---
title: Projector-Splitting Time Evolution
url: https://www.emergentmind.com/topics/projector-splitting-time-evolution
type: topic
---

# Projector-Splitting Time Evolution

Searching arXiv for recent and foundational papers on projector-splitting time evolution and closely related dynamical low-rank / TDVP methods.
Projector-splitting time evolution denotes a class of geometric time-integration procedures in which an evolution constrained by a projector is decomposed into simpler subflows that can be solved successively. In the dynamical low-rank setting, the projector is the orthogonal projector onto the tangent space of the manifold of fixed-rank matrices, and the resulting scheme propagates a low-rank approximation directly, rather than evolving a full matrix and recomputing a truncated factorization at every step. The formulation introduced for matrices by Lubich and Oseledets is explicit, computationally inexpensive, and notable for robustness when the core factor becomes ill-conditioned or nearly singular [1301.1058]. Across later literature, closely related constructions appear in tensor-network dynamics, multiconfiguration quantum propagation, kinetic equations, spatial projection methods for quasiperiodic PDEs, and low-rank optimization or training, although the phrase is not used in exactly the same sense in every subfield.

## 1. Variational and geometric foundation

The basic object in dynamical low-rank approximation is a time-dependent matrix \(A(t)\in\mathbb{R}^{m\times n}\), or the solution of a matrix ODE \(\dot A = F(A)\). One seeks an approximation \(Y(t)\) of fixed rank \(r\ll \min(m,n)\), constrained to the manifold
\[
\mathcal{M}_r=\{Y\in\mathbb{R}^{m\times n}:\operatorname{rank}(Y)=r\}.
\]
The defining variational principle requires \(\dot Y(t)\) to be the best tangent approximation to \(\dot A(t)\), or to \(F(Y(t))\) in the ODE setting. This yields the projected evolution
\[
\dot Y(t)=P(Y(t))\,F(Y(t)),
\]
where \(P(Y)\) is the orthogonal projector onto the tangent space \(T_Y\mathcal M_r\) [1301.1058].

Writing
\[
Y(t)=U(t)S(t)V(t)^T,
\]
with \(U\in\mathbb{R}^{m\times r}\) and \(V\in\mathbb{R}^{n\times r}\) column-orthonormal and \(S\in\mathbb{R}^{r\times r}\) invertible, together with gauge conditions
\[
U^T\dot U=0,\qquad V^T\dot V=0,
\]
gives the standard factor equations. Direct integration of those equations becomes problematic when \(S(t)\) is ill-conditioned or nearly singular, which corresponds to over-approximating the actual rank. The projector formulation avoids placing \(S^{-1}\) at the center of the numerical method [1301.1058].

A key identity is the explicit tangent-space projector
\[
P(Y)Z = ZVV^T - UU^T ZVV^T + UU^T Z,
\]
or equivalently
\[
P(Y)Z = ZP_{\mathcal R(Y^T)} - P_{\mathcal R(Y)} ZP_{\mathcal R(Y^T)} + P_{\mathcal R(Y)}Z.
\]
This splittable structure is the starting point of projector-splitting time evolution: the projected vector field is decomposed into three simpler pieces associated with row-space projection, overlap correction, and column-space projection [1301.1058].

## 2. Construction of the matrix projector-splitting integrator

The first-order integrator is a Lie–Trotter splitting of the projected evolution into three subproblems corresponding to the three terms of the tangent-space projector. In factored form, these become linear updates for \(US\), then for \(S\), then for \(VS\). For known matrix increments \(\Delta A=A(t_1)-A(t_0)\), the practical KSL ordering is
\[
K_1 = U_0S_0 + \Delta A\,V_0,
\]
factorized as \(U_1S_1=K_1\), followed by the core correction
\[
\tilde S_0 = S_1 - U_1^T\Delta A\,V_0,
\]
and then
\[
L_1 = V_0\tilde S_0^T + \Delta A^T U_1,
\]
factorized as \(V_1S_1^T=L_1\), after which
\[
Y_1 = U_1S_1V_1^T.
\]
This algorithm never uses \(S^{-1}\) or \((S^T)^{-1}\); it uses matrix additions and QR or SVD on thin matrices only [1301.1058].

The same splitting admits a symmetric second-order version by composing the first-order step with its adjoint at half step size. For known \(A(t)\), Lubich and Oseledets give an explicit formula involving \(A_0\), \(A_{1/2}\), and \(A_1\); the method is second-order accurate for the projected evolution and still requires no matrix inversions [1301.1058].

The construction extends directly to matrix ODEs \(\dot A=F(A)\). In that case, \(\Delta A\) is replaced by a numerical approximation based on \(F(Y)\): first order by \(hF(Y_0)\), and second order by a predictor–corrector using a linear approximation to \(\dot A(t)\) inside the symmetric splitting. This keeps the scheme fully explicit while preserving the geometric structure of the fixed-rank manifold [1301.1058].

## 3. Exactness, robustness, and adaptive-rank consequences

A central exactness result states that if \(A(t)\) has rank at most \(r\) for all \(t\), then with \(Y_0=A(t_0)\) the KSL splitting step is exact, \(Y_1=A(t_1)\). This property depends on the KSL ordering; other orderings such as KLS generally lose it [1301.1058].

The more distinctive result concerns over-approximation. If the true matrix is \(A'(t)\) of rank \(q<r\), perturbed by a small full-rank term \(A(t)=A'(t)+\varepsilon A''(t)\), then the rank-\(r\) and rank-\(q\) projector-splitting trajectories satisfy
\[
\|Y_n-\tilde Y_n\|\le C(\varepsilon+h),
\]
uniformly in time, with \(C\) independent of \(n\), \(h\), and \(\varepsilon\). In other words, the excess-rank components remain small even when the core \(S\) becomes very ill-conditioned [1301.1058].

This is not the behavior of standard integrators applied directly to factor equations containing \(S^{-1}\) and \(S^{-T}\). In the numerical experiments of Lubich and Oseledets, the implicit midpoint rule becomes unstable or fails in over-approximation regimes, and a KLS variant also loses robustness, whereas KSL and its symmetrized version maintain stable, accurate results and the expected order. For example, with \(\varepsilon=10^{-6}\) and over-approximation \(r=20\), midpoint fails, KLS-based methods lose nominal behavior, and KSL-based methods remain stable [1301.1058].

Two algorithmic consequences follow directly. First, robustness under over-approximation makes adaptive rank increase feasible, because the new degrees of freedom can be introduced through very small singular directions without ever evaluating \(S^{-1}\). Second, one projector-splitting step along the linear path \(A(t)=A+t\Delta A\) can serve as an efficient truncation or retraction back to the rank-\(r\) manifold in low-rank optimization methods [1301.1058].

## 4. Tensor-network and multiconfiguration formulations

In multiconfiguration time-dependent Hartree, the tangent-space projector of the MCTDH manifold can be written as
\[
\mathcal P(\Psi)=\mathcal P_0(\Psi)+\sum_\kappa \mathcal P_\kappa(\Psi),
\]
and, after orthogonalizing single-hole functions and introducing non-orthogonal single-particle functions, the SPF part becomes
\[
\mathcal P_\kappa(\Psi)=\mathcal P_\kappa^+(\Psi)-\mathcal P_\kappa^-(\Psi).
\]
The resulting projector-splitting algorithm yields linear equations for the transformed SPFs and avoids direct inversion of the single-particle density matrix. In this setting, projector splitting refers not to a Hamiltonian splitting but to a splitting of the tangent-space projector of the Tucker/MCTDH manifold [1802.01058].

For matrix product states, Haegeman and collaborators formulated TDVP time evolution by splitting the projector onto the MPS tangent space rather than using a Suzuki–Trotter splitting of the Hamiltonian. The resulting algorithm resembles finite-system DMRG so closely that it can be implemented by changing only a few lines of code, is compatible with arbitrary Hamiltonians including long-range interactions, and yields DMRG as a special case of imaginary time evolution with infinite time step [1408.5056].

The same geometric idea persists at deeper tensor-network levels. In ML-MCTDH, a projector-splitting integrator based on a gauge in which both SPFs and SHFs are orthonormal removes the singular reduced-density inverses responsible for weak-entanglement instabilities. On large spin-boson benchmarks, this implementation was reported to require roughly \(3\)–\(4\) orders of magnitude fewer Hamiltonian evaluations and \(2\)–\(3\) orders of magnitude fewer Hamiltonian applications than standard ML-MCTDH, with calculations including up to \(10^6\) bath modes and wavefunctions containing up to \(\sim 1.3\times 10^9\) parameters [2109.03134].

In two-dimensional tensor networks, an iPEPS time-step algorithm from first principles applies a short-time evolution that enlarges the bond dimension and then variationally projects back to the fixed-\(D\) iPEPS manifold by maximizing overlap using CTMRG environments. This is a projector step in a variational tensor-network manifold rather than a tangent-space KSL step, but it is explicitly cast as a projector-based time-evolution strategy and provides proof of principle for real-time and imaginary-time evolution on an infinite 2D lattice [1804.03872].

A TD-DMRG benchmark on the Fenna–Matthews–Olson complex found that TDVP-MU and TDVP-PS yield the same result when the time step size is converged, that both are more accurate than P\&C-RK4, and that TDVP-PS tolerates a larger time step size than TDVP-MU; GPU acceleration of the heavy tensor contractions sped up TDVP-MU and TDVP-PS by up to \(73\) times [1907.12044].

## 5. PDEs, kinetic equations, and stability theory

For the Vlasov–Poisson equation, a low-rank projector-splitting integrator constrains the dynamics to a manifold
\[
f(t,x,v)\approx \sum_{i,j=1}^r X_i(t,x)S_{ij}(t)V_j(t,v)
\]
and splits the tangent-space projector into three subprojections. In four and six phase-space dimensions, one time step is reduced to two systems of two- or three-dimensional advection equations and a small matrix ODE; with hierarchical dynamical low-rank approximation, the same step is further reduced to a set of four or six systems of one-dimensional advection equations whose size is still equal to the rank [1801.01103].

A different but related use of the phrase appears in the nonlinear quasiperiodic Schrödinger equation, where the method combines a spatial projection onto a finite quasiperiodic spectral space with Strang splitting in time. The fully discrete update has the form
\[
\Psi_m = F(\tfrac{\tau}{2})I_N \circ S(\tau)\circ F(\tfrac{\tau}{2})I_N \Psi_{m-1},
\]
and the main error estimate is
\[
\|\Psi_m-\psi(\cdot,t_m)\|\le C(\tau^2+N^{-\alpha}),
\]
which gives second-order accuracy in time and spectral accuracy in space under the stated regularity assumptions [2411.06641].

For linear hyperbolic and parabolic equations, recent stability analysis of the low-rank projector-splitting integrator distinguishes discretize-then-project and project-then-discretize formulations. For hyperbolic equations, the full upwind–Euler scheme is \(L^2\)-stable if \(\nu\le 1\), whereas Lie–Trotter PSI with Euler gives \(\nu\le 1/3\) for both DtP and PtD; Strang splitting with SSP-RK2 significantly enlarges the stability region, to approximately \(\nu\lesssim 0.866\) for DtP and \(\nu\lesssim 2\) for PtD. For parabolic equations, despite the negative S-step, unconditional stability is recovered with Crank–Nicolson in all substeps or with a hybrid scheme using backward Euler in K and L and forward Euler in S [2507.15192].

These PDE examples show two distinct roles for projection. In kinetic low-rank dynamics, the projector is the tangent-space projector of a low-rank manifold of functions. In quasiperiodic Schrödinger discretization, the projector \(I_N\) is a spatial spectral projector intertwined with time splitting. The formal similarity is real, but the underlying geometric object is different.

## 6. Optimization, machine learning, and broader meanings of the term

Projector-splitting ideas have moved from approximation of prescribed dynamics to optimization and training. In the matrix setting, one splitting step along \(A(t)=A+t\Delta A\) already acts as an efficient retraction onto the fixed-rank manifold, which is why the original paper explicitly points to optimization algorithms for low-rank matrices as a second application area besides time evolution [1301.1058].

In dynamical low-rank training of neural networks, the projected gradient flow
\[
\dot Y(t) = -P(Y(t))\nabla \ell(Y(t))
\]
inherits the same geometric stiffness issues as matrix ODEs. A recent augmented backward-corrected projector-splitting integrator introduces
\[
\widehat U_1=\mathrm{ortho}([U_0,K_1])
\]
so that \((I-\widehat U_1\widehat U_1^T)Y_0=0\), removes the descent obstruction of backward-corrected PSI, reduces the number of QR factorizations per step from two to one, and proves both a robust global error bound
\[
\|Y(t_n)-W(t_n)\|\le \varepsilon + c_1 h + c_2\delta + \frac{\vartheta}{h}
\]
and a loss decrease inequality for sufficiently small \(h\) [2502.03006].

The expression also appears in open quantum systems in a more abstract projector-operator sense. There, a time-dependent superoperator projector \(\mathcal P(t)\) splits relevant and irrelevant sectors of Liouville space, and generalized TCL theory produces time-local master equations for \(\mathcal P(t)\rho(t)\). With generalized Kawasaki–Gunton projectors, this leads to nonlinear master equations on ansatz manifolds determined by relevant observables. In that setting, “projector-splitting time evolution” refers to following a moving relevant subspace in operator space rather than splitting the tangent projector of a fixed-rank matrix manifold [2307.00607].

A persistent misconception is therefore that projector-splitting time evolution always means splitting the Hamiltonian. In the foundational low-rank, MPS, and MCTDH literature, what is split is the projector onto a tangent space, not the Hamiltonian itself [1301.1058]. In other literatures, the phrase can denote operator splitting combined with a spatial or statistical projector. The common core is geometric restriction of dynamics by projection; the specific projector, manifold, and meaning of the split depend on the application domain.

Source: https://www.emergentmind.com/topics/projector-splitting-time-evolution