate – inset (a)Fig. 1D), which has its resonance at the longer wavelength. With increasing separation this mode is spectrally shifted and attenuated (inset (b) Fig. 1D). For larger separations, the pSP is scattered by the single particle at the lSP mode resonance, which is present at shorter wavelengths than the gap-

modes for the case of spherical gold NP. The largest peak shifts occur for the smallest separations making this effect especially interesting for applications where small spatial changes have to be sensed, such as detection of reactions or conformational changes of molecules with sizes of a few nanometers in total. For comparison, we carried out FEM calculations to analyze the dependency of the peak spectral position on the NP-substrate separation for the case of the pSP excitation and the commonly used propagating plane wave excitation (Supp. Note 2). The magnitude of the incident field was normalized to unity at the position of the nanoparticle. Both types of the excitation exhibit similar high sensitivity of the peak shift on the separation. This implies that the electrical excitation can be combined with gap-mode sensing techniques and its capability of measuring changes of the NP-substrate separation with sub-nanometer resolution. For the design of the system several aspects are important. It is necessary that the LETJ generates sufficiently strong current fluctuations at optical frequencies. This is connected with the profile of the tunnel barrier and can be adjusted by the material and thickness of the heterostructure. The electrons have to tunnel in the FowlerNordheim regime, which can be verified by the measurement of the current-voltage characteristics. Further, in-

ACS Paragon Plus Environment

3

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 11

Figure 2: (A) Sample fabrication steps: a) Cleaning and dry thermal oxidation of the wafer, b) Wet chemical etching of trenches to define the tunneling junctions. c) Deposition of thin-film, high quality isolators as tunneling barrier by plasma enhanced atomic layer deposition. d) Structuring of the working and ground electrode contacts by physical vapor deposition. (B) Fabricated LETJ structure in cross-section by SEM showing the edge of the trench with the junction area and the contact pad. (C) SEM image of the cross-section of the LETJ structure with a deposited spherical gold NP.

tense pSP on the top layer should be generated, requiring that the top electrode supports generation of pSP, which is achieved by the choice of the material and thickness of the electrode. Simultaneously, the electrode has to be sufficiently thin so that the radiation originating in the oxide layer is not completely absorbed; and the electrode has to be sufficiently smooth so that the pSP are not scattered by the surface roughness. Ultimately, the system has to be electrically, optically and mechanically stable over the measurement time, which restricts the parameters such as thickness of the insulator layer, lateral size of the junction, and choice of the metal layers. Based on the design principle of the sensor, we have experimentally investigated different geometries and structural parameters of the LETJ chip (full list of the parameters is given in Supp. Note 3). A layout for microstructured junctions was designed as illustrated in Fig. 2A. As substrate, an n-doped silicon wafer was used. The 6′′ wafer was cleaned and a 190 nm thick barrier of SiO2 was grown by dry thermal oxidation (a). This layer served as leakage current inhibitor and prevented damage of the electrodes by mechanical stress during sample bonding. (b) Definition of the 2 × 2 mm² wide tunneling junctions was done by exposing a covering resist layer and subsequent wet-etching of the undeveloped areas down to the silicon substrate. This step resulted in trenches for each of the different junctions and a separate area for the ground contact of the sample. Atomic layer deposition (ALD) followed to deposit the tunneling barrier whereas the region of the ground-electrode was protected with an additional resist layer. Plasma enhanced ALD as described in the methods section was chosen since it guarantees high quality oxides at the nanoscale (c). After removal of the resist layer that covered the ground-electrode, all

metal electrodes were deposited simultaneously by physical vapor deposition in a final lift-off process(d). The resulting geometry of one tunneling junction is shown in Fig. 2B as cross-sectional SEM image. On each Si-chip of 15x15 mm2, eight independent tunnel junctions with one common electrode were prepared (Supp. Note 3). The 80 nm gold NPs were deposited on the LETJ chip by spotting an aqueous solution of dispersed NP and subsequent drying under ambient laboratory conditions. An exemplary SEM image of the deposited NP on the LETJ structure (in cross-section) is shown in Fig. 2C. To investigate the influence of the different parameters on generation and emission of light, the chips were electrically and optically characterized. The aim was to find an optimal set of parameters for the LETJ structure. First, the optimal thickness of the oxide barrier was investigated. Four LETJ junction types with different alumina thickness (2, 5, 7, 10, and 20 nm) were fabricated on an ndoped Si substrate and 30 nm gold electrode thickness. Subsequently, the electrical properties of the LETJ chips were measured by recording their I-V-characteristics. TheI-V-curves (Fig. 3A) approximately follow the expected profile of a tunneling junction (see Supp. Note 1), which indicates successful fabrication of the tunneling junction necessary for the light emission.The graph shows that the tunneling current decreases substantially with increasing oxide thickness. The current passing through the chips is higher than predicted, which probably originates from parasitic resistances parallel to the junction. Thus, by choosing a certain oxide thickness or material, the electrical properties can be tailored to desired specifications. Moreover, the LETJ emission intensity depends on the tunneling current, so it is important to find an optimal operating point for measurements.

ACS Paragon Plus Environment

4

Page 5 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Figure 3: (A) I-V-characteristics for different tunneling oxide materials and thicknesses in the Fowler-Nordheimregime. With increasing thickness of the tunneling barrier, the current is decreased accordingly. (B) Spectra of the directly emitted light from LETJ chip with a 7 nm thick Al2O3 tunneling barrier for several tunneling currents. Inset: Graph of the integral emitted intensity (400-900 nm) as a function of the tunneling current

Our research goal was the generation of pSP on the surface of the chip, but in the first instance, as a measure for the pSP generation, the amount of the directly emitted light from the LETJ chips was taken. Therefore, all the samples were optically characterized by far-field measurements. No emission was recorded for samples with 2 and 5 nm tunneling barriers, due to immediate breakdown of the junctions caused by electrical shortcuts after biasing, despite efforts to stabilize the measurements with compensating electric circuits. Samples with 7 or 10 nm oxide emitted light continuously according to the current. However, since the 10 and 20 nm samples did not emit sufficiently at reasonable voltages, we decided to focus on the 7 nm junctions, which were long-term stable for operating voltages up to 5 V. Additionally, we investigated the influence of the electrode thickness and material on the light emission. Hence, samples with 30 and 50 nm thick electrodes were fabricated. Both sample types emitted light with similar spectral features, but the measured intensities dropped with increasing electrode thickness, due to the lower coupling efficiency of the dipole moment at the junction boundaries. For this reason, we settled on samples with 30 nm electrode as the best tradeoff between film homogeneity and coupling efficiency. By implementation of different electrode materials, the spectral range of the direct emission could be varied (Supp. Note 4), which is caused by the altered tunneling conditions and the plasma frequency of the various metals. The spectral emission from the LETJ chips with the gold electrode was observed between 400-900 nm. In the other materials the spectral range was shifted to longer wavelengths. Because the expected spectral position of gap-modes is around 600 nm, we restricted our LETJ chip to gold electrodes. From these measurements we concluded that the best parameters for application as nanoruler are an n-doped silicon wafer with 7 nm Al2O3 insulating layer and 30 nm thick gold top electrode with positive bias on the top

electrode. The LETJ chips were biased in the exponential profile of the I-V curve. The spectra of the direct emission for different operating currents are shown in Fig. 3B. The spectra exhibit broad band emission with the spectral range over almost the whole range of the visible light, whereas the emitted light intensity increases with the tunneling current as indicated by the integral emitted light intensity as a function of tunnel current in the inset of Fig. 3B. The curve exhibits a current cut-off value, below which there is no observable light emission. Above this value the total emitted light is increasing non-linearly with the tunneling current.. The lower cut-off of the spectra is shifted to shorter wavelengths with increasing bias as clarified in Supp. Note 5. This spectral behavior agrees with the expectations from LETJ theory and indicates the quantum shot noise based generation of the emitted light. With the following experiments, we establish that the LETJ structures are a valid source of pSP, which are outcoupled by metal NP present on the top electrode. The roughness of the top gold electrode is rather low with RMS = 1.9 nm, which assures that the generated pSP are not significantly scattered by the intrinsic roughness, but mainly by the plasmonic gap-modes induced by the metallic NP (Supp. Note 6). First, direct emission from the LETJ chip without NP was measuredspectrally for several input powers. Afterwards, 80 nm spherical gold NP were adsorbed on the top electrode of the same LETJ chip with a density of around 7·105 particles/mm2. Now, the emission from this NP modified LETJ chip was again spectrally recorded with the identical input power range as in the previous case. The two spectral emission curves for the bare LETJ and the LETJ with NP at the operating voltage of 3.3 V are plotted in Fig. 4A. The NP-modified LETJ structure emits substantially more light than the bare LETJ structure. The improvement in the light emission efficiency due to the adsorbed NP on the surface can be adjusted by their density. The dependence of the inte-

ACS Paragon Plus Environment

5

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 11

Figure 4: (A) Emission efficiency of LETJ structures is enhanced by adsorbed nanoparticles. They serve as additional scattering centers which increase the scattering of pSP into visible radiation. This scattering process is enhanced at the system’s selfresonance, in this case 720 nm. (B) Comparison of the optically emitted power before (red) and after (blue) NP adsorption. An increase of emission efficiency is observed, while electrical power consumption of the LETJ structures stays constant. (C) Darkfield image of emission pattern of nanoparticles imaged with external illumination (bottom) compared with LETJ emission (top). The position of emission hot-spots match well with the location of certain nanoparticles acting as scattering centers as indicated by arrows (the white spot in the lower image probably originates from a dust particle adsorbed between the measurements).

grated emitted power on the power consumption for both LETJ types is plotted in Fig. 4B. It shows that the emission enhancement is consistent over the whole range of operating powers. It proves that the extra generated light is not caused by additional power to the system, but only by an increase of the out coupling efficiency of the generated light. Moreover, the spectral profile of the emitted light is changed as well. To confirm gap-modes as the origin of the enhanced light emission, the unbiased sample was also spectrally measured under external illumination in a dark field setup. In this way, the spectral profile of the gap-modes formed between the added NPs and the gold electrode could be recorded as plotted in Fig. 4A.This spectrum shows a distinct peak around 720 nm, which is also imprinted in the enhanced spectrum of the NPmodified LETJ structure (see Supp. Note 7). This indicates that the extra light comes from coupling of the pSP on the localized plasmon gap-modes between the NP and the gold electrode. To analyze the topographical origin of the light emission, the surface of the NP modified LETJ chip was imaged by a microscope. The image (see Fig. 4C) of the biased junction reveals that most of the light originates from distinct spots on the surface of the chip. As previously, the same area of the unbiased sample was then externally illuminated and imaged, so that the positions of the NPs are located. Correlation of these two images (see arrows in Fig. 4C)showsthat the extra light indeed originates from certain NPs on the surface of the LETJ chip. The contribution of individual particles varies in the emission process which we attribute to the different efficiency of pSP scattering due to the separation induced attenuation of gap-mode coupling.The demonstrated LETJ chips can be utilized as a novel type of sensor for small biomolecules. Since we experimentally verified that the enhanced emission from the NP-modified LETJ structure originates from the gap-modes, its spectral dependence on the NP-substrate separation31,62 can be exploited as a detection principle for biomolecule sensing. In this case, the NP-substrate acts as a physical transduc-

er, translating adsorbed target biomolecule layer thicknesses into respective spectral shifts. Therefore, NPs modified with a recognition layer (specific binding of the target molecules) are incubated with an analyte solution containing target molecules. If the target molecules are present, they bind and thereby form an additional shell around the NP, resulting in an increased thickness. Then, the NPs are immobilized on the LETJ chip and the spectral emission from the LETJ chip is measured. By the analysis of the emission peak position, the separation between NPs and electrode surface can be precisely determined. This value is then related to the presence of the target molecules in the analyte solution. To demonstrate the feasibility of this principle as well as to experimentally confirm the distance dependency of the LETJ emission peak position, three sample types were prepared. Thus, NPs without modification, with a single and with a double layer of molecules were prepared, respectively. The first layer consisted of BSA proteins (thickness ~ 2.5 nm), which easily form a monolayer around the NPs. This layer represents in our example the bio-recognition element. The analyte is represented by antibodies (IgG) specifically binding to BSA (therefore called anti-BSA IgG), with a dimension of about 5 nm. The binding would result in a total layer thickness of ~ 2.5 + 5 = 7 nm. Prior to the NP adsorption on the LETJ chips, the direct emission from these chips was recorded to assure that they are stable and have similar emission characteristics (see Supp. Note 8). Afterwards, each of the NP solutions was adsorbed on a separate LETJ chip and their spectral emission was recorded (similar to Fig. 4). These spectra exhibit, as expected, enhanced emission in comparison to the emission from bare LETJ chips and have the characteristic peak originating from the coupling of the pSP with the gap-modes. But most importantly, the peak positions exhibit blue-shifts of 12 and 13 nm for each step, respectively. As depicted in Fig. 5A, this can be assigned to the enlarged molecular layers around the NPs. This is in agreement to the simulated spectral position

ACS Paragon Plus Environment

6

Page 7 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Figure 5: (A) Emission spectra of equivalent LETJ structures after subsequent adsorption of bovine serum albumin (BSA) and anti-BSA IgG antibody layers on the particle surfaces, respectively. Shift of the spectral center of gravity due to the different number of molecular layers and therefore shell thickness around the nanoparticle. (B) Experiments with external illumination and LETJ measurements are shown in comparison. The sensitivities are analogous which validates the underlying coupling mechanism.

dependence (Fig. 1D). Moreover, the successful verification of the individual adsorptions of both steps is the demonstration of the LETJ chips sensing capability for antibodies, as needed e.g. in the detection of autoimmune or other rheumatic diseases. Moreover, the direct emission generated from the LETJ structures could be used as reference channel, which allows quantification of the adsorbed NP based on the intensity of the measured peaks. To validate the obtained shifts in the gap-mode resonance, reference experiments with identical measurement procedures but with external dark-field illumination instead of the LETJ chip illumination - were conducted. For this, a planar gold film of 100 nm thickness was deposited on a glass slide, and the same three NP solutions (as previously used) were adsorbed on it. As shown in Fig. 5B, the dependency of the peak position (see Supp. Note 8 for peak determination) on the protein shell for the two different systems is similar, which is backing the results of LETJ emission, since far-field excitation and near-field excitation should exhibit the same spectral response for equivalent systems.

Conclusion We have shown that the ultra-compact LETJ structures can be used as a source of surface plasmons. The plasmonscouple efficiently with gap-modes localized between the top metal electrode of the LETJ structure and adsorbed metal nanoparticles and are scattered into free space radiation. The NP-substrate separation sensitivity of the gap-modes can be exploited in the concept of the molecular nanoruler in our LETJ chips. We demonstrated

this novel sensor concept and showed that its sensitivity (12 nm peak shift for a target molecule layer) is sufficient for the detection of small biomolecules (~60 kDa). Advances in engineering of the bio-molecular functionalization of the substrate and the nanoparticles will increase the sensitivity in order to detect low concentration or even conformational changes of molecules placed between the nanoparticles and the top electrode. Therefore, e.g. LETJ-based compact label-free sensors exploiting on conformational changes of the recognition molecules upon attachment of target molecules29,63 could be realized. Although the intensity of the emitted light from the LETJ chips is substantially lower than the light scattered from an external optical source, the advantage is the spatially confined light generation (pSP waves) on top of the LETJ chip. Therefore, the LETJ chips do not require any extra optical component for generation and focusing light as in the case of the external illumination, which makes them ideal for fabrication of compact and integrated sensors. We showed that a single nanoparticle can be detected by employing LETJ chips. This suggests that ultrasensitive, maybe even single molecule detection could be achieved with this techniquein the future. Due to the advances of silicon fabrication technology, the LETJ chips can be easily scaled up to large arrays, which allows multiplex detection scheme. Additionally, due to the set of parameters (e.g. implemented materials, film thickness) that is available for tuning of the LETJ, the systems light emission characteristics are easily adaptable in a wide range. We believe that the presented sensor concept will provide a platform for exploring quantum shot noise based light generation in tunneling junctions and its interaction with the gap-mode plasmons, and will lead to further advances in this field.

ACS Paragon Plus Environment

7

Nano Letters

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 11

layer (see Fig. 3A). The thickness was varied between 2 and 10 nm, and was measured using an ex-situ ellipsometry measurement setup (M-2000, Woollam Co., Lincoln, NE).

Materials and Methods 2D FEM simulation For the evaluation of the interaction between the LETJ surface and adsorbed NPs, two-dimensional finite element method simulations (COMSOL Multiphysics) were conducted. Following the formulation of Dionne et al.64 an air-gold boundary was established that was fed with the analytical expression of a propagating surface plasmon. The whole domain was enclosed by a perfectly matched layer to suppress influences of backscattering effects at the outer domain boundaries. A virtual detector was placed parallel to the system boundary to record the scattered energy over a detection angle of 120°. The scattered energy was calculated by the real part of the time averaged pointing vector anddeconvoluted in its components to match the detectors alignment. Two different cases were then studied. First, an ideal planar boundary without any perturbation source for the pSP mode was set up. Naturally, no scattered intensity should be obtained, but with this approach, the numerical uncertainty of Hz was determined to be