Light-Ray Transform of Vector Fields

• The light-ray transform is an integral transform that computes the cumulative effect of a vector field along null geodesics in Lorentzian or Minkowski space.
• It reveals crucial properties regarding the kernel, gauge freedoms, and injectivity, with applications in inverse problems, tensor tomography, and hyperbolic PDEs.
• Recent developments include momentum-weighted transforms and explicit inversion methods, ensuring stable recovery in complex geometric and analytic settings.

A light-ray transform of a vector field is an integral transform that assigns to a vector field its integrals along null geodesics (lightlike lines) of a Lorentzian manifold or, in flat spacetime, Minkowski space. This transform arises as a mathematical abstraction of measuring the cumulative effect of a vector field along propagation directions of light, and it plays a central role in inverse problems for hyperbolic PDEs, tensor tomography, and geometric rigidity questions. The kernel and invertibility properties of the light-ray transform are governed by natural gauge freedoms, geometric assumptions, and—in specific cases—by analytic microlocal structure. Advanced results connect the light-ray transform to magnetic geodesic flows, X-ray transforms, and their weighted (momentum) variants.

1. Formal Definition and Global Geometric Setting

The standard setup for the light-ray transform of a vector field involves a smooth Lorentzian manifold $(N,g)$ of dimension $1+n$ and signature $(-,+,\dots,+)$, endowed with a time orientation. A null (lightlike) geodesic $\gamma(s)$ is a curve satisfying $g(\dot\gamma(s),\dot\gamma(s))=0$. Given a compactly supported vector field $X \in C_c^\infty(N;TN)$, its light-ray transform $L X$ is defined as

$$L X(\gamma) = \int_{-\infty}^\infty g\bigl(X(\gamma(s)), \dot\gamma(s)\bigr)\, ds,$$

where $\gamma$ ranges over all maximal null geodesics. In Minkowski space $\mathbb{R}^{1+n}$ with metric $(-,+,\ldots,+)$, identifying a null line through $(t,x)$ in direction $\theta\in S^{n-1}$ as $\gamma_{x,\theta}(s) = (s, x+s\theta)$, the light-ray transform for $$f \in C_0^\infty(\mathbb{R}^{1+n};\mathbb{R}^{1+n})$$ is

$$L f(x, \theta) = \int_{-\infty}^\infty f(s, x + s\theta)\cdot (1, \theta)\, ds.$$

This integral is well-defined for fields with suitable decay or compact support, since null lines escape any compact set in finite affine time (RabieniaHaratbar, 2016).

2. Gauge Structure and Characterization of the Kernel

The kernel of the light-ray transform reflects intrinsic gauge freedoms: any field of the form $X = \nabla\phi$ annihilates under the transform, since

$$g(\nabla\phi, \dot\gamma) = \frac{d}{ds} \bigl(\phi \circ \gamma(s)\bigr).$$

The integral telescopes and vanishes for compactly supported $\phi$. Consequently,

$$\ker L \supseteq \{ \nabla\phi : \phi \in C_c^\infty(N) \}.$$

Thus, injectivity can only be sought modulo these potentials (Feizmohammadi et al., 2019).

In the Minkowski case for $n\geq3$, full-data injectivity yields that $L f \equiv 0$ implies $d f = 0$, and thus (by Poincaré lemma for simply connected domains) $f = d\phi$ (RabieniaHaratbar, 2016). In two dimensions, the kernel is described in terms of $\mathrm{curl}\,\phi$.

For transforms with additional structure—such as the mixed ray transform with a Hodge-star factor (Ilmavirta et al., 2020)—the kernel consists of fields of the form $\nabla^\perp \phi = \star d\phi$, with $\phi$ vanishing at the boundary.

3. Injectivity and Conditions for Recovery

Injectivity of the light-ray transform on vector fields depends on manifold geometry, support assumptions, and analytic or dynamical properties. Several foundational results delineate these scenarios:

• Stationary Lorentzian manifolds: When the metric $g$ admits a timelike Killing field, null geodesics project to magnetic geodesics in the spatial domain. On $\overline M = \mathbb{R}_t \times M_x$ with metric

$\overline g = - (dt + \omega)^2 + g,$

the light-ray transform acting on $\alpha \in C^\infty(S^m T^*\overline M)$ integrates $\alpha$ along the flows governed by $X = (1-\omega(v)) \partial_t + G$, with $G$ the magnetic generator induced by $-d\omega$ (Oksanen et al., 6 Feb 2025). Injectivity holds up to the obstruction of potentials and traces, provided $G$ satisfies the finite degree property: for finite-degree functions on the unit sphere bundle $SM$, the transport equation $G u = f,~u|_{\partial_-(SM)}=0$ has solutions of degree at most $m-1$.

• Globally hyperbolic manifolds: Under convex foliation or real-analyticity with no null cut points, $L X=0$ implies $X=\nabla\phi$ for some scalar potential $\phi$ (Feizmohammadi et al., 2019).
• Static case: If the spatial X-ray transform on one-forms is solenoidally injective (i.e., vanishes only for exact forms vanishing at the boundary), then light-ray transform injectivity follows modulo potentials. The transport reduction method and Fourier slicing in time are central to this argument.
• Support theorems: On Minkowski space, vanishing of $L f$ on an open set of rays ensures that the support of $d f$ (or curl) is disjoint from the corresponding region traversed by those rays. Full-data injectivity means $f$ is uniquely determined up to a potential (RabieniaHaratbar, 2016).

4. Weighted and Momentum Light-Ray Transforms

The momentum light-ray transform (MLRT) generalizes the classical transform by integrating with polynomial weights: $$L^{1,k} f((t,x), \omega) = \sum_{i=0}^n \widetilde\omega_i \int_{\mathbb{R}} s^k f_i(t+s, x+s\omega)\, ds, \qquad \omega \in \mathbb{R}^n \setminus \{0\},$$ with $\widetilde\omega = (1,\omega)$ and $k$ indicating the moment order. For $m=1$ (vector fields), the first moment $L^{1,1}f$ is already fully injective: $L^{1,1} f \equiv 0 \implies f \equiv 0,$ with no gauge obstruction (Bhattacharyya et al., 21 Oct 2025). This injectivity persists for data restricted to any nonempty open subset of directions when $n\geq3$.

MLRT admits explicit analytic inversion algorithms under partial data. Fourier-slice relations connect the transform data in $(t,x,\omega)$ to the Fourier transform of $f$ in space-time, allowing reconstruction via linear systems in the orthogonal complement of propagation directions. The normal operator associated to MLRT is elliptic, ensuring stable inversion on Sobolev scales.

5. Reduction to X-ray and Magnetic Ray Transforms

On stationary Lorentzian manifolds, the light-ray transform relates closely to magnetic X-ray transforms on the spatial part. The null geodesic flow decomposes into time and magnetic spatial evolution, reducing the light-ray integral to an overdetermined transport problem for functions of finite vertical Fourier degree on the unit sphere bundle.

Algebraically, in low-dimensional settings (e.g., orientable surfaces), the light-ray transform can be represented as a composition $L = I \circ A$, where $I$ is the classical geodesic X-ray transform and $A$ an invertible bundle map (e.g., Hodge star). This structure facilitates explicit inversion through known solenoidal reconstruction operators (Ilmavirta et al., 2020).

Table: Light-ray Transform Scenarios and Injectivity Conditions

Geometric Setting	Injectivity Modulo	Key Condition
Stationary Lorentzian manifold	Potentials + traces	Finite degree property for $G$
Globally hyperbolic, convex foliation	Potentials	Strictly convex foliation
Real analytic, no null cut points	Potentials	Analyticity, absence of conjugate points
Static Lorentzian manifold	Potentials	Solenoidal injectivity for spatial X-ray
Minkowski space	Potentials (gradient fields)	Full data, simply connected domain
MLRT ($m=1$)	Trivial (no nonzero kernel)	Full or partial data, $k=1$ moment

6. Stability Estimates and Explicit Inversion

The composition of the light-ray transform with its formal adjoint yields a normal operator that, under appropriate geometric hypotheses (e.g., simple surfaces, nontrapping domains), is elliptic of order $-1$ on vector fields. Analogous statements hold for the MLRT. Consequently, Sobolev norms of the field are controllable by the norms of transform data. For divergence-free fields, there is no loss of derivatives, and boundary data admits sharp Sobolev trace inequalities.

Explicit inversion schemes exploit the Fourier-slice relationships and the decomposability of the transform in terms of geometric or potential theoretic components, with analytic continuation and reconstruction in the determined region (Bhattacharyya et al., 21 Oct 2025).

7. Applications and Extensions

The light-ray transform on vector fields underlies the uniqueness and recovery of time-dependent vector potentials in hyperbolic PDEs from boundary measurements, as in inverse wave problems. Its integral geometric features connect to rigidity theorems, geometric optics, and control for PDEs. The transform's kernel is invariant under conformal scaling of the metric, preserving injectivity modulo potentials.

Momentum-weighted transforms extend utility in tensor tomography and inverse problems under restricted measurements, admitting stable analytic inversion and further connecting to microlocal analysis. The link to magnetic transport and sphere bundle decompositions enables generalizations to magnetic and mixed X-ray settings. Stability results guarantee robust recovery in the presence of geometric or analytic complexity.

In summary, the light-ray transform of vector fields is a central construct in geometric tomographic analysis, admitting rigorous injectivity and inversion under precise analytic and dynamical conditions, with explicit relevance for inverse problems, rigidity, and integral geometry.