Graphene-Based Optoelectronic Mixer Device for Time-of-Flight Distance Measurements for Enhanced 3D Imaging Applications

A large and growing number of applications benefit from innovative and powerful 3D image sensors. Graphene photodetectors can achieve 3D sensing functionalities by intrinsic optoelectronic frequency mixing due to the nonlinear output characteristics of the sensor. In first proof of principle distance measurement demonstrations, we achieve modulation frequencies of 3.1 MHz, signal-to-noise ratios of ∼40 dB, distance detection up to at least 1 m, and a mean accuracy of 25.6 mm. The scalable More than Moore detector approach enables geometrical fill factors close to 100% and can easily complement powerful functionalities by simple back-end integration on top of CMOS electronics.

ABSTRACT: A large and growing number of applications benefit from innovative and powerful 3D image sensors. Graphene photodetectors can achieve 3D sensing functionalities by intrinsic optoelectronic frequency mixing due to the nonlinear output characteristics of the sensor. In first proof of principle distance measurement demonstrations, we achieve modulation frequencies of 3.1 MHz, signal-to-noise ratios of ∼40 dB, distance detection up to at least 1 m, and a mean accuracy of 25.6 mm. The scalable More than Moore detector approach enables geometrical fill factors close to 100% and can easily complement powerful functionalities by simple back-end integration on top of CMOS electronics. KEYWORDS: graphene, time-of-flight (ToF), range sensor, distance measurement, superheterodyning, 3d imaging, optoelectronics, optoelectronic mixer, 3d sensor 3D vision is a fundamental cornerstone for highly complex tasks in the future fields of autonomous driving for scene recognition 1,2 or novel manufacturing and quality tests for Industry 4.0. 3−5 The medical sector, 6 human recognition 7 and scanning, 8 and scene analysis in security applications for crowd screening 9 can benefit from high-performance 3D camera systems as well as the entertainment sector, 10,11 which pioneered widespread 3D imaging systems and awareness with the first generation of Microsoft's Kinect. 12 Because most applications differ in their requirements, research and industry developed various depth measurement and 3D imaging concepts, each with its own advantages and challenges. However, the objective of all systems is to improve overall performance characteristics, especially lateral precision and resolution, speed of image acquisition for real-time applications, reliability, and integrability, e.g., for mobile or spacelimited systems.
State-of-the-art 3D imaging systems typically use either triangulation 13,14 or the time-of-flight (ToF) 15 principle. Systems based on triangulation utilize geometries and special configurations to compute depth information from optical data. These systems are subdivided into (I) passive systems, such as stereo vision 8 or dynamic vision 13 entailing significant drawbacks and limitations in demanding applications, and (II) active systems that compensate for many of these limitations. The structured light approach, for example, uses specific light patterns to illuminate the scene, using diffraction of the reflected pattern to compute depth information, achieving depth resolutions of at least a several tenths of a micometer. 16 However, triangulation-based systems inevitably require multiple active elements, preventing highly integrated devices or miniaturized systems. This approach is also error-prone, as it has difficulty detecting uniformly shaped or monochromatic objects.
ToF methods such as the light detection and ranging (LiDAR) 17 or photonic mixer devices (PMD) 18 exploit the finite speed of light and utilize the difference between light emission and detection of photons reflected from the scene. The propagation time results in the distance. ToF principles range from (I) direct methods to accurately measure propagation time requiring very precise timing and fast sensors, (II) indirect methods that modulate the emitted light and measure the phase delay due to the propagation time, and (III) frequency modulated methods 2 that allow depth resolutions in the range of a few millimeters for less complex devices 19 down to several tenths of a micrometer for very complex and time-consuming techniques. 20 These approaches allow for very large-scale integration, e.g., enabling 3D in mobile applications, and do not suffer from scene-dependent uncertainties such as objects. However, today's approaches rely on highly complex sensor architectures or systems, such as scanning systems involving sensitive micromirrors, linked with extensive readout complexity and signal processing (e.g., mixing and circuitry).
The claim of this paper is the use of graphene 21 for indirect ToF paving the way to overcome bandwidth limitations of silicon-based technology and simplifying the sensor architecture by exploiting nonlinearities resulting in internal optoelectronic frequency mixing. Graphene can exhibit charge carrier mobilities above 200 000 cm 2 V −1 s −1 , 22 allowing bandwidths far beyond that of state-of-the-art silicon devices. Furthermore, the linear dispersion relation of graphene allows for an unlimited choice of illumination wavelengths and thus does not restrict applications due to light or laser safety limitations. 23 Although mixing of two optical 24 or two electrical signals 25 has already been demonstrated in graphene as well as optoelectronic mixing at modulation frequencies above 65 GHz 26,27 in highly optimized device structures, previous articles only mention 3D imaging to be a potential application lacking further demonstration.
In this work, we demonstrate for the first time a metal− graphene−metal photonic mixer device (MGM-PMD) for optical distance measurements exploiting internal optoelectronic frequency mixing. Proof of concept measurements show that operation frequencies in the megahertz range can easily be achieved without relying on complex device architectures or ebeam lithography.
Conventional PMDs require a highly precise gate control and complex readout circuitry (I) to separate charge carriers between two integrated readout diodes and (II) to detect a phase shift of the incoming amplitude-modulated light. An electrical circuit-based cross-correlation of both the modulated gate and the optical signal reflected from the scene results in 3D sensing capabilities. In contrast, the PMD functionality demonstrated in this work is provided by the MGM-PMD itself because an intrinsic frequency mixing of the electrical gate modulation with the optical modulation 27 occurs due to nonlinearities in graphene. Besides reduced integration efforts (no differential readout of two diodes is necessary), the complexity of an electrical circuitry for signal generation and to detect and evaluate accumulated charge on the respective readout diodes is significantly reduced. In the following section, the device technology and fabrication steps are presented. Figure 1 shows the process chain of the MGM-PMD fabrication. At first, trenches for the buried gate contacts have been wet chemically etched into the glass substrate ( Figure 1a) followed by chromium (Cr) gate contact deposition ( Figure  1b) and aluminum oxide (Al 2 O 3 ) passivation (Figure 1c). In a next step, interdigitated metal finger structures consisting of chromium/gold (Cr/Au) have been deposited (Figure 1d) prior to the monolayer graphene transfer (Figure 1e). Finally, a further Al 2 O 3 passivation layer has been deposited on top of the MGM-PMD (Figure 1f) in order to protect the device from environmental influences, which, for example, cause unwanted doping. 28 Further details on the MGM-PMD fabrication are given in the Methods section. A false-color Raman mapping image of the metal−graphene−metal photodetector on a glass substrate (purple) with graphene channel (cyan), source and drain Cr/Au contacts (yellow), and the buried gate Cr contact (black) without passivation layer is shown in the close-up (black rectangle) in Figure 1.
2D material qualities (crystallinity and defect density) have been monitored by Raman spectroscopic analysis (cf. Figure  2a). The small D-peak indicates a low defect density in monolayer graphene stating the good initial graphene quality and thorough PMMA-assisted graphene transfer.
Prior to the distance measurements, current−voltage (I−V) measurements were performed to electrically characterize the devices. First, the source current (I S ) has been determined to be a function of the gate voltage (V G ) (Figure 2b) showing the expected nonlinear behavior. From the results presented in Figure 2b, charge carrier mobilities have been extracted with the direct transconductance method (DTM) as proposed by Zhong et al. 31 considering the GFET as a parallel connection of transistors. Taking this consideration into account, the effective channel width W eff is given by the channel width W Ch multiplied by the number of gaps between the fingers (details on the finger structure are given below). Utilizing the DTM, electron and hole mobilities of μ e = 140 cm 2 V −1 s −1 and μ h = 266 cm 2 V −1 s −1 have been extracted for V DS = 100 mV, respectively. Compared to conventional semiconductor or other graphene-based devices, the mobilities presented in this work are comparatively low because CVD graphene has been used in conjunction with additional encapsulation of graphene in between Al 2 O 3 and SiO 2 that reduce mobilities further. Furthermore, source current vs drain−source voltage (V DS ) curves have been acquired, showing resistance behavior (Figure 2c). For the first-ever distance measurement demonstration, the gate has been modulated because the behavior of I S as a function of V G is highly nonlinear (Figure  2b), whereas I S linearly depends on V DS (Figure 2c). Previously, it has been shown that V DS modulation can also

Nano Letters pubs.acs.org/NanoLett
Letter be used to generate optoelectronic mixing in a graphene device, but at considerably higher channel bias voltages of at least 3 V and optical power exceeding 20 mW. 32 In order to estimate device and design specific modulation frequencies prior to experimental 3D measurements, the gate− drain, gate−source, and drain−source capacitance of the structure have been determined in a material-and architecture-specific manner. In this context, the drain−source capacity C DS for MSM finger structures can be estimated within 2% accuracy for the sensor geometry presented here taking into account the following relationship: 33 where N is the finger pair number, W Ch is the width of the graphene channel, ε r is the permittivity number, W F is the finger width, and L is the finger spacing. This interpolation considers the layer thickness of the contact material t Con to be much smaller than the finger width W (t Con ≪ W). The drain− source capacitance of the MGM-PMD prototype is C DS = 7.2 fF (N = 5, W Ch = 29.02 μm, ε r,Graphene = 6.9, 35 W = 4.17 μm, L = 11.22 μm, t = 135 nm). According to Mao et al., 36 the gate− source/gate−drain capacitance is to a first approximation C GS/GD = ∼12 pF (ε r,Ald 2 Od 3 = 9.5, t Ald 2 Od 3 = 20 nm). In this first ever demonstration, the modulation frequency of the graphene PMD achievable in the current design is significantly dominated by the gate capacitance. Considering the experimentally determined maximum sensor resistance of R = 427 Ω at the Dirac point, the 3 dB cutoff frequency is found to be f g ≈ 31 MHz. According to this estimation, electrical and optical modulation frequencies have been chosen far below the 3 dB cutoff frequency in first distance measurement demonstrations utilizing graphene-based mixing devices to ensure sufficiently high signal-to-noise ratios of the sensor output signal as demonstrated in the next section (cf. Figure 4). A schematic of the optical distance measurement setup utilizing 2D material-based MGM photodetectors is presented in Figure 3. The setup includes a 405 nm amplitude modulated laser, a motorized linear stage with a retroreflector, an electromechanical shutter, a function generator, a current− voltage amplifier, a lock-in amplifier, and a voltage source. The device under test (DUT) is contacted electrically using probe needles. The drain contact is biased utilizing a DC voltage supply to ensure charge transport across the channel region by spanning up an electrical field. The optical modulation signal f opt from the function generator serves as an input for the laser source. The same signal is fed into a lock-in amplifier, which internally generates a difference frequency of f opt and a freely adjustable frequency f el that is provided by the lock-in amplifier itself. The signal f el serves as the electrical modulation signal for the gate of the MGM-PMD. This measurement technique ensures that the lock-in amplifier reliably detects the differential frequency |f opt − f el | being generated by the DUT. In a next stage, the optoelectronically mixed current signal of the DUT is amplified (gain 10 4 ) and converted to a voltage which is then used as input signal for the lock-in amplifier. The motorized linear stage is used to move the retroreflector and utilized to change the distance from the sample to the light source. The laser beam can be blocked by means of an electromechanical shutter. A high-speed photoreceiver with a crystalline Si-PIN photodiode (iC212, iC-Haus GmbH) connected to a conventional oscilloscope is used to monitor the optically modulated signal.   . Schematic representation of the measurement setup for ToF distance measurements using a graphene-based MGM-PMD. In this configuration, the lock-in amplifier has been set to lock onto the difference frequency internally generated from the electrical modulation signal (f el ) and the optical modulation frequency (f opt ).

Nano Letters pubs.acs.org/NanoLett
Letter for a fixed distance d = 0 mm (cf. Figure 3). The distance measurement has been conducted using a gate modulation voltage of V G = 3 V pp (DC offset V DC = −2 V) and a drain− source bias of V DS = 10 mV. From the incident laser power (P laser = 10 mW) and a detected mixing current of 0.09 μA, a sensitivity of ∼0.1 mA W −1 for the mixed signal can be derived at that operating point. The signal located at the differential frequency f mix exhibits an excellent phase stability of 0.56°c orresponding to a depth resolution of ∼75 mm at 3.13 MHz with the shutter open. In time-of-flight optical distance measurements, a high phase stability of the sensor is a prerequisite to achieve a high depth resolution. The signal-tonoise ratio of the differential frequency components amplitude signal of the MGM-PMD exceeds values of 40 dB, indicating that in further experiments light intensities might be reduced or distances increased. In a next step, the nonlinear output characteristic of the graphene sensor has been exploited to demonstrate time-of-flight optical distance measurements. The distance measurement functionality of the MGM-PMD is verified experimentally in Figure 5, where Figure 5a represents the raw data and Figure 5b the mean values extracted from it. Here, the theoretical (expected) and the measured phase shift of the MGM sensor output are shown as a function of the distance d. As expected, the phase of the mixed signal is almost proportional to the distance, here shown for distances up to at least 1 m. The theoretical phase as a function of depth for ToF imagers has been calculated according to 37 where c is the speed of light, ϕ is the phase, and f m is the modulation frequency. The mean accuracy, 2 i.e., the deviation from the theoretical phase, is found to be 25.6 mm. The experimental results presented here clearly verify and demonstrate that the nonlinear behavior of 2D material graphene enables optical distance measurements by optoelectronic frequency mixing. The simple sensor design approach can achieve fill factors close to 100% whereas conventional time-of-flight detector concepts stagnate at 22%. 38,39 A huge benefit is the simple back-end integration on top of semiconductor electronics or flexible substrates, enabling additional sensor functionalities and performance on-chip, making the MGM-PMD a More than Moore device. 40 The detector presented here is thus a promising candidate to compete with alternative high-performance More than Moore devices in the future. The equipment used for distance measurement, such as function generator, lock-in amplifier, current amplifier, or stand-alone laser, can later be replaced by integrated devices. For example, if the differential frequency is fixed (defined by the two modulation frequencies f el and f opt ), the lock-in amplifier can be implemented by integrating either simple narrow bandwidth filters or more sophisticated phase sensitive detector (PSD) and low-pass filter below the sensor as part of the readout electronics.
In conclusion, intrinsic optoelectronic frequency mixing in graphene-based MGM-PMDs has been systematically investigated and utilized for optical distance measurements. The mixing effect thereby is enabled by the nonlinear output and transfer characteristics of the detector. The MGM-PMDs enable range measurements, hence 3D measurement capabilities, at modulation frequencies up to at least 3.1 MHz and mixing frequencies of 6 Hz in first proof of principle demonstrations. A distance of up to at least 1 m and an accuracy of ∼26 mm have been successfully demonstrated. A signal-to-noise ratio of ∼40 dB indicates that the current detection limit of 10 mW, corresponding to an irradiance of 328 mW/mm 2 , might be reduced in further experiments. Utilizing the innovative 2D material technology approach, sensor architectures and material compositions can further be developed towards fast, highly accurate, and sensitive distance measurement detectors and systems. The very simple device architecture and fabrication is scalable and allows for sensor integration on top of silicon electronics with fill factors close to 100%. This approach can enable significant performance   After that, source and drain areas have been defined in another lithography step. RF sputtering followed by a lift-off process has been applied to realize interdigitated Cr/Au (15 nm/120 nm) finger contacts. Single-layer graphene has been grown on top of a copper foil in a Moorfield NanoCVD-8G system by thermal assisted chemical vapor deposition. The graphene monolayers have been transferred by a conventional wet method using spin-coated poly(methyl methacrylate) (PMMA) as a support for the graphene. The 2D film has been separated from the growth substrate by etching the copper in sodium persulfate (Na 2 S 2 O 8 ). The PMMAsupported graphene has been transferred onto deionized water for cleaning and fished out with the prepatterned target substrate. The PMMA has been thoroughly removed by a multistep process using ultraclean solvents. In a next step, graphene channels have been patterned and etched by reactive ion etching with an oxygen plasma. Finally, a passivation layer of SiO 2 /Al 2 O 3 (5 nm/80 nm) has been deposited on top of the MGM-PMD graphene channel serving as a protective layer against ambient influences (the evaporated SiO 2 is required as a barrier layer to avoid damaging the graphene during the ALD process). Raman studies have been performed using a confocal microscope (TE 300, Nikon) and a laser (COMPASS 315M-100, Coherent) with a wavelength of 532 nm and a maximum excitation power of 100 mW. The signal has been collected through a 40× objective (NA 0.6, Nikon) and was detected using a monochromator (Triax 320, Horiba Jobin Yvon GmbH) and an EMCCD camera (Newton DU970P-BV, Andor Oxford Instruments).
Electrical device characterization comprises I−V measurements that have been performed using a SUSS microprobe station, connected to a Keithley 4200-SCS parameter analyzer.
Phase sensitive signals have been converted using a FEMTO DLPCA-200 I−V converter and amplifier and acquired via lock-in technique (SR865A, Stanford Research Systems). Distance measurements have been performed using an Omicrometer LDM405D.180.150 laser diode module with a peak wavelength of 405 nm. The power on the detector has been determined by a crystalline silicon reference detector (OP-2 VIS, Coherent) and a laser power meter (FieldMaxII-TO, Coherent).