1. Introduction

Surface plasmon polaritons (SPPs) in deep subwavelength structures [1] have gained much interest in various applications including extraordinary optical transmission (EOT) [2], manipulation of cold atoms [3], wavelength filtering [4], plasmonic devices [5], solar cell energy harvesting [6], metamaterials [7], ultrafast on-chip photonic information processing [8], molecular sensing and spectroscopy [9,10]. In particular, the exceptional ability to enhance light-matter interaction by confining light into extremely small volume has been demonstrated and it allows significant photocurrent enhancement in photodetectors by facilitating strong light absorption. Plasmonic enhanced photodetectors have been widely reported in visible and near infrared wavelength range [11]. For middle wavelength infrared (MWIR) range (3-5 µm), plasmonic-grating enhanced nBn photodetectors have been reported at 120 K [12]. For 5-10 µm range, InAs QDs [13,14] integrated with plasmonic structures have been developed for narrow band photodetection, showing a potential for sensitive large FPAs at low temperatures.

In the vicinity of surface plasmon resonance (SPR), the optical field is highly enhanced at the metal-dielectric interface, with its penetration depth into the dielectric characterized by the SPP skin depth. The interaction between the SPP fields and the absorption area results in the increase of light-absorption, thereby enhancing detector photoresponse. This gives us the motivation to electrically control the performance of a plasmonic photodetector through adjusting the interaction volume between SPP and absorption layer. This kind of tuning of performance is different from those tunings based on indium tin oxide (ITO) [15–17], Graphene [18,19], and doped semiconductors [20,21], where usually the plasmonic nature of the materials is directly modulated.

In this work, we propose an InAsSb-based n-i-p photodiode with operating wavelength in the MWIR atmosphere window range (λ = 3-5 µm), where plasmonic nanostructures based on two-dimensional nanohole arrays are integrated on top. The performance of the plasmonic devices is found to be improved compared to the reference ones which have no metal nanostructure patterns. Our experimental results demonstrate maximum electrically controlled enhancement of ~5x and ~6x for room-temperature (293 K) and 77 K operation, respectively. A tuning depth of 380% over 0.32 V change in the voltage bias at room temperature is achieved, giving a tuning factor of 11.9. Much higher tuning factor of 26 is demonstrated as the operating temperature is decreased to 77 K. This electrically controlled enhancement is also found to be in good agreements with the depletion of intrinsic InAsSb layer characterized by the RA product and energy band diagram at different temperature and voltage bias.

2. Design of the plasmonic photodiode

In order to demonstrate this electrically controlled enhancement, we first need to consider a photodetector platform where the active absorption volume can be easily modulated. Photovoltaic devices with external bias and temperature controlled active depletion region can thus be used as a possible platform since the light-absorption mainly takes place in the depletion region. This photodiode is designed to have a broad photocurrent response over the whole mid-infrared wave range with peak value at 3.5 µm. To realize plasmonic resonance in the photodiode, we design gold nanostructure based on two-dimensional array of periodic square holes with geometrical characteristics optimized for electric field enhancement at λ = 3.5 µm. Figure 1(a) shows the schematic of the plasmonic photodiode based on an InAsSb based n-i-p photodiode integrated with gold nanohole array. The n-i-p structure is epitaxially grown on an n-type GaSb substrate and consists of a 1000 nm thick InAs_0.91Sb_0.09 absorption layer sandwiched by 20 nm thick n-type and p-type AlInAsSb layers. The purpose of introducing p-doped AlGaSb and n-doped AlInAsSb layers is for reducing the diffusion dark current. A p-doped AlInAsSb is inserted to limit type II electron-hole transitions. A p-type GaSb buffer layer is used as the bottom contact layer. The purpose of having only 20 nm thick top n-type layer is to give access to SPPs to fully interact with the active InAsSb layer. The top and bottom metal contacts are 15 nm thick titanium (Ti) followed by 200 nm thick gold (Au). The 300-µm sized square mesas are defined by wet etching, followed by deposition of SiO₂ passivation layers (PECVD) to reduce the surface leakage current and protect the mesas. Detailed growth and fabrication process have been described elsewhere [22]. For the gold nanohole array integrated on top of this photodiode, the surface plasmon resonance (SPR) wavelengths are given by [23]:

 

Fig. 1 (a) Schematic of the plasmonic InAsSb-based n-i-p photodiode. Transmittance mapping as functions of (b) gold thickness, (c) incident light angle, and (d) hole width at different wavelength with specific constant parameters.

Download Full Size | PDF

To obtain full parameters for the plasmonic nanostructure, finite difference time domain simulations were performed to optimize the design. As expected from Eq. (1), the nanohole array exhibits extraordinary optical transmission (EOT) [2] around SPR wavelengths [Figs. 1(b) and 1(c)]. For the nanohole array with 900 nm periodicity, we present the mapping of transmittance spectra as functions of gold thickness (w = 450 nm, Angle = 0) [Fig. 1(b)] polarization angle (w = 450 nm, h = 70 nm) [Fig. 1(c)], and hole width (Angle = 0, h = 70 nm) [Fig. 1(c)]. We found that the nanohole array with a gold thickness of h ~70 nm and hole width of w ~450 nm allows the best performance (highest transmittance) [26], indicating best confining for the incident radiations at the optimal SPR wavelengths. The nanohole array also exhibit no polarization-dependence due to its square lattice arrangement, which is desirable for our application. Figure 2 presents the distributions of |E|-field at λ = 3.5 µm in XY and XZ planes for nanohole array with different periodicities, showing that the largest field enhancement occurs at p = 900 nm. The SPP penetration depth inside the absorption layer can be expressed as [27]

 

Fig. 2 Normalized |E|-field distribution of nanohole arrays with different hole periods at 3.5 µm. In all simulations, the width of the hole is designed as half of the period.

Download Full Size | PDF

The fabrication of the gold nanohole array on the n-i-p photodiode was carried out by electron beam lithography (EBL), followed by metal physical deposition and standard lift-off pattern transfer. The gold thickness is chosen according to our optimization in Fig. 1(b). Prior to 70 nm thick gold deposition, 3 nm thick titanium was first deposited as adhesion layer. The gold thickness is also much larger than the skin depth [28], where direct light transmission through the gold film can be avoided. The skin depth in gold is defined as ${\delta }_s=\sqrt{\rho /(\pi f\mu )}$, where $\rho$ is the resistivity, f is the frequency, and $\mu$ is the relative permeability, and is found to be in the range of ${\delta }_s=4.3-10.7$ nm at λ = 1-6 µm.

In experiments, we first fabricated top metal nanohole arrays with different periods on the InAsSb based n-i-p sample to demonstrate SPRs on it. We present in Fig. 3 the measured and calculated reflectance characteristics of the gold nanohole arrays as the periodicity is changed from p = 550 nm to p = 1550 nm. The reflection spectra of InAsSb n-i-p samples were measured by Fourier Transform Infrared Spectroscopy (FTIR) integrated by 36x objective lens (NA = 0.5), where the reflection spectra were normalized with that of the gold pad on the same chip. As evident from the spectral dips in Fig. 3(a), the reflectance spectra of all nanohole arrays exhibit clear SPRs, which are consistent with the simulated transmittance spectra shown in Figs. 1(b) and 1(c). Specifically, for the nanohole array with p = 900 nm, the fundamental resonance position at ~3.5 µm can be observed, which is in a good overlap with the photocurrent peak of the InAsSb based n-i-p photodetector. The measured results show slight red shift and increased broadness of resonance, which can be attributed to the presence of the scattering losses associated with nanohole roughness and the off-normal incidence angles from the condenser in the experimental set-up [29,30]. It is also noted here that the measured reflectance are smaller than their simulated counterparts. This is mainly due to the nonideal focus in the measurements and the ignored imaginary part of the permittivity of semiconductor in simulation.

 

Fig. 3 Measured (blue lines) and simulated (black lines) reflectance spectra for gold nanohole arrays with different periods on n-i-p sample.

Download Full Size | PDF

3. Electrically-controlled enhancement of the plasmonic photodiode

Prior to evaluating the electrically-controlled enhancement, we first characterized the photocurrent spectra of plasmonic photodiodes with different periodicity under photovoltaic mode (zero bias). The photocurrent spectra were characterized by the same FTIR system, where the MCT photodetector was replaced by our n-i-p detectors. As presented in Fig. 4(a), all the photodiodes with nanohole arrays show obvious photocurrent enhancements over 2-5 µm wavelength range, in comparison to the reference photodiode (black curve). For device with p = 900 nm, over 2-5 µm response range (for room-temperature operation), the average photocurrent enhancement factor, which defined as the ration between the photocurrent of the plasmonic device I and that of the reference I_ref, is found as I/I_ref = 4. For the devices with p ≠ 900 nm, on the other hand, the enhancement factors are only in the range of 1-2. This qualitative difference illustrates the plasmonic nature of the photocurrent enhancement as p = 900 nm is the optimized parameter for the |E|-field enhancement at λ = 3.5 µm (see Fig. 2). It is noted here that plasmonic n-i-p devices with p ≠ 900 nm still show photocurrent enhancement and the plasmonic device with p = 900 nm does not show obvious SPR characteristics (especially for high modes) as presented in Fig. 3. These indicate there are also other mechanisms that contribute to the photocurrent enhancement even though their contributions are small compared to SPP effect. One of the contribution is from the isolated metal holes. They can also offer electric field enhancement by coupling free space radiation into localized surface plasmons (LSP) [31]. The other one is the high SPR modes originated from the plasmonic nanohole array structures as demonstrated in Fig. 3. They can also offer enhancement for the plasmonic electric field in InAsSb. In addition to the LSP and higher order plasmon modes, Fabry-Perot (FP) resonance [32] could be another cause for the broadband enhancement. This is possible as the integrated metallic nanostructure enhances the reflection for thoes unabsorpted radiations which move towards it after being reflected back from the substrate (the refractive index is a little larger than that of InAsSb), and also the thickness of the intrinsic layer is comparable to the operational wavelength. Besides, the scattering of the light by the metallic structure would increase the interaction length between incident radiation and active absorptional materials. This will also facilitate the generation of photocarriers. Finally, the loss [33] from the gold and scattering from the metal array [31] can also contribute to the broad enhancement spectra. We have discussed all of the contributions elsewhere [22].

 

Fig. 4 (a) Photocurrent spectra of plasmonic n-i-p photodiodes with different hole periods under zero bias. The top panel is a typical SEM image of the plasmonic device with p = 900 nm and the scale bar represents 100 µm. Photocurrent mapping of the devices at different temperatures and bias situations (b) T = 293 K, reference device, (c) T = 293 K, plasmonic devices, (d) T = 77 K, reference device, (e) T = 77 K, plasmonic devices.

Download Full Size | PDF

In the following, we focus our attention to p = 900 nm devices for further investigating the electrically-controlled enhancements. Figures 4(b)-4(e) present the photocurrent mapping of the plasmonic and reference devices at different bias under room-temperature and 77 K operations. The reference devices show photocurrent peaks at around 3.5 µm at both temperatures, which keeps good overlap with the optimal SPP wavelength. At 77 K, the photodiode exhibits a shorter cut-off wavelength at ~4 µm due to bandgap widening of InAsSb layers.

To characterize the electrically-controlled enhancements, we derived the averaged enhancement factors (I /I_ref, over 2 µm to the cutoff wavelengths at different temperatures) of the plasmonic photodiode under different voltage biases for 293 K and 77 K operations. As shown in Fig. 5(a), the enhancement factors reach maxima at −150 mV bias at room temperature (~5x) and at 50 mV bias at 77 K temperature (~6x). We measured the maximum tuning depth (defined as the enhanced percentage) of 380% over 0.32 V change (−0.15 - ~0.17 V, this voltage range is defined as the voltage difference between minimum and maximum enhanced factors) in voltage bias for room temperature operation, giving tuning factor (defined as the ratio between the maximum tuning depth and the voltage range) of 11.9. Similarly, for 77 K operation, we measured tuning depth of 520% over 0.2 V change (−0.15 - ~0.05 V) in the voltage bias, giving tuning factor of 26.

 

Fig. 5 (a) Electrically controlled enhancement of the plasmonic device at 293 K and 77 K. (b) RA product of the plasmonic device at 293 K and 77 K under different applied biases. Schematics of the energy band diagram of the plasmonic photodiode at (c) 293 K, zero bias, (d) 293 K, reverse bias, and (e) 77 K, slight forward bias.

Download Full Size | PDF

To have a deep insight on the electrically-controlled enhancement, we measured the current-voltage characteristics of the plasmonic devices and derived its RA product (defined as the dynamic resistance multiplied by cross-section area of the photodiode) at both room temperature and 77 K. The RA value is directly related to the depletion and energy band diagram and usually, the more depletion the structure has, the larger the RA product. As shown in Fig. 5(b), the RA product shows consistent trend as the enhancement factor with respect to voltage bias in Fig. 5(a), where the largest RA product occurs at ~-150 mV for 293 K and ~50 mV for 77 K. This suggests that the photocurrent enhancement at different bias voltages is directly related to the depletion region and band structure of the n-i-p photodiode.

First, as the light absorption mainly takes place in the depletion layer, the overlap between the SPPs and the depletion layer would determine the enhancement factor. Since the SPP penetration depth is about 1000 nm (at λ₀ = 3.5 μm), which is the same as the thickness of the whole intrinsic layer, the enhancement strongly depends on the width of the depletion layer W, which is usually a function of bias voltage V, i.e., $W\propto \sqrt{V_{\text{bi}}-V}$ (V_bi is the built-in voltage). Therefore, increasing the forward bias (V is positive) leads to the decrease of the SPP overlap with depletion. It is translated to smaller absorption, and hence smaller enhancement. This can explain the non-enhancement of photocurrent at relative larger forward voltage bias for both room-temperature and 77 K operations.

For the n-i-p photodetector, the whole potential drop will ideally be confined to the i zone along which the electric field is constant. However, in our case, the intrinsic carrier density is high at high temperature and the i zone cannot be fully depleted. This can be explained by Debye length, $L_{\text{D}}=\sqrt{\epsilon {\epsilon }_0{k}_{B}T/(q^{2}N)}$ [34], where ε is the dielectric constant, ${\epsilon }_0$ is the permittivity in vacuum, k_B is the Boltzmann’s constant, T is the absolute temperature in kelvins, q is the elementary charge, and N is the intrinsic carrier density (~5 × 10¹⁶ cm⁻³). For such a high carrier concentration in our case, the Debye length (~20 nm) is found to be much shorter than the thickness of the i zone (1000 nm). This will result in a region of flat band in i zone at room temperature under zero bias (Fig. 5(c)). With a reverse bias added to the structure, the i zone becomes fully depleted (Fig. 5(d)). The overlap between SPP and depletion will become maximum, leading to largest enhancement factor as reflected in Fig. 5(a). As the Debye length and V_bi increase at 77 K [35] (Fig. 5(e)) [29], the i zone will be fully depleted at 77 K even under a slightly forward bias (Fig. 5(e)), leading to maximum overlap between SPP filed and absorption layer. This can explain the maximum enhanced factor at around 50 mV at 77 K.

When the structure is fully depleted (around the points for maximum enhanced factor), the velocity of carriers (mainly for electrons in our case) is very close to its saturated velocity. Further increase of voltage bias over the points cannot further increase the drift velocity [36]. On contrary, it will increase the thermal oscillations of lattices (phonons), resulting in more scattering for those photocarriers and therefore lower enhancement. This is reflected on the decreased enhancement factor on the left side of the maximum points at both room temperature and 77 K in Fig. 5(a).

Finally, we make an evaluation for detecting performance of the plasmonic device. A 700 °C blackbody radiation source was used to characterize the detectivity of the plasmonic n-i-p photodetector with p = 900 nm. The thermal-noise limited detectivity $D=R_i/\sqrt{2qJ+4kT/(RA)}$(q is the electronic charge, J is the dark current density, R is the dynamic resistance, A is the area, and R_i is the photocurrent responsivity which can be expressed as R_i = I_s/P, I_s is the signal current, P is incident radiation power on the detector which can be calibrated by a standard power meter (OPHIR PHOTONICs)) of the plasmonic n-i-p photodetector is 0.8 × 10¹⁰ Jones at −150 mV for room-temperature operation. It can be increased to 5 × 10¹¹ Jones at 50 mV at 77 K. These detectivities are very good compared to recently reported middle infrared photodetectors as well as commercially available products [37–39]

4. Conclusions

We have proposed and demonstrated a plasmonic photodetector based on the interaction between SPP and the absorpion layer in InAsSb-based mid-infrared n-i-p photodiode. By matching the intrinsic layer of the n-i-p photodiode with the SPP penetration depth, we have demonstrated electrically-controlled plasmonic photocurrent enhancement through controlling the overlap between the SPP and the depletion layer by means of applying external bias. Maximum enhancement factors of ~5x and ~6x have been achieved for room-temperature (293 K) and 77 K operation, respectively, corresponding to electrical tuning factors of 11.9 and 26. The maximum detectivities obtained at the two temperatures are 0.8 × 10¹⁰ Jones and 5 × 10¹¹ Jones, respectively, which are good among middle infrared photodetectors.