---
title: Hamiltonian Truncation Effective Theory
url: https://www.emergentmind.com/topics/hamiltonian-truncation-effective-theory
type: topic
---

# Hamiltonian Truncation Effective Theory

Hamiltonian Truncation Effective Theory (HTET) is the effective-field-theory formulation of Hamiltonian truncation, in which a quantum field theory Hamiltonian is written as \(H=H_0+V\), the Hilbert space is split into retained and discarded sectors by projectors \(P\) and \(Q=1-P\), and the effect of the discarded high-energy states is encoded in an effective Hamiltonian acting on the truncated subspace. In its standard form,
\[
H_{\mathrm{eff}}(E)=P H P + P H Q \,(E-Q H Q)^{-1} Q H P,
\]
so the truncation is treated as a low-energy EFT whose cutoff artifacts are removed, or at least organized, by local counterterms and controlled nonlocal corrections. In this sense, HTET is the organizing principle behind renormalized Hamiltonian truncation and related effective-interaction approaches, and it has become a framework for spectra, correlators, real-time dynamics, entanglement, and tensor-network algorithms in continuum QFT [2201.11696, 2110.08273, 2212.07266].

## 1. Formal construction and historical development

The basic HTET construction begins with a solvable reference Hamiltonian \(H_0\), usually a free theory or a CFT Hamiltonian, and a deformation \(V\). The full Hilbert space is decomposed as \(\mathcal H=\mathcal H_l\oplus \mathcal H_h\), with \(P\) projecting onto the truncated low-energy sector and \(Q\) onto its complement. Rewriting the Schrödinger equation in block form and formally integrating out \(Q\)-space yields the Schur-complement or Bloch–Feshbach effective Hamiltonian quoted above. Expanded around \(H=H_0+V\), this produces the standard sequence of corrections
\[
\delta H^{(2)}=P V Q\,(E-QH_0Q)^{-1}Q V P,\qquad
\delta H^{(3)}=P V Q\,(E-QH_0Q)^{-1}Q V Q\,(E-QH_0Q)^{-1}Q V P,
\]
together with higher-order terms [2201.11696].

A second, matching-based formulation constructs the effective Hamiltonian from a finite-volume transition matrix. In that approach one defines \(V_{\rm eff}=H_1+H_2+H_3+\cdots\) and fixes the operators order by order by requiring equality of the full and effective transition amplitudes. The resulting matrix elements are
\[
\langle f|H_1|i\rangle=\langle f|V|i\rangle,
\]
\[
\langle f|H_2|i\rangle=\sum_\alpha^{>} \frac{\langle f|V|\alpha\rangle\langle \alpha|V|i\rangle}{E_f-E_\alpha},
\]
\[
\langle f|H_3|i\rangle=
\sum_{\alpha,\beta}^{>} \frac{\langle f|V|\alpha\rangle\langle \alpha|V|\beta\rangle\langle \beta|V|i\rangle}{(E_f-E_\alpha)(E_f-E_\beta)}
-\sum_{\alpha}^{<}\sum_{\beta}^{>} \frac{\langle f|V|\alpha\rangle\langle \alpha|V|\beta\rangle\langle \beta|V|i\rangle}{(E_\alpha-E_\beta)(E_f-E_\beta)}.
\]
This formulation was presented explicitly as “Hamiltonian Truncation Effective Theory” and used to exhibit separation of scales in \(2\)D and \(3\)D \(\lambda\phi^4\) [2110.08273].

An alternative nonperturbative representation uses the interaction-picture operator \(\Sigma\), giving
\[
H_{\mathrm{eff}} = (\Sigma_l)^{-1}\,[\Sigma(H_0+V)\Sigma^\dagger]_l\,\Sigma_l.
\]
This representation makes spectral matching manifest: if \(\Sigma_l\) is computed exactly, diagonalizing \(H_{\mathrm{eff}}\) reproduces the low-energy spectrum of the renormalized UV theory. It also clarifies why the effective Hamiltonian is generally non-Hermitian: projection and matching are not symmetric operations on the retained and discarded sectors [2212.07266].

Historically, earlier Fock-space studies of two-dimensional \(\phi^4\) already treated truncation errors by an analytic renormalization procedure inspired by EFT, integrating out discarded high-energy states and replacing them by local counterterms in the truncated Hamiltonian [1412.3460]. The later HTET literature systematized that viewpoint, supplied explicit matching observables, and extended it beyond leading local counterterms.

## 2. Renormalization, power counting, and operator structure

HTET treats the truncation scale \(\Lambda\) or \(E_{\rm max}\) as an EFT cutoff. High-energy contributions from \(Q\)-space are reorganized into operator corrections
\[
\delta H = \sum_i c_i(\Lambda)\,O_i,
\]
with \(O_i\) local operators allowed by the symmetries of \(H_0\), and coefficients whose cutoff dependence can be organized through RG flow equations of the form
\[
\frac{d c_i}{d\log\Lambda}=\beta_i(\{c\}).
\]
For scalar theories, the dominant local operators typically include \(\phi^2\), \(\phi^4\), and derivative terms such as \((\partial\phi)^2\) [2201.11696].

In \(1+1\) dimensional \(\lambda\phi^4\), the EFT structure can be organized explicitly as an expansion in \(1/E_{\rm max}\). A matching analysis establishes that \(H_2\) contains local leading-order terms of order \(\lambda^2/E_{\rm max}^2\) and controlled nonlocal next-to-leading-order terms of order \(\lambda^2/E_{\rm max}^3\), while \(H_3\) starts at order \(\lambda^3/E_{\rm max}^4\). The same analysis identifies the unique NLO nonlocal correction to \(H_2\) as
\[
H_{2}^{(\rm NLO)} = \zeta H_0
+ \frac{1}{2}\alpha_1^{(1)}\!\int R\,d\theta\,\{H_0,:\phi^2:\}
+ \frac{1}{2}\alpha_2^{(2)}\!\int R\,d\theta\,[H_0,:\phi^2:]
+ \frac{1}{4!}\beta_1^{(1)}\!\int R\,d\theta\,\{H_0,:\phi^4:\}
+ \frac{1}{4!}\beta_2^{(2)}\!\int R\,d\theta\,[H_0,:\phi^4:].
\]
Including these nonlocal matching terms removes the \(1/E_{\rm max}^3\) error and leaves residual errors scaling as \(1/E_{\rm max}^4\), in agreement with EFT power counting [2507.15941].

The higher-order structure is richer still. A later analysis of \(2\)D \(\lambda\phi^4\) derived all-order resummations of infinite classes of local corrections to the mass and quartic coupling and extended the nonlocal sector to the next-to-next-to-local level at \(\mathcal O(E_{\rm max}^{-4})\). In that formulation the effective operator basis includes not only \(H_2=\int dx:\phi^2:\) and \(H_4=\int dx:\phi^4:\), but also composites such as \(H_0^p H_2 H_0^q\), \(H_0^p H_4 H_0^q\), and frequency-weighted structures represented by double commutators with \(H_0\). The continuum-first matching procedure introduced there computes the distributional coefficients in infinite volume and only then re-compactifies to finite volume, specifically to avoid ambiguities associated with \(\Theta\), \(\Theta'\), and \(\Theta''\) terms at discrete energies [2602.13019].

This hierarchy makes clear that HTET is not merely a prescription for adding a few counterterms. It is an order-by-order expansion in which locality is the leading approximation, while nonlocality appears in a controlled way through insertions of \(H_0\), external energies, and, at higher orders, additional operator classes.

## 3. Bases, truncation schemes, and symmetry realization

HTET is independent of a single basis choice. What changes from scheme to scheme is the definition of the projector \(P\), the kinematic organization of states, and the form of matrix elements. The Snowmass white paper emphasizes three principal continuum truncation schemes. In the Truncated Conformal Space Approach (TCSA), \(P\) projects onto CFT states organized by radial quantization on \(S^{d-1}\), with cutoff set by conformal energy or scaling dimension. In Discrete Light-Cone Quantization (DLCQ), \(P\) retains states at fixed total light-cone momentum \(K\). In Lightcone Conformal Truncation (LCT), \(P\) is a basis of CFT primary operators in Minkowski light-cone quantization truncated by scaling dimension [2201.11696].

These basis choices determine which symmetries are manifest. TCSA and LCT implement CFT data directly, and the white paper stresses that TCSA is intrinsically gauge-invariant when formulated entirely in terms of local gauge-invariant operators. DLCQ preserves gauge invariance in \(d=2\), where gauge fields carry no local degrees of freedom. Across all schemes, parity, global internal symmetries, and some Lorentz structure can be implemented exactly inside the truncated space, organizing states by irreducible representations [2201.11696].

Equal-time massive Fock-space truncation remains central in scalar theories. In two-dimensional \(\phi^4\), one works in a finite-volume Fock basis of free-theory eigenstates and imposes an \(H_0\)-energy cutoff. Early analytic renormalization studies used exactly this setup and already treated the discarded sector as an EFT contribution built from local operators such as \(V_0\), \(V_2\), and \(V_4\) [1412.3460]. In later next-to-leading-order renormalized truncation, the same basis was enlarged by “tail states”
\[
|\Psi_i\rangle=(\mathcal E_* - H_0)^{-1}V_{hl}|i\rangle,
\]
which package the UV-dominant part of the discarded sector in a variationally controlled way [1706.09929].

For UV-divergent deformations of \(1+1\) dimensional CFTs, the basis is constructed from Virasoro descendants and the truncated eigenproblem takes the generalized form \(H v=E G v\), with \(G\) the non-diagonal Gram matrix. In that context, local renormalization with a short-distance regulator \(\epsilon\) is followed by an effective-Hamiltonian construction that analytically removes the local regulator and leaves nonlocal \(K_{\rm eff}\) terms fixed by CFT OPE data. This strategy was implemented for deformed Ising, Tricritical Ising, and the non-unitary minimal model \(M(3,7)\), including multi-operator flows and coupling-constant renormalization [2312.09221].

The choice of basis is therefore not secondary. It fixes how locality, symmetry, and UV data are encoded, and it determines whether the most natural HTET language is one of local fields, conformal blocks, light-front kinematics, or Fock-space mode sums.

## 4. Spectra, correlators, dynamics, and entanglement

The most developed applications of HTET are spectral. In two-dimensional \(\phi^4\), next-to-leading-order renormalized truncation produced infinite-volume estimates
\[
g=0.2:\quad m_{\rm ph}=0.979733(5),\quad \Lambda=-0.0018166(5),
\]
\[
g=1.0:\quad m_{\rm ph}=0.7494(2),\quad \Lambda=-0.03941(2),
\]
\[
g=2.0:\quad m_{\rm ph}=0.345(2),\quad \Lambda=-0.1581(1),
\]
together with a critical coupling
\[
g_c=2.76(3),
\]
and finite-size gaps at criticality consistent with Ising CFT scaling dimensions [1706.06121]. A later systematic HTET analysis based on order-by-order matching reported
\[
g_c=\lambda_c/4! = 2.752(5),
\]
again for the \(1+1\) dimensional theory flowing to the \(2\)D Ising CFT [2507.15941].

In three-dimensional \(\phi^4\), the white paper emphasizes that adding the appropriate local counterterms made Hamiltonian truncation practical and predictive at strong coupling: spectra obeyed strong–weak duality and correlators near criticality exhibited universal scaling consistent with \(3\)D Ising exponents. The same review also summarizes applications to \(2\)D scalar theories, sine-Gordon dualities, deformed Ising models, and \(2\)D gauge theories, including spectral functions for \(T_{\mu\nu}\) and extractions of scattering information from truncation eigenstates through analytic S-matrix methods [2201.11696].

In \(d=2+1\), the effective-theory role of HTET becomes especially explicit. For scalar \(\phi^4\) on a spatial torus, the sharp energy cutoff produces state-dependent missing-state effects that spoil vacuum-bubble cancellations. The effective Hamiltonian was therefore supplemented not only by local vacuum and mass counterterms but also by state-dependent corrections \(\delta V^{I}\) and \(\delta V^{II}\) that restore the factorization of disconnected bubbles under the truncation regulator. With these HTE corrections included, spectra converged as \(E_T\to\infty\) and passed a weak/strong self-duality test [2003.08405].

HTET also supports real-time and entanglement observables. Once \(H_{\rm eff}\) is known, Lorentzian evolution follows from \(\exp(-iH_{\rm eff}t)\), enabling studies of spectral form factors, eigenstate thermalization, hydrodynamic transport, and level-statistics diagnostics such as Wigner–Dyson versus Poisson behavior [2201.11696]. For reduced density matrices, an explicit isomorphism \( \mathcal H_F \simeq \mathcal H_A\otimes\mathcal H_B \) can be constructed in a free basis by a multimode Bogoliubov transform. This makes possible direct computation of
\[
S(A)=-{\rm Tr}(\rho_A\log\rho_A),\qquad
S_n(A)=\frac{1}{1-n}\log{\rm Tr}(\rho_A^n),
\]
as well as mutual information and logarithmic negativity, for ground states, thermal states, and real-time quenches. Benchmarks were given for the free Klein–Gordon theory and for the interacting sine-Gordon model [2202.11113].

The scope of HTET observables is thus broader than energy levels alone. Its effective Hamiltonians also organize correlators, spectral densities, entanglement measures, and out-of-equilibrium dynamics in a common finite-volume continuum framework.

## 5. Algorithmic realizations and tensor-network extensions

A distinctive feature of HTET is that its effective corrections can be realized algorithmically in several inequivalent ways. One influential route is the “tail-state” construction. Starting from the exact Feshbach equation, next-to-leading-order renormalized truncation enlarges the variational space by the states
\[
|\Psi_i\rangle=(\mathcal E_* - H_0)^{-1}V_{hl}|i\rangle,
\]
eliminates them, and obtains the effective correction
\[
\Delta \widetilde H = \Delta H_2 \,(\Delta H_2-\Delta H_3)^{-1}\, \Delta H_2.
\]
Its series expansion reproduces \(\Delta H_2+\Delta H_3+O(V^4)\), but the matrix inverse stabilizes the large matrix elements that make a naive cubic truncation unreliable. In \(2\)D \(\phi^4\) this led to smoother UV extrapolations and \(E_T^{-3}\) scaling, rather than the \(E_T^{-2}\) behavior of raw or leading-order truncation [1706.09929].

A complementary development is Hamiltonian Truncation Tensor Networks (HTTN), which combines truncation with momentum-space MPS/MPO methods. The Hilbert space is truncated by a sharp momentum cutoff \(|k|\le k_{\max}\) together with nonuniform local occupation cutoffs \(n(k)\), and total momentum conservation is imposed by an exact global projector represented as a symmetric MPS. The bond dimension of this \(\delta\)-MPS satisfies
\[
D_\delta=K\le 4 k_{\max} n_{\max}+1,
\]
and the resulting interacting MPO bond dimension obeys
\[
D=2K+2\le 8 k_{\max} n_{\max}+4.
\]
The method uses two-site DMRG for spectra and TDVP with global Krylov expansion for dynamics. Reported truncated Hilbert spaces reach \(\dim(\mathrm{HT})\sim 10^{10}\), the sine-Gordon gap agrees with integrability after \(1/k_{\max}\) extrapolation, and for the bosonized massive Schwinger model at \(\theta=\pi\) the critical point was located at
\[
m_c=0.333(2),
\]
consistent with prior numerics \(0.3335(2)\) [2312.12506].

These two constructions exemplify different HTET philosophies. Tail-state methods integrate out a selected class of high-energy states explicitly and resum their effect into a low-dimensional operator correction. HTTN instead pushes the truncation window itself to much larger sizes, implements the projector exactly at the operator level, and controls the discarded sector mainly through RG-informed parameter choices and cutoff extrapolation. Both remain compatible with the general HTET language, in which the physics of \(Q\)-space must be represented inside the truncated theory, whether by explicit Schur-complement corrections, by matched counterterms, or by high-capacity variational representations.

## 6. Locality, non-Hermiticity, and open problems

A central technical issue in HTET is locality. The white paper emphasizes that a hard cutoff in the total energy of the system is a nonlocal condition, so counterterms often cannot be represented as purely local interactions without further analysis [2201.11696]. This problem was sharpened by a study of conformal perturbation theory up to fourth order, which found that in Hamiltonian truncation extra UV divergences appear once
\[
\Delta \ge \frac{d}{2}+\frac{1}{4}.
\]
Above this threshold, the mismatch between connected and subtraction terms behaves as
\[
S_A-S_B \sim \mathcal A_4\, \Delta_T^{\,4\Delta-2d-1},
\]
signaling a regulator-induced breakdown of locality. In that regime, a purely local HTET is insufficient; one must introduce nonlocal or bilocal counterterms, modify the truncation prescription, or restrict attention to deformations below the threshold [2112.09049].

This does not imply that HTET fails to describe a local QFT. A complementary analysis argues that one should first renormalize the UV theory with a local regulator, then construct the effective Hamiltonian for the renormalized theory, and only afterward remove the local regulator analytically. The resulting \(K\) terms may be nonlocal as operators on the truncated Hilbert space, yet still ensure that the low-energy spectrum matches that of the local theory [2212.07266]. In this sense, nonlocality in HTET is often a property of the regulator and of the reduced description, not of the underlying QFT itself.

Non-Hermiticity is another persistent feature. The transition-matrix matching construction produces an effective Hamiltonian that is generally non-Hermitian, and the matching paper argues that complete Hermiticity appears incompatible with preserving separation of scales in that framework [2110.08273]. Later works on systematic \(1/E_{\rm max}\) improvement confirmed that some nonlocal matching terms are generically non-Hermitian, especially when commutators with \(H_0\) first enter at order \(E_{\rm max}^{-3}\) [2507.15941].

Open problems follow directly from these structural facts. The Snowmass review identifies operator proliferation at strong coupling, rapidly growing basis size, the development of conformal three-point kinematics for LCT in \(d>2\), and the treatment of gauge invariance and Gauss’s law in \(d\ge 3\) as major challenges. It also singles out full LSZ scattering, improved RG schemes for \(c_i(\Lambda)\), and integration with bootstrap, tensor-network, and quantum-simulation methods as important directions [2201.11696]. Work on UV-divergent CFT deformations likewise shows that multi-operator flows, nonunitary theories, and PT-symmetric Hamiltonians can be handled, but only with increasingly elaborate \(K_{\rm eff}\) structures and careful control of Gram-matrix and cutoff effects [2312.09221].

HTET therefore occupies a technically delicate position. It is a systematic EFT of truncation, not a guarantee that a naive cutoff will preserve locality, Hermiticity, or gauge symmetry automatically. Its strength lies precisely in making those failures explicit, classifying them in operator language, and providing a controlled route to correct them.

Source: https://www.emergentmind.com/topics/hamiltonian-truncation-effective-theory