1. Introduction

The past decade has witnessed significant advances in the development of mid-infrared optoelectronic devices. In particular, there have been great improvements of intersubband devices, namely quantum well infrared photodetectors (QWIPs) [1–3] and quantum cascade lasers (QCLs) [4–6]. These innovations were driven by the increasing demand for thermal imaging [7], chemical sensing [8] and infrared data transmission [9]. The performance of all these devices not only depends on a good quantum mechanical design, but also on well designed optical wave guides or resonators. For QCLs, a wide range of resonators have been successfully implemented, offering qualities like high power [10], low divergence surface emission [11] or beam steering capabilities [12, 13].

A key factor affecting QWIP performance is the light coupling scheme. Standard mesa QWIPs are insensitive to surface incident light, since an electronic intersubband transition is only possible with light polarized perpendicular to the quantum wells (QWs) [14]. Without additional measures, a photon with electric field in-plane with the QWs will not be absorbed. For characterization, QWIPs are usually illuminated through a 45° polished surface, resulting in part of the electric field being normal to the QWs. For 2D pixel arrays, the coupling to surface incident light is usually achieved by diffraction gratings [15, 16]. Enhanced capabilities can be provided by using more complex structures like photonic crystals (PCs) or plasmonic arrays. In addition to simple light coupling these devices use resonant absorption to improve detector functionality [17, 18] and performance [19–21].

Previously, we have presented a concept for a free standing photonic crystal slab (PCS) QWIP [22]. In a PCS-QWIP the light is not only guided by the photonic crystal in the in-plane direction, but also in the out-of-plane direction by the free standing slab surrounded by air. This design features high Q-factors of the PC resonances and, therefore, should have the potential to significantly improve detector performance. In this article, we report performance measurements of a PCS-QWIP over the large temperature range from 20 K up to 200 K and compare it to a standard mesa QWIP. To accurately characterize the detector, it is illuminated with single mode QCLs. The temperature dependence of the refractive index, which plays a minor role in standard mesa QWIPs, has a major impact on the position of the resonance peaks in PCS-QWIPs. Especially under narrow-band or single-mode laser excitation it is crucial to consider this peak shift. Here we present experimental data and a peak shift model based on the revised plane wave expansion method (RPWEM) [23].

2. Design and fabrication

The QWIP is grown by molecular beam epitaxy and designed to operate at a wavelength of 7.6 µm. State-of-the-art QWIPs at this wavelength are optimized to have a peak detectivity up to 10¹¹ cm Hz^1/2/W at liquid nitrogen temperature. A high doping concentration of around 4 x 10¹¹ cm⁻² is needed to achieve these high detectivity values [3]. This however limits the temperature performance. QCLs already operate with good efficiency up to room temperatures [5]. To enable monolithically integrated systems, QWIPs with good performance at higher temperatures are desirable. For this reason, the QWs in our QWIPs are doped to an equivalent sheet carrier density of 4 x 10⁹ cm⁻². The low QW doping reduces the dark current and the intersubband absorption of the QWIP, hereby reducing detector noise and permitting the PCS resonances to build up [22].

The QWIP is grown by molecular beam epitaxy. The QWIP layer sequence is a GaAs substrate, followed by a 2 µm thick Al_0.85Ga_0.15As sacrificial layer, a 500 nm thick GaAs (2x10¹⁸ cm⁻³) bottom contact layer, the 1.4 µm thick active region, and a 100 nm GaAs (2x10¹⁸ cm⁻³) top contact layer. The active region consists of 26 periods, each with a well and barrier width of w = 4.5 nm and b = 45 nm, respectively. The PCS photodetectors and, for comparison, standard mesa photodetectors are fabricated from the same QWIP material. First, the PC is defined by anisotropic reactive ion etching, then mesas (148x148 µm) are etched, and finally a SiN_x sidewall isolation and extended contacts are deposited. Finally, the sacrificial layer below the detector region is removed by selective wet chemical underetching (24% HCl solution) to release the free standing PCSs. For fabrication reasons it is not possible to etch holes into the entire device area, hence, the optically sensitive area is smaller (100x100 µm), which needs to be considered for all measurements. A schematic illustration of the finished device is shown in Fig. 1(a) .

 

Fig. 1 PCS-QWIP design. (a) Cross section through the PCS-QWIP structure. (b) SEM image of the PCS overlaid with the zeroth and first order TM-like slab mode profiles.

Download Full Size | PDF

PCs support many optical modes, which results in different resonance frequencies. The slab design provides excellent vertical wave guiding, but introduces additional higher order slab modes [24]. For this device, the PCS is designed such that the lowest order TM-like resonance coincides with the QWIP absorption peak. This resonance was found to exhibit the largest resonant absorption enhancement. TM-like resonances are favorable as most of the electric field is polarized in the out-of-plane direction, which is necessary to be absorbed by the QWIP [25]. The zeroth-order resonance is preferred, since the largest part of the electric field overlaps with the detector active region, in the center of the slab (Fig. 1(b)). The dimensions of the used PCS are a lattice constant of a = 4.0 µm, normalized hole radius r/a = 0.26 and a slab thickness of d = 2 µm.

3. Experimental

The samples are illuminated with a Globar and the spectral response is measured with a Fourier transform infrared spectrometer. Standard QWIPs are measured at a 45° angle of illumination as they are insensitive to surface normal incidence light. The photocurrent spectrum of the standard QWIP has one broad absorption peak at 1310 cm⁻¹ (Fig. 2 , dashed line). Photons below this frequency do not have sufficient energy to excite electrons from the bound state into the continuum. Photons above this frequency exhibit a lower electronic transition probability, hence the absorption is reduced but never vanishes (inset Fig. 2).

 

Fig. 2 Photocurrent response of a PCS-QWIP (solid line) and a standard QWIP (dashed line). The normal incidence light is coupled into a PCS mode and absorbed by the QWIP. The increased photon lifetime at the PCS resonance frequency causes a pronounced photocurrent peak. Inset: The same spectra in log scale. At the peak center at 20 K the responsivity is more than 10x enhanced.

Download Full Size | PDF

PCS-QWIPs are illuminated at surface normal incidence, as the photonic crystal provides the polarization conversion of the electric field. The external radiation couples to symmetry matched PCS modes [26], which then have an electric field component in the out-of-plane direction and can be absorbed by the QWIP. The spectral response of the PCS-QWIP (Fig. 2, solid line) shows a pronounced resonance peak. At the resonance frequency, the photon lifetime in the active region is significantly longer, which causes enhanced absorption and sharp photocurrent peaks. The peak width of the PCS resonances is quantified by the quality factor Q, which is defined as Q = f/∆f. The measured Q-factors in our devices are around 135 for resonances at the QWIP absorption maximum. For resonances far off the absorption maximum much higher Q-factors are possible, as the Q-factor in our devices is still dominated by the intersubband absorption in the quantum wells [22].

To estimate the resonant photocurrent enhancement the detectors are illuminated with a continuous-wave single-mode QCL. The laser beam is spatially and spectrally very well confined and, hence, ideal to measure the detector performance. The peak responsivity of a PCS-QWIP is compared to a standard mesa QWIP in Fig. 3(a) . At 20K the PCS-QWIP signal is more than 10 times larger than the standard QWIP signal. The magnitude of the enhancement is limited by the total absorption in the device, which is caused also by free-carrier absorption in the contact layers. A reduced contact layer doping would permit even higher photocurrent enhancements. In general, the QWIP responsivity increases with increasing bias voltage, as the electron extraction probability increases with the electric field. This, however, also raises the dark current noise [3].

 

Fig. 3 Temperature dependent responsivity, dark current and noise. (a) The PCS-QWIP peak responsivity is increased compared to a standard mesa QWIP by resonant absorption enhancement. (b) The dark current of the PCS-QWIP (solid lines) is lower than of the standard QWIP (dashed lines) (c) Correspondingly, the noise current spectral density S_n of the PCS-QWIP is lower than S_n of the standard QWIP.

Download Full Size | PDF

A benefit of the PCS-QWIP design is that the responsivity is resonantly enhanced without creating additional dark current. In fact, the holes in the PCS reduce the active detector volume, which in turn reduces the dark current noise compared to a standard mesa QWIP (Fig. 3(b)). Noise measurements are performed using a Stanford Research SR570 low noise current preamplifier to amplify the noise spectrum and are recorded with a fast Fourier transform spectrum analyzer. The noise current spectral density S_n of the PCS-QWIP is less than half of S_n of a standard QWIP, although the detector volume is reduced by only 25% by the PC hole etching (Fig. 3(c)). The additional noise current decrease can be attributed to Fermi level pinning of the electronic potential in the PCS [27]. The exposed surface of a PCS-QWIP causes bending of the electronic band structure and depletes the GaAs surface. With a lower number of electrons in the active region the noise current is also lower.

The performance of photodetectors is usually expressed by the specific detectivity D*, which is defined as

 

Fig. 4 Comparison of the specific detectivity D*. (a) Measured detectivities of a PCS-QWIP (solid symbols) and a standard QWIP (open symbols). The increased responsivity and the reduced dark current noise of the PCS-QWIP together cause a significantly larger detectivity D*. At 200 K and 3V bias the PCS-QWIP detectivity shows a kink, which corresponds to a measurement range switch of the preamplifier. The dark current noise is then dominated by the preamplifier and not by the QWIP. (b) The detectivity enhancement is calculated from the PCS-QWIP detectivity normalized by the standard QWIP detectivity. (c) Comparison of the resonant signal enhancement. The detectivity enhancement (black squares) is equal to the responsivity enhancement (open triangles) in the range below 70 K, where detector noise is dominated by background black body radiation. Above 70 K the dark current is dominated by thermal carrier excitation, which is lower in a PCS-QWIP, and the detectivity enhancement becomes larger than the responsivity enhancement.

Download Full Size | PDF

For any measurement with resonant dielectric structures over a large temperature range, it is essential to consider temperature dependence of the refractive index. In standard mesa QWIPs, this effect plays only a minor role as no resonant structures are used. In a PCS-QWIP, the resonance frequencies are determined by the refractive index. When the devices are operated over the temperature range from 20 K to 200 K, the induced resonance shifts can be on the same order of magnitude as the resonance widths. To characterize the temperature dependence, photocurrent spectra are measured over this temperature range. Above 200 K the internal Globar of the spectrometer is not bright enough to generate a large enough photocurrent to exceed the thermal dark current. The temperature behavior is estimated by fitting a Lorentzian peak function to each photocurrent peak (Fig. 5(a) ). Within a temperature range of 180 K the resonance peak shifts over a range of 9 cm⁻¹ while the resonance width stays constant.

 

Fig. 5 Temperature dependent PCS resonance shift. (a) Lorentzian fitted photocurrent spectra measured at temperatures from 20 to 200 K. The resonance peak shifts over a range of 9 cm⁻¹. For accurate characterization the PCS-QWIPs are illuminated with QCLs (Laser 1-3). (b) The spectral resonance shift is modeled with the revised plane wave expansion method.

Download Full Size | PDF

For accurate characterization, the PCS-QWIPs are illuminated with QCLs. Laser excitation allows excellent focusing, is spectrally well confined and provides high power density. By switching on the Laser which overlaps best with the PCS resonance, it is possible to measure the detector responsivity while maintaining a large signal-to-noise ratio over the full temperature range. The emission spectra of the three used lasers are overlaid in Fig. 5(a). The optical power was between 400 and 900 µW on the detector size of 148 x 148 µm². This laser power would be strong enough to operate the QWIPs up to room temperature. However, at temperatures above 200 K the dark current become large enough to saturate the preamplifier and ultimately even heats up the detectors.

The temperature dependence of the PCS resonances can be modeled by using the revised plane wave expansion method [23, 24]. This method allows calculation of the photonic band structure and the resonance frequencies. For the photonic crystal slab consisting of a GaAs/AlGaAs heterostructure and at 80 K, an average refractive index of 3.12 is used. The refractive index change is approximated by ∆n = 4.5x10⁻⁵ T[K] for GaAs [28]. From the simulation we extracted a spectral shift of −0.055 cm⁻¹/K, which fits well to the measured resonance peak shifts (Fig. 5(b)). Only at very low temperatures, where the approximation for the refractive index change is not valid, the simulation deviates significantly from the measured spectral shift. The magnitude of the peak shift caused by the refractive index change depends also on the PCS geometry and the material. The influence is larger when the overlap of the optical PCS mode with the dielectric material is larger. For example for very thin slabs, compared to the wavelength, the optical mode would extend far into the surrounding air and exhibit only little changes of the effective refractive index.

7. Conclusion

In conclusion, we characterized the performance of a quantum well infrared photodetector, which was fabricated as a photonic crystal slab resonator. We report a resonant enhancement of the specific detectivity D* up to 20 times. This enhancement is a combined effect of increased responsivity and noise current reduction. For accurate characterization the PCS-QWIPs were illuminated using QCLs. We observed a resonance peak shift of −0.055 cm⁻¹/K, caused by the refractive index change of the dielectric material with temperature.

Future research on this topic will include optimization of the electronic design of the QWIP and designs to integrate the PCS-QWIP into backside-illuminated focal plane arrays. We envision that this advanced device concept will yield multi-color high temperature mid-infrared optoelectronic systems for applications including thermal imaging or high speed data transmission.