arXiv:1706.01578v3 [math.PR] 12 Jan 2018 n rniinprobabilities transition and (1.1) with indt epstv.Ti sas h rwinbig odtoe ost to conditioned bridge ze hit Brownian to the conditioned also process is Bessel This 3-dimensional the or positive. positive, be to tioned nte akvpoes eoe y( by Brownian denoted of process, distributions Markov finite-dimensional another of (20 transforms Janson of Laplace and analysis (2006) asymptotic Pitman in example appears for also see It problems, and (1999). binatorial Bertoin Yor example and for Revuz See and literature. the in investigated tensively n t nt-iesoa itiuin r ie yfrua(.)belo (2.1) formula by given are distributions B finite-dimensional its and rcadWn 21,Scin3). Section (2016, c a Wang approach, (1998) and The Biane’s Biane Bryc following in drift. function, density construction appropriate probability the including (2 transition from Wang by L´evy obtained and process free be 2.2) 1/2-stable also Proposition can (2016, It Wang tange 3.1). and the Bryc as see arises so-called that processes; of process support Markov of boundary self-similar the positive a is This 21) hr eivsiae,b setal optn h Laplac the computing essentially by investigated, we where (2017), process. ex ULRPEETTOSO ALC RNFRSOF TRANSFORMS LAPLACE OF REPRESENTATIONS DUAL RWINECRINADGNRLZDMEANDERS GENERALIZED AND EXCURSION BROWNIAN osdrteBona xuso tnadzdt aelnt and 1 length have to standardized excursion Brownian the Consider h ups fti oei oitouea“ulrpeetto”t representation” “dual a introduce to is note this of purpose The 2010 u anrsl stefloigiett hc a eddi rcadWan and Bryc in needed was which identity following the is result main Our e od n phrases. and words Key ( = ahmtc ujc Classification. Subject Mathematics B xrse steLpaetasomo h aeaxlayMar auxiliary law. me sim same initial Brownian the A different of generalized a transform a interchanged. Laplace of the times as transform and expressed Laplace arguments the of for roles holds the with ( the and of transform Laplace the as expressed is excursion nian itiuino naxlayMro rcs,satdfo a from started process, Markov auxiliary an of distribution Abstract. t ex ) t ∈ [0 p P , t 1] ( ( X ,y x, eoeti rcs hogot rwinecrinhsbe ex been has excursion Brownian throughout. process this denote h alc rnfr fthe of transform Laplace The t ∈ = ) OZMEZBY N IA WANG YIZAO AND BRYC W LODZIMIERZ dy rwinecrin rwinmadr alc transform; Laplace meander; Brownian excursion; Brownian π | [( X y s − = 1. x x ) = ) 2 Introduction q 2( + Bona oin and motions -Brownian 2 t p 60J25. X √ t − x t 1 y ) s t + ( ≥ ,y x, d 0 y dmninldsrbto fBrow- of distribution -dimensional ) hogot ihsaesae[0 space state with throughout, t ) 2 dy, + t 0 4 ] ≤ > t , ,x t, < s d 0 o rcs,with process, kov +1)-dimensional σ q y , ne,wihis which ander, -Ornstein–Uhlenbeck fiiemeasure -finite ≥ lridentity ilar ≥ eiaino the of derivation 0 0 . 07). imn(1994) Pitman nb on in found be an xuso and excursion 1,Theorem 017, transforms, e oat ro tpoesat process nt pidt the to pplied aiu com- various a isthe ties hat .W let We w. t Markov condi- 1, = , ∞ ay g ) - 2 WLODZIMIERZ BRYC AND YIZAO WANG the fluctuations of asymmetric simple exclusion processes with open boundaries in the steady state. These processes are representative non-equilibrium models Derrida (2007); Derrida et al. (2004) that have attracted much attentions recently in probability and mathematical physics. Let E [ ] denote the expectation with respect to the law of (X ) starting at x · t t≥0 X0 = x> 0.

Theorem 1.1. For d N, let s0 =0

Theorem 1.2. For d N, let s0 =0

The paper is organized as follows. In Section 2 we recall some known facts about Brownian excursion and prove Theorem 1.1. In Section 3 we prove the dual representations for generalized Brownian meanders, with Brownian meander as a special case.

2. Proof of dual representation of Brownian excursion We first recall some facts on Brownian excursion Bex. For our purposes it is Bex Bex convenient to define it as a Markov process that starts at 0 = 0, ends at 1 = 0, and has transition probabilities

√8πℓt(y)ℓ1−t(y) if s =0 0 Bex Bex P ( t dy s = x)=  ∈ |  ℓ1−t(y) gt−s(x, y) if 0 0,y> 0, ℓ1−s(x)  with  1 y2 ℓt(y)= y exp 1{y>0} √ 3 · − 2t 2πt   and 2 2 1 (y1 y2) (y1 + y2) gt(y1,y2)= exp − exp 1{y ,y >0}. √2πt − 2t − − 2t 1 2      Equivalently, the joint probability density function at time points 0 < t < < 1 ··· td < 1 is d−1 √ (2.1) ft1,...,td (y1,...,yd)= 8πℓt1 (y1)ℓ1−td (yd) gtk+1−tk (yk,yk+1). kY=1 See Itˆoand McKean (1965, page 76), Durrett et al. (1977), or Revuz and Yor (1999, page 464). Proof of Theorem 1.1. The proof consists of rewriting the right-hand side of (1.2). We first assume t1 > 0 and td < 1. With t0 = 0, we have, using transition density functions of the process X and letting d~x denote dx dx , 0 ··· d ∞ d E 1 x0 exp (tk+1 tk)Xsk √x0 dx0 0 " −2 − !# Z Xk=0 1 d d = d~x√x0 exp (tk+1 tk)xk psk−sk 1 (xk−1, xk) Rd+1 −2 − − + ! Z Xk=0 kY=1 d d d+1 1 2 2 2 2 (2.2) =2 d~x exp (tk+1 tk)xk x0 xkpsk−sk 1 (xk−1, xk), Rd+1 −2 − ! × − Z + kX=0 kY=1 2 where we apply in the last equality changes of variables from xk to xk, k =0,...,d. We derive a different expression of the products of transition density functions, thanks to the factorization. Namely, recalling (1.1), we write 2sy p (x2,y2)= s π(s2 + (y x)2)(s2 + (y + x)2) − 1 s s = . 2πx s2 + (y x)2 − s2 + (y + x)2  −  4 WLODZIMIERZ BRYC AND YIZAO WANG

We now use the elementary Laplace transform s ∞ = e−sy cos(ay)dy s2 + a2 Z0 to get y ∞ yp (x2,y2)= e−sz [cos((y x)z) cos((y + x)z)] dz s 2πx − − Z0 1 y ∞ = e−sz sin(xz) sin(yz)dz. π x Z0 Therefore, writing d~y = dy dy , 1 ··· d d d 2 2 2 1 x0 xkpsk−sk 1 (xk−1, xk)= d d~y exp (sk sk−1)yk − π Rd − − + ! kY=1 Z Xk=1 d−1 x sin(x y ) x sin(x y ) sin(x y ) sin(x y ). × 0 0 1 × d d d × k k+1 k k kY=1 Then, noticing that the integrand is non-negative and interchanging the order of integrations, we rewrite the right-hand side of (2.2) as

d 2d+1 (2.3) d d~y exp (sk sk−1)yk π Rd − − Z + k=1 ! ∞ X ∞ −t x2/2 −(1−t )x2 /2 x e 1 0 sin(x y )dx x e d d sin(x y )dx × 0 0 1 0 × d d d d Z0 Z0 d−1 ∞ 2 −(tk+1−tk)x /2 e k sin(xkyk+1) sin(xkyk)dxk. × 0 kY=1 Z By straightforward calculation we have

∞ ∞ 2 y 2 xe−tx /2 sin(xy)dx = e−tx /2 cos(xy)dx = πℓ (y). t t Z0 Z0 In the last step above we used the formula for the cosine transform

∞ 2 π 1 2 e−tx /2 cos(ax)dx = e−a /(2t), 2 t1/2 Z0 r which follows from the characteristic function of a Gaussian distribution with mean zero and variance 1/t. We also have

∞ −tx2/2 e sin(xy1) sin(xy2)dx 0 Z ∞ 1 2 = e−tx /2 [cos(x(y y )) cos(x(y + y ))] dx 2 1 − 2 − 1 2 Z0 π 1 2 2 π = e−(y1−y2) /(2t) e−(y1+y2) /(2t) = g (y ,y ). 8 t1/2 − 2 t 1 2 r   LAPLACE TRANSFORM OF BROWNIAN EXCURSION AND MEANDER 5

Thus, recalling (2.1), expression (2.3) now becomes,

d d−1

4π d~y exp (sk sk−1)yk ℓt1 (y1)ℓ1−td (yd) gtk+1−tk (yk,yk+1) Rd − − ! Z Xk=1 kY=1 d √ E Bex = 2π exp (sk sk−1) tk . " − − !# Xk=1 We have proved the theorem with 0 0. Recall that 1 =0. If d = 1, then the left-hand √ −1 ∞ −x/2 side of (1.2) becomes 1, and the right-hand side becomes ( 2π) 0 e √xdx = 1. If d> 1, we see that the desired identity is reduced to the same type of identity with fewer arguments, namely with 0 0). For d 2, consider {x≥0} ≥ ≥ sk := sk+1 s1, tk := tk+1, k = 0,...,d 1 and td := 1. In this way, the left- − E d−1 − Bex hand side of (1.2) becomes [exp( k=1(sk sk−1) e )], and the right-hand side − − tk becomese e e P ∞ d−1 e e 1 E 1 x exp (tk+1 tk)Xsek +s1 √x dx √2π 0 " −2 − !# Z Xk=0 e ∞ e d−1 1 E 1 = x exp (tk+1 tk)Xsek √x dx. √2π 0 " −2 − !# Z kX=0 Above in the last step we used the stationarity of Xeagain.e Since necessarily t1 = t2 > 0, when d 2 the desired identity (1.2) becomes an identity for Laplace transform of (d 1)-dimensional≥ distribution of Brownian excursion at time points − e0 < t1 < < td−1 1 that we have already proved before. This completes the proof. ··· ≤  e e 3. Application to generalized Brownian meanders In this section, we let ν(dv) be a probability measure on [0, 1] such that ν( 0 )= B(ν) { } 0, and consider the generalized Brownian meander ( t )0≤t≤1 defined by 1 (3.1) B(ν) = Bex , t [0, 1], t √V V t ∈ where V is a random variable with law ν and independent from Bex. Explicit examples, including Brownian meander, will be discussed later in this section. Noting that the function v e−x/vv−3/2 is bounded on (0, 1] for x > 0, we define 7→ x/2 x √xe − −3/2 (3.2) ϕν (x)= e 2v v ν(dv). √2π Z(0,1] 6 WLODZIMIERZ BRYC AND YIZAO WANG

We have the following dual representation of B(ν).

Theorem 3.1. For d N, let s0 =0

d E B(ν) exp (sk sk−1) 1−tk " − − !# Xk=1 ∞ d E 1 = x exp (tk+1 tk)Xsk ϕν (x)dx. 0 " −2 − !# Z Xk=0 Proof. The key step is the following identity

d E sk sk−1 Bex (3.3) exp − v(1−tk ) " − √v !# Xk=1 x d 1 ∞ e− 2v 1 √ x/2 E = xe 3/2 x exp (tk+1 tk)Xsk dx √2π 0 v " −2 − !# Z Xk=0 for 0 < v 1. Once we establish (3.3), integrating with respect to ν on both sides yields the≤ desired result, by Fubini’s theorem. Now we prove (3.3). We first rewrite (1.2), with tk replaced by 1 tk, as processes Bex Bex − ( t )t∈[0,1] and ( 1−t)t∈[0,1] have the same law. So (1.2) becomes

d E Bex (3.4) exp (sk sk−1) 1−tk " − − !# Xk=1 ∞ d 1 E 1 1 = x exp (tk+1 tk)Xsk (1 (1 t1))X0 √x dx. √2π 0 " −2 − − 2 − − !# Z Xk=1

This identity remains valid, if we replace the increasing sequence (sk)k=0,...,d with (sk/√v)k=0,...,d and replace the increasing sequence (tk)k=1,...,d+1 with (tk)k=1,...,d+1 with tk := 1 v(1 tk) (formally we just replace sk by sk/√v and 1 t by v(1 t ), but we also− need− to verify the same monotonicity of the replaced − k − k sequences,e and thatet 0, t 1). In this way we get 1 ≥ d ≤

d e e E sk sk−1 Bex exp − v(1−tk ) " − √v !# Xk=1 d 1 ∞ 1 1 = Ex exp v(tk+1 tk)X sk (1 v(1 t1))X0 √x dx. √2π 0 " −2 − √v − 2 − − !# Z kX=1 Thanks to the self-similarity of process X,

d 2 ((Xλs)s≥0, Px) = λ (Xs)s≥0, Px/λ2 ,x> 0,λ> 0,  LAPLACE TRANSFORM OF BROWNIAN EXCURSION AND MEANDER 7 where P ( )= P( X = x) (see Bryc and Wang (2016)), we have x · · | 0 ∞ 1 d 1 Ex exp v(tk+1 tk)X sk (1 v(1 t1))X0 √xdx 0 " −2 − √v − 2 − − !# Z Xk=1 ∞ d E 1 1 X0 = vx exp (tk+1 tk)Xsk (1 v(1 t1)) √xdx 0 " −2 − − 2 − − v !# Z Xk=1 ∞ d E 1 1 1 = vx exp (tk+1 tk)Xsk 1 X0 √xdx 0 " −2 − − 2 v − !# Z Xk=0   ∞ d E 1 −x(1/v−1)/2 √x = x exp (tk+1 tk)Xsk e 3/2 dx. 0 " −2 − !# v Z Xk=0 This yields (3.3), and completes the proof.  Next, we specialize Theorem 3.1 to the case of Beta distribution. The formula shall involve the confluent hypergeometric function 1 ∞ ψ(α, β, x)= e−xuuα−1(1 + u)β−α−1du,α > 0,β R,x> 0. Γ(α) ∈ Z0 See for example (Lebedev, 1965, page 268).

Corollary 3.2. For d N, let s0 = 0 < s1 < s2 < < sd and t0 = 0 t1 < 0, then≤ ··· d ≤ d+1 − d E B(ν) (3.5) exp (sk sk−1) 1−tk " − − !# Xk=1 ∞ d Γ(α + β) E 1 5 x = x exp (tk+1 tk)Xsk ψ β, α, √x dx. √2πΓ(α) 0 " −2 − !# 2 − 2 Z kX=0   Proof. We apply Theorem 3.1 with 1 1 √x −x(1/v−1)/2 α−5/2 β−1 ϕν (x)= e v (1 v) dv √2π B(α, β) − Z0 1 √x ∞ (3.6) = e−ux/2(1 + u)3/2−(α+β)uβ−1du √2π B(α, β) Z0 √xΓ(β) 5 x = ψ β, α, , √2πB(α, β) 2 − 2   where in the second step we applied a change of variable u =1/v 1.  − Our results are also related to generalized Bessel meanders that were introduced in Mansuy and Yor (2008, Sections 3.6 and 3.7). There are three equivalent ways to define a generalized Bessel meander B(δ,δ′) for δ, δ′ > 0. (This is the process under the law M δ,δ′ in Mansuy and Yor (2008), and strictly speaking the authors did not give it a name there but only mentioned ‘generalized meanders’ in the section titles.) Firstly, it can be defined as a randomized Bessel bridge of dimension δ + δ′, using Beta(δ/2,δ′/2) distribution. For this approach, see Mansuy and Yor (2008, 2 ′ 2 1/2 Theorem 3.12). Secondly, it can be defined as ((Rt + (Rt) ) )t∈[0,1], where R is 8 WLODZIMIERZ BRYC AND YIZAO WANG

′ ′ a δ-dimensional Bessel bridge (pinned down at R0 = 0 and R1 = 0), R is a δ - dimensional starting from 0, and R and R′ are independent. Thirdly, the law M δ,δ′ of B(δ,δ′), viewed as a probability measure on C([0, 1]), is absolutely BES(δ+δ′) continuous with respect to the probability measure P0 on C([0, 1]) induced by a (δ + δ′)-dimensional Bessel process starting from 0; more precisely

δ,δ′ cδ,δ′ BES(δ+δ′) ′ (3.7) M = δ P0 , δ,δ > 0, X1 with 2δ/2Γ((δ + δ′)/2) (3.8) c = M δ,δ′ (Xδ)= . δ,δ′ 1 Γ(δ′/2)

Here, for ω from the canonical space C([0, 1]), X1(ω) = ω1. For the second and third characterizations, see Mansuy and Yor (2008, Theorem 3.9); the correspond- ing formula (3.8) in Theorem 3.9 therein has a typo and is corrected here. The generalized Bessel meanders Bδ,δ′ and the generalized Brownian meanders B(ν) with Beta distribution ν in Corollary 3.2 are in general different processes. However, since a Bessel bridge of dimension 3 is a Brownian excursion, generalized Bessel meanders (3.7) with (3.9) δ (0, 3) and δ′ =3 δ, ∈ − become a special case of the generalized Brownian meanders introduced in (3.1). This case covers a couple of examples investigated in the literature. In partic- ular, it is known that B(1,2) is the Brownian meander Bme and in this case the relation (3.7) is originally due to Imhof (1984). The process B(2,1), known as Brow- nian co-meander, has also been investigated before, and the relation (3.7) is due to Biane et al. (1987). See also Yen and Yor (2013, Theorem 7.4.1). Our Corollary 3.2 simplifies and takes the following form. Corollary 3.3. In the notations of Corollary 3.2, we have for all δ (0, 3), ∈ d E B(δ,3−δ) (3.10) exp (sk sk−1) 1−tk " − − !# Xk=1 ∞ d 1 E 1 δ/2−1 = δ/2 x exp (tk+1 tk)Xsk x dx. 2 Γ(δ/2) 0 " −2 − !# Z Xk=0 Proof. From (3.6) with α = δ/2, β =3/2 α it follows that − ∞ β √x Γ(α + β) −ux/2 β−1 2 Γ(α + β) 1/2−β ϕν (x)= e u du = x √2π Γ(α)Γ(β) √2πΓ(α) Z0 xδ/2−1 = . 2δ/2Γ(δ/2) 

Remark 3.4. When d = 1, expression (3.10) becomes ∞ B(δ,3 δ) 1 E −s t − E −tXs/2 −(1−t)x/2 δ/2−1 e = δ/2 x e e x dx, 2 Γ(δ/2) 0 h i Z h i LAPLACE TRANSFORM OF BROWNIAN EXCURSION AND MEANDER 9

This differs from Bertoin and Yor (2001, Eq. (2)), who developed a dual represen- tation of Laplace transforms of univariate distributions in the form of univariate integrals E e−sUt = E e−tRs ,s,t> 0. But, clearly, there is some similarity  in how the roles of arguments and times are interchanged. Remark 3.5. It is natural to interpret Beta(0, 3/2) and Beta(3/2, 0) as degenerate laws. The limiting case of Beta(δ/2, 3/2 δ/2)as δ 3 is V = 1, so with B(3,0) = Bex Corollary 3.3 can be viewed as an extension− of Theorem↑ 1.1. On the other hand, as δ 0, one can check that B(δ,3−δ) converges to the 3-dimensional Bessel process. In this↓ limit, the right-hand side of (3.5) becomes an undefined expression. The 3-dimensional Bessel process is also considered in Hirsch and Yor (2012, Theorem 6.1). Laplace transforms of squared Bessel processes have been investigated in Pitman and Yor (1982). However, we do not see immediate connection between our identities to the results there. Acknowledgement. The authors thank Jim Pitman for insightful comments on several early versions of the paper, and a few key references including in particular Mansuy and Yor (2008), which helped us improve significantly the paper. The authors also thank Joseph Najnudel for several inspiring discussions. WB’s research was supported in part by the Charles Phelps Taft Research Center at the University of Cincinnati. YW’s research was partially supported by NSA grant H98230-16-1- 0322 and Army Research Laboratory grant W911NF-17-1-0006.

References Bertoin, J. and Pitman, J. (1994). Path transformations connecting , excursion and meander. Bull. Sci. Math., 118(2):147–166. Bertoin, J. and Yor, M. (2001). On subordinators, self-similar Markov processes and some factorizations of the exponential variable. Electron. Comm. Probab., 6:95–106. Biane, P. (1998). Processes with free increments. Math. Z., 227(1):143–174. Biane, P., Le Gall, J.-F., and Yor, M. (1987). Un processus qui ressemble au pont brownien. In S´eminaire de Probabilit´es, XXI, volume 1247 of Lecture Notes in Math., pages 270–275. Springer, Berlin. Bryc, W. and Wang, Y. (2016). The local structure of q-Gaussian processes. Probab. Math. Statist., 36(2):335–352. Bryc, W. and Wang, Y. (2017). Limit fluctuations for density of asymmetric simple exclusion processes with open boundaries. arxiv http://arxiv.org/abs/1707.07350. Bryc, W. and Wesolowski, J. (2017). Asymmetric simple exclusion process with open boundaries and quadratic harnesses. J. Stat. Phys., 167(2):383–415. Derrida, B. (2007). Non-equilibrium steady states: fluctuations and large deviations of the density and of the current. Journal of Statistical Mechanics: Theory and Experiment, 2007(07):P07023. Derrida, B., Enaud, C., and Lebowitz, J. (2004). The asymmetric exclusion process and Brownian excursions. Journal of statistical physics, 115(1):365–382. Durrett, R. T., Iglehart, D. L., and Miller, D. R. (1977). Weak convergence to Brownian meander and Brownian excursion. Ann. Probability, 5(1):117–129. 10 WLODZIMIERZ BRYC AND YIZAO WANG

Hirsch, F. and Yor, M. (2012). On temporally completely monotone functions for Markov processes. Probab. Surv., 9:253–286. Imhof, J.-P. (1984). Density factorizations for Brownian motion, meander and the three-dimensional Bessel process, and applications. J. Appl. Probab., 21(3):500– 510. Itˆo, K. and McKean, H. P. (1965). Diffusion processes and their sample paths. Grundlehren der Mathematischen Wissenschaften, 125. Janson, S. (2007). Brownian excursion area, Wright’s constants in graph enumer- ation, and other Brownian areas. Probab. Surv., 4:80–145. Lebedev, N. N. (1965). Special functions and their applications. Revised Eng- lish edition. Translated and edited by Richard A. Silverman. Prentice-Hall, Inc., Englewood Cliffs, N.J. Mansuy, R. and Yor, M. (2008). Aspects of Brownian motion. Universitext. Springer-Verlag, Berlin. Pitman, J. (1999). Brownian motion, bridge, excursion, and meander characterized by sampling at independent uniform times. Electron. J. Probab., 4:no. 11, 33. Pitman, J. (2006). Combinatorial stochastic processes, volume 1875 of Lecture Notes in Mathematics. Springer-Verlag, Berlin. Lectures from the 32nd Summer School on held in Saint-Flour, July 7–24, 2002, With a foreword by Jean Picard. Pitman, J. and Yor, M. (1982). A decomposition of Bessel bridges. Z. Wahrsch. Verw. Gebiete, 59(4):425–457. Revuz, D. and Yor, M. (1999). Continuous martingales and Brownian motion, volume 293 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]. Springer-Verlag, Berlin, third edition. Wang, Y. (2017). Extremes of q-Ornstein–Uhlenbeck processes. Stochastic Processes and their Applications, to appear. Yen, J.-Y. and Yor, M. (2013). Local times and excursion theory for Brownian motion, volume 2088 of Lecture Notes in Mathematics. Springer, Cham. A tale of Wiener and Itˆomeasures.

W lodzimierz Bryc, Department of Mathematical Sciences, University of Cincinnati, 2815 Commons Way, Cincinnati, OH, 45221-0025, USA. E-mail address: [email protected]

Yizao Wang, Department of Mathematical Sciences, University of Cincinnati, 2815 Commons Way, Cincinnati, OH, 45221-0025, USA. E-mail address: [email protected]