arXiv:2609.38739
Geometric estimation of NV charge-state contributions from a low-dimensional spectral representation
Yuto Yamakawa, Keisuke Oshimi, Keigo Arai
Abstract
Nitrogen-vacancy (NV) centers in diamond exist in neutral (NV0) and negatively charged (NV-) states, and quantifying their respective photoluminescence (PL) contributions is important for charge-state-based measurements. Existing methods either require additional experimental control or may suffer from limited physical identifiability. Here, we introduce a geometric method for determining NV charge-state contributions from PL spectra acquired at different excitation intensities. Each spectrum is mapped into a low-dimensional space through broadband spectral weighting, where changes in the relative NV0 and NV- contributions trace a one-dimensional trajectory. Zero-phonon-line (ZPL) information from two spectra physically calibrates this trajectory, enabling the NV-PL contribution to be determined by geometric projection. Measurements on two bulk single-crystal diamond samples yielded NV- contributions in close agreement with those obtained using an independent dual-excitation reference-spectrum method, with root-mean-square errors of 1.11 and 0.31 percentage points. Under additive spectral noise, the proposed method exhibited more than an order of magnitude less variation than a ZPL-only method. Fisher-information analysis further showed that the three-dimensional CIE XYZ representation retained approximately 81% of the information about the NV- contribution available in the full spectrum. These results establish a physically calibrated approach to low-dimensional spectral estimation that combines reduced experimental overhead with robust parameter estimation and provides a general framework for spectral sensing governed by a small number of physical degrees of freedom.
AI Summary
The paper addresses estimation of the negatively charged NV center’s photoluminescence (PL) contribution, $r=P_-/(P_0+P_-)$, from overlapping NV$^0$ and NV$^-$ spectra without additional excitation wavelengths or separately measured reference spectra. It estimates PL fractions rather than charge-state populations. Assuming fixed component spectral shapes, area-normalized spectra form an affine line under any linear spectral mapping. The implementation projects spectra onto three CIE XYZ weighting functions. Zero-phonon-line (ZPL) areas from two excitation-power settings determine a sample-specific conversion factor and calibrate the line’s pure-component endpoints; subsequent fractions follow by geometric projection. This calibration avoids literature-derived Debye–Waller factors. Measurements comprise nine excitation-power settings on each of two bulk single-crystal diamonds. Estimates agree with a dual-excitation reference-spectrum method with root-mean-square differences of 1.11 and 0.31 percentage points. PCA of the full spectra assigns 99.81% and 99.83% of variance to the first component. Monte Carlo tests adding white Gaussian spectral noise, with calibration held fixed, give approximately 14–15-fold smaller estimation standard deviations than a ZPL-only method. A local, first-order Fisher-information calculation incorporating normalization-induced covariance finds that CIE XYZ retains approximately 81% of full-spectrum information, versus 0.80% for the ZPL observables, for one test spectrum. The noise analysis excludes calibration uncertainty; calibration becomes sensitive when the selected spectra have similar ZPL areas. Temperature, strain, background, or additional emitters can invalidate the fixed two-component model. CIE weights are not information-optimal, and implementation with spectrally weighted detection channels remains a proposed extension.
OpenAI ChatGPT (GPT-5.6 Sol) was used as an assistive tool in portions of the mathematical and computational analysis
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 19 pages, 11 figures