Monolithic Electronic-Biophotonic System-on-Chip for Label-Free Real-Time Molecular Sensing

Abstract

Label-free miniaturized optical sensors can have a tremendous impact on highly sensitive and scalable Point-of-Care (PoC) diagnostics by monitoring in real-time molecular interactions without any labels. However, current biophotonic platforms are limited by complex optical and external readout equipment, precluding their use in a PoC setting. In this work, we address this challenge by developing a first-of-its-kind fully integrated electronic-photonic label-free molecular sensor utilizing micro-ring resonators (MRRs) co-integrated with on-chip electronics in a high-volume advanced electronic process. In particular, we present an arrayed electronic-photonic system-on-chip (EPSoC) in GlobalFoundries (GF) 45-nm RFSOI with sixty 5-um radius MRRs connected to on-chip receivers, approaching an idealized limit of detection (LoD) equivalent to a single 140-nm viral particle. To further enhance the LoD, we propose a dual-ring phase-based sensing architecture, boosting the system’s sensitivity by 3.7x compared to our previously reported intensity-based single MRR scheme. An integrated heater embedded in the design of the ring and an on-chip controller lock the ring’s resonance at the desired point of operation, eliminating the need for a tunable laser. The inherent intrinsic limitations of MRRs due to ambient temperature variations are addressed with an on-chip differential scheme using sensing and reference rings to cancel common mode errors. We demonstrate the sensing capabilities of the EPSoC by monitoring in real-time binding events of Bovine Serum Albumin (BSA), anti-BSA molecules and streptavidin-coated nanoparticles, unlocking the door towards self-contained fully integrated Lab-on-Chip (LoC) photonic sensors for PoC applications.

I. INTRODUCTION

Immediate and widely accessible Point-of-Care diagnostic technology is the key for halting rapidly spreading and potentially fatal pathogens. A multiparametric, highly scalable, portable, and sensitive biosensing platform monitoring complex dynamic molecular parameters will transform our ability to provide early diagnosis and understand a disease’s progress by supplying accurate, multidimensional and on-demand information [1], [2]. Current biosensing platforms heavily rely on labeling techniques [3], [4]. However, despite their accuracy, these label-based technologies require clinical laboratories and infrastructure not easily deployed in a PoC setting. In addition, they are unable to provide real-time information about molecular interactions such as binding kinetics, which is key for continuous monitoring of dynamic disease processes. Therefore, a label-free sensing approach is critically needed.

Refractive Index (RI) chip-scale biophotonic sensors have shown increasing promise over the last decade in monitoring real-time label-free molecular interactions [5]–[10]. However, despite the label-free operation, to date, evanescent field biosensors have been hindered by the need for complex, bulky optics and external readout equipment precluding both mass production and miniaturization towards PoC testing, thus remaining confined to special laboratory settings [11]–[15]. In this work, we address these challenges by using a first of its kind electronic-photonic label-free sensor with nanophotonic micro-ring resonators (MRRs) and electronics on the same chip of a high volume CMOS 45-nm advanced electronic process. The deep integration of planar nanophotonic ring resonators with on-chip electronics unlocks the door towards smart, monolithic, and scalable lab-on-chip (LoC) photonic devices, capable of monitoring multi-molecular dynamics in real-time and in a PoC setting, as illustrated in Fig. 1.

Fig. 1.

Overview of the fully-integrated PoC platform for multi-molecular sensing. (a) Body samples are loaded on a cartridge, which plugs into a wearable device. (b) Different molecules run through multiple fluidic channels, co-packaged on a monolithic electronic-photonic Lab-on-Chip system. (c) Each channel of the fluidic network attached to the EPSoC has photonic sensors functionalized with different biorecognition molecules. Multi-analyte detection provides unique fingerprints of disease progression over time.

Advances in label free nanophotonic molecular sensing have made significant strides, employing a wide range of highly sensitive optical transducers [16]–[18]. Interferometric biosensors, which rely on splitting the light into a sensing and reference waveguide [19]–[26], have gained increasing interest due to their high sensitivity, reporting limits of detection of ≈10⁻⁸ RI Units. However, this occurs at the cost of large mm-scale sensing transducers and therefore large chip area.

The dependence of sensitivity on the waveguide’s length can be circumvented through the use of miniaturized resonant-based sensors which rely instead on the sharpness of their resonant spectrum, expressed by the quality factor ($Q$) of the transducer [27]. Numerous resonant-based biosensors have been introduced over the last years demonstrating real-time molecular kinetics and multiplexed sensing of different analytes [7], [9], [13], [28]–[33]. However, despite the significant promise of these efforts, most of the focus has been on optimizing the sensitivity of the transducers resulting in complex optics - in some occasions non planar [7] - therefore precluding their fabrication in a high volume commercial process. Additionally, most prior work relies on sweeping a bulky tunable laser to interrogate the sensors, therefore precluding miniaturization and setting a limit on the measurement speed which is restricted by the sweeping rate of the laser. Another major bottleneck towards the large-scale commercialization of passive photonic platforms is the need for an off-chip bulky readout equipment, thus hindering the development of a true monolithic and compact LoC system.

In this work, we solve these challenges by relying on a key breakthrough technology that allows zero-change co-integration of planar 5 μm radius MRRs with high performance state of the art on-chip transistors. To overcome the aforementioned limitations, we effectively shift the cumbersome requirements for complex optics into the electronic domain by introducing several key innovations.

First we eliminate the need for a bulky tunable laser and external readout equipment by relying on the direct integration of on-chip electronics with planar MRRs. Integrated heaters actively tune the ring’s resonance at a fixed wavelength, enabling co-packaging of single $\lambda$ low cost light sources. The optical sensing information is processed with on-chip receivers connected to the MRRs, drastically miniaturizing the LoC system. Benefiting from the slow nature of molecular kinetics and the averaging of high frequency noise sources, we approach an idealized LoD (without biosamples) equivalent to a single 140-nm viral particle.

Second, we enhance the system’s sensing performance by introducing various photonic detection schemes. Single MRR-based photodetectors can be used as sensing transducers eliminating the need for a separate sensor and photodetector. However, to leverage the use of higher $Q$ sensors we also implement dual-ring phase-based sensing using a ring-assisted interferometric architecture that presents higher sensitivity [34]. Additionally, we address the inherent limitations of MRRs due to environmental drift, using an on-chip differential common mode cancellation scheme, comprised of sensing and reference rings. In order to further suppress ambient large RI changes, we implement an on-chip tuning controller that locks the MRR at the desired point of operation.

We develop a highly scalable arrayed EPSoC, co-packaged with multi-channel microfluidics for monitoring molecular dynamics. We demonstrate real-time molecular kinetics of anti-BSA, BSA molecules and streptavidin-coated nanoparticles, unlocking the door towards multiplexed sensing in a high volume commercial disposable platform.

The remainder of this paper is organized as follows. In Section II, we describe the sensing principle of MRR-based label-free photonic sensors. In Section III, we introduce the fully integrated electronic-photonic platform in 45-nm RFSOI, describing the process, the MRR-based transducer and the sensing circuit implementation. The system architecture is discussed in Section IV. Finally, in Section V, we demonstrate the experimental results of the EPSoC.

II. LABEL-FREE PHOTONIC SENSORS

A. Overview of the Evanescent Field Sensing Principle

Label-free sensing with silicon photonics (SiPh) has shown increasing promise over the last two decades in monitoring real-time molecular kinetics without requiring complex labeling techniques. The key operating principle is based on evanescent field sensing (Fig. 2a). The optical mode propagating through an integrated waveguide interacts with any material placed on the surface of the photonic platform within a decaying tail of hundreds of nm range. This interaction between the tail of the evanescent field and the ambient environment induces changes in the effective index $n_{eff}$ of the optical mode. To leverage this sensing capability, the surface of the platform can be coated with biorecognition molecules that specifically bind with the target antigens. The molecular binding events thus alter the RI at the sensing interface and manifest as a change in the $n_{eff}$.

Fig. 2.

(a) Overview of the Evanescent Field sensing principle. (b) Structure of a micro-ring resonator (MRR). (c) MRR-based label-free sensing mechanism. Molecular binding events change the effective index of the mode, resulting in a shift in resonance. Operating at a fixed wavelength, the resonant shift generates a power and phase fluctuation. (d) A photodetector and a sensing receiver convert the optical information in the electronic domain.

B. MRR-Based Label-Free Sensors

Ring-based transducers have been the driving force for SiPh-based biosensing applications. Their μm-scale footprint offers unique advantages towards the design of highly scalable multiplexed photonic arrays for simultaneous detection of different target analytes [28]. Fig. 2b shows the basic structure of an MRR device, comprised of a bus waveguide coupled to a circular waveguide. The key mechanism of photonic sensing using MRRs is based on the shift of the resonant wavelength ${\lambda }_{\mathit{res}}$. At the well-established condition of resonance the optical round-trip length of the ring is equal to a multiple integer of the light’s wavelength and can be written as:

$${\lambda }_{\mathit{res}}=\frac{n_{eff}L}{m}$$ (1)

where m is the resonance order, $L$ is the perimeter of the ring and $n_{eff}L$ is the optical round trip length. The change of the effective index $\mathrm{\Delta }{n}_{eff}$ – induced by molecular binding taking place on the sensor’s surface – shifts the resonant wavelength of the MRR, as shown in Fig. 2c. The intrinsic sensitivity $S_{int}$ of the platform can be defined as the change of the resonant wavelength ${\lambda }_{\mathit{res}}$ due to some measurand $\mathrm{X}$ which represents molecular binding events in label-free sensing. Specifically, by manipulating (1), we can write:

$$S_{int}=\frac{\partial {\lambda }_{\mathit{res}}}{\partial X}=\frac{{\lambda }_{res}}{n_g}\frac{\partial n_{eff}}{\partial X}$$ (2)

where $n_g=n_{\mathit{eff}}-{\lambda }_{\mathit{res}}\frac{\partial n_{\mathit{eff}}}{\partial {\lambda }_{\mathit{res}}}$ is the group index of the MRR.

A number of publications have proposed label-free biophotonic sensing based on resonant photonic structures. Most prior work performs multiple scans of the whole thru-port transfer function in order to detect the MRR’s resonant shift in real-time and estimate the rate of molecular binding. However, all of these techniques heavily rely on a bulky and expensive tunable laser for precise optical scanning. This in turn limits the rate and accuracy at which measurements can be taken and leaves the quality factor $Q$ of the photonic device as the only parameter through which a higher fidelity resolution can be achieved.

To address this challenge, alternative approaches have been proposed [35] [36]. These methods employ a broadband light source and implement thermal tuning of the ring filter on-chip to select the part of the light spectrum that aligns with the sensor ring. While they eliminate the use of the tunable laser, they have inherently very low power-efficiency due to a large mismatch between the broadband source bandwidth and the bandwidth of the sensing rings. In this paper, we propose the use of the narrow-band source, coupled with thermal tuning lock of the sensor ring. Thus, detecting molecular binding events is achieved by monitoring the optical power circulating inside the ring waveguide at a fixed wavelength ${\lambda }_0$. This approach opens the door to co-packaging of low cost, single wavelength light sources. While, the locking of the MRR’s resonance to the laser’s fixed wavelength is a challenge in passive photonics platforms, it is readily achievable in the monolithic electronic-photonic proces platform used in this work.

When operating the laser at a fixed bias point of the Lorentzian spectrum, a shift in the MRR’s resonance wavelength changes the power circulating inside the ring waveguide, and creates an amplitude and phase change in the MRR’s electric field (Fig. 2c). For intensity-based sensing schemes, the system’s photonic sensitivity $S_{phot}$ can be defined as the amount of change in the circulating optical power $P_{\text{ring}}$ per induced resonant shift. Hence, we can write:

$$S_{phot}={\left.\frac{\partial P_{\mathit{ring}}}{\partial {\lambda }_{\mathrm{res}}}\right|}_{\lambda ={\lambda }_0}={\left.\frac{\partial {\gamma }^2}{\partial {\lambda }_{\mathrm{res}}}{P}_{in}\right|}_{\lambda ={\lambda }_0}=2\gamma \left({\lambda }_o\right)S_{\gamma }\left({\lambda }_o\right)P_{\mathit{in}}$$ (3)

where $S_{\gamma }\left({\lambda }_0\right)=\frac{\partial \gamma }{\partial {\lambda }_0}$ is the slope of the amplitude transfer coefficient $\gamma$ inside the ring waveguide at fixed wavelength ${\lambda }_o$. The sharpness of the Lorentzian spectrum is expressed by the slope $S_{\gamma }\left({\lambda }_0\right)$ which depends on the quality factor $Q$ of the MRR [27]. As intuitively expected, resonators with lower losses and hence higher quality factors have a higher photonic sensitivity.

Finally the optical signal is converted in the electronic domain using a photodetector (PD) with rensponsivity $R_{pd}$ and a sensing receiver of gain $G$, as illustrated in Fig. 2d. Therefore the overall electronic-photonic sensitivity can be written as:

$$S_{tot}=\frac{\partial V_{out}}{\partial X}=S_{int}{S}_{phot}{R}_{pd}G.$$ (4)

III. MONOLITHIC ELECTRONIC-PHOTONIC PLATFORM FOR LABEL-FREE MOLECULAR SENSING

Co-integration of nanophotonic sensors with state of the art electronics in a high volume and low cost commercial monolithic platform will create a breakthrough towards self contained, affordable PoC photonic sensors. Instead of relying on bulky optical and electronic equipment for optical scanning and readout processing, co-integrated electronics can unlock the door to low cost single $\lambda$ laser diodes and on-chip processing of the optical signal embedding the molecular information. To achieve this, we rely on the zero-change integration of photonic components in a high volume, advanced electronic process node (GF 45-nm RF-SOI CMOS) [37].

A. Fully Integrated Electronic-Photonic Process

The chip cross-section is illustrated in Fig. 3. As a first step, after the flip-chip packaging of the die on a PCB we fully etch the Si handle to expose the photonic sensors to the fluidic samples [38] and prevent leakage of the waveguide mode into the silicon handle. Undoped Si-crystalline is used as the waveguide layer, while various available transistor dopings are leveraged to create pn, pin, etc., diode structures in the same layer, allowing the interaction between the electrical carriers and light due to the carrier-plasma effect [39]. The 200nm thin buried oxide layer (BOX) allows relatively low confinement of the optical mode propagating in the thin Si-body waveguide layer. As a result of this, the effective index of the guided mode is affected by molecular binding events taking place on the BOX layer after the handle release.

Fig. 3.

Zero-Change 45nm SOI cross-section.

B. MRR-Based Photodetector with Embedded Heater

The structure of the MRR transducer in the zero-change 45-nm RFSOI process is shown in Fig. 4. Its 5 μm radius enables densely packed ring arrays for multianalyte detection. Conversion of the optical power to the electrical domain is achieved with 30 interleaved pin diodes located along the perimeter of the ring cavity that fully deplete the ring waveguide from carriers. The p-n junctions use the same n and p-type implants as source/drain. Spokes of n and p-types are used to connect the p-n junctions with their respective contacts in the center of the ring. The responsivity of the MRR-based sensor is further enhanced with a doped SiGe strip that is responsible for carrier generation [40]. This enables the use of the same ring element as both a sensing transducer and a ring photodetector (PD) resulting in a simple scheme without the need of a separate sensor and PD.

Fig. 4.

MRR-based photodetector in 45nm platform. An integrated heater embedded in the design of the ring eliminates the need for a tunable laser.

The MRR sensor is embedded with an integrated heater resistance, located at close proximity to the ring cavity. By relying on the strong thermo-optic effect, the thermal power dissipated by the heater resistance selectively heats the ring waveguide, thus actively tuning its resonance and eliminating the need for a tunable laser. A 10-bit Pulse Duration Modulation (PDM) driver generates pulses at an 800 MHz rate, which are averaged by the low pass thermal response of the ring (Fig. 4). The PDM driver includes a 10-bit accumulator, and its carryout bit regulates the gate of the NMOS. This in turn manages the current density flowing through a resistive heater that is integrated into the MRR.

C. Sensing Circuitry

Co-integration of on-chip electronics and photonics on the same die eliminates the need for bulky and external readout equipment. The on-chip PD is connected in a reverse bias configuration to an on-chip sensing receiver and converts the optical power fluctuations – induced by molecular binding events – into a change in photocurrent. The strong background current originating from the optical power circulating in the ring requires a large dynamic range receiver capable of detecting small <10 nA photocurrent changes at a large 1–100 μA background signal. Additionally, gain programmability is desired in order to adjust between coarse and fine sensing modes of operation required for detecting a wide range of molecular concentrations. The analog front-end of the sensing receiver, illustrated in Fig. 5 is composed of a pseudo-differential shunt feedback Transimpedance Amplifier (TIA) followed by a fully differential postamplifier with a gain of 2. A 9-bit asynchronous SAR ADC operating at 50 MSa/s digitizes the signal. The ADC’s output is averaged on-chip by 512 cycles in order to further enhance the SNR and provide an averaged value to the on-chip tuning controller. The 10-bit PDM module receives the controller’s output and based on the digital code selectively red or blue shifts the MRR. The clock path includes a Current Mode Logic (CML) receiver and a distribution network providing an 800 MHz clock to the PDM logic. The clock signal is then divided by 16 times in order to provide a 50 MHz clock to the SAR ADC.

Fig. 5.

Architecture of the on-chip receiver. The photocurrent generated from MRR-PD is passed as an input to a pseudo-differential TIA. A post-amplifier converts the single-ended output of the sensing TIA into a fully differential signal, digitized by a 9-bit SAR ADC. An on-chip tuning controller receives the averaged digital output and locks the MRR at the desired bias point.

The large background photocurrent is suppressed by input current DACs, which prevent saturation of the TIA. This is achieved by a 5-bit coarse and a 5-bit fine current DAC connected at the input of the sensing TIA (Fig. 6a,b), canceling a maximum bias photocurrent of ≈130 μA and providing a 10-bit biasing resolution for the sensing TIA. By leveraging the input current DAC’s dynamic range, we allow operation at high input optical power levels, therefore improving the SNR. However, increasing the optical power can reach a diminishing return due to self-heating effects which decrease the effective slope of the ring spectrum [41].

Fig. 6.

Circuit diagrams of (a) input current DAC, (b) shunt-feedback pseudo-differential TIA, and (c) fully differential post-amplifier.

The first stage of the analog frontend is a pseudo-differential inverter-based shunt feedback TIA, as illustrated in Fig. 6b. A digitally controlled large feedback resistance (50–800 kΩ), enabled by monolithic integration, lowers the input-referred ADC quantization error and thermal noise floor, allowing the detection of nA current changes induced from single nanoparticle binding events while providing the required gain programmability of the receiver, necessary for wide dynamic range operation. A reference-dummy TIA in conjunction with a 3-bit push-pull current DAC connected at its input is used to provide a reference voltage for biasing a fully differential postamplifier, and also suppresses supply noise. The single-ended (SE) output of the sensing TIA is converted to a differential signal by a fully differential amplifier (Fig. 6c). The PMOS transistors M₃ and M₄ adjust the amplifier’s common mode output and a 3-bit offset current DAC corrects the amplifier’s offset using NMOS transistors M₅ and M₆. SW₊ and SW₋ determines the offset correction strength of NMOS transistors M₅ and M₆. The common mode of the differential amplifier is controlled through scan.

IV. SYSTEM ARCHITECTURE

The μm-scale footprint of the MRR allows integration of multiple rings into arrays for multiplexed sensing of different analytes. Figure 7a shows the overall architecture of the 5.5×3 mm² arrayed EPSoC, consisting of 60 MRRs. The rings are arranged into 4 dual-row intensity and 2 phase-based sensing instances. Input laser light of a fixed wavelength is coupled to each instance through grating couplers. For two of the intensity-based instances, an on-chip Y-splitter splits the light into two rows of 5 MRR sensing units each. Each ring has a radius increment that sets the nominal resonant wavelength spacing of ≈100 GHz in the O-band region, which minimizes the optical crosstalk, and enables simultaneous sensing with multi-wavelength laser source.

Fig. 7.

(a) Architecture of the biosensing EPSoC. A total of 60 MRRs are arranged in intensity and phase-based sensing schemes, offering the advantage of multiplexed sensing and leveraging the amplitude and phase information of the rings. (b) Single MRR-PD intensity detection. A single MRR can be used as both a sensor and a photodetector, simplifying the complexity of the sensing scheme. (c) Dual-MRR intensity scheme. A high Q resonator - used as the sensing transducer - is interrogated by a separate MRR-PD. (d) Phase detection architecture of the Ring Assisted MZI. (e) Dual-row instance comprised of sensing and references MRRs for common mode error cancellation.

A. Single MRR-PD Intensity Detection

In intensity-based detection schemes, the molecular binding events manifest as a change in the amplitude transfer function of the MRR, resulting in a power fluctuation inside the ring waveguide. The arrayed system-on-chip consists of two types of intensity-based schemes. In the first, a Si-Ge enhanced single ring (Fig. 7b) can act as both a sensing transducer and a PD. This eliminates the need for a separate PD and reduces the complexity of the intensity scheme by requiring control of only one ring. However, this happens at the cost of a lower photonic sensitivity due to the limited $Q$ of the MRR-PD.

B. Dual MRR Intensity-based Scheme

In order to circumvent this limitation and leverage higher quality factor MRRs, a dual-ring topology is also implemented. In this scheme, power fluctuations $d{P}_{hQ}$ at the thru-port of a higher $Q$ biosensing ring are detected by an MRR-PD which acts only as a photodetector (Fig. 7c). This dual MRR architecture unlocks the path towards the use of ultra-high $Q$ sensing rings [42] that can be interrogated by high responsivity PDs, enhancing the photonic sensitivity and thus the overall system’s performance. However, the need to thermally stabilize two rings increases the level of complexity.

C. Ring Assisted Phase Detection Architecture

A resonant shift induced from molecular binding events results in a change in the amplitude and phase of the electric field of the MRR. Detecting phase instead of amplitude has received increased interest over the last decade due to its superior sensitivity compared to intensity based schemes [34], [43], [44]. In this work, we implement Ring-Assisted Mach Zehnder Interferometer (RAMZI) phase detection architectures to further enhance the sensing performance of the EPSoC. In Fig. 7d, the topology of a RAMZI is shown. Light of a fixed wavelength is split into a reference arm and a sensing arm with high $Q$ modulator MRRs. The electric fields E₁, E₂ of both arms are then recombined using an adiabatic 2×2 coupler [44]. The initial waveguide widths $W_1,W_2$ and tapering length $L$ of the coupler are determined with mode solver simulations. The optical power is then evenly split in two output waveguides. The output power at ports 3 (${\mathrm{P}}_3$) and 4 (${\mathrm{P}}_4$) can be written as:

$$P_3=\frac{P_1}{2}+\frac{P_2}{2}+\sqrt{P_1{P}_2}\mathrm{c}\mathrm{o}\mathrm{s}\left({\varphi }_{\mathit{sense}}-{\varphi }_{ref}\right)$$ (5)

$$P_4=\frac{P_1}{2}+\frac{P_2}{2}-\sqrt{P_1{P}_2}\mathrm{c}\mathrm{o}\mathrm{s}\left({\varphi }_{\mathit{sense}}-{\varphi }_{ref}\right)$$ (6)

where ${\varphi }_{\text{sense}}$ and ${\varphi }_{\text{ref}}$ are the phases of the biosensing ring and reference arm respectively and $P_1={E_1}^2$ and $P_2={E_2}^2$ are the optical power of the sensing and reference input waveguides. As observed, the RAMZI’s single-ended output power expressions include both amplitude and phase related terms. By taking a differential measurement of (5) and (6) and leveraging the output power from both arms, the phase terms can be isolated as following:

$$P_{diff}=P_3-P_4=2\sqrt{P_1{P}_2}\mathrm{c}\mathrm{o}\mathrm{s}\left({\varphi }_{\mathit{sense}}-{\varphi }_{\mathit{ref}}\right)$$ (7)

The sensitivity of the RAMZI scheme can be defined as the change of the differential power $P_{diff}$ due to a resonant shift of the sensing ring. This can be written as:

$$\begin{gathered} \frac{\partial P_{\mathit{diff}}}{\partial {\lambda }_{\mathit{res}}}=\frac{\partial P_1}{\partial {\lambda }_{\mathit{res}}}\sqrt{\frac{P_2}{P_1}}\text{cos}\left({\varphi }_{\mathit{sense}}-{\varphi }_{\mathit{ref}}\right) \\ -2\sqrt{P_1{P}_2}\text{sin}\left({\varphi }_{\mathit{sense}}-{\varphi }_{\mathit{ref}}\right)\frac{\partial {\varphi }_{\mathit{sense}}}{\partial {\lambda }_{\mathit{res}}} \end{gathered}$$ (8)

Sensitivity is maximized at the quadrature point of the sine wave, where the phase slope $\frac{\partial {\varphi }_{\mathit{sense}}}{\partial {\lambda }_{\mathit{res}}}$ dominates [34]. A thermo-optic phase shifter, located in the reference arm, is used to optimize the reference phase offset and operate at the quadrature point.

Two MRR photodetectors in the upper and bottom output arms convert the optical power at ports 3 and 4 into the electronic domain. By obtaining a differential measurement between the upper and bottom MRR-PD we further enhance the scheme’s sensitivity compared to a single-ended RAMZI operation. Overall, the proposed RAMZI structure offers a two-fold advantage: (a) the use of higher $Q$ ring resonators as sensing rings (like in the dual-ring intensity scheme) and (b) the increased sensitivity of the photonic architecture due to phase instead of amplitude-based detection. However, in order to achieve this maximum sensitivity compared to a single MRR PD detection scheme, 3 photonic structures need to be optimally biased: the high-Q MRR, the MRR-PD, and the thermal phase shifter in the reference arm.

D. On-chip Common Mode Error Cancellation

One of the challenges of MRR-based label-free sensing is the high intrinsic sensitivity of MRRs to temperature as well as pressure variations [45]. These sensitivities to changes in the ambient environment, along with the occurrence of nonspecific binding events, cause erroneous signals to be detected. To cancel these interferers, each functionalized sensing ring is paired with a non-functionalized ring that acts as a reference structure. In doing so, ambient fluctuations and non-specific binding events appear as a common mode signal to both rings. Thus the differential response suppresses these interferers, isolating the signal induced from specific molecular binding events.

In the system architecture of Fig. 7e, each sensing dual row instance includes a sensing and a reference ring row. The top row consists of sensing rings to be functionalized with receptor molecules, whereas the bottom row provides reference ring structures for each of the sensors to cancel common mode offsets from the environment and non-specific binding. The rings are physically separated by a 300 μm pitch to facilitate high channel density microfluidic integration and allow a low cost simultaneous in-channel functionalization of multiple rings with different biorecognition molecules. The reference rings are interdigitated in the horizontal axis and not placed directly below their paired sensing MRRs in order to prevent contamination from functionalization solutions flowing vertically across the fluidic channels. As a result, the pairing of the sensing and reference ring in Fig. 7e is represented by (MRR<i>, MRR<i+8>) and (MRR<i+1>, MRR<i+9>).

E. Electronic-Photonic and Fluidic Packaging

One of the key challenges towards efficient multiplexed sensing of different biomarkers in a SiPh platform is simultaneous electronic, photonic and fluidic coupling. The flat surface needed in order to accommodate multiplexed microfluidic networks requires custom packaging techniques for the microfluidics [46] and large chip surfaces that significantly increase cost. Here, we address these challenges using (a) a multilayer PDMS structure to interface with the small footprint of the silicon photonic chip and (b) a 3D printed transfer molding technique to create high resolution molds with densely packed channels down to <200 μm spacing [47]. The fabricated microfluidic package is aligned and attached on the silicon-photonic die, as illustrated in Fig. 8, providing the capability to flow multiple solutions in parallel over the ring sensors. Single optical fibers couple light to the grating couplers of the photonic array. The fully integrated EPSoC is flipchip attached to an FR4 chip-board using 250 μm C4 bumps that provide electrical coupling, as shown in Fig. 8. This packaging technique offers unique advantages in multiplexed sensing arrays as it facilitates routing of multiple sensing signals, which remains a challenge in wirebond-based packaging required in systems using hybrid electronic-photonic integration. The chip-board is then plugged into a socket of a hostboard which provides regulated power and external connectivity.

Fig. 8.

Simultaneous electronic-photonic and fluidic coupling using a co-packaged multi-channel microfluidic device fabricated with 3D printed molds. Single fibers couple light to the photonic array and flipchip of the silicon die on a PCB provides electrical coupling.

V. MEASUREMENT RESULTS

In this section, we evaluate the sensing performance of the fully integrated EPSoC. In all experiments, a fixed wavelength laser light with an optical power of ≈−4dBm is coupled onto the chip from a standard single-mode optical fiber through vertical grating couplers. The waveguide loss in this process is 3.7dB/cm, and the grating coupler’s coupling loss is ≈6dB.

A. Electronic Characterization

As a first step, we characterize the performance of the on-chip receiver. Figure 9a shows the dc characteristic of the receiver at a TIA gain of 50 kΩ. This is obtained by sweeping the digital code of the fine current DAC, thus creating incremental current changes at the input of the receiver. Using the simulated gain of the post-amplifier, the DNL of the SAR ADC can be characterized for each current DAC step (Fig. 9b), which creates an incremental change of ≈12 mV at the SAR’s input. We then evaluate the noise performance of the receiver by monitoring the ADC’s output at a fixed input current DAC code. For clinically relevant molecular concentrations in the fM and pM regime, binding events can occur on the order of a minute. By leveraging the slow nature of molecular kinetics, the LoD can be lowered with further off-chip post-averaging, enabled by the 50 ms readout sampling rate. The contribution of white noise sources is therefore significantly decreased, resulting in an averaged output that is mostly dominated by the flicker noise. Fig. 9c shows the on-chip averaged SAR output at a TIA gain of 400 kΩ and a fixed input DAC code. After applying a moving average window of 1 minute, the post averaged output, shown in Fig. 9d., results in a standard deviation (σ) of 0.117 LSB. Dividing by the gain of the receiver results in a σ in input current of 170 pA. This minimum detectable current combined with the maximum current detected at 50 kΩ, results in an 85 dB electronic dynamic range. The power dissipation from each receiver unit is ≈3.4 mW. The analog front-end consumes ≈2.7 mW, while the digital back-end and the integrated heater dissipate ≈0.5 mW and ≈0.2 mW respectively.

Fig. 9.

Electronic Characterization of the receiver. (a) SAR dc characteristic. (b) Extracted DNL. (c) On-chip averaged SAR output at 400kΩ TIA gain. (d) Post averaged output.

B. Intrinsic Sensitivity Characterization

In order to evaluate the sensing capability of a photonic transducer, the initial step involves characterizing the inherent sensitivity of the device, which is significantly influenced by the waveguide’s geometry and the process technology. To illustrate the intrinsic sensitivity of our platform, we introduced three distinct sodium chloride (NaCl) solutions ($NaC{l}_1,{NaCl}_2$, and ${NaCl}_3$) with known refractive index units (RIU) into three distinct microring channels using our microfluidic device. The three NaCl solutions $NaC{l}_1,NaC{l}_2$, and $NaC{l}_3$ have RIU of 1.3356, 1.3492, and 1.3650, respectively. We then examined the resonant shift of each microring by adjusting the heat code, as depicted in Figure 12. Based on the maximum tuning range at $H{V}_{\mathrm{D}\mathrm{D}}=1.1\mathrm{V}$ and the 10-bit resolution of the PDM modulation, each heating step corresponds to a step of 3 pm. [48] Importantly, it should be noted that the relationship between resonance shift and temperature is linear and does not introduce any distortion. [27]

Fig. 12.

Intrinsic sensitivity analysis of our platform with three NaCl solutions with known RIU [48]

In order to consider the influence of liquid cooling, we first measure the absolute heating step shift from the air and then check the relative shift between two microrings. The relative heat step shift was measured to be 10 LSB between $NaC{l}_1$ and ${NaCl}_2$, and 9 LSB between ${NaCl}_2$ and ${NaCl}_3$. With the RIU of each NaCl solution, we concluded that our platform has an intrinsic sensitivity of 2nm/RIU. Furthermore, we employed a mode solver to validate that our measurement results align with our simulation data.

C. Single MRR-PD Lorentzian Spectrum

We convert the optical power into the electronic domain using on-chip reverse-biased single MRR-PDs, with a maximum responsivity of $0.2\frac{A}{W}$ and a Q-factor of 6k. The photocurrent generated by the MRR-PD is then passed as an input to the on-chip receiver. We first evaluate the overall electronic-photonic performance of single ring-based PDs by characterizing the Lorentzian spectrum of the MRR device using the on-chip receiver. Figure 10a shows the ring spectrum of a MRR-PD at different TIA gains and a fixed wavelength. Instead of tuning the laser to obtain the Lorentzian shape of the ring, we tune the ring’s resonance by sweeping the input heating code of the PDM driver by an LSB step within a range of 200 codes. Using a heater $H{V}_{\mathrm{D}\mathrm{D}}=1.1\mathrm{V}$ makes each heat code step to be 3pm resonant shift, resulting in an equivalent wavelength sweep of 600 pm. The gain programmability of the receiver allows an increase of the Lorentzian slope. By taking the derivative of each ring spectrum over the stable, blue side of resonance, an effective electronic photonic sensitivity can be established, as illustrated in Fig. 10b. The linear dependence of the maximum effective slope on the TIA’s gain is presented in Fig. 10c.

Fig. 10.

(a) Lorentzian spectrum of a single MRR-PD for different TIA gains. (b) Slope of the ring spectrum for different TIA gains. (c) Electronic-photonic sensitivity versus TIA gain. (d) Ring spectrum of the higher-Q sensing/modulator MRR(MRR Mod) and the interrogating MRR-PD. (e) Lorentzian slope of the higher-Q modulator/sensing MRR(MRR Mod) and MRR-PD. A 24% increase in sensitivity is observed.

D. Dual-MRR Demonstration

In order to leverage higher $Q$ rings, a dual ring scheme is needed. In this detection mechanism, we use a separate non SiGe-doped lower-loss MRR modulator(MRR Mod) acting as the sensor and an MRR-PD used only as a photodetector. In order to examine the performance of the dual ring scheme, the tuning heater of a ring modulator sweeps its resonance and the power fluctuations of the thru port are captured by the receiver of an MRR-PD which has its resonance located on the laser’s wavelength. Figures 10d,e show the Lorentzian spectrum and the effective sensitivity of a non-SiGe doped MRR modulator and a ring PD. A 24% increase in sensitivity is observed which results from the higher $Q$ of the MRR modulator. This detection mechanism unlocks the path toward the use of up to 200k $Q$ rings, previously shown in this platform [42], thus enhancing the system’s sensitivity by at least an order of magnitude compared to 12k-18k $Q$ MRRs measured in this chip.

E. Phase-Based Scheme Characterization

In order to further improve the system’s sensitivity we have implemented ring-assisted MZI structures. By sweeping the heating code of the phase shifter, the optimal sensitivity point can be selected at the quadrature point of the sine wave in order to maximize the photonic sensitivity, as discussed in Section IV–C.

The improved performance of phase-based sensing is illustrated in Fig. 11, where the sensitivity of the RAMZI’s differential response is 49% higher compared to the slope of the MRR’s amplitude spectrum. In the design implemented in this work, the optical power is further split for monitoring purposes before entering the adiabatic coupler, resulting in an extra 3-dB power loss which is not fundamentally required in a RAMZI-based scheme. Therefore, the differential phase-based architecture can be up to $\approx 3x$ more sensitive compared to intensity-based sensing with a ring modulator and $3x$ more sensitive compared to a single MRR-PD by benefiting from the higher quality factor of an MRR modulator.

Fig. 11.

Phase detection RAMZI characterization. Comparison of measured sensitivity between differential RAMZI output and intensity based detection of the MRR modulator. A 49% enhancement is observed.

F. On-chip Tuning Controller

By leveraging the in-ring integrated heater, we implement an on-chip tuning controller that automatically locks the ring at a user-defined bias point of the Lorentzian spectrum. Specifically, once the tuning controller detects the optimal bias point, the ADC output of that bias point is stored. Once there is any fluctuation of the power circulating in the ring by a user-defined power threshold amount, the on-chip tuning controller enters a “recover” state, during which it searches the heat code at which the desired ADC output of the optimal bias point is found.

In order to cancel thermal variations, the digital logic of the tuning controller needs to run faster than the thermal time constant of the MRR which is ≈ 30us. In this work, the controller is updated by the averaged output of the SAR ADC at ≈ 98kHz. The on-chip tuning mechanism offers the additional benefit of re-locking the ring at its nominal point when large bulk RI changes introduce offsets. On-chip tuning controller has two modes, PD-lock mode, and sense-lock mode, and the user defines which mode to use depending on the microring’s usage [45].

In PD-lock mode, the on-chip tuning controller performs a sweep of the heat code and monitors the first-order differential output of the averaged ADC signal. It actively seeks the heat code by identifying changes in the sign of this first-order differential output.

Alternatively, in the sense-lock mode, the on-chip tuning controller also conducts a sweep of the heat code but focuses on detecting changes in the sign of the second-order differential output of the averaged ADC signal to determine the appropriate heat code.

For the single MRR-PD Intensity Detection scheme, we use the sense-lock mode and the on-chip tuning controller locks the heat code at the maximum slope of the Lorentzian. For the dual MRR Intensity-based scheme, we use the sense-lock mode for the high-Q biosensing ring to lock the ring at the maximum slope, and the PD-lock mode for the MRR-PD to maximize the signal. Finally, for the Ring Assisted Phase Detection Architecture, both the high-Q biosensing ring and the MRR-PD will be in the PD-lock mode as the sensitivity maximizes at the resonance of the sensing ring for phase detection.

After pre-characterizing the Lorentzian spectrum of the MRR and selecting the desired bias point (Fig. 13a), the controller is enabled and locks the ring, as shown in Fig. 13b. Starting from off-resonance, the integrated heater blue-shifts the Lorentzian spectrum at a scanning step selected by the user. Once the desired bias point of the Lorentzian is detected, the controller enters the locking mode in which the ring is locked within an LSB heater precision, as shown in Fig. 13c. In addition to the ring’s locking, we verify the functionality of the controller by introducing ambient heating events. As illustrated in Fig. 13d, the tuning controller adjusts the heating code of the PDM driver and cancels the ambient-induced shift.

Fig. 13.

On chip tuning controller operation. (a) Pre-characterized Lorentzian Spectrum of the MRR. (b) Locking of the MRR. Starting from off-resonance the PDM heating decreases and blue-shifts the ring until the bias point is detected. (c) Locking resolution. (d) On-chip re-locking of the MRR at its nominal bias point, cancelling any temperature fluctuations.

G. Differential Readout Demonstration

We suppress the environmental common-mode drift by using a differential scheme of a sensing and reference ring located 425 μm apart. To characterize the effectiveness of this architecture, the MRRs – not exposed to any biosamples – are monitored in real-time at a fixed wavelength and heating code, and the differential shift is calculated by subtracting the reference shift from the sensing response.

First, the Lorentzian spectra of the sensing and reference rings are pre-characterized, as shown in Fig. 14a,b. By monitoring the ADC responses of the sensing and reference receiver at a fixed wavelength (Fig. 14c) and multiplying the relative changes with the slope of the sensing and reference MRR spectrum around the region of operation (Fig. 14a,b), the electronic responses are converted into resonant shifts. Subtracting the reference from the sensing signal results in an effective differential shift, as illustrated in Fig. 14d. As can be observed from the post-averaged differential shift (moving average with 1 minute window) in Fig. 14e, the ambient temperature variations result in a peak sensing and reference resonant shift of ≈ 9pm, while the effective differential response is < 1pm, thus suppressing the ambient common mode fluctuations by ≈10x. The same experiment is repeated at an increased optical power and more stable temperature conditions. After subtracting the effect of the reference ring, the differential shift is obtained, as shown in Fig. 14f. The standard deviation and peak to peak value of the effective differential signal are 100 fm and 396 fm respectively, thus approaching the detection of a single nanoparticle of 140-nm size and 1.6 RI, similar to the characteristics of a viral particle, which induces a 170 fm shift. Additionally, based on the $2\frac{nm}{RIU}$ (RIU: Refractive Index Unit) intrinsic sensitivity of the platform, characterized in Section V–B, the peak-to-peak differential shift corresponds to 1.98*10⁻⁴ RIU. The LoD can be further improved by 3.7x, leveraging the sensitivity enhancement offered by the dual-ring phase detection scheme, thus resulting in an equivalent sub-nanoparticle resolution. By combining the minimum detectable shift with the ≈ 7nm maximum tuning range at a heater supply voltage of 1.8V [48], a system dynamic range of 79 dB is achieved.

Fig. 14.

On-chip differential scheme for ambient common mode error cancellation. (a),(b) Characterization of the Lorentzian spectrum of the sensing and reference ring. (c) Monitoring of the sensing and reference ADC responses without exposure to fluidic solutions. (d) Sensing, reference and effective differential resonant shift. (e) Post-averaged differential shift. (f) Post-averaged differential shift at increased input optical power. The peak to peak differential signal corresponds to the detection of ≈ 2 140nm particles, similar to the size of a viral particle.

H. Label-Free Real-Time Molecular Kinetics

In [48] we demonstrated for the first time real-time molecular kinetics in a fully integrated electronic-photonic platform using crosslinker-labeled biotin molecules and streptavidin. Here we modify the coating chemistry and activate the surface of the chip instead of the receptors, in order to directly immobilize molecules, not conjugated with any crosslinkers. To verify the modified chemistry, we present real-time kinetics of anti-BSA and BSA binding without using conjugated receptors.

After uniformly coating the surface of the chip with an aminosilane layer of APTES, the die was immersed in 2.5% glutaraldehyde solution and incubated for 1hr at RT (Room Temperature). This results in the formation of a crosslinker layer that allows direct immobilization of non-conjugated antibodies [49]. The sensing ring was functionalized in-channel with anti-BSA molecules in order to detect the specific binding events of BSA of ≈7 nm size (Fig. 15a). After coating with glutaraldehyde, the Lorentzian spectrum of the sensing and reference ring is first characterized, as shown in Fig. 15b and an initial bias point is selected for both rings. Once the thermal tuner sets the desired bias point, the thermal tuner is deactivated and the environmental fluctuations are tracked using a dual row sensing scheme and a differential readout. The sensing ring is then functionalized in-channel with anti-BSA molecules. A 3.3 μM anti-BSA solution is flown through the sensing microfluidic channel at a $5\frac{\mu L}{min}$ flow rate while PBS is flown over the reference ring. The ADC responses of both MRRs are monitored in real-time at a fixed wavelength and heating code, as shown in Fig. 15c. After converting the electronic responses into resonant shifts (Fig. 15d) and subtracting the reference from the sensing response, an effective differential shift is established, as illustrated in Fig. 15e. The binding curve is characterized by an initial slope of $3.5\frac{pm}{min}$ which depends on the analyte concentration and describes the binding rate of the reaction [13]. Kinetics are then followed by an equilibrium between association and dissociation of anti-BSA molecules which results in a steady state of ≈15 pm.

Fig. 15.

(a)-(e) Real-time molecular kinetics of a 3.3 uM anti-BSA solution. The SiO₂ BOX surface is coated with an aminosilane layer and functionalized with a glutaraldehyde crosslinker layer that allows direct immobilization of non-conjugated anti-BSA on the surface (a). The Lorentzian spectrum of the sensing and reference MRR is characterized (b) and the sensing and reference ADC responses are monitored (c). After subtracting the reference from the sensing shift (d), an effective differential signal is obtained (e), representing anti-BSA kinetics. (f) Differential shift from 560 nM and 15 uM BSA kinetics. (g) Differential shift from 150 nm streptavidin-coated nanoparticle binding.

The specific binding of BSA-antiBSA is then monitored for 560 nM and 15 μM BSA solutions, as shown in Fig. 15f. An overall steady-state shift of ≈12 pm is measured. The reduced increase of the initial binding slope at 15 μM concentration can be attributed to the partial coverage of the anti-BSA receptor sites from the previous 560 nM BSA solution.

I. Nanoparticle detection

After establishing the first proof of concept for real-time molecular sensing, we investigated the detection of streptavidin coated 150 nm nanoparticles with biotin (SuperMag Streptavidin Magnetic Beads). Nanoparticles conjugated with target analytes have been used in label-free sensing for amplifying the detected binding signal [50]. Additionally, their >100 nm size is similar to the size of viral particles, therefore making them good candidates for mimicking viruses. We demonstrate real-time kinetics of streptavidin coated nanoparticles (Ocean NanoTech) at a 50 pM concentration ($3\mathrm{*}{10}^{10}\frac{\mathit{particles}}{ml}$), establishing the versatility of the platform in detecting various types of molecules, ranging from proteins to nanoparticles. A sensing ring is first functionalized with biotin molecules that act as receptors for the specific detection of streptavidin (Fig. 15g). After flowing the conjugated particle solution through the sensing fluidic channel and using a similar detection methodology of subtracting the reference from the sensing ring response, an overall effective differential shift is monitored (Fig. 15g). Based on the simulated resonant shift of ≈600 fm induced from a single 150 nm particle binding event with RI≈2.2, the overall measured shift of ≈23 pm corresponds to the detection of ≈38 nanoparticles.

Table I summarizes the performance of the EPSoC and compares it with other state-of-the-art label-free photonic platforms. This work is the world’s first fully integrated biophotonic sensor, enabling nanophotonic label-free sensing and readout processing on the same die of a high volume commercial process. The measured LoD without biosamples corresponds to the equivalent physical presence of ≈two 140 nm viral particles and can be further lowered to a sub-nanoparticle resolution using phase detection. It should also be noted that partial etching of the SiO₂-BOX layer by 100 nm can further boost sensitivity by x7 [38] due to a lower confinement of the optical mode leading to a more exposed evanescent field to the sample.

TABLE I.

Comparison Table with State of the Art

	This work*			Zhu [7] Nature ’10	Iqbal [13] IEEE Quant. EI. ’10	Flueckinger [30] Optics Express ’16	Leuermann [23] Sensors ’19
Sensor	MRR			Micro-Toroid	MRR	MRR	MZI
Photonic Technology	CMOS RFSOI 45nm			Passive	Passive	Passive	Passive
On-Chip Readout Processing	Yes			No	No	No	No
Fixed Wavelength Laser	Yes			No	No	No	Yes
Limit of Detection	Single MRR-PD	RAMZI	Etched BOX
Equivalent to RIU	1.98*10−4	5.35*10−5	7.6*10−6	N/A	7.6*10−7	2*10−6	~10−8
Equivalent to # of 140nm particles	~2	sub 1**		sub 1	sub 1	sub 1	sub 1
Transducer Area (um2)	78 (5um radius)			3D (15um radius)	707 (15um radius)	707 (15um radius)	~400 (folded 6mm wguide)

^*The reported LoD is derived from rings rot exposed to any biosamples

^**Estimated LoD after architectural and process-dependent sensitivity enhancement

VI. CONCLUSION

We demonstrate the world’s first fully integrated electronic-biophotonic label-free system-on-chip in a high volume SOI process. We achieve this by developing an array of 60 rings connected to on-chip receivers, approaching an idealized LoD (without biosamples) equivalent to a single 140 nm viral particle. However, we recognize that with the addition of biosamples, additional sources of noise such as temperature and pressure fluctuations, as well as non-specific binding, will degrade this limit of detection. We also introduce a dual-ring phase detection sensing scheme to further enhance the system’s sensitivity compared to an intensity-based single ring detection. An embedded integrated heater combined with an on-chip tuning controller tunes the ring’s resonance, eliminating the need for a tunable laser and locks the MRR at the desired point of operation. By suppressing ambient common mode errors with an on-chip differential scheme using a sensing and a reference MRR, we demonstrate real-time molecular interactions in a monolithic high volume electronic-photonic process, unlocking the door towards highly scalable and self-contained photonic LoC systems.

Biographies

Christos Adamopoulos (M’21) received the B.S. degree in electrical and computer engineering from National Technical University of Athens, Athens, Greece, in 2015, and the M.S. degree in electrical engineering and computer sciences from the University of California at Berkeley, Berkeley, CA, USA, in 2019, where he is currently working towards a PhD degree.

He has held internship positions at Infineon Technologies, Munich, Germany and Apple, Cupertino, CA, USA. His current research interests include analog and mixed-signal integrated circuits and the design and implementation of electronic-photonic integrated systems for biosensing lab-on-chip applications.

Hyeong-Seok Oh received the B.S. (summa cum laude) degree in electrical and computer engineering from Seoul National University, Seoul, South Korea, in 2021. He is currently pursuing the Ph.D. degree in electrical engineering and computer sciences with the University of California at Berkeley, Berkeley, CA, USA. His current research interests include analog and mixed-signal IC design, the design and modeling of electronic-photonic integrated systems for biomedical application, and hardware design methodologies.

Panagiotis Zarkos (S’21) received the B.S. degree in electrical and computer engineering from National Technical University of Athens, Athens, Greece, in 2015, and the M.S. degree in electrical engineering and computer sciences from the University of California at Berkeley, Berkeley, CA, USA, in 2019, where he is currently working towards a Ph.D. degree.

He held an internship position at the PHY Research Laboratory, Intel Corporation, Hillsboro, OR, USA, where he worked on energy-efficient, high-speed transceiver architectures based on silicon photonics. His current research interests include analog and mixed-signal IC design and the design and modeling of electronic-photonic integrated systems for biomedical sensors, focusing on endoscopic ultrasound receivers.

Sidney Buchbinder received the B.S. degree in electrical engineering from the California Institute of Technology (Caltech), Pasadena, CA, USA, in 2015. He is currently pursuing the Ph.D. degree at the University of California at Berkeley, Berkeley, CA, USA.

His research interests include photonic design automation and verification, and the design of analog electronic-photonic integrated systems. He has held an internship position at Ayar Labs, where he worked on integrated photonic system design.

Pavan Bhargava (S’14) received the B.S. degree in electrical engineering from the University of Maryland at College Park, College Park, MD, USA, in 2014. He is currently pursuing the Ph.D. degree in electrical engineering from the University of California at Berkeley, Berkeley, CA, USA. He held internships at the U.S. Army ResearchLaboratory, Adephi, MD, USA, where he developed low-power biosensors. He is with the AyarLabs, Emeryville, CA, USA, where he focuses on high-speed analog/mixed signal design for optical receivers. His current research interests include modeling and designing fully integrated optical beam-steering systems for solid-state LiDAR applications.

Asmaysinh Gharia received a B.A. degree in molecular and cellular biology from the university of California at Berkeley, Berkeley, CA, USA in 2018 and conducted research in the department of electrical engineering and computer science there. He also conducted post baccalaureate research at the University of California San Francisco, San Francisco, CA, USA until 2020 when he began his PhD at the National Institutes of Health, Bethesda, MD, USA and the department of engineering at the University of Cambridge, Cambridge, UK.

His current interests include cellular immunotherapies, microfluidics, and the use of electrode arrays for high through-put cell engineering.

Ali Niknejad (Fellow, IEEE) received the Ph.D. degrees in electrical engineering from the University of California, Berkeley, in 2000 where he now holds the Donald O. Pederson Distinguished Professorship chair in the EECS department at UC Berkeley and he is a faculty co-director of the Berkeley Wireless Research Center (BWRC). He is also the Associate Director of the Center for Converged Tera-Hertz Communications and Sensing (ComSenTer). Prof. Niknejad received the 2020 SIA/SRC University Research Award, recognized “for noteworthy achievements that have advanced analog, RF, and mm-wave circuit design and modeling, which serve as the foundation of 5G+ technologies.” Prof. Niknejad is the recipient the 2017 IEEE Transactions On Circuits And Systems Darlington Best Paper Award, the 2017 Most Frequently Cited Paper Award from 2010 to 2016 of the Symposium on Very Large-Scale Integration Circuits, the CICC 2015 Best Invited Paper Award, and the 2012 ASEE Frederick Emmons Terman Award. He is also the co-recipient of the 2013 and 2010 Jack Kilby Award for Outstanding Student Paper, and the co-recipient of the Outstanding Technology Directions Paper at ISSCC 2004. He is a co-founder of LifeSignals and RF Pixels, a 5G technology startup. His research interests lie within the area of wireless and broadband communications and biomedical imaging and sensors, integrated circuit technology (analog, RF, mixed-signal, mm-wave), device physics and compact modeling, and applied electromagnetics.

Mekhail Anwar (Member, IEEE), MD PhD, is a physician-scientist with over 15 years of experience in bioengineering and integrated circuit design experience, with an active clinical practice in Radiation Oncology. He received his BA in Physics from UC Berkeley, where he graduated as the University Medalist, and went on to receive his Ph.D. degree from the Massachusetts Institute of Technology (Cambridge MA) in Electrical Engineering and Computer Sciences followed by an MD from the University of California, San Francisco, completing residency in Radiation Oncology at UCSF. He joined the faculty at the UCSF in 2014, where he focuses on the use of precision radiotherapy for challenging to treat malignancies, such as pancreas cancer. He is a Core member of the UC Berkeley and UCSF Bioengineering group, and recipient of the Department of Defense Prostate Cancer Research Program Physician Research Award, the AACR Career Development Award and the NIH’s Trailblazer Award and New Innovator Award. His research is centered on developing new technologies to guide modern surgery and radiotherapy, as well as interrogate fundamental biological processes. His primary focus is in developing micro-devices to interact with and image the tumor microenvironment. Key to these efforts is the introduction of integrated circuits (ICs or “computer chip technology”) to biological detection. With the ability to integrate both sensors, computation and communication in a millimeter-scale form factor with features easily in the sub-micron range, extremely versatile sensors can be developed and embedded into biological systems, including in vivo.

Vladimir Stojanović (Senior Member, IEEE) received the Dipl.Ing. degree from the University of Belgrade, Belgrade, Serbia, in 1998, and the Ph.D. degree in electrical engineering from Stanford University, Stanford, CA, USA, in 2005. He was with Rambus, Inc., Los Altos, CA, USA, from 2001 to 2004, and the Massachusetts Institute of Technology, Cambridge, MA, USA, as an Associate Professor, from 2005 to 2013. He is currently a Professor of electrical engineering and computer sciences with the University of California at Berkeley, Berkeley, CA, USA and a faculty co-director of the Berkeley Wireless Research Center (BWRC). His current research interests include design, modeling, and optimization of integrated systems, from CMOS-based VLSI blocks and interfaces to system design with emerging devices such as NEM relays and silicon photonics, design and implementation of energy-efficient electrical and optical networks, and digital communication techniques in high-speed interfaces and high-speed mixed-signal IC design. Dr. Stojanović was a recipient of the 2006 IBM Faculty Partnership Award, the 2009 NSF CAREER Award, and the 2008 ICCAD William J. McCalla, the 2008 IEEE TRANSACTIONS ON ADVANCED PACKAGING and the 2010 ISSCC Jack Raper Best Paper and 2020 ISSCC Best Forum Presenter Awards. He was an IEEE Solid-State Circuits Society Distinguished Lecturer from 2012 to 2013.