SEARCH

SEARCH BY CITATION

Keywords:

  • quantum dots;
  • infrared photodetectors;
  • photonic crystals;
  • optical filters

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Device fabrication
  5. Results and discussion
  6. Conclusions
  7. Acknowledgements
  8. References

Quantum dot infrared photodetectors can be coupled with micro-structured filters to create narrowband sensors. Guided-mode resonance filters based on a high-index dielectric slab can exhibit bandpass characteristics that are suitable for monolithic integration with focal-plane arrays. Here, patterned Ge filters were integrated with InGaAs/GaAs quantum dot detectors to linearly tune their 77 K photoresponse peaks from 5.6 µm to 6.2 µm. The dark current was not influenced by these filters but the ability to narrow the photoresponse linewidth was limited by substrate scattering, which is often encountered with front-side illumination architectures. (© 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Device fabrication
  5. Results and discussion
  6. Conclusions
  7. Acknowledgements
  8. References

Over the last decade or so, significant progress has been made in the capabilities of III–V quantum dot infrared photodetectors (QDIPs) 1–3. Focal-plane arrays (FPAs) based on QDIPs have been demonstrated with two-colour detection 4 or operation at elevated temperatures 5. Yet challenges associated with growing self-assembled quantum dots (QDs) still hinder many of the anticipated merits of QDIPs, particularly in comparison to quantum well infrared photodetectors (QWIPs) 2, 6. The intersubband photoresponse in QDIPs and QWIPs is well-suited to multicolour infrared (IR) sensing in the mid-wavelength infrared (MWIR) and other regimes. The spectral linewidth of these devices is narrower than in HgCdTe detectors and this is beneficial in, for example, remote temperature measurements 2.

Hyperspectral imaging may be used in biomedicine 7 or various remote-sensing scenarios 2. Hyperspectral systems require very narrow photoresponse linewidths throughout the relevant spectral band. The full-width at half-maximum (FWHM) may be as low as $ {\rm \Delta} \lambda /\lambda _{\rm p} = 1{\rm %},$ which can be achieved by coupling with bulky components such as diffraction gratings 2 or acousto-optic tunable filters 7. QDIPs and QWIPs that operate through bound-to-bound absorption can exhibit narrow linewidths, however, equation image is still limited to about 10% 8, 9 and these devices typically exhibit low responsivities. Dots-in-a-well (DWELL) QDIPs 1 offer control over the peak wavelength equation image however, equation image is still too large for hyperspectral imaging without enhanced functionality.

To realise compact, spectrally-enhanced IR sensors, narrowband filtering should be implemented at the pixel level. Multi-layer dielectric filters do not suit many spectral lines as the film thicknesses would have to vary from pixel-to-pixel. Actively-tunable filters can overcome this limitation, although additional electronics are required for actuation 2. DWELL-QDIPs have been filtered by patterning a two-dimensional (2D) photonic crystal (PC) through the QD layers 10 and suspended-membrane PCs have resulted in significant spectral narrowing of QWIPs 11. The latter approach could also be applied to QDIPs, however wet under-etching may not be suited to large FPAs. Whereas PCs are fabricated with dielectric materials, surface plasmon-polaritons can be exploited in patterned metal films. The spectral response of QDIPs can also be modified using these plasmonic filters 12. Metallic filters can suffer from absorption losses and lead to lower device performance. In addition, plasmonic filters should be fabricated close to the active region, which may require modifications to the device structure.

The devices presented in this Letter utilise guided-mode resonance filters (GMRFs) to tune the photoresponse of InGaAs/GaAs QDIPs. As shown schematically in Fig. 1, these filters are based on an all-dielectric PC array but the holes do not extend into the III–V material. The normally-incident signal is filtered by interaction with guided-mode resonances (GMRs). That is, diffracted orders from the grating-like pattern couple to leaky modes of the PC slab. GMRs are fundamentally reflectance peaks, however, two interacting resonances can produce a bandpass response that is suitable for monolithic integration 13. The GMRF consists of a patterned Ge slab and a CaF2 cladding layer, as described elsewhere 14. The filter is fabricated onto a completed QDIP so it is essentially detector-agnostic. Numerical simulations have shown that when the period of a triangular lattice is fixed at a = 3.4 µm, the transmitted wavelength is tunable with the hole radius (r) such that equation image 14.

thumbnail image

Figure 1. Layout of an integrated GMRF-QDIP (not to scale). A patterned Ge slab and a CaF2 cladding sit on the ten-layer QDIP; only four QD layers are shown for clarity. The cladding and the mesa are each on the order of 1 µm thick and the actual substrate is about 350 µm thick.

Download figure to PowerPoint

These GMRFs are inherently scalable and could be optimised for other detector technologies by altering the film thicknesses and PC dimensions. Within a spectral band, each narrowband pixel can be tuned by varying only a and r. Different PC geometries can be etched in parallel, so these GMRFs may be compatible with high-density FPAs. Small detector arrays can be used for sensing noxious gases whereas large arrays might be developed for hyperspectral imaging applications.

Device fabrication

  1. Top of page
  2. Abstract
  3. Introduction
  4. Device fabrication
  5. Results and discussion
  6. Conclusions
  7. Acknowledgements
  8. References

The QDIP structure was grown by metal-organic chemical vapour deposition and has been previously reported 15. The undoped active region included ten In0.5Ga0.5As QD layers separated by 50 nm GaAs barriers. The top and bottom n+ contact layers were 300 nm and 1 µm thick, respectively. Large (850 µm × 850 µm) mesas were chosen to facilitate characterisation and these were defined by photolithography and wet etching. To minimise the coupling of unfiltered light, the surface-areas covered by the metal contacts were maximised and this is indicated in Fig. 1. Ohmic contacts were fabricated using photolithography, electron-beam evaporation of Au/Ni/Ge and rapid-thermal annealing.

The GMRF pass-band was designed for equation image ≈ 6 µm to suit the spectral response of these bound-to-continuum QDIPs. The filter areas were defined by photolithography and the dielectric layers were deposited on selected devices. To provide a low transmittance background, the slab and cladding were designed with quarter-wavelength thicknesses at equation image CaF2 deposition with minimal heating was found to be somewhat complicated. Sputter-deposited CaF2 exhibited high MWIR absorption whereas evaporated CaF2 exhibited lower real (n) and imaginary (k) components of the refractive index. On the other hand, the latter contained large grains and provided a poor substrate for Ge deposition 16. For the filtered QDIPs, evaporated CaF2 was initially deposited and then RF sputtered CaF2 was used to planarise the surface. RF sputtering was also used to deposit the Ge slab, exhibiting n = 4.37 at 6 µm. This is higher than the value 17 of n = 4.0 used in the simulations, which should benefit the the fabricated filters. To pattern each Ge slab, ZEP520A resist was exposed by electron beam lithography. The resist was used as a soft mask during inductively-coupled plasma reactive-ion etching with CHF3 gas. Finally, the samples were cleaned and wire-bonded into packages. It is well known that substrate scattering in IR detectors can increase optical crosstalk and the coupling of background light 18. The edges of one sample were entirely surrounded by silver epoxy in order to minimise the latter. This sample contained several filtered and unfiltered QDIPs, including two GMRF-QDIPs with measured hole radii of r = 0.49 µm and r = 0.71 µm. The other two devices that are presented in this Letter (a GMRF-QDIP with r = 0.58 µm and an unfiltered device) were mounted in separate packages with minimal epoxy.

Results and discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Device fabrication
  5. Results and discussion
  6. Conclusions
  7. Acknowledgements
  8. References

Fourier-transform infrared spectroscopy (FTIR) was used to measure the spectral response from these QDIPs and a modulated blackbody was used for the responsivity and detectivity measurements 15, 16. Figure 2(a) shows the photoresponse spectra for the unfiltered QDIP and the three GMRF-QDIPs. Each filtered device exhibits a peak that is tunable (indicated by the arrow) with the GMRF design. The r = 0.71 µm QDIP exhibits equation image= 5.61 µm, whereas the r = 0.58 µm device peaks at 5.92 µm. The r = 0.49 µm device is dominated by a 6.15 µm peak. This is a similar wavelength to equation image= 6.14 µm of the unfiltered QDIP, which exhibits a FWHM of 1.4 µm. The r = 0.49 µm spectrum has a FWHM of 0.78 µm, which is about half of the unfiltered equation image If each linewidth is instead described by the full-width at 60% of the maximum, then the r = 0.49 µm width is only 32% of the reference and this is a clear improvement.

thumbnail image

Figure 2. (a) Normalised photoresponse spectra (arbitrary units) at 77 K and –0.4 V of one unfiltered and three integrated devices. Nominal hole radii are indicated for the three GMRFs and two spectra have been offset. (b) Relative transmittance spectra, extracted from the unnormalised response of each GMRF-QDIP compared with the unfiltered device.

Download figure to PowerPoint

A close examination of Fig. 2(a) shows that the r = 0.49 µm response also includes a broadband component, even though the edges of this sample were covered to minimise substrate scattering. This background is instead attributed to optical crosstalk from an adjacent unfiltered QDIP (not presented here). The single-pass absorption through ten QD layers is relatively low, so some unfiltered signal can be scattered through the substrate to reach the r = 0.49 µm device. Conversely, a sharp spectral feature was evident in the response (not shown) of an unfiltered pixel and this was also attributed to crosstalk.

Aside from the GMRF peak, the r = 0.58 µm photoresponse in Fig. 2(a) resembles the unfiltered QDIP because neither sample was surrounded by silver epoxy. Hence, a significant amount of light was coupled through the substrate in each case.This substrate scattering was mitigated in the r = 0.49 µm device but the use of silver epoxy introduced attenuation dips that are clear in Fig. 2(a). The r = 0.49 µm and 0.71 µm curves exhibit clear dips near 5.7 µm and 6.6 µm, but these dips are less significant in the other two spectra. FTIR measurements on the silver epoxy itself indicated some specific attenuation at 5.78 µm and 6.63 µm, as well as other regions. Attenuation would be enhanced if internally-scattered light interacts with the silver epoxy multiple times at the substrate edges or backside. These features are not detrimental to the r = 0.49 µm GMRF design, however the r = 0.71 µm peak at r = 5.61 µm overlaps with a strong dip and this peak height has been reduced. As a result, the normalised photoresponse for the r = 0.71 µm device in Fig. 2(a) has actually become broader. Without a strong GMR peak, this filter behaves more like a distributed Bragg reflector than a bandpass filter. Fabricating a GMRF with larger holes (r ≈ 0.78 µm) might be able to shift equation image below the attenuation dip and also increase the tuning range of these filters.

The transmittance spectra in Fig. 2(b) were extracted from the unnormalised response of each GMRF-QDIP, compared with the reference device. The magnitude of each spectrum is indicative of the filter transmission ratio, although scattering effects mean that it is difficult to accurately quantify the transmittance in this manner. Clearly, the r = 0.71 µm curve actually exhibits the lowest background because of abundant silver epoxy coverage. The background for the r = 0.49 µm GMRF is slightly higher because of optical crosstalk and the transmittance associated with the r = 0.58 µm device was generally highest. These backgrounds are not entirely related to the filter designs. Rather, the r = 0.58 µm spectrum in Fig. 2(b) indicates that this device received considerable IR signal through the substrate edges, as did the unfiltered QDIP.

The main spectral features in Fig. 2(b) clearly confirm the GMRF performance. All filters exhibit broad minima near 3.1 µm, broad peaks near 4.2 µm and low transmittance past 6 µm. These features result from Fabry–Pérot resonances in the dielectric layers. The actual layers were slightly thicker than quarter-wavelengths at λ = 6 µm, resulting in some red-shift of the broad background. Figure 2(b) also exhibits the GMRs from these filters, which are sensitive to the PC geometry. As well as the targeted peaks near 6 µm (coloured arrows), all spectra display weak dips at λ ≈ 4.5 µm and strong peaks between 3.5 µm and 3.8 µm. The latter GMRs do not affect the GMRF free-spectral range as they fall outside the QDIP photoresponse.

The QDIP equation image values were independent of bias voltage and the filter tunability is compared with the simulated trend in Fig. 3. The actual fabricated hole radius was measured from scanning electron micrographs (SEM) of each PC slab. The least-squares regression (not shown) of the three points in Fig. 3 is given by equation image = –2.49r + 7.36 µm and the correlation coefficient indicates a strong trend. The regression line is steeper than the simulated trend (dashed line) and this discrepancy is related to the high index of the sputtered Ge. The three measured peaks are blue-shifted compared to the simulations. This may also be due to the angle of the air-hole sidewalls, which was found to be about 10° from vertical 16.

thumbnail image

Figure 3. Tunability of the photoresponse peak (λp) with respect to the hole radius that was measured from SEM. Simulations were performed on stand-alone filters 14.

Download figure to PowerPoint

The dark current was measured at various temperatures and similar results were obtained for all devices, suggesting that the filter fabrication steps have not affected the electronic properties of the QDs. The peak responsivity equation image and specific detectivity (D *) were determined using standard procedures 15 and the results are shown in Fig. 4. These quantities depend on the measured photocurrent and hence they are sensitive to substrate scattering. For all bias voltages,the relative responsivities of the three GMRF-QDIPs in Fig. 4(a) correspond to the relative backgrounds in Fig. 2(b). For example, the r = 0.58 µm device exhibits equation image values similar to the unfiltered QDIP because in both samples, a significant proportion of the light was coupled through the uncovered substrate edges.

thumbnail image

Figure 4. (a) Peak responsivity (Rp) and (b) specific detectivity (D *), both at 77 K against bias voltage.

Download figure to PowerPoint

These effects are also carried through to the D * values, so the same relative differences are clear in Fig. 4(b). In general, the peak detectivities are observed with biases of about ±0.4 V. The unfiltered device exhibits the highest value of D * = 5.7 × 108 cm Hz1/2/W at 0.45 V where equation image = 1.9 mA/W. The r = 0.49 µm GMRF-QDIP is characterised by equation image = 0.64 mA/W and D * = 4.0 × 108 cm Hz1/2/W at 0.35 V and also exhibited a narrower spectral response. Discontinuities are present in all D * curves near ±0.5 V and these are artefacts of the measurement setup rather than the QDIP devices.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Device fabrication
  5. Results and discussion
  6. Conclusions
  7. Acknowledgements
  8. References

This work has demonstrated the integration of GMR filters with MWIR QDIPs, which produced a linearly tunable response with the hole radius. Clearly, the equation image and D * values were governed by scattering effects. In fact, the detector area used in such calculations can lead to an overestimation of equation image in some cases 18. To avoid this, flip-chip bonding and substrate-removal might be considered, with backside-illumination employed for characterisation 12. Removal of the thick substrate is used to eliminate crosstalk in commercial FPAs and these QDIPs would also benefit from this approach 2. The filter reflection of off-resonance wavelengths may still result in a reduced D *, however shrinking the contact layer between the QDs and the GMRF could enhance the filter coupling. Differential-mode operation would reduce the common background and also enhance the spectral capabilities of a GMRF-QDIP array.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Device fabrication
  5. Results and discussion
  6. Conclusions
  7. Acknowledgements
  8. References

This work has been supported by the ACT and WA Nodes of the Australian National Fabrication Facility and by the Australian Research Council. The authors are also grateful to Professors Mariusz Martyniuk and Thuyen Nguyen for IR ellipsometry measurements.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Device fabrication
  5. Results and discussion
  6. Conclusions
  7. Acknowledgements
  8. References
  • [1]
    T. E. Vandervelde, M. C. Lenz, E. Varley, A. Barve, S. Jiayi, R. V. Shenoi, D. A. Ramirez, J. Wooyong, Y. D. Sharma, and S. Krishna, IEEE J. Sel. Top. Quantum Electron. 14(4), 11501161 (2008).
  • [2]
    A. Rogalski, J. Antoszewski, and L. Faraone, J. Appl. Phys. 105(9), 091101 (2009).
  • [3]
    A. V. Barve and S. Krishna, in: Advances in Infrared Photodetectors, edited by S. D. Gunapala, D. R. Rhiger, and C. Jagadish (Academic Press, San Diego, 2011), pp. 153–193.
  • [4]
    S. Krishna, D. Forman, S. Annamalai, P. Dowd, P. Varangis, T. Tumolillo, Jr., A. Gray, J. Zilko, K. Sun, M. Liu, J. Campbell, and D. Carothers, Appl. Phys. Lett. 86(19), 193501 (2005).
  • [5]
    S. F. Tang, C. D. Chiang, P. K. Weng, Y. T. Gau, J. J. Luo, S. T. Yang, C. C. Shih, S. Y. Lin, and S. C. Lee, IEEE Photon. Technol. Lett. 18(8), 986988 (2006).
  • [6]
    H. Liu, Opto-Electron. Rev. 11(1), 15 (2003).
  • [7]
    T. Vo-Dinh, D. L. Stokes, M. B. Wabuyele, M. E. Martin, J. M. Song, R. Jagannathan, E. Michaud, R. J. Lee, and X. Pan, IEEE Eng. Med. Biol. Mag. 23(5), 4049 (2004).
  • [8]
    T. Asano, H. Chong, Z. Yi, L. Mingguo, J. C. Campbell, and A. Madhukar, IEEE J. Quantum Electron. 46(10), 14841491 (2010).
  • [9]
    S. Gunapala and S. Bandara, in: Properties of Gallium Arsenide, edited by M. Brozel and G. Stillman, 3rd ed. (INSPEC, London, 1996), pp. 906–917.
  • [10]
    K. T. Posani, V. Tripathi, S. Annamalai, N. R. Weisse-Bernstein, S. Krishna, R. Perahia, O. Crisafulli, and O. J. Painter, Appl. Phys. Lett. 88(15), 151104 (2006).
  • [11]
    S. Kalchmair, R. Gansch, S. I. Ahn, A. M. Andrews, H. Detz, T. Zederbauer, E. Mujagi, P. Reininger, G. Lasser, W. Schrenk, and G. Strasser, Opt. Express 20(5), 56225628 (2012).
  • [12]
    S. J. Lee, Z. Ku, A. Barve, J. Montoya, W. Y. Jang, S. R. J. Brueck, M. Sundaram, A. Reisinger, S. Krishna, and S. K. Noh, Nature Commun. 2, 286 (2011).
  • [13]
    Y. Ding and R. Magnusson, Opt. Lett. 29(10), 11351137 (2004).
  • [14]
    I. R. McKerracher, L. Fu, H. H. Tan, and C. Jagadish, J. Phys. D 46(9), 095104 (2013).
  • [15]
    L. Fu, P. Lever, K. Sears, H. H. Tan, and C. Jagadish, IEEE Electron Dev. Lett. 26(9), 628630 (2005).
  • [16]
    I. R. McKerracher, Post-growth spectral tuning of InGaAs/ GaAs quantum dot infrared photodetectors, Ph.D. thesis, The Australian National University, 2011.
  • [17]
    W. L. Wolfe and G. J. Zissis(eds.), The Infrared Handbook, 3rd ed. (Environmental Research Institute of Michigan, 1989).
  • [18]
    J. Montoya, A. Barve, R. Shenoi, M. Naydenkov, H. Kim, Z. Ku, S. R. J. Brueck, S. Krishna, S. J. Lee, and S. K. Noh, Proc. SPIE 7660, 76603P (2010).