WorksPhysics and Astronomy
Freezing transition and shock-like transients in ballistic random walks under threshold resetting
Abstract
We investigate the transient dynamics of a one-dimensional ballistic random walker under threshold resetting (TR). The walker starts at , moves towards a threshold at the origin with a velocity drawn from a distribution , and on reaching the threshold is reset to with a new sampled velocity. Whenever the mean first-passage time to the threshold is finite, the walker reaches a steady state that is uniform on and independent of . The relaxation towards it, in contrast, is highly sensitive to the behavior of . For a large class of -s, we find, rather remarkably, that the relaxation rate depends explicitly on the spatial coordinate, in contrast to conventional relaxation phenomena, where the rate is typically position-independent. Furthermore, this space-dependent relaxation rate eventually freezes to a constant, position-independent value beyond a critical position, a phenomenon we identify as freezing transition. Additionally, for a power-law velocity distribution, the system exhibits an anomalously slow, algebraic relaxation towards the steady state. We also showed that a sharp cutoff in generates a shock-like discontinuity in the particle propagator that travels towards the threshold, with a jump height that decays with time. These results demonstrate how velocity fluctuations, despite leaving the steady state unchanged, can profoundly influence the transient spatial structures and showcase rich non-equilibrium behavior. Our analytical results are corroborated by Monte Carlo simulations, showing an excellent agreement.
Introduction
Stochastic resetting, whereby an ongoing stochastic process is interrupted and returned to a prescribed configuration, has emerged as an important paradigm in statistical physics since its introduction in the context of diffusion [1, 2]. Despite its conceptual simplicity, resetting can profoundly alter the dynamics of a system, breaking detailed balance and generating non-equilibrium steady states (NESS) [1, 2, 3, 4, 5, 6, 7, 8], as well as rich transient phenomena, including anomalous relaxation [5, 9, 10]. Beyond its role in shaping non-equilibrium dynamics, resetting has proved to be a powerful strategy for optimizing first-passage processes, offering improvements in both search efficiency and reliability [1, 2, 5, 11, 12, 13, 14, 15, 16, 17, 18, 19]. The scope of resetting has since expanded well beyond its original setting in diffusion, finding applications in rather unrelated topics such as chemical reaction kinetics [20, 21, 22], biophysical search processes [23, 24, 25, 26], operations research [27, 28], and even in economics [29, 30, 31]. In parallel, resetting protocols have been realized experimentally using optical traps, enabling the investigation of a variety of non-equilibrium phenomena [32, 33, 34, 35, 36, 37, 38, 39], and have also been implemented using mechanical robots [40, 41]. For broader perspectives on the development of and emerging directions of the field, we refer the reader to the reviews [42, 43, 44, 45, 46].
In the conventional paradigm of stochastic resetting, reset events occur at random times determined by an external stochastic process (say, an external timer), independently of the instantaneous state of the system. However, many natural and engineered systems instead reset themselves when an internal variable crosses a critical value. An integrate-and-fire neuron discharges and returns to its resting potential once the membrane potential reaches a firing threshold [47]; a stop-loss order liquidates a position when a price limit is breached [48]; a loaded fibre bundle sheds a fibre once its rupture threshold is exceeded [49]; a cell divides once a critical amount of some internal variable has accumulated [50, 51]. This event-driven protocol, in which the reset is triggered by the system itself is referred to as threshold resetting (TR) [52, 53, 54], and is also known as first-passage resetting in [55, 56, 57]. Naturally, the reset events are, by construction, correlated with the trajectory that produces them.
Threshold resetting remains a relatively nascent area of research, with only a handful of studies addressing its effects in single- and many-particle systems. The existing literature has explored a few key aspects of this protocol across different settings. For instance, De Bruyne et al analyzed the steady state in a finite system and the reset-reward cost objective function of a single diffusing particle under TR [55, 56]. Biroli et al computed the exact steady-state joint position distribution for Brownian searchers with threshold resetting, showed evidence of dynamically emergent correlations in the steady state and, moreover, computed the exact statistics of several physical observables in this steady state analytically, including the extreme value statistics, the gap statistics and the full-counting statistics [57]. On the other hand, Biswas et al. carried out a first-passage study involving diffusing and ballistic particles subjected to TR and found non-trivial optimization behavior in the mean search time and further showed that optimal resetting can forestall larger cost [52, 53]. García-Valladares et al. studied first-passage-time optimization of a single one-dimensional Brownian searcher to distributed targets in the presence of both threshold resetting and Poissonian resetting [58]. Recent work by Chen and Bressloff on bounded two-dimensional domains with small interior targets has shown that the effectiveness of boundary resetting depends sensitively on the choice of reset position, the target configuration, and the waiting time at the boundary [59]. Frimpong et al. studied position distribution, first-passage- and occupation-time statistics of a sluggish random walker (discrete and continuous) in one dimension under threshold resetting [60], generalizing earlier works on sluggish walker with Poissonian resetting [61] and with uniform memory resetting [62].
In a preceding work, we investigated non-equilibrium dynamics of a general stochastic process subjected to TR [54]. We derived a general criterion for the emergence of a NESS, showing that whenever the mean first-passage time (MFPT) to the threshold is finite, the steady state is the ratio of the mean local time (MLT) to the MFPT. To illustrate this result, we analyzed a drift-diffusion process under TR, computed its NESS and further examined its transient behavior. The aim of the current work is to examine non-equilibrium dynamics of a different stochastic process namely a ballistic walker with randomized velocities. As illustrated schematically in Figure 1, the walker starts at and moves ballistically towards the threshold at the origin with a velocity drawn from a prescribed distribution . Upon reaching the threshold, the walker is instantaneously returned to its initial position, where a new velocity is independently sampled from the same distribution . A distinctive feature of this model is that the motion remains entirely deterministic between successive resetting events, with stochasticity entering solely through the random selection of velocities at the beginning of each excursion. We would like to understand how these velocity fluctuations lead to a non-equilibrium steady state and the relaxation mechanism towards stationarity.
Figure 1. Schematic illustration of a ballistic particle under threshold resetting (TR). The particle starts from an initial position and moves toward the threshold at the origin with a random velocity drawn from a velocity distribution . As soon as it reaches the threshold, it resets back to the initial position and renews its dynamics. Repeated resetting may lead to a steady-state distribution in the long-time limit. In this system, the particle always remains confined in the spatial regime .
We now summarize the key results of our work. Whenever the MFPT to the threshold is finite, TR drives the ballistic walker into a universal steady state, uniform on and independent of the velocity distribution , whereas the approach to it depends strongly on the low-velocity behavior of . Most strikingly, the relaxation rate itself can depend on position. This is unlike in most systems where the relaxation rate is a constant that does not depend on space variables. This space-dependent rate arises when as (). Intriguingly, the rate can undergo a freezing transition, i.e., it depends on position up to a critical point and then freezes onto a constant plateau, thus becoming space-independent. We derive a necessary
and sufficient condition on to observe this transition. In addition, a power law () as instead yields a slow relaxation. We also find that a sharp cutoff (lower or upper) in the support of generates a shock-like discontinuity in the particle propagator, travelling towards the threshold at speed . Its jump height at the discontinuity decays exactly as , so the shock directly measures the velocity distribution at its cutoff. In a strict deterministic limit , the system is found to attain a periodic steady state. Overall, velocity fluctuations leave rich and striking signatures in the transient dynamics while preserving the universality of the steady state.
For the readers’ convenience, we now provide the structure of the paper. Section II recalls the renewal framework for TR and the representation of the steady state as a ratio of the MLT to the MFPT. Section III introduces the ballistic model, derives its propagator in Laplace space, and establishes the universal uniform steady state. The transients are taken up next: the freezing transition in Sec. IV, the power-law relaxation in § 2.1, and the shock-like fronts in Sec. VI. Section VII treats the deterministic limit, and we conclude in Sec. VIII. Supporting derivations are collected in the Appendices.
General framework: threshold resetting
We consider a particle moving in one dimension that starts at and evolves under some intrinsic stochastic dynamics. Whenever the particle reaches the origin (which we designate as the threshold) it is instantaneously reset to its initial position , after which it resumes its dynamics anew. The central object of interest is the probability density of finding the particle at position at time starting from initial position .
The renewal structure of TR was introduced in Ref. [55] and was developed in Ref. [54] into a general characterization of the steady state for an arbitrary underlying dynamics. We recall below only those elements that are needed for the ballistic problem, and refer the reader to Ref. [54] for the detailed derivations.
The renewal structure
Because each reset erases the memory of the trajectory and restarts it from , TR is a renewal process, and the propagator obeys the renewal equation [54, 55, 57]
where is the propagator in the presence of an absorbing boundary at the origin (here, stands as an indicator that the particle stays in the positive half-line) and is the first-passage time (FPT) density to the threshold. The two terms have a transparent meaning. The first accounts for trajectories that have not reset at all in : the threshold acts as an absorbing boundary because a particle that reaches it stops and is reset, never continuing past. The second term accounts for trajectories that have reset at least once: the first reset occurs in with probability , after which the particle restarts from and propagates for the remaining time .
A Laplace transform, , converts the convolution in Eq. (1) into an algebraic relation
The two inputs are not independent. Integrating the constrained propagator over the final position gives the survival probability , and the FPT density follows as ; using , this yields [54]
Consequently, Eq. (2) becomes
Hence the full resetting dynamics is fixed entirely by the reset-free constraint propagator . For a wide class of underlying processes, this is known in closed form, which makes the propagator under TR analytically accessible.
Steady state under TR
If a steady-state (SS) exists, becomes time-independent at long times, i.e., . Consequently, its Laplace transform with respect to behaves for small as
This gives rise to the following relation (also known as the final value theorem) [54, 57]
When the mean first-passage time (MFPT) to the threshold is finite, equation (6) then collapses to the compact form derived in [54]
where the quantities in the above equation can be computed from the reset-free propagator
Here is the mean local time (MLT), i.e. the mean time the particle spends at position during a single excursion between two successive threshold crossings; it follows by averaging over the absorbing-boundary ensemble [54]. Eq. (7) therefore states that is the average time spent at in one cycle divided by the average duration of a cycle, a cycle being the interval between two threshold-driven first-passage events.
Two consequences of Eq. (7) are worth mentioning. First, Eqs. (8) and (9) imply , so that is automatically normalized to unity whenever it exists. Second, this ratio form constrains the admissible shape of the steady state: a finite forces the MLT either to be finite for every , giving a continuous , or to develop integrable singularities at isolated points, giving a steady state that diverges there while remaining normalizable [54]. The ballistic model studied below falls in the first class, and in fact realizes an interesting version of it: the MLT is not merely finite but independent of throughout the accessible region, which is the origin of the universal steady state obtained in Sec. III B.
A finite MFPT thus provides a sufficient criterion for the existence of a NESS in a noisy system. Interestingly, when the underlying dynamics is deterministic, successive renewal cycles never dephase, and the long-time state is strictly periodic rather than stationary even though is finite. We return to this case in Sec. VII; in Secs. III-VI the randomness of the velocity supplies the dephasing between cycles and Eq. (7) applies.
Ballistic Particle under TR: Model and universal steady state
We now specialize the underlying process to purely ballistic motion. A particle is released at and moves toward the origin at a constant velocity drawn from a probability density . When it reaches the threshold at the origin, it is reset to , and a new velocity is sampled from ; the cycle then repeats (see Figure 1). We take , i.e. every excursion is directed toward the threshold. Restricting to guarantees that each excursion terminates at the threshold. Otherwise, the particle escapes to infinity, TR fails to confine it, and no non-equilibrium steady state exists.
Other simple systems include many particles with simultaneous resetting after hitting the threshold or a single particle with a drift towards the threshold, ensuring finite MFPT and consequently a steady-state under TR, which are studied in Ref. [52, 53, 54, 57].
Reset-free propagator and first-passage density
During a single excursion starting from the initial position , the particle travels deterministically, i.e., until it hits the origin at time . The propagator with an absorbing boundary at the origin is therefore
where the Heaviside functions confine the particle to be between and . As the propagator under TR can be computed exactly in Laplace space following
(4), we need to compute the Laplace transform of the above quantity, which reads
The full expression of the propagator under TR in Laplace space immediately follows Eq. (4)
Now we are going to rewrite Eq. (12) with a change of variable , which will be useful in Sec. IV and § 2.1. After the change in variable, the Laplace space propagator can be re-expressed as
where is the Laplace transform of and is defined in terms of as follows
also . Consequently, and are singular at the same points in the complex- plane, although the nature of the singularity differs: a pole of of order is a pole of of order , and a branch point of one is a branch point of the other. This fact will be quite useful when we discuss the relaxation properties later.
Universal steady state: uniform distribution
By substituting the result from Eq. (10) into Eq. (8) and Eq. (9), we obtain the MFPT and MLT for a ballistic particle with a general velocity distribution ,
Both are controlled by the same integral , which converges only if is integrable near ; that is, must vanish sufficiently fast as . Physically, this is a statement about slow excursions: if there is a non-negligible chance of drawing a near-zero velocity, the corresponding excursion takes an anomalously long time; consequently, the MFPT to find the threshold diverges. Provided fast enough that the integral converges, the ratio in Eq. (7) is finite and, strikingly, independent of the shape of :
The stationary state is a universal uniform distribution on . This is a direct consequence of Eqs. (15),(16) and Eq. (7): for ballistic motion, the local time is the same at every point along an excursion, so the particle is, on average, equally likely to be found anywhere between the reset point and the threshold, regardless of how the speed is distributed.
The uniformity of the steady state admits a natural geometric interpretation through a change in geometry. By reversing the direction of motion and identifying the two endpoints of the interval, the original dynamics can be mapped onto a ballistic random walker moving around a circle of circumference . The angular velocity is independently drawn from a distribution after each completed revolution. During each cycle, the walker traverses the entire circle, spending a time in an angular interval . Consequently, the mean residence time per unit angle is , while the mean cycle duration is . The stationary angular density therefore follows from the ratio of the mean residence time per unit angle to the mean cycle duration
provided the mean cycle duration is finite and the renewal process admits a stationary state. This geometric construction is related in spirit to the idea of equidistribution, as exemplified in Weyl’s theorem [63, 64], and the result is a simple consequence of ergodicity via Weyl’s theorem. The geometric mapping thus provides an intuitive interpretation of the uniform steady state while the distribution of cycle durations can still govern the relaxation towards this universal stationary density.
Freezing transition
While the steady state is universal, the approach to it is not: the detailed shape of leaves a strong imprint on the transient dynamics. Depending on , we get different types of relaxation behaviors, e.g., exponential, power-law, etc. In this section, we discuss only exponential relaxations, specifically when the slowest relaxation rate depends on particle position . We find that there exists a freezing transition in the relaxation rate which, quite interestingly, changes from an independent to space-dependent form, governed by the properties of . Such a freezing transition was identified earlier in [65] in the escape rate of an overdamped diffusing particle from a trap.
An exact case
To expose the physics of the relaxation properly, we adopt the velocity distribution of a one-parameter family,
which is normalised and has a convergent for every , so the universal steady state Eq. (17) always holds. What distinguishes the members of the family is the singularity structure of the propagator in Laplace space for different values, and hence the relaxation. Substituting Eq. (19) into Eqs. (11) gives,
and, through Eq. (4),
The explicit finite-time propagator, obtained by inverse Laplace transform of Eq. (21), is given in Appendix A. The spatial behavior of the propagator for at different times is shown in Figure 3(c); non-monotonic relaxations of different spatial points to the steady state are shown in Figure 3(e).
Figure 3. A comprehensive plot for case, i.e., case (see appendix B and Sec. IV A) and its verification with simulation results. The freezing transition happens at point [Eq. (25)]. There exist two relaxation rates: by Eq. (26), one of which is space-dependent and the other is not. Panel (a) shows the space-independent relaxation for different spatial points , whereas Panel (b) shows the space-dependent relaxation for spatial points . We here defined . The dashed lines are exact theoretical results, and the markers are simulation results, where the propagator is generated through Monte Carlo simulations. Panel (c) shows the propagator behaviour with respect to at different time points , and how it reaches the steady state at long times. Panel (d) shows the relaxation rate as a function of space , which increases as until , then freezes at [Eq. (26)]. Panel (e) presents the non-monotonic relaxation of different spatial points [exact expression given in Eq. (78)]. Parameters: .
To study the steady state and relaxation behavior, it is quite important to analyze the singularity structure of the propagator in Eq. (21). Depending on whether is an integer or non-integer, this structure changes, giving rise to different relaxation behavior. Both cases are discussed below.
Integer case: Due to the pole present in the constrained propagator in Laplace space given in Eq. (20), we find a space-dependent -th order pole in the TR propagator in Eq. (21) at
which sweeps along the line depending on the position of the particle . The other denominator in , involving the Laplace-space first-passage (FP) density, gives rise to space-independent simple poles
with . These lie on a circle of radius in the complex- plane centered at that touches the origin; the pole at produces the steady state, while introduces complex-conjugate pairs and hence damped oscillatory relaxation. The positioning of the poles in the complex plane is shown in Figure 2.
Figure 2. The pole positions of and for [from Eq. (22) and (23)]. As we increase the value of , more poles from appear and lie on the circle centered at . There is always a pole at the origin, which is responsible for the steady state in the long-time limit. gives rise to complex conjugate poles. There also exists a pole at only for odd values.
The rightmost non-zero singularity dictates the slowest relaxation mode, and the rate is defined by the absolute real part of its position in the complex- plane. A freezing transition occurs when the space-dependent singularity [Eq. (22)] crosses the rightmost non-zero space-independent singularity [Eq. (23), ] at a spatial point . Equating their real parts gives the exact position of the crossing, namely the freezing point
which depends on but not on . Explicitly,
For there is no space-independent pole other than the one at the origin, so the relaxation rate is space-dependent everywhere, , and no freezing transition occurs. For the crossing has moved out of the domain , and the relaxation is space-independent everywhere. A genuine freezing transition with a position-dependent regime giving way to a plateau inside exists only for and . The behavior of the relaxation rate for these cases is as follows:
Figure 3(d) plots the behaviour of for , whereas the simulation verification of these relaxations for different spatial points is shown in Figure 3(a), (Figure 3b).
Non-integer case: We do not observe any freezing transition for non-integer values. The relaxation rate is space-independent for all . This is evident from the singularity structure of , given in Eq. (21). There exists
a space-independent branch point at ,
a space-dependent branch point at ,
poles arising from the roots of the denominator .
Notice that moves for varying , but is bounded on the right by as . There would always be a pole at , giving rise to a steady state, whereas other poles may or may not exist on the principal Riemann sheet depending on the specific value of [Appendix A § 2.6.2]. As the rightmost singularity (except the one) is always space-independent for any in this case, we do not observe any freezing transition in the relaxation rate .
General theory of the space-dependent rate and the freezing transition
To obtain the general criterion on the velocity distribution responsible for a freezing transition in the relaxation rate, we examine the rightmost singularity of in the complex- plane, other than , which is responsible for the steady state. For this discussion, we are going to use the expression in Eq. (13). A freezing transition occurs when the space-dependent singularity crosses the rightmost space-independent (non-zero) one at a spatial point .
Space-dependent singularity: Since , the rightmost singularity of its Laplace transform lies on the real axis [[66]; call it with . Consequently, its derivative also has a singularity there. Through , the singularity of the propagator by Eq. (13) is located at
This singularity lies on the real axis for any in the complex- plane. This singularity in makes the relaxation rate space-dependent. The singularity at must be a pole for a freezing transition to exist, as discussed in the next subsection.
Space-independent singularity: The denominator of Eq. (13) contributes in two distinct ways. Each of them contributes to the space-independent singularities.
(i) Zeros of : Normalization , guarantees the trivial root , which is the pole responsible for the steady state. The maximum-valued (real part) non-trivial root gives rise to the rightmost space-independent pole(s) in , whose position on the real axis of the complex- plane is as follows
where the root is defined as
This depends on the full velocity distribution .
(ii) The rightmost singularity of itself is at : This survives in only if it is a branch point. If it is a pole, then there and Eq. (13) develops a zero rather than a singularity.
If there exists a branch point at in , it bounds the space-dependent singularity on the right side of the complex- plane so that they can’t cross each other for any . Consequently, there does not exist any freezing transition. This is the same reason we did not find a freezing transition for the non-integer case in Sec. IV A. This indicates that the rightmost singularity of has to be a pole, then large- behavior of is as follows
where denotes the order of the pole ( is for a simple pole). Note: this condition fixes the nature of the space-dependent singularity discussed earlier as a pole of order , as it arises due to . Returning to the velocity variable , Eq. (30) sets a condition on the low-velocity behavior of ,
where . This is a hard constraint on in the small- limit, acting as a necessary condition for the existence of a freezing transition, but it is not sufficient. This property in ensures the existence of a space-dependent relaxation mode, as we have already seen for the example used in Sec. IV A for the integer- case – there, the representative case from Eq. (19) connects to Eq. (31) through the relation .
Condition for a freezing transition: A freezing transition occurs when crosses at the transition point , i.e.,
For , the space-dependent singularity dictates the relaxation; for , it shifts to the space-independent one . A genuine transition therefore lies inside the domain, , if and only if
where is set by the essential singularity in low- behavior of as shown in Eq. (31) and depends on the full distribution through Eq. (29). Eq. (31) and Eq. (33) together constitute the necessary and sufficient condition for the existence of a freezing transition in the ballistic threshold resetting model.
Power-law relaxation
The freezing transition analysis in Sec. IV assumed an essential singularity in the limit, giving rise to exponential relaxation. A slower approach to the steady state occurs when the low-velocity () behavior is a power law instead, such as
The first-passage (FP) time density inherits the low-velocity behavior of directly,
showing a heavy-tail behavior. For , the second and higher moments of the FP time density diverge, but the MFPT to find the threshold at the origin is finite, and the universal steady state of Eq. (17) exists. For , the first two moments of the FP density are finite, and so on (narrow density). We focus here only on the broad FP time density regime, i.e., .
The relaxation behavior can be studied using the small- (or large ) behavior of the Laplace space propagator . In this section, we are again going to use the form given in Eq. (13) in terms of the Laplace transform of . Recall that is the changed variable mentioned. The small- behavior in Eq. (34) translates into algebraic large- tail of the function
The corresponding Laplace space expansion of this function and its derivative in the small- limit reads
where and we have used the property of gamma function that . By substituting the above results into Eq. (13), we find the small- behavior of the propagator
where we have used the MFPT definition . The first term corresponds to the steady state , and the second term contributes to the power-law relaxation. After the Laplace inversion, the propagator in the long-time limit behaves as follows (Tauberian Theorem)
where we have defined . The amplitude satisfies , since and are separately normalized in the domain . Simulation verification of the above result is shown in Figure 4 for with . The physical reason behind this slow relaxation is the abundance of slow excursions with that populate arbitrarily long first-passage times (even though their mean stays finite), and it is this heavy supply of slow modes that governs the late stage of relaxation.
Figure 4. Power-law relaxation towards the steady state for with . Consequently, the mean FP time is for this velocity distribution. This plot shows how the spatial behavior of collapses to at long times, justifying Eq. (41). Markers denote the simulation results and solid lines denote theory obtained using numerical inverse Laplace transform of the exact Laplace space propagator given by Eq. (12). Parameters used: , .
Shock-like transients
A qualitatively different transient arises when the velocity distribution has sharp bounds, e.g., defined in . Such a cutoff produces moving spatial discontinuities, i.e., shock-like fronts in the propagator .
The mechanism is transparent from the reset-free propagator. Because of the presence of the cut-off , the particle can’t be in the region at time in the case when the particle hasn’t gone through any resetting, where is defined as follows
Note that this front propagates towards the origin with time and leaves the domain after . Since the positions with (which would require ) are unreachable in the reset-free case, the probability density drops abruptly to zero for the absorbing boundary propagator
where the jump height is defined as
This is the exact jump discontinuity we find in the propagator under TR. It can be explained simply by rewriting Eq. (1) in the last renewal approach [55]
where is the first-passage time distribution for reaching the threshold for the -th time (so that ). The jump discontinuity in at arises from the first term in Eq. (45), and the convolution term does not contribute to the jump as has a discontinuity at , but is continuous at . Consequently, the jump height in the TR propagator is the same as in the propagator,
The jump height depends only on the velocity distribution evaluated at the cutoff , and it decays with time. Simulation verification of these results is shown in Figure 5.
Figure 5. Shock-like transient in the propagator under threshold resetting for a velocity distribution with a lower cutoff, . is the mode of this distribution. For the density is discontinuous at the receding front , drawn as the vertical segment in each curve; the step has height . This jump height is verified with simulation in the inset log-log plot. For , there is no discontinuity in the propagator. Solid lines are obtained using numerical inverse Laplace transform of Eq. (12), and markers denote the Monte-Carlo simulation results. Parameters used: , , .
A similar discussion holds in the presence of an upper cutoff, i.e., is defined for . In this case, the jump height becomes , but now the discontinuity happens at position . The same goes for with both upper and lower cut-offs, i.e., . In this case, discontinuities occur at both and positions with the jump heights mentioned above.
Deterministic limit: Periodic steady state
It is also interesting to ask what happens when there is no randomness in the system, i.e., when the velocity distribution is just a Dirac delta function,
The particle always starts each excursion with the same velocity . The propagator under TR follows from Eq. (12)
Every pole of Eq. (48) lies on the imaginary -axis, so the final-value theorem can be applied to compute the steady state. By expanding the denominator as a geometric sum, we can compute the inverse Laplace transform. This gives a train of delta peaks
a single peak sweeping from to the origin at speed and reappearing at each time crosses a multiple of . The state is strictly periodic
with period equal to the deterministic MFPT. The purely imaginary poles in correspond to undamped oscillations in every observable; hence, they are also periodic.
Conclusion
In this work, we have investigated a ballistic random walker under threshold resetting (TR). The walker starts at and moves deterministically towards the threshold at the origin, with its speed independently redrawn from the distribution after each reset. The renewal structure of the process allows its dynamics under TR to be characterized entirely in terms of the corresponding reset-free quantities. Our analysis of the long-time propagator reveals an interesting universality. Whenever the velocity is random and the mean first-passage time (MFPT) to the threshold, , is finite, the system approaches a uniform nonequilibrium steady state (NESS) [Eq. (17)], irrespective of the form of . This universality follows from the mean local time (MLT), which is spatially uniform on [Eq. (16)], so the steady state, being the ratio of the MLT to the MFPT [Eq. (7)], is necessarily uniform.
The transients, on the other hand, depend strongly on the velocity distribution. Different features of map onto distinct singularity structures of the Laplace-space propagator , and hence onto qualitatively different relaxation behaviors. Most strikingly, the relaxation rate depends on the position , unlike in typical relaxation phenomena, where the rate is usually independent of spatial coordinate. Such a space-dependent rate arises when as with . Quite interestingly, the rate can undergo a transition and freezes onto a constant plateau beyond a critical point , giving rise to a freezing transition. Equations (31) and (33) together constitute a necessary and sufficient criterion for this transition – a representative case was then fully discussed in Sec. IV A. We also studied another case when vanishes as a power law when and showed that such fluctuations instead preclude any freezing transition and lead to a slow approach to the steady state. Intriguingly, a lower cutoff in the support of generates a shock that travels towards the threshold along . The discontinuity remains till time , and its height decays exactly as , and an upper cutoff also produces an analogous front. In the deterministic limit , the uniform steady state is replaced by a strictly periodic steady state.
Several promising directions emerge from the present study. For an -particle system, it is quite important to study various observables such as spatial correlations, order and gap statistics, and counting statistics in NESS. For a diffusive system under threshold resetting, these observables were extensively studied in Ref. [67]. It would be interesting to extend this study to particles with the ballistic dynamics studied here. A further direction is to relax the assumption of ballistic motion with a fixed velocity between successive resets and allow the velocity to change at random times. Depending on the statistics of these run times, this would connect threshold resetting to run-and-tumble dynamics [68] and Lévy walks [69], where reversals or changes in speed can drive the walker away from the threshold and raise new questions about the existence and properties of NESS. Beyond these theoretical extensions, our results suggest that the transient density may provide a means of probing the underlying velocity statistics: the shock structure and relaxation law encode information about the upper cutoff and the low-velocity behavior of , respectively. These observations are possibly realizable in controlled resetting experiments [32, 33, 34, 36, 40, 70]. More broadly, our findings can help connect threshold-resetting theory with threshold-driven physical and biological systems, including integrate-and-fire neuronal dynamics [47], and motivate new ways of extracting dynamical information from transient observations.
Acknowledgments
The numerical calculations reported in this work were carried out on the Kamet cluster, which is maintained and supported by the Institute of Mathematical Science’s High-Performance Computing Center. SNM acknowledges support from ANR Grant No. ANR23-CE30-0020-01 EDIPS. AP acknowledges research funding under the scheme ANRF/ARGM/2025/001623 from ANRF, India and research support from the Department of Science and Technology, India, SERB Start-up Research Grant Number SRG/2022/000080. SNM and AP also acknowledge the International Research Project (IRP) titled “Classical and quantum dynamics in out of equilibrium systems” by CNRS, France. RD and AP gratefully acknowledge research support from the Department of Atomic Energy, Government of India via the Apex project “Modelling Structure & Dynamics of Soft Matter”.
Finite-time propagator for the family
The simplest way to compute the inverse Laplace transform of Eq. (21) in Sec. IV A is to use Laplace inversion tables. After expressing in geometric sum, i.e., , we can rewrite as following
The following inverse Laplace transform is known [71]
where is the confluent hypergeometric function[72] (or Kummer’s function). By comparing the values for our function in Eq. (51), we find the inverse Laplace transform in a summation form
This result is valid for any , for both integer and non-integer cases. One can also compute this Laplace inversion using the Bromwich integral method.
where is the line of integration in the complex -plane such that all the singularities of remain on the left side of it (see Figure 6). Because the singularity structure depends on whether is an integer or non-integer, we discuss each case separately.
Figure 6. Bromwich contours used to invert , traversed counter-clockwise; lies to the right of all singularities. (a) Integer ( taken as an example)[Eq. (55)]: the simple poles (filled) lie on the dotted circle , with (red) giving the steady state, and (blue) is a space-dependent pole of order . We discuss the exact positioning of these poles in more detail in Sec. IV A for integer . (b) Non-integer ( taken as an example)[Eq. (58)]: becomes a branch point with a cut to (shaded) on which lies; only the poles with survive on the principal sheet and contribute to the Bromwich integral. These singularities are discussed in Sec. IV A for non-integer in the main text and Sec. A 2 in the appendix.
Integer n
The integrand has an -th-order space-dependent pole at and simple space-independent poles at . Exact expressions are given in Eq. (22) and Eq. (23) in the main text. Using the Cauchy integral theorem, we can write
where the contour lines are shown in Figure 6(a). The contribution due to vanishes by Jordan’s Lemma. After computing all the residues, we find the full time-dependent expression of the propagator
where is the -th derivative of with respect to , and is defined as follows
The steady-state contribution comes from the pole . Both () and have a negative real part; they give rise to exponentially decaying relaxation towards the steady state.
Non-integer n
For non-integer cases, the multi-valued terms present in the propagator give rise to branch points instead of full set of poles: a space-independent branch point at and a space-dependent one at , with for all . The principal branch point is always bound from right, so the long-time decay is space-independent and no freezing transition survives for non-integer . Poles from appear on the principal sheet only when (none for , one conjugate pair for , and so on). Using the Cauchy integral theorem, we write
where is the greatest integer less or equal to and the contours are shown in Figure 6(b). The contribution due to contour line vanishes[73]. contribution arises from the branch cut from to .
where is chosen on the RHS simply for convenience. After computing the residues and the branch point contribution, we can express as follows
where the steady state arises due to the pole at (or the contribution of the residue in Eq. (58)) and is given by the residues of the rest of the poles other than the one at existing on the principal sheet. The exact expressions of these quantities are given below.
where, for the simplicity of the expression, we have defined
The branch-cut contribution can be computed using , where denotes the upper () and lower () branch-cut lines and .
The value of the Laplace space propagator on the integration lines is as follows
Due to another second branch point at , we find a critical point defined as
There are two regimes and , due to the term present in the denominator in Eq. (66) as it produces an extra phase factor for , which makes the discontinuities different between these regimes. For simplicity of notation, let us define
By using the above-mentioned variables, we can express the different discontinuities in both regions as below
After doing a few steps, we find the exact expression of the above quantities,
Using the above results, the final expression of the branch-cut integral reads
Exactly at , the above integral diverges for any . We can overcome this problem by computing the finite part integral, which is nothing but computing the above integral without considering the point . For a detailed discussion and comparison with the Cauchy principal value integrals, see Ref. [74]. A finite part integral can be defined in this way
where is a differentiable function at . The RHS integral is finite. In our case, can be any non-integer value, which may require multiple derivatives with respect to ; e.g., the above operation is applicable when , where we can directly compare with . For an arbitrary , we define and . Using these, we can compute the finite part integral of Eq. (74) as below
where the function is times differentiable at . Figure 7 shows a verification plot of the different analytical results for the inverse Laplace transform for non-integer values, along with simulation.

Figure 7. Finite-time propagator for non-integer with given in Eq. (19): (a) , (b) , (c) , (d) , at , , (blue, orange, green). Solid lines are the sum using Laplace inversion tables [Eq. (53)], circles the Bromwich-integral result [Eq. (60)], crosses a direct numerical Laplace inversion of Eq. (21), and squares Monte Carlo simulations. The three analytic and simulation results agree well. Parameters: , .
Exactly solvable example:
To expose the physics of the exponential relaxation and freezing transition, we adopt the velocity distribution
where is a characteristic speed. Due to the presence of an essential singularity in the small- limit, the MFPT to find the threshold is finite, and consequently there exists a steady state. After carrying out the Laplace inversion using the Bromwich integral method mentioned in Appendix A, the propagator under TR can be expressed as follows
with the scale function defined as
where
Although Eq. (79) appears singular at , the singularities cancel, and the limit is finite,
so the density is smooth throughout . As , at every , recovering the universal uniform steady state (17). At any fixed , the relaxation of is non-monotonic in time—the density overshoots and undershoots its stationary value before settling—reflecting the interplay of the several decay modes present in Eq. (79). These non-monotonic relaxations of different spatial points are shown in Figure 3(e).
Freezing transition in the relaxation rate
At long times, the slowest decay mode dominates, and the density relaxes as
where is the steady state, is algebraic in and is the relaxation rate. Reading off the dominant exponential in Eq. (79), we find
The relaxation is thus spatially heterogeneous. In one part it is space-dependent, and in the rest it is space-independent. The crossover point , where the rate stops depending on position and freezes to a plateau value, is called a freezing transition—of the same type identified in the barrier crossing rate of a diffusing particle [65]. Its mathematical origin lies in the competition between two families of singularities of , which we analyze in full generality in Sec. IV in the main text.
Simulation verification in Figure 3(a), (b)
For verification purposes in Figure 3, we represented the propagator as
where and can be defined using the dimensionless coefficients and as below
References
References
- [1]Martin R. Evans and Satya N. Majumdar. Diffusion with Stochastic Resetting. Physical Review Letters, 106:160601, 2011.DOI
- [2]Martin R Evans and Satya N Majumdar. Diffusion with optimal resetting. Journal of Physics A: Mathematical and Theoretical, 44(43):435001, 2011.DOI
- [3]Shamik Gupta, Satya N Majumdar, and Gregory Schehr. Fluctuating interfaces subject to stochastic resetting. Physical Review Letters, 112(22):220601, 2014.pubmed.ncbi.nlm.nih.gov/24949748
- [4]Arnab Pal. Diffusion in a potential landscape with stochastic resetting. Physical Review E, 91(1):012113, 2015.pubmed.ncbi.nlm.nih.gov/25679576
- [5]Arnab Pal, Anupam Kundu, and Martin R Evans. Diffusion under time-dependent resetting. Journal of Physics A: Mathematical and Theoretical, 49:225001, 2016.DOI
- [6]Stephan Eule and Jakob J Metzger. Non-equilibrium steady states of stochastic processes with intermittent resetting. New Journal of Physics, 18(3):033006, 2016.
- [7]Apoorva Nagar and Shamik Gupta. Diffusion with stochastic resetting at power-law times. Physical Review E, 93(6):060102, 2016.DOI
- [8]Gabriele de Mauro, Marco Biroli, Satya N Majumdar, and Gregory Schehr. Dynamically emergent correlations in brownian particles subject to simultaneous non-poissonian resetting protocols. Physical Review E, 113(1):014120, 2026.DOI
- [9]Satya N Majumdar, Sanjib Sabhapandit, and Grégory Schehr. Dynamical transition in the temporal relaxation of stochastic processes under resetting. Physical Review E, 91(5):052131, 2015.DOI
- [10]Rishu Kumar Singh, Ralf Metzler, and Trifce Sandev. Resetting dynamics in a confining potential. Journal of Physics A: Mathematical and Theoretical, 53(50):505003, 2020.DOI
- [11]Martin R Evans, Satya N Majumdar, and Kirone Mallick. Optimal diffusive search: nonequilibrium resetting versus equilibrium dynamics. Journal of Physics A: Mathematical and Theoretical, 46(18):185001, 2013.DOI
- [12]Arnab Pal and Shlomi Reuveni. First Passage under Restart. Physical Review Letters, 118:030603, 2017.DOI
- [13]Shlomi Reuveni. Optimal Stochastic Restart Renders Fluctuations in First Passage Times Universal. Physi-DOI
- [14]Lukasz Kusmierz, Satya N Majumdar, Sanjib Sabhapandit, and Grégory Schehr. First order transition for the optimal search time of Lévy flights with resetting. Physical Review Letters, 113(22):220602, 2014.
- [15]Priyo Shankar Pal, Arnab Pal, Hyunggyu Park, and Jae Sung Lee. Thermodynamic trade-off relation for first passage time in resetting process. Physical Review E, 108(4):044117, 2023.arxiv.org/abs/2305.04562
- [16]Arnab Pal and V. V. Prasad. Landau-like expansion for phase transitions in stochastic resetting. Physical Review Research, 1(3):032001, 2019.DOI
- [17]Martin R Evans and Satya N Majumdar. Effects of refractory period on stochastic resetting. Journal of Physics A: Mathematical and Theoretical, 52(1):01LT01, 2018.DOI
- [18]Arnab Pal, Iddo Eliazar, and Shlomi Reuveni. First passage under restart with branching. Physical Review Letters, 122(2):020602, 2019.DOI
- [19]Gourab Kumar Sar, Arnob Ray, Dibakar Ghosh, Chittaranjan Hens, and Arnab Pal. Resetting-mediated navigation of an active brownian searcher in a homogeneous topography. Soft Matter, 19(24):4502–4518, 2023.DOI
- [20]Shlomi Reuveni, Michael Urbakh, and Joseph Klafter. Role of substrate unbinding in michaelis–menten enzymatic reactions. Proceedings of the National Academy of Sciences, 111(12):4391–4396, 2014.DOI
- [21]Tal Robin, Shlomi Reuveni, and Michael Urbahk. Single-molecule theory of enzymatic inhibition. Nature Communications, 9(1):1–9, 2018.DOI
- [22]Arup Biswas, Arnab Pal, Debasish Mondal, and Somrita Ray. Rate enhancement of gated drift-diffusion process by optimal resetting. The Journal of Chemical Physics, 159(5), 2023.DOI
- [23]Édgar Roldán, Ana Lisica, Daniel Sánchez-Taltavull, and Stephan W Grill. Stochastic resetting in backtrack recovery by rna polymerases. Physical Review E, 93(6):062411, 2016.DOI
- [24]Srikanth Budnar, Kabir B Husain, Guillermo A Gomez, Maedeh Naghibosadat, Amrita Varma, Suzie Verma, Nicholas A Hamilton, Richard G Morriss, and Alpha S Yap. Anillin promotes cell contractility by cyclic re- setting of rhoa residence kinetics. Developmental Cell, 49(6):894–906, 2019.
- [25]Jitin Rajoria and Arnab Pal. Broad distributions of sliding times are fingerprints of efficient target search on dna. bioRxiv, pages 2026–03, 2026.DOI
- [26]Jane Kondev, Marc Kirschner, Hernan G Garcia, Gabriel L Salmon, and Rob Phillips. Biological processes as exploratory dynamics. Biophysical journal, 125(12):2812–2837, 2026.arxiv.org/abs/2506.04104
- [27]Ofek Lauber Bonomo, Arnab Pal, and Shlomi Reuveni. Mitigating long queues and waiting times with service resetting. PNAS Nexus, 1(3), 2022.DOI
- [28]Reshmi Roy, Arup Biswas, and Arnab Pal. Queues with resetting: a perspective. Journal of Physics: Complexity, 5(2):021001, 2024.DOI
- [29]Viktor Stojkoski, Trifce Sandev, Ljupco Kocarev, and Arnab Pal. Geometric brownian motion under stochastic resetting: a stationary yet nonergodic process. Physical Review E, 104(1):014121, 2021.DOI
- [30]Petar Jolakoski, Arnab Pal, Trifce Sandev, Ljupco Kocarev, Ralf Metzler, and Viktor Stojkoski. A first passage under resetting approach to income dynamics. Chaos, Solitons & Fractals, 175:113921, 2023.DOI
- [31]Viktor Stojkoski, Petar Jolakoski, Arnab Pal, Trifce Sandev, Ljupco Kocarev, and Ralf Metzler. Income inequality and mobility in geometric brownian motion with stochastic resetting: theoretical results and empirical evidence of non-ergodicity. Philosophical Transactions of the Royal Society A, 380(2224):20210157, 2022.DOI
- [32]Ofir Tal-Friedman, Arnab Pal, Amandeep Sekhon, Shlomi Reuveni, and Yael Roichman. Experimental realization of diffusion with stochastic resetting. The Journal of Physical Chemistry Letters, 11(17):7350–7355, 2020.
- [33]Benjamin Besga, Alfred Bోవon, Artyom Petrosyan, Satya N Majumdar, and Sergio Ciliberto. Optimal mean first-passage time for a brownian searcher subjected to resetting: experimental and theoretical results. Physical Review Research, 2(3):032029, 2020.DOI
- [34]Felix Faisant, Benjamin Besga, Artyom Petrosyan, Sergio Ciliberto, and Satya N Majumdar. Optimal mean first-passage time of a brownian searcher with resetting in one and two dimensions: experiments, theory and numerical tests. Journal of Statistical Mechanics: Theory and Experiment, 2021(11):113203, 2021.DOI
- [35]Rémi Goerlich, Minghao Li, Luís Barbosa Pires, Paul-Antoine Hervieux, Giovanni Manfredi, and Cyriaque Genet. Experimental test of landauer’s principle for stochastic resetting. arXiv preprint arXiv:2306.09503, 2023.
- [36]Sandip Kundu, Dibyendu Mondal, Arup Biswas, Arnab Pal, and Manas Khan. Emulating microbial run-and-tumble and tactic motion by stochastically reorienting synthetic active brownian particles. arXiv preprint arXiv:2509.21903, to appear in npj Soft Matter, 2026.arxiv.org/abs/2509.21903
- [37]Marco Biroli, Sergio Ciliberto, Manas Kulkarni, Satya N Majumdar, Artyom Petrosyan, and Grégory Schehr. Experimental evidence for strong emergent correlations between particles in a switching trap. Physical Review Letters, 137(3):037102, 2026.DOI
- [38]Rémi Goerlich, Kristian Stølevik Olsen, Hartmut Löwen, and Yael Roichman. Time-energy trade-off in stochastic resetting using optimal control. Physical Review E, 113(1):014103, 2026.DOI
- [39]Nir Sherf, Rémi Goerlich, Barak Hirshberg, and Yael Roichman. Stochastic resetting versus thermal equilibration: Faster relaxation, different destination. Physical Review Research, 8(2):L022028, 2026.arxiv.org/abs/2510.26629
- [40]Somnath Paramanick, Arup Biswas, Harsh Soni, Arnab Pal, and Nitin Kumar. Uncovering universal characteristics of homing paths using foraging robots. PRX Life, 2(3):033007, 2024.DOI
- [41]Somnath Paramanick, Arnab Pal, Harsh Soni, and Nitin Kumar. Programming tunable active dynamics in a self-propelled robot. The European Physical Journal E, 47(5):34, 2024.DOI
- [42]Martin R. Evans, Satya N. Majumdar, and Grégory Schehr. Stochastic resetting and applications. Journal of Physics A: Mathematical and Theoretical, 53:193001, 2020.DOI
- [43]Arnab Pal, Sarah Kostinski, and Shlomi Reuveni. The inspection paradox in stochastic resetting. Journal of Physics A: Mathematical and Theoretical, 55(2):021001, 2022.DOI
- [44]Shamik Gupta and Arun M Jayannavar. Stochastic resetting: A (very) brief review. Frontiers in Physics, page 130, 2022.DOI
- [45]Arnab Pal, Viktor Stojkoski, and Trifce Sandev. Random resetting in search problems. In Target Search Problems, pages 323–355. Springer, 2024.DOI
- [46]Apoorva Nagar and Shamik Gupta. Stochastic resetting in interacting particle systems: A review. Journal of Physics A: Mathematical and Theoretical, 2023.DOI
- [47]Anthony N Burkitt. A review of the integrate-and-fire neuron model: I. homogeneous synaptic input. Biological cybernetics, 95:1–19, 2006.pubmed.ncbi.nlm.nih.gov/16622699
- [48]Albert N Shiryaev. Optimal stopping rules, volume 8. Springer Science & Business Media, 2007.
- [49]Alex Hansen, Per Christian Hemmer, and Srutarshi Pradhan. The fiber bundle model: modeling failure in materials. John Wiley & Sons, 2015.
- [50]Mattia Corgliano, Kuheli Biswas, Matteo Bocchiola, Daniele Montagnani, Ariel Amir, and Marco Cosentino Lagomarsino. Essential role of extrinsic noise in models of e. coli division control. arXiv preprint arXiv:2605.07007, 2026.
- [51]Aanjaneya Kumar and James Holehouse. Branching under first-passage resetting. arXiv preprint arXiv:2605.16693, 2026.arxiv.org/abs/2605.16693
- [52]Arup Biswas, Satya N. Majumdar, and Arnab Pal. Target search optimization by threshold resetting. Phys. Rev. Lett., 135:227101, Nov 2025.DOI
- [53]Arup Biswas, Satya N Majumdar, and Arnab Pal. Optimal threshold resetting in collective diffusive search. Journal of Physics A: Mathematical and Theoretical, 59(38):385002, 2026.DOI
- [54]Rahul Das, Satya N Majumdar, and Arnab Pal. Nonequilibrium dynamics of drift-diffusion process under threshold resetting. arXiv preprint arXiv:2609.24438, 2026.arxiv.org/abs/2609.24438
- [55]B De Bruyne, J Randon-Furling, and S Redner. Optimization in first-passage resetting. Physical Review Letters, 125(5):050602, 2020.DOI
- [56]B De Bruyne and S Redner. Optimization and growth in first-passage resetting. Journal of Statistical Mechanics: Theory and Experiment, 2021(1):013203, 2021.DOI
- [57]Marco Biroli, Satya N Majumdar, and Grégory Schehr. First-passage resetting gas. Europhysics Letters, 153(3):31002, 2026.
- [58]Gregorio García-Valladares, Antonio Prados, Alessandro Manacorda, and Carlos A Plata. How target distributions shape optimal stochastic resetting. Physical Review E, 113(2):024107, 2026.
- [59]Kevin Chen and Paul C Bressoff. 2d target search with boundary-induced resetting. arXiv preprint arXiv:2609.31085, 2026.arxiv.org/abs/2609.31085
- [60]Theophilus Dwamena Frimpong, Léa Bresque, and Édgar Roldán. Sluggish random walks with resetting: nonequilibrium aspects. arXiv preprint arXiv:2610.21962, 2026.arxiv.org/abs/2610.21962
- [61]Giuseppe Del Vecchio Del Vecchio, Manas Kulkarni, Satya N Majumdar, and Sanjib Sabhapandit. Proxidaxis: An adaptive search strategy based on proximity and stochastic resetting. Physical Review E, 113(4):L042101, 2026.DOI
- [62]Denis Boyer and Satya N Majumdar. Ultra slow sub-logarithmic diffusion of a sluggish random walker subject to resetting with memory. Journal of Statistical Mechanics: Theory and Experiment, 2026(6):063201, 2026.DOI
- [63]Hermann Weyl. Über die gleichverteilung von zahlen mod eins. Mathematische Annalen, 77:313–352, 1916.
- [64]Lauwerens Kuipers and Harald Niederreiter. Uniform Distribution of Sequences. Wiley-Interscience, New York, 1974.
- [65]Sanjib Sabhapandit and Satya N Majumdar. Freezing transition in the barrier crossing rate of a diffusing particle. Physical Review Letters, 125(20):200601, 2020.DOI
- [66]David V. Widder. The Laplace Transform, volume 6 of Princeton Mathematical Series. Princeton University Press, Princeton, NJ, 1941. Ch. II, Thm. 5b (page 58).
- [67]Marco Biroli, Hernán Larralde, Satya N Majumdar, and Grégory Schehr. Exact extreme, order, and sum statistics in a class of strongly correlated systems. Physical Review E, 109(1):014101, 2024.DOI
- [68]Martin R Evans and Satya N Majumdar. Run and tumble particle under resetting: a renewal approach. Journal of Physics A: Mathematical and Theoretical, 51(47):475003, 2018.DOI
- [69]Vasily Zaburdaev, Sergey Denisov, and Joseph Klafter. Lévy walks. Reviews of Modern Physics, 87(2):483–530, 2015.
- [70]Amy Altshuler, Ofek Lauber Bonomo, Nicole Gorohovsky, Shany Marchini, Eran Rosen, Ofir Tal-Friedman, Shlomi Reuveni, and Yael Roichman. Environmental memory facilitates search with home returns. Physical Review Research, 6(2):023255, 2024.arxiv.org/abs/2306.12126
- [71]Harry Bateman and Bateman Manuscript Project. Tables of Integral Transforms. McGraw-Hill Book Company, 1954.
- [72]Milton Abramowitz and Irene A Stegun. Handbook of mathematical functions: with formulas, graphs, and mathematical tables, volume 55. National Bureau of Standards Washington, DC, 2006.
- [73]George B Arfken, Hans J Weber, and Frank E Harris. Mathematical methods for physicists: a comprehensive guide. Academic press, 2011.
- [74]Giovanni Monegato. Definitions, properties and applications of finite-part integrals. Journal of computational and applied mathematics, 229(2):425–439, 2009.





