Toward a CMOS-integrated quantum diamond biosensor based on NV centers
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | Open MIND |
| Published | 2026-02-24 |
| Authors | Ioannis Varveris, Gianni D. Aliberti, Felix J. Barzilaij, Zhi Jin, Samantha A. van Rijs |
| DOI | 10.48550/arxiv.2602.20437 |
The proposed 40 nm CMOS SPAD biosensor is estimated to reach approximately 90 nT/√Hz per pixel for continuous-wave ODMR. The estimate uses a shallow ensemble NV layer, a 10 micrometer diamond slab, and modeled collection and detection efficiencies. Most system results are simulations or design calculations, not a completed cell measurement.
Sensor architecture and chip status
Section titled “Sensor architecture and chip status”The sensor head places an NV-containing diamond substrate above a 16 × 16 SPAD array designed in TSMC 40 nm CMOS. Each pixel includes active quenching, programmable hold-off, recharge, counting, and multiplexing. The array was the largest that fit with its circuitry in slightly less than 1 mm². A 0.9 layout shrink reduced the 1085 × 1085 micrometer design to a 976.5 × 976.5 micrometer die.
Each detector must exceed 3 Mcounts/s, which requires a cycle time below 333 ns. The targeted quench time is 20 ns, with recharge within 1 ns. An earlier chip iteration showed a 20 ns cycle with active quenching and hold-off, with negligible degradation. Row serialization required 30 ns across 16 pixels. Checks on the current chip covered DRC and partial LVS, with the SPAD excluded from LVS.
Readout and microwave integration
Section titled “Readout and microwave integration”The external chain uses an FPGA, Arduino, and lab computer. The FPGA decodes 16 serial streams and accumulates 32-bit counters with capacity above 4 × 10^9 events. A logic mismatch restricted FPGA operation to 1 MHz, so the system is not yet a self-contained readout component. The computer reconstructs each 16 × 16 count image for ODMR analysis.
Microwave drive targets the 2.87 GHz NV resonance. Radiators should sit within a few hundred micrometers of the NV layer and produce a homogeneous B1 field. Distributed single-turn loops, parallel loop stacks, and wire grids are proposed. Their uniformity and coupling remain unmeasured in the reported platform.
Excitation and collection efficiency
Section titled “Excitation and collection efficiency”Green excitation is typically 532 nm, while NV fluorescence spans 600 to 800 nm. A side-coupled TIR method-of-images simulation used a 10 micrometer diamond, a 1 mm width, a 20 nm deep NV layer, and 53 degree divergence. Illumination became uniform after about 750 micrometers. The proposed 1.75 × 1.0 mm diamond reserves 0.75 mm for spreading and supports a 1 × 1 mm sensing area.
An integrated metal grating is proposed to suppress the 532 nm pump while transmitting 600 to 800 nm fluorescence. A cited prior grating measured -2.1 dB at 700 nm and -21.8 dB at 532 nm, with -3 dB from covering half the SPAD. The present paper does not report measured transmission for its own grating.
Classical ray tracing with refraction and Fresnel transmission produced the following modeled collection results.
| Optical design | Assumed condition | Modeled peak collection | Effective collection area | Main result |
|---|---|---|---|---|
| Simple slab | 10 micrometer diamond, 10 micrometer air gap, 12 micrometer SPAD | 0.75% | 4.6 µm² | Broad collection profile |
| Quartz micropillar | 10 micrometer diamond, critical angles of 37° and 43° | 4.7% | 5.2 µm² | Narrow collection profile |
| Diamond-quartz-diamond | 10 micrometer diamond and quartz layers | 1.5% | 11.4 µm² | Largest equivalent collection area |
| Objective | Air, numerical aperture 0.95 | Approximately 4% | Not reported | 23.1 degree internal half-angle |
The diamond-quartz-diamond calculation neglects pillar sidewall coupling. The simple design was retained because its 4.6 micrometer squared equivalent area was sufficient to saturate SPADs in the sensitivity model.
At the chip’s 55 micrometer pitch, optical crosstalk from the simple design is negligible. Quartz micropillars keep crosstalk negligible down to about 27 micrometer pitch. A hemispherical diamond microlens can collect about 29% under non-TIR excitation, but that geometry conflicts with planar TIR.
Continuous-wave sensitivity estimate
Section titled “Continuous-wave sensitivity estimate”The 90 nT/√Hz result comes from a two-dimensional sweep over Rabi frequency and optical saturation. Its minimum occurs with a 0.15 MHz ODMR linewidth. The model assumes 500 NV centers per SPAD, a surface density of 38 per square micrometer, or 3.8 × 10^9 per square centimeter. It also assumes about 40% SPAD photon detection efficiency and a detected rate near 1 Mcounts/s. At that rate, the paper treats shot noise as dominant and dark counts as negligible.
Five hundred NV centers reach 1 Mcounts/s at a saturation parameter of 7 × 10^-3. The authors retain 500 centers to stay below SPAD saturation. They cite 1 ppm nitrogen as a coherence-oriented choice, with T2* near 10 microseconds in isotopically controlled 12C diamond. Their modeled density corresponds to tens to hundreds of ppb nitrogen before activation.
SPION-labeled HEK293T target
Section titled “SPION-labeled HEK293T target”The biological case uses SPION-labeled HEK293T cells as a uniform magnetized sphere with a 9 micrometer radius and a volume of 3 × 10^-9 mL. At 25 emu/g, the estimated cell dipole moment is 3 × 10^-12 emu, or 3 × 10^-15 A m². A concentration of 40 micrograms/mL produces a surface field of 0.4 to 0.8 microtesla. The modeled ODMR shift is near 0.01 MHz, against ensemble linewidths up to 0.15 MHz. The field and shift are dipole estimates, not measured cellular signals.
The SPIONs were randomly dispersed and not functionalized for a biomarker. Magnetization still requires an external bias. A 5 nm magnetite particle with 25 emu/g has a stated moment of 8.5 × 10^-21 A m². The paper says tesla-scale fields would stabilize its magnetization against 4 × 10^-21 J thermal energy.
Larger 10 to 100 nm particles need only a few mT to several hundred mT for near saturation. The cited toxicity threshold often starts above 100 micrograms/mL, although formulation changes it.
Performance limits and engineering implications
Section titled “Performance limits and engineering implications”The current 55 micrometer pitch gives an 880 × 880 micrometer field of view. The authors say this is insufficient for cellular-precision cancer detection, but suitable for cardiography, which needs millimeter-scale sensing. A smaller CMOS pitch could improve spatial resolution, yet optical and electrical crosstalk increase. The 16 × 16 array is modest beside prior 40 nm arrays of 96 × 40 and 192 × 128 pixels. A separate 3D-stacked SPAD sensor reached 3.2 megapixels. The paper cites MRI at 60 micrometers and 9.4 T, while attributing finer imaging to smaller CMOS pixels.
The paper does not close the experimental sensitivity budget. The authors list end-to-end optical throughput, wavelength-dependent PDE, TIR interaction with subsurface micro-optics, thermal and magnetic drift, and long-term SPION effects. Their next tests are calibrated sensitivity, reference phantoms, fixed HEK293T cells with known loading, and co-design of microwave delivery with fluorescence collection.
For material selection, the authors favor ensemble NV because it provides more signal and simpler integration than single NV. The chip still requires an external laser, microwave source, FPGA, Arduino, and computer. The practical result is a compact sensor head with a defined diamond specification, not yet a standalone biosensor. The sensing diamond must place the ensemble close to the sample, while total internal reflection and index matching control excitation and collection losses.
Materials and support
Section titled “Materials and support”The CMOS-integrated biosensor needs a shallow ensemble-NV layer in a thin diamond sensing plate, with 10 micrometers modeled and up to 120 micrometers proposed. 6CCVD’s diamond materials catalog lists single-crystal and polycrystalline MPCVD diamond plates and wafers with thicknesses from 0.1 to 500 micrometers, covering the paper’s dimensional range. The NV creation, 20 nm near-surface placement, and optical stack should be specified through engineering support.
View Original Abstract
We report progress toward a CMOS-integrated quantum diamond biosensing platform that combines nitrogen-vacancy (NV) centers in diamond with a custom 40 nm CMOS Single-Photon Avalanche Diode (SPAD) array. The system integrates on-chip active quenching and digital readout with external FPGA-based photon counting, compact microwave delivery, and practical optical excitation and collection schemes to support widefield optically detected magnetic resonance (ODMR). System-level design considerations spanning fluorescence collection efficiency, detector count-rate capability, and microwave homogeneity are analyzed with biological compatibility and scalability in mind. Using superparamagnetic iron oxide nanoparticle (SPION)-labeled HEK293T cells as a representative use case, simple dipole-field estimates indicate that sub-$μ$T sensitivity is required to resolve ODMR shifts within typical ensemble linewidths. Based on the proposed architecture and efficiency analysis, a magnetic field sensitivity of approximately 90 nT/$\sqrt{\mathrm{Hz}}$ per pixel is estimated. These results outline a practical path from optics-heavy quantum diamond microscopes toward compact, CMOS-integrated NV-based biosensors for quantitative magnetic imaging in complex biological environments.
High Concentration Quantum Grade Single Crystal Diamond
View Specs →
Single Crystal Cvd Diamond Plate
View Specs →