betaFeedback

WorksMathematics

Steady-state of Lévy process with general resetting: unreflected and reflected cases

Abstract

We study (unreflected and reflected) Lévy processes that are modified at independent Poisson review epochs by a random state-dependent reset map. The first model is unreflected: between resettings the level evolves as a Lévy process, and at a resetting epoch the new level is FF of the level seen just before the resetting. The second model is the reflected Lévy process at 00, but with reset maps chosen from a wider transform-solvable class. In this work, we give contraction-type ergodicity criteria for the existence of the steady-state. We identify this steady-state of the system as the solution of certain stochastic fixed-point equation. This allows to find the Laplace-transform, moments recursions, and regularly-varying-tail formulae. For non-reflected case, the examples include affine deterministic resets F(x)=cx+bF(x)=cx+b, random-coefficient resets F(x)=Cx+bF(x)=Cx+b with two-point CC for the finite-activity Lévy process, spectrally positive Lévy flight, reflected Brownian process. For the reflected cased we analyse affine resetting mechanism F(x)=cxF(x)=cx, the alternating resetting F(x)=(B−x)+F(x)=(B-x)^+, and uniform proportional resetting F(x)=CxF(x)=Cx with CC having uniform distribution.

Introduction

There has been a vast amount of papers related to the resetting mechanism. This mechanisms arise in search processes (see, e.g., [2, 9, 16]), physics (see, e.g., [6]), biology (see, e.g., [15]), queueing theory (see, e.g., [12]) and in many other research fields; see also the survey [11]. The usual resetting mechanism is based either by sending the process to 0 (or any other fixed state) or by multiplying by possibly random amplitude at independent Poisson epochs. Lately, there have been efforts to consider more general resetting mechanisms, see for example [1] and references therein. Based on [14], we also consider more general possibly random resetting mechanisms F(x)F(x), where xx is state prior resetting, that is applied to Lévy processes XX and its reflected version YY at 0.

For unreflected Lévy process XX, we give a contraction ergodicity criterion for the resetting mechanism to get steady-state of our system. For the stationary law we give a general transform equation and moment recursions for arbitrary reset maps FF whenever powers of F(Z)F(Z) can be expressed through the moments of ZZ. For the resetting mechanism F(x)=cx+bF(x)=cx+b (where c,bc,b are constants), using a perpetuity representation, we obtain an infinite-product transform of the stationary law. We also identify the moment recursions. Finally, we also present the behaviour of the asymptotic tail of the law of this stationary distribution under regular variation of the positive jumps. We managed also to analyse another interesting case when F(x)=Cx+bF(x)=Cx+b with two-point and three-point distribution of CC.

For the reflection at 0 case we consider Lévy process without negative jumps XX. Such process is called spectrally positive. For spectrally positive Lévy process YY reflected at 0 we also give sufficient conditions for the existence of the steady-state, and wrire an appropriate recursion identifying the limiting law. We analyse in detail the reſtting maps F(x)=cxF(x)=cx, F(x)=(B−x)+F(x)=(B-x)^{+} with exponential BB, and F(x)=CxF(x)=Cx with CC having uniform distribution U(0,1)U(0,1). We also include the tail of distribution of the steady-state in the case when jumps have regularly varying jumps.

The present paper has two aims. Firstly, for very general process between resettings and quite general resetting mechanisms, we give the ergodicity conditions and show that the limiting stationary distributions satisfies a one-step recursive equation. This equations codes some recursive procedures to find moments of this stationary distribution and allows to identify the tail behaviour when jumps of the process between resettings is regularly varying. The second aim is to analyse the process reflected at 0 and then reset it. We show that, to some extend, the same ideas as for the process without reflection will still work.

The paper is organised as follows. In the next section we analyse the case without reflection and then we focus on the reflected case. Throughout this paper random reset coefficients are denoted by capital letters, such as CC; deterministic reset constants are denoted by lower-case letters, such as cc and bb.

The unreflected reset model

Notations

For a real-valued Lévy process XX, write

ψ(s):=log⁡⟨e−sX(1)⟩,\psi(s) := \log\langle e^{-sX(1)} \rangle,

whenever the expectation is finite. Then, for an independent exponential random variable eq∼Exp⁡(q)e_q \sim\operatorname{Exp}(q) with intensity qq, we denote

m(s):=⟨e−sX(eq)⟩=∫0∞qe−qtetψ(s) dt=qq−ψ(s),(1)m(s) := \langle e^{-sX(e_q)} \rangle= \int_0^\infty qe^{-qt}e^{t\psi(s)}\,\mathrm{d}t = \frac{q}{q-\psi(s)}, \tag*{(1)}

for ss in the domain

Dq:={s:Re⁡(q−ψ(s))>0}.D_q := \{s : \operatorname{Re}(q-\psi(s))>0\}.

In the reflected spectrally positive part of the paper, we use the standard fluctuation notation

X‾(t):=sup⁡0≤u≤tX(u),X‾(t):=inf⁡0≤u≤tX(u).\overline{X}(t) := \sup_{0\leq u\leq t} X(u), \qquad\underline{X}(t) := \inf_{0\leq u\leq t} X(u).

Let Φ(q)\Phi(q) be the largest non-negative solution of

ψ(Φ(q))=q.(2)\psi(\Phi(q))=q. \tag*{(2)}

For a spectrally positive Lévy process, the Wiener–Hopf factorisation gives that −X‾(eq)-\underline{X}(e_q) has the exponential law with intensity Φ(q)\Phi(q) and

ℓq(s):=⟨e−sX‾(eq)⟩=q(s−Φ(q))Φ(q)(ψ(s)−q).(3)\ell_q(s) := \langle e^{-s\underline{X}(e_q)} \rangle= \frac{q(s-\Phi(q))}{\Phi(q)(\psi(s)-q)}. \tag*{(3)}

The expression has the usual removable singularity at s=Φ(q)s=\Phi(q).

Embedded chain

The unreflected process with resetting evolves as follows. Immediately after a resetting Poissonian epoch TnT_n with intensity qq, the level is reset to FF of the level just before that epoch. During the following inter-review interval, it evolves by adding an independent Lévy increment. We denote this Lévy process with resetting by Z(t)Z(t). Let

Zn:=Z(Tn−),Z_n := Z(T_n-),

be the pre-resetting level. Then

Zn+1=Fn+1(Zn)+ξn+1,ξn+1=dX(eq),(4)Z_{n+1}=F_{n+1}(Z_n)+\xi_{n+1}, \qquad\xi_{n+1}\mathrel{\overset{d}{=}}X(e_q), \tag*{(4)}

with ξn+1\xi_{n+1} independent of Fn+1F_{n+1} and ZnZ_n. It is so-called one step analysis based on observing the process at resetting times. Assume that stationary law exists and is unique. Then, at stationarity,

Z=dF(Z)+ξ,ξ=dX(eq),(5)Z\mathrel{\overset{d}{=}}F(Z)+\xi, \qquad\xi\mathrel{\overset{d}{=}}X(e_q), \tag*{(5)}

where ZZ is a steady-state and it is independent of FF and ξ\xi on the right-hand side. This identity is a direct consequence of so-called PASTA property.

Ergodicity conditions

We start from identifying the sufficient ergodicity conditions for the existence of the stationary distribution. that follows directly from the iterated-random-functions theorem.

Theorem 2.1 (Unreflected contraction criterion). Assume that there are i.i.d. nonnegative random variables AnA_n such that

∣Fn(x)−Fn(y)∣≤An∣x−y∣for all x,y,\lvert F_n(x)-F_n(y)\rvert\le A_n\lvert x-y\rvert\qquad\text{for all }x,y,

and that, for some x0x_0,

⟨log⁡A1⟩<0,⟨log⁡+∣F1(x0)+ξ1∣⟩<∞.(6)\langle\log A_1\rangle< 0,\qquad\langle\log^{+}\lvert F_1(x_0)+\xi_1\rvert\rangle< \infty. \tag*{(6)}

Then the chain (4) has a unique stationary distribution. Moreover, for any initial state Z0Z_0, ZnZ_n converges in distribution to the unique solution of (5).

Proof. Let Gn(x)=Fn(x)+ξnG_n(x)=F_n(x)+\xi_n. Then

∣Gn(x)−Gn(y)∣=∣Fn(x)−Fn(y)∣≤An∣x−y∣.\lvert G_n(x)-G_n(y)\rvert=\lvert F_n(x)-F_n(y)\rvert\le A_n\lvert x-y\rvert.

Thus GnG_n is a random Lipschitz map with negative mean logarithmic Lipschitz constant. The second condition in (6) is the usual integrability condition. The standard backward iteration

G1∘G2∘⋯∘Gn(x0)G_1\circ G_2\circ\cdots\circ G_n(x_0)

converges almost surely to a finite limit that is independent of x0x_0, and the forward chain converges weakly to the law of that limit. □

For affine reset maps, the condition is straightforward: the contraction depends only on the multiplicative coefficient, while the additive part only enters through the integrability condition.

Corollary 2.2 (Affine and random-affine resets). Let

Fn(x)=Cnx+Bn,F_n(x)=C_nx+B_n,

where (Cn,Bn)(C_n,B_n) are i.i.d. and independent of ξn\xi_n. If

⟨log⁡∣C1∣⟩<0,⟨log⁡+∣B1+ξ1∣⟩<∞,\langle\log\lvert C_1\rvert\rangle<0,\qquad\langle\log^{+}\lvert B_1+\xi_1\rvert\rangle<\infty,

then the unreflected reset chain has a unique stationary distribution. In particular, if F(x)=cx+bF(x)=cx+b with ∣c∣<1\lvert c\rvert<1 and Elog⁡+∣b+X(eq)∣<∞\mathbb{E}\log^{+}\lvert b+X_{\mathrm{(eq)}}\rvert<\infty, then the stationary distribution exists and is unique.

Laplace transform of steady-state and moments recursion

Let

v(s):=⟨e−sZ⟩,hF(s):=⟨e−sF(Z)⟩.v(s):=\langle e^{-sZ}\rangle,\qquad h_F(s):=\langle e^{-sF(Z)}\rangle.

By (5) and independence,

v(s)=m(s)hF(s)=qq−ψ(s)⟨e−sF(Z)⟩,s∈Dq.(7)v(s)=m(s)h_F(s)=\frac{q}{q-\psi(s)}\langle e^{-sF(Z)}\rangle,\qquad s\in D_q. \tag*{(7)}

Assume that the necessary moments exist. Define

νn:=⟨ξn⟩=⟨X(eq)n⟩,mn:=⟨Zn⟩,ηn:=⟨F(Z)n⟩.\nu_n:=\langle\xi^n\rangle=\langle X_{\mathrm{(eq)}}^n\rangle,\qquad m_n:=\langle Z^n\rangle,\qquad\eta_n:=\langle F(Z)^n\rangle.

Expanding the fixed-point equation Z=dξ+F(Z)Z\mathrel{\mathop{=}\limits^{d}}\xi+F(Z) gives

mn=∑k=0n(nk)νn−kηk,n≥0,(8)m_n=\sum_{k=0}^{n}\binom{n}{k}\nu_{n-k}\eta_k,\qquad n\ge0, \tag*{(8)}

with m0=η0=ν0=1m_0=\eta_0=\nu_0=1.

If FF is polynomial or analytic and powers of F(Z)F(Z) can be expressed through the moments of ZZ, then (8) is a recursive coefficient scheme.

Affine reset: product transform and perpetuity representation

Let

F(x)=cx+b,∣c∣<1,F(x) = cx + b,\qquad|c| < 1,

be an affine resetting mechanism based on multiplying by fixed proportion c∈(−1,1)c \in(-1,1) and shifting by b∈Rb \in\mathbb{R}. Then (7) becomes

v(s)=qq−ψ(s)e−bsv(cs).(9)v(s) = \frac{q}{q-\psi(s)}e^{-bs}v(cs). \tag*{(9)}

Iterating (9) gives

v(s)=exp⁡{−bs1−c}∏j=0∞qq−ψ(cjs).(10)v(s) = \exp\left\{-\frac{bs}{1-c}\right\}\prod_{j=0}^{\infty}\frac{q}{q-\psi(c^{j}s)}. \tag*{(10)}

Equivalently,

Z=db1−c+∑j=0∞cjξj,(11)Z \mathrel{\overset{d}{=}} \frac{b}{1-c}+\sum_{j=0}^{\infty}c^{j}\xi_j, \tag*{(11)}

where ξ0,ξ1,…\xi_0,\xi_1,\ldots are i.i.d. copies of X(eq)X(e_q). In this case, when F(x)=cx+bF(x)=cx+b, we also have

ηk=E(cZ+b)k=∑r=0k(kr)crbk−rmr.\eta_k=\mathbb{E}(cZ+b)^k=\sum_{r=0}^{k}\binom{k}{r}c^r b^{k-r}m_r.

Thus, using (8), the term containing mnm_n appears only when k=nk=n and r=nr=n and

(1−cn)mn=∑k=0n(nk)νn−k∑r=0min⁡(k,n−1)(kr)crbk−rmr,n≥1.(12)(1-c^n)m_n=\sum_{k=0}^{n}\binom{n}{k}\nu_{n-k}\sum_{r=0}^{\min(k,n-1)}\binom{k}{r}c^r b^{k-r}m_r,\qquad n\geq1. \tag*{(12)}

This gives moments in increasing order. The first two are

m1=b+ν11−c.(13)m_1=\frac{b+\nu_1}{1-c}. \tag*{(13)}
MSD⁡(Z)=MSD⁡(ξ)1−c2.(14)\operatorname{MSD}(Z)=\frac{\operatorname{MSD}(\xi)}{1-c^2}. \tag*{(14)}

Unreflected Brownian under affine resetting

In this section we focus on very special case of the Brownian motion with affine resetting, that is, Let

X(t)=dt+σB(t),d∈R,σ>0,X(t)=dt+\sigma B(t),\qquad d\in\mathbb{R},\quad\sigma>0,

where BB is standard Brownian motion. Then

ψ(s)=−ds+σ2s22,m(s)=qq+ds−σ2s2/2.\psi(s)=-ds+\frac{\sigma^2s^2}{2},\qquad m(s)=\frac{q}{q+ds-\sigma^2s^2/2}.

For F(x)=cx+bF(x)=cx+b, ∣c∣<1|c|<1, (10) becomes

v(s)=exp⁡{−bs1−c}∏j=0∞qq+dcjs−12σ2c2js2.(15)v(s)=\exp\left\{-\frac{bs}{1-c}\right\}\prod_{j=0}^{\infty}\frac{q}{q+dc^js-\frac{1}{2}\sigma^2c^{2j}s^2}. \tag*{(15)}

The product is valid for ss in a neighbourhood of the origin contained in the intersection of the domains of all factors. The increment ξ=X(eq)\xi=X(e_q) has

⟨ξ⟩=dq,MSD⁡(ξ)=σ2q+d2q2.\langle\xi\rangle=\frac{d}{q},\qquad\operatorname{MSD}(\xi)=\frac{\sigma^2}{q}+\frac{d^2}{q^2}.

Consequently,

⟨Z⟩=b+d/q1−c,MSD⁡(Z)=σ2/q+d2/q21−c2.(16)\langle Z\rangle=\frac{b+d/q}{1-c},\qquad\operatorname{MSD}(Z)=\frac{\sigma^2/q+d^2/q^2}{1-c^2}. \tag*{(16)}

Unreflected finite-activity Lévy flight under affine resetting

Here “Lévy flight” is used in the finite-activity sense: jumps occur at Poisson times and have a regularly varying displacement distribution. Let

X(t)=at+∑i=1Nλ(t)Ji(17)X(t)=at+\sum_{i=1}^{N_{\lambda}(t)}J_i \tag*{(17)}

where NλN_{\lambda} is Poisson with rate λ\lambda, J1,J2,…J_1,J_2,\ldots are i.i.d., and a∈Ra\in\mathbb{R}. Let

G^(s):=⟨e−sJ1⟩.\widehat{G}(s):=\langle e^{-sJ_1}\rangle.

Then

ψ(s)=−as+λ(G^(s)−1),m(s)=qq+as−λ(G^(s)−1).(18)\psi(s)=-as+\lambda(\widehat{G}(s)-1),\qquad m(s)=\frac{q}{q+as-\lambda(\widehat{G}(s)-1)}. \tag*{(18)}

For F(x)=cx+bF(x)=cx+b, ∣c∣<1|c|<1,

v(s)=exp⁡{−bs1−c}∏j=0∞qq+acjs−λ(G^(cjs)−1).(19)v(s)=\exp\left\{-\frac{bs}{1-c}\right\}\prod_{j=0}^{\infty}\frac{q}{q+ac^{j}s-\lambda(\widehat{G}(c^{j}s)-1)}. \tag*{(19)}

Assume JJ has finite first two moments. Then

⟨ξ⟩=a+λ⟨J⟩q,\langle\xi\rangle=\frac{a+\lambda\langle J\rangle}{q},

and, by conditioning on eqe_q,

MSD⁡(ξ)=E[MSD⁡(X(eq)∣eq)]+MSD⁡(⟨X(eq)∣eq⟩)=λ⟨J2⟩q+(a+λ⟨J⟩)2q2.(20)\begin{aligned} \operatorname{MSD}(\xi)&=\mathbb{E}\left[\operatorname{MSD}(X(e_q)\mid e_q)\right]+\operatorname{MSD}\left(\langle X(e_q)\mid e_q\rangle\right)\\ &=\frac{\lambda\langle J^2\rangle}{q}+\frac{(a+\lambda\langle J\rangle)^2}{q^2}. \tag*{(20)} \end{aligned}

Therefore

⟨Z⟩=b+(a+λ⟨J⟩)/q1−c,MSD⁡(Z)=λ⟨J2⟩/q+(a+λ⟨J⟩)2/q21−c2.(21)\langle Z\rangle=\frac{b+(a+\lambda\langle J\rangle)/q}{1-c},\qquad\operatorname{MSD}(Z)=\frac{\lambda\langle J^2\rangle/q+(a+\lambda\langle J\rangle)^2/q^2}{1-c^2}. \tag*{(21)}

If JJ is nonnegative and regularly varying with index −α-\alpha, α>0\alpha>0, i.e.

P(J>x)=x−αL(x),\mathbb{P}(J>x)=x^{-\alpha}L(x),

where LL is slowly varying, then ξ=X(eq)\xi=X(e_q) has the one-big-jump tail

P(ξ>x)∼λqP(J>x).(22)\mathbb{P}(\xi>x)\sim\frac{\lambda}{q}\mathbb{P}(J>x). \tag*{(22)}

Using the perpetuity representation (11) and standard subexponential closure,

P(Z>x)∼∑j=0∞P(cjξj>x)∼λq11−cαP(J>x),x→∞,(23)\mathbb{P}(Z>x)\sim\sum_{j=0}^{\infty}\mathbb{P}(c^j\xi_j>x)\sim\frac{\lambda}{q}\frac{1}{1-c^\alpha}\mathbb{P}(J>x),\qquad x\to\infty, \tag*{(23)}

for 0<c<10<c<1.

Unreflected spectrally-positive α\alpha-stable flight under affine resetting

Let XX be a spectrally-positive stable Lévy flight with index α∈(1,2)\alpha\in(1,2) and negative linear component, normalised so that

ψ(s)=μs+κsα,μ>0,κ>0.(24)\psi(s)=\mu s+\kappa s^\alpha,\qquad\mu>0,\quad\kappa>0. \tag*{(24)}

The positive jumps have tail

Π(x,∞)∼Kαx−α,(25)\Pi(x,\infty)\sim K_\alpha x^{-\alpha}, \tag*{(25)}

for a constant $K_\alpha>0 determined by the normalisation of the stable flight.

For F(x)=cx+bF(x)=cx+b, 0<c<10<c<1, the transform is

v(s)=exp⁡{−bs1−c}∏j=0∞qq−μcjs−κcαjsα.(26)v(s)=\exp\left\{-\frac{bs}{1-c}\right\}\prod_{j=0}^{\infty}\frac{q}{q-\mu c^j s-\kappa c^{\alpha j}s^\alpha}. \tag*{(26)}

The domain is 0≤s<Φ(q)0\le s<\Phi(q), where Φ(q)\Phi(q) is the positive root of μs+κsα=q\mu s+\kappa s^\alpha=q. Since the positive jumps are regularly varying, the exponential-time increment satisfies

P(X(eq)>x)∼1qΠ(x,∞)∼Kαqx−α.(27)\mathbb{P}(X(e_q)>x)\sim\frac{1}{q}\Pi(x,\infty)\sim\frac{K_\alpha}{q}x^{-\alpha}. \tag*{(27)}

Therefore

P(Z>x)∼Kαq(1−cα)x−α.(28)\mathbb{P}(Z>x)\sim\frac{K_\alpha}{q(1-c^\alpha)}x^{-\alpha}. \tag*{(28)}

Unreflected two-point random coefficient

2.9. Unreflected two-point random coefficient. Let

F(x)=Cx+b,P(C=c)=p,P(C=0)=1−p,(29)F(x)=Cx+b,\qquad\mathbb{P}(C=c)=p,\qquad\mathbb{P}(C=0)=1-p, \tag*{(29)}

with 0≤c<10\le c<1, 0≤p≤10\le p\le1, and b≥0b\ge0. Then

v(s)=m(s)e−bs((1−p)+pv(cs)).(30)v(s)=m(s)e^{-bs}\bigl((1-p)+pv(cs)\bigr). \tag*{(30)}

Iterating gives the explicit series

v(s)=(1−p)∑n=0∞pnexp⁡{−bs1−cn+11−c}∏j=0nm(cjs).(31)v(s)=(1-p)\sum_{n=0}^{\infty}p^n\exp\left\{-bs\frac{1-c^{n+1}}{1-c}\right\}\prod_{j=0}^{n}m(c^js). \tag*{(31)}

For p=1p=1, the geometric termination disappears and (31) reduces to the affine product (10). The corresponding perpetuity representation is

Z=d∑j=0N(∏i=0j−1Ci)(b+ξj),(32)Z\mathrel{\overset{d}{=}}\sum_{j=0}^{N}\left(\prod_{i=0}^{j-1}C_i\right)(b+\xi_j), \tag*{(32)}

where NN is geometric with

P(N=n)=(1−p)pn,n=0,1,2,…,\mathbb{P}(N=n)=(1-p)p^n,\qquad n=0,1,2,\ldots,

and, on {N=n}\{N=n\}, C0=⋯=Cn−1=cC_0=\cdots=C_{n-1}=c and Cn=0C_n=0.

For moments, let γr:=⟨Cr⟩=pcr\gamma_r:=\langle C^r\rangle=pc^r for r≥1r\ge1, with γ0=1\gamma_0=1. Then

(1−γn)mn=∑k=0n(nk)νn−k∑r=0min⁡(k,n−1)(kr)E(Crbk−r)mr,n≥1.(33)(1-\gamma_n)m_n=\sum_{k=0}^{n}\binom{n}{k}\nu_{n-k}\sum_{r=0}^{\min(k,n-1)}\binom{k}{r}\mathbb{E}(C^r b^{k-r})m_r,\qquad n\ge1. \tag*{(33)}

Because bb is deterministic here,

⟨Crbk−r⟩=bk−rγr.\langle C^r b^{k-r}\rangle=b^{k-r}\gamma_r.

The same recursion works for any finite-support CC. For a three-point reset coefficient

P(C=ci)=pi,i=1,2,3,\mathbb{P}(C=c_i)=p_i,\qquad i=1,2,3,

one only replaces γr\gamma_r by p1c1r+p2c2r+p3c3rp_1c_1^r+p_2c_2^r+p_3c_3^r. The transform equation becomes

v(s)=m(s)e−bs∑i=13piv(cis),v(s)=m(s)e^{-bs}\sum_{i=1}^{3}p_i v(c_i s),

which is a branching functional equation.

If ξ\xi is regularly varying with index −α-\alpha and B≡bB\equiv b is light-tailed, then

P(Z>x)∼11−⟨Cα⟩P(ξ>x).(34)\mathbb{P}(Z>x)\sim\frac{1}{1-\langle C^\alpha\rangle}\mathbb{P}(\xi>x). \tag*{(34)}

For the two-point coefficient (29), this gives

P(Z>x)∼11−pcαP(ξ>x).(35)\mathbb{P}(Z>x)\sim\frac{1}{1-p c^{\alpha}}\mathbb{P}(\xi>x). \tag*{(35)}

The reflected spectrally positive reset model

Model and embedded chain

Assume now that XX is spectrally positive and not a subordinator. Given an initial level x≥0x\geq0, define the reflected process at 00 by

Y(t)=x+X(t)−inf⁡0≤u≤t(x+X(u))∧0.(36)Y(t)=x+X(t)-\inf_{0\leq u\leq t}(x+X(u))\wedge0. \tag*{(36)}

At resetting Poissonian epochs with intensity q>0q>0, the resetting using functional FF is performed. Let

Yn:=Y(Tn−).Y_n:=Y(T_n-).

Then

Yn+1=dmax⁡{Fn+1(Yn)+X(eq),X‾(eq)}.(37)Y_{n+1}\mathrel{\mathop{=}^{d}}\max\{F_{n+1}(Y_n)+X(e_q),\overline{X}(e_q)\}. \tag*{(37)}

and at stationarity (assuming that it exists and is unique),

Y=dmax⁡{F(Y)+X(eq),X‾(eq)},(38)Y\mathrel{\mathop{=}^{d}}\max\{F(Y)+X(e_q),\overline{X}(e_q)\}, \tag*{(38)}

where by YY we denote the steady-state. Equivalently, by Wiener–Hopf factorisation,

Y=dX‾(eq)+(F(Y)−EΦ)+,(39)Y\mathrel{\mathop{=}^{d}}\overline{X}(e_q)+(F(Y)-E_{\Phi})^{+}, \tag*{(39)}

where EΦE_{\Phi} has the exponential distribution with parameter Φ(q)\Phi(q) and X‾(eq)\overline{X}(e_q), EΦE_{\Phi}, and F(Y)F(Y) are independent on the right-hand side.

Ergodicity conditions

The following is the Palmowski-Vlasiou [14] comparison criterion, written in the notation used here.

Theorem 3.1 (Reflected comparison criterion). Assume that XX is not a subordinator and that there exist nonnegative i.i.d. random variables (An,Bn)(A_n,B_n) such that

Fn(x)≤Anx+Bn,x≥0.(40)F_n(x)\leq A_nx+B_n,\qquad x\geq0. \tag*{(40)}

Then the reflected reset chain has a unique stationary limiting distribution if one of the following holds:

(a) An=0A_n=0 a.s.;

(b) An=1A_n=1 a.s. and ⟨B1⟩+⟨X(1)⟩/q<0\langle B_1\rangle+\langle X(1)\rangle/q<0;

(c) ⟨log⁡A1⟩<0\langle\log A_1\rangle<0 and ⟨log⁡+B1⟩<∞\langle\log^{+}B_1\rangle<\infty.

Here, if B1=0B_1=0 in (c), one may dominate by Anx+1A_nx+1 without changing the conclusion.

Proof sketch. Under (40), the original chain is stochastically dominated by the affine reflected chain

Yn+1a=dmax⁡{An+1Yna+Bn+1+X(eq),X‾(eq)}.Y^{a}_{n+1}\mathrel{\mathop{=}^{d}}\max\{A_{n+1}Y^{a}_n+B_{n+1}+X(e_q),\overline{X}(e_q)\}.

Cases (a)–(c) imply return to a compact set. Since XX is not a subordinator, the reflected process can hit zero, and zero is a regeneration state in the Harris-chain sense. The minorisation on a compact set then gives positive Harris recurrence and uniqueness of the stationary distribution. □\square

Remark 3.2. For the partial restting F(x)=CxF(x)=Cx with 0≤C≤10\leq C\leq1 and P(C<1)>0\mathbb{P}(C<1)>0, condition (c) holds after domination by Cx+1Cx+1 whenever ⟨log⁡C⟩<0\langle\log C\rangle<0. Note that this holds for example when CC has a uniform law U(0,1)U(0,1), because in this case ⟨log⁡C⟩=−1\langle\log C\rangle=-1.

Laplace transform, moments and tail of steady-state

Let

v(s):=⟨e−sY⟩,hF(s):=⟨e−sF(Y)⟩.v(s) := \langle e^{-sY} \rangle,\qquad h_F(s) := \langle e^{-sF(Y)} \rangle.

Using (39), independence, and the elementary identity

⟨e−s(a−EΦ)+⟩=Φ(q)e−sa−se−Φ(q)aΦ(q)−s,a≥0,(41)\left\langle e^{-s(a-E_{\Phi})^+} \right\rangle= \frac{\Phi(q)e^{-sa}-se^{-\Phi(q)a}}{\Phi(q)-s},\qquad a\geq0, \tag*{(41)}

we get

v(s)=ℓq(s)Φ(q)hF(s)−shF(Φ(q))Φ(q)−s,(42)v(s)=\ell_q(s)\frac{\Phi(q)h_F(s)-sh_F(\Phi(q))}{\Phi(q)-s}, \tag*{(42)}

where ℓq(s)\ell_q(s) is defined in (3). Substituting (3) and simplifying yields the reset-isolated form

v(s)=qΦ(q)(q−ψ(s)){Φ(q)hF(s)−shF(Φ(q))}.(43)v(s)=\frac{q}{\Phi(q)(q-\psi(s))}\left\{\Phi(q)h_F(s)-sh_F(\Phi(q))\right\}. \tag*{(43)}

Equivalently,

v(s)=a(s)(hF(s)−sΦ(q)hF(Φ(q))),a(s):=qq−ψ(s).(44)v(s)=a(s)\left(h_F(s)-\frac{s}{\Phi(q)}h_F(\Phi(q))\right),\qquad a(s):=\frac{q}{q-\psi(s)}. \tag*{(44)}

Let ξ:=X(eq)\xi:=X(e_q) and write

νn:=⟨ξn⟩,mn:=⟨Yn⟩,ηn:=⟨F(Y)n⟩,KF:=hF(Φ(q))=⟨e−Φ(q)F(Y)⟩.\nu_n:=\langle\xi^n\rangle,\qquad m_n:=\langle Y^n\rangle,\qquad\eta_n:=\langle F(Y)^n\rangle,\qquad K_F:=h_F(\Phi(q))=\langle e^{-\Phi(q)F(Y)}\rangle.

Expanding (44) gives, for n≥1n\geq1,

mn=∑k=0n(nk)νn−kηk+nKFΦ(q)νn−1.(45)m_n=\sum_{k=0}^{n}\binom{n}{k}\nu_{n-k}\eta_k+\frac{nK_F}{\Phi(q)}\nu_{n-1}. \tag*{(45)}

This is not yet closed for a general FF, but it becomes recursive for affine or random-affine reset maps. For F(x)=cxF(x)=cx, 0≤c<10\leq c<1, ηk=ckmk\eta_k=c^km_k and

(1−cn)mn=∑k=0n−1(nk)νn−kckmk+nKcΦ(q)νn−1,Kc:=v(cΦ(q)).(46)(1-c^n)m_n=\sum_{k=0}^{n-1}\binom{n}{k}\nu_{n-k}c^km_k+\frac{nK_c}{\Phi(q)}\nu_{n-1},\qquad K_c:=v(c\Phi(q)). \tag*{(46)}

In particular,

⟨Y⟩=ν1+Kc/Φ(q)1−c=⟨X(eq)⟩+v(cΦ(q))/Φ(q)1−c.(47)\langle Y\rangle=\frac{\nu_1+K_c/\Phi(q)}{1-c}=\frac{\langle X(e_q)\rangle+v(c\Phi(q))/\Phi(q)}{1-c}. \tag*{(47)}

Since ν1=⟨X(1)⟩/q=−ψ′(0)/q\nu_1=\langle X(1)\rangle/q=-\psi'(0)/q, this may also be written as

⟨Y⟩=v(cΦ(q))/Φ(q)−ψ′(0)/q1−c.\langle Y\rangle=\frac{v(c\Phi(q))/\Phi(q)-\psi'(0)/q}{1-c}.

For F(x)=CxF(x)=Cx with CC independent of YY,

(1−⟨Cn⟩)mn=∑k=0n−1(nk)νn−k⟨Ck⟩mk+nKCΦ(q)νn−1,KC:=⟨v(CΦ(q))⟩.(48)(1-\langle C^n\rangle)m_n=\sum_{k=0}^{n-1}\binom{n}{k}\nu_{n-k}\langle C^k\rangle m_k+\frac{nK_C}{\Phi(q)}\nu_{n-1},\qquad K_C:=\langle v(C\Phi(q))\rangle. \tag*{(48)}

For CC having a uniform distribution U(0,1)U(0,1), we have ⟨Cn⟩=(n+1)−1\langle C^n\rangle=(n+1)^{-1} and KC=Φ(q)−1∫0Φ(q)v(u) duK_C=\Phi(q)^{-1}\int_0^{\Phi(q)}v(u)\,du.

Following Palmowski–Vlasiou [14], we record the regularly varying tail expression for the most relevant partial resetting case with random proportion.

Proposition 3.3 (Regularly varying reflected Lévy process under partial resetting). Assume that XX is spectrally positive, not a subordinator, and that its Lévy measure satisfies

Π‾(x):=Π(x,∞)∈RV⁡−α,α>0.\overline{\Pi}(x):=\Pi(x,\infty)\in\operatorname{RV}_{-\alpha},\qquad\alpha>0.

If F(x)=CxF(x)=Cx, where 0≤C≤10\le C\le1, then

P(Y>x)∼1q(1−⟨Cα⟩)Π‾(x).(49)\mathbb{P}(Y>x)\sim\frac{1}{q(1-\langle C^\alpha\rangle)}\overline{\Pi}(x). \tag*{(49)}

Proof sketch. The Wiener–Hopf representation (39) gives

Y=dX‾(eq)+(CY−EΦ)+.Y\overset{d}{=}\overline{X}(e_q)+(CY-E_\Phi)^+.

The exponential EΦE_\Phi is light-tailed and does not change a regularly varying tail. The stochastic recurrence is therefore tail-equivalent to the random-coefficient affine recursion

Y≈X‾(eq)+CY.Y\approx\overline{X}(e_q)+CY.

For regularly varying additive term with index α\alpha and ⟨Cα⟩<1\langle C^\alpha\rangle<1, the standard one-large-term argument gives

P(Z>x)∼∑j=0∞P(C1⋯CjX‾j(eq)>x)∼∑j=0∞⟨Cα⟩jP(X‾(eq)>x).\mathbb{P}(Z>x)\sim\sum_{j=0}^{\infty}\mathbb{P}(C_1\cdots C_j\overline{X}_j(e_q)>x)\sim\sum_{j=0}^{\infty}\langle C^\alpha\rangle^j\mathbb{P}(\overline{X}(e_q)>x).

We also know, by principle of one big jump, that

P(X‾(eq)>x)∼1qΠ‾(x).(50)\mathbb{P}(\overline{X}(e_q)>x)\sim\frac{1}{q}\overline{\Pi}(x). \tag*{(50)}

Using (50) gives (49). □

For deterministic C=c∈[0,1)C=c\in[0,1) above proposition gives

P(Y>x)∼1q(1−cα)Π‾(x),(51)\mathbb{P}(Y>x)\sim\frac{1}{q(1-c^\alpha)}\overline{\Pi}(x), \tag*{(51)}

and for CC having uniform law U(0,1)U(0,1),

P(Y>x)∼α+1αqΠ‾(x).(52)\mathbb{P}(Y>x)\sim\frac{\alpha+1}{\alpha q}\overline{\Pi}(x). \tag*{(52)}

The partial resetting

Let F(x)=cxF(x)=cx, 0≤c<10\le c<1. Then hF(s)=v(cs)h_F(s)=v(cs) and (44) gives

v(s)=a(s)(v(cs)−sΦ(q)v(cΦ(q))),a(s)=qq−ψ(s).(53)v(s)=a(s)\left(v(cs)-\frac{s}{\Phi(q)}v(c\Phi(q))\right),\qquad a(s)=\frac{q}{q-\psi(s)}. \tag*{(53)}

Let

K:=v(cΦ(q)),A∞(s):=∏j=0∞a(cjs),K:=v(c\Phi(q)),\qquad A_\infty(s):=\prod_{j=0}^{\infty}a(c^j s),

and

B∞(s):=∑k=0∞ck∏j=0ka(cjs).B_\infty(s):=\sum_{k=0}^{\infty}c^k\prod_{j=0}^{k}a(c^j s).

Iterating (53) yields

v(s)=A∞(s)−KsΦ(q)B∞(s).(54)v(s)=A_\infty(s)-K\frac{s}{\Phi(q)}B_\infty(s). \tag*{(54)}

The constant KK is obtained by setting s=cΦ(q)s=c\Phi(q):

K=A∞(cΦ(q))1+cB∞(cΦ(q)).(55)K=\frac{A_\infty(c\Phi(q))}{1+cB_\infty(c\Phi(q))}. \tag*{(55)}

Equations (54)–(55) give a completely explicit transform in terms of the Lévy exponent ψ\psi.

For completeness, again using Palmowski–Vlasiou [14], we record the exponential-type asymptotics of the tail of the law of YY. Suppose there exists θ∗>0\theta_* > 0 such that

ψ(θ∗)=q,(56)\psi(\theta_*) = q, \tag*{(56)}

and the corresponding tilted ladder-height mean mqm_q is finite. Let

⟨eθ∗F(Y)⟩<∞.(57)\langle e^{\theta_*F(Y)}\rangle< \infty. \tag*{(57)}

Then the reflected Lévy process under resetting the asymptotic tail is given by

P(Y>x)∼De−θ∗x,(58)\mathbb{P}(Y>x)\sim De^{-\theta_*x}, \tag*{(58)}

with

D=P(EΦ>F(Z))κ(q,0)θ∗mq,EΦ∼Exp⁡(Φ(q)),(59)D=\mathbb{P}(E_{\Phi}>F(Z))\frac{\kappa(q,0)}{\theta_*m_q},\qquad E_{\Phi}\sim\operatorname{Exp}(\Phi(q)), \tag*{(59)}

where κ(q,0)\kappa(q,0) is LST of ladder time. For bounded resets and for F(x)=(B−x)+F(x)=(B-x)^+ with light-tailed BB, the condition (57) is immediate. For F(x)=CxF(x)=Cx, it becomes a condition on v(−θ∗C)v(-\theta_*C), and therefore requires finite positive exponential moments of YY.

Reflected Brownian motion under partial resettting

Consider F(x)=cxF(x)=cx. Let

X(t)=σB(t)−μt,μ>0,σ>0.X(t)=\sigma B(t)-\mu t,\qquad\mu>0,\quad\sigma>0.

Then

ψ(s)=μs+σ2s22,Φ(q)=−μ+μ2+2qσ2σ2.\psi(s)=\mu s+\frac{\sigma^2s^2}{2},\qquad\Phi(q)=\frac{-\mu+\sqrt{\mu^2+2q\sigma^2}}{\sigma^2}.

Thus

a(s)=qq−μs−σ2s2/2.(60)a(s)=\frac{q}{q-\mu s-\sigma^2s^2/2}. \tag*{(60)}

Substituting this aa in (54) gives

v(s)=∏j=0∞qq−μcjs−12σ2c2js2−KsΦ(q)∑k=0∞ck∏j=0kqq−μcjs−12σ2c2js2.(61)v(s)=\prod_{j=0}^{\infty}\frac{q}{q-\mu c^j s-\frac{1}{2}\sigma^2c^{2j}s^2} -K\frac{s}{\Phi(q)}\sum_{k=0}^{\infty}c^k\prod_{j=0}^{k}\frac{q}{q-\mu c^j s-\frac{1}{2}\sigma^2c^{2j}s^2}. \tag*{(61)}

where KK is given by (55). The first moment follows from (47):

⟨Y⟩=K/Φ(q)−μ/q1−c.(62)\langle Y\rangle=\frac{K/\Phi(q)-\mu/q}{1-c}. \tag*{(62)}

The second moment follows from (46). Since

ν1=⟨X(eq)⟩=−μq,ν2=⟨X(eq)2⟩=σ2q+2μ2q2,\nu_1=\langle X(e_q)\rangle=-\frac{\mu}{q},\qquad \nu_2=\langle X(e_q)^2\rangle=\frac{\sigma^2}{q}+\frac{2\mu^2}{q^2},

we have

(1−c2)m2=ν2+2cν1m1+2KΦ(q)ν1.(63)(1-c^2)m_2=\nu_2+2c\nu_1m_1+\frac{2K}{\Phi(q)}\nu_1. \tag*{(63)}

Together with (62), this gives MSD⁡(Y)=m2−m12\operatorname{MSD}(Y)=m_2-m_1^2.

Reflected reset F(x)=(B−x)+F(x)=(B-x)^{+} with exponential BB

Let F(x)=(B−x)+F(x)=(B-x)^{+}, where BB has an exponential law Exp⁡(β)\operatorname{Exp}(\beta) that is independent of YY. The memoryless property closes the reset transform. Conditional on Y=zY=z,

(B−z)+=0on {B≤z},(B-z)^{+}=0 \quad\text{on } \{B\le z\},

and, on {B>z}\{B>z\}, the excess B−zB-z is exponential with rate β\beta. Therefore

P((B−Y)+>y)=⟨e−β(Y+y)⟩=v(β)e−βy,y≥0.(64)\mathbb{P}\left((B-Y)^{+}>y\right)=\left\langle e^{-\beta(Y+y)}\right\rangle=v(\beta)e^{-\beta y}, \qquad y\ge0. \tag*{(64)}

Thus (B−Y)+(B-Y)^{+} has an atom 1−v(β)1-v(\beta) at zero and an exponential density v(β)βe−βyv(\beta)\beta e^{-\beta y} on (0,∞)(0,\infty). Hence

hF(s)=⟨e−s(B−Z)+⟩=1−v(β)+v(β)ββ+s=1−v(β)sβ+s.(65)h_F(s)=\left\langle e^{-s(B-Z)^{+}}\right\rangle=1-v(\beta)+v(\beta)\frac{\beta}{\beta+s}=1-v(\beta)\frac{s}{\beta+s}. \tag*{(65)}

Let r:=v(β)r:=v(\beta) and Φ:=Φ(q)\Phi:=\Phi(q). Substituting (65) into (43) gives

v(s)=qΦ(q−ψ(s))[Φ(1−rsβ+s)−s(1−rΦβ+Φ)].(66)v(s)=\frac{q}{\Phi(q-\psi(s))}\left[\Phi\left(1-r\frac{s}{\beta+s}\right)-s\left(1-r\frac{\Phi}{\beta+\Phi}\right)\right]. \tag*{(66)}

The remaining scalar r=v(β)r=v(\beta) is obtained by setting s=βs=\beta in (66). With

Dβ:=qΦ(q−ψ(β)),D_{\beta}:=\frac{q}{\Phi(q-\psi(\beta))},

we get

r=Dβ[Φ(1−r2)−β(1−rΦβ+Φ)]=Dβ(Φ−β)−DβrΦ(12−ββ+Φ).\begin{aligned} r=D_{\beta}\left[\Phi\left(1-\frac{r}{2}\right)-\beta\left(1-r\frac{\Phi}{\beta+\Phi}\right)\right] \\ &=D_{\beta}(\Phi-\beta)-D_{\beta}r\Phi\left(\frac{1}{2}-\frac{\beta}{\beta+\Phi}\right). \end{aligned}

Thus, provided β<Φ\beta<\Phi and β\beta is in the transform domain,

r=Dβ(Φ−β)1+DβΦ(Φ−β)/(2(β+Φ)).(67)r=\frac{D_{\beta}(\Phi-\beta)}{1+D_{\beta}\Phi(\Phi-\beta)/(2(\beta+\Phi))}. \tag*{(67)}

Equations (66) and (67) provide the LST explicitly.

Partial resetting with uniform proportion

Let F(x)=CxF(x)=Cx with CC having uniform distribution U(0,1)U(0,1). Then for CC, independent of YY we have

hF(s)=⟨e−sCY⟩=∫01v(su) du=1s∫0sv(y) dy.(68)h_F(s)=\left\langle e^{-sCY}\right\rangle=\int_0^1v(su)\,du=\frac{1}{s}\int_0^sv(y)\,dy. \tag*{(68)}

Write

F0(s):=∫0sv(y) dy,Φ:=Φ(q).F_0(s):=\int_0^s v(y)\,dy,\qquad\Phi:=\Phi(q).

Then (44) becomes

v(s)=F0(s)/s−(s/Φ2)F0(Φ)1−ψ(s)/q.(69)v(s)=\frac{F_0(s)/s-(s/\Phi^2)F_0(\Phi)}{1-\psi(s)/q}. \tag*{(69)}

Since v=F0′v=F_0', this is the first-order inhomogeneous differential equation

F0′(s)=F0(s)−b0s2/2s(1−ψ(s)/q),b0:=2F0(Φ)Φ2.(70)F_0'(s)=\frac{F_0(s)-b_0s^2/2}{s(1-\psi(s)/q)},\qquad b_0:=\frac{2F_0(\Phi)}{\Phi^2}. \tag*{(70)}

Solving (70) with the boundary condition F0(0)=0F_0(0)=0 and the regularity condition at s=Φs=\Phi gives final form. Define

g(s):=∫0sexp⁡{−∫0xψ(y)y(q−ψ(y)) dy} dx.(71)g(s):=\int_0^s\exp\left\{-\int_0^x\frac{\psi(y)}{y(q-\psi(y))}\,dy\right\}\,dx. \tag*{(71)}

Then

v(s)=1−g(s)/g(Φ)g′(s)(1−ψ(s)/q),0≤s<Φ.(72)v(s)=\frac{1-g(s)/g(\Phi)}{g'(s)(1-\psi(s)/q)}, \qquad0\le s<\Phi. \tag*{(72)}

Moreover b0=1/g(Φ)b_0=1/g(\Phi).

Reflected Brownian motion under uniform partial resetting

Take again

X(t)=σB(t)−μt,ψ(s)=μs+σ2s22.X(t)=\sigma B(t)-\mu t,\qquad\psi(s)=\mu s+\frac{\sigma^2s^2}{2}.

Let

y1:=Φ(q)=−μ+μ2+2qσ2σ2,y2:=−μ−μ2+2qσ2σ2<0.(73)y_1:=\Phi(q)=\frac{-\mu+\sqrt{\mu^2+2q\sigma^2}}{\sigma^2},\qquad y_2:=\frac{-\mu-\sqrt{\mu^2+2q\sigma^2}}{\sigma^2}<0. \tag*{(73)}

Then

−ψ(y)y(q−ψ(y))=D1y−y1+D2y−y2,D1=−y2y1−y2,D2=y1y1−y2.(74)-\frac{\psi(y)}{y(q-\psi(y))}=\frac{D_1}{y-y_1}+\frac{D_2}{y-y_2},\qquad D_1=\frac{-y_2}{y_1-y_2},\quad D_2=\frac{y_1}{y_1-y_2}. \tag*{(74)}

Hence

g(s)=∫0s(x−y1−y1)D1(x−y2−y2)D2 dx,0≤s≤y1.(75)g(s)=\int_0^s\left(\frac{x-y_1}{-y_1}\right)^{D_1}\left(\frac{x-y_2}{-y_2}\right)^{D_2}\,\mathrm{d}x,\qquad0\le s\le y_1. \tag*{(75)}

Substitution into (72) gives

v(s)=(y1y1−s)1+D1(−y2−y2+s)1+D2(1−g(s)g(y1)),0≤s<y1.(76)v(s)=\left(\frac{y_1}{y_1-s}\right)^{1+D_1}\left(\frac{-y_2}{-y_2+s}\right)^{1+D_2}\left(1-\frac{g(s)}{g(y_1)}\right),\qquad0\le s<y_1. \tag*{(76)}

The integral gg may be written through incomplete beta functions. With

z(s)=s−y2y1−y2,z(s)=\frac{s-y_2}{y_1-y_2},

we have

∫0s(x−y1)D1(x−y2)D2 dx=(−1)D1(y1−y2)D1+D2+1[B(z(s);1+D2,1+D1)−B(z(0);1+D2,1+D1)].(77)\int_0^s(x-y_1)^{D_1}(x-y_2)^{D_2}\,\mathrm{d}x=(-1)^{D_1}(y_1-y_2)^{D_1+D_2+1}\left[B(z(s);1+D_2,1+D_1)-B(z(0);1+D_2,1+D_1)\right]. \tag*{(77)}

where B(z;a,b)=∫0zta−1(1−t)b−1 dtB(z;a,b)=\int_0^z t^{a-1}(1-t)^{b-1}\,\mathrm{d}t.

Reflected spectrally positive stable flight under partial restting

Let XX have exponent

ψ(s)=μs+κsα,α∈(1,2),\psi(s)=\mu s+\kappa s^\alpha,\qquad\alpha\in(1,2),

as in (24). For deterministic proportion of the partial resetting, hence for F(x)=cxF(x)=cx, 0<c<10<c<1, the transform is given by (54) with

a(s)=qq−μs−κsα(78)a(s)=\frac{q}{q-\mu s-\kappa s^\alpha} \tag*{(78)}

and Φ(q)\Phi(q) solving μΦ(q)+κΦ(q)α=q\mu\Phi(q)+\kappa\Phi(q)^\alpha=q. Note that the Lévy tail satisfies (25). Then Proposition 3.3 gives

P(Y>x)∼Kαq(1−cα)x−α.(79)\mathbb{P}(Y>x)\sim\frac{K_\alpha}{q(1-c^\alpha)}x^{-\alpha}. \tag*{(79)}

For uniform partial resetting, one should replace (1−cα)−1(1-c^\alpha)^{-1} by (α+1)/α(\alpha+1)/\alpha.

Conclusion

The unreflected model with resetting gives a clean baseline: one step analysis reduce the description of stationary state ZZ to the fixed-point equation Z=dX(eq)+F(Z)Z \mathrel{\mathop{=}\limits^{d}} X(e_q) + F(Z) for the steady-state ZZ. The reflected model keeps the same reset structure but inserts the Wiener–Hopf correction term, producing the stationary state identity (39). The presented method is quite robust. For example, we can analyse other processes under resetting: unreflected Lévy process under affine random resetting, that is, F(x)=Cx+BF(x)=Cx+B with CC having finite support, beta distributions, or mixtures with an atom at zero. Similarly, we can handle polynomial resetting mechanism with F(x)=b0+b1x+⋯+bdxdF(x)=b_0+b_1x+\cdots+b_dx^d. Then (8) expresses mnm_n through moments up to order dndn. Very interesting seems to be the case the resetting is based on finite-support random functions. In this case FF takes maps f1,…,frf_1,\ldots,f_r with probabilities p1,…,prp_1,\ldots,p_r. Then

v(s)=m(s)∑i=1rpi⟨e−sfi(Z)⟩.v(s)=m(s)\sum_{i=1}^{r}p_i\langle e^{-sf_i(Z)}\rangle.

The same resetting mechanisms can be considered for the reflected Lévy process. Within this context quite intriguing is the case of Brownian plus compound Poisson with phase-type positive jumps which yields rational ψ\psi. Then all transform formulae reduce to products or finite sums involving rational factors. This is the natural setting for numerical inversion.

Faculty of Pure and Applied Mathematics, Wrocław University of Science and Technology Wyb. Wyspiańskiego 27, 50-370 Wrocław, Poland

Email address: [email protected]

University of Twente, the Netherlands

Email address: [email protected]

References

  1. [1]J. Assefa, M. Keller-Ressel, Z. Palmowski. Moments of polynomial processes under affine resetting. arXiv:2610.01363, 2026.arxiv.org/abs/2610.01363
  2. [2]W.J. Bell. Searching behaviour: The behavioural ecology of finding resources. Chapman and Hall, 1991.
  3. [6]M. Dahlenburg, A. Checkin, R. Schumer, R. Metzler. Stochastic resetting by a random amplitude. Phys. Rev. E 103:052123, 2021.DOI
  4. [9]M.R. Evans and S.N. Majumdar. Diffusion with stochastic resetting. Phys.Rev. Lett. 106:160601, 2011.DOI
  5. [11]S. Gupta and A.M. Jayannavar. Stochastic resetting: A (very) brief review. Front. Phys. 10:789097, 2022.DOI
  6. [12]F.F. Guillemin, Ph. Robert, B. Zwart. AIMD algorithms and exponential functionals. Ann. Appl. Probab., 14:90–117, 2004.DOI
  7. [14]Z. Palmowski and M. Vlasio. A Lévy input model with additional state-dependent services. Stochastic Processes and their Applications, 121:1546–1564, 2011.DOI
  8. [15]E. Roldan, A. Lisica, D. Sanchez-Taltavull, S.W. Grill. Stochastic resetting in backtrack recovery by RNA polymerases. Phys. Rev. E 93(6):062411, 2016.DOI
  9. [16]J. Whitehouse, M.R. Evans, and S.N. Majumdar. Effect of partial absorption on diffusion with resetting. Phys. Rev. E 8:022118, 2013.DOI

Paper details

vlnc:W_zsvyurbu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents