An integrated silicon nanophotonic coherent imager (NCI), with a 4 × 4 array of coherent pixels is reported. In the proposed NCI, on-chip optical processing determines the intensity and depth of each point on the imaged object based on the instantaneous phase and amplitude of the optical wave incident on each pixel. The NCI operates based on a modified time-domain frequency modulated continuous wave (FMCW) ranging scheme, where concurrent time-domain measurements of both period and the zero-crossing time of each electrical output of the nanophotonic chip allows the NCI to overcome the traditional resolution limits of frequency domain detection. The detection of both intensity and relative delay enables applications such as high-resolution 3D reflective and transmissive imaging as well as index contrast imaging. We demonstrate 3D imaging with 15μm depth resolution and 50μm lateral resolution (limited by the pixel spacing) at up to 0.5-meter range. The reported NCI is also capable of detecting a 1% equivalent refractive index contrast at 1mm thickness.
© 2015 Optical Society of America
Three-dimensional (3D) imaging has many applications in bio-medical imaging, nano-particle characterization, security, robotics, and gesture recognition. Although 3D imaging systems have been studied, they are usually realized either as single pixel detectors [1–4] or as bench-top systems with simple detector arrays [5–7]. Silicon nanophotonic processes use similar steps as their electronic counterparts, leading to higher yields, and compatibility with electronic integrated circuits. This combined with the low loss and high confinement of silicon waveguides  make the silicon photonics an ideal platform for coherent 3D imaging and projection systems [9–11] with high resolution and pixel count.
Infrared (IR) optical electromagnetic waves are promising candidates for high-resolution 3D imaging due to their short wavelength. These wavelengths can be processed efficiently by integrated silicon-on-insulator (SOI) nanophotonic platforms with high confinement and relatively low loss. A recent demonstration of phased array transmitter [9, 11] is an example of the advantage of SOI platforms in implementing complex nanophotonic systems. The high fabrication yields and potential for integration with silicon electronics necessary for additional signal processing make SOI nanophotonic processes a suitable platform for realization of NCIs with large number of pixels in a small area at low cost.
In this paper we report a nanophotonic coherent imager that uses an array of integrated silicon photonics antennas to couple the impinging infrared electromagnetic wave into a series of nanophotonic waveguides, which coherently process the optical signals to measure the instantaneous phase and amplitude of the incident wave on each pixel to form a 3D image of the target object. An electrically generated frequency chirp is used to produce a linearly chirped optical wave that illuminates the object. A part of this optical signal serves as a reference and is coherently combined with the received signal by each pixel on the silicon nanophotonic chip to produce an electrical output. The proposed NCI uses time-domain signal detection and processing of this electrical output to resolve the relative phase and power of the incident wave at each pixel. This simultaneously enables high depth resolution and large dynamic range, overcoming some of the limitations of conventional frequency domain FMCW approaches. The NCI chip can be used in 3D reflective, transmissive, and index contrast imaging modes with a high spatial resolution of about 15μm over a large dynamic range of 0.5m without use of a wide range highly linear tunable laser source. To our knowledge, this is the first coherent imager that enables multiple phase sensitive imaging schemes via a single nanophotonic chip.
2. Coherent imager architecture
The structure of the 4 × 4 NCI is depicted in Fig. 1. A part of the laser output that is used to illuminate the target object being imaged is coupled into the input grating coupler through an optical fiber and then is guided to a Y-junction splitter network through silicon nano-waveguides. This coupled coherent light, the reference signal, is split into 16 equal intensity optical signals. The incoming light from the object is coupled into nano-waveguides through an array of grating couplers  serving as pixels. The received light from each pixel is then combined with a part of the reference signal in a Y-junction combiner  and is converted to electrical current using a silicon-germanium photo-detector . Figure 1(b) shows the NCI micro-photograph implemented on IME SOI photonic process . The implemented NCI can be used in the transmissive, reflective, and index contrast imaging modes.
2.1. Theory of operation
Figure 2(a) shows the imaging system for ith grating coupler. The output electric field of the laser emitting with power P0 at angular frequency ω0 is represented by . The laser output is intensity modulated with a frequency-chirped electrical voltage v(t) using a Mach-Zehnder modulator (MZM). The use of an electrically generated frequency chirp eliminates the need for a tunable laser source with highly linear chirp response. An erbium-doped fiber amplifier (EDFA) is placed after the modulator. The parameter G accounts for the gain of the EDFA and the loss of the MZM. The EDFA output is split into two branches using a fiber optic fusion coupler. The light in the top branch is collimated and used as the illumination source. Part of the light coming from the target object is coupled into the NCI chip by the ith grating coupler. The light in the bottom branch, the reference signal, is also coupled into the NCI chip. The coupled reference signal is split into 16 branches and combined with the signal from the ith grading coupler in a Y-junction. The Y-junction output is then photo-detected and the photo-current i(t) is wire-bonded to the electronic circuit. The grating coupler efficiency and the on-chip waveguide loss in the top branch are represented by γ, while β accounts for the excess loss of the 1-to-16 on-chip splitter and the grating coupler efficiency. The total delay difference between the top and the bottom arms, τ0, consists of the delay in the optical fiber, τof, and the free-space delay, τf. As shown in Fig. 2(a), in the transmissive mode, the target object which is substantially transparent at the operating wavelength is placed between the collimator and the grating coupler. In this case, τf increases as the light slows down in the target object. In the reflective mode, the light is reflected from the object and forms an image on the pixel array (not shown). For this case, τf contains information on the spatial depth of the target object.
In order to study the principle of operation of the imager, it is useful to consider the case that the collimator is directly illuminating the pixel array and therefore, τf represents the free space delay and depends on the distance between the collimator and the pixel array. In this case, the electric field of the combined signal, Eout, can be written asEq. (2) into Eq. (1) and using the Jacobi-Anger expansion (i.e., ), the DC and AC components of the photodiode current can be written as Equation 5 shows that the frequency of the detected signal is a linear function of the delay difference between the received signal and the optical reference signal. This detection scheme is similar to that of a FMCW radar , since the optical reference signal and the detected signal at each pixel will be at different optical frequencies due to the difference in their propagation delays. Therefore, different time delays between the reference signal and the receive signal at each pixel results in different electrical frequencies detected at corresponding photodiode of the NCI chip.
Defining the frequency sweep range, Δfc = αT, and using Eq. (5) the period of VT (t) can be calculated asFig. 2(b). In case 1, the total delay difference between top and bottom arms in Fig. 2(a) is set to τ0. In case 2, the distance between the collimator and the ith grating coupler of the NCI is slightly increased, such that this total delay difference increases to τ0 + Δτ. For case 1 and case 2, electrical signals VT1(t) and VT2(t) with different frequencies appear at the output of the filter, respectively. By knowing the chirp rate α and counting the number of zero crossings in VT1 or VT2, a coarse estimate for the delay difference τ0 can be achieved. However, the amount of the small delay difference between cases 1 and 2, Δτ, can not be estimated by counting the number of zero crossings of VT1 or VT2. In order to find a fine estimate for Δτ, the time difference between the last zero crossing in VT1(t) and VT2(t) and the end of the frequency ramp (i.e., ΔT1 and ΔT2) can be used. First, from Fig. 2(b) the chirp period, T, can be written in terms of the periods of VT1 and VT2 as Eq. (6) the difference between the periods of VT1(t) and VT2(t) can be written as Fig. 2(b), is calculated. In practice, τ0 is much smaller than the chirp period T and hence, is assumed which together with Eqs. (6), (8), and (9) can be used to find Δτ as
One important challenge in the implemented NCI is that the phase difference between the top and bottom arms in Fig. 2(a), ϕ, may vary slowly with time due to the presence of a thermal gradient between two arms. This results in a slow amplitude fluctuation for the detected signal as Eq. (3) suggests. However, the DC component of the photo-current in Eq. (4) also experiences the same thermal fluctuation and is used to form a phase correction loop by placing a thermal phase modulator on the optical reference signal path. This feedback loop together with an automatic gain control (AGC) loop in electronic detection circuit keeps the amplitude of the detected electrical signal constant. Also, note that the variations in the term ( ) in Eq. (5) are negligible compared to ΔT1 (or ΔT2) and do not affect the performance of the reported imager.
2.2. Transmissive imaging mode
The block diagram of the transmissive measurement setup is illustrated in Fig. 3(a). A laser source is intensity modulated with a frequency-chirped electrical signal. In transmissive imaging mode, the collimated beam passes through the object and impinges on the NCI chip. The generated photo-current resulting from beating the reference optical signal and the received signal by each pixel is amplified using an array of TIAs, filtered, and captured by the detection system. In this mode, the object being imaged is substantially transparent at the laser operating wavelength. The object is placed between the collimator and the NCI chip. In this mode, a motorized micro-positioner moves the object by equivalently 16-pixel increments to emulate a larger pixel array. At each position, different parts of the collimated beam that are passing through the parts of the object with different spatial depths experience different time delays when arriving at the corresponding pixels of the NCI chip.
As discussed in section 2.1, the coarse spatial depth of the object being imaged is automatically calculated by counting the number of zero-crossings per chirp period and the fine spatial depth of the object is calculated by measuring the time difference between the last zero-crossing in one chirp period to the end of the chirp period. By combining these coarse and fine depth detection schemes, a very fine depth resolution over a large spatial dynamic range can be achieved.
2.3. Reflective imaging mode
In the reflective measurement setup, a collimator is used to coherently illuminate the object and the image is formed on the pixel array using a lens, as depicted in Fig. 3(b). The detection scheme is identical to that of the transmissive mode. However, the optical power at the output of the collimator is increased to improve the signal-to-noise ratio at the input of the electronic detection system. In this mode, to emulate a larger pixel array, the pixel array is moved using an automated micro-positioner with increments of 4 × 4 pixels while keeping the lens stationary.
3. Measurement results
The NCIs were fabricated in IME silicon-on-insulator technology node  where silicon-on-insulator wafers with a 0.22μm top silicon layer and 2μm buried oxide are used. Three levels of silicon etching were available. The pixels are lensed grating couplers with effective area of 17μm×17μm and with 40% average coupling efficiency that were designed using two levels of etching. Travelling wave photodiodes were designed based on epitaxial growth of germanium on top of silicon waveguides and have measured responsivity of 0.7A/W. An average optical loss of 5loss of 1.8dB/cm for a 500nm wide and 220nm thick silicon waveguide were measured. Two aluminum metal layers with thickness less than 1.5μm were available. No post processing was done on the fabricated chips. The fabricated chips were mounted on a printed circuit board (PCB) and electrically connected to the PCB using bond wires while the optical fiber carrying the reference optical signal was attached to the reference grating coupler. No temperature control was required. All measurements have been carried out at the room-temperature.
For all measurements, a laser source emitting at λ0=1.55μm is intensity modulated with a frequency-chirped electrical signal. The frequency of the electrical signal increases linearly with time from 1MHz to 3.8GHz, over 500μs, periodically. To measure the minimum depth resolution in the transmissive mode, first the collimator output power is set to 10mW. In this case each grating coupler receives about 4μW. Then, the free-space distance between the collimator and the NCI is changed using the motorized micro-positioner. For each position, 10 measurements per pixel were conducted. The measured distance vs. actual distance is depicted in Fig. 4(a). The error bars on this graph represent the range of all 160 measurements for each distance setting. The highest range of error occurs for the 100μm distance measurements (shown as an inset) where a 15μm deviation from the actual value can be observed corresponding to a 15μm depth resolution. This depth resolution is in close agreement with the predicted value from Eq. (11), for (ΔT2 − ΔT1)min ≈ 5ns, τ0 ≈ 5ns, and T = 0.5ms.
In a different experiment, a 40mm thick Plexiglas cube was placed next to two small pedestals with approximate thickness of 150μm and 225μm, respectively (Fig. 4(b)-left). The 3D transmissive image shown in Fig. 4(b)-right demonstrates that a high spatial resolution over a large dynamic range can be achieved with a single measurement and without modifying the measurement setup and parameters. In another experiment, the 3D transmissive image of a hollow pyramid made of Plexiglas (Fig. 4(c)-top) was formed which is depicted in Fig. 4(c)-bottom. For all transmissive measurements, the output power of the collimator is set to 10mW. Due to scattering from abrupt edges of the target object, the SNR of the received signal at corresponding pixel may be reduced significantly resulting in inaccurate zero-crossing detection. To avoid this uncertainty, the pixels with low SNR are tagged automatically during the imaging stage and in the post-processing stage, their values are automatically replaced by the average of the surrounding pixels. Figure 4(d) shows the refractive index contrast imaging that is conducted using the transmissive imaging setup. In Fig. 4(d)-top, a sheet of Polycarbonate with typical index of 1.56 is placed next to a sheet of Plexiglas with typical index of 1.47. Both sheets have thickness of 1mm. The image in Fig. 4(d)-bottom shows a clear difference in the detected thickness corresponding to the contrast in the index of refractions of two materials.
In reflective mode measurements, the collimator output power is set to 120mW and each pixel receives about 0.2μW in presence of the lens. Fig. 4(e)-bottom shows the reflective depth image of a US one-cent coin (in Fig. 4(e)-top) where in the setup in Fig. 3(c), a micro-positioner is used to move the 4 × 4 NCI with increments of 4 × 4 pixels (without moving the lens) to mimic a larger coherent pixel array. The 3D image in Fig. 4(e)-bottom has a depth resolution of about 15μm and lateral resolution of 50μm limited by the on-chip pixel spacing.
We have demonstrated a nanophotonic coherent imager implemented on a silicon photonics platform where all optical signal processing and detection is done on a single photonic chip. Use of low-cost and robust silicon photonics platform combined with modified FMCW detection scheme enables 3D imaging with high depth resolution of 15μm over a large dynamic range of 0.5m. Using the implemented integrated NCI, we have conducted high resolution index of refraction contrast imaging, 3D transmissive, and 3D reflective imaging. Use of silicon photonics platform enables realization of NCIs with large number of pixels making low-cost high performance NCIs available for varieties of applications in near future.
We thank OpSIS foundry for the fabrication of the chips and Profs. Micheal Hochberg and Tom Baehr-Jones from University of Delaware for valuable discussions on the fabrication process.
References and links
2. A. McCarthy, X. Ren, A. D. Frera, N. R. Gemmell, N. J. Krichel, C. Scarcella, A. Ruggeri, A. Tosi, and G. S. Buller, “Kilometer-range depth imaging at 1550 nm wavelength using an InGaAs/InP single-photon avalanche diode detector,” Opt. Express 21(19), 22098–22113 (2013). [CrossRef] [PubMed]
3. M. D. Learmouth, “High resolution laser lidar utilizing two-section distributed feedback semiconductor laser as a coherent source,” Electron. Lett. 26(9), 577–579 (1990). [CrossRef]
6. R. M. Marino, T. Stephens, R. E. Hatch, J. L. McLaughlin, J. G. Mooney, M. E. O’Brien, G. S. Rowe, J. S. Adams, L. Skelly, R. C. Knowlton, S. E. Forman, and W. R. Davis, “A compact 3D imaging laser radar system using Geiger-mode APD arrays: system and measurements,” Proc. SPIE 5086, 1–15 (2003). [CrossRef]
7. D. P. Hutchinson, “Coherent infrared imaging camera,” Proc. SPIE 2540, 204–209 (1998). [CrossRef]
8. M. Hochberg and T. Baehr-Jones, “Towards fabless silicon photonics,” Nature Photon. 4, 492–494 (2010). [CrossRef]
9. B. Abiri, F. Aflatouni, A. Rekhi, and A. Hajimiri, “Electronic two-dimensional beam steering for integrated optical phased arrays,” Optical Fiber Communication Conference (OFC), paper M2K.7 (2014).
10. F. Aflatouni and H. Hashemi, “An electronically controlled semiconductor laser phased array,” International Microwave Symposium Digest (IMS), 1–3 (2012).
13. K. W. Ang, T. Y. Liow, Q. Fang, Jun-F. Song, Y. Z. Xiong, M. B. Yu, G. O. Lo, and D. L. Kwong, “Low thermal budget monolithic integration of evanescent-coupled Ge-on-SOI photodetector on Si CMOS platform,” IEEE J. Sel. Top. Quantum Electron. 16(1), 106–113 (2010). [CrossRef]
14. A. Novack, Y. Liu, R. Ding, M. Gould, T. Baehr-Jones, Q. Li, Y. Yang, Y. Ma, Y. Zhang, K. Padmaraju, K. Bergmen, A. E. J. Lim, G. Q. Lo, and M. Hochberg, “A 30 GHz silicon photonic platform,” in proc. SPIE8781 (2013). [CrossRef]
15. I. V. Komarov and S. M. Smolskiy, Fundamentals of short-range FM radar (Artech House, 2003).