Artificial intelligence enhanced two-dimensional nanoscale nuclear magnetic resonance spectroscopy

At a Glance

Metadata	Details
Publication Date	2020-09-16
Journal	npj Quantum Information
Authors	Xi Kong, Leixin Zhou, Zhijie Li, Zhiping Yang, Bensheng Qiu
Institutions	Nanjing University, Hefei National Center for Physical Sciences at Nanoscale
Citations	18
Analysis	Full AI Review Included

Technical Documentation & Analysis: AI-Enhanced Nanoscale 2D NMR Spectroscopy

This document analyzes the research demonstrating AI-enhanced two-dimensional nanoscale Nuclear Magnetic Resonance (NMR) using Nitrogen-Vacancy (NV) centers in diamond. It outlines the technical achievements and connects the material requirements directly to 6CCVD’s advanced MPCVD diamond solutions, positioning 6CCVD as the essential supplier for replicating and scaling this quantum sensing technology.

Executive Summary

The research successfully integrates deep learning (DL) and matrix completion (MC) algorithms (DLMC) with NV-center quantum sensing to dramatically accelerate 2D nanoscale NMR, a critical tool for molecular structure analysis.

Core Achievement: Demonstrated 2D nanoscale NMR using a single NV center in CVD diamond coupled to a ¹³C nuclear spin cluster.
Efficiency Breakthrough: The DLMC protocol successfully reconstructed the full 2D spectrum from only 10% sampled data, reducing experimental acquisition time by an order of magnitude.
Sensitivity Improvement: The Signal-to-Noise Ratio (SNR) was enhanced by 5.7 ± 1.3 dB compared to conventional Fourier Transform (FFT) reconstruction.
Noise Suppression: The AI-enhanced protocol intrinsically suppresses observation noise, leading to improved sensitivity and fidelity (lowest Root Mean Square Error, RMSE).
Material Requirement: Requires ultra-high quality, low-defect Single Crystal Diamond (SCD) substrates to host stable NV centers with long coherence times.
Application Potential: Enables high-speed, high-resolution 2D NMR for single-molecule structure determination under ambient conditions.

Technical Specifications

The following hard data points were extracted from the experimental results and analysis of the DLMC method applied to 2D nanoscale NMR.

Parameter	Value	Unit	Context
SNR Enhancement (DLMC)	5.7 ± 1.3	dB	At 10% sampling coverage, compared to original FFT
Minimum Sampling Coverage	10	%	Required for successful 2D spectrum recovery
Experimental Time Reduction	Order of magnitude	N/A	Achieved via sparse sampling and AI reconstruction
External Magnetic Field (B)	1580 (158)	Gauss (mT)	Applied along the main axis of the NV sensor
¹³C Larmor Frequency ($\omega_L$)	1.69	MHz	Frequency of the ¹³C nuclear spin
Bond Length Sensitivity	0.3	nm/√Hz	Enhanced from 0.8 nm/√Hz due to noise suppression
Spectrum Matrix Size	50 x 50	N/A	Data points collected in $t_1$ and $t_2$ time domains
Reconstruction Fidelity (RMSE)	0.018 ± 0.002	N/A	Achieved on simulated multiple nuclear spin systems (10% sampled)

Key Methodologies

The experiment utilized a quantum sensor based on a single NV center in diamond, controlled via microwave and laser pulses, with data reconstruction accelerated by a hybrid AI approach.

Material and Setup: A single NV center was utilized in a CVD-grown diamond with natural ¹³C abundance (1.1%). An external magnetic field (158 mT) was applied along the NV axis.
NMR Protocol: A two-dimensional protocol analogous to conventional COSY (Correlation Spectroscopy) was performed to reveal coupling between two nuclear spins (a coupled nuclear spin dimer).
Pulse Sequence: The protocol involved nuclear spin initialization (via laser and microwave pulses), free evolution periods ($t_1$ and $t_2$), a half $\pi$ pulse, and a final correlation readout pulse sequence.
Data Acquisition: The time parameters $t_1$ and $t_2$ were swept from 4 µs to 0.9 ms, generating a 50 x 50 spectrum matrix. Due to time constraints, only 80% of the data was initially collected.
DLMC Algorithm Training: A deep learning encoder-decoder Convolutional Neural Network (DLNet) was trained using simulated 2D NMR spectra to learn complex non-linear mappings from partially filled maps to full-resolution maps.
Hybrid Reconstruction: The DLMC method combined the DLNet output with a post-processing step using the classical Singular Value Thresholding (SVT) Matrix Completion (MC) algorithm. This step enhanced the low-rank property of the reconstructed map and mitigated the domain shift problem between simulated training data and real experimental data.

6CCVD Solutions & Capabilities

Replicating and advancing this cutting-edge nanoscale quantum sensing research requires diamond materials with exceptional purity, precise isotopic control, and superior surface quality. 6CCVD is uniquely positioned to supply the necessary MPCVD diamond substrates.

Applicable Materials

To achieve the long coherence times ($T_2$) and high sensitivity required for NV-center quantum sensing, researchers need diamond with minimal defects and controlled isotopic composition.

Optical Grade Single Crystal Diamond (SCD): Essential for hosting stable NV centers. 6CCVD provides high-purity SCD substrates (up to 500 µm thick) with extremely low nitrogen and defect concentrations, ensuring maximum NV center stability and long $T_2$ coherence times.
Isotopically Controlled Diamond: The paper used natural abundance ¹³C (1.1%). For next-generation experiments requiring ultra-long coherence times or targeted nuclear spin environments, 6CCVD offers ultra-high purity ¹²C enriched SCD (e.g., < 0.1% ¹³C). This minimizes background nuclear spin noise, crucial for extending $T_2$ and improving sensitivity further.
Boron-Doped Diamond (BDD): While not used in this specific NV sensing application, 6CCVD offers BDD films for related electrochemical or quantum device applications requiring conductive diamond.

Customization Potential

The success of nanoscale NMR relies on precise device integration and surface engineering. 6CCVD offers comprehensive customization services to meet the exact needs of quantum engineers.

Research Requirement	6CCVD Customization Service	Technical Benefit
Ultra-Smooth Surface	Precision Polishing (Ra < 1 nm)	Critical for minimizing surface-related decoherence and ensuring the target molecule is placed within nanometers of the NV sensor. Our SCD polishing achieves Ra < 1 nm.
Integrated Control Lines	Custom Metalization (Au, Pt, Ti, W, Cu)	We provide in-house deposition of thin-film metal contacts for microwave strip lines and electrical gates, necessary for applying the complex $\pi/2$ and $\pi$ pulse sequences used in the COSY protocol.
Unique Dimensions	Custom Dimensions & Laser Cutting	We supply SCD plates and PCD wafers up to 125 mm in diameter, with custom laser cutting capabilities to produce specific chip geometries required for integration into quantum setups.
Thick Substrates	Substrates up to 10 mm	For robust mechanical support or specific thermal management needs in advanced quantum setups, we offer substrates up to 10 mm thick.

Engineering Support

6CCVD’s in-house PhD team specializes in MPCVD growth and diamond material science for quantum applications. We offer expert consultation to assist researchers in selecting the optimal diamond specifications (purity, isotopic ratio, thickness, and surface termination) necessary to replicate or extend this nanoscale NMR research.

For custom specifications or material consultation, visit 6ccvd.com or contact our engineering team directly.

View Original Abstract

Abstract Two-dimensional nuclear magnetic resonance (NMR) is indispensable to molecule structure determination. Nitrogen-vacancy center in diamond has been proposed and developed as an outstanding quantum sensor to realize NMR in nanoscale or even single molecule. However, like conventional multi-dimensional NMR, a more efficient data accumulation and processing method is necessary to realize applicable two-dimensional (2D) nanoscale NMR with a high spatial resolution nitrogen-vacancy sensor. Deep learning is an artificial algorithm, which mimics the network of neurons of human brain, has been demonstrated superb capability in pattern identifying and noise canceling. Here we report a method, combining deep learning and sparse matrix completion, to speed up 2D nanoscale NMR spectroscopy. The signal-to-noise ratio is enhanced by 5.7 ± 1.3 dB in 10% sampling coverage by an artificial intelligence protocol on 2D nanoscale NMR of a single nuclear spin cluster. The artificial intelligence algorithm enhanced 2D nanoscale NMR protocol intrinsically suppresses the observation noise and thus improves sensitivity.

Tech Support

Original Source

DOI: https://doi.org/10.1038/s41534-020-00311-z