Nuclear spin relaxation in zero- to ultralow-field magnetic resonance spectroscopy

Extracted source text · arXiv 2511.08517

% ****** Start of file apssamp.tex ******
%
%   This file is part of the APS files in the REVTeX 4.2 distribution.
%   Version 4.2a of REVTeX, December 2014
%
%   Copyright (c) 2014 The American Physical Society.
%
%   See the REVTeX 4 README file for restrictions and more information.
%
% TeX'ing this file requires that you have AMS-LaTeX 2.0 installed
% as well as the rest of the prerequisites for REVTeX 4.2
%
% See the REVTeX 4 README file
% It also requires running BibTeX. The commands are as follows:
%
%  1)  latex apssamp.tex
%  2)  bibtex apssamp
%  3)  latex apssamp.tex
%  4)  latex apssamp.tex
%
\documentclass[%
 reprint,
 superscriptaddress,
%groupedaddress,
%unsortedaddress,
%runinaddress,
%frontmatterverbose, 
%preprint,
%preprintnumbers,
%nofootinbib,
%nobibnotes,
%bibnotes,
 amsmath,amssymb,
 aps,
%pra,
%prb,
%rmp,
%prstab,
%prstper,
%floatfix,
]{revtex4-2}

\usepackage{graphicx}% Include figure files
\usepackage{dcolumn}% Align table columns on decimal point
\usepackage{bm}% bold math
\usepackage{braket}
\usepackage{xcolor}
\usepackage{amsmath}
\usepackage{upgreek}
\usepackage{multirow} % put in preamble
\usepackage{textalpha}
\usepackage{xr}  % for cross-referencing between Main and SI

\externaldocument{supplementary}

\newcommand{\FT}[1]{\textcolor{teal}{#1}} % Florin
\newcommand{\AMF}[1]{\textcolor{purple}{#1}} %Anne
\newcommand{\DB}[1]{{\color{blue}{{#1}}}} %Dima
\newcommand{\GC}[1]{{\color{red}{{#1}}}} %Gary
%\usepackage{hyperref}% add hypertext capabilities
%\usepackage[mathlines]{lineno}% Enable numbering of text and display math
%\linenumbers\relax % Commence numbering lines

%\usepackage[showframe,%Uncomment any one of the following lines to test 
%%scale=0.7, marginratio={1:1, 2:3}, ignoreall,% default settings
%%text={7in,10in},centering,
%%margin=1.5in,
%%total={6.5in,8.75in}, top=1.2in, left=0.9in, includefoot,
%%height=10in,a5paper,hmargin={3cm,0.8in},
%]{geometry}


\begin{document}

\preprint{APS/123-QED}

\title{Nuclear spin relaxation in zero- to ultralow-field magnetic resonance spectroscopy}%

\author{Florin Teleanu}
\affiliation{Department of Chemistry, New York University, New York, NY 10003, United States}
\affiliation{ELI-NP, “Horia Hulubei” National Institute for Physics and Nuclear Engineering, 30 Reactorului Street, Bucharest-Magurele, 077125, Ilfov, Romania}
\author{Anne M. Fabricant}
\altaffiliation{Current address: Department of Biosignals, Physikalisch-Technische Bundesanstalt (PTB), 10587 Berlin, Germany}
\affiliation{Institute of Physics, Johannes Gutenberg University of Mainz}
\affiliation{Helmholtz Institute Mainz, 55099 Mainz, Germany}
\affiliation{GSI Helmholtzzentrum für Schwerionenforschung, 64291 Darmstadt, Germany}
\author{Chengtong Zhang}
\affiliation{Department of Chemistry, New York University, New York, NY 10003, United States}
\author{Gary P. Centers}
\affiliation{Institute of Physics, Johannes Gutenberg University of Mainz}
\affiliation{Helmholtz Institute Mainz, 55099 Mainz, Germany}
\affiliation{GSI Helmholtzzentrum für Schwerionenforschung, 64291 Darmstadt, Germany}
\author{Dmitry Budker}
\affiliation{Institute of Physics, Johannes Gutenberg University of Mainz}
\affiliation{Helmholtz Institute Mainz, 55099 Mainz, Germany}
\affiliation{GSI Helmholtzzentrum für Schwerionenforschung, 64291 Darmstadt, Germany}
\affiliation{Department of Physics, University of California, Berkeley, CA 94720, USA}
\author{Danila A. Barskiy}
\altaffiliation{Current address: Frost Institute for Chemistry and Molecular Science, 
Department of Chemistry, University of Miami, Coral Gables, FL 33146, USA}
\affiliation{Institute of Physics, Johannes Gutenberg University of Mainz}
\affiliation{Helmholtz Institute Mainz, 55099 Mainz, Germany}
\affiliation{GSI Helmholtzzentrum für Schwerionenforschung, 64291 Darmstadt, Germany}
\author{Alexej Jerschow}
\email{Contact author: alexej.jerschow@nyu.edu}
\affiliation{Department of Chemistry, New York University, New York, NY 10003, United States}

\date{\today}% It is always \today, today,
             %  but any date may be explicitly specified

\begin{abstract}
Nuclear-magnetic-resonance experiments can interrogate a broad spectrum of molecular-tumbling regimes and can accurately measure interatomic distances in solution with sub-nanometer resolution. In the zero- to ultralow-field (ZULF) regime, population and coherence decay reveal nontrivial behavior due to strong coupling between nuclear spins. We note, in particular, the surprising effects that different resonances show different relaxation rates, depending on  (i) the (pre)polarizing magnet field and the shuttling trajectory to the detection region at nano- and microtesla fields, (ii) the strength of the measurement field, (iii) the detection method (single-channel or quadrature), and even (iv) the nutation angle induced by the excitation pulse. We describe herein experimental data of relaxation rates measured for a $^{13}$C-labeled formic acid sample, with an atomic-magnetometer-based ZULF setup, and develop a theoretical framework to explain the detected effects and extract molecular properties. The observed effects could be used for spectral assignment, for the establishment of specific motional regimes, for image contrast, and for the characterization of relaxation processes at nano- to microtesla magnetic fields.   

\end{abstract}

\maketitle


% \section{\label{sec:level1} Introduction}

In nuclear magnetic resonance (NMR) experiments of liquids, molecular tumbling produces fluctuating interactions, which induce nuclear spin flips and ultimately establish equilibrium spin populations\cite{abragam_principles_1961,levitt_spin_2013,kuprov_spin_2023,kowalewski_nuclear_2019}. %Consequently, the coupling between spin and motion degrees of freedom dictates the amount of detectable NMR signal as well as the lifetimes of induced perturbations following radio-frequency pulses. 
Monitoring these recovery processes, described by measured relaxation rates, provides a characterization of sub-millisecond molecular dynamics\cite{charlier_protein_2016,stenstrom_how_2022}, inter-particle distances via the Overhauser effect\cite{overhauser_polarization_1953,noggle_nuclear_1971,kumar_two-dimensional_1980}, relative orientations and angles through cross-correlated interactions\cite{kumar_cross-correlations_2000}, and the local electric and magnetic fields impacting quadrupolar nuclei\cite{jerschow_nuclear_2005}.

The relaxometry framework is typically used in the context of high-field NMR, since chemical shift resolution and sensitivity improve significantly with the strength of the external field\cite{cavanagh_protein_1995,abragam_principles_1961}. Field-cycling (FC) experiments\cite{kimmich_field-cycling_2018,bolik-coulon_comprehensive_2023} can be used to monitor relaxation rates (usually $T_1$) at much lower fields (down to hundreds of nT) using mechanical shuttling to move the sample between the field at which relaxation processes are allowed to progress, and the field at which signals are detected. 
%In this manner, the spectral density profile of the fluctuating interactions that are resonsible for nuclear spin transitions can be mapped over a large range of magnetic fields (and resonance frequencies) spanning several orders of magnitude\cite{bolik-coulon_comprehensive_2023}. 
%The Bloch-Redfield-Wangsness (BRW) model\cite{goldman_formal_2001} is generally used for evaluating the transition probability between energy levels in spin systems leading to the decoherence of spin order and thermalization.

Here, we analyze the relaxation rates measured at zero-to-ultralow fields (ZULF), where scalar coupling dominates over the Zeeman interaction\cite{barskiy_zero-_2025}, and build a theoretical framework for their interpretation and straightforward connection to internuclear distances and molecular tumbling. In contrast to high-field FC experiments, our experiments are performed by shuttling a prepolarized sample through a guiding field to the detection region for storage and subsequent measurement. The detection of nuclear spin signals is performed using optical atomic magnetometers\cite{budker_optical_2007,fabricant_proton_2024} and takes place at nT--$\upmu$T fields where Zeeman and scalar coupling interactions have similar strength. We observe several surprising and counter-intuitive effects in this regime, including the dependence of peak lifetimes on both pulse flip angles and on spin-system preparation by magnetic-field cycling, as well as the observation of different rates for different transitions in the ZULF spectra. Such behavior is distinctly different from that observed in high-field FC experiments\cite{bodenstedt_fast-field-cycling_2021,alcicek_zero-_2023,zhukov_field-cycling_2018,zhukov_assessment_2019}, and therefore represents a missing piece in regimes where inductive coils are inefficient for detection. We also provide a model-based description of polarization lifetimes in experiments performed on $^{13}$C-labeled formic acid at different ultralow fields and show how to derive molecular properties from observed effects.

% \newpage
\begin{figure}[htb]
\includegraphics[page=1, trim=30pt 25pt 50pt 50pt, clip, width=0.5\textwidth]{Figures/Figures_ZULF.pdf}\caption{Breit-Rabi diagram showing the shifts of energy levels as a function of magnetic field. The populations have been adiabatically transferred from high to low field. Filled and empty circles represent overpopulated and underpopulated states, respectively. The diagram was calculated for  the [$^{13}$C-$^{1}$H] pair in formic acid ($J_{IS} = 222$\,Hz).}
\label{Breit_Rabi}
\end{figure}


Consider a heteronuclear two-spin-1/2 system with spin operator labels \textit{\{I, S\}} characterized by the $\gamma_{I}$ and $\gamma_{S}$ gyromagnetic ratios (in $\text{rad} \cdot \text{s}^{-1} \cdot $T$^{-1}$) and a scalar coupling constant $J_{IS}$ (in Hz). The active space can be described in terms of the field-dependent eigenstates $\{\ket{\psi_{i}}_{i=1,4}\}$ (See Appendix \ref{appenx:eigenstates}) which range from the high-field Zeeman basis to the singlet-triplet basis at zero field \cite{levitt_spin_2013}. In ZULF experiments, it is common to shuttle the sample from a (pre)polarizing region characterized by a higher homogeneous field $B_{p}$ to a region shielded from external sources of magnetic fields. Here, the main field $B_0$ is generated with a guiding solenoid that is controlled by the user. At the beginning of the experiment, the populations of the Zeeman states are:
\begin{equation}
    p_{\alpha\alpha}=\frac{1}{4}+\delta;p_{\alpha\beta}=\frac{1}{4}+\epsilon ; p_{\beta\alpha}=\frac{1}{4}-\epsilon; p_{\beta\beta}=\frac{1}{4}-\delta \,,
    \label{eq:populations}
\end{equation}
where $\delta=\frac{B_{p}(\gamma_{I}+\gamma_{S})}{8k_{B}T}$ and $\epsilon=\frac{B_{p}(\gamma_{I}-\gamma_{S})}{8k_{B}T}$ with $k_B$ as the Boltzmann constant and $T$ the temperature. The thermal equilibrium density operator at this field is given by $\hat{\rho}_{\text{eq}} (B_{p})  \approx \frac{1}{4k_{B}T}( - \gamma_{I}B_{p}\hat{I}_{z}-\gamma_{S}B_{p}\hat{S}_{z})$, where we neglect the identity matrix as it does not contribute to the observed signal. During shuttling, the high-field populations can be transferred adiabatically or suddenly to the eigenstates at the final field $B_0$. Based on our analyses (see Sections I and II in Supporting Information), we estimate that our experimental shuttling setup transfers populations adiabatically (Fig.\,\ref{Breit_Rabi}), such that the high-field density operator rewritten as $\hat{\rho}_{\text{eq}}{} =(\delta+\epsilon)\hat{I}_{z} +(\delta-\epsilon)\hat{S}_{z}$ transforms to

\begin{equation}
    \hat{\rho}_{\text{ad}} = \delta \hat{D}_{z} - \epsilon\left(\sin{(2\theta)}\hat{Z}_{x}+\cos{(2\theta)}\hat{Z}_{z}\right ) \,.
    \label{eq:rho_adiabatic}
\end{equation}
Here, $\theta = \frac{1}{2}\arctan\left (\frac{2\pi J_{IS}}{|\gamma_{I}-\gamma_{S}|B_{0}}\right )$ is known as the `mixing angle'\cite{pileio_singlet_2017} and we adopt the notation\cite{harris_zero-_2016} $\hat{D}_z = \hat{I}_z + \hat{S}_z$, $\hat{Z}_x = 2\hat{I}_x\hat{S}_x + 2\hat{I}_y\hat{S}_y$, $\hat{Z}_y = 2\hat{I}_x\hat{S}_y - 2\hat{I}_y\hat{S}_x$, $\hat{Z}_z = \hat{I}_z - \hat{S}_z$. 


%Experimentally, the magnetic-field profile along the shuttling axis is mapped using classical magnetometers, and the spatial transfer of the sample can be tuned to achieve desired rates of change in the experienced magnetic fields. Implementing the 3D spatial profile of the magnetic fields along the shuttling path and approximating the velocity profile as constant, we ran Spinach\cite{hogben_spinach_2011} simulations to evaluate how populations are transferred in our setup (See Section \ref{sec:SI_shuttle} in Supporting Information). Our results show that for a $^{13}$C-$^{1}$H heteronuclear spin system characteristic for the [$^{13}$C]-formic acid sample ($J_{\text{CH}}=222$\,Hz), shuttling is performed adiabatically, such that final density operator is given by Eq.\,\ref{eq:rho_adiabatic}. Depending on how the shuttling from the pre-polarizing to the detection region is performed, operators characterized by different lifetimes will be populated. This will directly impact the observed effective decay rates of various peaks as a function of storage time.


%It is of note that this differential line broadening arises from a single interaction (the dipolar coupling), unlike known differential line-broadening effects in high-field NMR, which require the interplay between two different interactions\cite{whipham_cross-correlated_2024,ling_relaxation-allowed_2007}.
%In the limiting case of high field ($\theta \rightarrow 0$), all four transitions have the same linewidth, as expected for a dipolar- and scalar-coupled heteronuclear two-spin-1/2 system\cite{freeman_highresolution_1970}. Figure \ref{nutation_R1_R2_DD_RFF}a highlights the field dependence of the relaxation rates due to dipolar interactions described above.

\begin{figure*}[htb]
\includegraphics[trim=0pt 85pt 0pt 80pt, clip, width=\textwidth]{Figures/Figure2_big.pdf}\caption{a) Energy-level diagram for the [$^{13}$C-$^{1}$H] system of $^{13}$C-labeled formic acid at ultralow field ($B_{0}=2.7\,\upmu$T and $J_{CH}=222$\,Hz). Observed transitions $\nu_{ij}$ connect nuclear-spin eigenstates, satisfying the selection rule $\Delta m_{f} = \pm 1$. b) Example of the simulated ULF spectrum corresponding to either single-channel detection or dual-channel quadrature detection. Reduced spectral crowding, absolute determination of peak frequencies, and increased signal amplitude are achieved using quadrature detection
%\AMF{the SNR enhancement is sensor-dependent and mainly comes from the gradiometry; the sim here shows single-channel quadrature}.
c) Field dependence of the observable peak frequencies. d) Simulated nutation profile (Eqs.\,\ref{eq:nut13}--\ref{eq:nut24}) of the observable coherences $\hat{\rho}_{ij}=\ket{\psi_{i}}\bra{\psi_{j}}$ under a DC pulse along the $y$-axis, characterized by a magnetic field $B_{y}=30\,\upmu$T and total duration $t_p=1.6$\,ms; Theoretical relaxation rates of populations (e) and observable coherences (f) under dipolar interactions (DD) as a function of mixing angle $\theta$ for a [$^{13}$C-$^{1}$H] system in the fast-tumbling regime. The internuclear distance is $r_{\text{CH}}=1.1$\,\AA \ and the rotational correlation time is $\tau_{c}^{\text{DD}}=3.24$\,ps. Theoretical relaxation rates of populations (g) and observable coherences (h) under random field fluctuations (RFF), as a function of the mixing angle $\theta$ for a [$^{13}$C-$^{1}$H] system in the fast-tumbling regime. Parameters used for the simulations are $\omega_{\text{rms}}^{\text{C}}=85.32\cdot 10^3\,$rad$;\omega_{\text{rms}}^{\text{H}}=49.63 \cdot10^{3}\,$rad$; \kappa_{\text{CH}}=0.49$, and $\tau_{c}^{\text{RFF}}=19.34$\,ps.}
\label{nutation_R1_R2_DD_RFF}
\end{figure*}

%Short microsecond DC pulses applied along the $y$-axis characterized by a magnetic field strength $B_{y}$ and a duration of $t_{p}$ are described by the Hamiltonian $\hat{H}_{p} = (\gamma_{I}\hat{I}_{y} + \gamma_{S}\hat{S}_{y})B_{y}$ and convert populations into coherences.

Figure \ref{nutation_R1_R2_DD_RFF}a shows the energy-level diagram (along with population) of the {$^{13}$C-$^1$H} spin system, highlighting the observed transitions generated by superpositions of states satisfying the selection rule $\Delta m_f = \pm 1 $, with $m_f$ being the projection of the total spin angular momentum $F$ along the quantization axis\cite{stern_simulation_2023}. 
In order to convert the stationary density operator from Eq.\,\ref{eq:rho_adiabatic} into detectable magnetization, we apply short microsecond DC pulses along axes perpendicular to that of the small background magnetic field $B_{0}$ and measure the precession of polarization in the transverse plane. 
%To compute the evolution of the density operator during a DC pulse along the $y$-axis after adiabatic shuttling, we solve the sandwich equation $\hat{\rho}_{1}(t_{p}) = e^{-i(\hat{H}_{p}+\hat{H}_{0})t_{p}}\hat{\rho}_{ad}e^{i(\hat{H}_{p}+\hat{H}_{0})t_{p}}$. Then, the spectrum is computed by propagating $\hat{\rho}_{1}(t_p)$ under the free-precession Hamiltonian $\hat{H}_{0}$ using the sandwich relation $\hat{\rho}_{\text{det}}=e^{-i\hat{H}_{0}t_{2}}\hat{\rho}_{1}(t_p)e^{i\hat{H}_{0}t_{2}}$, where $t_2$ is the acquisition time. 
Depending on the experimental setup, detection can be performed with one or multiple magnetometers in configurations that allow for enhanced signal-to-noise ratios and unequivocal determination of the direction of precession with respect to the laboratory frame. If a single-axis detector is placed near the sample, only a cosine-modulated signal is detected with no information regarding the direction of precession. This measurement setup results in spectral crowding as all detected transitions lead to both positive and negative peaks (Fig.\,\ref{nutation_R1_R2_DD_RFF}b). To overcome these issues, we implemented a quadrature detection scheme, as described previously\cite{fabricant_proton_2024}. Thus, we are able to discriminate the absolute sign of the detected peaks as well as increase the overall SNR and simplify the spectrum. The two different configurations are exemplified in Fig.\,\ref{nutation_R1_R2_DD_RFF}b, where the simulated spectra are obtained by choosing either the `single-channel' operator $\hat{\rho}_{\text{det}}=\gamma_{I}\hat{I}_{x} + \gamma_{S}\hat{S}_{x}$ or `quadrature' operator $\hat{\rho}_{\text{det}}=\gamma_{I}\hat{I}_{+} + \gamma_{S}\hat{S}_{+}$, for detection. The maximum number of observable peaks is four, provided that the spectral separation between peaks is larger than the characteristic linewidths. We label these peaks as `near-zero-frequency' (nZF) peaks ($\nu_{13}$ and $\nu_{34}$) and $J$ peaks ($\nu_{12}$ and $\nu_{24}$). At zero field, only the zero-frequency (overlapping $\nu_{13}$ and $\nu_{34}$ peaks) and $\pm J$ peaks are observed; by increasing the main field, the splitting between energy levels increases (Fig.\,\ref{nutation_R1_R2_DD_RFF}c) such that peaks shift to higher frequencies, ultimately leading to two doublets centered around $^1$H and $^{13}$C Larmor frequencies at high field. 

In the limiting case of zero field ($\theta \rightarrow \pi/4$), we compute analytical expressions for the theoretical nutation curves of the corresponding peaks $A_{\nu_{ij}}(t_p)$ as a function of pulse duration $t_{p}$ (assuming $t_{p}\ll 1/J_{IS}$). The expressions for the nutation curves shown in Fig. \ref{nutation_R1_R2_DD_RFF}d are given by 

\begin{widetext}
\begin{align}
    A_{\nu_{13}}(t_p)  &\propto  p_{\psi_1}(t_s)(\sin{( B_{y}\gamma_{\text{C}}t_{p})} + \sin{( B_{y}\gamma_{\text{H}}t_{p}})) + p_{\psi_3}(t_s) \sin{( B_{y}(\gamma_{\text{H}} + \gamma_{\text{C}})t_{p})}\label{eq:nut13} \\
    A_{\nu_{34}}(t_p) &\propto p_{\psi_4}(t_s)(\sin{( B_{y}\gamma_{\text{C}}t_{p})} + \sin{( B_{y}\gamma_{\text{H}}t_{p}})) - p_{\psi_3}(t_s)\sin{( B_{y}(\gamma_{\text{H}} + \gamma_{\text{C}})t_{p})} \label{eq:nut34} \\
    A_{\nu_{12}}(t_p) &\propto p_{\psi_1}(t_s)(\sin{( B_{y}\gamma_{\text{C}}t_{p})} - \sin{( B_{y}\gamma_{\text{H}}t_{p}})) + p_{\psi_2}(t_s)\sin{( B_{y}(\gamma_{\text{H}} - \gamma_{\text{C}})t_{p})}\label{eq:nut12} \\
    A_{\nu_{24}}(t_p) &\propto p_{\psi_4}(t_s)(\sin{(B_{y}\gamma_{C}t_{p})} - \sin{( B_{y}\gamma_{H}t_{p}})) - p_{\psi_2}(t_s)\sin{( B_{y}(\gamma_{\text{H}} - \gamma_{\text{C}})t_{p})} \,.
    \label{eq:nut24}
\end{align}
\end{widetext}
where $p_{\psi_i}(t_s)$ represents the time-dependent population of eigenstate $\ket{\psi_i}$ after decay during a storage time $t_s$.

As the eigenstates of the Hamiltonian characteristic to both storage and measurement time are field-dependent, their initial populations subsequent to adiabatic shuttling $\{p_{\psi 1}(0),p_{\psi 2}(0),p_{\psi 3}(0),p_{\psi 4}(0) \}=\{\delta, \epsilon, -\epsilon, -\delta \}$ will decay at different rates and contribute to the observed transitions in a complex manner. The general equations describing population evolution in zero-to-ultralow fields after adiabatic shuttling for a storage time $t_{s}$, due solely to $^{13}$C-$^1$H dipole-dipole (DD) interaction, are
\begin{flalign}
&p_{\psi_{1}}(t_s)=\delta \cdot e^{-R_{1}^{\text{fast}}t_s}+\frac{\epsilon}{3}\left(e^{-R_{1}^{\text{inter}}(\theta)t_s}-e^{-R_{1}^{\text{slow}}(\theta)t_s}\right)
\label{eq:p1_decay} \\
&p_{\psi_{2}}(t_s)=\epsilon \cdot e^{-R_{1}^{\text{slow}}(\theta)t_s}
\label{eq:p2_decay} \\
&p_{\psi_{3}}(t_s)=-\frac{\epsilon}{3}\left(2e^{-R_{1}^{\text{inter}}(\theta)t_s}+e^{-R_{1}^{\text{slow}}(\theta)t_s}\right)
\label{eq:p3_decay} \\
&p_{\psi_{4}}(t_s)=-\delta \cdot e^{-R_{1}^{\text{fast}}t_s}+\frac{\epsilon}{3}\left (e^{-R_{1}^{\text{inter}}(\theta)t_s}-e^{-R_{1}^{\text{slow}}(\theta)t_s}\right )
\label{eq:p4_decay}
\end{flalign}
with $R_{1}^{\text{fast}}=\frac{15b_{IS}^{2}\tau_{c}^{\text{DD}}}{10}$, $R_{1}^{\text{inter}}(\theta)=\frac{b_{IS}^{2}\tau_{c}^{\text{DD}}}{20}(11-2\sin2\theta)(1+\sin2\theta)$ and $R_{1}^{\text{slow}}(\theta)=\frac{b_{IS}^{2}\tau_{c}^{\text{DD}}}{10}(5+2\sin2\theta)(1-\sin2\theta)$, where $b_{IS}=-\frac{\mu_{0}}{4\pi}\frac{\gamma_{I}\gamma_{S}\hbar}{r_{IS}^{3}}$ is the dipole-dipole coupling constant between the two spins at a distance $r_{IS}$, with  $\tau_{c}^{\text{DD}}$ being the rotational correlation time characterizing molecular tumbling (See Appendix~\ref{appex:relax}). These features represent a new source of relaxation dispersion compared to standard FC experiments, where changes in decay rates originate normally from the field dependence of the spectral function $J(\omega)$, given that polarization and detection are done at the same high field.

In the limiting case of zero field ($\theta \rightarrow \pi/4$), the $\psi_{2}$ state becomes the dipolar-immune singlet state $\ket{S_{0}}$. Meanwhile, the triplet populations equilibrate quickly toward thermal equilibrium (essentially equalized populations at zero field). 
Thus, assuming only dipolar interaction at zero field, the system should reach a hyperpolarized stationary state (immune to both coherent and incoherent evolution) given by the population imbalance between the singlet state and the average of the triplet manifold, known as a long-lived state\cite{pileio_long-lived_2020,emondts_long-lived_2014}. 
In practice, additional interactions such as intermolecular dipolar couplings will lead to equilibration of the singlet-triplet imbalance. These can be modeled by the Random Field Fluctuation (RFF) mechanism which also leads to multi-exponential field-dependent population dynamics during storage time similar the to dipole interaction (see Appendix \ref{appex:relax}). 

Additional information regarding relaxation mechanisms can be obtained by measuring peak linewidths, which are proportional to the auto-relaxation rates of individual coherences. At ultralow field, dipolar interactions induce differential line broadening between the inner ($\ket{\psi_{3}}\bra{\psi_{1}}$ and  $\ket{\psi_{3}}\bra{\psi_{4}}$) and outer ($\ket{\psi_{2}}\bra{\psi_{1}}$ and $\ket{\psi_{2}}\bra{\psi_{4}}$) transitions. The auto-relaxation rates of coherences connecting $\ket{\psi_{i}}\leftrightarrow\ket{\psi_j}$ eigenstates, denoted  $R_{2}^{\nu_{ij}}$, are 
\begin{flalign}
& R_{2}^{\nu_{13},\nu_{34}}(\theta)=\frac{b_{IS}^{2}\tau_{c}^{\text{DD}}}{20}(17+7\sin2\theta)\\
& R_{2}^{\nu_{12},\nu_{24}}(\theta)=\frac{b_{IS}^{2}\tau_{c}^{\text{DD}}}{20}(17-7\sin2\theta) \,.
\end{flalign}

In Fig.~\ref{nutation_R1_R2_DD_RFF}e--h, we highlight the different impact and field-scaling profile of the two relaxation mechanisms, DD and RFF, implying that no universal constant rate can be added to the rates based on the dipolar interaction in order to fit experimental data. In this work, we use a general field-dependent model to fit simultaneously the set of observable decays (see Section III in Supporting Information) with fitting variables describing molecular tumbling ($\tau_{c}^{\text{DD}},\tau_{c}^{\text{RFF}}$), internuclear distances ($r_{\text{CH}}$), local field fluctuations ($\omega_{\text{rms}}^{\text{C}},\omega_{\text{rms}}^{\text{H}}$), and their correlations ($\kappa_{\text{CH}}$).

As all detectable peaks originate from linear combinations of time-dependent populations (Eqs.\,\ref{eq:p1_decay}--\ref{eq:p4_decay}), the observed signals will display a multi-exponential decay. At the same time, the apparent peak decays can be modulated by the nutation angle induced by the detection pulse, unveiling different evolution processes during storage time. From Eqs.\,\ref{eq:nut13}--\ref{eq:nut24}, we see that a suitable choice of pulse duration $t_p$ can lead to selective observation of just one type of population, each with a different decay rate.
%In summary, relaxation phenomena in ZULF NMR depend on a series of factors: (i) the shuttling profile dictates the density operator upon reaching the measurement field and hence the populations of states (Eq.\,\ref{eq:rho_adiabatic}); (ii) the measurement field $B_0$ influences both auto- and cross-relaxation phenomena during storage time, leading to multi-exponential evolution (Eqs.\,\ref{eq:p1_decay}--\ref{eq:p4_decay}); and (iii) different nutation angles induced by the detection pulse can unveil different processes during the storage time (Eqs.\,\ref{eq:nut13}--\ref{eq:nut24}). 
The model described above can be easily extended to larger spin systems like the $^{13}$CH$_3$ spin system in [$^{13}$C]-methanol (see Section V of Supporting Information).  

\begin{figure*}[htb]
\includegraphics[trim=0pt 30pt 0pt 50pt, clip, width=\textwidth]{Figures/Figure3_big.pdf}
\caption{a) Schematic representation of the experimental setup (lateral view in the laboratory frame): the sample is shuttled from a prepolarizing 1\,T Halbach magnet array through a distance of around 36\,cm in a piercing solenoid into the magnetically shielded region for storage, pulse and detection. b) Top view of gradiometric quadrature detection scheme, consisting of two optical atomic magnetometers (dual-axis vector detectors) with four acquisition channels. c) Pulse sequences used for nutation (indicated by pulse time $t_p$) and polarization-lifetime (indicated by storage time $t_s$) experiments in nT--$\upmu$T fields. Spin polarization in the Halbach array lasted 10\,s. d) Nutation curves of the three observed peaks for the [$^{13}$C] formic-acid sample, using a detection pulse with a constant background field amplitude of $B_y = -30\,\upmu$T for storage time $t_{s} = 1 s$. Equations \ref{eq:nut13}--\ref{eq:nut24} were used to fit individual nutation profiles. e) Signal decay for $+J$ and $-J$ peaks as a function of storage time $t_s$ using two different pulse lengths $t_p$ prior to detection. Decay profiles were fitted with a mono-exponential function $S(t_s)=A\exp(-R_1\,t_s)$, with characteristic rates shown as inset. All experiments were performed at a measurement field of $B_0=0.1\,\upmu$T. f) Nutation curves of the four observed peaks nZF1 ($\nu_{34}$), nZF2 ($\nu_{13}$), $+J$ ($\nu_{24}$), and $-J$ ($\nu_{12}$) and g) Signal decay for $+J$ and $-J$ peaks as a function of storage time $t_s$ for similar [$^{13}$C] formic-acid experiments performed at a measurement field of $B_0=5\,\upmu$T.
%\AMF{Something is wrong here, need to refer to f) and g) and avoid repetition.}
} 
\label{pulse_seq}
\end{figure*}

We now focus on the experimental results obtained for a [$^{13}$C-$^{1}$H] spin system characteristic to [$^{13}$C] formic-acid solutions. The experiments were performed with a custom setup as described in \cite{fabricant_proton_2024} and schematically represented in Fig.\,\ref{pulse_seq}a. Samples are first polarized in a 1\,T Halbach magnet and shuttled through a piercing solenoid into a shielded region for storage, pulse, and detection. We developed a gradiometric quadrature detection method which utilizes two orthogonal dual-axis magnetometers (Fig.\,\ref{pulse_seq}b), with a total of four acquisition channels from which a complex signal is constructed\cite{fabricant_proton_2024}.
%\AMF{here there is redundancy}. 
Figure \ref{pulse_seq}c shows the pulse sequence used for all  experiments, where the pulse duration ($t_p$) for measuring nutation curves and the storage time ($t_s$) for extracting polarization lifetimes were varied independently. Each experiment represents the average of signals from a total of 64 automated repetitions, including shuttling from high field to low field. 



As mentioned above, certain choices of nutation angles can help isolate the contributions of different populations to the observed signal. 
Figure \ref{pulse_seq}d shows the nutation curves of the three observable peaks of $^{13}$C-labeled formic acid at $B_0=0.1\,\upmu$T. Due to the small ($<1$\,mHz) frequency difference and finite linewidths, the $\nu_{13}$ and $\nu_{34}$ transitions appear as a single peak which is labeled `nZF'. The other two transitions ($\nu_{24}$ and $\nu_{12}$) are labeled `$+J$' and `$-J$', respectively. The plot shows the signal amplitude of individual peaks as a function of pulse duration, with a constant pulse field amplitude ($B_y=-30\upmu$T) applied after shuttling ($t_s=1$\,s). Equations \ref{eq:nut13}--\ref{eq:nut24} were used to fit the experimental points, showing good agreement (average $R^2>0.95$). The different nutation profiles for $+J$ and $-J$ peaks arise from the $\hat{Z}_x$ component of the density operator $\hat{\rho}_{\text{ad}}$ (Eq.\,\ref{eq:rho_adiabatic}) after shuttling (see Section II in Supporting Information). For a sudden transfer to ultralow fields, the density operator becomes $\hat{\rho}_{\text{sud}}=\frac{\delta-\epsilon}{2}\hat{I}_{z}+\frac{\delta+\epsilon}{2}\hat{S}_{z}$, and the $+J$ and $-J$ nutation profiles would therefore be different from the adiabatic case considered here.

\renewcommand{\arraystretch}{1.3} % slightly less than 1.5 for tighter fit

\begin{table}[htb]
\centering
\caption{Best-fit relaxation parameters extracted from a global fit of experimental data at $0.1\ \upmu\text{T}$ and $5\ \upmu\text{T}$ fields using the theoretical model described above (see Section III in Supporting Information for computational details).
}
\label{tab:fit}
\resizebox{\linewidth}{!}{%
\begin{tabular}{c|cc|cccc}
\multirow{2}{*}{Field ($\upmu$T)} &
\multicolumn{2}{c|}{Dipolar} &
\multicolumn{4}{c}{Random Field Fluctuations} \\
\cline{2-7}
& $d_{\text{CH}}$\,(\AA) & $\tau_{C}^{\text{DD}}$\,(ps) & $\omega_{\text{rms}}^{\text{H}}$\,(rad) & $\omega_{\text{rms}}^{\text{C}}$\,(rad) & $\kappa_{\text{CH}}$ & $\tau_{c}^{\text{RFF}}$\,(ps) \\
\hline
% 0.1  & 1.1\footnotemark[1] & 5.94 & $10.05\cdot10^{3}$ & $13.06\cdot10^{3}$ & 0.06 & 9.97 \\
% 5.0  & 1.1\footnotemark[1] & 3.35 & $10.41\cdot10^{3}$ & $13.55\cdot10^{3}$ & 0.28 & 20.32 \\
0.1\,$\upmu$T \& 5\,$\upmu$T  & 1.1\footnotemark[1] & 3.24 & $49.63\cdot10^{3}$  & $85.32\cdot10^{3}$ & 0.49 & 19.34 \\
\hline
\end{tabular}%
}
\footnotetext[1]{Fixed value taken from literature.}
\end{table}

Thus, we have further evidence (in addition to the simulated spin dynamics during field shuttling from Fig. S1) based on the nutation data that our shuttling was performed adiabatically, which is an essential aspect for properly analyzing polarization lifetimes in ultralow fields (See Sections II and VI in Supporting Information where we consider the case of sudden transfer in more detail).


We next performed polarization-lifetime measurements by increasing the storage time before the pulse ($t_s \in [0;10\,\text{s}]$) and tracked individual peak decays using different pulse durations ($t_p \in [0; 1 \,\text{ms}]$). Firstly, we observed that the measured decay rates depended strongly on the nutation angle used for observation (Fig.\,\ref{pulse_seq}e). This phenomenon stems from the modulation of each individual population's contribution to the observed signal: for $t_p=300\,\upmu$s both $+J$ and $-J$ peaks originate from the $\psi_{2}$ population (Eqs.\,\ref{eq:nut12}--\ref{eq:nut24}) which has a slow decay (Eq.\,\ref{eq:p2_decay}), while at $t_p=500\,\upmu$s the two peaks come from the populations of $\psi_4$ and $\psi_1$, respectively, both having enhanced relaxation (Eqs.\,\ref{eq:p1_decay} and \ref{eq:p4_decay}). Thus, the apparent decay rate can be experimentally tuned by adjusting the pulse length (or amplitude), targeting desired features such as extending storage time before detection. This approach may be relevant for extended pulse sequences as well as experiments featuring hyperpolarized samples, which require chemical or physical manipulation prior to signal acquisition\cite{van_dyke_relayed_2022}. 


Furthermore, we conducted both nutation and polarization lifetime experiments at $B_0 = 5\,\upmu$T (Fig.\,\ref{nutation_R1_R2_DD_RFF}f--g) and observed a similar trend in decay rates at two different pulse lengths. The extracted rates characteristic to the $\psi_2$ population decay (for $t_p = 350\,\upmu$s) are significantly higher than the corresponding ones at $B_0=0.1\,\upmu$T, as predicted by theory (Fig.\,\ref{nutation_R1_R2_DD_RFF}e). 

In order to derive relevant molecular parameters such as dipolar couplings, local field fluctuations, or tumbling rates, we performed a global fit on all datasets combining nutation and decay profiles at both $0.1\,\upmu$T and $5\,\upmu$T fields (see Section III of Supporting Information). The extracted parameters are listed in Table\,\ref{tab:fit}. Using this set of parameters, we predict, based on our model, the relaxation rates of the coherences generating $+J$ and $-J$ peaks and compared them with experimental values derived from linewidth fitting at $B_0=0.1\,\upmu$T. The expected values of the relaxation rate for the two transitions (due to the additive contributions of DD and RFF mechanisms) are $R_{2,\pm J}^{\text{model}}\approx0.27 $\,s$^{-1}$, smaller than the experimentally measured ones of $R_{2,+J}^{*,\text{exp}}=0.405 \pm0.003 $\,s$^{-1}$ and $R_{2,-J}^{*,\text{exp}}=0.392 \pm 0.005 $\,s$^{-1}$. This mismatch can be ascribed to  field inhomogeneities detected in our experimental setup, in particular, gradients produced by the solenoid as well as by the internal compensation coils of the zero-field magnetometers used for detection (see Section IV of Supporting Information).

In summary, we highlight in this work the counterintuitive relaxation effects that occur at ultralow field, where measurements can only be performed with quantum magnetometers, not with classical induction coils. The presented theoretical model provides a framework by which the observed effects can be described in terms of  molecular parameters. Importantly, we have highlighted a rather unexpected feature of relaxation rate measurements at ultralow field: the nutation angle induced by the detection pulse modulates the apparent relaxation rates measured through the individual peaks, thereby providing additional contrast, spectral editing, and spectral assignment mechanisms.
%such that polarization lifetimes and not only frequency separation can function as a contrast mechanism.

In the ULF regime, spectra reach their maximum complexity in terms of the total number of individual peaks\cite{appelt_paths_2010}, each decaying differently. This effect makes the ULF regime richer in information than the high-field regime, as the same relaxation mechanism can be tracked with multiple different observables. Potentially, this process could allow for a better quantification  of each relaxation mechanism's contribution. Close to zero field, differences in the precession frequencies in the laboratory frame become of the same order as the relaxation rates, such that the `secular approximation' is not applicable anymore. Furthermore, relaxation processes can also be incorporated into magnetic resonance imaging experiments, thereby producing innovating image contrast.
%\AMF{Please add back one short sentence about imaging.}
%Using the analyses presented here, one can also use the relaxation properties as an authentication or assignment tool in crowded spectra.  Such experiments need not be performed under ZULF conditions, just the preparation can go through that regime for a specified amount of time to produce desired relaxation effects that are reflected in image contrast. 
Finally, experiments performed under ZULF conditions are compatible with examining samples enclosed in (thick) conductive casings, such as, for example, batteries\cite{fabricant_enabling_2025}. Monitoring relaxation rates in batteries could be used to report on the dynamic processes in functioning devices. 


\textit{Acknowledgment}--A.J. acknowledges funding from the US National Science Foundation, award no. CHE 2505792, and an award from the ACS PRF 68117-ND6. F.T. acknowledges funding from National Medical Project MySMIS 326475 and Project ELI-RO/RDI/2024/14 SPARC funded by the Institute of Atomic Physics (Romania). This work was supported in part by the Deutsche Forschungsgemeinschaft (DFG,German Research Foundation) in the framework to the collaborative research center ``Defects and Defect Engineering in Soft Matter''
(SFB1552) under Project No. 465145163 and under the DFG/ANR grant BU 3035/24-1.  

\appendix 

\section{Eigenstates at ultralow fields}
\label{appenx:eigenstates}
The master equation describing both coherent and incoherent evolution of the density operator $\hat{\rho}(t)$ in Bloch-Redfield-Wangsness  theory has the following form:

\begin{equation}
    \frac{\partial\hat{\rho}(t)}{\partial t} = -i\left[\hat{H}_{0},\hat{\rho}(t)-\hat{\rho}_{\text{eq}}\right] + \hat{\hat{\Gamma}}\left(\hat{\rho}(t)-\hat{\rho}_{\text{eq}}\right) \,.
    \label{eq:Lioville_Equation}
\end{equation}
Here, the static Hamiltonian $\hat{H}_{0}$ determines the coherent evolution experienced by all spins in the sample, which for molecules in isotropic liquids amounts to the Zeeman and scalar-coupling interactions. The second term represents the relaxation superoperator $\hat{\hat{\Gamma}}$ driving the system towards the thermal equilibrium density operator $\hat{\rho}_{\text{eq}}$.

For a heteronuclear two-spin-1/2 \textit{\{I, S\}} system characterized by $\gamma_{I}$ and $\gamma_{S}$ gyromagnetic ratios (in $\text{rad} \cdot \text{s}^{-1} \cdot $T$^{-1}$) and a scalar coupling constant $J_{IS}$ (in Hz), the Hamiltonian describing the interaction of this system with a magnetic field $B_0$ takes the form
\begin{equation}
    \hat{H}_{0} = -\left(\gamma_{I}\hat{I}_{z}+\gamma_{S}\hat{S}_{z}\right)B_{0} + 2\pi J_{IS} \boldsymbol{\hat{I}}\cdot\boldsymbol{\hat{S}} \,,
    \label{eq:Total_Hamiltonian}
\end{equation}
with the eigenstates 
\begin{flalign}
& \psi_{1}=\ket{\alpha\alpha}\\
& \psi_{2} = -\sin{\theta}\ket{\alpha\beta} + \cos{\theta}\ket{\beta\alpha}\\
& \psi_{3} = \cos{\theta}\ket{\alpha\beta} + \sin{\theta}\ket{\beta\alpha}\\
& \psi_{4}=\ket{\beta\beta}
\label{eq:eigenstates}
\end{flalign}
and their eigenvalues
\begin{flalign}
&E_{1} = \frac{1}{2}\left((\gamma_{I}+\gamma_{S})B_{0}+ \pi J_{IS}\right)\\
&E_{2}  = -\frac{1}{2}\left(\sqrt{((\gamma_{I}-\gamma_{S})B_{0})^{2}+(2\pi J_{IS})^{2}} +\pi J_{IS}\right)\\
& E_{3}  = \frac{1}{2}\left(\sqrt{((\gamma_{I}-\gamma_{S})B_{0})^{2}+(2\pi J_{IS})^{2}} -\pi J_{IS}\right)\\
& E_{4} = \frac{1}{2}\left (-(\gamma_{I}+\gamma_{S})B_{0} + \pi J_{IS}\right ) \,.
\label{energy_levels}
\end{flalign}

Two limiting cases are relevant for our discussion: in the weakly coupled regime ($\theta \rightarrow 0$), the system is described by the Zeeman basis $\boldsymbol{Q_{Z}}=\{\ket{\alpha\alpha},\ket{\alpha\beta},\ket{\beta\alpha},\ket{\beta\beta}\}$, while in the strongly coupled regime ($\theta \rightarrow \frac{\pi}{4}$), the eigenbasis is the singlet-triplet basis $\boldsymbol{Q_{ST}} = \{\ket{\alpha\alpha},\frac{1}{\sqrt{2}}(\ket{\alpha\beta}-\ket{\beta\alpha}),\frac{1}{\sqrt{2}}(\ket{\alpha\beta}+\ket{\beta\alpha}),\ket{\beta\beta}\}$. 

Because the singlet state $\ket{S_{0}}=\frac{1}{\sqrt{2}}\left(\ket{\alpha\beta}-\ket{\beta\alpha}\right)$ is antisymmetric with respect to the permutation operator ($\hat{P}_{\alpha\leftrightarrow\beta}\ket{S_{0}}=-\ket{S_{0}}$), while the other triplet states are symmetric ($\hat{P}_{\alpha\leftrightarrow\beta}\ket{T_{\mu}}=\ket{T_{\mu}}$ with $\mu \in \{-1,0,1\}$), no symmetric interaction like the dipolar coupling between \textit{I} and \textit{S} can mix these states\cite{pileio_long-lived_2020}. Moreover, the singlet-state population is known to be immune to the intra-pair dipolar interaction and therefore has an extended lifetime\cite{carravetta_beyond_2004}. However, the Zeeman states follow different transformations under permutation, and no population is immune to the dipolar coupling. These previously known results suggest that the relaxation rates of spin-state populations and coherences are directly impacted by the structure of the eigenstates and, thus, by the field at which storage and detection are performed. 

\section{Relaxation Dynamics during Storage Time}
\label{appex:relax}
The non-equilibrium polarization following adiabatic shuttling commutes with the main Hamiltonian $\hat{H}_{0}$, meaning that detection pulses are needed to detect spin-state transitions. At ZULF, where measurements are performed, the equilibrium polarization is negligible. Thus, the (hyper)polarized nuclear spin order will gradually decay until a new equilibrium (near-zero-polarization) state is achieved. The irreversible polarization loss is caused by various relaxation mechanisms via random molecular tumbling that induces spin transitions.

In liquid-state NMR, the dominant interactions that lead to loss of spin order are intramolecular dipole-dipole (DD) interactions, chemical shift anisotropy (CSA), and electric-field-gradient (EFG) interactions with quadrupolar nuclei\cite{kowalewski_nuclear_2019}.
For our heteronuclear two-spin-1/2 system at ultralow field, the CSA relaxation contribution is negligible as it scales as $B_{0}^{2}$\cite{kowalewski_nuclear_2019}. At the same time, EFGs only impact nuclei with spin greater or equal to 1. Some other mechanisms (e.g. intermolecular dipolar couplings\cite{hwang_dynamic_1975} or spin-rotation interaction\cite{singer_nmr_2018}) or unknown relaxation sources can be further modeled by random-field fluctuations (RFF). In the fast-tumbling approximation, the relaxation superoperator describing the intramolecular dipolar interaction is\cite{levitt_spin_2013}

\begin{equation}
    \hat{\hat{\Gamma}}_{\text{DD}}=-\frac{6}{5}b_{IS}^{2}\tau_{c}^{\text{DD}}\sum_{m=-2}^{2}(-1)^{m}\left [\hat{T}_{2m},[\hat{T}_{2-m},\cdot\,]\right ]\,,
    \label{eq:RDD_relax}
\end{equation}
where $\hat{T}_{rm}$ is the spherical tensor operator of rank \textit{r} and order \textit{m}. The fast-tumbling approximation is adequate in ultralow fields, as the spectral density $J(\omega)\propto \frac{\tau_{c}^{\text{DD}}}{1+(\omega\tau_{c}^{\text{DD}})^{2}}$ will always be field-independent ($\omega\tau_{c}^{\text{DD}}\ll1$). Field-dependent rates describing populations' auto- and cross-relaxation are computed from Liouville products\cite{cavanagh_protein_1995} of the form 

\begin{equation}
    \Gamma_{ij}=\frac{\braket{\ket{\psi_i}\bra{\psi_i}|\hat{\hat{\Gamma}} |\ket{\psi_j}\bra{\psi_j}}}{\braket{\ket{\psi_i}\bra{\psi_i}|\ket{\psi_i}\bra{\psi_i}}}\,, i,j \in \{1, 2, 3, 4\} \,.
    \label{eq:Liouville_product_pop}
\end{equation}

Using these $\Gamma_{ij}$ elements, we build the matrix representation of the system of coupled differential equations describing both population decay and mixing in our system during the storage time:

\begin{widetext}
\begin{align}
 \frac{d}{dt}\begin{pmatrix}
p_{\psi1}(t)  \\ p_{\psi2}(t)  \\p_{\psi3}(t)  \\p_{\psi4}(t)  
\end{pmatrix}   =
\frac{b_{IS}^{2}\tau_{c}^{\text{DD}}}{20}
\begin{pmatrix}
-18 & 3(1-\sin2\theta) &3(1+\sin2\theta) &12\\
3(1-\sin2\theta) &2(4+\sin2\theta)(1-\sin2\theta) & 2(1-\sin2\theta)(1+\sin2\theta)& 3(1-\sin2\theta)\\
3(1+\sin2\theta) & 2(1+\sin2\theta)(1-\sin2\theta) &2(-4+\sin2\theta)(1+\sin2\theta)  &3(1+\sin2\theta)\\
12 & 3(1-\sin2\theta) &3(1+\sin2\theta) &-18
\end{pmatrix}
\begin{pmatrix}
p_{\psi1}(t) \\ p_{\psi2}(t)  \\p_{\psi3}(t) \\p_{\psi4}(t)
\label{eq:matrix_eq}
\end{pmatrix}\,,
\end{align}
\end{widetext}
whose solutions are given by Eqs.~\ref{eq:p1_decay}--\ref{eq:p4_decay}.

The RFF mechanism can be invoked to describe the effects of other mechanisms that are often difficult to quantify, such as the intermolecular dipolar coupling or the spin-rotation interaction\cite{whipham_cross-correlated_2024}. Similarly to Eq.\,\ref{eq:RDD_relax}, we can build the corresponding relaxation superoperator for RFF as 
\begin{equation}
\hat{\hat{\Gamma}}_{\text{RFF}}=\sum_{i,j=1}^{2}\kappa_{ij}\omega_{\text{rms}}^{i}\omega_{\text{rms}}^{j}\tau_{c}^{\text{RFF}}\sum_{m=-1}^{1}(-1)^{m}[\hat{T}_{1m},[\hat{T}_{1-m},\cdot\,]]
\end{equation}
with $\tau_{c}^{\text{RFF}}$ being the correlation time of the RFF and $\omega_{\text{rms}}^{i,j}$ being the root-mean-square amplitudes of the local field fluctuations associated with spins \textit{i} and \textit{j}; $-1\leq\kappa_{ij}\leq1$ is the coefficient describing the degree of correlation of fluctuations experienced by the two spins. By definition, $\kappa_{ii}=\kappa_{jj}=1$, while $\kappa_{ij}=\pm1$ represents a perfectly correlated ($+1$) or anti-correlated ($-1$) interaction between the two spins. The system of coupled differential equations describing population dynamics during storage time due to RFF is similar to the one in Eq. \ref{eq:matrix_eq} and has an identical solution. A Mathematica notebook summarizing step-by-step derivations of spin dynamics and analytical expressions of relaxation rates is attached as Supporting Information. 

\bibliography{references}% Produces the bibliography via BibTeX.

\end{document}
%