Exchange-Time Integral (ETI)

• Exchange-Time Integral (ETI) is a field-dependent concept representing various time-integrated measures, from exposure time indicators in clinical trials to running integrals in neutron scattering.
• It captures dynamic phenomena by integrating temporal correlations, offering flexibility to model delayed onset, ramp-up effects, and nonlocal interactions across disciplines.
• ETI frameworks provide actionable insights for experimental design and power calculations, yet require careful contextual qualification to avoid misinterpretation as a universal parameter.

Searching arXiv for papers relevant to "Exchange-Time Integral (ETI)" and its distinct uses across fields. Tool unavailable in this environment; proceeding with the supplied arXiv records and citing the relevant papers directly. Exchange-Time Integral (ETI) is not a single standardized quantity across the arXiv literature. In the materials considered here, the acronym is used explicitly in biostatistics as Exposure Time Indicator, while in neutron scattering, open quantum systems, diffusion MRI, and equilibrium theory the closest analogues are time-integrated or time-indexed exchange observables rather than a named ETI object. The resulting concept is therefore field-dependent: an ETI may denote an exposure-time parametrization, a running time-integral of a dynamical correlation function, a nonlocal exchange kernel in a path integral, or an implicit intertemporal valuation rule. This heterogeneity is itself a substantive feature of the term’s usage (Kenny et al., 14 Mar 2025, Benedetto et al., 2019, Carrega et al., 2014, Kiselev et al., 28 Jan 2026, Riane, 14 Oct 2025).

1. Terminological status and cross-disciplinary scope

Within the cited papers, only one work uses ETI as a formal acronym: the Exposure Time Indicator model for stepped wedge cluster randomized trials (SW-CRTs). That paper also defines an explicit averaging functional,

$$\text{TATE}(s_1,s_2) \equiv \frac{1}{s_2-s_1}\int_{s_1}^{s_2}\delta(s)\,ds,$$

which is the closest literal “integral ETI” construct in the supplied literature (Kenny et al., 14 Mar 2025).

Other domains use ETI-like structures without naming them ETI. In neutron scattering, the directly measured quantity is the running time-integral of the van Hove function,

$$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$

with the derivative recovering $I(Q,t)$ (Benedetto et al., 2019). In open quantum systems, heat exchange is encoded by a counting-field generating functional and an explicitly nonlocal time kernel (Carrega et al., 2014). In diffusion MRI, by contrast, the relevant paper argues that the FEXI-derived $\tau_x$ is an effective, protocol-dependent summary statistic, not a true exchange-time integral (Kiselev et al., 28 Jan 2026). In decentralized equilibrium theory, time enters through debt rollover, discounting, and the integral of interest rates,

$$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$

but no formal ETI is defined (Riane, 14 Oct 2025).

Context	Quantity closest to ETI	Status in the source
SW-CRT biostatistics	ETI model and TATE integral	Explicit ETI acronym
Neutron scattering	$vHI(t)$	Direct running time-integral
Open quantum systems	$G_\nu(t)$, $\Phi^{(1)}$	ETI-like functional structure
Diffusion MRI	FEXI-derived $\tau_x$	Not a true exchange-time integral
General equilibrium theory	$\exp(\int r(s)\,ds)$	Implicit intertemporal representation

A plausible implication is that “Exchange-Time Integral” functions best as an umbrella label only when the surrounding field is specified. Without that specification, the term is ambiguous.

2. Exposure Time Indicator in stepped wedge trials

In SW-CRTs, the Exposure Time Indicator model is designed for situations in which treatment effects vary with exposure time, meaning time since a cluster crossed from control to intervention. The mean model is

$$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$0

with $$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$1, and the mixed-effects ETI specification is

$$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$2

Here ETI replaces a single post-treatment effect with exposure-time-specific parameters $$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$3, allowing the effect curve to vary arbitrarily over time since implementation (Kenny et al., 14 Mar 2025).

This framework induces two primary estimands. The Point Treatment Effect is

$$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$4

while the Time-Averaged Treatment Effect is

$$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$5

The paper explicitly rejects a design-dependent discrete average of ETI parameters in favor of this integral definition because the integral is valid for both discrete and continuous time and for varying period lengths. In this literature, if “Exchange-Time Integral” is used informally, TATE is the closest exact match.

The ETI model is contrasted with the Immediate Treatment model,

$$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$6

which assumes an immediate and constant effect after treatment initiation. The paper states that using IT when treatment effects vary with exposure time can produce “severely misleading results.” ETI accommodates delayed onset, ramp-up, washout, long-term accumulation, and different short-term versus long-term benefits.

The principal cost of this flexibility is statistical power. For common SW-CRT designs, maintaining $$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$7 power when switching from IT to ETI for the full-study TATE requires a sample-size increase of roughly 2.5 to 3, with the paper also reporting factors of about 2.3 to 2.8 across many settings. Estimand choice is decisive: in a six-sequence design, the sample size ratio is about 2.7 for $$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$8, about 2.1 for $$vHI(t_{\mathrm{obs}})\equiv \int_0^{t_{\mathrm{obs}}} I(Q,t)\,dt,$$9, and 1.4 for $I(Q,t)$0. For PTEs, about 100 individuals per cluster-period are needed for $I(Q,t)$1, but nearly 800 are needed for $I(Q,t)$2. The asymmetry reflects the design fact that all sequences contribute at exposure time $I(Q,t)$3, whereas only one sequence contributes at the largest exposure time.

Design recommendations follow directly. Power for ETI-based TATE estimands is improved by adding time points to the start of the study or increasing baseline sample size, while adding time points to the end yields little gain. Increasing the number of clusters is usually more effective than increasing individuals per cluster-period. The paper also gives a Delayed Constant Treatment model for washout settings,

$I(Q,t)$4

and a Natural Cubic Spline model,

$I(Q,t)$5

These structured alternatives reduce flexibility when scientifically justified; the paper states that DCT is preferred when a constant post-washout effect is plausible.

3. Running time-integrals in neutron scattering

In neutron scattering, the paper “Dynamics from elastic neutron-scattering via direct measurement of the running time-integral of the van Hove distribution function” introduces a genuinely integral observable in the time domain. The relevant quantity is not ETI by name, but the van Hove integral,

$I(Q,t)$6

where $I(Q,t)$7 is the van Hove self-correlation function or intermediate scattering function in the time domain. The core result is that, under the appropriate mismatch between primary and secondary spectrometer resolutions, the measured elastic intensity at zero energy transfer is precisely this cumulative integral, and

$I(Q,t)$8

This makes ETI-like reasoning literal: dynamics are recovered by differentiating a measured time-integral (Benedetto et al., 2019).

The general count-rate formula is written as

$I(Q,t)$9

with $\tau_x$0 and $\tau_x$1 the primary and secondary response functions. In the elastic channel, $\tau_x$2, and if one response is much narrower than the other, the experiment reduces to a single time integral. In the idealized step-function limit,

$\tau_x$3

The paper interprets this signal as counting “species at rest” within the observation time.

A practical calibration links instrumental energy width to observation time: $\tau_x$4 with the unit conversion

$\tau_x$5

This mapping is central because the method converts an elastic-intensity-versus-resolution scan into a time-domain dynamical observable.

The method is presented as a conceptual alternative to conventional quasielastic analysis. Standard neutron spectroscopy measures $\tau_x$6, deconvolves resolution, and then Fourier transforms to obtain $\tau_x$7. Here the route is reversed: the experiment remains in the elastic channel, varies resolution, measures the cumulative integral $\tau_x$8, and recovers $\tau_x$9 by differentiation. The paper states that prior approaches failed to identify that the elastic intensity versus resolution is the running time integral of $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$0, not $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$1 itself.

The numerical validation uses single exponential, double exponential, and stretched exponential test dynamics. Reported parameters include $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$2 ps for the single exponential, $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$3 ps and $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$4 ps for the double exponential, and $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$5 ps with $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$6 for the stretched exponential. The simulated primary resolution spans approximately $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$7, corresponding to about $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$8 ps, with fixed analyser widths $$v^i(t_1)=p_{L+1}(t_0)\exp\!\left(\int_{t_0}^{t_1} r^i(s)\,ds\right),$$9 in one set of tests and $vHI(t)$0 in the McStas example. The paper also emphasizes limitations: current spectrometers do not implement the method directly, long-time points are count-time expensive, and differentiation is sensitive to noise.

4. Path-integral exchange kernels in open quantum systems

In open quantum systems, the paper “Functional Integral approach to time-dependent heat exchange in open quantum systems” develops a formally exact path-integral framework for the statistics of heat transferred from a driven quantum system to a thermal reservoir. The central object is the moment generating function

$vHI(t)$1

from which all heat moments follow. Heat is defined operationally by two projective measurements of bath energy at $vHI(t)$2 and $vHI(t)$3. Although the paper does not name this construction ETI, it is ETI-like in the precise sense that exchange is represented by a time-nonlocal functional over histories (Carrega et al., 2014).

After integrating out the reservoir, the problem reduces to a generalized influence functional $vHI(t)$4, and in the sum-and-difference variables

$vHI(t)$5

the characteristic functional becomes

$vHI(t)$6

The $vHI(t)$7-dependent correction $vHI(t)$8 contains the heat-counting information and is explicitly nonlocal in time.

For the first heat moment, the relevant kernel is

$vHI(t)$9

which yields the average heat as a path integral weighted by $G_\nu(t)$0. In this sense, $G_\nu(t)$1 is the paper’s closest ETI-like object: the exchanged quantity is not a local observable at one instant but a time-integrated functional of the forward and backward trajectories.

For the dissipative two-state system, the average heat power $G_\nu(t)$2 is written in weak damping as a convolution integral involving system population correlations $G_\nu(t)$3, coherence correlations $G_\nu(t)$4, and the bath polarization correlation $G_\nu(t)$5. The paper emphasizes that heat transfer depends jointly on population dynamics, coherence dynamics, and bath memory. This rules out an interpretation of heat exchange as a purely local system observable.

The formalism is then specialized to several regimes. In the white-noise or Markov regime, $G_\nu(t)$6, and the heat power simplifies. At lower temperature, colored quantum noise produces genuinely non-Markovian behavior, and the heat power splits as $G_\nu(t)$7, where $G_\nu(t)$8 contains Matsubara sums and represents a purely quantum contribution absent in the Markov limit. The paper states that the method achieves a complete description of the average heat transfer from the classical regime down to zero temperature.

5. Apparent exchange times in diffusion MRI

The diffusion MRI paper “What Does FEXI Measure in Neurons?” addresses a concept that superficially resembles ETI but is explicitly argued not to be a unique integral measure. Standard FEXI analysis fits the recovery of apparent diffusivity after a strong filter block with

$G_\nu(t)$9

where $\Phi^{(1)}$0 is the mixing time and $\Phi^{(1)}$1 is the apparent exchange time. The paper’s central conclusion is that, in realistic ramified neurons, FEXI does not directly measure a single intrinsic membrane exchange time unless the geometry is effectively well mixed and membrane-limited. Instead, the recovery is multiexponential, and the fitted $\Phi^{(1)}$2 is a protocol-dependent effective summary statistic (Kiselev et al., 28 Jan 2026).

The mathematical reason is the decomposition of Bloch–Torrey dynamics into Laplace eigenmodes,

$\Phi^{(1)}$3

so that many modes contribute to the recovery. In the bent-cylinder analytical model, the crossover scale is identified with

$\Phi^{(1)}$4

With $\Phi^{(1)}$5 and $\Phi^{(1)}$6, the corresponding timescale is of order $\Phi^{(1)}$7 s, explaining why a monoexponential asymptote appears only at long mixing times.

This has direct consequences for interpretation. In the simulations, fitting only short mixing times $\Phi^{(1)}$8 ms gives $\Phi^{(1)}$9 ms. Extending the fit to $\tau_x$0 ms yields $\tau_x$1 ms, and using the full simulated range up to seconds gives $\tau_x$2 ms. The fast short-time component is attributed to geometric exchange, meaning intraneuronal redistribution of spins between microdomains with different local diffusivities, rather than to transmembrane water passage.

To estimate membrane-limited exchange, the paper reinterprets preexchange-lifetime data using a two-compartment model,

$\tau_x$3

with equilibrium condition $\tau_x$4. Using $\tau_x$5 and $\tau_x$6 ms yields $\tau_x$7 ms, and matching the slow escape rate gives a membrane permeability

$\tau_x$8

The paper’s interpretive point is precise: exchange times in the tens of milliseconds inferred from fast FEXI-style fits are too short to be interpreted as neuronal membrane permeability. In this context, any ETI language would be misleading if it suggested a unique geometry-independent exchange integral.

6. Implicit exchange-time frameworks in decentralized equilibrium theory

The paper “Rethinking Arrow–Debreu: A New Framework for Exchange, Time, and Uncertainty” also does not define ETI as a named object, but it develops an explicitly time-indexed theory of exchange in which bilateral feasibility, production lags, debt dynamics, and uncertainty are unified. Exchange is decentralized: each agent chooses a subjective price vector $\tau_x$9 and intended bilateral trades $\exp(\int r(s)\,ds)$0, while effective trades are given by

$\exp(\int r(s)\,ds)$1

Trade therefore occurs only when both sides are willing to transact. This replaces global market clearing by bilateral feasibility (Riane, 14 Oct 2025).

Time enters in several distinct ways. Debt evolves with interest rates $\exp(\int r(s)\,ds)$2, production includes an explicit delay vector $\exp(\int r(s)\,ds)$3, and uncertainty is handled not by rational expectations but by a conditional mode,

$\exp(\int r(s)\,ds)$4

Intertemporal utility uses the discount factor

$\exp(\int r(s)\,ds)$5

The most literal ETI-like formula in the paper is the future-value relation

$\exp(\int r(s)\,ds)$6

which makes the integral of the interest rate the determinant of intertemporal value.

The framework is cast as a generalized game with feasibility correspondences and subjective pricing. Under continuity, compactness, convexity, and quasi-concavity conditions, the paper proves the existence of a Nash equilibrium, termed a transaction equilibrium. It also draws stronger welfare conclusions than standard general equilibrium theory: autarky is always an equilibrium; transaction equilibria need not be Pareto optimal; Pareto-optimal allocations need not be transaction equilibria; and both welfare theorems fail in this framework. Here the exchange-time structure is implicit rather than integral in the narrow measurement sense used in neutron scattering or open quantum heat transfer.

7. Comparative interpretation and recurring misconceptions

The surveyed literature supports a consistent negative conclusion: ETI is not a universally defined scientific object. The same three letters denote the Exposure Time Indicator model in SW-CRTs, while in other fields the closest counterparts are only ETI-like by analogy. This suggests that any use of “Exchange-Time Integral” requires immediate contextual qualification.

A second recurring issue concerns the difference between a summary statistic and a time-integrated dynamical quantity. In diffusion MRI, the FEXI-derived $\exp(\int r(s)\,ds)$7 compresses a multiexponential recovery into one number and depends strongly on the fitting window (Kiselev et al., 28 Jan 2026). In neutron scattering, by contrast, the measured quantity is literally a running integral whose derivative reconstructs the underlying dynamics (Benedetto et al., 2019). In open quantum systems, the relevant quantity is not a scalar exchange time at all, but a history-dependent functional encoded by $\exp(\int r(s)\,ds)$8 and $\exp(\int r(s)\,ds)$9 (Carrega et al., 2014).

A third misconception concerns immediacy. In stepped wedge trials, an IT model assumes the treatment effect is immediate and constant, whereas ETI allows delayed or time-varying effects (Kenny et al., 14 Mar 2025). In quantum dissipation, memory kernels and Matsubara terms make heat exchange explicitly nonlocal in time (Carrega et al., 2014). In decentralized equilibrium theory, intertemporal value depends on debt rollover, production delay, and mode-based anticipation rather than on instantaneous market clearing (Riane, 14 Oct 2025). Across these domains, the common theme is that exchange phenomena are often distributed over time, and methods that collapse them to a single parameter risk conflating observability, identifiability, and mechanism.

The broadest defensible synthesis is therefore limited and field-sensitive. ETI may refer to: a flexible exposure-time parametrization with an integral estimand in SW-CRTs; a directly measurable running time-integral of the van Hove function in neutron scattering; a path-integral heat-exchange functional in open quantum systems; or an implicit intertemporal valuation structure in decentralized economic exchange. It does not denote a single canonical mathematical object shared across these literatures.