Tomographic Regime

• Tomographic regime is a framework that organizes information through lower-dimensional projections, enabling stable recovery in inverse problems, electron transport, imaging, and quantum protocols.
• It leverages mathematical tools like the Fourier-slice theorem and symmetry-based twirling to reconstruct high-dimensional structures from reduced projections.
• The regime spans diverse applications—from cryo-EM and mesoscopic electron fluids to imaging modalities and quantum state tomography—highlighting its practical and theoretical significance.

The term tomographic regime is used across several research areas to denote a setting in which information is organized, propagated, or reconstructed through lower-dimensional projections, slices, or marginals. In inverse problems it refers to observation models where projections retain enough structure for exact or stable recovery; in two-dimensional electron systems it denotes an intermediate transport window between ballistic or collisionless motion and hydrodynamics; in signal analysis, field theory, cosmology, and display engineering it denotes parameterized projection families that interpolate between representations or encode depth and structure; and in generalized probabilistic theories it is tied to the question of whether local measurements suffice for state determination (Balanov et al., 9 Apr 2026, Ledwith et al., 2017, Centeno et al., 2024, Lee et al., 2018).

1. Projection models and exact moment lifting

A mathematically precise use of the term appears in the study of randomly rotated lower-dimensional projections of an unknown object $f:\mathbb{R}^n\to\mathbb{R}$, observed as

$Y=P(R\cdot f),$

with $R\sim \mathrm{Haar}(\mathrm{SO}(n))$ and $P$ the tomographic projection onto a fixed $m$-dimensional subspace. In coordinates $$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$,

$P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$

The central tool is the Fourier-slice theorem,

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$

which identifies the Fourier transform of a tomographic projection with the restriction of the ambient Fourier transform to a rotated $m$-dimensional central slice (Balanov et al., 9 Apr 2026).

The relevant observables are the full Haar-orbit $d$-th Fourier moment

$Y=P(R\cdot f),$0

and the projected $Y=P(R\cdot f),$1-th moment

$Y=P(R\cdot f),$2

For all $Y=P(R\cdot f),$3,

$Y=P(R\cdot f),$4

The main theorem states that if $Y=P(R\cdot f),$5, then the projected $Y=P(R\cdot f),$6-th order moment tensor determines the full $Y=P(R\cdot f),$7-th order Haar-orbit moment tensor on all of $Y=P(R\cdot f),$8, and the argument is dimension-free in $Y=P(R\cdot f),$9 because it uses only the fact that any $R\sim \mathrm{Haar}(\mathrm{SO}(n))$0 vectors span a space of dimension at most $R\sim \mathrm{Haar}(\mathrm{SO}(n))$1 together with Haar invariance (Balanov et al., 9 Apr 2026).

The recovery is constructive. Given $R\sim \mathrm{Haar}(\mathrm{SO}(n))$2 with $R\sim \mathrm{Haar}(\mathrm{SO}(n))$3, one takes $R\sim \mathrm{Haar}(\mathrm{SO}(n))$4, chooses an orthonormal basis of an $R\sim \mathrm{Haar}(\mathrm{SO}(n))$5-dimensional $R\sim \mathrm{Haar}(\mathrm{SO}(n))$6, extends it to an orthonormal basis of $R\sim \mathrm{Haar}(\mathrm{SO}(n))$7, forms $R\sim \mathrm{Haar}(\mathrm{SO}(n))$8, reads off $R\sim \mathrm{Haar}(\mathrm{SO}(n))$9 from $P$0, and evaluates

$P$1

A direct consequence is that any identifiability theorem for the unprojected orbit-recovery model that uses only the $P$2-th order Haar-orbit moment automatically extends to the tomographic model when $P$3. For $P$4, this recovers the classical cryo-EM statement that the covariance of random $P$5D projections determines the second-order rotationally invariant moment of the $P$6D object (Balanov et al., 9 Apr 2026).

2. Intermediate transport window in two-dimensional electron systems

In two-dimensional Fermi liquids, the tomographic regime is an intermediate transport regime created by the odd-even hierarchy of quasiparticle relaxation rates. At $P$7, Pauli blocking suppresses generic large-angle scattering and momentum and energy conservation strongly constrain two-body collisions. Head-on scattering becomes the dominant angular-relaxation channel, relaxing the even-parity part of the distribution rapidly while leaving odd-parity harmonics anomalously long-lived. In a harmonic expansion

$P$8

the model rates are

$P$9

with $m$0 and $m$1 exact zero modes from conservation laws (Ledwith et al., 2017).

The resulting hierarchy defines

$m$2

and the spatial regimes

$m$3

Within the tomographic window, motion consists of fast $m$4D spatial diffusion along the current velocity direction and slow angular diffusion or superdiffusion that randomizes the direction only gradually. Eliminating the even sector adiabatically yields the master equation

$m$5

which makes explicit the coexistence of rapid propagation along fixed directions and slow angular relaxation (Ledwith et al., 2017).

This regime is neither ordinary diffusion nor standard hydrodynamics. The conductivity and viscosity become scale dependent,

$m$6

and for a strip of width $m$7 the current profile becomes fractional-power-like rather than parabolic,

$m$8

with conductance scaling

$m$9

These fractional exponents are a characteristic signature of the tomographic transport window (Ledwith et al., 2017).

An analytically solvable interacting model shows that the longitudinal response crosses from zero sound in the collisionless regime to first sound in the tomographic and hydrodynamic regimes, whereas the transverse response acquires two purely imaginary odd-parity collective modes specific to the tomographic regime. The characteristic scale is

$$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$0

and the lower transverse branch exists only for sufficiently strong repulsion,

$$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$1

For charged liquids, the longitudinal modes become plasmons and the odd-parity damping appears only as a strongly suppressed correction (Hofmann et al., 2022).

A magnetic field suppresses tomographic transport rapidly because it breaks time-reversal invariance, which is a prerequisite for the odd-even parity effect. In a minimal model, the suppression criterion is

$$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$2

so the regime is destroyed at fields much smaller than those required to suppress ordinary hydrodynamics. A collective-mode analysis further shows that in zero field the transverse conductivity has two diffusive tomographic collective modes, and that at a critical magnetic field one of the two disappears; which branch survives depends on the Landau parameters, and the remaining mode becomes increasingly dominated by hydrodynamic harmonics at high field (Rostami et al., 2024, Maki et al., 6 Mar 2026).

A separate high-level account connects the same terminology to the superballistic paradox in electron fluids. There, replacing classical dynamics with tomographic dynamics, where only head-on collisions are allowed between electrons, is presented as the mechanism that strengthens superballistic conduction and explains why the resistance can decrease with increasing temperature starting at close-to-zero temperatures (Estrada-Álvarez et al., 14 Feb 2025).

3. Imaging regimes in physical tomography

In inverse problems and imaging physics, tomographic regime often denotes the operating condition under which projection data can be interpreted as localized or otherwise controlled probes of a target. In interior tomography, the decisive shift is from global reconstruction from un-truncated projections to exact reconstruction of a local region of interest from truncated data that directly pass through that region. The foundational relations are written in Radon-transform and Hilbert-transform form, and exact recovery is established when a known subregion is available or when the region of interest is piecewise constant or piecewise polynomial and is reconstructed by minimizing total variation or high-order total variation. The same local-reconstruction principle is elevated into omni-tomography, where multiple modalities are integrated so that diversified features are captured simultaneously with tomographic synchrony (Wang et al., 2013).

In laser-induced strong-field ionization tomography, the tomographic regime is the condition in which a focused strong-field laser acts as a spatially localized probe beam and the ion signal can be treated as a tomographic projection of the target density. The measurement is

$$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$3

so the signal is the convolution of the target density with the laser-induced ionization profile rather than a direct density map. The operational parameter is the Keldysh parameter

$$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$4

which separates multiphoton, tunneling, and intermediate strong-field regimes. The useful tomography regime is the one in which ionization is strong enough to be measurable, localized enough to be spatially selective, and narrow compared with the features one wants to resolve. The effective resolution is set by the ionization probability $$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$5, not by the beam waist, and the paper emphasizes a three-way trade-off among resolution, localization extent, and signal-to-noise ratio (Frumker, 2024).

In mesoscopic optical tomography, the target regime is a specimen scale of roughly one transport mean free path, where scattering and absorption are strong enough that projections become blurred and angle dependent. The phase-retrieved tomography protocol addresses this by moving the reconstruction problem into autocorrelation space: the autocorrelation of each projection is stacked into an autocorrelation sinogram, backprojected into a $$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$6D autocorrelation volume, and then inverted by a $$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$7D Gerchberg-Saxton family phase retrieval method using Hybrid Input-Output followed by Error-Reduction. Because autocorrelation is shift invariant, the reconstruction is registration free, and the method is presented as especially suited to tumor spheroids in the mesoscopy regime (Ancora et al., 2016).

For highly scattering media, the tomographic regime is one in which scattering is not a perturbation but the signal itself. Radiance satisfies the radiative transfer equation

$$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$8

with arbitrary-order scattering handled by Monte Carlo photon transport. This formulation is designed for in-situ imaging, where cameras may be inside or near the medium, and where forward models must remain stable despite severe conditioning problems. The proposed voxelized forward Monte Carlo model is intended to be both in-situ stable and highly parallelizable (Holodovsky et al., 2015).

4. Reconstruction algorithms and data-limited tomography

A related use of tomographic regime concerns the computational setting induced by dynamic, sparse, or nonstandard data acquisition. In four-dimensional tomography, the classical requirement that the sample remain static during one tomographic rotation is relaxed by modeling the object as

$$\mathbb{R}^n\cong\mathbb{R}^m\times\mathbb{R}^{n-m}$$9

with $P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$0 linking angle and time. The reconstruction is formulated with a coupled spatiotemporal regularizer,

$P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$1

and solved by a Chambolle–Pock primal-dual method on GPUs. This turns dynamic changes from artifacts into explicit degrees of freedom of the inverse problem (Nikitin et al., 2018).

For broad classes of static inverse problems, generalized SART shows that many tomographic Kaczmarz iterations can be reduced to projection-space optimization. Starting from a block-iterative subproblem

$P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$2

the update is recast as a lower-dimensional minimization in detector coordinates followed by backprojection. The framework covers nontrivial image-formation models, nonquadratic and nonconvex data fidelities, parallel- and cone-beam geometries, and applications including polychromatic CT and X-ray phase contrast tomography (Maretzke, 2018).

In longitudinal sparse-view tomography, the regime is defined by repeated scans of the same object, where previous reconstructions are used as priors to reconstruct a current scan from far fewer projection views. An unweighted global eigenspace prior,

$P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$3

is used when the goal is mainly to track the location of change, while a weighted version

$P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$4

is used when new structures absent from the templates must be preserved. The weight map is estimated from multiple pilot reconstructions so that new regions are downweighted relative to the prior (Gopal et al., 2019).

A quantum-optimization version appears in quantum compressed sensing tomographic reconstruction, where the reconstruction energy is cast as a QUBO combining sinogram fidelity and total variation,

$P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$5

In the reported experiments, the method obtained a solution within $P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$6 projection images for $P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$7 image samples and within $P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$8 projection images for $P(f)(u)=\int_{\mathbb{R}^{n-m}} f(u,v)\,dv.$9 image samples, reconstructing error-free CT images, using sinograms from Shepp-Logan and body CT examples (Ryou et al., 16 May 2025).

5. Tomographic representations in signals, fields, cosmology, and displays

Outside physical reconstruction, the term also denotes families of parameterized marginals or slices that serve as alternative representations. In ECG analysis, the tomogram is defined as the Radon transform of the Wigner–Ville distribution,

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$0

equivalently

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$1

The representation is nonnegative and equals the modulus squared of the fractional Fourier transform. In the reported ECG example, the most informative differences appeared near $$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$2, where the patient with early ischemic heart disease showed additional clearly resolved peaks, including a peak at zero frequency and a third harmonic (Belousov et al., 2018).

In the tomographic description of classical fields, a state is represented not directly by a phase-space density but by a family of probability densities over affine hyperplanes. For finite systems, the center-of-mass tomogram is

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$3

with inverse Radon transform

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$4

For free classical scalar fields, the field is treated as an infinite collection of harmonic oscillators, the tomogram becomes an infinite-dimensional Radon transform, and the Liouville equation is rewritten directly in tomographic variables (Ibort et al., 2012).

In quantum information geometry, the tomographic picture refers to representing a quantum state by tomograms

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$5

so that the tomographic metric on the space of tomograms is the Fisher-Rao metric,

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$6

For qubits, changing the tomographic scheme induces a diffeomorphism $$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$7, and any two Petz-form metrics can be related through a first-order, second-degree differential equation for the tomographic parameters. The paper also exhibits a nonlinear tomographic map that takes a monotone metric to a non-monotone one (Laudato et al., 2017).

In weak-lensing cosmology, a new tomographic formalism for aperture-mass statistics supplements the usual auto-slice analysis with cross-slice information. For three redshift bins, the maps

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$8

are augmented by

$$\widehat{P(R\cdot f)}(\eta)=\widehat f(R^{-1}\iota(\eta)), \qquad \iota(\eta)=(\eta,0),$$9

The reported forecasts show that auto-$m$0 precision is improved by $m$1 when including cross-$m$2, and that combining $m$3D $m$4 with the shear two-point correlation function yields a factor of three reduction of the statistical error on $m$5 compared to the $m$6-2PCF alone (Martinet et al., 2020).

In tomographic displays, the term denotes a display architecture composed of a $m$7D display panel, focus-tunable optics, and a fast spatially adjustable backlight. Depth is encoded by synchronizing the focus sweep and the backlight so that each pixel is illuminated only when the optics place that pixel’s image at the target depth. The TomoReal prototype reports more than $m$8 tomographic layers spanning $m$9 D to $d$0 D, spatial resolution $d$1, refresh rate $d$2 Hz, eye-box $d$3 mm, and field of view $d$4 (Lee et al., 2018).

6. Tomographic locality, symmetry restriction, and nonlocality

In generalized probabilistic theories, tomographic locality is the principle that local product effects suffice to distinguish all bipartite states. Formally, for distinct bipartite states $d$5, there exists a product effect $d$6 such that

$d$7

A theory that satisfies this is in a tomographic regime in which local measurements determine composite states; failure of this property is called tomographic nonlocality (Centeno et al., 2024).

A general mechanism for leaving this regime is twirling under a collective symmetry. Starting from a tomographically local theory and restricting states and effects to those invariant under a symmetry group $d$8, with transformations restricted to the covariant ones, produces a twirled world. The twirling map is

$d$9

The key theorem states that for any GPT and any physical symmetry group $Y=P(R\cdot f),$00 with an action that is nontrivial on at least one state, the corresponding twirled world fails tomographic locality. The mechanism is that globally twirled correlated states and products of locally twirled marginals are generally distinct, yet invariant product effects cannot distinguish them (Centeno et al., 2024).

The result is exhibited in quantum, classical, and post-quantum examples. In a phase-shift-twirled bosonic world, local invariant measurements are blind to the relative phase in states such as

$Y=P(R\cdot f),$01

In a bit-flip-twirled classical-bit world, local tomography fails even though the ontic state space still factorizes as $Y=P(R\cdot f),$02. This supports the conclusion that tomographic nonlocality does not imply ontological holism, and it makes superselection rules central to any reconstruction program that uses tomographic locality as an axiom (Centeno et al., 2024).

The phrase tomographic regime therefore does not denote a single universal notion. In the cited literature it names a family of projection-based informational settings: exact moment lifting under random projections, parity-split transport in $Y=P(R\cdot f),$03D electron fluids, localized and data-limited operating windows in imaging, parameterized marginal representations in signals and fields, and locality properties of composite systems under symmetry restriction. What is common across these uses is the central role of lower-dimensional slices, marginals, or covariant subspaces as the carriers of recoverable structure.