Bosch: Quantum Sensing with NV-Center Magnetometry for Precision Manufacturing

The Problem

Bosch manufactures precision components at scale: fuel injectors with sub-micron tolerances, MEMS inertial sensors, power electronics, and PCB assemblies for automotive systems. Quality control at this precision requires detecting defects that classical sensors are approaching their fundamental limits on.

Two specific problems motivated Bosch’s quantum sensing research:

Non-destructive inspection: Current-carrying defects in PCB traces (hairline cracks, voids in solder joints, delamination) produce weak magnetic field signatures. Classical Hall-effect sensors and fluxgate magnetometers can detect these, but their noise floor limits detection of small or deep defects without destructive cross-sectioning. A sensor with 10x better sensitivity could catch defect classes that currently escape inspection.

Inertial navigation: Automotive and industrial systems increasingly require navigation without GPS (tunnels, underground facilities, GPS-denied environments). Classical MEMS accelerometers and gyroscopes suffer from drift: errors accumulate over time. Quantum inertial sensors (atom interferometers) do not drift in the same way; their fundamental accuracy is set by atomic physics, not mechanical wear.

Both problems are amenable to quantum sensing, specifically the quantum mechanical properties of physical systems used as precision measurement devices rather than gate-based quantum computing.

NV-Center Magnetometry

The nitrogen-vacancy (NV) center is a point defect in diamond: a nitrogen atom adjacent to a lattice vacancy. Its electron spin has a coherence time of milliseconds at room temperature (extraordinary for a solid-state system) and its spin state can be read out optically using a green laser.

An NV center’s resonance frequency shifts in proportion to an external magnetic field (Zeeman effect). By driving the spin with microwave pulses and measuring the fluorescence, you measure the magnetic field with high precision.

Bosch used the Ramsey sequence: prepare the NV spin, let it precess in the field for a fixed time, then read out. The phase accumulated during free precession encodes the field strength.

import numpy as np
import matplotlib.pyplot as plt
from scipy.optimize import curve_fit

# Physical constants
GAMMA_NV = 28.0e9        # NV gyromagnetic ratio: 28 GHz/T
T2_STAR = 1e-6           # NV dephasing time (microseconds, typical ensemble)
T2_COHERENCE = 1e-3     # Hahn echo coherence time (milliseconds)

def ramsey_signal(tau, B_field, T2_star=T2_STAR, contrast=0.3, noise_level=0.01):
    """
    Simulate NV Ramsey fringes as a function of free precession time tau.

    The spin precesses at frequency omega = gamma_NV * B_field.
    Fluorescence oscillates as cos(omega * tau) with Gaussian envelope decay.

    Parameters:
        tau: free precession time array (seconds)
        B_field: external magnetic field (Tesla)
        T2_star: NV dephasing time (seconds)
        contrast: fringe visibility (fluorescence contrast)
        noise_level: photon shot noise amplitude
    """
    omega = 2 * np.pi * GAMMA_NV * B_field
    envelope = np.exp(-(tau / T2_star) ** 2)
    fringes = contrast * envelope * np.cos(omega * tau)
    noise = noise_level * np.random.randn(len(tau))
    # Baseline fluorescence + fringes + noise
    return 1.0 - fringes + noise


def fit_ramsey_field(tau, signal):
    """
    Fit Ramsey fringe data to extract the B-field magnitude.
    """
    def model(t, B, T2, contrast, offset):
        omega = 2 * np.pi * GAMMA_NV * B
        return offset - contrast * np.exp(-(t / T2) ** 2) * np.cos(omega * t)

    p0 = [1e-4, T2_STAR, 0.3, 1.0]
    bounds = ([0, 1e-8, 0, 0.5], [1e-2, 1e-3, 1.0, 1.5])
    popt, pcov = curve_fit(model, tau, signal, p0=p0, bounds=bounds, maxfev=5000)
    B_fit = popt[0]
    B_uncertainty = np.sqrt(pcov[0, 0])
    return B_fit, B_uncertainty


# Simulate a defect detection scan
B_background = 50e-6    # 50 microtesla background (Earth's field)
B_defect = 5e-6         # 5 microtesla anomaly from a defect current

tau = np.linspace(0, 2e-6, 500)  # 0 to 2 microseconds

signal_clean = ramsey_signal(tau, B_background)
signal_defect = ramsey_signal(tau, B_background + B_defect)

B_measured_clean, uncertainty = fit_ramsey_field(tau, signal_clean)
B_measured_defect, _ = fit_ramsey_field(tau, signal_defect)

print(f"Background field: {B_measured_clean*1e6:.2f} uT (expected {B_background*1e6:.1f} uT)")
print(f"Defect field: {B_measured_defect*1e6:.2f} uT (expected {(B_background+B_defect)*1e6:.1f} uT)")
print(f"Field difference: {(B_measured_defect - B_measured_clean)*1e9:.1f} nT")
print(f"Measurement uncertainty: {uncertainty*1e9:.2f} nT")

Sensitivity Estimate

The fundamental sensitivity limit of an NV magnetometer is set by the coherence time and the number of NV centers:

def nv_sensitivity_estimate(T2_star, n_nv_centers, eta_collection=0.01):
    """
    Compute shot-noise-limited magnetic field sensitivity.

    eta_B = 1 / (gamma_NV * sqrt(T2* * N_NV * eta))
    Units: Tesla / sqrt(Hz)

    Parameters:
        T2_star: NV dephasing time (seconds)
        n_nv_centers: number of NV centers in sensing volume
        eta_collection: photon collection efficiency (typically 0.01-0.1)
    """
    sensitivity = 1.0 / (2 * np.pi * GAMMA_NV * np.sqrt(T2_star * n_nv_centers * eta_collection))
    return sensitivity


# Bosch lab setup: dense NV ensemble in diamond chip
n_nv = 1e12          # 1 trillion NV centers in a 1mm^3 chip
nv_sens = nv_sensitivity_estimate(T2_STAR, n_nv)

# Classical Hall sensor sensitivity
hall_sensitivity = 1e-8   # ~10 nT/sqrt(Hz) for best room-temp Hall sensors

print(f"NV magnetometer sensitivity: {nv_sens:.2e} T/sqrt(Hz)")
print(f"Classical Hall sensor: {hall_sensitivity:.2e} T/sqrt(Hz)")
print(f"Improvement factor: {hall_sensitivity / nv_sens:.1f}x")

Atom Interferometer for Inertial Navigation

In the second research thread, Bosch built an atom interferometer prototype for accelerometry. Cold rubidium-87 atoms are laser-cooled to microkelvin temperatures, then a sequence of laser pulses (beam splitter, mirror, recombiner in the Mach-Zehnder geometry) creates a quantum superposition of two trajectories. The phase difference at recombination is proportional to acceleration.

The advantage over MEMS: the scale factor is set by atomic physics constants (laser wavelength, atomic mass, pulse timing), not mechanical fabrication tolerances. There is no drift caused by material creep or temperature cycling.

def atom_interferometer_phase(acceleration, T_interrogation, k_eff):
    """
    Compute phase shift in Mach-Zehnder atom interferometer.

    Phase = k_eff * acceleration * T^2
    where T is the interrogation time between pulses.

    Parameters:
        acceleration: input acceleration (m/s^2)
        T_interrogation: time between pi/2 and pi pulses (seconds)
        k_eff: effective wavevector of Raman laser (rad/m)
    """
    phase = k_eff * acceleration * T_interrogation ** 2
    return phase


# Rb-87 D2 line parameters
LAMBDA_RB = 780e-9     # 780 nm wavelength
K_EFF = 4 * np.pi / LAMBDA_RB  # double-diffraction geometry

T_interrogation = 0.1   # 100 ms interrogation time
g = 9.81                # Earth gravity (m/s^2)

phi_gravity = atom_interferometer_phase(g, T_interrogation, K_EFF)
sensitivity = 1.0 / (K_EFF * T_interrogation ** 2)  # rad per m/s^2

print(f"Phase shift from gravity: {phi_gravity:.2f} rad")
print(f"Acceleration sensitivity: {sensitivity:.4e} m/s^2 per radian")
print(f"Equivalent to: {sensitivity/g*1e6:.3f} micro-g per radian")

Results

Bosch’s NV-center magnetometry setup demonstrated a factor of 10 improvement in sensitivity over their benchmark classical Hall sensor at room temperature. Detection of individual wire-level defects in a 4-layer PCB (current anomalies from a 100-micron crack) was achieved without contact or destructive sectioning.

The atom interferometer prototype reached sensitivity of approximately 50 nano-g per root Hz in lab conditions, matching the best commercially available fiber optic gyroscope (FOG) performance on a prototype system that does not yet have FOG-level packaging maturity.

Bosch filed patents on diamond chip integration for inline PCB inspection in 2023. NV-center sensor products for automotive quality control are in pre-commercial development, with industrial availability targeted before 2027.

Context: Quantum Sensing vs. Quantum Computing

This case study involves quantum sensing, not gate-based quantum computing. The NV center and atom interferometer exploit quantum coherence and superposition as measurement resources, not as computational resources. This distinction matters commercially: quantum sensing products are closer to market, can be manufactured today, and do not require error correction or large qubit counts.

Bosch’s quantum strategy is layered: quantum sensing for near-term commercial value, quantum computing partnerships (with IBM and others) for longer-term chemistry and optimization applications.

Learn more: IBM Quantum Reference