1. Introduction

Bio/chemical sensors are widely used in various application areas such as environmental monitoring, food testing, medical diagnosis, and biopharmaceutical analysis [1–4]. As a sort of bio/chemical sensors, optical sensors have been attracting a great deal of attention due to the salient capabilities of label-free and point-of-care testing [5,6]. In particular, plasmonic resonance-based optical sensors that utilize strong electromagnetic near-field enhancements around metal-dielectric interfaces exhibited large optical sensitivity (S = Δλ/Δn, where Δλ is the wavelength shift and Δn is the change in refractive index of environmental medium) and high figure-of-merit (FOM = S/FWHM, where FWHM is full width at half-maximum) [7–12]. However, in practice, the operation of plasmonic optical sensors often relies on an affiliated instrument (i.e., spectrometer) that is used to analyze spectral responses and to give test results. Motivated by compact and integrated optical sensing, hot-electron optical sensing has been proposed and demonstrated as an advanced optical sensing technology with a unique capability of direct electrical readout [13]. Instead of assessing Δλ to detect the change of surrounding environment that is quantitatively characterized by Δn in plasmonic optical sensing, one can conveniently evaluate Δn by measuring the change in responsivity (ΔR) of hot-electron optical sensors.

Unfortunately, the performances of both plasmonic and hot-electron optical sensors are lack of tunability. In other words, the performances of these devices are almost fixed once they were fabricated. If performance improvements are required, firstly, one has to optimize device configurations, including structure parameters and material properties, to enhance coupling between the device and incident light (for plasmonic optical sensors) and to boost photoelectric conversion efficiency (for hot-electron optical sensors); then according to the optimization results, the optimized devices would be prepared and characterized. The lack of strategies that can be used to tailor the performances of finished devices limits the development of optical sensing technology. In addition, previously reported hot-electron optical sensors used nanostructured metal-semiconductor (M-S) Schottky junctions to excite plasmonic resonances [13,14]. From the optical perspective, the excitation of plasmonic resonance is an advantage because strong confinement of optical fields within a tiny volume with dimensions much smaller than the wavelength of light enhances the interaction between device and environmental medium, results in significant Δλ under small Δn, and creates strong absorption in metallic materials [15]. However, from the electrical perspective, the hot-electron collection efficiency was scarified because the bending, folding, and truncation of metallic materials inevitably decreases of the range of hot-electron diffusion angle (θ) towards the Schottky interface, declining the population of hot electrons arriving at M-S interfaces [16–21]. Indeed, for hot-electron optical sensing with nanostructured M-S junctions, the requirements in optical and electrical domains are often contradictory. The same dilemma also exists in the studies of hot-electron photodetection. Actually, the photoelectric conversion processes of hot-electron photodetection and optical sensing are identical due to the same mechanism of internal photoemission [18].

In this work, inspired by strategies for achieving tunable hot-electron photodetection [22], we provide a design to realize tunable hot-electron optical sensing by applying variable voltage to a planar hot-electron harvesting junction consisting of two gold (Au) layers separated by a molybdenum disulfide (MoS₂) layer. Photoelectric calculations show that bias voltage between two Au layers is a critical factor for the proposed device because the device responsivity is strongly related with bias voltage. Furthermore, the electrical sensitivity increased by 2 folds when bias voltage is 1 V compared to that for the case of zero-bias voltage. Through a detailed comparison, we found the electrical performances of the device with a planar Au-MoS₂-Au junction are better than that with a nanostructured Au-MoS₂-Au junction because the design of planar junction ensures the extraction of photoexcited hot electrons as much as possible. Our design is expected to promote the technology of hot-electron optical sensing.

2. Results and discussions

As schematically shown in the left panel of Fig. 1(a), the proposed device is composed of a one-dimensional grating consisting of periodically arranged silicon nitride (Si₃N₄) stripes and a MoS₂ layer that is sandwiched by two Au layers. The device is mounted on a silica substrate. The cross-sectional view of the structure unit cell is shown in the right panel of Fig. 1(a). The thicknesses of MoS₂, top, and bottom Au layers are 5 nm, 25 nm, and 150 nm, respectively. The height and width of Si₃N₄ stripe are denoted with h and w, respectively. The periodicity of the grating is denoted with p. Under a normal illumination with polarization along the x direction, surface plasmons (SPs) can be launched with the aid of momentum compensation provided by the scattering of Si₃N₄ grating. The excitation of SPs gives rise to environment-sensitive optical responses. The produced optical spectra probably carry information on analyte in which the device is immersed. In the scenario of hot-electron optical sensing, the refractive index of analyte that is denoted with n can be directly obtained by transforming optical spectra into electrical responsivity spectra. This photoelectric conversion process can be qualitatively described by consecutive electronic processes [23–25], as depicted in Fig. 1(b). When SPs are launched, the optical fields of incident light are concentrated into a sub-wavelength region. On the other hand, the lifetime of SPs is extremely short in 10-fs regime, and the nonradiative decay of SPs results in the transition of electrons in the top (bottom) Au layer below the Fermi level E_ft (E_fb) to higher energy levels [22]. The generated energetic carries (hot electrons) with energy (E) exceeding Fermi levels diffuse towards two Au-MoS₂ interfaces, experiencing thermalization loss caused by electron-electron and electron-phonon scatterings. After arriving at the Au-MoS₂ interface, hot electrons in an Au layer with sufficient energy to surmount Au-MoS₂ barrier undergo two interfacial reflection losses and a transport loss in MoS₂ before they are collected by the opposite Au layer that acts as a counter electrode [26]. The net collected hot electrons, i.e., the difference between the collected hot electrons by top and bottom Au layers, contribute to a measurable photocurrent with the aid of external circuit. When bias voltage (V_a) between two Au layers serving as two electrodes is available, one can use V_a to manipulate the hot-electron collection efficiency, and then the responsivity.

 

Fig. 1. (a) Schematic of a three-dimensional (left panel) and a side view (right panel) of the proposed hot-electron optical sensor. (b) Energy band diagram with a bias voltage (e.g., V_a > 0) between two Au layers that results in an uplift of E_ft relative to E_fb, keeping constant the height (Φ_b = 0.5 eV) of Au-MoS₂ barriers. The charge on the electron is denoted with e. (c) Absorption and impedance spectra in the absence of analyte. (d) Spatial profile of a normalized electric filed at the central cross section (i.e., xoy plane) for the resonant wavelength of 1251 nm.

Download Full Size | PDF

To begin, we first performed optical simulations on a finite element platform with reported optical indexes to evaluate the optical properties of SPs excitation in the absence of analyte [27–30]. When p = 1100 nm, h = 100 nm, and w = 600 nm, Fig. 1(c) shows a nearly perfect absorption in two Au layers (A_Au ∼ 0.99) when the wavelength of incident light is 1251 nm. In specific, A_Au is the sum of optical absorption in top (A_top) and bottom (A_bot) Au layers (i.e., A_Au = A_top + A_bot). We found that the absorption in top Au layer at 1251 nm is significantly larger than that in bottom Au layer. It is because that the skin depth of SPs in Au at 1251 nm is less than the thickness of top Au layer [31]. To gain more insights into the optical responses, we calculated the effective impedance spectrum of the device [32]. It is found that the impedance at 1251 nm is very close to 1 (i.e., the impedance of free space), indicating that the reflection of incident light at 1251 nm is negligible. Considering the bottom Au layer is optically thick, the device transmission in the infrared band is also negligible. In consequence, the incident light with wavelength of 1251 nm is almost totally trapped in the device and nearly perfect absorption in Au layers is achieved. Figure 1(d) depicts the spatial distribution of trapped electric field strength (|E_e|²) at 1251 nm. It is found that electric field is mainly confined around the Si₃N₄ stripe, indicating the strong coupling between device and surroundings when SPs are launched. We believe that the resonantly optical responses of our device are highly sensitive with the refractive index of environmental medium.

Therefore, when the proposed device is used as an optical sensor, we investigated the analyte-sensitive optical responses of the device, as shown in Fig. 2(a). It is found that with structure parameters of p = 1000 nm, w = 350 nm, and h = 200 nm, when n increases from 1.33 (the refractive index of water as a reference) to 1.45 (the refractive index of trichloromethane) with a step of 0.03, the dip position of reflection spectrum shifts toward longer wavelength while the magnitude of reflection dip maintains below 0.1. It is known that the physical periodicity of a grating that is the product of geometrical periodicity (p) and n increases with the increase of n. Generally, the resonance wavelengths of a grating-assisted plasmonic system are in positive correlation with physical periodicity. Therefore, the redshift of dip position when n increases can be explained. Particularly, one can use this phenomenon to detect the changes in analyte because the dip position is relevant with n when p is fixed. We also studied the relationships between analyte-dependent dip positions and p. As shown in Fig. 2(b), for the case of w = 350 nm and h = 200 nm, when p increases from 950 nm to 1150 nm with a step of 50 nm, the dip position exhibits a red shift for a fixed n due to the increase of physical periodicity. Moreover, for all five periodicities, dip positions show a fine linear dependence on n and the fitted slope is equal to S. When p = 950, 1000, 1050, 1100, and 1150 nm, S are 700, 783, 857, 933, and 1017 nm/RIU (RIU: refractive index unit), respectively. We found that S increases with the increase of p. However, one can see in Fig. 2(c) that for a fixed n, FWHM decreases with the increase of p. Meanwhile, FWHM exhibits nonlinear dependence on n. Thus, for five periodicities, we calculated average values of five FWHMs for different analytes (i.e., n = 1.33, 1.36, 1.39, 1.42, and 1.45), as shown in Fig. 2(d). We found the mean of FWHM is in a negative correlation with p. Moreover, FOM as a function of p was obtained by dividing S with the mean of FWHM. Overall, one can look forward to better performances in two quantities of S and FOM by increasing p before the device fabrication. However, one can hardly improve device performances once the device was fabricated.

 

Fig. 2. (a) Reflection spectra of the system immersed in trichloromethane-water solution with varying compositions that are characterized by analyte refractive index (n) at TM (transverse magnetic) normal incidence. (b) Resonant wavelength and (c) FWHM versus n for five cases of p = 950, 1000, 1050, 1100, and 1150 nm. (d) Average value of FWHM and calculated FOM as a function of p.

Download Full Size | PDF

After detailed optical evaluations, we turned our attention to the device electrical responses. In order to measure stable and remarkable photocurrent for hot-electron harvesting devices, strong optical absorption in metals is prerequisite [33–35]. As shown in Fig. 3(a), when p = 1000 nm, w = 350 nm, and h = 200 nm, A_Au spectrum exhibits a prominent peak which absorbance is larger than 0.95 and which position is dependent with n. The photoexcited hot electrons in two Au layers can be extracted by Au-MoS₂-Au junction and this photoelectric conversion process can be quantitatively described with a probability-based numerical model in which the power of incident light is set to 1 W [23–26]. In the absence of bias voltage, the fluxes of net collected hot electrons (N_col) as a function of E and light wavelength when n = 1.33 and 1.36 were calculated, as shown in Figs. 3(b) and 3(c), respectively. The dashed lines represent Au-MoS₂ barrier. It is known that only above-barrier hot electrons can be extracted and the massive hot-electron harvesting is on the condition of strong absorption in two Au layers [36]. Therefore, the bright spots shown in Figs. 3(b) and 3(c) that represent efficient hot-electron extraction appear above the dashed lines and around the positions of absorption peaks when n = 1.33 and 1.36, respectively. Based on wavelength-dependent N_col, we obtained responsivity (R) spectra, as depicted in Fig. 3(d). When n increases from 1.33 to 1.36 (i.e., Δn = 0.03), the peak position of responsivity spectrum shifts to longer wavelength. In order to detect the change in analyte, one uses a laser with working wavelength of 1436 nm to measure ΔR (14.94 nA/mW). In terms of hot-electron optical sensing, a quantity of electrical sensitivity (S_PER) is defined to characterize the device performance [13]. S_PER can be written as

 

Fig. 3. (a) Wavelength-dependent optical absorption in two Au layers (A_Au) when n = 1.33 and 1.36. The flux of the net collected hot electrons (N_col) as a function of hot-electron energy (E) and wavelength for two cases of (b) n = 1.33 and (c) n = 1.36. (d) The calculated responsivity (R) spectra with different analyte refractive indexes. The change in responsivity (14.94 nA/mW) at the working wavelength of 1436 nm is also indicated.

Download Full Size | PDF

Thus, S_PER is 498 nA/(mW·RIU) in above-mentioned circumstance. It is suggested that if enhanced S_PER is required, enhanced ΔR at 1436 nm is necessary. We believe that bias voltage between two Au layers can be harnessed to boost hot-electron collection efficiency and then enhanced ΔR would be obtained.

Therefore, we investigated the influences of V_a on N_col at 1436 nm when n = 1.33 and 1.36, as shown in Figs. 4(a) and 4(b), respectively. It is noted that N_col at 1436 nm is significantly enhanced by positive V_a because positive V_a raises the potential difference between top surface of bottom Au layer and top surface of MoS₂ layer, cutting down the fluxes of collected hot electrons from bottom to top Au layers. Meanwhile, positive V_a boosts the probability that a hot electron in top Au layer transmits through MoS₂ layer and be collected by bottom Au layer. In addition, the polarity of N_col at 1436 nm switches if negative bias voltage is sufficiently high. It is because that negative V_a raises the potential difference between bottom surface of top Au layer and bottom surface of MoS₂ layer. As a result, the fluxes of collected hot electrons from top Au layer to bottom Au layer sharply decrease, but at the same time negative V_a enhances the hot-electron collection efficiency from bottom to top Au layers. Thus, the fluxes of collected hot electrons from bottom to top Au layers exceed that from top to bottom Au layers. We considered that the underlying physics of these bias voltage-manipulated electronic processes are unrelated to optical absorption strength. In consequence, the profiles of the curves of R versus V_a at 1436 nm for two cases of n = 1.33 and 1.36 are similar, expect for the difference between magnitudes of R, as shown in Fig. 4(c). Then, when n increases from 1.33 to 1.36, we obtained S_PER that can be tuned by V_a. It is found that when V_a = 1 V, S_PER increased up to 998 nA/(mW·RIU) that is twice as high as that when V_a = 0 V. It is noted that we took no account of the responsivity polarity in the calculations of S_PER, leading to a positive S_PER when large negative V_a switches the polarity. Moreover, Fig. 4(c) exhibits minor enhancements of S_PER for sufficiently high negative V_a. It is because that optical absorption (i.e., hot-electron generation rate) in bottom Au layer is relatively small and large negative V_a enhances hot-electron collection efficiency from bottom to top Au layers. In traditional optical sensing, S is almost independent with n because Δλ is often linearly proportional with Δn, as shown in Fig. 2(b). However, in hot-electron optical sensing, S_PER is relevant with n because ΔR is a function of Δn with a complicated manner in which ΔR is mainly determined by the change in absorption efficiency at the working wavelength that is associated with wavelength shift due to Δn. Specifically, S_PER degrades as the increase of n, as shown in Fig. 4(d). Therefore, we utilized bias voltage-tuned strategy to improve the n-dependent performance. It is shown that S_PER that is in inversely proportional to n can be enhanced by applying positive V_a.

 

Fig. 4. E-dependent N_col tailored by bias voltage (V_a) when (a) n = 1.33 and (b) n = 1.36. (c) R and electrical sensitivity (S_PER) as a function of V_a. (d) S_PER as a function of n enhanced by positive V_a. All calculations were performed at the working wavelength of 1436 nm.

Download Full Size | PDF

For comparison purposes, we also designed a hot-electron optical sensor based on a nanostructured Au-MoS₂-Au junction that is composed of a top grating consisting of periodically arranged Au stripes, intermediate MoS₂ layer, and bottom Au layer, as schematically shown in the inset of Fig. 5(a). For convenience, the bias voltage between Au grating and bottom Au layer is also denoted with V_a. As similar as Si₃N₄ grating, Au grating assists the plasmonic excitation which resonance wavelength is closely related with grating structure parameters. Therefore, we optimized the structure parameters to ensure strong optical absorption in metallic materials at 1436 nm. Optimization results shows when the height and width of Au stripes, and the periodicity of Au grating are 45, 800, and 1043 nm, respectively, distinct optical absorption in Au grating (A_ga ∼ 0.52) and in bottom Au layer (A_ba ∼ 0.45) at 1436 nm can be achieved, as shown in Fig. 5(a). The sum of A_ga and A_ba at 1436 nm is over 0.97. When n increases from 1.33 to 1.36, Fig. 5(b) shows that the positions of three absorption peaks shift to longer wavelengths. Then, we calculated the corresponding responsivity spectra, as shown in Fig. 5(c). It is found that when n = 1.33, the responsivity at 1436 nm is 5.26 nA/mW which is less than that of device with planar Au-MoS₂-Au junction. It can be ascribed to three factors: (1) The net optical absorption (i.e., A_ga − A_ba) at 1436 nm for this nanostructured junction is less than that (i.e., A_top − A_bot) for planar junction, leading to a decreased net hot-electron generation rate at 1436 nm [16]. (2) The use of Au grating reduces the range of θ, declining the populations of hot electrons reaching two Au-MoS₂ interfaces, as shown in the inset of Fig. 5(c) [26]. (3) The height of top Au stripe is larger than the thickness of top Au layer for the case of planar junction, resulting in an increase of hot-electron transport loss [36]. Overall, the photoelectric conversion efficiency of the hot-electron device with Si₃N₄ grating is better than that of device with Au grating. It is generally recognized that the photoelectric conversion efficiency of a hot-electron device is fundamentally determined by hot-electron dynamic losses in the hot-electron harvesting junction that are made up of several parts, including barrier loss, transport loss, and interfacial reflection loss [37]. For our device, we used Au-MoS₂-Au junction in which the barrier of Au-MoS₂ contact is as low as 0.5 eV to reduce barrier loss. Moreover, the adoption of planar junction and bias voltage-manipulation relieve the transport loss and interfacial reflection loss, respectively [38]. When n increases from 1.33 to 1.36, the position of peak responsivity exhibits a red shift. The change in responsivity at 1436 nm is 4.84 nA/mW. Electrical sensitivity is 161.3 nA/(mW·RIU) at V_a = 0 V that is less than that of proposed system with planar junction. Figure 5(d) shows the responsivity can be enhanced by the positive V_a for two cases of n = 1.33 and n = 1.36. Therefore, electrical sensitivity can also be enhanced by positive V_a. But this enhanced performance is lower than that of device with planar junction.

 

Fig. 5. The hot-electron optical sensor with nanostructured Au-MoS₂-Au junction is considered in this figure. The device configuration is inserted in (a). The detailed optical responses, including optical absorption in top Au grating (A_ga), bottom Au layer (A_ba), and the sum of both absorptions when (a) n = 1.33 and (b) n = 1.36. (c) Calculated responsivity spectra for two cases of n = 1.33 and 1.36. The inset of (c) indicates the adoption of Au grating limits the range of hot-electron diffusion angle (θ). (d) Device responsivity and electrical sensitivity as a function of V_a at 1436 nm.

Download Full Size | PDF

The proposed device can be fabricated by the following suggested processes. First, the bottom Au film is deposited by electron-beam evaporator on a quartz substrate that has been ultrasonically cleaned. Then, a continuous 5 nm MoS₂ layer can be prepared by atomic layer deposition (ALD) on the bottom Au layer in which the thickness of MoS₂ layer is precisely controlled by ALD cycles [39]. After the growth of MoS₂ layer, the top Au film is evaporated onto the MoS₂ layer. Next, one-dimensional aligned Si₃N₄ stripes are obtained by electron beam lithography technology. Finally, two electrodes/probes contacting separately the top and bottom Au films are created to enable V_a-dependent photocurrent measurements.

3. Conclusions

To summarize, a novel hot-electron optical sensor with tunable performances is demonstrated. The proposed device is composed of a planar Au-MoS₂-Au junction integrated with a front Si₃N₄ grating that is used to launch SPs and to realize SPs-assisted strong absorption in two Au layers. The light-induced hot electrons in Au layers can be harnessed to detect the change of analyte through measuring the change in responsivity originating from the hot-electron extraction by planar Au-MoS₂-Au junction. By investigating the influences of bias voltage between two Au layers on device performances, we have demonstrated a positive bias voltage significantly boosts electrical sensitivity because the positive bias voltage improves the hot-electron dynamics. Detailed results show that when V_a = 1 V, the electrical sensitivity at the working wavelength increased 2 folds compared to that when bias voltage is absent. By comparing the electrical performances of the device with planar Au-MoS₂-Au junction to that with nanostructured junction, we have shown planar system is preferred due to more efficient hot-electron harvesting. The employments of bias voltage and planar junction may provide a promising route towards compact, highly sensitive, and tunable bio/chemical sensing.