Photon-trapping array for enhanced midwave infrared photoresponse

Incorporation of photonic structures is widely regarded as an effective way to realize high-performance photodetection as the raised light–matter interaction inside detection layer can facilitate light absorption without sacrificing the response speed [1–4]. Integrating the plasmonic structure is one of the most popular strategies, since at the metal–dielectric interface, the optical field can be enhanced by the surface plasmon resonance [5–9], thus resulting in the increase of absorption. Another approach to improve the photoresponse is to utilize Fabry–Pérot or other resonant structure surrounding absorption layer [10]. While it does enhance quantum efficiency, the wavelength selectivity of cavity resonance behavior requires the perfect match between the cavity resonance wavelength and bandgap of device absorption layer.

An embedded photon-trapping (PT) structure is considered to be a good choice for better light manipulation. With the help of PT structure, the combined effects of the resulted lateral propagating modes and reduced effective refractive index can bring various advantages, such as polarization-independent broadband absorption enhancement and reduced bulk dark current. Recently, PT structure has been proved useful to improve optical performance in visible wavelengths for solar cell, optical communication links [11, 12], as well in near infrared regime for single-photon avalanche detectors [13].

The atmosphere midwave infrared (MIR) range (2–5 µm) is of great significance for security, medical diagnosis and military applications [14, 15]. The drive towards high-sensitivity, high-operating-temperature and robust MIR photodetection has stimulated considerable research efforts, which mainly lies in adopting external plasmonic or other resonant structures. In this work, we design an InAsSb-based hetero-n–i–p photodiode with an etched PT hole-array structure to improve absorption efficiency, thus obtaining a MIR photodetector with enhanced detection performance. Using a set of simulated optimal geometric parameters to fabricate, a 38% absorption enhancement is obtained at 3 µm. As such, a broadband photoresponse improvement of 26% is achieved in the wavelength range of 2–5 μm. Besides, dark current density is also reduced due to decreased bulk volume by the etched PT structure. In this way, a zero-bias room-temperature detectivity enhancement of 29% higher than the reference detector without PT structure is obtained, where the enhanced detectivity increases to 2.09 × 10⁹ Jones. When the temperature drops to thermoelectric cooling (TEC) temperatures of 260 K and 243 K, the detectivity enhancement can even reach to more than two- and three-times. A response speed as fast as that of the reference is also observed. To this end, the etched PT structure does realize the polarization-independently broadband photodetection improvement for robust applications.

2.1. Device design

The schematic diagram of the hetero n–i–p photodetector integrated with PT hole array is shown in figure 1(a), where the cylindrical hole array is in hexagonal lattice. The targeted device comprises a InAsSb-based heterojunction n–i–p architecture grown on GaSb substrate, where the absorption i-layer is composed of InAs_0.912Sb_0.088 with a thickness of 1.5 μm. In a conventional homojunction photodiode, large dark current is always the major obstacle for obtaining good performance. To improve on that, the basic device not only adopts the heterojunction architecture, but also includes a highly doped wide-bandgap p-AlGaSb layer between the GaSb contact layer and InAsSb absorption i-layer to prevent the minority carrier (electron) diffusion. Such high doping also facilitates high electron concentration without inducing obvious bandgap narrowing. The lattice-matched layers with different bandgaps are grown by RIBER-32 molecular beam epitaxy system, where the specific layer parameters are indicated in table 1. High resolution x-ray diffraction and photoluminescence spectra indicate high-quality interfaces and absorption layer [16]. In this way, the basic heterojunction photodetector obtains relatively low dark current and high sensitivity for MIR detection.

Zoom In Zoom Out Reset image size

Table 1. Layer parameters of the hetero-n–i–p photodetector.

Material	Composition	Thickness	Doping
GaSb	—	50 nm	ND ∼ 2 × 1018 cm−3
Alx In1−x Asy Sb1−y	x = 0.15; y = 0.009	20 nm	ND ∼ 2 × 1018 cm−3
InAsy Sb1−y	y = 0.912	1500 nm	Intrinsic
Alx In1−x Asy Sb1−y	x = 0.2; y = 0.912	20 nm	NA ∼ 2 × 1018 cm−3
Alx Ga1−x Sb	x = 0.42	40 nm	NA ∼ 2 × 1018 cm−3
Alx Ga1−x Sb	x from 0.42 to 0.1	30 nm	NA ∼ 2 × 1018 cm−3
Buffer GaSb (Be)	—	1000 nm	NA ∼ 2 × 1018 cm−3
Substrate GaSb (Te)	—	600 μm	ND ∼ 7 × 1017 cm−3

We then manage to integrate the hetero photodetector with the PT hole-array structure, where microscale holes are etched through the top contact layer to reach the absorber, for promoting absorption efficiency and thus increasing the photoresponse capability. The corresponding cross-sectional diagram is shown in figure 1(b). To quantify the performance enhancement of the PT structure, we regard the detector without PT structure as the reference (ref) detector and the device integrated with the PT structure as 'PT detector', both with the same absorption area.

2.2. Simulations

Finite-difference time-domain method is used to analyze the light confinement capability and wave propagation condition in PT detector. It is turned to calculate the absorption of patterned photodetector through transmittance and reflectance, as well as the electric field distribution in the structure. The detailed simulation setting is described in appendix A. Firstly, to investigate the maximum light confinement capability of PT structure, geometrical parameters of PT structure (hole array period p, ratio of hole diameter to period (r_d), defined by d/p, and etched hole depth D) are scanned to obtain an optimal set of parameters that can achieve the maximum absorption spectra in the range of 2–5 μm. As shown in appendix A, for specific parameter optimization results (see figure 7), the corresponding optimal parameters are as following: p= 3 μm; r_d= 0.67; D= 1 μm. Besides, the polarization-independent characteristic of PT structure is also confirmed by the simulation result shown in appendix A (see figure 7(c)).

Figure 2(a) illustrates the simulated absorption spectra of the PT detector in its optimal geometrical parameters, and the ref detector. It is observed that the absorption is obviously enhanced by integrating the PT structure, where the maximum enhancement rate of 55% is obtained at wavelength of 3 μm. The integrated absorption value, as an index to quantify the optical performance enhancement, is defined as the ratio of absorbed optical power to incident power over the wavelength range of 2–5 μm [17]. The simulation results indicate a maximal integrated absorption value of 56% for the PT detector. Compared with the 40% for the ref detector, a significant 38% absorption improvement is achieved. Meanwhile, the absorption value is quite robust with the change of geometrical parameters as a 5% variation in geometrical parameters only results in about 1% change in the integrated absorption. This characteristic is helpful in facilitating large-scale production. As a confirmation, we also performed simulation on the absorption by integrating per unit volume, as indicated in figure 9. The results obtained from integration (P_{abs_tota}) are almost the same as those derived from transmission and reflection.

Zoom In Zoom Out Reset image size

To explore the light trapping mechanism in the PT structure, the simulated reflection spectra are presented in figure 2(b). It is observed that the reflection of PT detector is obviously lower than that of the ref detector, while the simulated transmission spectra are almost same (figure 2(c)). In this way, the reduction of reflection can almost directly reflect the absorption increase in the PT detector. The integrated reflection value between 2 and 5 μm for the PT detector is only 15%, much smaller than the 34% of ref detector. Thus, one of the roles of the PT-structure is to reduce reflection due to the reduced volume filling factor (ff = volume of air/volume of material) which is described as:

$$\begin{equation}{\text{ff}} = { }\frac{{{ }{d^2}}}{{2{p^2}{\text{tan}}\left( {60^\circ } \right)}}.\end{equation} \tag{ 1 }$$

The corresponding effective refractive index (n_eff) of the PT hole array is thus reduced which is expressed in the following formula [18]:

$$\begin{equation}{n_{{\text{eff}}}} = \sqrt {{\varepsilon _{\text{b}}}\frac{{{\varepsilon _{\text{i}}}\left( {1 + {\text{ff}}} \right) + {\varepsilon _{\text{b}}}\left( {1 - {\text{ff}}} \right)}}{{{\varepsilon _{\text{i}}}\left( {1 - {\text{ff}}} \right) + {\varepsilon _{\text{b}}}\left( {1 + {\text{ff}}} \right)}}} { }\end{equation} \tag{ 2 }$$

where _i and _b are the dielectric constants of air and background material (hetero InAsSb-based n–i–p, ∼15 [19]), respectively. As a result, the reduced refractive indices will lead to a reduced reflection, as derived from the equation (3):

$$\begin{equation}R = \frac{{{{\left( {{n_{\text{i}}} - {n_{{\text{eff}}}}} \right)}^2}}}{{{{\left( {{n_{\text{i}}} + {n_{{\text{eff}}}}} \right)}^2}}}{{\boldsymbol{\,}}}{{\boldsymbol{.}}}\end{equation} \tag{ 3 }$$

By taking the optimal parameters (p = 3 µm, d = 0.67p) into the mathematic model, the calculated R is 21% with a n_eff of 2.67, which is consistent with the value of wavelengths below 2.5 µm and above 4.5 µm for PT detector. While the calculated R of the ref detector is 34% with a n_eff of √_b equal to 3.88, which corresponds to that of wavelengths from 2 to 5 µm for ref detector. Also, it is remarkable to note the effect of reflection reduction is more significant at wavelengths between 3 and 4 µm, where the reflection value is around 10%.

Figure 2(d) shows the E_x component of electric field distribution changing with time in the PT detector under a normal plane wave illumination. To better observe the light propagation, the unit cell in simulation contains eight holes, whose area is expressed as: 2p × 2p × tan60°, since the separate distance between two neighboring holes in y-direction of the hexagonal lattice is ptan60°. The transient evolution of E_x distribution from t = 1.0 to 11.1 fs is depicted in the figure. The results show that the waves in lateral direction gradually appear around holes and form the collective lateral propagating modes. In order to clearly present the bending of light propagation direction, Poynting vector distribution is shown in appendix A (see figure 8), where the vector arrows are bent to lateral direction around the PT holes. This characteristic is almost the same at wavelengths between 3 and 4 μm, while it never appears in ref detector without the PT structure, suggesting that the laterally propagating modes are indeed generated by the PT hole array. These modes simultaneously move in downward and horizontal direction, indicating that the PT structure can support a set of modes with wavevectors in z direction (k_z ) and in-plane direction (k_|| ). For the modes with large k_|| , they are lateral propagating modes [20]. Since the photodiode dimension in x–y plane is much larger than the thickness of i-region, it leads to longer optical propagating path and longer dwell time before light exiting the absorption layer. In contrast to light propagating in z-direction only, the lateral propagating modes can result in more light absorption [21].

Besides, it is reported when the period of PT hole array is set close to the wavelength of the absorbed light, the group velocity of the modes can be treated as slow modes in photonic crystal, leading to more chances for stimulating laterally propagating modes [22, 23]. It thus explains why we can achieve an obviously enhanced absorption at wavelengths around 3 µm with the period of PT structure set to 3 µm [16]. In this way, the incident light propagating in PT hole arrays can obtain a large k_|| which makes light interaction in absorption layer more efficient [24].

Therefore, in the case of hetero-n–i–p detector assisted by the PT structure, the absorption enhancement can be attributed to the combination effects of lateral propagating modes around the resonant wavelengths and reduced reflection due to lower effective refractive index of the PT structure. In addition, the etched hole array also reduces the physical volume of the device. It can help to limit the bulk dark current, which is the main obstacle for MIR detector operating at room temperature [25]. As such, the decreased bulk dark current can balance or even overwhelm the increased surface leakage by extra exposed areas induced from the etched PT structure.

The PT hole array in its simulated optimal geometrical parameters was fabricated on the InAsSb-based heterojunction n–i–p photodiode. The corresponding SEM profile is shown in figure 3(a). The absorption spectra are then obtained by measuring the transmission and reflection via Fourier transform infrared (FTIR) spectroscopy, as shown in figure 3(b). Compared to ref detector, it is demonstrated that a 38% absorption enhancement is obtained at 3 µm in PT detector. Due to the continually distributed incident angles which are up to about 30° mainly from the condenser in the FTIR experimental set-up (36× objective lens (NA = 0.5)), the measured reflection spectra demonstrate broadband feature with a wider peak, compared with the simulation results under normal incident light [26]. Meanwhile, the absorption of ref and PT detectors at different polarization angles were measured (figure 3(c)) and the results show the polarization-independent property of the PT structure, which is in consistent with the simulation result shown in appendix A (see figure 7(c)).

Zoom In Zoom Out Reset image size

For device fabrication, the corresponding schematic diagrams of detailed fabrication process and SEM image of the fabricated device are shown in appendix B (see figure 10) and figure 4(a), respectively. The PT structure has covered about 70% of the light absorption area. The photocurrent spectra of the devices were measured at different temperatures and bias voltages. More details about the performance characterization are shown in appendix C. To gain clear insights into the enhancement from the PT structure, the enhancement factor (η) is defined as the ratio of measured values of PT detector to those of ref detector. As such, the η of photocurrent (I_PH) spectra, measured using a FTIR spectrometer, is defined as η = I_{PH, PT}/I_{PH, Ref}. Figures 4(b) and (c) show the η of the PT detector measured at different biases and temperatures and the measured I_PH spectra for ref and PT detectors are shown in appendix C (see figure 11). The η values under zero bias at 293 K and 260 K, extracted from the black dash lines in figures 4(b) and (c) are shown in figures 4(d) and (e) respectively. The results indicate that the photoresponse enhancement becomes stronger with the decrease of temperature. To further quantify the enhancement at specific bias and temperature, we calculated the average value of the η integral from 2 to 5 μm at room temperature for different bias and under zero bias for different temperatures, as shown in figures 4(f) and (g), respectively. A 26% photocurrent enhancement is obtained under zero bias at room temperature, and the enhancement can reach to about 40% under the bias of −0.05 V. Meanwhile, it is observed that the zero-bias enhancement would be increased to 170% with the temperature decreasing to 243 K and become stable below 243 K. The observation will be explained in the following section. Besides, the polarization-independent character of the PT assisted photoresponse, is confirmed by the I_PH at 3 μm of two detectors measured under different polarization angles, as shown in appendix C (see figure 11(e)).

Zoom In Zoom Out Reset image size

We used a 1000 K blackbody radiation source to measure the detector responsivity in the mid-infrared range. The responsivity is defined as R_A = I_s/P_inc, where P_inc is the incident power radiated on detector and I_s is the signal current generated by fabricated detectors. Appendix C (see figures 11(b) and 12(a)) shows the change of R_A under the bias from −0.3 V to 0.1 V at temperatures from 78 K to 293 K for the ref and PT detectors, respectively. The obvious enhancement is observed at different temperatures and voltage biases. The extracted results at 293 K and 0 V are separately presented in figures 5(a) and (b), respectively. It is noted that a maximum room-temperature responsivity is obtained at −0.05 V for both the ref (0.39 A W⁻¹) and PT (0.48 A W⁻¹) detectors mainly due to the increased electric field. For the bias more negative than −0.05 V, the saturation effect suppresses the carrier velocity and increases the thermal recombination which lead to a decrease in responsivity [27]. Also, the response enhancement is not apparent at positive bias due to narrow depletion region and reduced electric field. As such, a 40% room-temperature response enhancement is obtained under the reverse bias of −0.05 V, which is in consistent with the measured result of photocurrent spectra. In addition, figure 5(b) shows an increased enhancement factor in responsivity with the decrease of temperature. The better improvement in the responsivity of the PT device than that of the reference at low temperature may be ascribed to three factors. One of them is the increased absorption efficiency as more photons can reach the PT structure due to the smaller effective refractive index and they travel longer path due to the laterally propagating modes in the PT device [18, 21, 28–30]. Another possible factor is the suppressed surface recombination within the holes at low temperature, which helps the improvement in the absorption [31]. Enhanced responsivity by cooling was also reported by others [32, 33]. The other factor is the increased collection efficiency in the PT structure as more photons will be absorbed at/near the bottom of the hole array and the photogenerated carriers can be collected faster than in the [34, 35]. In this way, the PT detector which owns more absorbed photons would expect a larger enhancement factor as observed. In the temperature region <243 K, the response is saturated due to the limited light absorption and almost unchanged carrier density. The maximum enhancement of the zero-bias responsivity at temperatures below 240 K is about 170%.

Zoom In Zoom Out Reset image size

For the noise (v_n) characteristic shown in appendix C (see figures 11(d) and 12(c)), it considers dark current shot noise, thermal noise and background noise as expressed in v_n = (2q(J_d + J_b)+ 4kT/(RA))^1/2, where J_d and J_b are the dark current density and background noise current density, R is dynamic resistance and A is the mesa area [35]. It is noted that the noise near the zero-bias region is obviously reduced. Figures 5(c) and (d) shows the extracted noise data at room temperature under different biases and under zero bias at different temperatures of the PT and ref devices, respectively. As shown in figure 5(c), a similar but slightly lower noise is observed for PT detector at room temperature. It means that the reduced bulk dark current makes up for the potentially increased surface leakage, which also confirms the effectiveness of surface passivation we adopted in fabrication. The relatively smaller J_d for PT detector, as indicated in figure 5(e), also confirms the role. It is noted that with the decrease of temperature (figure 5(d)), the noise of the PT detector owns a more significant reducing trend than that of the ref detector. It is mainly attributed to the stronger temperature dependence of the diffusion dark current, compared with the Shockley–Read–Hall recombination current [36, 37]. From that, it can be derived that the diffusion current with stronger temperature dependence is dominant in a wider temperature range for PT detector. Thus, the noise reduction with the decrease of temperature is faster in the PT detector. The detailed detectivity (D*) data of the ref and PT detectors derived from R_A/noise are shown in appendix C (see figures 11(f) and 12(e)), and figures 5(f) and (g) present the extracted data at 293 K and 0 V for the two devices, respectively. The room-temperature D* of the PT detector is about 2.09 × 10⁹ Jones under zero-bias, which is 29% higher than that of ref detector (1.62 × 10⁹ Jones). When temperature decreases, the detectivity enhancement becomes more pronounced. At TEC temperatures of 260 K and 243 K, which are practically achievable in low-cost cooling technology, the detectivities of the PT detector are increased to 5.75 × 10⁹ Jones and 7.97 × 10⁹ Jones, respectively. These detectivity values correspond to about two and three times of those of the ref detector. In addition, it is remarkable that such high performances are all achieved under zero bias, indicating that the PT detector is truly suitable for the low-power-consumption and high-efficiency detection.

Finally, the response waveform in reaction to a 2.94 μm laser source is shown in figure 6(a), where the corresponding rise and fall time are around 1.4 μs and 1.2 μs for ref and PT detectors, respectively. It is relatively lower than the current high-speed detection, which may be originated from the different detection mechanism and RC time delay [38]. Figure 6(b) shows the amplitude–frequency response (f_−3dB) values which are 2.4 × 10⁵ Hz and 3 × 10⁵ Hz for the ref and PT devices, respectively. The slightly better frequency response is mainly due to the faster collection of the photogenerated carriers as the physical holes shorten the transportation path for the carriers generated after absorbing the photons reaching the bottom of the holes.

Zoom In Zoom Out Reset image size

This article presents the design and fabrication of PT detector which is an InAsSb-based hetero-n–i–p MIR photodetector with etched hole array structure. With assistance of this polarization-independent PT structure, the resulted lateral propagating modes and reduced effective refractive index can enhance light absorption, leading to a broadband absorption improvement of 43% in the wavelength range from 2 to 5 μm. The PT hole array also reduces bulk dark current. A room-temperature detectivity of 2.09 × 10⁹ Jones is obtained under zero bias, which is 29% higher than that of reference detector without photonic cavity structure. In addition, at TEC temperatures of 260 K and 243 K, the detectivities are about 5.75 × 10⁹ Jones and 7.97 × 10⁹ Jones without applied bias, which correspond to two- and three-times enhancement compared to that of the reference detector. The work provides an effective alternative for enhancing mid-infrared photodetection.

This work was supported by A*Star (SERC A1883c0002 and 1720700038), Singapore. Fei Suo and Jinchao Tong contributed equally to this work.

The data that support the findings of this study are available upon reasonable request from the authors.

A.1. Simulation setting

A.2. Parameter optimization for PT hole array structure

To optimize parameters of PT structures for realizing maximum MIR light absorption, we use Lumerical FDTD solution to simulate light absorption in the designed hetero n–i–p photodetector with the PT cylindrical hole array in hexagonal lattice (CH).

Firstly, the absorption of hetero-n–i–p photodiode with the PT structure at λ= 3 μm is simulated by simultaneously scanning hole array period (p) and ratio of hole diameter to period (r_d), defined by d/p, where the hole depth is initially set at 1 μm. The scanning range of p is set around the peak wavelength response of our InAsSb-based photodiode, and the r_d range is chosen in between 0.4 and 0.7 as we consider the practicability of the detector. The corresponding simulation results are shown in figure 7(a). Furthermore, figure 7(b) shows the range of hole depth (D) scanned from 0.5 μm to 1.3 μm at per step of 0.1 μm in the wavelength (λ) ranging from 2 μm to 5 μm under the previously simulated optimal p and r_d.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Those calculations are performed by selecting a set of optimal topological parameters of PT hole array structure for obtaining a maximized absorption in the photodetector, which can then be used as the guideline for the PT hole array fabrication. The corresponding optimal parameters derived from the simulation results are p= 3 μm, r_d from 0.67 to 0.72, and D⩾ 1 μm. Considering fabrication and effective absorption of light, we select 1 μm as the hole depth. Likewise, the optimal r_d is chosen with the smallest value among the selected range.

In addition, to verify the polarization independence of the PT structure, the absorption simulation is implemented by simultaneously scanning polarization angle (θ_pol) and λ, with optimized p, r_d and h values. As shown in figure 7(c), the PT structure possess the feature of polarization independence of the incident light due to the structure symmetry, which is good for practical applications.

A.3. Poynting vector distribution

A.4. Simulated absorption per unit volume

The detailed fabrication process schematic diagrams are described in figure 10. In order to check the effectiveness of PT structure, the hetero-n–i–p photodiode without the PT hole array is also fabricated as the reference device. The fabrication of reference device starts from the step of lithography for mesa pattern define, followed by the same steps as PT devices.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

B.1. PT hole array pattern define and surface passivation

B.2. Mesa etch and surface passivation

B.3. Window open for light absorption and electrodes

B.4. Electrode deposition

The Fourier transform infrared (FTIR) spectrometer equipped with a microscope (Vertex 70, Bruker) was used to characterize the reflection and transmission spectra of photodiodes w/o PT structures. In the measurement setup, a MIR source, an infrared polarizer, a KBr beam splitter, a mercury–cadmium–telluride (MCT) detector and the HYPERION Series FTIR Microscopy and Imaging System with a 36× objective (NA = 0.41) were used to collect the optical signal in 2–5 μm range. Then the collected light signal was normalized with the background spectra of gold pad (for reflection) and GaSb substrate (for transmission). To facilitate the photodetection performance characterization, the fabricated devices were mounted and wire-bonded to the ceramic chip carrier. The current–voltage (I–V) characteristics were performed by the SC-200 mm probe station with the B1500A semiconductor parameter analyzer. The MIR responsivities were obtained from the blackbody light source (at 1000 K) passing through the optical bandpass filter (2.5–4.8 μm) and mechanical chopper, where the blackbody photocurrent signal generated by photodetector was read out by a Lock-in amplifier (EG&G 7260 DSP) after being converted into a photovoltage by a low noise current pre-amplifier (model SR570, Stanford Research Systems). The photocurrent spectra in MIR range were also measured by the FTIR using a MIR source and focusing lens, but with our fabricated detectors w/o PT structures instead of its internal MCT detector to convert optical signal into electrical signal. The response waveform was measured using a 2.94 μm laser with pulse repetition frequency varied from 1 kHz to 1 MHz. The pulse width of the reference signal was thus varied with the repetition frequency while the rise and fall times are both fixed at ∼20 ns. The output power of the laser source for different modulation frequencies was calibrated by a standard power meter. For the measurement performed at different temperatures (78–293 K), the samples were mounted in a cryogenic thermostat in front of the ZnSe window (Optistat DN-V, Oxford Instruments), with the high-vacuum condition (below 10⁻⁵ mTorr).