betaFeedback

WorksMathematics

Compensation Formulas for Branching Lévy Processes

Abstract

This paper aims to establish two compensation formulas for branching Lévy processes. The first expresses expected sums over spatial jumps as integrals against the Lévy measure, with the future family of the jumping particle started from its position immediately after the jump. The second expresses expected sums over reproduction events as time integrals weighted by the branching rate and the offspring distribution, with each event marked by the ordered tuple of branching processes initiated by its offspring. Combined with the fixed-time many-to-one formula, it recovers the whole-tree many-to-one identity for functionals of ancestral spatial paths.

Introduction

Consider a one-dimensional branching Lévy process with local reproduction defined on a probability space (Ω,F,P)(\Omega,\mathcal{F},\mathbb{P}). The system starts with a single particle that moves according to a Lévy process with triplet (β,σ2,ν)(\beta,\sigma^2,\nu). Its lifetime is exponentially distributed with rate γ>0\gamma> 0 and is independent of its motion. At the end of its lifetime, the particle is replaced by kk offspring with probability pkp_k, where p={pk}k≥0p = \{p_k\}_{k\geq0} is the offspring distribution. All offspring are born at the parent’s position at death and evolve independently according to the same motion and reproduction rules. We assume that only finitely many branching events occur in each bounded time interval almost surely.

This branching particle system can also be constructed by equipping a continuous-time Galton–Watson tree with independent Lévy motions along the lifetimes of its individuals; see [1] and § 2. We label the initial particle by ∅\varnothing and other particles using the Ulam–Harris system. The collection T\mathcal{T} of all labels forms a subset of U\mathcal{U} defined by

U:=⋃n=0∞N+n.\mathcal{U} := \bigcup_{n=0}^{\infty}\mathbb{N}_{+}^{n}.

with N+={1,2,…}\mathbb{N}_{+} = \{1,2,\ldots\} and the convention N+0={∅}\mathbb{N}_{+}^{0} = \{\varnothing\}; see [1, 3] for details. Let Nf(R)N_f(\mathbb{R}) denote the space of finite counting measures on R\mathbb{R}, equipped with the topology of weak convergence. Let DfD_f be the space of Nf(R)N_f(\mathbb{R})-valued càdlàg paths, equipped with the Skorokhod J1J_1-topology. For z∈Rz \in\mathbb{R}, write δz\delta_z for the unit point mass at zz. For t≥0t \geq0, denote by VtV_t the set of particles alive at time tt. For u∈Vtu \in V_t and 0≤s≤t0 \leq s \leq t, let Xu(s)X_u(s) be the position of uu or its ancestor at time ss. The Nf(R)N_f(\mathbb{R})-valued process Z={Zt}t≥0Z = \{Z_t\}_{t\geq0} defined by

Zt:=∑u∈VtδXu(t)(1)Z_t := \sum_{u\in V_t} \delta_{X_u(t)} \tag*{(1)}

has sample paths in DfD_f and records the positions of all living particles over time. It is usually referred to as a branching Lévy process. We denote its natural filtration after the usual completion by {Ft}t≥0\{\mathcal{F}_t\}_{t\geq0}. For x∈Rx\in\mathbb{R}, we write Px\mathbb{P}_x and Ex\mathbb{E}_x for the law and expectation of this process with the initial particle started from x∈Rx\in\mathbb{R}, and abbreviate P=P0\mathbb{P}=\mathbb{P}_0 and E=E0\mathbb{E}=\mathbb{E}_0.

Our first result extends the Lévy process compensation formula to the spatial jumps of all particles in the branching Lévy process. For t≥0t\geq0 and u∈Vtu\in V_t, let ΔXu(s):=Xu(s)−Xu(s−)\Delta X_u(s):=X_u(s)-X_u(s-) denotes the jump of XuX_u at time s∈(0,t]s\in(0,t], and Zt,u={Zst,u}s≥0Z^{t,u}=\{Z_s^{t,u}\}_{s\geq0} denotes the future family of uu, with its time origin at tt and its positions in the original spatial coordinates, i.e.,

Z0t,u=δXu(t)andZst,u=∑v∈Vt+s, u⪯vδXv(t+s),s>0,Z_0^{t,u}=\delta_{X_u(t)}\quad\text{and}\quad Z_s^{t,u}=\sum_{v\in V_{t+s},\,u\preceq v}\delta_{X_v(t+s)},\qquad s>0,

where u⪯vu\preceq v means that u=vu=v or uu is an ancestor of vv. In particular, Zgu,uZ^{g_u,u} is the sub branching Lévy process initiated by uu at its birth.

Theorem 1.1. Given a (Ft)(\mathcal{F}_t)-predictable process F={Ft:t≥0}\mathbf{F}=\{\mathbf{F}_t:t\geq0\} taking values in the space of non-negative measurable functions on U×R\mathcal{U}\times\mathbb{R} such that Ft(u,0)=0\mathbf{F}_t(u,0)=0 for all t≥0t\geq0 and u∈Uu\in\mathcal{U}, and a deterministic measurable function G={Gt:t≥0}\mathbf{G}=\{\mathbf{G}_t:t\geq0\} taking values in the space of non-negative measurable functions on U×R×Df\mathcal{U}\times\mathbb{R}\times D_f. For any x∈Rx\in\mathbb{R}, we have

Ex[∑t>0∑u∈VtFt(u,ΔXu(t))Gt(u,ΔXu(t),Zt,u)]=∫0∞Ex[∑u∈Vt∫RFt(u,y)⋅EXu(t)+y[Gt(u,y,Z)]ν(dy)]dt.(2)\mathbb{E}_x\left[\sum_{t>0}\sum_{u\in V_t}\mathbf{F}_t(u,\Delta X_u(t))\mathbf{G}_t(u,\Delta X_u(t),Z^{t,u})\right] = \int_0^\infty\mathbb{E}_x\left[\sum_{u\in V_t}\int_{\mathbb{R}}\mathbf{F}_t(u,y)\cdot\mathbb{E}_{X_u(t)+y}\left[\mathbf{G}_t(u,y,Z)\right]\nu(\mathrm{d}y)\right]\mathrm{d}t. \tag*{(2)}

Remark 1.2. When p1=1p_1=1, exactly one particle is alive at each time, and its position evolves as a Lévy process. When p0=1p_0=1, the initial particle dies without offspring, so its motion is killed at an independent exponential time of rate γ\gamma. In these two cases, the formula (2) with G≡1G\equiv1 recovers the compensation formulas for a Lévy process and for a Lévy process killed at rate γ\gamma, respectively.

For n≥1n\geq1, let DfnD_f^n be the space of ordered nn-tuples of paths in DfD_f endowed with the product Borel σ\sigma-algebra. Set

Df:=⋃n=0∞DfnwithDf0:={()},D_f:=\bigcup_{n=0}^{\infty}D_f^n\quad\text{with}\quad D_f^0:=\{()\},

and equip DfD_f with the disjoint-union σ\sigma-algebra. Here ()() denotes the empty tuple. For x∈Rx\in\mathbb{R} and k≥1k\geq1, let Px⊗k\mathbb{P}_x^{\otimes k} denote the joint law on DfkD_f^k of kk independent copies Z1,…,ZkZ^1,\ldots,Z^k of the branching Lévy process started from a single particle at xx. Write Ex⊗k\mathbb{E}_x^{\otimes k} for the corresponding expectation. For k=0k=0, let Px⊗0\mathbb{P}_x^{\otimes0} be the unit mass at ()().

Our second result gives a compensation formula for the reproduction events of the branching Lévy process. For each individual u∈Tu\in\mathcal{T}, let gug_u, dud_u, and ℓu\ell_u denote its birth time, death time, and lifetime, respectively. We make the convention that g∅=0g_{\varnothing}=0. At its death time dud_u, particle uu produces kuk_u offspring, where kuk_u has law pp. Under the Ulam–Harris labeling, the ii-th child of uu is labeled uiui for 1≤i≤ku1\leq i\leq k_u.

Theorem 1.3. Given a (Ft)(\mathcal{F}_t)-predictable process F={Ft:t≥0}\mathbf{F}=\{\mathbf{F}_t:t\geq0\} taking values in the space of non-negative measurable functions on U×N\mathcal{U}\times\mathbb{N}, and a deterministic measurable function G={Gt:t≥0}\mathbf{G}=\{\mathbf{G}_t:t\geq0\} taking values in the space of non-negative measurable functions on U×Df\mathcal{U}\times D_f. For any x∈Rx\in\mathbb{R}, we have

Ex[∑u∈TFdu(u,ku)Gdu(u,(Zdu,ui)i=1ku)]\mathbb{E}_x\left[\sum_{u\in\mathcal{T}}\mathbf{F}_{d_u}(u,k_u)\mathbf{G}_{d_u}\left(u,\left(Z^{d_u,ui}\right)_{i=1}^{k_u}\right)\right]
=Ex[∑t>0∑u∈Vt−Ft(u,ku)⋅1{du=t}⋅Gt(u,(Zt,ui)i=1ku)]=γ∫0∞Ex[∑u∈Vt∑k=0∞pkFt(u,k)EXu(t)⊗k[Gt(u,(Zi)i=1k)]]dt.(3)\begin{aligned} = \mathbb{E}_x\left[\sum_{t>0}\sum_{u\in V_{t-}} F_t(u,k_u)\cdot\mathbf{1}_{\{d_u=t\}}\cdot G_t\left(u,\left(Z^{t,ui}\right)_{i=1}^{k_u}\right)\right] \\ &= \gamma\int_0^\infty\mathbb{E}_x\left[\sum_{u\in V_t}\sum_{k=0}^{\infty}p_k F_t(u,k)\mathbb{E}_{X_u(t)}^{\otimes k}\left[G_t\left(u,(Z^i)_{i=1}^k\right)\right]\right]\mathrm{d}t. \tag*{(3)} \end{aligned}

Remark 1.4 The proof of Theorem 1.3 remains valid when the Lévy motion is replaced by a strong Markov motion with a Borel measurable path law as a function of the starting point. In particular, let ff be a non-negative measurable function on R+×D([0,∞),R)\mathbb{R}_+ \times D([0,\infty),\mathbb{R}) such that f(t,ω)f(t,\omega) depends only on ω∣[0,t)\omega|_{[0,t)}. We write f(t,ω∣[0,t))f(t,\omega|_{[0,t)}) for its value. Taking G≡1G \equiv1, Ft(u,k)=f(t,Xu∣[0,t))F_t(u,k)=f(t,X_u|_{[0,t)}) for u∈Vt−u\in V_{t-}, k≥0k\geq0, and setting Ft(u,k)=0F_t(u,k)=0 elsewhere, we obtain from (3) that

Ex[∑u∈Tf(du,Xu∣[0,du))]=γ∫0∞Ex[∑u∈Vtf(t,Xu∣[0,t))]dt.\mathbb{E}_x\left[\sum_{u\in\mathcal{T}}f(d_u,X_u|_{[0,d_u)})\right] =\gamma\int_0^\infty\mathbb{E}_x\left[\sum_{u\in V_t}f(t,X_u|_{[0,t)})\right]\mathrm{d}t.

If m:=∑k≥0kpk<∞m:=\sum_{k\geq0}kp_k<\infty, applying the fixed-time many-to-one formula to the right-hand side gives

Ex[∑u∈Tf(du,Xu∣[0,du))]=γ∫0∞eγ(m−1)t⋅Exξ[f(t,ξ∣[0,t))]dt,\mathbb{E}_x\left[\sum_{u\in\mathcal{T}}f(d_u,X_u|_{[0,d_u)})\right] =\gamma\int_0^\infty e^{\gamma(m-1)t}\cdot\mathbb{E}_x^\xi\left[f(t,\xi|_{[0,t)})\right]\mathrm{d}t,

where Exξ\mathbb{E}_x^\xi denotes expectation for a single copy ξ\xi of the spatial motion started at xx. This is the whole-tree many-to-one formula of [1 Proposition 3.5] specialized to local reproduction and functionals of the spatial path.

Remark 1.5 For any x0∈Rx_0\in\mathbb{R}, consider a branching Lévy process Z~\widetilde{Z} with particles being killed as they enter (−∞,x0)(-\infty,x_0). The two compensation formulas (2) and (3) remain valid for Z~\widetilde{Z}, since Z~\widetilde{Z} can be constructed as a functional of ZZ, i.e., for any tt and u∈V~tu\in\widetilde{V}_t, we have X~u(t)=Xu(t)\widetilde{X}_u(t)=X_u(t) and

Z~t=∑u∈Vt1{inf⁡{Xu(s):s∈[0,t]}≥x0}⋅δXu(t)andZ~st,u=∑v∈V~t+su⪯v1{inf⁡r∈[t,t+s]Xv(r)≥x0}⋅δXv(t+s),s≥0.\widetilde{Z}_t=\sum_{u\in V_t}\mathbf{1}_{\{\inf\{X_u(s):s\in[0,t]\}\geq x_0\}}\cdot\delta_{X_u(t)} \quad\text{and}\quad \widetilde{Z}^{t,u}_s=\sum_{\substack{v\in\widetilde{V}_{t+s}\\u\preceq v}}\mathbf{1}_{\{\inf_{r\in[t,t+s]}X_v(r)\geq x_0\}}\cdot\delta_{X_v(t+s)},\quad s\geq0.

Let F~t≥0\widetilde{\mathcal{F}}_t\geq0 denote the natural filtration of Z~\widetilde{Z}, completed in the usual way. Obviously, it is a subfiltration of {Ft}t≥0\{\mathcal{F}_t\}_{t\geq0}. As an illustration, let F\mathbf{F} and G\mathbf{G} be two functionals satisfying the assumptions of Theorem 1.1, with F~t\widetilde{\mathcal{F}}_t replaced by F~t\widetilde{\mathcal{F}}_t. Then

Ex[∑t>0∑u∈V~tFt(u,ΔX~u(t))Gt(u,ΔX~u(t),Z~t,u)]=Ex[∑t>0∑u∈Vt1{Xu(t)≥x0}Ft(u,ΔXu(t))Gt(u,ΔXu(t),Z~t,u)]=∫0∞Ex[∑u∈Vt−∫R1{Xu(t−)≥x0}1{Xu(t−)+y≥x0}Ft(u,y)EXu(t−)+y[Gt(u,y,Z~)] ν(dy)]dt=∫0∞Ex[∑u∈V~t∫R1{X~u(t)+y≥x0}Ft(u,y)EX~u(t)+y[Gt(u,y,Z~)] ν(dy)]dt.\begin{aligned} \mathbb{E}_x\left[\sum_{t>0}\sum_{u\in\widetilde{V}_t}F_t(u,\Delta\widetilde{X}_u(t))G_t(u,\Delta\widetilde{X}_u(t),\widetilde{Z}^{t,u})\right] \\ &=\mathbb{E}_x\left[\sum_{t>0}\sum_{u\in V_t}\mathbf{1}_{\{X_u(t)\geq x_0\}}F_t(u,\Delta X_u(t))G_t(u,\Delta X_u(t),\widetilde{Z}^{t,u})\right] \\ &=\int_0^\infty\mathbb{E}_x\left[\sum_{u\in V_{t-}}\int_{\mathbb{R}}\mathbf{1}_{\{X_u(t-)\geq x_0\}}\mathbf{1}_{\{X_u(t-)+y\geq x_0\}}F_t(u,y)\mathbb{E}_{X_u(t-)+y}[G_t(u,y,\widetilde{Z})]\,\nu(\mathrm{d}y)\right]\mathrm{d}t \\ &=\int_0^\infty\mathbb{E}_x\left[\sum_{u\in\widetilde{V}_t}\int_{\mathbb{R}}\mathbf{1}_{\{\widetilde{X}_u(t)+y\geq x_0\}}F_t(u,y)\mathbb{E}_{\widetilde{X}_u(t)+y}[G_t(u,y,\widetilde{Z})]\,\nu(\mathrm{d}y)\right]\mathrm{d}t. \end{aligned}

Proofs

Before proving the two main theorems, we first describe in detail the construction of the branching Lévy process ZZ started from a single particle at x∈Rx\in\mathbb{R} by assigning independent Lévy motions to the individuals of a continuous-time Galton–Watson tree. To each potential label u∈Uu\in\mathcal{U}, we assign an exponential lifetime ℓu\ell_u of rate γ\gamma, an offspring count kuk_u with law pp, and a Lévy process Yu=(Yu(s))s≥0Y_u=(Y_u(s))_{s\geq0} started at zero with triplet (β,σ2,ν)(\beta,\sigma^2,\nu). The parameter ss denotes the age of the particle. All these random elements are mutually independent, both within and across labels.

The realized tree T\mathbf{T} contains the root ∅\varnothing. For a label u=(u1,…,un)u=(u_1,\ldots,u_n) with n≥1n\ge1, it belongs to T\mathbf{T} if and only if ui≤ku(i−1)u_i\le k_{u^{(i-1)}} for all i=1,…,ni=1,\ldots,n, where u(0)=∅u^{(0)}=\varnothing and u(i−1)=(u1,…,ui−1)u^{(i-1)}=(u_1,\ldots,u_{i-1}) for i≥2i\ge2. For each u∈Tu\in\mathbf{T}, its birth and death times are

gu:=∑v≺uℓvanddu:=gu+ℓu,g_u:=\sum_{v\prec u}\ell_v \qquad\text{and}\qquad d_u:=g_u+\ell_u,

where v≺uv\prec u means that vv is a strict ancestor of uu. The empty sum gives g∅=0g_{\varnothing}=0. Thus the set of particles alive at time t≥0t\ge0 is Vt={u∈T:gu≤t<du}V_t=\{u\in\mathbf{T}:g_u\le t<d_u\}. For u∈Vtu\in V_t, its position at time tt can be given by

Xu(t)=x+∑v≺uYv(ℓv)+Yu(t−gu).X_u(t)=x+\sum_{v\prec u}Y_v(\ell_v)+Y_u(t-g_u).

Each strict ancestor contributes its displacement over its entire lifetime, and uu contributes its displacement up to age t−gut-g_u. Substituting VtV_t and Xu(t)X_u(t) into (1) gives a realization of ZZ.

Proof of Theorem 1.1. Let T={(u,ℓu)}u∈T\mathcal{T}=\{(u,\ell_u)\}_{u\in\mathbf{T}} be the whole genealogy of all labels and lifetimes. Almost surely, no jump occurs at gug_u for all u∈Tu\in\mathbf{T}, since the lifetime is independent of the motion and a Lévy process has no jump at a fixed age. Reindexing all jumps gives

S:=∑t>0∑u∈VtFt(u,ΔXu(t))Gt(u,ΔXu(t),Zt,u)=∑u∈T∑t∈[gu,du)Ft(u,ΔXu(t))Gt(u,ΔXu(t),Zt,u).(4)\begin{aligned} S&:=\sum_{t>0}\sum_{u\in V_t}F_t(u,\Delta X_u(t))G_t(u,\Delta X_u(t),Z^{t,u})\\ &=\sum_{u\in\mathbf{T}}\sum_{t\in[g_u,d_u)}F_t(u,\Delta X_u(t))G_t(u,\Delta X_u(t),Z^{t,u}). \tag*{(4)} \end{aligned}

Taking expectations on both sides of the second equality, then using the tower property and conditional Tonelli theorem, we obtain

Ex[S]=Ex[∑u∈TSuT]withSuT:=Ex[∑t∈[gu,du)Ft(u,ΔXu(t))Gt(u,ΔXu(t),Zt,u)∣T].(5)\mathbb{E}_x[S]=\mathbb{E}_x\left[\sum_{u\in\mathbf{T}}S_u^{\mathcal{T}}\right]\quad\text{with}\quad S_u^{\mathcal{T}}:=\mathbb{E}_x\left[\left.\sum_{t\in[g_u,d_u)}F_t(u,\Delta X_u(t))G_t(u,\Delta X_u(t),Z^{t,u})\right|\mathcal{T}\right]. \tag*{(5)}

For ϵ>0\epsilon>0 and u∈Tu\in\mathbf{T}, set τu,0ϵ=gu\tau_{u,0}^{\epsilon}=g_u and define recursively τu,jϵ:=inf⁡{s∈(τu,j−1ϵ,du):∣ΔXu(s)∣≥ϵ}\tau_{u,j}^{\epsilon}:=\inf\{s\in(\tau_{u,j-1}^{\epsilon},d_u):|\Delta X_u(s)|\ge\epsilon\} for j≥1j\ge1 with the convention inf⁡∅=∞\inf\varnothing=\infty. The finite terms of {τu,jϵ}j≥1\{\tau_{u,j}^{\epsilon}\}_{j\ge1} are the successive times at which particle uu makes an upward or downward jump of size at least ϵ\epsilon during its lifetime. Let

SuT(ϵ):=Ex[∑j=1∞Fτu,jϵ(u,ΔXu(τu,jϵ))Gτu,jϵ(u,ΔXu(τu,jϵ),Zτu,jϵ,u)∣T].S_u^{\mathcal{T}}(\epsilon):=\mathbb{E}_x\left[\left.\sum_{j=1}^{\infty}F_{\tau_{u,j}^{\epsilon}}(u,\Delta X_u(\tau_{u,j}^{\epsilon}))G_{\tau_{u,j}^{\epsilon}}(u,\Delta X_u(\tau_{u,j}^{\epsilon}),Z^{\tau_{u,j}^{\epsilon},u})\right|\mathcal{T}\right].

The monotone convergence theorem yields that SuT(ϵ)→ϵ→0+a.s.SuTS_u^{\mathcal{T}}(\epsilon)\xrightarrow[\epsilon\to0+]{{\rm a.s.}}S_u^{\mathcal{T}} as ϵ→0+\epsilon\to0+. For t≥0t\ge0, let Gt:=Ft∨σ(T)\mathcal{G}_t:=\mathcal{F}_t\vee\sigma(\mathcal{T}) reveal T\mathcal{T} at time zero and each particle’s spatial motion up to time tt. It is obvious that {τu,jϵ}j≥1\{\tau_{u,j}^{\epsilon}\}_{j\ge1} is a sequence of (Gt)(\mathcal{G}_t)-stopping times. Since σ(T)⊂Gτu,1ϵ\sigma(\mathcal{T})\subset\mathcal{G}_{\tau_{u,1}^{\epsilon}}, we have

SuT(ϵ)=Ex[Ex[SuT(ϵ)∣Gτu,1ϵ]∣T]=Ex[Ex[∑j=1∞Fτu,jϵ(u,ΔXu(τu,jϵ))Gτu,jϵ(u,ΔXu(τu,jϵ),Zτu,jϵ,u)∣Gτu,1ϵ]∣T].(6)\begin{aligned} S_u^{\mathcal{T}}(\epsilon) &=\mathbb{E}_x\left[\mathbb{E}_x\left[S_u^{\mathcal{T}}(\epsilon)\mid\mathcal{G}_{\tau_{u,1}^{\epsilon}}\right]\mid\mathcal{T}\right]\\ &=\mathbb{E}_x\left[\left.\mathbb{E}_x\left[\left.\sum_{j=1}^{\infty}F_{\tau_{u,j}^{\epsilon}}(u,\Delta X_u(\tau_{u,j}^{\epsilon}))G_{\tau_{u,j}^{\epsilon}}(u,\Delta X_u(\tau_{u,j}^{\epsilon}),Z^{\tau_{u,j}^{\epsilon},u})\right|\mathcal{G}_{\tau_{u,1}^{\epsilon}}\right]\right|\mathcal{T}\right]. \tag*{(6)} \end{aligned}

Conditioned on a given genealogy T\mathbf{T}, for any particle uu and t∈[gu,du)t\in[g_u,d_u), let Qt,uT(x,dz)Q_{t,u}^{\mathbf{T}}(x,dz) be the law on DfD_f of branching Lévy process started at x∈Rx\in\mathbb{R} at time tt, using the branching schedule of T\mathbf{T} from time tt and fresh independent Lévy motions. Define

Kt,uT(x,y):=∫DfGt(u,y,z)Qt,uT(x,dz),x,y∈R.(7)K_{t,u}^{\mathbf{T}}(x,y):=\int_{D_f}G_t(u,y,z)Q_{t,u}^{\mathbf{T}}(x,dz),\qquad x,y\in\mathbb{R}. \tag*{(7)}

On the event τu,1ϵ∈[gu,du)\tau_{u,1}^{\epsilon}\in[g_u,d_u), we have Fτu,1ϵ(u,ΔXu(τu,1ϵ))F_{\tau_{u,1}^{\epsilon}}(u,\Delta X_u(\tau_{u,1}^{\epsilon})) is Gτu,1ϵ\mathcal{G}_{\tau_{u,1}^{\epsilon}}-measurable and

Ex[Gτu,1ϵ(u,ΔXu(τu,1ϵ),Zτu,1ϵ,u)∣Gτu,1ϵ]=Kτu,1ϵ,uT(Xu(τu,1ϵ),ΔXu(τu,1ϵ)).\mathbb{E}_x\left[\left.G_{\tau_{u,1}^{\epsilon}}(u,\Delta X_u(\tau_{u,1}^{\epsilon}),Z^{\tau_{u,1}^{\epsilon},u})\right|\mathcal{G}_{\tau_{u,1}^{\epsilon}}\right] =K_{\tau_{u,1}^{\epsilon},u}^{\mathbf{T}}\left(X_u(\tau_{u,1}^{\epsilon}),\Delta X_u(\tau_{u,1}^{\epsilon})\right).

Substituting these into the last expression in (6) yields

SuT(ϵ)=Ex[Fτu,1ϵ(u,ΔXu(τu,1ϵ))Kτu,1ϵ,u(Xu(τu,1ϵ),ΔXu(τu,1ϵ)) | T]+Ex[∑j=2∞Fτu,jϵ(u,ΔXu(τu,jϵ))Gτu,jϵ(u,ΔXu(τu,jϵ),Zτu,jϵ,u) | T].\begin{aligned} S_u^T(\epsilon)=\mathbf{E}_x\left[F_{\tau_{u,1}^{\epsilon}}\left(u,\Delta X_u(\tau_{u,1}^{\epsilon})\right)K_{\tau_{u,1}^{\epsilon},u}\left(X_u(\tau_{u,1}^{\epsilon}),\Delta X_u(\tau_{u,1}^{\epsilon})\right)\,\middle|\,\mathcal{T}\right] \\ &+\mathbf{E}_x\left[\sum_{j=2}^{\infty}F_{\tau_{u,j}^{\epsilon}}\left(u,\Delta X_u(\tau_{u,j}^{\epsilon})\right)G_{\tau_{u,j}^{\epsilon}}\left(u,\Delta X_u(\tau_{u,j}^{\epsilon}),Z^{\tau_{u,j}^{\epsilon},u}\right)\,\middle|\,\mathcal{T}\right]. \end{aligned}

Using the preceding argument recursively to the last conditional expectation gives

SuT(ϵ)=Ex[∑j=1∞Fτu,jϵ(u,ΔXu(τu,jϵ))Kτu,jϵ,u(Xu(τu,jϵ),ΔXu(τu,jϵ)) | T]=Ex[∑t∈[gu,du)1{∣ΔXu(t)∣≥ϵ}⋅Ft(u,ΔXu(t))Kt,uT(Xu(t−)+ΔXu(t),ΔXu(t)) | T],\begin{aligned} S_u^T(\epsilon)=\mathbf{E}_x\left[\sum_{j=1}^{\infty}F_{\tau_{u,j}^{\epsilon}}\left(u,\Delta X_u(\tau_{u,j}^{\epsilon})\right)K_{\tau_{u,j}^{\epsilon},u}\left(X_u(\tau_{u,j}^{\epsilon}),\Delta X_u(\tau_{u,j}^{\epsilon})\right)\,\middle|\,\mathcal{T}\right] \\ &=\mathbf{E}_x\left[\sum_{t\in[g_u,d_u)}\mathbf{1}_{\{|\Delta X_u(t)|\geq\epsilon\}}\cdot F_t\left(u,\Delta X_u(t)\right)K_{t,u}^{T}\left(X_u(t-)+\Delta X_u(t),\Delta X_u(t)\right)\,\middle|\,\mathcal{T}\right], \end{aligned}

which converges almost surely as ϵ→0+\epsilon\to0+ to

SuT=Ex[∑t∈[gu,du)Ft(u,ΔXu(t))Kt,uT(Xu(t−)+ΔXu(t),ΔXu(t)) | T].S_u^T=\mathbf{E}_x\left[\sum_{t\in[g_u,d_u)}F_t\left(u,\Delta X_u(t)\right)K_{t,u}^{T}\left(X_u(t-)+\Delta X_u(t),\Delta X_u(t)\right)\,\middle|\,\mathcal{T}\right].

Conditioned on T\mathcal{T}, the process XuX_u is a Lévy process on [0,du][0,d_u] started at xx with Lévy measure ν\nu. By the compensation formula for jumps of Lévy process; see e.g. [2 p.7],

SuT=Ex[∫gududt∫RFt(u,y)Kt,uT(Xu(t−)+y,y)ν(dy) | T].S_u^T=\mathbf{E}_x\left[\int_{g_u}^{d_u}dt\int_{\mathbb{R}}F_t(u,y)K_{t,u}^{T}\left(X_u(t-)+y,y\right)\nu(dy)\,\middle|\,\mathcal{T}\right].

Substituting this into (5) gives

Ex[S]=Ex[∑u∈TEx[∫gududt∫RFt(u,y)Kt,uT(Xu(t−)+y,y)ν(dy) | T]]=Ex[∑u∈T∫gududt∫RFt(u,y)Kt,uT(Xu(t−)+y,y)ν(dy)]=Ex[∫0∞dt∑u∈Vt∫RFt(u,y)Kt,uT(Xu(t−)+y,y)ν(dy)].\begin{aligned} \mathbf{E}_x[S]=\mathbf{E}_x\left[\sum_{u\in\mathcal{T}}\mathbf{E}_x\left[\int_{g_u}^{d_u}dt\int_{\mathbb{R}}F_t(u,y)K_{t,u}^{T}\left(X_u(t-)+y,y\right)\nu(dy)\,\middle|\,\mathcal{T}\right]\right] \\ &=\mathbf{E}_x\left[\sum_{u\in\mathcal{T}}\int_{g_u}^{d_u}dt\int_{\mathbb{R}}F_t(u,y)K_{t,u}^{T}\left(X_u(t-)+y,y\right)\nu(dy)\right] \\ &=\mathbf{E}_x\left[\int_0^\infty dt\sum_{u\in V_t}\int_{\mathbb{R}}F_t(u,y)K_{t,u}^{T}\left(X_u(t-)+y,y\right)\nu(dy)\right]. \end{aligned}

Since Ft(u,y)F_t(u,y) is Ft\mathcal{F}_t-measurable, by Tonelli’s theorem and the tower property we have

Ex[S]=∫0∞dt∫REx[∑u∈VtFt(u,y)Kt,uT(Xu(t−)+y,y)]ν(dy)=∫0∞dt∫REx[∑u∈VtFt(u,y)Ex[Kt,uT(Xu(t−)+y,y) | Ft]]ν(dy).(8)\begin{aligned} \mathbf{E}_x[S]=\int_0^\infty dt\int_{\mathbb{R}}\mathbf{E}_x\left[\sum_{u\in V_t}F_t(u,y)K_{t,u}^{T}\left(X_u(t-)+y,y\right)\right]\nu(dy) \\ &=\int_0^\infty dt\int_{\mathbb{R}}\mathbf{E}_x\left[\sum_{u\in V_t}F_t(u,y)\mathbf{E}_x\left[K_{t,u}^{T}\left(X_u(t-)+y,y\right)\,\middle|\,\mathcal{F}_t\right]\right]\nu(dy). \tag*{(8)} \end{aligned}

By the exponential memorylessness and the independent future motions and reproduction, the branching property at time tt yields that conditional on Ft\mathcal{F}_t, the process Zt,uZ^{t,u} is a branching Lévy process started at Xu(t−)+yX_u(t-)+y. This along with (7) induces that for u∈Vtu\in V_t,

Ex[Kt,uT(Xu(t−)+y,y) | Ft]=Ex[∫DfGt(u,y,ω)QtT(Xu(t−)+y,dω) | Ft]=∫DfGt(u,y,ω)Ex[QtT(Xu(t−)+y,dω) | Ft]=∫DfGt(u,y,ω)PXu(t−)+y(Z∈dω)=EXu(t−)+y[Gt(u,y,Z)].\begin{aligned} \mathbf{E}_x\left[K_{t,u}^{T}\left(X_u(t-)+y,y\right)\,\middle|\,\mathcal{F}_t\right] =\mathbf{E}_x\left[\int_{D_f}G_t(u,y,\omega)Q_t^{T}\left(X_u(t-)+y,d\omega\right)\,\middle|\,\mathcal{F}_t\right] \\ &=\int_{D_f}G_t(u,y,\omega)\mathbf{E}_x\left[Q_t^{T}\left(X_u(t-)+y,d\omega\right)\,\middle|\,\mathcal{F}_t\right] \\ &=\int_{D_f}G_t(u,y,\omega)\mathbf{P}_{X_u(t-)+y}(Z\in d\omega) =\mathbf{E}_{X_u(t-)+y}\left[G_t(u,y,Z)\right]. \end{aligned}

The desired formula (2) follows directly by plugging this into the right-hand side of the second equality in (8) and using the stochastic continuity of Lévy processes. This completes the proof. □

Proof of Theorem 1.3. The first equality in (3) follows directly from the fact that each particle contributes to the sum only at its death time. For the second equality, we first have

Ex[∑u∈TFdu(u,ku)Gdu(u,(Zdu,ui)i=1ku)]=Ex[Fd∅(∅,k∅)Gd∅(∅,(Zd∅,i)i=1k∅)]+Ex[∑u∈Tu≠∅Fdu(u,ku)Gdu(u,(Zdu,ui)i=1ku)].(9)\mathbb{E}_{x}\left[\sum_{u\in\mathcal{T}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_i}\right)_{i=1}^{k_u}\right)\right] = \mathbb{E}_{x}\left[F_{d_{\varnothing}}(\varnothing,k_{\varnothing})G_{d_{\varnothing}}\left(\varnothing,\left(Z^{d_{\varnothing},i}\right)_{i=1}^{k_{\varnothing}}\right)\right] + \mathbb{E}_{x}\left[\sum_{\substack{u\in\mathcal{T}\\u\ne\varnothing}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_i}\right)_{i=1}^{k_u}\right)\right]. \tag*{(9)}

For the first expectation of the right-hand side, since Fd∅(∅,k∅)F_{d_{\varnothing}}(\varnothing,k_{\varnothing}) is Fd∅\mathcal{F}_{d_{\varnothing}}-measurable, by the tower property and the branching property we have

Ex[Fd∅(∅,k∅)Gd∅(∅,(Zd∅,i)i=1k∅)]=Ex[Fd∅(∅,k∅)⋅Ex[Gd∅(∅,(Zd∅,i)i=1k∅)∣Fd∅]].\mathbb{E}_{x}\left[F_{d_{\varnothing}}(\varnothing,k_{\varnothing})G_{d_{\varnothing}}\left(\varnothing,\left(Z^{d_{\varnothing},i}\right)_{i=1}^{k_{\varnothing}}\right)\right] = \mathbb{E}_{x}\left[F_{d_{\varnothing}}(\varnothing,k_{\varnothing})\cdot\mathbb{E}_{x}\left[G_{d_{\varnothing}}\left(\varnothing,\left(Z^{d_{\varnothing},i}\right)_{i=1}^{k_{\varnothing}}\right)\mid\mathcal{F}_{d_{\varnothing}}\right]\right].

The parent’s motion has no jump at its independent exponential death age almost surely. Hence every child of ∅\varnothing starts at X∅(d∅−)X_{\varnothing}(d_{\varnothing}-). Consequently,

Ex[Gd∅(∅,(Zd∅,i)i=1k∅)∣Fd∅]=EX∅(d∅−)⊗k∅[Gd∅(∅,(Zi)i=1k∅)],\mathbb{E}_{x}\left[G_{d_{\varnothing}}\left(\varnothing,\left(Z^{d_{\varnothing},i}\right)_{i=1}^{k_{\varnothing}}\right)\mid\mathcal{F}_{d_{\varnothing}}\right] =\mathbb{E}_{X_{\varnothing}(d_{\varnothing}-)}^{\otimes k_{\varnothing}}\left[G_{d_{\varnothing}}\left(\varnothing,\left(Z^i\right)_{i=1}^{k_{\varnothing}}\right)\right],

which follows that

Ex[Fd∅(∅,k∅)Gd∅(∅,(Zd∅,i)i=1k∅)]=Ex[Fd∅(∅,k∅)⋅EX∅(d∅−)⊗k∅[Gd∅(∅,(Zi)i=1k∅)]].(10)\mathbb{E}_{x}\left[F_{d_{\varnothing}}(\varnothing,k_{\varnothing})G_{d_{\varnothing}}\left(\varnothing,\left(Z^{d_{\varnothing},i}\right)_{i=1}^{k_{\varnothing}}\right)\right] =\mathbb{E}_{x}\left[F_{d_{\varnothing}}(\varnothing,k_{\varnothing})\cdot\mathbb{E}_{X_{\varnothing}(d_{\varnothing}-)}^{\otimes k_{\varnothing}}\left[G_{d_{\varnothing}}\left(\varnothing,\left(Z^i\right)_{i=1}^{k_{\varnothing}}\right)\right]\right]. \tag*{(10)}

For i=1,…,k∅i=1,\ldots,k_{\varnothing}, let Ti\mathcal{T}^{i} collects labels of the ii-th child of ∅\varnothing and all its descendants. By the tower property and the mutual independence among descendants, the second expectation on the right-hand side of (9),

Ex[Ex[∑u∈T,u≠∅Fdu(u,ku)Gdu(u,(Zdu,uj)j=1ku)∣Fd∅]]=Ex[∑i=1k∅Ex[∑u∈TiFdu(u,ku)Gdu(u,(Zdu,uj)j=1ku)∣Fd∅]].(11)\mathbb{E}_{x}\left[\mathbb{E}_{x}\left[\sum_{\substack{u\in\mathcal{T},u\ne\varnothing}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_j}\right)_{j=1}^{k_u}\right)\mid\mathcal{F}_{d_{\varnothing}}\right]\right] =\mathbb{E}_{x}\left[\sum_{i=1}^{k_{\varnothing}}\mathbb{E}_{x}\left[\sum_{u\in\mathcal{T}^{i}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_j}\right)_{j=1}^{k_u}\right)\mid\mathcal{F}_{d_{\varnothing}}\right]\right]. \tag*{(11)}

Similarly as in the preceding argument, for any i=1,…,k∅i=1,\ldots,k_{\varnothing}, we have

Ex[∑u∈TiFdu(u,ku)Gdu(u,(Zdu,uj)j=1ku)∣Fd∅]=Ex[Fdi(i,ki)Gdi(i,(Zdi,ij)j=1ki)∣Fd∅]+Ex[∑u∈Tiu≠iFdu(u,ku)Gdu(u,(Zdu,uj)j=1ku)∣Fd∅],\mathbb{E}_{x}\left[\sum_{u\in\mathcal{T}^{i}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_j}\right)_{j=1}^{k_u}\right)\mid\mathcal{F}_{d_{\varnothing}}\right] =\mathbb{E}_{x}\left[F_{d_i}(i,k_i)G_{d_i}\left(i,\left(Z^{d_i,i_j}\right)_{j=1}^{k_i}\right)\mid\mathcal{F}_{d_{\varnothing}}\right]+\mathbb{E}_{x}\left[\sum_{\substack{u\in\mathcal{T}^{i}\\u\ne i}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_j}\right)_{j=1}^{k_u}\right)\mid\mathcal{F}_{d_{\varnothing}}\right],

and further the first expectation on the right-hand side equals

Ex[Fdi(i,ki)EXi(di−)⊗ki[Gdi(i,(Zj)j=1ki)]∣Fd∅].\mathbb{E}_{x}\left[F_{d_i}(i,k_i)\mathbb{E}_{X_i(d_i-)}^{\otimes k_i}\left[G_{d_i}\left(i,\left(Z^j\right)_{j=1}^{k_i}\right)\right]\mid\mathcal{F}_{d_{\varnothing}}\right].

Substituting these back into the right-hand side of (11) gives

Ex[Ex[∑u∈T,u≠∅Fdu(u,ku)Gdu(u,(Zdu,ui)i=1ku)∣Fd∅]]=Ex[∑i=1k∅Fdi(i,ki)EXi(di−)⊗ki[Gdi(i,(Zj)j=1ki)]]+Ex[∑i=1k∅∑u∈Tiu≠iFdu(u,ku)Gdu(u,(Zdu,ui)i=1ku)].\mathbb{E}_{x}\left[\mathbb{E}_{x}\left[\sum_{\substack{u\in\mathcal{T},u\ne\varnothing}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_i}\right)_{i=1}^{k_u}\right)\mid\mathcal{F}_{d_{\varnothing}}\right]\right] =\mathbb{E}_{x}\left[\sum_{i=1}^{k_{\varnothing}}F_{d_i}(i,k_i)\mathbb{E}_{X_i(d_i-)}^{\otimes k_i}\left[G_{d_i}\left(i,\left(Z^j\right)_{j=1}^{k_i}\right)\right]\right]+\mathbb{E}_{x}\left[\sum_{i=1}^{k_{\varnothing}}\sum_{\substack{u\in\mathcal{T}^{i}\\u\ne i}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_i}\right)_{i=1}^{k_u}\right)\right].

Using the preceding argument recursively to the last expectation, then plugging this and (10) back into the right-hand sides of (9), we obtain

Ex[∑u∈TFdu(u,ku)Gdu(u,(Zdu,ui)i=1ku)]=Ex[∑u∈TFdu(u,ku)EXu(du−)⊗ku[Gdu(u,(Zj)j=1ku)]].\mathbb{E}_{x}\left[\sum_{u\in\mathcal{T}}F_{d_u}(u,k_u)G_{d_u}\left(u,\left(Z^{d_u,u_i}\right)_{i=1}^{k_u}\right)\right] =\mathbb{E}_{x}\left[\sum_{u\in\mathcal{T}}F_{d_u}(u,k_u)\mathbb{E}_{X_u(d_u-)}^{\otimes k_u}\left[G_{d_u}\left(u,\left(Z^j\right)_{j=1}^{k_u}\right)\right]\right].

We now continue considering the expectation on the right-hand side of the last equation. For convenience, we set

Ht(u,k):=Ft(u,k)EXu(t−)⊗k[Gt(u,(Zj)j=1k)],t≥0, u∈Vt−, k≥0,H_t(u,k) := F_t(u,k) \mathbb{E}_{X_u(t-)}^{\otimes k}\left[G_t\left(u,(Z^j)_{j=1}^k\right)\right], \qquad t \ge0,\ u \in V_{t-},\ k \ge0,

and Ht(u,k)=0H_t(u,k)=0 elsewhere. The indicator of {u∈Vt−}={gu<t≤du}\{u \in V_{t-}\}=\{g_u<t\le d_u\} is (Ft)(\mathcal{F}_t)-predictable, with gu=du=∞g_u=d_u=\infty for u∉Tu\notin\mathbb{T}. Thus Ht(u,k)H_t(u,k) is (Ft)(\mathcal{F}_t)-predictable and

Ex[∑u∈TFdu(u,ku)Gdu(u,(Zidu,u)i=1ku)]=Ex[∑u∈THdu(u,ku)]=∑u∈UEx[1{u∈T}Hdu(u,ku)].\begin{aligned} \mathbb{E}_x\left[\sum_{u\in\mathbb{T}}F_{d_u}(u,k_u)G_{d_u}\left(u,(Z_i^{d_u,u})_{i=1}^{k_u}\right)\right]=\mathbb{E}_x\left[\sum_{u\in\mathbb{T}}H_{d_u}(u,k_u)\right] \\ &=\sum_{u\in\mathcal{U}}\mathbb{E}_x\left[\mathbf{1}_{\{u\in\mathbb{T}\}}H_{d_u}(u,k_u)\right]. \end{aligned}

For each u∈Uu\in\mathcal{U}, let Au:=σ((Yv)v∈U,(ℓv,kv)v∈U∖{u})\mathcal{A}_u:=\sigma\left((Y_v)_{v\in\mathcal{U}},(\ell_v,k_v)_{v\in\mathcal{U}\setminus\{u\}}\right). Hence both the event {u∈T}\{u\in\mathbb{T}\} and the time gug_u are Au\mathcal{A}_u-measurable, since they depend only on the strict ancestors of uu. Conditionally on Au\mathcal{A}_u, the lifetime ℓu\ell_u and the offspring count kuk_u are still independent. For 0≤a<b<∞0\le a<b<\infty, A∈FaA\in\mathcal{F}_a and k≥0k\ge0, we first consider the contribution of particle uu when AA occurs and a<du≤ba<d_u\le b. On the event {u∈T,du>a}\{u\in\mathbb{T},d_u>a\}, particle uu has not died by time aa, so the history up to aa is the same if we replace ℓu\ell_u by a+1a+1 and kuk_u by zero while keeping all other variables fixed. Consequently, on this event, 1A=1Au\mathbf{1}_A=\mathbf{1}_{A^u} for some Au∈AuA^u\in\mathcal{A}_u, obtained by evaluating AA in that modified history. Since du=gu+ℓud_u=g_u+\ell_u on {u∈T}\{u\in\mathbb{T}\}, the density of ℓu\ell_u and the law of kuk_u give

Ex[1A⋅1{a<du≤b,ku=k}∣Au]=1Aupk∫a∨gubγe−γ(t−gu) dt=γpkEx[1Au∫ab1{gu<t≤du} dt∣Au].(12)\begin{aligned} \mathbb{E}_x\left[\mathbf{1}_A\cdot\mathbf{1}_{\{a<d_u\le b,k_u=k\}}\mid\mathcal{A}_u\right]=\mathbf{1}_{A^u}p_k\int_{a\vee g_u}^{b}\gamma e^{-\gamma(t-g_u)}\,dt \\ &=\gamma p_k\mathbb{E}_x\left[\mathbf{1}_{A^u}\int_a^b\mathbf{1}_{\{g_u<t\le d_u\}}\,dt\mid\mathcal{A}_u\right]. \tag*{(12)} \end{aligned}

where the middle integral is taken to be zero if a∨gu≥ba\vee g_u\ge b. Here the last equality follows from Tonelli’s theorem and the fact that

Px(du∈dt,ku=k∣Au)=pk⋅γe−γ(t−gu)⋅1{gu≤t} dt,t≥0, k≥0.\mathbb{P}_x(d_u\in dt,k_u=k\mid\mathcal{A}_u)=p_k\cdot\gamma e^{-\gamma(t-g_u)}\cdot\mathbf{1}_{\{g_u\le t\}}\,dt,\qquad t\ge0,\ k\ge0.

When u∉Tu\notin\mathbb{T}, both conditional expectations in (12) are zero. Taking expectations on both sides of (12) yields the desired identity for a predictable rectangle (a,b]×A(a,b]\times A with mark kk. For A∈FaA\in\mathcal{F}_a, we apply the same identity on (a+1/n,b](a+1/n,b] using an event in Fa+1/n\mathcal{F}_{a+1/n} that agrees with AA almost surely, and then let n→∞n\to\infty. By the monotone class theorem and the monotone convergence theorem, we obtain

Ex[1{u∈T}Hdu(u,ku)]=γEx[∫0∞1{gu<t≤du}∑k=0∞pkHt(u,k) dt].\mathbb{E}_x\left[\mathbf{1}_{\{u\in\mathbb{T}\}}H_{d_u}(u,k_u)\right]=\gamma\mathbb{E}_x\left[\int_0^\infty\mathbf{1}_{\{g_u<t\le d_u\}}\sum_{k=0}^\infty p_kH_t(u,k)\,dt\right].

Finally, summing this identity over u∈Uu\in\mathcal{U} and then using Tonelli’s theorem, we have

Ex[∑u∈THdu(u,ku)]=γ∫0∞Ex[∑u∈Vt−∑k=0∞pkHt(u,k)] dt.\mathbb{E}_x\left[\sum_{u\in\mathcal{T}}H_{d_u}(u,k_u)\right]=\gamma\int_0^\infty\mathbb{E}_x\left[\sum_{u\in V_{t-}}\sum_{k=0}^\infty p_kH_t(u,k)\right]\,dt.

The second equality in (3) follows directly by substituting the definition of Ht(u,k)H_t(u,k) into the right-hand side of the last equality, then using the stochastic continuity of VV and Lévy processes. This completes the proof. □

References

  1. [1]V. Bansaye, J.F. Delmas, L. Marsalle, and V.C. Tran. Limit theorems for Markov processes indexed by continuous time Galton–Watson trees. Ann. Appl. Probab., 21(6):2263–2314, 2011.
  2. [2]J. Bertoin. Lévy Processes. Cambridge University Press, Cambridge, 1996.
  3. [3]J.F. Le Gall. Random trees and applications. Probab. Surv., 2:245–311, 2005.DOI

Paper details

vlnc:W_c2haetqn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 4

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