High-speed programmable photonic circuits in a cryogenically compatible, visible-NIR 200 mm CMOS architecture

Recent advances in photonic integrated circuits (PICs) have enabled a new generation of"programmable many-mode interferometers"(PMMIs) realized by cascaded Mach Zehnder Interferometers (MZIs) capable of universal linear-optical transformations on N input-output optical modes. PMMIs serve critical functions in photonic quantum information processing, quantum-enhanced sensor networks, machine learning and other applications. However, PMMI implementations reported to date rely on thermo-optic phase shifters, which limit applications due to slow response times and high power consumption. Here, we introduce a large-scale PMMI platform, based on a 200 mm CMOS process, that uses aluminum nitride (AlN) piezo-optomechanical actuators coupled to silicon nitride (SiN) waveguides, enabling low-loss propagation with phase modulation at greater than 100 MHz in the visible to near-infrared wavelengths. Moreover, the vanishingly low holding-power consumption of the piezo-actuators enables these PICs to operate at cryogenic temperatures, paving the way for a fully integrated device architecture for a range of quantum applications.


Introduction
For many applications proposed for programmable optical systems, [1][2][3] the key requirements are (i) large scale , (ii) fast response --sub-µs for applications in machine learning, 4-7 quantum control, [8][9][10][11] and (iii) power efficiency to enable operation at cryogenic environment for integration with superconducting detectors 12,13 and artificial atoms. [14][15][16] Leading PMMI platforms 17 consist of cascaded MZIs to perform the general SU(N) unitary requiring N(N-1)/2 MZIs, 18 as illustrated in Fig.1a for a 4-mode programmable SU(4) in a rectangular mesh arrangement. 19 To meet the scalability requirement (i), for even modest mode count N, the most demanding technical challenge is reliable manufacturing and a clear path to electronics co-integration. As presently only modern CMOS-compatible VLSI processes offer these capabilities through foundry services, material choices are practically limited to Si 20,21 and silicon nitride technologies. 22 For the high-bandwidth requirement (ii) within these materials, silicon free carrier modulation 23 is problematic for PMMIs because of inherently coupled phase and amplitude. Silicon electric-field induced 24 modulators are promising but have not been realized at scale and restrict operating wavelength power. χ (2) A recently introduced alternative relies on piezo-optomechanical actuation of SiN waveguides, 25-28 which in the case of 25 has enabled narrow-band operation up to 250 MHz with SiN waveguides for visible-NIR operation and high optical power handling. Finally, the energy requirement (iii) is incompatible with any thermal modulation schemes, which dissipate an average of >800 mW for doped SiO2, 21 >20 mW per modulator for SOI 29 , and >40 mW per modulator for SiN. 17 Electrostatic MEMS devices meet the power requirement, 30,31 but modulation timescales have thus far been limited to the µs regime. Fig. 1 (b) summarizes this power-modulation bandwidth trade-space for SiN and Si platforms compatible with requirement (i). For cross-platform comparison, the figure plots the power dissipated for -phase shifts as a function π of circuit reconfiguration rate. In the thermal regime, the power dissipated is constant up to the maximum reconfiguration bandwidth, while piezo-actuation contains a trade-off between faster programming speeds and power dissipated on-chip. The piezo-optomechanical approach offers up to three orders of magnitude improvement in programming time with less power consumption as that of thermal and three to five orders of magnitude improvement in power dissipation at the maximum thermal reconfigurability rate, showing the exceptional suitability of this technology for PMMI systems. Motivated by these considerations, we advance our wafer-scale process for SiN PICs (Fig. 1c) with phase-only piezo-optomechanical tuners, whose basic design and operation are shown in Fig. 1d, e, through co-design and development of photonics, electromechanics, electrical and mechanical control systems, and driver software. We combine the best attributes of different devices in our previous work 25 into a proof-of-principle PMMI architecture capable of 20 ns reconfiguration time, on-chip dissipation per modulator below 200 µW when switching on average every 1 µs and 6 nW power to hold, operation at 700 -780 nm with optical transmission up to 1550 nm, and improved power efficiency at 5 K.

A. Device fabrication, design, and theory of operation
Our devices are designed and fabricated on a 200 mm wafer following an improved process flow based on our previous work. 25 The optical layer consists of a 300 nm thick silicon nitride (SiN) layer in a silicon dioxide cladding, located atop the AlN piezo-electric actuators with functionality for selective release and metal routing; see Supplementary Section 1 for more fabrication details. This process enables high-speed and broadband active modulators. Specifically, Fig.2a shows a full MZI with two strain-actuated phase-shifters internal and external to the directional couplers, as in the unfolded layout of Fig.1d. Each phase-shifter consists of a 400 nm width SiN waveguide passing through an adiabatic taper, expanding to 5 µm width for improved strain transfer, 25 and then through a second adiabatic taper back down to 400 nm-wide single-mode turn. This loop repeats nine times as shown in Fig. 2b. The cross-section in Fig. 2c, d shows a slightly undercut 10 µm-wide pillar around the oxide-clad 5 µm-wide SiN waveguide and AlN actuators. Applying a potential difference across the AlN transfers strain to the waveguide (see finite-element simulation in Fig. 2e), imparting a refractive index shift predominantly due to the strain-optic effect 32 in addition to moving boundary effects. 33 When applying a potential difference, the strain imparts a phase change to the optical mode for a given length L ∆θ phase-shifter is , where is the free-space wavelength, is the static effective mode index, is the strain-induced path-length λ ∆ change of the phase-shifter, and is the total effective strain-induced refractive index change: ∆ signed quantity depending on the direction of the induced strain, meaning opposite strains produced by a negative applied voltage produce an opposite-shift effective index. This effect allows for a push-pull or differential operation of the phase-shifts in Fig. 2a for an applied , defined as the Θ potential relative to ground.

B. Room temperature operation
Fig . 3 summarizes the modulator performance at room temperature (300 K). Fig. 3a, b show operation across a broadband range of wavelengths (700 nm-780 nm). The plots shown here take advantage of the push-pull operation for optical loss balance (-3.5 dB typical insertion loss at 737 nm, see Supplementary Section 2 for loss characterization details) of the phase-shifters by applying opposite polarity voltages to each arm. We estimate the from sinusoidal fits, yielding values from 50 cm (700 nm) to 65 cm (780 nm) in the push-pull π · · configuration or a corresponding 100 cm to 130 cm per single phase-shifter. We attribute the variance in · · π to material and waveguide dispersion. This cutoff is consistent with the RC constant during measurement, consisting of the device capacitance of 17 pF and series resistances of ~80 Ω arising from the voltage source and on-chip routing metal. We attribute the peak at 241 MHz to the fundamental mechanical resonance, which our finite-element model puts at 253 MHz (Fig. 2e). The shape of the peak exhibits a characteristic resonance and anti-resonance feature typical of that in mechanical resonators. 34,35 We note that the AWG used for these tests does not produce the full 20 V swing across its entire bandwidth (500 MHz) which accounts for the atypical falling edge and distorted Gaussian envelope seen in Fig. 3e, d respectively. The contrast in these measurements is low, as this device is optimized for cryogenic operation, as described next.

C. Cryogenic operation at 5 K
We performed the cryogenic characterization in a 4 K closed-cycle cryostat (Montana Instruments). Using a fiber array and RF probe mounted on nanopositioners inside the cryostat (see Supplementary section 5), we applied various voltage waveforms. The transmission curves in Fig. 4 show no degradation in switching performance at a base temperature of 5 K (Fig. 4a, b) compared to room temperature. Fig. 4c indicates an extinction of approximately 30 dB. on the minutes time-scale. The corresponding holding-power of 2 nW is nearly 8 orders of magnitude smaller than the Montana cryostat cooling power (90 mW). This low power consumption is a critical benefit, as even state of the art thermal SiN WG modulators (Fig. 1b) would easily overwhelm the cryostat's cooling power. We note the actual extinction ratio of the modulator is larger than 30 dB; however, limited volume in the cryostat prevented the use of differential driving as demonstrated at room temperature and limited the phase change to approximately , rather π/2 than . π

D. 4x4 programmable interferometer
Having characterized the MZIs, we now consider a proof-of-principle SU(4) programmable unitary transformation, which consists of N(N-1)/2 = 6 MZIs. The PMMI is fully programmable, with all phase-shifters computer-controlled (see Supplementary Section 3).  Fig. 5c) here for each plotted MZI show maximum power differences when actuating -a full data set of all twelve phase-shifters is shown in Supplementary Section 4. Fig. 5d plots the transmission of an exemplary MZI: "MZI3," which indicates an optical extinction ratio in excess of 40 dB. The total loss through the PMMI ranges from -14 dB to -21 dB at 737 nm, depending on the optical path taken, primarily due to waveguide bending losses in the phase-shifters, which could be completely eliminated with an unfolded implementation or larger radii waveguide bends. We confirm that our PMMI maintains the 100 MHz bandwidth as demonstrated in the single MZI devices by applying a 10 V, 100 MHz sinusoid to the internal phase-shifter of MZI2 and monitor the optical outputs 5, 6 with a laser coupled through optical input 3. We utilize a lock-in detection scheme on the output photodiodes to extract the signal at 100 MHz (Fig. 5e, f) for both channels and a relative phase offset of 0.59π; the small deviation from π/2 is likely due to the 125 MHz bandwidth of the photodiode being close to the driving frequency. Although results in Fig. 5 demonstrate the programmability of our PMMI as a single packaged device, voltage handling limitations of our on-chip vias (see Discussion section below) currently prevent accessing the full scope of U(4) operations.

Discussion
We have demonstrated a 4-mode programmable interferometer in a 200 mm wafer-scale CMOS process. The PMMI is comprised of cryogenically compatible, >100 MHz bandwidth piezo-optomechanical phase-shifters operating in the visible to NIR wavelengths, which should be immediately applicable to many hybrid photonic architectures, 16 quantum protocols, 8,36 and machine-learning algorithms. 6 While the devices presented here show promise for universal linear-optic programmable photonics in new performance regimes and application spaces, there are still some non-idealities that will be improved in future work. First and foremost, the dominant source of loss in all devices presented is due to bends in waveguide meanders and directional couplers, measured to be -2.7 dB of the -3.5 dB MZI insertion loss (see Supplementary Section 2). Waveguide propagation losses (-0.39 dB/cm) and adiabatic taper losses -0.022 dB + 0.038 dB per taper, account for the remaining -0.8 dB of loss. We estimate that implementing less tightly folded versions of the same devices that have fewer bends and tapers would allow us to achieve MZI insertion losses significantly less than -1 dB. Second, in these particular devices, small voids in the AlN in the vicinity of electrical vias produced metal filamentation that degraded the device breakdown voltage and prevented application of voltages above 25 V. While we found = individual vias in test structures on the same wafer that allowed 90 V , the yield was insufficient for an MZI > comprising 90 vias . After systematic failure analysis and cross-sectional imaging, the process yield improved greatly so that full SU(N) devices appear likely with 90 V. Moreover, high-scandium-concentration Sc (1-x) Al x N > has shown a five-fold increase in the piezoelectric responsivity, 37 promising a further reduction in V π .

×
Finally, the layer-stack of our PIC (Table S1 and Fig. 2c), which places the optical layers on top, enables many extensions to the platform. The optical layer, being above all metal and piezoelectric layers, allows for alternative optical layers to expand the transparency window into the UV regime. [38][39][40] Moreover, heterogeneously integrated thin films for either photon generation, 14,41,42 detection, 13 or nonlinear interactions [43][44][45] could be placed directly on top of the optical layer with photons coupling evanescently to the rest of the integrated photonics. The AlN-actuator based photonic platform is also entirely post-CMOS compatible and can thus be directly fabricated on fully formed CMOS integrated circuits 46 for direct biasing or control of hybrid systems; 47-49 thus, the architecture presented here can be scaled to both very large photonic circuit sizes and very small photonic circuit pitches without electrical control bottlenecks.

Holding power consumption and reconfiguration energy calculations.
We first calculate the device capacitance and on-chip routing metal resistance based upon the measured RC roll-off in our frequency response curve, finding pF and 30 Ω in addition to a series voltage-source resistance of 50 Ω. The device leakage resistance = 17 ℎ = based on previous measurements 25 is estimated to be 500 MΩ and 200 GΩ at 300 K and 5 K respectively. The holding power is found for two phase-shifters holding 25 V to maintain a π phase-shift. The reconfiguration = energy dissipated on-chip is calculated via the formula where is the total series resistance = ( ℎ / ) 2 of the circuit including external resistors in the AWG and V. This equation accounts for two phase-shifters = 50 with a 50 V swing. To reduce on-chip energy dissipation for slower reconfiguration rates, is adjusted such that the corner frequency matches the current reconfiguration frequency (which is done off-chip at the 1/(2π ) voltage source). This minimizes the on-chip energy dissipation while maintaining the necessary reconfiguration rate. The total power dissipated (plotted in Fig. 1b) is simply the holding power plus the reconfiguration energy times the reconfiguration rate.
Device characterization. We characterize individual MZI modulators both at room and cryogenic temperatures with 250-um pitch optical fiber arrays grating coupled to the on-chip waveguides. We use a 150-um pitch RF probe (in GSGSG configuration for room temperature, GSG for cryogenic temperature) to apply high-frequency electrical signals.

Digital lock-in amplifier.
A time-trace of both output channels 5 and 6 of the PMMI is digitized directly by a high-speed oscilloscope. The time-traces are digitally integrated (for 4 ms) with sinusoids of varying frequencies and phases, whose resulting amplitudes form the data for the plots in Fig. 5e, f. The two photodiodes used for both channels had 125 MHz and 600 MHz bandwidths respectively.

Data availability.
The data that support the plots within this paper are available from the corresponding authors upon reasonable request.
Please see Supplementary Information for more detailed experimental methods. We fabricated all devices at Sandia National Laboratories. The nominal film thicknesses with labels corresponding to Fig. 2c are listed in Table S1. Starting with an oxide film on silicon, we sequentially deposited and patterned the M1 routing (Al/Ti), amorphous silicon (a-Si) release layer, tungsten vias, M2 bottom electrode, AlN, M3 top electrode, bottom oxide cladding, SiN, and top oxide cladding. Several chemical-mechanical polishing steps were introduced to ensure wafer flatness and more uniform film thicknesses, especially before critical layers such as M2 for AlN growth and SiN for reliable waveguide etching. We then performed an oxide etch to open all metal pads for probe or wire-bond access. To complete the fabrication, we etched deep trenches through all layers until reaching the a-Si to define all piezo-optomechanical structures, then subsequently performed a XeF 2 release, etching away the a-Si and undercutting all devices.

High-speed programmable photonic circuits in a cryogenically compatible, visible-NIR 200 mm CMOS architecture: Supplementary Information
We note several major improvements in our fabrication since our earlier work: 25 all processes have been recalibrated to operate on 200 mm wafer technology (from previous 150 mm); we greatly improved the deep trench process and utilized narrower trenches for flatter device topography; and we added additional CMP processes for more uniform definitions of pillars and trenches.

Optical loss characterization
We characterize the propagation loss in our single-mode SiN waveguides with simple ring-resonators by extracting quality factors at both visible (780 nm) and NIR (1550 nm) wavelengths. As seen in Fig. S1, we estimate the intrinsic Q to be ~1.6 million at 780 nm and ~30000 at 1550 nm, corresponding to propagation losses of -0.39 dB/cm and -9.1 dB/cm respectively. are used to allow custom fiber array holders to edge couple an 8-channel fiber array to our 4x4 PMMI, as shown in Fig. S4. The edge coupling is currently unoptimized, with efficiencies of -15 dB per facet.
Electronic control of the PICs is done using two Analog Devices 5535B Evaluation Boards. These boards use the AD5535B 32-Channel, 14-bit DAC with on-chip high voltage output amplifiers which are capable of amplifying each channel up to 200 V at up to 550 µA of driving current. The evaluation kit routes all channels to an IDC40 40-pin connector, so an IDC40 ribbon cable connects the evaluation boards to a custom chip carrier PCB. The two AD5535B Evaluation Boards are each connected to the Analog Devices SDP-S System Demonstration Platform board, which allows voltages to be set on the AD5535B DAC over Serial Peripheral Interface (SPI) and is connected to a control laptop via a mini-USB cable. Power is supplied to the board from a 5 V supply for digital logic and low-voltage analog power and a 200 V supply for the amplifier. When set to a maximum output voltage of 200V, the AD5535B will have a minimum step size of 12.5 mV. The channels can exhibit crosstalk of up to 4 volts, so a delay of 250 ms is used after setting a voltage to allow signals and crosstalk effects time to settle before making a measurement of the optical response.
Custom python drivers are written to set voltages on each channel from a control laptop, and an additional layer of software is written to allow a user to set voltages with respect to the individual devices on the 4x4 PMMI. We run top-level scripts utilizing these drivers to sweep the actuation voltage of devices while measuring the response on an optical power meter.
A custom printed circuit board and chassis are designed to hold the 4x4 PMMI and secure it for edge coupling on an optics table. The PCB routed control voltages and ground signals from a 2x20 pin connector to wirebond pads, which were then wirebonded to the PIC. The PIC was placed on a copper mount in the middle of the board.
that would maximize the power output of one measured channel before sweeping the voltage on one specific phase shifter. These values are recorded as the "optimized" power measurements, with a comparison shown in Fig. S6. Characterization sweeps are limited from -15V to +15V to avoid the breakdown of the vias at high voltages. We note that when changing the input or output channels, the optical test setup is realigned to maximize the measured power before collecting new baseline and optimized measurements.

Cryogenic test setup
We perform cryogenic testing of our photonics in a Montana, closed-cycle cryostat equipped with a three nanopositioner stack. The PIC is cooled down to ~5.1 K and optical and electrical I/O are provided with a fiber array and RF probe (Fig. S7a). The optical and electrical connections are routed through the chamber via built-in feedthroughs ( (Fig. S7b). The monitored temperature remained at roughly 5.1 K throughout the entire duration of the cryogenic characterization measurements, including DC and AC actuation tests.