From Cantilevers to Membranes: Advanced Scanning Protocols for Magnetic Resonance Force Microscopy

TL;DR

The paper tackles slow, fidelity-limited 3D MRFM imaging by comparing cantilever and membrane resonators, and by introducing multislice scanning and compressed sensing to dramatically accelerate acquisition. It develops ADMM-based reconstructions that exploit forward models linking spin density to force-variance measurements through PSFs, while leveraging 2D Fourier diagonalization for computational efficiency. The main findings show that two-dimensional spatial sampling combined with a frequency dimension improves reconstruction fidelity by 2–5×, and that 50% subsampling via compressed sensing can achieve similar fidelity with proportional speed-ups; membranes outperform cantilevers in many scenarios due to deeper, better-conditioned PSFs and lower noise. These advances point toward practical, high-resolution volumetric MRFM of biological nanostructures with substantially reduced acquisition times and improved robustness to noise.

Abstract

Magnetic Resonance Force Microscopy (MRFM) enables three-dimensional imaging of nuclear spin densities in nanoscale objects. Based on numerical simulations, we evaluate the performance of strained SiN resonators as force sensors and show that their out-of-plane oscillation direction improves the quality of the reconstructed sample. We further introduce a multislice, compressed-sensing scan protocol that maximizes the information obtained for a given measurement time. Our simulations predict that these new scanning protocols and optimized algorithms can shorten the total acquisition time by up to two orders of magnitude while maintaining the reconstruction fidelity. Our results demonstrate that combining advanced scanning protocols with state-of-the-art resonators is a promising path toward high-resolution MRFM for volumetric imaging of biological nanostructures.

Figures (6)

• Figure 1: Analyzed geometries. (a) Illustration of MRFM setup utilizing cantilever-style force sensors. The spin ensemble (pink) is attached to the tip of the resonator and positioned above the magnetic field gradient source (gray). The resonator's oscillation direction is along $x$ and parallel to the gradient source's top surface. When spin inversion pulses are applied via an external antenna, all spins inside the resonance slice (purple) are inverted periodically. (b) Illustration of a setup utilizing tensioned resonators as force sensors. The sample is still attached to the resonator, but the oscillation direction is along $z$, perpendicular to the gradient source's top surface, changing the direction of the detected force. (c) PSF of a cylindrical nanomagnet when the cantilever geometry is used. Its magnitude is proportional to $|\partial B_z/\partial x|^2$. The nanomagnet's radius is $50nm$ and it has a rounded edge with radius $r_\mathrm{edge} = 10nm$. (d) PSF of the identical magnet for a membrane resonator geometry. The strength of the PSF is determined by the derivative $|\partial B_z/\partial z|^2$. (e) Timing diagram of spin inversions. Twice per resonator oscillation, a spin inversion pulse (blue) is applied. This results in a force generated by the ensemble (pink) that is resonant with the resonator's movement (orange) leading to a detectable driving force.
• Figure 2: Mid-plane of the spin density ground truth $\mathbf{O}_\mathrm{GT}$ of the spherical test. The object has a diameter of $40nm$ and a mean hydrogen spin density of $60\per nm\cubed$. The spin density varies with a standard deviation of $\sigma = 18.9\per nm\cubed$ over a characteristic distance of $5nm$. The inset shows the three-dimensional position of the displayed slice in the sample.
• Figure 3: Comparison of scanning protocols. (a-d) Mid-plane $(x,y)$ slices of the reconstructed three-dimensional volumes for an effective integration time of $T_\mathrm{m}=30s$ per measurement point (at the reconstruction iteration that produces minimal RMSE error). (a-b) utilize cantilever resonators with the XYZ scanning protocol (a) and the proposed multislice protocol (b). (c-d) employs membrane resonators, again following the XYZ (c) and multislice (d) protocols. In (e) reconstruction results are compared between scanning protocols and measurement geometries for resonators with thermal force noise density of $S_\mathrm{F} =10aN\per\sqrt Hz$ over varying measurement times $T_\mathrm{m}$. (f) shows the reconstruction error using the membrane geometry and the multislice scanning protocol [blue line in (e)] for resonators with decreasing $S_\mathrm{th}$ between $S_\mathrm{th} = 10aN\per\sqrt Hz$ and $S_\mathrm{th} = 0aN\per\sqrt Hz$, with decrements of $2aN\per\sqrt Hz$.
• Figure 4: Impact of compressed sensing on the multislice scanning protocol. (a) Mid-plane of the spin density ground truth. (b-f) Mid-plane $(x,y)$ slices of reconstructed volume, obtained by applying compressed sensing to a simulated measurement employing membrane resonators, a data point integration time of $T_\mathrm{m}=60s$, and sampling ratios (b) $p = 1$, (c) $p= 1/2$, (d) $p= 1/3$, (e) $p= 1/6$, and (f) $p= 1/10$. (g) Relative reconstruction errors of the sub-sampled measurement compared to a scan with identical point integration times $T_\mathrm{m}$ and sampling ratio $p=1$.
• Figure 5: Computed multislice SNR enhancement as a function of the variance ratio $r_\mathrm{var} = \sigma^2_\mathrm{spin}/\sigma^2_\mathrm{th}$ for several values of addressed inversion slice $n_\mathrm{f}$ per resonator position.