---
title: Witnessed Quantum Time Evolution (WQTE)
url: https://www.emergentmind.com/topics/witnessed-quantum-time-evolution-wqte
type: topic
---

# Witnessed Quantum Time Evolution (WQTE)

Witnessed Quantum Time Evolution (WQTE) is used in the cited literature for several related constructions in which quantum evolution is certified, quantified, or reconstructed through an explicit witness rather than taken as primitive. In these constructions, the witness is variously realized as system–clock entanglement in a history state, an ancilla interferometric signal built from controlled real-time evolution, a coherence test under a conservation law, a speed-based indicator of open-system memory effects, or a macroscopic record defining an intrinsic time observable [1512.07313][2603.17953][2205.00198][2003.12681][2607.01296].

## 1. Terminological scope and principal usages

In the cited literature, WQTE does not denote a single formalism. It appears in discrete history-state models of time, in relational Page–Wootters constructions, in quantum-cosmological discussions of internal or intrinsic time, in open-system witnessing protocols, and in a 2026 single-ancilla algorithm for spectral estimation [1512.07313][1310.4691][1011.4444][2003.12681][2603.17953].

| Usage | Witness | Representative source |
|---|---|---|
| History-state evolution | $E(S,T)$ or $E_2(S,T)$ between system and clock | [1512.07313] |
| Optical parallel-in-time simulation | Measured system-time entanglement and single-shot time averages | [1901.04808] |
| Relational time | Conditional system state relative to clock outcomes | [1310.4691] |
| Spectral algorithm | $Q(t)=\langle\psi_{\mathrm{ref}}|\cos(Ht)|\psi_{\mathrm{ref}}\rangle$ | [2603.17953] |
| Non-Markovian open dynamics | $\chi(t)=\frac{d}{dt}HSS(\rho_\varphi(t))$ | [2003.12681] |
| Temporal non-classicality witness | $C_{\ell_1}(\rho_Q(t))>0$ under a conservation law | [2205.00198] |

A common structure nonetheless recurs. A system is coupled to a clock, ancilla, mediator, or record subsystem, and evolution is inferred from an observable quantity built from that coupling. In the history-state line, the quantity is entanglement between a system and a quantum clock. In the algorithmic line, it is an oscillatory signal whose Fourier spectrum resolves eigen-energies. In the temporal-witness line, it is a coherence or speed increase that cannot arise under a specified classical or Markovian null model [1512.07313][2603.17953][2205.00198][2003.12681].

## 2. History states, system–time entanglement, and distinguishable evolution

A central history-state construction introduces a system $S$ of dimension $M$ and a finite-dimensional clock $T$ of dimension $N$ with orthonormal basis $\{|t\rangle_T:t=0,\dots,N-1\}$. The discrete evolution is encoded in the joint state
$$
|\Psi\rangle_{ST}=\frac{1}{\sqrt{N}}\sum_{t=0}^{N-1} |\psi_t\rangle_S\otimes |t\rangle_T,
$$
with $|\psi_t\rangle=U_t|\psi_0\rangle$ and, for a constant Hamiltonian,
$$
|\psi_t\rangle=U^t|\psi_0\rangle,\qquad U=e^{-iH\Delta t}.
$$
Projective measurement of the clock in the $\{|t\rangle\}$ basis returns the conditional system state at step $t$ [1512.07313].

The same framework defines the global translation super-operator
$$
\mathcal{U}=\sum_{t=1}^{N} U_{t,t-1}\otimes |t\rangle\langle t-1|,\qquad U_{t,t-1}=U_tU_{t-1}^\dagger,
$$
with cyclic condition $|N\rangle\equiv |0\rangle$, such that
$$
\mathcal{U}|\Psi\rangle=|\Psi\rangle.
$$
Writing $\mathcal{U}=e^{-i\mathcal{J}}$, one obtains $\mathcal{J}|\Psi\rangle=0$, presented as a discrete counterpart of the Wheeler–DeWitt equation [1512.07313].

The history state turns distinguishability of evolution into entanglement. For pure histories,
$$
E(S,T)=S(\rho_S)=S(\rho_T),\qquad S(\rho)=-\mathrm{Tr}(\rho\log_2\rho),
$$
and the quadratic monotone is
$$
E_2(S,T)=2\big(1-\mathrm{Tr}\,\rho_S^2\big).
$$
For the history state,
$$
E_2(S,T)=2\frac{N-1}{N}\left[1-\frac{1}{N(N-1)}\sum_{t\neq t'}|\langle\psi_t|\psi_{t'}\rangle|^2\right],
$$
so $E_2$ is a decreasing function of the average pairwise fidelity between visited states [1512.07313].

The extremal cases are explicit. If $|\psi_t\rangle=e^{i\gamma_t}|\psi_0\rangle$ for all $t$, then $E(S,T)=0$. If $\langle\psi_t|\psi_{t'}\rangle=0$ for all $t\neq t'$, then $E(S,T)=\log_2N$. The mapping
$$
\tau_m=2^{E(S,T)}-1
$$
therefore yields $\tau_m=0$ for stationary states and $\tau_m=N-1$ for a fully orthogonal trajectory [1512.07313].

For cyclic constant-Hamiltonian evolution with $U^N=\mathbbm{1}$ and equally spaced spectrum $E_k=2\pi k/N$ up to a constant shift, the history state becomes Schmidt-diagonal in the energy basis and the system–time entanglement is exactly the entropy of the energy-basis distribution,
$$
E(S,T)=-\sum_k |c_k|^2\log_2|c_k|^2,
$$
while
$$
E_2(S,T)=2\sum_k |c_k|^2(1-|c_k|^2).
$$
In this cyclic case the entropic energy–time uncertainty relation
$$
E(S,T)+\tilde E(S,T)\ge \log_2N
$$
holds, where $\tilde E$ is the entropy of the discrete Fourier-transformed amplitudes $\tilde c_l$ [1512.07313].

A two-step clock makes the dependence on state distance especially transparent. For
$$
|\Psi\rangle=\frac{1}{\sqrt{2}}\big(|0\rangle_T|\psi_0\rangle_S+|1\rangle_T|\psi_1\rangle_S\big),
$$
the eigenvalues of $\rho_S$ are $p_\pm=\frac12(1\pm|\langle\psi_0|\psi_1\rangle|)$, and
$$
E_2(T,S)=1-|\langle\psi_0|\psi_1\rangle|^2.
$$
For pure states with fidelity $F=|\langle\psi(0)|\psi(t)\rangle|^2$, this yields
$$
E(T,S)=H_2\!\left(\frac{1+\sqrt F}{2}\right),\qquad E_2(T,S)=1-F.
$$
For two qubits, the same literature further states that increasing internal entanglement speeds up the evolution and makes the system more entangled with time [2306.11675].

## 3. Operational realizations and experimental demonstrations

An experimental optical realization of the history-state program uses photon polarization as system $S$ and transverse spatial modes on a reflective LCoS spatial light modulator as clock $T$. The controlled operation is
$$
\mathcal{W}=\sum_{t=0}^{N-1} U_t\otimes |t\rangle\langle t|,
$$
applied after preparation of an equal superposition of clock modes, so that
$$
|\Psi\rangle=\mathcal{W}(\mathbbm{1}\otimes H)|\psi_0\rangle|0\rangle
=\frac{1}{\sqrt N}\sum_{t=0}^{N-1}(U_t|\psi_0\rangle)\otimes |t\rangle.
$$
In this platform, the time average of an observable $O$ is obtained directly from
$$
\langle\Psi|O\otimes\mathbbm{1}_T|\Psi\rangle=\frac1N\sum_t\langle\psi_t|O|\psi_t\rangle,
$$
so that one single global measurement yields a time average that would otherwise require separate measurements over all time steps [1901.04808].

The optical implementation used a 660 nm solid state laser, a HoloEye Lc-R 2500 SLM with $1024\times 768$ pixels, and an Andor Zyla 4.2 sCMOS camera. Mueller–Stokes characterization and Lu–Chipman decomposition were used to extract effective polarization unitaries for 52 gray levels. The practical modulation range was reported as gray levels $0$–$40$, while above $40$ LCoS pulse-width-modulation flicker induced depolarization and phase fluctuations. History states with $N=2,4,8$ were generated, time averages obtained in the image plane and Fourier plane showed excellent agreement with theory, and the operator entanglement of the implemented $\mathcal{W}$ over all 52 extracted unitaries was reported as $E^{(2)}(\mathcal{W})=0.712$; Monte Carlo over 1000 random inputs verified
$$
\langle E^{(2)}(S:T)\rangle=\frac{d_S}{d_S+1}E^{(2)}(\mathcal{W}),\qquad d_S=2,
$$
to within $\le 0.01$ [1901.04808].

A distinct photonic experiment illustrated the Page–Wootters mechanism with two entangled photons. The singlet state
$$
|\Psi\rangle=\frac{1}{\sqrt2}\big(|H\rangle_c|V\rangle_r-|V\rangle_c|H\rangle_r\big)
$$
was prepared so that, with local Hamiltonians $H_C=H_S=\hbar\omega\sigma_y$, the global constraint
$$
(H_C+H_S)|\Psi\rangle=0
$$
held. The internal observer conditioned on the clock photon and reconstructed time-dependent conditional statistics, while the external “super-observer” erased which-time information and performed full two-photon tomography. In the GPPT-style two-time version, the reported conditional probabilities were
$$
p(3|t_f=t_1)=\frac{1+2\cos^2(\omega\tau)}{4},\qquad
p(3|t_f=t_2)=\frac{1+2\sin^2(\omega\tau)}{4},
$$
so that the internal observer witnessed oscillatory evolution, whereas the external observer found the global fidelity $F(T)=\langle\Psi|\rho_{\mathrm{out}}(T)|\Psi\rangle$ constant and close to $1$ across all external times $T$ [1310.4691].

These experiments establish two complementary operational meanings of WQTE. In one, the witness is system–time entanglement together with single-shot time averaging. In the other, the witness is relational dynamics recovered from conditioning on a physical clock while the global state remains stationary [1901.04808][1310.4691].

## 4. WQTE as a quantum algorithm for eigen-energy spectra

A 2026 formulation defines WQTE as a spectral algorithm using one ancillary qubit $q_0$ and $n$ target qubits. After preparing a reference state $|\psi_{\mathrm{ref}}\rangle$ on the target register, a Hadamard is applied to the ancilla, followed by a controlled real-time evolution $U(t)=e^{-iHt}$ and a second Hadamard. The circuit states are
$$
|T1\rangle=\frac{|0\rangle\otimes|\psi_{\mathrm{ref}}\rangle+|1\rangle\otimes|\psi_{\mathrm{ref}}\rangle}{\sqrt2},
$$
$$
|T2\rangle=\frac{|0\rangle\otimes|\psi_{\mathrm{ref}}\rangle+|1\rangle\otimes e^{-iHt}|\psi_{\mathrm{ref}}\rangle}{\sqrt2},
$$
$$
|T3\rangle=|0\rangle\otimes\frac{(1+e^{-iHt})|\psi_{\mathrm{ref}}\rangle}{2}
+|1\rangle\otimes\frac{(1-e^{-iHt})|\psi_{\mathrm{ref}}\rangle}{2}.
$$
Measuring Pauli-$Z$ on the ancilla yields
$$
Q(t)=P_0(t)-P_1(t)=\langle\psi_{\mathrm{ref}}|\cos(Ht)|\psi_{\mathrm{ref}}\rangle.
$$
In the energy basis $H|E_i\rangle=E_i|E_i\rangle$ and $|\psi_{\mathrm{ref}}\rangle=\sum_i c_i|E_i\rangle$,
$$
Q(t)=\sum_i |c_i|^2\cos(E_it),
$$
so a single time series carries all eigen-energies with nonzero overlap [2603.17953].

Sampling at $t_n=n\Delta$ with $n=0,1,\dots,N$ and $T_{\max}=N\Delta$, the algorithm applies a discrete Fourier transform
$$
R(k)=\sum_{n=0}^{N-1} q(n)e^{-i2\pi kn/N},
$$
with $q(n)=Q(n\Delta)$. The energy magnitudes are extracted from peak locations according to
$$
|E_i|=2\pi x_{k_i}= \frac{2\pi k_i}{N\Delta},
$$
and the peak heights approximate the overlap weights,
$$
|c_i|^2\approx a_{k_i}.
$$
Because the signal uses $\cos(Ht)$, it is insensitive to the sign of $E_i$; the paper therefore proposes shifting the Hamiltonian by a positive offset $S_0I$, with the recommendation $S_0\approx 5\delta$–$10\delta$, where $\delta=2\pi/T_{\max}$ [2603.17953].

The precision claim is Heisenberg-limited. The spectral resolution is set by
$$
\delta=\frac{2\pi}{T_{\max}},
$$
and numerical fits gave
$$
E_{\mathrm{err}}\approx \frac{5.964}{T_{\max}}
$$
with $R^2\approx 0.987$, described as very close to the $2\pi/T_{\max}$ law. For chemical accuracy $1.6$ mHa, the paper states the guideline $T_{\max}>3927$ in its units [2603.17953].

Sampling must satisfy the Nyquist bound
$$
\Delta\le \Delta_{\max}=\frac{\pi}{|E|_{\max}},
$$
and the paper reports concrete thresholds. For H$_4$ at $1.0$ Å, $\Delta$ must not exceed approximately $1.4$; at $2.0$ Å, $\Delta<1.6$. For a one-dimensional Heisenberg model with 10 sites and $J=h=1$, $\Delta\lesssim 0.18$, while for a $4\times 3$ Heisenberg model with $J=h=1$, $\Delta\lesssim 0.1$ [2603.17953].

The noise model distinguishes sampling noise, compilation error, and circuit noise. Sampling noise produces a white-noise floor but does not shift true peak positions. A half-normal model for the DFT amplitudes leads to the practical rule that peaks larger than $4\sigma$ are extremely unlikely to be due to sampling noise, with probability approximately $0.0063\%$. Circuit noise is modeled by a Pauli channel after each gate,
$$
\mathcal{E}(\rho)=(1-\gamma)\rho+\frac{\gamma}{3}(X\rho X+Y\rho Y+Z\rho Z),
$$
which attenuates the signal by a factor $f(\gamma)\approx (1-\gamma)^{N_G}$ but, according to the analysis and simulations, does not shift frequencies. Second-order Suzuki–Trotter compilation preserves the time-reversal symmetry needed by the cosine signal, and the peak shifts scale quadratically with the Trotter step [2603.17953].

The resource comparison is explicit. WQTE requires only nonzero overlap with eigenstates of interest, rather than eigenstate preparation. It resolves multiple eigenvalues in parallel from one time series, and the abstract states that compared to VQE, QPE, and their variants, it exhibits superior circuit depth efficiency, resource economy, and noise resilience. Numerical demonstrations were carried out for H$_4$ chains and Heisenberg spin systems, while an NMR implementation on a 3-qubit SPINQ Triangulum processor recovered correct peak positions for a two-site Heisenberg model despite visible time-domain noise [2603.17953].

## 5. Alternative witness constructions

A speed-based notion of WQTE arises in open quantum systems through the Hilbert–Schmidt speed,
$$
v_{\mathrm{HS}}(t)=\sqrt{\frac12\,\mathrm{Tr}\!\left[(\dot\rho(t))^2\right]},
$$
or, for a parameterized state family $\rho(\varphi)$,
$$
HSS(\rho_\varphi)=\sqrt{\frac12\,\mathrm{Tr}\!\left[\left(\frac{d\rho_\varphi}{d\varphi}\right)^2\right]}.
$$
The witness criterion is
$$
\chi(t):=\frac{d}{dt}HSS(\rho_\varphi(t))>0\quad\Rightarrow\quad \text{non-Markovianity},
$$
and the same framework gives a quantum-speed-limit bound
$$
\tau\ge \frac{D_{\mathrm{HS}}(\rho(0),\rho(\tau))}{\overline v_{\mathrm{HS}}},
$$
with $\overline v_{\mathrm{HS}}=\frac1\tau\int_0^\tau v_{\mathrm{HS}}(t)\,dt$. The paper reports complete qualitative agreement between this HSS witness and the Breuer–Laine–Piilo trace-distance witness across phase-covariant qubit noise, Pauli channels, leaky-cavity two-qubit models, and V- and $\Lambda$-type qutrit systems [2003.12681].

A different temporal witness addresses non-classicality of a mediator. With a probe $Q$, mediator $M$, conserved quantity
$$
K_{\mathrm{tot}}=K_Q+K_M,
$$
and interaction Hamiltonian
$$
H_{\mathrm{int}}=gA_M\otimes B_Q,
$$
the conservation law
$$
[H,K_{\mathrm{tot}}]=0
$$
implies
$$
[A_M\otimes B_Q,K_M+K_Q]=[A_M,K_M]\otimes B_Q+A_M\otimes[B_Q,K_Q]=0.
$$
If $Q$ acquires coherence in a basis that does not commute with $K_Q$, then $[B_Q,K_Q]\neq 0$, forcing $[A_M,K_M]\neq 0$. The paper formulates this as: if a physical system $Q$ evolves from an eigenstate of one of its observables to an eigenstate of a different, non-commuting observable under the conservation law of a global quantity of the composite $Q\oplus M$, then $M$ must be non-classical. Operationally, with $Q$ initially diagonal in the $K_Q$ basis, the observation
$$
C_{\ell_1}(\rho_Q(t))=\sum_{i\neq j}|\langle i|\rho_Q(t)|j\rangle|>0
$$
witnesses mediator non-classicality under the stated assumptions [2205.00198].

A third line treats an effect $a$ that occurs at $t=0$ but is not read out as the generator of subsequent evolution. For effects $a,b\in E(H)$,
$$
b(t|a)=e^{-ita}\,b\,e^{ita},
$$
and the time-dependent sequential product is
$$
a[t]b=a^{1/2}e^{-ita}be^{ita}a^{1/2}.
$$
These obey
$$
\frac{d}{dt}b(t|a)=i[b(t|a),a],\qquad
\frac{d}{dt}a[t]b=i[a[t]b,a].
$$
The central theorem states that $a[t]b$ is constant in time if and only if either $[a,b]=0$ or $a=\lambda p$ with $p$ a projection and $\lambda\in[0,1]$ [2105.07046].

These constructions are not identical, but each defines a witness quantity whose nontrivial value certifies a specific form of quantum temporal structure: information backflow, mediator non-classicality, or measurement-conditioned dynamics [2003.12681][2205.00198][2105.07046].

## 6. Relational time, q-number calculus, and quantum cosmology

Several sources place WQTE within relational or timeless formulations of quantum theory. In a Heisenberg-picture Page–Wootters framework, a clock $C$ carries a q-number time observable $T$ and conjugate generator $P_T$ satisfying
$$
[T,P_T]=i\hbar I_C.
$$
For bounded operator functions $f(\omega T)$, the q-number derivative is defined spectrally and obeys
$$
\frac{df(\omega T)}{dT}=\frac{i}{\hbar}[P_T,f(\omega T)].
$$
Relational observables are then functions of the clock’s q-number time, and the constraint
$$
[P_T-H_S,q(\omega T)]=0
$$
is equivalent to the relational Heisenberg equation
$$
\frac{d q(\omega T)}{dT}=\frac{i}{\hbar}[H_S,q(\omega T)].
$$
This formulation explicitly argues that c-number time is unnecessary in the Heisenberg picture as well as in the Schrödinger picture [2108.02771].

In quantum cosmology, one influential alternative introduces an internal, unobservable “test clock” to define evolution relative to the proper time along a specific observer’s worldline. The clock has Hamiltonian $h$ and origin variable $t_0$ satisfying
$$
\{t_0,h\}=-1\qquad\Rightarrow\qquad [\hat h,\hat t_0]=i\hbar.
$$
For the homogeneous FRW minisuperspace model with a massless scalar, the total classical constraint is
$$
C=H_0+h\approx 0,
$$
and quantum Dirac observables evolve as
$$
\widehat F(T)=e^{\frac{i}{\hbar}\hat H_0(t_0+T)}\hat f\,e^{-\frac{i}{\hbar}\hat H_0(t_0+T)},
\qquad
\frac{d}{dT}\widehat F(T)=\frac{i}{\hbar}[\hat H_0,\widehat F(T)].
$$
The stated role of the test clock is to witness evolution without itself being a Dirac observable available to the observer [1011.4444].

A complementary minisuperspace analysis argues that quantum evolution can be described in four distinct but agreeing ways. For a flat FRW universe with a massless scalar field, the Wheeler–DeWitt equation is
$$
\left[\frac{\partial^2}{\partial \alpha^2}-\frac{\partial^2}{\partial \phi^2}\right]\Psi(\alpha,\phi)=0,
$$
with $\alpha=\ln a$. The paper compares deparametrized Schrödinger evolution, conditional probabilities in the Wheeler–DeWitt picture, a semiclassical WKB construction, and group averaging or path-integral methods with Grassmann variables, and reports that the approaches yield the same time dependence for key observables such as $\langle a\rangle$, $\langle a^2\rangle$, and $\langle\phi^2\rangle$ [2003.11284].

A recent intrinsic-time construction develops the same theme under a positive external Hamiltonian. Here the Schrödinger parameter $t$ is declared not to be physical time, and the cumulative timeless state is defined by group averaging,
$$
|\Psi\rangle:=\int_{\mathbb R} dt\,|\psi(t)\rangle,
$$
leading to
$$
H|\Psi\rangle=0.
$$
Intrinsic time is a macroscopic pointer observable $T$ with $\tau$-uniform spectral measure, and the intrinsic generator is
$$
K:=-i\hbar\frac{\partial}{\partial\tau},\qquad [T,K]=i\hbar.
$$
The resulting intrinsic Schrödinger-like equation is
$$
i\hbar\,\partial_\tau |\psi(\tau)\rangle=K|\psi(\tau)\rangle.
$$
Macroscopic pointer records are said to witness a strict temporal order within each branch, while large-scale time-reversing or discontinuous transitions are not internally observable in the records [2607.01296].

Taken together, these relational constructions treat the witness as a clock observable, test clock, or record structure that makes evolution meaningful inside a globally stationary description [2108.02771][1011.4444][2003.11284][2607.01296].

## 7. Assumptions, limitations, and scope

The literature imposes strong model-dependent assumptions. In the discrete history-state framework, the exact identity between system–time entanglement and the entropy of the energy-basis distribution requires a cyclic equally spaced spectrum $E_k=2\pi k/N$; for arbitrary spectra the result is only the majorization bound
$$
E(S,T)\le -\sum_k |c_k|^2\log_2|c_k|^2.
$$
The model also presumes a finite clock dimension and stepwise evolution, even though nonconstant $U_t$ is allowed [1512.07313].

Experimental realizations inherit platform-specific limits. In the SLM-based optical simulation, finite ROI and mode orthogonality constrain clock dimension, while gray levels above $40$ induce depolarization and phase fluctuations. The setup also uses two degrees of freedom of a single photon, so the nonseparable mode structure is local rather than nonlocal entanglement, although it is still used as a witness of simulated dynamics [1901.04808].

The algorithmic WQTE formulation has its own constraints. Because it measures $\cos(Ht)$, it does not directly determine the sign of eigen-energies without an additional Hamiltonian offset. Accurate reconstruction also requires Nyquist-compliant sampling, sufficiently small Trotter step, and nonzero overlap with the targeted eigenstates. The paper’s noise analysis emphasizes that circuit noise suppresses peak heights but not peak positions, whereas poor overlap can bury weak peaks under the sampling-noise floor [2603.17953].

The temporal non-classicality witness depends critically on a conservation law and on exclusion of hidden quantum mediators. The cited criterion $C_{\ell_1}(\rho_Q(t))>0$ certifies non-classicality only if $[H,K_{\mathrm{tot}}]=0$ is operationally satisfied and no auxiliary quantum system is responsible for the observed coherence [2205.00198].

Relational-clock programs also face idealization issues. The q-number calculus assumes an ideal clock with $\mathrm{Sp}(T)=\mathbb R$, while the paper itself notes that physical clocks must be unideal and finite-dimensional. The intrinsic-time construction requires rigged Hilbert space methods, $\tau$-uniformity may fail for pathological $T$, and recovery of the standard subsystem Schrödinger equation from the internal perspective of the rest of the universe is identified as an open direction [2108.02771][2607.01296].

These limitations indicate that WQTE is best read as a cluster of witness-based approaches to quantum temporality rather than a single standardized theory. Across that cluster, however, a stable technical theme remains: evolution is certified through entanglement, conditioned states, interferometric signals, conserved-quantity constraints, or macroscopic records, and the witness itself becomes part of the dynamical description [1512.07313][2603.17953][2205.00198][2607.01296].

Source: https://www.emergentmind.com/topics/witnessed-quantum-time-evolution-wqte