We demonstrate a room-temperature semiconductor-based photodetector where readout is achieved using a resonant radio-frequency (RF) circuit consisting of a microstrip split-ring resonator coupled to a microstrip busline, fabricated on a semiconductor substrate. The RF resonant circuits are characterized at RF frequencies as function of resonator geometry, as well as for their response to incident IR radiation. The detectors are modeled analytically and using commercial simulation software, with good agreement to our experimental results. Though the detector sensitivity is weak, the detector architecture offers the potential for multiplexing arrays of detectors on a single read-out line, in addition to high speed response for either direct coupling of optical signals to RF circuitry, or alternatively, carrier dynamics characterization of semiconductor, or other, material systems.
The rapid growth of wireless communication technologies has resulted in an abundance of new chip-scale radio-frequency (RF) structures and devices, with a corresponding decrease in RF device cost. New technologies and the scaling of RF components hold the potential to continue the shrinking of the “THz gap” between optical and optoelectronic devices (typically operating at >100 THz frequencies) and electronic devices, now operating up to 100's of GHz and even THz frequencies.1–7 There is thus increasing interest in devices and architectures that are able to merge RF and optical capabilities, linking these two technologies that make the foundation of our modern communication infrastructure. At the same time, the greater accessibility of RF components offers new opportunities for materials and device metrology at lower costs. These components could enable efficient and cost-effective high-frequency measurement of device performance and material response across a range of optical materials and optoelectronic device architectures. The RF frequency range between 1 and 20 GHz is particularly intriguing since several low-cost options for RF sources and detectors are readily available that operate at these frequencies, commensurate with the time constants associated with state-of-the-art electronic and optoelectronic devices, and the charge carrier dynamics in these devices.
The utilization of RF circuitry for optical detection is perhaps best demonstrated in the microwave- and lumped element- kinetic inductance detectors (MKID and LEKID, respectively).8–10 These detectors utilize superconducting RF resonant circuits with high-Q (>106) resonances resulting from the effectively lossless superconducting transmission lines.11–13 Radiation with energy greater than the superconducting gap, incident on a co-planar waveguide λ/4 resonator (MKID) or a lumped element LC resonator consisting of a meander inductor with an interdigitated capacitor (LEKID) breaks the Cooper pairs in the superconductor and generates quasiparticle excitations. This in turn changes the surface impedance on the superconductor and has the effect of changing the inductance and resistance of the resonator. These effects can be measured in either the phase or amplitude of the transmitted RF signal. The -KID class of detectors, as a result of their ultra high-Q, can achieve single-photon sensitivity, coarse energy-resolution, and perhaps most importantly, multiplexing of 1000's of detectors (with slightly different resonances) along a single busline.14–16 In addition, this detection mechanism can exhibit sensitivity at photon energies much lower than traditional semiconductor photodetectors (wavelengths in the mm and sub-mm range).9,17 These features make -KIDs ideal for low photon flux imaging across a range of wavelengths, attractive for a number of astronomy, astrophysics, and particle physics applications. However, these detectors require operating temperatures below the Tc of superconductors, and typically below 1 K. The high-Q and long quasiparticle lifetimes, which depend on the superconducting material choice and quality as well as temperature, are typically in the 100's of microseconds,18,19 which limits the bandwidth of M- and LE-KIDS to low kHz frequencies. MKIDs utilizing superconductors with shorter quasiparticle lifetimes can exhibit faster responses (100's of kHz),20 while the -KID's sharp rise times have been explored for time of arrival resolutions on the order of single μs.21
While some advantages of the superconductor-based KIDs (high sensitivity, high-Q) do not extend to temperatures above the superconductor Tc, the broader operational principles of -KIDs can be utilized to design room-temperature, high-speed semiconductor-based detectors capable of similar multiplexing (if not quite as dense as the superconducting -KIDs). Recent work has demonstrated the utility of microwave probes for measuring the carrier dynamics in semiconductor materials.22 In Ref. 22, electron hole pairs (EHPs) are photo-excited, and the resulting change in the free carrier concentration of the material leads to a corresponding change in the material permittivity, as expected from the Drude formalism. This change is measured in the change of reflectivity for a probe microwave signal, allowing for time-resolved measurement of carrier concentrations in the semiconductor material. A similar effect has been used to demonstrate tunable THz meta-materials, where THz frequency optical transmission through split-ring resonator (SRR) arrays can be modulated by optical excitation of EHPs in the semiconductor-filled gap of the SRR.23–29 Here, combining the results of Refs. 22–29 with the fundamental approach employed by the -KIDs, we demonstrate high operating temperature detectors where incident light, absorbed by a semiconductor material in the SRR gap, alters the local conductivity (or complex permittivity) of the resonant microwave circuit and is read out with an RF signal.
Our basic device structure (Figs. 1(a) and 1(b)) consists of a microstrip busline coupled to a SRR structure with a resonance in the 10–16 GHz range. The devices are fabricated using a UV photolithography, metallization, and lift-off process to pattern 500 nm thick metal (Au) features on a double-side polished semi-insulating (SI) GaAs substrate. The back side of the GaAs wafer is also coated with 500 nm Au. Each metallization uses a thin layer of Ti (10 nm) for adhesion. The resulting circuit can be understood using an equivalent lumped element model (Fig. 1(c)) where the SRR is modeled as an LC resonator that is capacitively coupled (CS) to the busline. Such structures act as stopband filters and have been previously utilized as a platform to study left-handed materials in the microwave frequency range.30–32 In our structures, when light with energy above the GaAs band-edge is incident upon the gap in the SRR, the excited electron hole pairs modulate the SI-GaAs conductivity in the capacitive gap, effectively changing the R1 of the circuit shown in Figure 1(c) and thus tuning the RF-response (in particular the Q) of the resonant circuit (in a similar manner to the tuning of the free-space THz transmission in the metamaterials of Refs. 23–29). At resonance, the transmitted RF signal at the output port of the circuit can be used to measure the free carrier concentration, and therefore, the intensity of the incident light at the capacitive gap (C1), or alternatively, could allow for incident light to directly modulate the transmitted RF signal. The RF resonance of the SRR can be easily tuned by modifying the geometry and material properties of the unit cell such that multiple SRRs can be coupled to a single busline, offering the potential for RF-multiplexed detection/modulation using a single input and output port. The time-response of the detector element is determined by a combination of carrier lifetimes and the inherent time-response of the LC circuit. The capacitive gap of the SRR can be filled with a range of different materials using direct epitaxial growth on the substrate wafer or some form of pick and place or deposition techniques to achieve significant control of carrier lifetime and semiconductor band-gap. This offers a path towards multiple-wavelength and high-speed or high-sensitivity detection of incident radiation. In this manuscript, we demonstrate the performance of such detectors using SI GaAs and show control over the RF resonance of the SRR, in addition to the ability to detect light on multiple detectors coupled to a single busline. The optical response and sensitivity of the fabricated detectors are characterized, and potential applications for the demonstrated detectors are discussed. Our results are modeled analytically and simulated with good agreement.
Birdseye schematic (top), cross-sectional schematic (middle), and optical micrograph (bottom) of (a) single and (b) double SRR detector structures showing relevant dimensions. (c) Equivalent circuit model for the SRR RF circuit response and (d) schematic of optically pumped single-SRR detector structure.
Birdseye schematic (top), cross-sectional schematic (middle), and optical micrograph (bottom) of (a) single and (b) double SRR detector structures showing relevant dimensions. (c) Equivalent circuit model for the SRR RF circuit response and (d) schematic of optically pumped single-SRR detector structure.
Figure 2 shows the SRR RF response as a function of the circuit geometry measured with an Agilent 5230A Performance Network analyzer (PNA) and a probe station. A Short-Load-Open-Thru (SLOT) calibration was performed with on-chip standards to move the measurement reference planes to the tips of the GSG probes. The position of the RF resonance is primarily determined by its geometry. The resonant frequency can be fine-tuned () by controlling the capacitive gap of the SRR. Figures 2(a) and 2(b) show the (a) experimental and (b) HFSS simulated RF amplitude transmission for a SRR of side lengths 1 mm, and a spacer of , for a range of capacitive gap values . Good qualitative agreement is demonstrated between the experimental results and simulations with respect to the depth, linewidth, and spectral shift achieved with changing . Figure 2(c) shows the experimental RF amplitude transmission spectra for SRRs of side lengths 1 mm, capacitive gaps , for a range of coupling gap distances . The black line shows the response for a busline without the SRR, which as expected, shows no resonance. For all , strong resonant features are observed at , with decreasing magnitude and linewidth as a function of increasing coupling gap . As expected, in the absence of the superconducting metals used for the -KID devices, we observe significantly diminished Q's (∼10–20) in all of our SRR circuits, which will limit the density of detectors that can be coupled to a single busline. Finally, a significant shift of the RF resonant frequency (∼3 GHz), and a slight increase in Q, can be obtained by using a double-SRR structure, as shown in Fig. 2(c). Here, the inner SRR has side lengths of 0.8 mm, with and for both SRRs.
(a) Experimental and (b) simulated S21 spectral response of single SRR circuits with fixed spacer () and varying capacitive gaps (). (c) Experimental S21 spectra of single SRR circuit with fixed capacitive gap () and varying spacers (), S21 for busline only (no SRR) is shown in black. (d) Spectra for single- and double-SRRs with , , and . Inner SRR on the double-SRR has side length of 0.8 mm.
(a) Experimental and (b) simulated S21 spectral response of single SRR circuits with fixed spacer () and varying capacitive gaps (). (c) Experimental S21 spectra of single SRR circuit with fixed capacitive gap () and varying spacers (), S21 for busline only (no SRR) is shown in black. (d) Spectra for single- and double-SRRs with , , and . Inner SRR on the double-SRR has side length of 0.8 mm.
For optical characterization, a diode laser was focused onto the detectors via a 1 in. focal length lens after passing through a spatial filter. The laser spot size was measured to be ∼50 μm (FWHM), and incident laser powers were measured by replacing the RF detector device with a broadband thermal sensor. The laser light incident upon the surface was controlled by both the laser driving current and neutral density (ND) filters in the laser beam path, and the absorbed laser power determined using the calculated transmission at a GaAs/air interface. Because some light is reflected from the Au SRR, the values shown for absorbed laser power are likely slightly overestimated. In addition, because the laser beam spot is larger than the capacitive gap upon which the beam is focused, many of the generated EHPs will not contribute to the modulation of the circuit, leading to a further underestimation of our detector response. The detector's RF spectral response was measured using the PNA for varying laser excitation powers. Figure 3(b) shows representative scans of the detector response as a function of absorbed laser power 0 to ∼45 mW for a 785 nm laser diode in continuous wave operation. The spectrum for the SRR with no light incident on the gap is shown in black.
(a) HFSS-modeled S21 spectra of single-SRR circuit with inset showing contour plot of electric field magnitude at resonance, for σ = 0 S/m. The effect of absorbed light is modeled as a varying conductivity in a ∼ 700 nm thick layer of GaAs () under the split-gap, shown schematically in inset (i). (b) Experimental S21 spectra of single-SRR circuit (shown in inset (ii)) under CW excitation from 785 nm diode laser and (c) lumped circuit element modeled S21 spectra of single-SRR circuit with the effect of absorbed light modeled as a changing R1 (shown in inset (iii)).
(a) HFSS-modeled S21 spectra of single-SRR circuit with inset showing contour plot of electric field magnitude at resonance, for σ = 0 S/m. The effect of absorbed light is modeled as a varying conductivity in a ∼ 700 nm thick layer of GaAs () under the split-gap, shown schematically in inset (i). (b) Experimental S21 spectra of single-SRR circuit (shown in inset (ii)) under CW excitation from 785 nm diode laser and (c) lumped circuit element modeled S21 spectra of single-SRR circuit with the effect of absorbed light modeled as a changing R1 (shown in inset (iii)).
It is evident from these results that the generation of EHPs in the capacitive gap of the SRR has a significant effect on the response of the SRR circuit, effectively damping the SRR resonance, and giving a >5 dB change in the depth of the SRR resonance at absorbed powers of ∼45 mW. No response was observed when the laser beam was incident anywhere else on the sample surface. Figure 3(a) shows the numerical simulations (using HFSS) for the device in (b). Using the photon flux of our experiments, and assuming a pump laser absorption length of ∼700 nm and a carrier lifetime of in the SI:GaAs, we are able to calculate a bulk conductivity as a function of incident laser power for the 700 nm-thick section of GaAs in the capacitive gap of the SRR. These results show good agreement with the experimental results of Fig. 3(b). Figure 3(c) shows the results from the equivalent circuit model of the SRR, where the resistor () in the RLC structure represents the combination of the resonator Ohmic loss and the conductivity of the SRR gap. By reducing the value of , we can qualitatively reproduce our experimental results.
In order to better determine the detector sensitivity, the laser diode was modulated at 50 Hz with a 50% duty cycle pulse, and the transmitted RF signal was measured with a Pasternack 10 MHz-18.5 GHz zero-biased Schottky RF detector. We were able to measure the effect of the incident laser power (down to the ∼μW range) on the detector response by feeding the detector output into a lock-in amplifier (LIA), and measuring the LIA output (in Volts) as a function of laser power. The data in Fig. 4, linear over a wide range of incident optical powers, gives some indication of the detector responsivity in V/W, though this value will scale linearly with the RF output from the PNA (set to 3 dBm for this experiment). The results from Fig. 4 demonstrate that we can measure incident optical powers at the ∼μW range. While the sensitivity demonstrated in Fig. 4 does not exceed state-of-the-art photodetectors, there are yet potential advantages to our detector architecture when compared to standard photodetector devices, which we describe below.
Lock-in amplifier output signal from the SRR circuit as a function absorbed power, using 785 nm diode laser excitation, modulated at 50 Hz with 50% duty cycle square pulses. The inset shows schematic of SRR excitation.
Lock-in amplifier output signal from the SRR circuit as a function absorbed power, using 785 nm diode laser excitation, modulated at 50 Hz with 50% duty cycle square pulses. The inset shows schematic of SRR excitation.
Since the response of each detector element is relegated to the resonant frequency of the SRR associated with it, multiple detectors can be linked to a single busline. Therefore, only a single input and output are required for measuring multiple detectors' responses given that each detector occupies a separate range in the RF spectrum. Figure 5 shows an example of such a configuration, with three double-SRR detectors coupled to a single busline. The RF transmission of the circuit was measured in dark (no illumination) and with laser illumination on each of the SRR's inner and outer capacitive gaps ( and , respectively). As can be seen from Fig. 5, light incident on each SRR can be resolved in the RF spectrum, with minimal cross talk between detector elements. With the current SRR RF linewidth, approximately 10 SRRs could be read out from a single busline across a 10 GHz span of the RF spectrum. With improvements in resonator Q, or alternatively, by extending the RF operational range, additional detectors could be multiplexed on a single busline.
RF response of circuit with three double-SRR's on a single busline. All double-SRRs have an outer ring side length of 1 mm, with inner ring side lengths of (a) 0.5 mm (magenta), (b) 0.8 mm (red), and (c) 0.75 mm (blue). All SRRs have and . The RF spectra shown for each sample correspond to the detector RF response in dark (black, solid), with the inner capacitive gap () illuminated (dotted) and with the outer capacitive gap () illuminated (solid). Note that in (a) the inner gap illumination spectrum (dotted magenta) largely overlaps with the dark signal (solid black). The schematic inset shows the illumination positions, color coded for each SRR detector structure.
RF response of circuit with three double-SRR's on a single busline. All double-SRRs have an outer ring side length of 1 mm, with inner ring side lengths of (a) 0.5 mm (magenta), (b) 0.8 mm (red), and (c) 0.75 mm (blue). All SRRs have and . The RF spectra shown for each sample correspond to the detector RF response in dark (black, solid), with the inner capacitive gap () illuminated (dotted) and with the outer capacitive gap () illuminated (solid). Note that in (a) the inner gap illumination spectrum (dotted magenta) largely overlaps with the dark signal (solid black). The schematic inset shows the illumination positions, color coded for each SRR detector structure.
The detectors presented here may have potential applications for material characterization. As demonstrated in Ref. 22, the time response of a reflected microwave-frequency probe from a semiconductor surface allows for the measurement of the time-dependent photo-excited free-carrier concentration (and thus minority carrier lifetime) in an optoelectronic material. These experiments typically use large areas of materials and free-space microwave probes. Similar measurements can be achieved on material samples of areas ∼, with a direct readout from the SRR-circuit using the detectors presented here. However, it is important to note that free carriers in the semiconductor material (outside of the capacitive gap) will damp the circuit response and obscure the effects of interest. Thus, for metrology of conductive active regions, one must either use a “pick and place” approach to position a lifted-off active region in the gap, or alternatively, etch the sample down to an insulating substrate everywhere but between the gap. Finally, the demonstrated detectors offer a compact (on the microwave scale) mechanism for direct integration of optical signals to microwave circuitry.
The time response of the detector architecture will depend primarily on the excited carrier lifetimes in the capacitive gap and the quality factor Q of the resonator. Using the expression , where is the energy stored in the resonator and is the energy lost per cycle, we can write a characteristic time for the detector response as , where is the resonant frequency of our detector. For the case of short carrier lifetimes (), and using , extracted from the experimental response of our detector, we calculate . For a material with long carrier lifetimes (), however, the time response of the detectors will be dominated by the free carrier lifetime, such that . Experimental characterization of the time response of the detectors presented is ongoing.
In conclusion, we demonstrate a room-temperature optical photodetector based on a resonant RF circuit. Our RF circuits utilize SRRs, and can be tuned coarsely by changing the dimensions of the SRR (side length) and finely by controlling the capacitive gap. We show linear detector response over three orders of magnitude of absorbed laser power for light incident on the SRR capacitive gap of our RF circuits. The demonstrated devices were modeled using full-wave electromagnetic simulations in commercial software (HFSS) as well verified analytically using a lumped circuit element approach. Results from both sets of models showed good agreement to our experimental results. Finally, we demonstrate multiple detector elements multiplexed on a single busline and discuss the potential applications of the demonstrated detectors. In the current design, such detectors would not be able to compete with the sensitivity or speed of commercial photoconductive detectors but the design methodology developed and underpinning principles may benefit numerous applications, including materials characterization, direct integration of optical signals with microwave circuitry, and multiplexed, high speed read-outs of multiple detector arrays into RF electronic circuitry. In addition, while we use SRRs fabricated on GaAs in our demonstration, there are a plethora of other resonant structures and semiconductor substrates/structures that can be used in a similar configuration that have different characteristics (such as higher Q, faster recombination times, different optical absorption profiles, etc.) that would give the designer a larger design space for specific applications.
The authors would like to acknowledge funding from the National Science Foundation, Award Nos. DMR #1210398 (D.W. and R.L.) and DMR #1209761 (C.R.). The authors (J.W.A., M.S.A., and B.R.W.) would like to thank the 2015 AFRL/RW Corporate Venture Fund award (Dr. D. Lambert) and AFOSR Lab Task 14RY07COR (Dr. G. Pomrenke). D.W. was grateful to Professor J. Vieira, Professor J. Filippini, and Professor J. Eckstein (UIUC) for stimulating and illuminating discussion.