Abstract
We consider the q-totally asymmetric simple exclusion process (q-TASEP) in the stationary regime and study the fluctuation of the position of a particle. We first observe that the problem can be studied as a limiting case of an N-particle q-TASEP with a random initial condition and with particle dependent hopping rate. Then we explain how this N-particle q-TASEP can be encoded in a dynamics on a two-sided Gelfand–Tsetlin cone described by a two-sided q-Whittaker process and present a Fredholm determinant formula for the q-Laplace transform of the position of a particle. Two main ingredients in its derivation is the Ramanujan’s bilateral summation formula and the Cauchy determinant identity for the theta function with an extra parameter. Based on this we establish that the position of a particle obeys the universal stationary KPZ distribution (the Baik–Rains distribution) in the long time limit.
This is a preview of subscription content, access via your institution.
We’re sorry, something doesn't seem to be working properly.
Please try refreshing the page. If that doesn't work, please contact support so we can address the problem.





References
Aggarwal, A.: Current fluctuations of the stationary ASEP and six-vertex model. Duke Math. J. 167, 269–384 (2018)
Aggarwal, A., Borodin, A.: Phase transitions in the ASEP and stochastic six-vertex model. arXiv:1607.08684
Amir, G., Corwin, I., Quastel, J.: Probability distribution of the free energy of the continuum directed random polymer in \(1+1\) dimensions. Commun. Pure Appl. Math. 64, 466–537 (2011)
Anderson, G.W., Guionnet, A., Zeitouni, O.: An Introduction to Random Matrices. Cambridge University Press, Cambridge (2009)
Andjel, E.D.: Invariant Measures for the Zero Range Process. Ann. Probab. 10, 525–547 (1982)
Andrews, G.E., Askey, R., Roy, R.: Special Functions. Volume 71 of Encyclopedia of Mathematics and its Applications. Cambridge University Press, Cambridge (1999)
Baik, J., Ferrari, P.L., Péché, S.: Limit process of stationary TASEP near the characteristic line. Commun. Pure Appl. Math. 63, 1017–1070 (2010)
Baik, J., Rains, E.M.: Limiting distributions for a polynuclear growth model with external sources. J. Stat. Phys. 100, 523–541 (2000)
Barabási, A.L., Stanley, H.E.: Fractal Concepts in Surface Growth. Cambridge University Press, Cambridge (1995)
Barraquand, G.: A phase transition for \(q\)-TASEP with a few slower particles. Stoch. Process Their Appl. 125, 2674–2699 (2015)
Bertini, L., Giacomin, G.: Stochastic burgers and KPZ equations from particle systems. Commun. Math. Phys. 183, 571–607 (1997)
Borodin, A.: Biorthogonal ensembles. Nucl. Phys. B 536, 704–732 (1998)
Borodin, A.: Determinantal point processes. arXiv:0911.1153
Borodin, A.: Stochastic higher spin six vertex model and Madconald measures. arXiv:1608.01553
Borodin, A.: Schur dynamics of the Schur processes. Adv. Math. 228, 2268–2291 (2011)
Borodin, A., Corwin, I.: Macdonald processes. Probab. Theory Relat. Fields 158, 225–400 (2014)
Borodin, A., Corwin, I., Ferrari, P.L., Veto, B.: Height fluctuations for the stationary KPZ equation. Math. Phys. Anal. Geom. 18, 1–95 (2015)
Borodin, A., Corwin, I., Remenik, D.: Log-gamma polymer free energy fluctuations via a Fredholm determinant identity. Commun. Math. Phys. 324, 215–232 (2013)
Borodin, A., Corwin, I., Sasamoto, T.: From duality to determinants for \(q\)-TASEP and ASEP. Ann. Probab. 42, 2314–2382 (2014)
Borodin, A., Ferrari, P.L.: Anisotropic growth of random surfaces in \(2+1\) dimensions. Commun. Math. Phys. 325, 603–684 (2014)
Borodin, A., Ferrari, P.L., Prähofer, M.: Fluctuations in the discrete TASEP with periodic initial configurations and the \(\text{Airy}_1\) process. International Mathematics Research Papers (2007), rpm002
Borodin, A., Ferrari, P.L., Prähofer, M., Sasamoto, T.: Fluctuation properties of the TASEP with periodic initial configuration. J. Stat. Phys. 129, 1055–1080 (2007)
Bump, D.: The Rankin Selberg Method: A Survey, Number Theory, Trace Formulas and Discrete Groups. Academic Press, Cambridge (1989)
Burke, P.J.: The output of a queuing system. Oper. Res. 4, 699–704 (1956)
Calabrese, P., Le Doussal, P., Rosso, A.: Free-energy distribution of the directed polymer at high temperature. Eur. Phys. Lett. 90, 200002 (2010)
Carinci, G., Giardina, C., Redig, F., Sasamoto, T.: A generalized asymmetric exclusion process with \(U_q(sl_2)\) stochastic duality. Probab. Theory Relat. Fields 166, 887 (2016)
Corwin, I., Ferrari, P.L., Péché, S.: Universality of slow decorrelation in KPZ growth. Ann. Inst. H. Poincaré B 48, 134–150 (2012)
Corwin, I., Petrov, L.: The q-PushASEP: a new integrable model for traffic in \(1+1\) dimension. J. Stat. Phys. 160, 1005–1026 (2015)
Dotsenko, V.: Replica Bethe ansatz derivation of the Tracy-Widom distribution of the free energy fluctuations in one-dimensional directed polymers. J. Stat. Mech. 2010, P07010 (2010)
Ferrari, P.: TASEP hydrodynamics using microscopic characteristics. Probab. Surv. 15, 1–27 (2018)
Ferrari, P.A., Fontes, L.R.G.: The net output process of a system with infinitely many queues. Ann. Appl. Probab. 4, 1129–1144 (1994)
Ferrari, P.L., Spohn, H.: Scaling limit for the space–time covariance of the stationary totally asymmetric simple exclusion process. Commun. Math. Phys. 265, 1–44 (2006)
Ferrari, P.L., Spohn, H., Weiss, T.: Brownian motions with one-sided collisions: the stationary case. Electron. J. Probab. 20(69), 1–41 (2015)
Ferrari, P.L., Vető, B.: Tracy–Widom asymptotics for q-TASEP. Ann. Inst. H. Poincaré Probab. Stat. 4, 1465–1485 (2015)
Forrester, P.J.: Log Gases and Random Matrices. Princeton University Press, Princeton (2010)
Forster, D., Nelson, D.R., Stephen, M.J.: Large-distance and long-time properties of a randomly stirred fluid. Phys. Rev. A 16, 732–749 (1977)
Frobenius, G.: Ueber die elliptischen Functionen zweiter Art. J. für die reine und angew. Math. 93, 732–749 (1882)
Gasper, G., Rahman, M.: Basic Hypergeometric Series. Cambridge University Press, Cambridge (2004)
Goncalves, P.: On the asymmetric zero-range in the rarefaction fan. J. Stat. Phys. 154, 1074–1095 (2014)
Gwa, L.-H., Spohn, H.: Six-vertex model, roughened surfaces, and an asymmetric spin Hamiltonian. Phys. Rev. Lett. 68, 725–728 (1992)
Horn, R.A., Johnson, C.R.: Matrix Analysis, 2nd edn. Cambridge University Press, Cambridge (2012)
Imamura, T., Mucciconi, M., Sasamoto, T.: In preparation
Imamura, T., Sasamoto, T.: Free energy distribution of the stationary O’Connell-Yor polymer model. J. Phys. A 50, 285203 (2017)
Imamura, T., Sasamoto, T.: Fluctuations of the one-dimensional polynuclear growth model with external sources. Nucl. Phys. B 699, 503–544 (2004)
Imamura, T., Sasamoto, T.: Current moments of 1D ASEP by duality. J. Stat. Phys. 142, 919–930 (2011)
Imamura, T., Sasamoto, T.: Exact solution for the stationary KPZ equation. Phys. Rev. Lett. 108, 190603 (2012)
Imamura, T., Sasamoto, T.: Stationary correlations for the 1D KPZ equation. J. Stat. Phys. 150, 908–939 (2013)
Imamura, T., Sasamoto, T.: Determinantal structures in the O’Connell–Yor directed random polymer model. J. Stat. Phys. 163, 675–713 (2016)
Johansson, K.: Shape fluctuations and random matrices. Commun. Math. Phys. 209, 437–476 (2000)
Johansson, K.: Non-intersecting paths, random tilings and random matrices. Probab. Theory Relat. Fields 123, 225–280 (2002)
Johansson, K.: Discrete polynuclear growth and determinantal processes. Commun. Math. Phys. 242, 277–329 (2003)
Kajihara, Y., Noumi, M.: Mutiple elliptic hypergeometric series: an approach from the Cauchy determinant. Indag. Mathem. N.S. 14, 395–421 (2003)
Kardar, M., Parisi, G., Zhang, Y.C.: Dynamic scaling of growing interfaces. Phys. Rev. Lett. 56, 889–892 (1986)
Katori, M.: Elliptic determinantal process of type A. Probab. Theory Relat. Fields 162, 637–677 (2015)
Katori, M.: Elliptic Bessel processes and elliptic Dyson models realized as temporally inhomogeneous processes. J. Math. Phys. 57, 103302 (2016)
Kipnis, C., Landim, C.: Scaling Limits of Interacting Particle Systems. Springer, Berlin (1999)
Ledoux, M.: A Remark on Hypercontractivity and Tail Inequalities for the Largest Eigenvalues of Random Matrices. In: Séminaire de Pobalilités XXXVII, vol. 1832 of Lecture Notes in Mathematics. Springer, Berlin (2003)
Liggett, T.M.: An infinite particle system with zero range interactions. Ann. Probab. 1, 240–253 (1973)
Liggett, T.M.: Interacting Particle Systems. Springer, Berlin (1985)
Liggett, T.M.: Continuous Time Markov Processes: An Introduction (graduate studies in mathematics). American Mathematical Society, Providence (2010)
Macdonald, I.G.: Symmetric Functions and Hall Polynomials, 2nd edn. Oxford University Press, Oxford (1995)
Matveev, K., Petrov, L.: \(q\)-randomized Robinson–Schensted–Knuth correspondences and random polymers. Ann. Inst. Henri Poincaré D 4, 1–123 (2017)
Mehta, M.L.: Random Matrices, 3rd edn. Elsevier, Amsterdam (2004)
O’Connell, N.: Directed polymers and the quantum Toda lattice. Ann. Probab. 40, 437–458 (2012)
Orr, D., Petrov, L.: Stochastic higher spin six vertex model and q-TASEPs. Adv. Math. 317, 473–525 (2017)
Povolotsky, A.M.: Bethe ansatz solution of zero-range process with nonuniform stationary state. Phys. Rev. E 69, 061109 (2004)
Prähofer, M., Spohn, H.: Current fluctuations for the totally asymmetric simple exclusion process. In: Sidoravicius, V. (ed.) In and Out of Equilibrium, vol. 51 of Progress in Probability, pp. 185–204 (2002)
Prähofer, M., Spohn, H.: Scale invariance of the PNG droplet and the Airy process. J. Stat. Phys. 108, 1071–1106 (2002)
Prähofer, M., Spohn, H.: Exact scaling functions for one-dimensional stationary KPZ growth. J. Stat. Phys. 115, 255–279 (2004)
Reed, M., Simon, B.: Methods of Modern Mathematical Physics I: Functional Analysis. Academic Press, New York (1980)
Sasamoto, T.: Spatial correlations of the 1D KPZ surface on a flat substrate. J. Phys. A 38, L549–L556 (2005)
Sasamoto, T., Spohn, H.: Exact height distributions for the KPZ equation with narrow wedge initial condition. Nucl. Phys. B 834, 523–542 (2010)
Sasamoto, T., Spohn, H.: The crossover regime for the weakly asymmetric simple exclusion process. J. Stat. Phys. 140, 209–231 (2010)
Sasamoto, T., Spohn, H.: Universality of the one-dimensional KPZ equation. Phys. Rev. Lett. 834, 523–542 (2010)
Sasamoto, T., Wadati, M.: Exact results for one-dimensional totally asymmetric diffusion models. J. Phys. A 31, 6057–6071 (1998)
Schütz, G.M.: Duality relations for asymmetric exclusion processes. J. Stat. Phys. 86, 1265–1287 (1997)
Soshnikov, A.: Determinantal random point fields. Russ. Math. Surv. 55, 923–975 (2000)
Spohn, H.: Large Scale Dynamics of Interacting Particles. Springer, Berlin (1991)
Spohn, H.: KPZ scaling theory and the semi-discrete directed polymer model. In: MSRI proceedings (2013)
Stade, E.: Archimedean \(L\)-factors on \(\text{ GL }(\text{ n })\times \text{ GL }(\text{ n })\) and generalized Barnes integrals. Isr. J. Math. 127, 201–219 (2002)
Stanley, R.P.: Enumerative Combinatorics 2. Springer, Berlin (1999)
Takeuchi, K.A., Sano, M.: Evidence for geometry-dependent universal fluctuations of the Kardar–Parisi–Zhang interfaces in liquid-crystal turbulence. J. Stat. Phys. 147, 853–890 (2010)
Takeuchi, K.A., Sano, M.: Growing interfaces of liquid crystal turbulence: universal scaling and fluctuations. Phys. Rev. Lett. 104, 230601 (2010)
Tracy, C.A., Widom, H.: Level-spacing distributions and the Airy kernel. Commun. Math. Phys. 159, 151–174 (1994)
Tracy, C.A., Widom, H.: On orthogonal and symplectic matrix ensembles. Commun. Math. Phys. 177, 727–754 (1996)
Tracy, C.A., Widom, H.: Correlation functions, cluster functions, and spacing distributions for random matrices. J. Stat. Phys. 92, 809–835 (1998)
Tracy, C.A., Widom, H.: Asymptotics in ASEP with step initial condition. Commun. Math. Phys. 209, 129–154 (2009)
Whittaker, E.T., Watson, G.N.: A Course of Modern Analysis. Cambridge University Press, Cambridge (1927)
Acknowledgements
Part of this work was performed during the stay at KITP in Santa Barbara, USA. This research was supported in part by the National Science Foundation under Grant No. NSF PHY11-25915. The work of T.I. and T.S. is also supported by JSPS KAKENHI Grant Numbers JP25800215, JP16K05192 and JP25103004, JP14510499, JP15K05203, JP16H06338, JP18H01141, JP18H03672, respectively.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A: Some q-functions and q-formulas
In this appendix, we summarize a few q-notations, q-functions and q-formulas. The first is the q-Pochhammer symbol, or the q shifted factorial. For \(|q|<1\) and \(n\in \mathbb {N}\),
The q-binomial theorem will be useful in various places in the discussions,
In particular the \(a=0\) case appears in many applications. Another q-binomial formula reads (see e.g. Cor.10.2.2.(b) in [6])
There is yet another version of the q-binomial theorem (see e.g. Cor.10.2.2.(c) in [6]),
The q-exponential function, denoted as \(e_q(z)\) is defined to be
The second equality is by the above q-binomial theorem (A.2). From the series expansion expression, it is easy to see that this tends to the usual exponential function in the \(q\rightarrow 1\) limit.
The q-Gamma function \(\Gamma _q(x)\) is defined by
The q-digamma function is defined by \(\Phi _q(z)=\partial _z\log \Gamma _q(z)\). In the \(q\rightarrow 1\) limit, they tends to the usual \(\Gamma \) function and the digamma function respectively.
Ramanujan’s summation formula (cf [6, p502], [38, p138]) is a two-sided generalization of the above q-binomial theorem (A.2). For \(|q|<1, |b/a|<|z|<1\),
Appendix B: q-Whittaker functions and q-Whittaker process
In this appendix, we introduce the ordinary (skew) q-Whittaker functions labeled by partitions and the q-Whittaker process and then discuss some of their properties. Most of them are standard [61] but some of them are reformulated for the applications in the main text.
First a partition of length \(n\in \mathbb {N}\) is an n-tuple \(\lambda =(\lambda _1,\ldots ,\lambda _n)\) with \(\lambda _j\in \mathbb {Z}_+, 1\le j\le n\) s.t. \(\lambda _1\ge \ldots \ge \lambda _n\). The set of all partitions of length n is denoted by \(\mathcal {P}_n\). A partition \(\lambda \) can also be represented (and identified) as a Young diagram with n rows of length \(\lambda _1,\ldots ,\lambda _n\). For two partitions \(\lambda \in \mathcal {P}_n,\mu \in \mathcal {P}_m\) s.t. \(m\le n\) and \(\lambda _i-\mu _i\ge 0, 1\le i\le n\) (with the understanding \(\mu _i\equiv 0, m<i\le n)\), a pair \((\lambda ,\mu )\) is called a skew diagram and is denoted by \(\lambda /\mu \). The transpose \(\lambda '\) is the partition of length \(\lambda _1\) defined as \(\lambda _i'=\#\{j\in \mathbb {Z}_+ | \lambda _j\ge i\},1\le i\le \lambda _1\).
The Gelfand–Tsetlin (GT) cone for partitions, denoted by \(\mathbb {G}_N^{(0)},N\in \mathbb {N}\) here, is defined by
An element \(\underline{\lambda }_N\in \mathbb {G}_N^{(0)}\) is called a Gelfand–Tsetlin pattern. Next we define the (skew) q-Whittaker functions.
Definition B.1
Let \(\lambda \in \mathcal {P}_n,\mu \in \mathcal {P}_{n-1}\) be two partitions of order n and \(n-1\) respectively and a an indeterminate. The skew q-Whittaker function (with 1 variable) is defined as ([61, VI.7, (7.14) and Ex. 2.])
Using this, for a partition \(\lambda \in \mathcal {P}_N\) and \(a=(a_1,\ldots ,a_N)\) being N indeterminates, we define the q-Whittaker function with N variables as (cf. [61, VI.7, (7.9’)])
Here the sum is over the Gelfand–Tsetlin cone \(\mathbb {G}_N^{(0)}\) with the condition \(\lambda ^{(N)}=\lambda \) and \(\lambda ^{(0)}=\phi \).
It is known that the q-Whittaker function \(P_\lambda (a)\) forms a basis of \(\Lambda _N\), the space of N-variable symmetric polynomials with coefficients being rational functions in q (cf [61, VI]). There is an inner product \(\langle ~,~ \rangle \) in this space for which \(P_\lambda \)’s are orthogonal ([61, VI (6.19)]):
Using this, we also introduce \(Q_\lambda (x)\).
Definition B.2
For \(\lambda \in \mathcal {P}_N\) and for \(x=(x_1,\ldots ,x_N)\), we define ([61, VI (4.11), (4.12)])
Note that \(P_\lambda \) and \(Q_\lambda \) are orthonormal: \(\langle P_\lambda , Q_\mu \rangle = \delta _{\lambda ,\mu }\).
If we take a sum over all partitions of length N of a product of \(P_\lambda \) and \(Q_\lambda \), the following Cauchy identity for the q-Whittaker functions holds [61, VI(4.13), (2.5)],
Let us rewrite this identity in a form used in the main text. First one notices that one can express \(P_{\lambda }(x)\) in a form,
where \(X=x_1 \ldots x_N\) and \(\ell =(\ell _1,\ldots ,\ell _{N-1})\) with \(\ell _j=\lambda _{j}-\lambda _{j+1}, 1\le j\le N-1\) and \(R_\ell \in \Lambda _N\). This is seen as follows. Since \((\lambda ^{(1)},\ldots , \lambda ^{(N)})\) in (B.3) is an element of \(\mathbb {G}_N\), the order of each term in the sum in (B.3) is bigger than or equal to \(\lambda ^{(N)}_N\) and one can factor out this lowest order common factor. Then a partition \(\lambda \) is uniquely determined by \(\lambda ^{(N)}_N\) and \(\ell _j,j=1,...,N-1\) and the coefficient of each term depends only on \(\ell _j,j=1,\ldots ,N-1\). Then we have (B.7). We do not write explicitly the form of the function \(R_\ell (x)\) since it is not necessary in the following discussion. In terms of \(R_\ell \), the Cauchy identity (B.6) is equivalent to
with \(Y=y_1 \ldots y_N\). In fact if one substitutes (B.7) into lhs of (B.6) and uses (B.8) and the q-binomial theorem (A.2) with \(a=0\), we get rhs of (B.6).
There is another inner product \(\langle \cdot ,\cdot \rangle '\) in \(\Lambda _N\) called the torus scaler product: For the N-variable functions, \(f(z), g(z)\in \Lambda _N\), the torus scalar product is defined by [61, VI (9.10)]
where
is the q-Sklyanin measure. The q-Whittaker functions of N variables are known to satisfy the following orthogonality relations with respect to this inner product [61],
Here we rewrite the orthogonality relation (B.11). From the definition (B.3), we find that \(P_{\lambda }(e^{i\theta })=P_{\lambda }(e^{i\theta _1},\ldots ,e^{i\theta _N})\) can be expressed as
where \(\bar{\theta }=(\theta _1+\cdots +\theta _N)/N\) and \(\bar{\lambda }=(\lambda _1+\cdots +\lambda _N)/N\) represent the barycentric coordinates while relative coordinates are denoted by \(\tilde{\theta }=(\tilde{\theta }_1,\ldots ,\tilde{\theta }_{N-1})\) and \(\ell =(\ell _1,\ldots ,\ell _{N-1})\) with \(\tilde{\theta }_j=\theta _j-\theta _{j+1}\) and \(\ell _j=\lambda _j-\lambda _{j+1}, 1\le j\le N-1\). Considering this we see that (B.11) can be rewritten as, with \(m_j=\mu _j-\mu _{j+1}, 1\le j\le N-1\)
Hence for the difference variables \(\ell ,m\in \mathbb {Z}_{\ge 0}^{N-1}\), the orthogonality relation (B.11) can be restated as
One finds a representation of \(Q_\lambda \) using torus scalar product,
Lemma B.3
For \(y\in \mathbb {R}^N,\lambda \in \mathcal {P}_N\),
where 1 / z in \(P_\lambda \) is a shorthand notation for \((1/z_1,\ldots ,1/z_N)\).
Proof
It immediately follows from the Cauchy identity (B.6) and the orthogonality (B.11). \(\square \)
We can consider a generalization of the function \(Q_\lambda \) by modifying \(\Pi (z;y)\) in (B.15) to
with \(\alpha _j,\beta _j, \gamma \ge 0\) for \(j=1,\ldots ,M\). We call this \(Q_{\lambda } \left( \{\alpha ,\beta ,\gamma \}\right) \). The case where only some of \(\alpha _j\)’s (resp. \(\beta _j\)’s or \(\gamma \)) are positive is called the \(\alpha \)-specialization (resp. \(\beta \)-specialization or Plancherel specialization) in [16]. Note that the (B.16) for the \(\alpha \)-specialization is nothing but \(\Pi (z;\alpha )\) in (B.6). The (ascending) q-Whittaker process corresponding to this generalization is defined by
For instance we define \(Q^{(\beta )}_{\lambda }(y),y\in \mathbb {R}^M,M\in \mathbb {N}\) by (B.15) with \(\Pi (z;y)\) replaced by \(\Pi _M^{(\beta )}(z;y):=\Pi (z;{\alpha =0,\beta =y,\gamma =0})= \prod _{i=1}^N\prod _{j=1}^M(1+z_i y_j)\). We state two properties of the q-Whittaker functions with this \(\beta \)-specialization, which will be also useful to discuss our q-TASEP with a random initial condition (cf [61, VI 7]).
where in (B.19), \(Q^{(\beta )}_{\lambda /\mu }(r)\) with one variable is defined as
and we introduced a notation \(\{r\}_M=(r_1,\ldots ,r_M)\) to consider a change of M below.
In the rest of this appendix we consider some properties of the q-Whittaker process corresponding to this pure \(\beta \) specialization.
Definition B.4
For \(M\in \mathbb {N}\),
As mentioned above, the positivity of (B.21) is known. This process describes a distribution function at time M of a discrete time Markov process on \(\mathbb {G}_N^{(0)}\) described by the following Kolmogorov forward equation.
Proposition B.5
where the transition matrix \(G_M(\underline{\mu }_N,\underline{\lambda }_N)\) is
with \(\Delta (\lambda ,\mu )=\sum _{\kappa }P_{\kappa /\lambda }(a)Q^{(\beta )}_{\kappa / \mu }(r_M)\).
Remark
It is easy to see that each factor in (B.23) is a transition probability matrix from the definition of \(\Delta (\lambda ,\mu )\).
Proof
We rewrite \(G_{M}(\underline{\mu }_N,\underline{\lambda }_N)\) as
where
Using (B.18) and (B.19), we have
Thus we have
\(\square \)
Appendix C: Two lemmas regarding the Airy kernel and the GUE Tracy–Widom limit
Here we provide two lemmas for establishing the GUE Tracy–Widom limit when a kernel of a specific form is given. The Hermite kernel for the GUE is a simplest example. For GUE, one can also use a bound due to Ledoux, which holds for GUE and simplifies the proof for the case (cf [4, 57]), but in other applications such a bound is not available. Our lemmas do not rely on such extra information but focus only on properties of the kernel. The essential part of the arguments are given in [21, 49] but we reformulate and generalize them in a way which would be suited for various applications. The Airy kernel \(K(\xi ,\zeta )\) is defined as
Our discussions below are given in a continuous setting but can be also applied to a discrete setting as well.
Lemma C.1
Suppose that we have a kernel of the form,
where \(\varphi _n,\psi _n,n\in \mathbb {N}\) are complex functions on \(\mathbb {R}\). Assume that the functions \(\varphi _n(x)\) satisfy the followings for some \(a\in \mathbb {R}, \gamma>0, c>0\).
-
(a)
For \(\forall M>0\) and \(\forall L>0\),
$$\begin{aligned} \lim _{N\rightarrow \infty } \gamma _1N^{1/3} \varphi _{N-cN^{1/3}\lambda }(aN+\gamma N^{1/3}\xi ) = \mathrm{Ai}(\xi +\lambda ), \end{aligned}$$(C.3)uniformly for \(|\xi |<L\) and for \(\lambda \in [0,M]\).
-
(b)
For \(\forall L>0, \forall \epsilon >0\) and N large enough,
$$\begin{aligned} \gamma _1N^{1/3} \varphi _{N-cN^{1/3}\lambda }(aN+\gamma N^{1/3}\xi ) \le c_{b} e^{-(\xi +\lambda )}, \end{aligned}$$(C.4)for \(\xi ,\lambda \) satisfying \(\lambda >0, L\le |\xi +\lambda | \le \epsilon N^{2/3}\) and for some constant \(c_b\).
-
(c)
For \(\forall \epsilon >0\) and N large enough,
$$\begin{aligned} \gamma _1N^{1/3} \varphi _{N-cN^{1/3}\lambda }(aN+\gamma N^{1/3}\xi ) \le c_{c} e^{-(\xi +\lambda )} , \end{aligned}$$(C.5)for \(\xi ,\lambda \) satisfying \(\lambda>0, |\xi +\lambda |> \epsilon N^{2/3}\) and for some constants \(c_c\).
The functions \(\psi _n\) are also assumed to satisfy the same conditions (a), (b), (c) with the parameters \(a\in \mathbb {R}, c>0\), same as above and \(\gamma _2>0\). Define the rescaled kernel, with \(\gamma _1\gamma _2=c\gamma \),
Then we have
-
(i)
For \(\forall L>0\),
$$\begin{aligned} \lim _{N\rightarrow \infty } K_N(\xi ,\zeta ) = K(\xi ,\zeta ), \quad \text {uniformly (in}\ N)\ \text {on} \quad [-L,L]^2. \end{aligned}$$(C.7) -
(ii)
For \(\forall L>0\),
$$\begin{aligned} {|} K_N(\xi ,\zeta )| \le c_1 e^{-\max (0,\xi )-\max (0,\zeta )} , \quad \text {uniformly (in}\ N\text {) for} \quad \xi ,\zeta \ge -L \end{aligned}$$(C.8)for some constant \(c_1\).
Proof
First we divide the sum over n in (C.2) as \(\sum _{n=0}^{N-1} = \sum _{n=0}^{[N-cN^{1/3}M]} + \sum _{n=[N-cN^{1/3}M]+1}^{N-1}\), where [x] is the maximum integer which is less than or equal to \(x\in \mathbb {R}\). For the second sum, we have, due to the uniform convergence in (a),
for \(\xi ,\zeta \in [-L,L]\) for \(\forall L>0\). The first sum is, due to (c), (d), bounded by \(\int _M^\infty e^{-2\lambda } d\lambda = e^{-2M}/2\). By taking the \(M\rightarrow \infty \) limit, we have (C.7) in (i). The bound in (ii) easily follows from the uniform convergence in (a) (Note \(\mathrm{Ai}(\xi )\) is bounded by a constant as \(\xi \rightarrow -\infty \) and by \(e^{-\xi }\) as \(\xi \rightarrow \infty \)) combined with the bounds in (b), (c). \(\square \)
Lemma C.2
Suppose that a sequence of kernel \(K_N: \mathbb {R}\times \mathbb {R}\rightarrow \mathbb {C}, N\in \mathbb {N}\) satisfies
-
(i)
For \(\forall L>0\),
$$\begin{aligned} \lim _{N\rightarrow \infty } K_N(\xi ,\zeta ) = K(\xi ,\zeta ), \quad \text {uniformly in} \quad [-L,L]^2. \end{aligned}$$(C.10) -
(ii)
For \(\forall L>0\),
$$\begin{aligned} {|} K_N(\xi ,\zeta )| \le c_1 e^{-\max (0,\xi )-\max (0,\zeta )} , \quad \text {uniformly (in}\ N\text {) for} \quad \xi ,\zeta \ge -L, \end{aligned}$$(C.11)for some constant \(c_1\).
In addition suppose that a (sequence of ) function \(f_N: \mathbb {R}\rightarrow \mathbb {R},N\in \mathbb {N}\) satisfy
-
(iii)
The functions \(f_N,N\in \mathbb {N}\) are uniformly bounded and converges to \(1_{(s,\infty )}\) for a \(s\in \mathbb {R}\) in the \(L^1(\mathbb {R})\) norm,
$$\begin{aligned} \lim _{N\rightarrow \infty } \int _\mathbb {R}|f_N (\xi ) - 1_{(s,\infty )} (\xi )|d\xi =0. \end{aligned}$$(C.12) -
(iv)
For \(\forall L(>|s|)\),
$$\begin{aligned} \sqrt{|f_N(\xi )f_N(\zeta )|} |K_N(\xi ,\zeta )| \le e^{\xi +\zeta } \end{aligned}$$(C.13)and
$$\begin{aligned} \lim _{N\rightarrow \infty } \sqrt{|f_N(\xi )f_N(\zeta )|} |K_N(\xi ,\zeta )| =0 \end{aligned}$$(C.14)uniformly (in N) for \(\xi ,\zeta \) satisfying \(\xi \le -L\) or \(\zeta \le -L\).
Then for \(\forall s\in \mathbb {R}\),
$$\begin{aligned} \lim _{N\rightarrow \infty } \det (1-f_N K_N)_{L^2(\mathbb {R})} = F_2(s). \end{aligned}$$(C.15)
Remark
The condition (i), (ii) are the same as (i)(ii) in Proposition C.1. For \(f_N= 1_{(s,\infty )},\forall N\in \mathbb {N}\), (iii)(iv) are trivially satisfied. Hence the kernel \(K_N\) from Proposition C.1 with this special \(f_N\) automatically satisfies the above four conditions. This special case appears in many applications, for example in GUE and TASEP.
Proof
In this proof \(c_i,i=1,2,\ldots \) are some constants. We first prove the \(f_N= 1_{(s,\infty )},\forall N\in \mathbb {N}\) case. By the remark above, we will show that for a kernel satisfying (i)(ii) in Proposition C.1 we have
The general case will be treated by considering the difference to this special case later in the proof. For a given \(s\in \mathbb {R}\), take \(L(>|s|)\). By the Hadamard’s inequality,
which holds for general \(n\times n\) matrix [41], and (ii), we have for \(\xi ,\zeta \ge -L\),
and hence
Take \(\epsilon >0\). By (C.19), the Fredholm expansion,
converges and there exists \(l\in \mathbb {N}\) s.t.
By (ii), we have
where in the second inequality we used (C.18). Hence, for L large enough,
uniformly in N. Combining (C.21) and (C.23), we have
By the uniform convergence (i), for N large enough,
By the same argument as to get (C.24), we have
Combining (C.24), (C.25), (C.26), we have
This completes the proof for the \(f_N=1_{(s,\infty )},\forall N\in \mathbb {N}\) case.
Next we consider the general \(f_N\) case. It is enough to show
for large enough N. Indeed, taking the sum over k and using the triangular inequality on the Fredholm explansion, the bound (C.28) gives
Then taking \(N\rightarrow \infty \) limit, we find
This completes the proof for general \(f_N\) case. Coming back to the proof of (C.28), in the second term on lhs, the \(\infty \) as the upper limit of the integrals can be replaced by a large L with an error of the form \(c_1^k k^{k/2+1}\epsilon \) as in (C.22). Similarly, in the first term on lhs of (C.28), the \(\infty \) as the upper limit of the integrals can be replaced by L with an error of the form \(c_2^k k^{k/2+1}\epsilon \) and the \(-\infty \) as the lower limit by \(-L\) with an error of the form \(c_3^k k^{k/2+1}\epsilon \) due to (iv). Combining these, we can replace the integrals in (C.28) within \((-L,L]\) for large L with an error \(c_4^k k^{k/2+1}\epsilon \). By a similar argument as to get (C.22), we have
Using (C.18) and (iii), one observes that the integral over \(\xi _i, i\ne j\) in the first sum is finite. By using (iii), we see that the integral over \(\xi _j\) in the first sum is bounded for large N by \( c_5^k k^{k/2+1} \epsilon \). Similarly for the second term in (C.31), one first sees \(| \prod _{i=1}^k f_N(\xi _i)-1 | \le c_6 \sum _{i=1}^k |f_N(\xi _i)-1|\) since \(f_N\) is bounded. Then using (iii) and (C.18) one sees that it is bounded by \( c_7^k k^{k/2+1} \epsilon \) for large N. Combining all of the above, we get (C.28). \(\square \)
Appendix D: Inverse q-Laplace transforms
In this appendix we discuss inversion formulas of the q-Laplace transform. In Appendix B of [43], we discussed a few inversion formulas for the ordinary Laplace transform, which is summarized as follows. First the Laplace transform is defined as
(In [43] we used the notation \(\tilde{\varphi }\). Here we use \(\hat{\varphi }\) to keep the former for the transformed one, see below.) When \(\varphi (x)\) is a (probability) distribution function on \((0,\infty )\), the Laplace transform is analytic for \(\mathrm{Re}\,u>0\). The usual inversion formula is
where \(\delta \) should be taken so that the singularities of \(\hat{\varphi }\) are to the left of the integration contour. If \(\varphi \) is associated with a random variable X, we have \(G(u):=\langle e^{-uX} \rangle = u\hat{\varphi }(u)\). The formula can be restated for the random variable \(Y=\log X\). For instance for the distribution \(F(y)=\mathbb {P}[Y\le y] = \mathbb {P}[X\le e^y] = \varphi (e^y)\), (D.1) is rewritten as
For discussing the distribution of the O’Connell–Yor polymer model, the following inversion formula was useful.
Proposition D.1
For a random variable Y, set \(G(u) = \langle e^{-ue^Y}\rangle \). The distribution function of Y, which decays as \(F(y) = e^{a y}, a>0\) as \(y\rightarrow -\infty \), is recovered from G(u) as
where \(0<\delta <a\). The corresponding density function \(f(y)=F'(y)\), if it exists, is given by
The formulas discussed in this appendix are q-analogues of the above.
Suppose we have a function \(f(n),n\in \mathbb {Z}\) and denote by \(\tilde{f}_q(z)\) its q-Laplace transform,
In this appendix, we mainly consider the case where \(f(n),n\in \mathbb {Z}\) is a discrete probability density function, i.e., \(f(n)\ge 0, \sum _{n\in \mathbb {Z}} f(n)=1\), for which the q-Laplace transform converges and analytic except \(z\ne q^n, n\in \mathbb {Z}\). By using the fact that the q-exponential function tends to the usual exponential function in a \(q\rightarrow 1\) limit [see the comment after (A.5)], one sees that this formula goes to (D.1) as
where in the limit (\(\rightarrow \)) we set \(n=y/\log q\) and took \(q\rightarrow 1\). An inverse formula for the q-Laplace transform is given by
where \(\gamma \) is a osing \(\mathbb {R}_+\) clockwise, see Fig. 6. Because of the factor \((q^{n+1}z;q)_\infty \), the poles at \(z=q^{-k},k=n+1,n+2,\ldots \) vanish and the contour can be taken to be around the poles at \(z=1,2,\ldots , n\), but the infinite contour in Fig. 6 has an advantage that it can be used for any n.
In the \(q\rightarrow 1\) limit, with \(n=y/\log q\) and the change of variable \(z=-(1-q)u\), tends to (D.2) as
There is also an inversion formula corresponding to (D.5).
Proposition D.2
When \(f(n),n\in \mathbb {Z}\) is a discrete probability density function,
Here \(R_k(\tilde{f}_q)\) is the residue of \(\tilde{f}_q(z)\) at the pole \(z=q^{-k}\) and \(C_0\) is a small contour around the origin.
Proof
It is equivalent to showing
for \(x=e^{i\theta },\theta \in (-\pi ,\pi ]\). In fact considering the contour integral around the origin of both sides of (D.11) divided by \(x^{n+1}\), we get (D.10). By (D.8), this is further equivalent to showing
Taking the poles at \(z=q^{-k},k\in \mathbb {Z}\), one can rewrite lhs of (D.12) as
The last sum in (D.13) can be taken owing to the q-binomial theorem (A.2),
Substituting this into (D.13), we get (D.12). \(\square \)
The last sum in (D.10) formally seems to be a consequence of taking poles at \(z=q^{-k},k\in \mathbb {Z}\) of the integral, \(\int _{\gamma } \frac{dz}{2\pi i z} z^{\xi } \tilde{f}_q(z)\) (though this is not true since \(z^{\xi }\) has a cut along \(\mathbb {R}_+\)). Setting \(x=q^{-\xi }\) and taking the \(q\rightarrow 1\) limit of this and recalling the factor \((q;q)_\infty /(qx;q)_\infty \) can be written in terms of the q-Gamma function (A.6) which tends to the Gamma function as \(q\rightarrow 1\), one would observe that (D.10) may be regarded as a q-analogue of (D.5).
As a corollary of proposition D.2, we get a formula for the distribution function \(F(n)=\mathbb {P}[X\le n], n\in \mathbb {N}\). It reads
For the density function, \(f(n):=F(n)-F(n-1)\), this gives (D.10). (D.15) is an analogue of (D.4). If we introduce the generating function \(Q(x)=\sum _{n\in \mathbb {Z}} F(n) x^n\), clearly
and (D.15) is equivalent to
Suppose further that a random variable X having the density \(f(n),n\in \mathbb {Z}\) is written as \(X=X_0+\chi \) where \(X_0\) and \(\chi \) are independent random variables on \(\mathbb {Z}\). By the independence, the three quantities, Q(x), \(Q_0(x)=\sum _{n\in \mathbb {Z}} \mathbb {P}[X_0 \le n] x^n, g(x)=\sum _{n\in \mathbb {Z}} \mathbb {P}[\chi = n] x^n\), are related as \(Q(x) = g(x)Q_0(x)\). Combining this and (D.16), (D.17), we find a formula for the distribution of \(X_0\),
Noting \(\tilde{f}_q(\zeta )= \langle \frac{1}{(\zeta q^X;q)}\rangle \), this is a formula which gives the distribution function of \(X_0\) in terms of the q-Laplace transform of X.
Rights and permissions
About this article
Cite this article
Imamura, T., Sasamoto, T. Fluctuations for stationary q-TASEP. Probab. Theory Relat. Fields 174, 647–730 (2019). https://doi.org/10.1007/s00440-018-0868-3
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00440-018-0868-3
Keywords
- Asymmetric simple exclusion process
- Kardar–Parisi–Zhang universality
- Integrable probability
- Macdonald processes
- Frobenius determinant
Mathematics Subject Classification
- 82C22
- 60K35
- 33D52
