Nuclear Diffusion: Concepts and Mechanisms

• Nuclear Diffusion is a multifaceted term describing processes from spin transport via flip-flop interactions to barrier-mediated diffusion in nuclear structure and cell nuclei.
• Studies reveal that nuclear-spin diffusion is polarization dependent and influences dynamic nuclear polarization in quantum dots and semiconductor nanostructures.
• Diffusion NMR techniques and nuclear diffusion barrier analyses provide practical insights for imaging pore structures and understanding protein compartmentalization in cells.

Nuclear diffusion is a polysemous term rather than a single transport law. In contemporary literature it can denote transport of longitudinal nuclear magnetization through flip-flop exchange interactions between neighboring spins without actual mass transport; diffusion-controlled crossing of an internal barrier in collective nuclear shape coordinates during superheavy-nucleus formation; diffusion barriers that compartmentalize proteins within the nucleus or nuclear envelope; diffusion measurements performed by nuclear magnetic resonance; and transport processes in which nuclear quantum effects of light nuclei reshape free-energy barriers and self-diffusion (Wang et al., 2021, Cap et al., 2021, Zavala et al., 2014, Kuder et al., 2012, Eren et al., 24 Jun 2026). A closely related but distinct concept is nuclear surface diffuseness, where “diffuseness” denotes the smooth radial falloff of nuclear density rather than a transport coefficient (Hatakeyama et al., 2018).

1. Scope and terminology

The breadth of usage is best understood by separating the major research contexts in which the term appears.

Usage	Meaning in the source literature	Representative papers
Nuclear-spin diffusion	Transport of longitudinal nuclear magnetization through flip-flop exchange interactions	(Wang et al., 2021, Chessari et al., 2022, Millington-Hotze et al., 2022)
Diffusion NMR	Phase-encoded diffusion measurement for direct imaging of closed pores	(Kuder et al., 2012)
Nuclear diffusion barriers	Restriction of protein lateral exchange in the nucleus or nuclear envelope	(Zavala et al., 2014, Alas et al., 2024, Merino-Aceituno et al., 10 Oct 2025)
Fusion-by-diffusion	Diffusion over an internal barrier in collective nuclear shape space	(Cap et al., 2021)
Nuclear quantum effects in diffusion	Zero-point motion and tunneling of light nuclei in barrier crossing	(Yang et al., 2019, Eren et al., 24 Jun 2026)
Algorithmic naming	“Nuclear Diffusion” as nuclear-norm plus diffusion posterior sampling	(Stevens et al., 25 Sep 2025, Yang et al., 9 Jul 2025)

A recurrent source of confusion is the proximity between diffusion and diffuseness. In high-energy nucleon–nucleus scattering, the two-parameter Fermi density uses a diffuseness parameter $a$ to describe the transition region of the nuclear surface, and the first diffraction peak height is used to extract that structural quantity; this is not a transport problem (Hatakeyama et al., 2018).

2. Nuclear-spin diffusion as transport of magnetization

In rigid solids, nuclear-spin diffusion is the transport of longitudinal nuclear magnetization through flip-flop exchange interactions between neighboring spins, without actual mass transport. In the high-temperature Lowe-Gade theory, the spin-diffusion tensor component is written

$$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$

For short times the correlation function behaves as

$$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$

Extending Lowe-Gade theory to low temperature and high polarization changes the decay width to $(1-\overline{p}^2)\Delta_i$, so that

$$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$

and therefore

$D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$

This makes the diffusion coefficient polarization dependent rather than a strict constant of the lattice (Wang et al., 2021).

The finite-polarization correction has a clear limiting structure. When $\overline{p}\ll 1$, the factor $1/\sqrt{1-\overline{p}^2}$ approaches unity and the conventional high-temperature theory is recovered smoothly. When $\overline{p}\to 1$, the formula diverges, but the authors note that this is not a physical problem, because if all spins are perfectly polarized, the polarization is already uniform and diffusion is no longer needed. The agreement with linewidth scaling based on the Van Vleck second moment is described as suggestive, but not as a justification for simply inserting a polarized second moment into the high-temperature Redfield formula, since that derivation had already discarded polarization-dependent terms. The proposed scaling implies that transport of nuclear magnetization accelerates in hyperpolarized solids and can alter the shape and timescale of dynamic nuclear polarization buildup (Wang et al., 2021).

In driven quantum dots, spin diffusion often appears explicitly as the loss channel that competes with optical pumping. For a single charged InAs quantum dot, the reduced quasi-analytic equation

$$\frac{\partial\omega}{\partial t} = -\kappa \omega + \alpha\frac{\partial^2}{\partial \omega^2}[\omega C(\omega,t)]$$

separates a spin-diffusion loss term, $$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$0, from an optical nuclear pumping term. The balance between these two processes produces the hysteretic triangle-like free-induction-decay pattern, and the resulting feedback can dynamically tune the electron Larmor frequency to a value determined by the pulse timing (0912.5401).

3. Hyperfine-mediated transport, DNP, and semiconductor nanostructures

Hyperfine coupling, Knight-field inhomogeneity, and driven electron dynamics can enhance, suppress, or spatially redirect nuclear-spin diffusion. In the $$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$1 fractional quantum Hall domain phase, resistively detected NMR mediated by electrically driven domain-wall motion shows pumping-dependent signal saturation and a relatively homogeneous polarization profile extending away from domain-wall pinning centers. The interpretation is enhanced diffusion of nuclear spin polarization over micrometer-scale distances through hyperfine-mediated indirect nuclear coupling via the electronic domain-wall environment, modulated by a nonuniform Knight field (Miyamoto et al., 2016). In GaAs/Al$$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$2Ga$$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$3As double quantum wells, dynamic nuclear polarization at the $$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$4 spin phase transition resolves strong anisotropy: the in-plane diffusion coefficient is estimated as $$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$5, whereas perpendicular transport through the AlGaAs barrier gives $$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$6, about two orders of magnitude smaller; the detection quantum well continues to show an $$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$7 increase for approximately 2 h after the termination of DNP in the polarization quantum well (Hatano et al., 2014).

In electrostatically defined double quantum dots, nuclear spin diffusion does not merely drain polarization from the dots. Transport-driven polarization, nanomagnet field gradients, and dot-resolved EDSR reveal multiple fixed points and show that polarization diffuses into the surrounding material, producing a polarized reservoir that feeds back on the dots and lengthens decay times from about 3 minutes to almost 8 minutes. The environmental polarization is inferred from delayed EDSR peaks and from the fact that the effective field cancellation cannot be explained by a profile confined only to the dot interior (Forster et al., 2015). In neutral self-assembled quantum dots, a distinct mechanism appears: the effective heavy-hole non-collinear term

$$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$8

flips a nuclear spin without changing the hole pseudospin, thereby generating a hyperfine-mediated nuclear-spin diffusion mechanism and leaving the post-pumping nuclear bath in a complex mixed state rather than a simple highly polarized Overhauser state (Ribeiro et al., 2014).

Near paramagnetic dopants in DNP samples, electron polarization itself can be the control parameter for nuclear-spin transport. In HypRes experiments on 50 mM TEMPOL in H$$(D_0)_{\alpha,\beta} = \frac{\sqrt{\pi}}{2}\sum_{i\neq 0} A_i\, \alpha_i \beta_i\, F_i, \qquad F_i=\int_0^\infty \langle G_i(\tau)\rangle\, d\tau.$$9O:D$$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$0O:D$$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$1-glycerol $$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$2 at $$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$3 and $$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$4, proton spin diffusion is much faster when the electron polarization is reduced by microwave irradiation, and it becomes strongly suppressed—essentially vanishing in the extreme limit—when the electron polarization approaches unity. The corresponding two-nucleus–one-electron Lindblad model predicts that at the hidden/visible boundary the diffusion coefficient is about 7 times larger when the electron polarization is lowered from equilibrium to microwave-irradiated conditions (Chessari et al., 2022). By contrast, in a single GaAs/AlGaAs quantum dot, charge-controlled experiments show that a resident electron accelerates nuclear-spin diffusion without forming any Knight-field gradient barrier: $$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$5, $$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$6, and $$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$7, with diffusion-limited nuclear-spin lifetimes of about $$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$8 at low field and about $$G_i(\tau)\approx \exp(-\Delta_i \tau^2), \qquad \Delta_i=\frac12\sum_{j\neq 0,i}(B_j-B_{ij})^2.$$9 at high field (Millington-Hotze et al., 2022). This suggests that electron-controlled diffusion is regime-dependent rather than universally barrier-forming.

In crystalline nanoscale silicon particles, first-principles spin-diffusion calculations and finite-element simulations shift the emphasis from particle size to interfacial transport. For natural-abundance silicon, the nearest-neighbor estimate gives $(1-\overline{p}^2)\Delta_i$0, whereas the full lattice-sum estimate gives $(1-\overline{p}^2)\Delta_i$1. The zero-quantum linewidth governing flip-flop transport is found to be nearly identical to the experimentally accessible single-quantum linewidth, so measured SQ linewidths can serve as a proxy for the spin-diffusion coefficient. Core–shell simulations then indicate that buildup and decay are governed mainly by outer-shell relaxation near $(1-\overline{p}^2)\Delta_i$2 centers rather than bulk particle diameter; 20 nm and 50 nm particles show room-temperature $(1-\overline{p}^2)\Delta_i$3 min and $(1-\overline{p}^2)\Delta_i$4 min, respectively (Witte et al., 2024).

4. Diffusion NMR and pore imaging

In diffusion NMR, nuclear spins are probes of stochastic motion through the accumulated phase

$(1-\overline{p}^2)\Delta_i$5

For the conventional two-short-gradient experiment, the signal reduces to

$(1-\overline{p}^2)\Delta_i$6

so the phase of the pore space function is lost and arbitrary closed pores cannot be reconstructed uniquely. The long-narrow gradient scheme imposes

$(1-\overline{p}^2)\Delta_i$7

and in the long-time limit yields

$(1-\overline{p}^2)\Delta_i$8

which preserves the complex Fourier transform of the pore shape itself (Kuder et al., 2012).

The experimental demonstration used hyperpolarized $(1-\overline{p}^2)\Delta_i$9Xe generated by spin-exchange optical pumping, with a gas diffusion coefficient about $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$0, on a $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$1 clinical MRI scanner with maximum gradient strength $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$2. Parallel plates with distances of $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$3, $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$4, and $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$5, equilateral triangular cutouts with edge length about $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$6, mixed-geometry slit phantoms, and a phantom with 170 triangular pores established that signal from the whole sample can reconstruct the average shape of arbitrary closed pores rather than merely indirect microstructural indices. The method was proposed as a route to cell or pore shapes, cell density, and axon integrity in systems where conventional high-resolution imaging would suffer from severe signal-to-noise penalties (Kuder et al., 2012).

5. Nuclear diffusion barriers and transport at the nuclear envelope

In cell biology, nuclear diffusion barriers refer to restrictions on protein lateral exchange in or around the nucleus rather than to spin transport. During Saccharomyces cerevisiae closed mitosis, a spatial stochastic model constrained by FLIP data from the outer nuclear membrane, inner nuclear membrane, and nuclear pore complex predicts that in early anaphase a sphingolipid domain and a protein ring could constitute the barrier, whereas in late anaphase a sphingolipid domain spanning the bridge between lobes would be entirely sufficient (Zavala et al., 2014). The barrier is quantified through a transmission coefficient,

$$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$7

so permeability depends on partitioning into the ordered phase, diffusion inside the barrier, and barrier thickness rather than on viscosity alone.

At the inner nuclear membrane, diffusion can also organize membrane protein nanostructure. A reaction-diffusion model for emerin nanodomains distinguishes a slowly diffusing activating species $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$8 with diffusion coefficient $$G_i(\tau)\approx \exp\!\left[-(1-\overline{p}^2)\Delta_i \tau^2\right], \qquad F_i \propto \frac{1}{\sqrt{1-\overline{p}^2}},$$9 and a rapidly diffusing inhibitory species $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$0 with diffusion coefficient $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$1, governed by coupled equations with steric factor $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$2. For wild-type emerin, the inferred parameters $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$3 and $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$4 reproduce a nanodomain diameter $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$5 and area coverage $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$6. Under mechanical stress and in EDMD-associated mutants, changes in diffusion coefficients and higher-order assembly terms shift nanodomain size and occupancy in quantitatively predictable ways (Alas et al., 2024).

A separate envelope-transport problem concerns exchange between the endoplasmic reticulum and the nuclear envelope through narrow luminal junctions. A compartment model with junction density $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$7, effective area $D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$8, and transport rate

$D=\frac{D_0}{\sqrt{1-\overline{p}^{2}}}.$9

reduces the nuclear-envelope recovery to

$\overline{p}\ll 1$0

Using FRAP-derived diffusion coefficients $\overline{p}\ll 1$1 for moxGFP-KDEL and $\overline{p}\ll 1$2 for NusA-moxGFP×2-KDEL, together with ER–NE junctions that are typically $\overline{p}\ll 1$3 in diameter and occur at approximately $\overline{p}\ll 1$4 junctions/$\overline{p}\ll 1$5, the model accounts for small-reporter recovery within $\overline{p}\ll 1$6 and large-reporter recovery within $\overline{p}\ll 1$7. In this setting, simple passive diffusion is sufficient to explain rapid ER-to-NE transport (Merino-Aceituno et al., 10 Oct 2025).

6. Collective-shape diffusion, surface diffuseness, and extended usages

In heavy-ion reaction theory, diffusion appears in collective nuclear shape space rather than real space. The fusion-by-diffusion model factorizes the evaporation-residue cross section as

$\overline{p}\ll 1$8

where the fusion probability is obtained from the Smoluchowski diffusion equation with an inverted-parabola barrier. The barrier height is

$\overline{p}\ll 1$9

so rotational energies at the saddle and injection configurations directly modify fusion hindrance. For $1/\sqrt{1-\overline{p}^2}$0, $1/\sqrt{1-\overline{p}^2}$1, and $1/\sqrt{1-\overline{p}^2}$2 incident on a $1/\sqrt{1-\overline{p}^2}$3 target, the model reproduces the saturation of fusion probability above the interaction barrier through suppression of higher partial waves and the role of critical angular momentum (Cap et al., 2021).

This usage remains distinct from nuclear surface diffuseness in elastic scattering. In the two-parameter Fermi parametrization

$1/\sqrt{1-\overline{p}^2}$4

the first diffraction peak angle is almost independent of diffuseness and is determined mainly by the nuclear radius, whereas the differential cross section at that peak depends strongly on $1/\sqrt{1-\overline{p}^2}$5. Numerical experiments based on Skyrme-HF+BCS densities show that rms radii can be extracted to within about $1/\sqrt{1-\overline{p}^2}$6 and diffuseness to within about $1/\sqrt{1-\overline{p}^2}$7 from limited forward-angle data (Hatakeyama et al., 2018).

In atomistic transport, the adjective nuclear often refers to the nuclei of light atoms. On $1/\sqrt{1-\overline{p}^2}$8-PtO$1/\sqrt{1-\overline{p}^2}$9(001), hydrogen diffusion among neighboring O-top sites is analyzed with a WKB tunneling coefficient,

$\overline{p}\to 1$0

and nuclear quantum effects are significant for the surface diffusion of H at room temperature and play a dominant role in cryogenic conditions (Yang et al., 2019). In layered MoS$\overline{p}\to 1$1, machine-learning-enhanced path-integral simulations show that nuclear quantum effects substantially lower free-energy barriers for hydrogen diffusion at $\overline{p}\to 1$2, with a $\overline{p}\to 1$3 difference between hydrogen and deuterium quantum free-energy barriers; in the $\overline{p}\to 1$4 stacking, the quantum self-diffusion coefficient reaches $\overline{p}\to 1$5 (Eren et al., 24 Jun 2026). In nuclear fuels, data-driven kernel ridge regression models trained on experimental data and CENTIPEDE cluster-dynamics simulations predict diffusivities of defects and mobile species in $\overline{p}\to 1$6 and $\overline{p}\to 1$7 more accurately than current analytical models, with average percent error about $\overline{p}\to 1$8 for U and Xe diffusion in the $\overline{p}\to 1$9 space (Craven et al., 2023).

The phrase also appears as nomenclature in machine learning rather than as a physical transport mechanism. “Nuclear Diffusion” in video restoration denotes a hybrid framework that combines a nuclear norm prior on a low-rank background with diffusion posterior sampling for the foreground, and is reported to improve gCNR and KS statistic in cardiac ultrasound dehazing relative to traditional RPCA (Stevens et al., 25 Sep 2025). “DiffNMR” uses a conditional discrete diffusion model for de novo molecular structure elucidation from NMR spectra, with a diffusion autoencoder, contrastive alignment of spectral and molecular representations, retrieval initialization, and similarity filtering during inference (Yang et al., 9 Jul 2025).

Across these literatures, the unifying feature is not a single microscopic law but the use of diffusion to describe constrained transport, barrier crossing, or pattern formation in systems where “nuclear” may refer to nuclear spins, nuclear structure, the cell nucleus, or the nuclei of atoms. The main technical distinction is therefore contextual: whether the transported quantity is magnetization, protein density, collective shape probability, molecular occupancy, or a light nucleus itself.