Abstract
We show that the stationary density fluctuations of exclusion processes with long jumps, whose rates are of the form \(c^\pm |y-x|^{-(1+\alpha )}\) where \(c\pm \) depends on the sign of \(y-x\), are given by a fractional Ornstein–Uhlenbeck process for \(\alpha \in (0,\frac{3}{2})\). When \(\alpha =\frac{3}{2}\) we show that the density fluctuations are tight, in a suitable topology, and that any limit point is an energy solution of the fractional Burgers equation, previously introduced in Gubinelli and Jara (Stoch Partial Differ Equ Anal Comput 1(2):325–350, 2013) in the finite volume setting.
Similar content being viewed by others
Notes
In the physics literature, a stationary state is what in probability is called an invariant measure, and an equilibrium state corresponds to an invariant measure which is in addition reversible.
The fractional Burgers equation considered in [18] is defined in finite volume, and therefore the spatial scaling changes its domain.
We write \({\mathbb N} = \{1,2,\ldots \}\) and \({\mathbb N}_0 = \{0\}\bigcup {\mathbb N}\).
This property assures us that the application \(f \mapsto {\mathscr {Y}}_t(f)\) from \({\mathbb {S}}({\mathbb R}) \subseteq L^2({\mathbb R})\) into \(L^2(\Omega )\) is uniformly continuous, and equicontinuous on t. Any other property that would ensure this equicontinuity would serve as a substitute to stationarity; nonetheless stationarity will be a consequence of other hypotheses needed to prove our main results, so we will not give too much attention to this point.
Note that \({\mathscr {Y}}_t({\mathscr {L}}f)\) is well defined up to a set of null probability. This set depends on the choice of the function f and therefore we can not a priori think about \({\mathscr {Y}}_t({\mathscr {L}} f)\) as a distribution-valued random variable. Nevertheless, stationarity and Proposition 2.4 can be used to show that \({\mathscr {Y}}_t({\mathscr {L}} f)\) is indeed a distribution-valued random variable.
The theory of regularity structures of [20, 21] provides a uniqueness criterion for the stochastic Burgers equation (the case \(\alpha =2\)) and in principle this criterion could be extended to \(\alpha \) strictly larger than 3 / 2, at least in finite volume. The case \(\alpha = 3/2\), which is the relevant one for these notes, seems to be out of the reach of the current state of this theory.
Along these notes, we will denote by \(c_i(\rho )\) various constants which depend only on \(\rho \). The exact value of these constants will not be important; only the fact that they depend only on \(\rho \).
Note that the support of \(F_i\) has at most diameter \(\ell _i\).
We actually proved this with m replaced by \(\sum _{y=1}^{K_n-1} y a(y)\), but this last sum converges to m as \(n \rightarrow \infty \).
In order to avoid heavy notation, we decided to restrict the computations in previous sections to functions not depending on time.
The generalization of the arguments to time-dependent test functions can be done as explained in Sect. 5.1.
References
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)
Bernardin, C., Gonçalves, P., Jara, M.: 3/4-Fractional superdiffusion in a system of harmonic oscillators perturbed by a conservative noise. Arch. Ration. Mech. Anal. 220(2), 505–542 (2016)
Bernardin, C., Gonçalves, P., Sethuraman, S.: Occupation times of long-range exclusion and connections to KPZ class exponents. Probab. Theory Relat. Fields 166(1), 365–428 (2016)
Borodin, A.N., Ibragimov, I.A.: Limit theorems for functionals of random walks. In: Proceedings of the Steklov Institute of Mathematics. vol. 195, no. 2. American Mathematical Society (1995)
Breiman, L.: Probability, Classics in Applied Mathematics, 2nd edn. SIAM, Philadelphia (1992)
Brox, T., Rost, H.: Equilibrium fluctuations of stochastic particle systems: the role of conserved quantities. Ann. Probab. 12(3), 742–759 (1984)
Caputo, P.: Uniform Poincaré inequalities for unbounded conservative spin systems: the non-interacting case. Stoch. Process. Appl. 106(2), 223–244 (2003)
Caputo, P., Liggett, T., Richthammer, T.: Proof of Aldous’ spectral gap conjecture. J. Am. Math. Soc. 23, 831–851 (2010)
Chang, C.C., Landim, C., Olla, S.: Equilibrium fluctuations of asymmetric simple exclusion processes in dimension \(d\ge 3\). Probab. Theory Relat. Fields 119(3), 381–409 (2001)
Cocozza-Thivent, C.: Processus des misanthropes. Probab. Theory Relat. Fields 70(4), 509–523 (1985)
Corwin, I.: The Kardar–Parisi–Zhang equation and universality class. Random Matrices Theory Appl. 1, 1–76 (2012)
Corwin, I., Quastel, J., Remenik, D.: Renormalization fixed point of the KPZ universality class. J. Stat. Phys. 160, 815–834 (2015)
Gonçalves, P.: Central limit theorem for a tagged particle in asymmetric simple exclusion. Stoch. Proc. Appl. 118, 474–502 (2008)
Gonçalves, P., Jara, M.: Nonlinear fluctuations of weakly asymmetric interacting particle systems. Arch. Ration. Mech. Anal. 212(2), 597–644 (2014)
Gonçalves, P., Jara, M.: Crossover to the KPZ equation. Ann. Henri Poincaré 13(4), 813–826 (2012)
Gonçalves, P., Jara, M.: Scaling limits of additive functionals of interacting particle systems. Commun. Pure Appl. Math. 66(5), 649–677 (2013)
Gonçalves, P., Jara, M., Sethuraman, S.: A stochastic Burgers equation from a class of microscopic interactions. Ann. Probab. 43(1), 286–338 (2015)
Gubinelli, M., Jara, M.: Regularization by noise and stochastic Burgers equations. Stoch. Partial Differ. Equ. Anal. Comput. 1(2), 325–350 (2013)
Gubinelli, M., Perkowski, N.: Energy solutions of KPZ are unique. (2015). arXiv:1508.07764
Hairer, M.: Solving the KPZ equation. Ann. Maths 178(2), 559–664 (2013)
Hairer, M.: A theory of regularity structures. Invent. Math. 198(2), 269–504 (2014)
Hairer, M., Quastel, J.: A class of growth models rescaling to KPZ. (2015). arXiv:1512.07845
Jara, M.: Non-equilibrium scaling limit for a tagged particle in the simple exclusion process with long jumps. Commun. Pure Appl. Math. 62(2), 198–214 (2009)
Jara, M.: Current and density fluctuations for interacting particle systems with anomalous diffusive behavior. (2009). arXiv:0901.0229
Jara, M.: Spectral gap for random walks with long jumps and applications (in preparation)
Jara, M., Komorowski, T., Olla, S.: Superdiffusion of energy in a chain of harmonic oscillators with noise. Commun. Math. Phys. 339(2), 407–453 (2015)
Kipnis, C., Landim, C.: Scaling Limits of Interacting Particle Systems. Springer, New York (1999)
Kipnis, C., Varadhan, S.R.S.: Central limit theorem for additive functionals of reversible Markov processes and applications to simple exclusions. Commun. Math. Phys. 104(1), 1–19 (1986)
Komorowski, T., Landim, C., Olla, S.: Fluctuations in Markov Processes, Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences], vol. 345. Springer, Heidelberg (2012)
Landim, C., Panizo, G., Yau, H.T.: Spectral gap and logarithmic Sobolev inequality for unbounded conservative spin systems. Ann. I.H.Poincaré 38(5), 739–777 (2002)
Liggett, T.M.: Interacting Particle Systems. Springer, Berlin (2005)
Mitoma, I.: Tightness of probabilities on \(C([0,1];{\cal{S}}^{\prime } )\) and \(D([0,1];{{\cal{S}}}^{\prime } )\). Ann. Probab. 11(4), 989–999 (1983)
Walsh, J.B.: An introduction to stochastic partial differential equations. In: cole d’ete de probabilites de Saint-Flour, XIV 1984. Lecture Notes in Math, vol. 1180, pp. 265–439. Springer, Berlin (1986)
Whitt, W.: Proofs of the martingale FCLT. Probab. Surv. 4, 268–302 (2007)
Acknowledgements
P.G. thanks FCT/Portugal for support through the project UID/MAT/04459/2013. M.J. was supported by FAPERJ through the Grant E-26/103.051/2012 “Jovem Cientista do Nosso Estado”. M.J. was partially supported by NWO Gravitation Grant 024.002.003-NETWORKS. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovative programme (grant agreement No 715734).
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A: The spectral gap inequality
A classical problem in the theory of Markov chains is the study of the time that the chain needs to reach the equilibrium. In the case of a (continuous time) finite state ergodic Markov chain it is known that the convergence to equilibrium is exponentially fast. Therefore the relevant question is the exponential rate at which this happens. Let \(\{x(t); t \ge 0\}\) be an ergodic Markov chain on a finite state space E. Let \(\mu \) be its unique invariant measure. For \(f: E \rightarrow {\mathbb R}\) and \(x \in E\), let \(P_t f(x) = {\mathbb E} [ f(x(t))| x(0)=x]\). Let \(\langle \cdot \rangle _\mu \) denote the expectation with respect to \(\mu \). Then we define
The number \(1/\lambda \) is known as the relaxation time of the chain \(\{x(t); t \ge 0\}\). In the case on which the chain \(\{x(t); t \ge 0\}\) is reversible with respect to \(\mu \), the number \(\lambda \) is equal to the spectral gap of the generator A of the chain \(\{x(t); t \ge 0\}\), that is, the absolute value of the largest non-zero eigenvalue of A. In that case, we have the variational formula
When the chain is not reversible, this variational formula provides an upper bound for \(\lambda ^{-1}\). A natural question in the theory of Markov chains is to estimate the spectral gap of a Markov chain, or of a family of Markov chains of increasing complexity.
For the symmetric simple random walk restricted to \(\{1,\dots ,n\}\) it is well known that \(\lambda ^{-1} = {\mathscr {O}}(n^2)\). It turns out that this property of the random walk over finite intervals, by means of a computation of Nash type, allows one to show that in the case of the symmetric simple random walk on \({\mathbb Z}\),
where \(\mu \) is the counting measure on \({\mathbb Z}\). Therefore a sharp upper bound on the spectral gap of finite-state Markov chains gives valuable information even in the case of chains on infinite state spaces.
Consider the symmetric random walk restricted to the set \(\Lambda _\ell = \{1,\dots ,\ell \}\). In the case where the random walks have long jumps we have the following result:
Proposition A.1
Let \(p(\cdot )\) be given by (2.2). There exists \(\kappa >0\) such that
for any \(\ell \in {\mathbb N}\) and any \(f: \Lambda _\ell \rightarrow {\mathbb R}\) such that
Remark A.2
This proposition is telling us that the spectral gap of a Markov chain with jump rates given in (2.2) and defined on the interval \(\Lambda _\ell \), is bounded from below by \(\frac{1}{\kappa \ell ^\alpha }\). In addition, pairing together the two terms involving x and y, we see that only the behavior of the symmetric part of \(p(\cdot )\) is relevant for this proposition.
The proof of this proposition is in fact very simple. For that purpose note that
To conclude, use the fact that for f satisfying (A.3), it holds:
together with \(s(y-x) \ge \frac{c^++c^-}{2 \ell ^{1+\alpha }}\) for any \(x,y \in \Lambda _\ell \).
As a corollary of Proposition A.1 we can obtain a lower bound for the spectral gap of the exclusion process with transition rate \(p(\cdot )\):
Corollary A.3
Let \(p(\cdot )\) be defined by (2.2). Let \(f: \Omega \rightarrow {\mathbb R}\) be a local function with \({{\mathrm{supp}}}(f) \subseteq \Lambda _\ell \). Assume that \(\int f d\mu _\sigma = 0\) for any \(\sigma \in [0,1]\). Then,
for any \(\sigma \in [0,1]\).
The simplest way to prove this corollary is by means of the Aldous’ conjecture, proved in [8], which says that the spectral gap of an exclusion process with symmetric rates is equal to the spectral gap of the random walk with the same rates. Another proof using a comparison principle can be found in [25].
Appendix B: Proof of Proposition 2.2
We do the proof for the case \(\alpha >1\), the others being analogous. The proof of the proposition is elementary, but very tedious. Recall the definition of \({\mathscr {L}}_n\) and \({\mathscr {L}}\) from (2.6) and (2.4), respectively. First note that by the definition of \(m_n^\alpha \) in (2.3), for this regime of \(\alpha \) we rewrite \({\mathscr {L}}_n f(\tfrac{x}{n}) = n^\alpha \sum _{y} p(y)\psi _{x/n}(\tfrac{y}{n})\) and \({\mathscr {L}} f(x) = \int _{\mathbb R} p(y) \psi _x(y)dy \) where \( \psi _{u}(v) = f(u+v)-f(u)-vf'(u),\) for \(f\in {\mathscr {C}}^2({\mathbb R})\). Second, note that \({\mathscr {L}}_n^+ f(\tfrac{x}{n}) = n^\alpha \sum _{y \ge 1} p(y)\psi _{x/n}(\tfrac{y}{n})\) and \( {\mathscr {L}}^+ f(x) = \int _0^\infty \frac{c^+}{y^{1+\alpha }} \psi _x(y)dy \) are well-defined and that it is enough to show (2.7) for \({\mathscr {L}}_n^+\) and \({\mathscr {L}}^+\). For \(x \in {\mathbb N}\), define \(P(x) = \sum _{y \ge x} p(y)\) and \(a(x) = x^\alpha P(x)-\frac{c^+}{\alpha }\). Note that a(x) tends to 0, as \(x\rightarrow \infty \). The idea is to perform an integration by parts in the formula of \({\mathscr {L}}_n^+f\) in order to work with the more regular object \(P(\cdot )\). By writing \(p(y)=P(y)-P(y+1)\), performing a summation by parts and a Taylor expansion on \(\psi \), we see that \({\mathscr {L}}_n^+f(\tfrac{x}{n}) = n^{\alpha -1} \sum _{y\ge 1} P(y)\psi _{x/n}'(\tfrac{y}{n})+R_1^n(x)\), where \(R_1^n(x)\) is an error term which satisfies \(|R_1^n(x)| \le \Vert \psi ''_{x/n}\Vert _\infty n^{\alpha -2} \sum _{y \ge 1} P(y).\) Note that \(\Vert \psi _{x/n}'\Vert _\infty {\le 2\Vert f'\Vert _\infty }\), which does not depend on x. This last sum is equal to \(\sum _{y \ge 1} y p(y)<+\infty \) and since \(\alpha <2\), \(R^1_n(x)\) vanishes, as \(n\rightarrow \infty \). For \(y\ge 1\), let \(A(y) = \sup _{z \ge y} |a(z)|\). We have that
Note that \(\psi _u'\) is bounded and that \(\psi _u(0)=\psi '_u(0)=0\). Therefore, there exists a constant K such that \(|\psi _u'(v)| \le Kv\) for any \(v\in [0,1]\) and such that \(|\psi _u'(v)| \le K\) for any \(v >1\). In fact, we can choose \(K=\max \{2\Vert f'\Vert _\infty , \Vert f''\Vert _\infty \}\). Therefore, the second term on the right hand side of (B.1) is bounded by
for any \(k<n\). Choosing, for example, \(k = \sqrt{n}\) , the last sums vanish, as \(n \rightarrow \infty \). Note that
since \(\psi _x(y)\) is quadratic around \(y =0\) and linear for \(y \gg 1\). Moreover, the first sum on the right hand side of (B.1) is just a Riemann sum for this last integral. Since the function \(\frac{\psi _x'(y)}{y^\alpha }\) is continuous at \(y=0\) and it decays like \(\frac{1}{y^\alpha }\), this Riemann sum converges to the corresponding integral. Note that this convegence is uniform in x, since \(\frac{\psi _x'(y)}{y^\alpha }\) is equicontinuous in x. Finally, since for \(0<y<z\), we have that \(\Big |\frac{\psi '(z)}{z^\alpha }-\frac{\psi '(y)}{y^\alpha }\Big | \le \frac{C(z-y)}{y^\alpha } \) for some constant C which depends only on \(\Vert f''\Vert _\infty \), we conclude that
Since the last sum is finite and \(\alpha <2\), we have just shown that
Moreover, since all the constants above do not depend on x, the limit is uniform in x, showing the first half of the proposition. The second half can be proved in a similar way.
Appendix C: Proof of Proposition 2.7
For the reader’s convenience, we repeat here various definitions introduced in Sect. 2. Let \({\mathscr {L}}\) be a generator of a Lévy process in \({\mathbb R}\). Let \(\{{\mathscr {W}}_t; t \ge 0\}\) be a Brownian motion on \(L^2({\mathbb R})\) and let \({\mathscr {S}} = \frac{1}{2}({\mathscr {L}} + {\mathscr {L}}^*)\) be the symmetric part of the operator \({\mathscr {L}}\). We say that a stochastic process \(\{{\mathscr {Y}}_t; t \ge 0\}\) is a stationary solution of the infinite-dimensional Ornstein–Uhlenbeck equation
if for each \(t \in [0,T]\) the random variable \({\mathscr {Y}}_t\) is a white noise of variance \(\chi \) and for any differentiable function \(f: [0,T] \rightarrow {\mathbb S}({\mathbb R})\) the process
is a martingale of quadratic variation \(2 \chi \int _0^t \langle f_s , -{\mathscr {S}} f_s \rangle ds.\) We will prove following result:
Proposition C.1
Two stationary solutions of (C.1) have the same law.
Proof
Let f be a function in \({\mathbb S}({\mathbb R})\) and take \(t \ge 0\). Let \(\{P_t; t \ge 0\}\) the semigroup associated to the generator \({\mathscr {L}}\), that is, \(P_t = e^{t {\mathscr {L}}}\) for any \(t \ge 0\). Since \({\mathscr {L}}\) is the generator of a Lévy process, \(\{P_t; t \ge 0\}\) is a strongly continuous, contraction semigroup on \({\mathscr {C}}_b({\mathbb R})\). In particular \(f_s = P_{t-s} t\) is a differentiable trajectory on \({\mathscr {C}}_b({\mathbb R})\) satisfying \(\frac{d}{ds} f_s = - {\mathscr {L}} f_s\) for any \(s \le t\) and \(f_t = f\). Since \(\{P_t; t \ge 0\}\) is also a contraction in \(L^1({\mathbb R})\), it is a contraction in \(L^2({\mathbb R})\). Note that \(\{f_s; s \le t\}\) is not a legitimate test function, since although \(P_{t-s} f\) is infinitely differentiable, it does not satisfy the decay properties needed to be in \({\mathbb S}({\mathbb R})\). However, \(\{f_s; s \le t\}\) can be approximated in \(L^2\) by differentiable functions \(f_\epsilon : [0,t] \rightarrow {\mathbb S}({\mathbb R})\), justifying the use of \(\{f_s; s \le t\}\) as a test function. Since \((\partial _s + {\mathscr {L}}) f_s =0\), we conclude that
is a martingale of quadratic variation (with respect to s)
Since the quadratic variation of \({\mathscr {M}}_{s,t}(f)\) is deterministic, \(\{{\mathscr {M}}_{s,t}(f); s\in [0, t]\}\) and in consequence \(\{{\mathscr {Y}}_t; t\in [0, T]\}\) are Gaussian processes. Note that
Therefore
We conclude that \({\mathscr {Y}}_t(f)\) can be written as the sum of two independent Gaussian variables: \({\mathscr {Y}}_0(P_t f)\), which depends only on the initial law and \({\mathscr {M}}_{t,t}(f)\), which is independent of \({\mathscr {Y}}_0\). Since \(\{{\mathscr {Y}}_t; t\in [0, T]\}\) is a Gaussian process, it is characterized by its covariance structure. By stationarity, the computations above show that for any \(0 \le s \le t \le T\) and any \(f,g \in {\mathbb S}({\mathbb R})\),
which shows uniqueness in law of the process \(\{{\mathscr {Y}}_t; t\in [0, T]\}\). \(\square \)
Appendix D: Auxiliary computations
In this section we collect the proofs of some estimates that are needed in Sect. 4.
1.1 D.1 Bounds on Taylor expansions of test functions
In the computations below, it will be useful to control the decay at infinity of the error terms of the Taylor expansion for test functions. For \(g:{\mathbb R} \rightarrow {\mathbb R}\), \(x \in {\mathbb R}\) and \(M >0\), define
We have the following:
Lemma D.1
Let \(g \in {\mathbb S}({\mathbb R})\). Then, for any \(\ell \in {\mathbb N}\),
The proof of this lemma is elementary, so we omit it. We can apply this lemma to obtain the aforementioned bounds on Taylor expansions of test functions: let \(f \in {\mathbb S}({\mathbb R})\), \(x \in {\mathbb R}\) and \(\ell \in {\mathbb N}\). For any \(M >0\) and any \(y \in {\mathbb R}\) such that \(|y-x| \le M\), we have that
This is an immediate consequence of the Taylor formula with Lagrange error and Lemma D.1.
1.2 D.2 Estimation of (4.1)
In order to estimate the first term of (4.1) we first make a change of variables \(z=y-x\), we fix M and write it as
The second term in (D.2) can be bounded from above by
By making a change of variables, using the fact that \(f\in { {\mathbb S}}({\mathbb R})\) and since
we conclude that the two previous sums are of order \(O(n^{-2})\). To compute the first term in (D.2), we use Lemma D.1 to conclude that
Now we use (D.1) and since
we obtain that the second term in (D.2) is of order \(O(n^{2\alpha -5})\) and therefore vanishes as \(n\rightarrow \infty .\) The second term of (4.1), by a change of variables \(z=y-x\) and fixing M, can be written as
By repeating exactly the same computations as above, the first term of last expression is of order \(O(n^{\alpha -3})\) while the second one is of order \(O(n^{-1})\). This shows that the second term of (4.1) vanishes, as \(n\rightarrow \infty \).
1.3 D.3 Estimation of (4.3)
The term (4.3), by a change of variables \(z=y-x\) and fixing M, can be written as
Repeating the same procedure as above we can see that the second term above is of order \(O(n^{-2})\), while the first one can be bounded from above by
Last sum is convergent if \(\alpha <1/2\) and if \(\alpha \ge 1/2\), it is divergent. When \(\alpha =1/2\), the order of divergence is \(\log (n)\). Therefore, we can conclude that (4.3) is of order o(1) for \(\alpha <1\) and bounded for \(\alpha =1\).
1.4 D.4 Proof of Lemma 4.5
The expectation in (4.4) is bounded from above by
and, by the change of variables \(z=y-x\) and by fixing M, we can write it as
Now, doing the same arguments as above, we see that the first term in last expression if of order \(O(n^{-1})\), while the second one can be bounded from above by
which is of order \(\frac{n^{2\alpha -2}}{K_n^{2\alpha -1}}\) and vanishes if \(K_n \gg n^{\frac{2\alpha -2}{2\alpha -1}}\).
1.5 D.5 Proof of Lemma 4.6
Note that the expectation in (4.5) is bounded from above by
As above, by making the change of variables \(z=y-x\) and using the fact that \(f\in {\mathbb S}({\mathbb R})\) we can bound the previous expression by
which is of order \(\frac{K^{3-2\alpha }}{n^{4-2\alpha }}.\) Since \(K \gg n^{\frac{2\alpha -2}{2\alpha -1}}\), the expectation vanishes as \(n\rightarrow \infty \).
1.6 D.6 Proofs of Lemmas 4.10 and 4.11
We start with the proof of Lemma 4.10. By the Cauchy–Schwarz inequality and by (4.8), the expectation in (4.9) is bounded from above by
which vanishes, as \(n\rightarrow \infty \), if \(K \gg n^{2\alpha -2}\). Above we used the fact that \(\sum _{y=1}^{K-1} y a(y)<\infty \), since we are in the regime \(\alpha >1\). Now we prove Lemma 4.11. For each \(j=1,\ldots , K\), we use Proposition 4.9 with \(F(\eta )=\sum _{x \in {\mathscr {Z}}_j}F_x(\eta )\) and
Therefore, the expectation in (4.9) is bounded from above by
We note that above we used the fact that \(\sum _{y=1}^{K-1} y^2 a(y)^2<\infty \). We conclude that last term vanishes if \(K \ll n^{\frac{2-\alpha }{1+\alpha }}\).
1.7 D.7 Proof of Lemma 4.14
By the Minkowski’s inequality and the estimate in (4.12), the expectation in (4.13) is bounded from above by
1.8 D.8 Proof of Lemma 4.15
At at a first glance we note that by Proposition 4.9 and by (4.8), the expectation in (4.15) is bounded by
But this bound will not be sufficient for us. Therefore, in order to prove the lemma we use the multiscale structure introduced in [16]. For that purpose, let j be fixed and take the sequence of boxes \(\ell _0=K^j\), \(\ell _1=2\ell _0\) and for \(p\ge 2\), \(\ell _p=2^p\ell _0\). Suppose that there exists P sufficiently big such that \(2^P \ell _0=K^{j+1}\). Performing a telescopic sum and using the Minkowski’s inequality together with (D.6), the expectation on the left hand side of (4.15) is bounded from above by
This proves (4.15) for the case \(K^{j+1}=2^P \ell _0\). For the other cases, we take P sufficiently big such that \(2^P\ell _0\le K^{j+1}\le 2^{P+1}\ell _0\). Let \({\tilde{\Psi }}_x^i=\Psi _x^{2^i\ell _0}\). Then, by using the inequality \((x+y)^2\le 2x^2+y^2\), the expectation on the left hand side of (4.15) is bounded from above by
From (D.7), the first expectation in (D.8) is bounded from above by
while from (D.6) the second expectation is bounded from above by
Putting together the two previous estimates, the proof of (4.15) ends.
1.9 D.9 Proof of (4.18)
First we compute the price to double the size the box. For that purpose, let M be given. By Proposition 4.9 and (4.8) we have that
Define \(M_0=M\) and \(M_i = 2^i M\) for \(i \in {\mathbb N}\). By writing a telescopic sum, using Minkowski’s inequality and the previous estimate, we see that
Taking \(M= K_n\) and \(M_\ell = \varepsilon n\) the proof ends.
Rights and permissions
About this article
Cite this article
Gonçalves, P., Jara, M. Density fluctuations for exclusion processes with long jumps. Probab. Theory Relat. Fields 170, 311–362 (2018). https://doi.org/10.1007/s00440-017-0758-0
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00440-017-0758-0
Keywords
- Density fluctuations
- Exclusion with long jumps
- Fractional Burgers equation
- Fractional Ornstein–Uhlenbeck process