Abstract
The random-cluster model has been widely studied as a unifying framework for random graphs, spin systems and electrical networks, but its dynamics have so far largely resisted analysis. In this paper we analyze the Glauber dynamics of the random-cluster model in the canonical case where the underlying graph is an \(n \times n\) box in the Cartesian lattice \({{\mathrm{\mathbb {Z}}}}^2\). Our main result is a \(O(n^2\log n)\) upper bound for the mixing time at all values of the model parameter p except the critical point \(p=p_c(q)\), and for all values of the second model parameter \(q\ge 1\). We also provide a matching lower bound proving that our result is tight. Our analysis takes as its starting point the recent breakthrough by Beffara and Duminil-Copin on the location of the random-cluster phase transition in \({{\mathrm{\mathbb {Z}}}}^2\). It is reminiscent of similar results for spin systems such as the Ising and Potts models, but requires the reworking of several standard tools in the context of the random-cluster model, which is not a spin system in the usual sense.
This is a preview of subscription content, access via your institution.



Notes
We say that an event occurs with high probability if it occurs with probability approaching 1 as \(n \rightarrow \infty \).
For a pair of functions \(f,g:{{\mathrm{\mathbb {N}}}}\rightarrow {{\mathrm{\mathbb {R}}}}^+\), we say that \(f(n) = O(g(n))\) (resp., \(f(n) = \varOmega (g(n))\)) if there exists constants \(c,n_0 > 0\) such that \(f(n) \le c \cdot g(n)\) (resp., \(f(n) \ge c \cdot g(n))\) for all \(n > n_0\). We say that \(f(n) = \varTheta (g(n))\) when \(f(n) = O(g(n))\) and \(f(n) = \varOmega (g(n))\).
References
Alexander, K.S.: Mixing properties and exponential decay for lattice systems in finite volumes. Ann. Probab. 32(1A), 441–487 (2004)
Alexander, K.S.: On weak mixing in lattice models. Probab. Theory Relat. Fields. 110, 441–471 (1998)
Beffara, V., Duminil-Copin, H.: The self-dual point of the two-dimensional random-cluster model is critical for \(q \ge 1\). Probab. Theory Relat. Fields. 153, 511–542 (2012)
van den Berg, J.: A uniqueness condition for Gibbs measures, with applications to the 2-dimensional Ising antiferromagnet. Commun. Math. Phys. 152(1), 161–166 (1993)
Blanca, A., Sinclair, A.: Dynamics for the mean-field random-cluster model. In: Proceedings of the 19th International Workshop on Randomization and Computation (RANDOM), pp. 528–543 (2015)
Borgs, C., Chayes, J., Tetali, P.: Swendsen-Wang algorithm at the Potts transition point. Probab. Theory Relat. Fields. 152, 509–557 (2012)
Borgs, C., Frieze, A., Kim, J.H., Tetali, P., Vigoda, E., Vu, V.: Torpid mixing of some Monte Carlo Markov chain algorithms in statistical physics. In: Proceedings of the 40th Annual Symposium on Foundations of Computer Science (FOCS), pp. 509–557 (1999)
Duminil-Copin, H., Sidoravicius, V., Tassion, V.: Continuity of the phase transition for planar random-cluster and Potts models with \(1 \le q \le 4\). arXiv:1505.04159 (2015)
Dyer, M., Sinclair, A., Vigoda, E., Weitz, D.: Mixing in time and space for lattice spin systems: a combinatorial view. Random Struct. Algorithm 24, 461–479 (2004)
Edwards, R.G., Sokal, A.D.: Generalization of the Fortuin–Kasteleyn–Swendsen–Wang representation and Monte Carlo algorithm. Phys. Rev. D 38(6), 2009–2012 (1988)
Fortuin, C.M.: On the random-cluster model: III. The simple random-cluster model. Physica 59(4), 545–570 (1972)
Fortuin, C.M., Kasteleyn, P.W.: On the random-cluster model I. Introduction and relation to other models. Physica 57(4), 536–564 (1972)
Galanis, A., Štefankovič, D., Vigoda, E.: Swendsen–Wang algorithm on the mean-field potts model. In: Proceedings of the 19th International Workshop on Randomization and Computation (RANDOM), pp. 815–828 (2015)
Ge, Q., Štefankovič, D.: A graph polynomial for independent sets of bipartite graphs. Comb. Probab. Comput. 21(5), 695–714 (2012)
Grimmett, G.R.: The Random-Cluster Model. Springer-Verlag, Berlin (2006)
Hayes, T.P., Sinclair, A.: A general lower bound for mixing of single-site dynamics on graphs. Ann. Appl. Probab. 17(3), 931–952 (2007)
Laanait, L., Messager, A., Miracle-Solé, S., Ruiz, J., Shlosman, S.: Interfaces in the Potts model I: Pirogov–Sinai theory of the Fortuin–Kasteleyn representation. Commun. Math. Phys. 140(1), 81–91 (1991)
Levin, D.A., Peres, Y., Wilmer, E.L.: Markov Chains and Mixing Times. American Mathematical Society, Providence, Rhode Island (2008)
Long, Y., Nachmias, A., Ning, W., Peres, Y.: A power law of order 1/4 for critical mean-field Swendsen–Wang dynamics. MEMOIRS of the American Mathematical Society, 232(1092) American Mathematical Society, Providence, Rhode Island (2014)
Lubetzky, E., Sly, A.: Critical Ising on the square lattice mixes in polynomial time. Commun. Math. Phys. 313(3), 815–836 (2012)
Martinelli, F.: Lectures on glauber dynamics for discrete spin models. In: Bernard P (ed) Lectures on Probability Theory and Statistics. Lecture Notes in Mathematics, vol. 1717, pp. 93–191. Springer, Berlin, Heidelberg (1999)
Martinelli, F., Olivieri, E., Schonmann, R.H.: For 2-D lattice spin systems weak mixing implies strong mixing. Commun. Math. Phys. 165(1), 33–47 (1994)
Mossel, E., Sly, A.: Exact thresholds for Ising–Gibbs samplers on general graphs. Ann. Probab. 41(1), 294–328 (2013)
Propp, J., Wilson, D.: Exact sampling with coupled Markov chains and applications to statistical mechanics. Random Struct. Algorithm 9, 223–252 (1996)
Strassen, V.: The existence of probability measures with given marginals. Ann. Math. Stat. 36, 423–439 (1965)
Swendsen, R.H., Wang, J.S.: Nonuniversal critical dynamics in Monte Carlo simulations. Phys. Rev. Lett. 58(2), 86–88 (1987)
Ullrich, M.: Swendsen–Wang is faster than single-bond dynamics. SIAM J. Discrete Math. 28(1), 37–48 (2014)
Ullrich, M.: Rapid mixing of Swendsen–Wang dynamics in two dimensions. Diss. Math. 502 (2014). doi:10.4064/dm502-0-1
Ullrich, M.: Comparison of Swendsen–Wang and heat-bath dynamics. Random Struct. Algorithm 42(4), 520–535 (2013)
Acknowledgments
The authors would like to thank Fabio Martinelli and Allan Sly for helpful suggestions. We would like also to thank the anonymous referees whose valuable comments improved the exposition of the results.
Author information
Authors and Affiliations
Corresponding author
Additional information
Antonio Blanca was supported in part by an NSF Graduate Research Fellowship and NSF Grant CCF-1420934. Alistair Sinclair was supported in part by NSF Grant CCF-1420934.
Appendix: Proof of Lemma 6.2
Appendix: Proof of Lemma 6.2
We show first that the measure that results from conditioning on the state of a single edge maintains the exponential decay of finite volume connectivities (3).
Fact 7.2
Let \(p < p_c(q)\), \(q \ge 1\), and let \(\eta \) be a boundary condition for \({{\mathrm{{\Lambda }}}}= ({{\mathrm{{\Lambda }}}}_n,E_n)\). Consider a copy \(\{Y_t\}\) of the continuous time Glauber dynamics on \({{\mathrm{{\Lambda }}}}\), and assume \(Y_0\) is sampled from the distribution \(\mu ^\eta _{{{\mathrm{{\Lambda }}}},p,q}(\,\cdot \,|\,e=b)\), for some \(e \in E_n\) and \(b \in \{0,1\}\). Then, for all \(u,v \in {{\mathrm{{\Lambda }}}}_n\), there exists positive constant C and \({{\mathrm{{\lambda }}}}\) such that

where
denotes the event that u and v are connected by a path of open edges in \(Y_t\).
Proof
Let \(\{Z_t\}\) be a second instance of the continuous time Glauber dynamics. The evolution of \(\{Z_t\}\) is coupled with that of \(\{Y_t\}\) via the identity coupling, except that \(\{Z_t\}\) never updates the edge e. The initial configuration of \(\{Z_t\}\) is sampled according to the distribution \(\mu ^{\eta }(\,\cdot \,|\,e=1)\) such that \(Y_0 \subseteq Z_0\). This is always possible because \(\mu ^{\eta }(\,\cdot \,|\,e=1) \succeq \mu ^{\eta }\). Then, \(Y_t \subseteq Z_t\) and \(Z_t\) has law \(\mu ^{\eta }(\,\cdot \,|\,e=1)\) for all \(t \ge 0\). We establish that the measure \(\mu ^{\eta }(\,\cdot \,|\,e=1)\) has exponential decay of finite volume connectivities and thus so does the distribution of \(Y_t\) for all \(t \ge 0\). By (3), for all \(u,v \in {{\mathrm{{\Lambda }}}}_n\), we have
where \(C,{{\mathrm{{\lambda }}}}\) are positive constants. If \(p' = \frac{p}{q(1-p)+p}\), then \(\mu ^{\eta } \succeq \mu ^{\eta }_{{{\mathrm{{\Lambda }}}},p',1}\) (see, e.g., [11]), and thus \(\mu ^\eta (e=1) \ge p'\). Since \(q \ge 1\),
The result then follows immediately when \(p = \varOmega (1)\). Otherwise, the measure \(\mu ^{\eta }\) is stochastically dominated by any random-cluster measure \(\mu ^{\eta }_{{{\mathrm{{\Lambda }}}},p'',q}\) with \(p''= \varOmega (1)\), for which we just established exponential decay of finite volume connectivities; the result follows by monotonicity. \(\square \)
We are now ready to prove the lemma.
Proof
Let \(Q_t\) be the random time at which the t-th edge is updated by the identity coupling. For some fixed \(\ell \) to be chosen later, and each \(t \ge 0\), consider the event
where \(u\overset{\scriptscriptstyle {Y_{Q_t}}}{\nleftrightarrow }v\) denotes the event that u and v are not connected by a path in \(Y_{Q_t}(B)\). Also, let \(\mathcal{E}_\ell := \bigcap _{t:Q_t \le T} \mathcal{E}_{\ell ,t}\). Then,
[cf. Eq. (5)]. We bound each term on the right hand side of (20) separately.
Conditioned on the event \(\mathcal{E}_\ell \), a witness for the fact that \(X_{T}(e) \ne Y_{T}(e)\) can be constructed as in discrete time. However, the probability that a given witness of length L is updated by the continuous time dynamics is instead bounded using the following fact from [16]. \(\square \)
Fact 7.3
Consider L independent rate 1 Poisson clocks. Then, the probability that there is an increasing sequence of times \(0 \le t_1< \cdots < t_L \le T\) such that clock i rings at time \(t_i\) is at most \(\left( \frac{eT}{L}\right) ^L\).
Recall from Sect. 3 that the number of witnesses of length L is at most \((4(\ell +1)^2)^L\) (crudely). Hence, following the same steps as in the proof Lemma 3.1, and taking \(\ell = r^{1/4}-1\), we get
using the fact that \(T \le r^{1/4}/(4{{{\mathrm{\mathrm{e}}}}}^2)\) [cf. Eq. (7)].
To bound the second term on the right hand side of (20), let N be the number of edge updates in B up to time T. Observe that N is a Poisson random variable with rate \(M := |E(B,r)| = \varTheta (r^2)\). Using standard bounds for Poisson tail probabilities we get that \(\Pr [N > {{{\mathrm{\mathrm{e}}}}}^2 M T] = \exp (-\varOmega (M T))\) for all \(T \ge 1\). Therefore,
Also, \(\lnot \mathcal{E}_\ell := \bigcup _{t:Q_t \le T} \,\lnot \mathcal{E}_{\ell ,t}\), and if the edge update at time \(Q_t\) occurs outside B, we have \(\lnot \mathcal{E}_{\ell ,t} = \lnot \mathcal{E}_{\ell ,t+1}\). Hence, a union bound implies
Fact 7.2 establishes exponential decay of finite volume connectivities (3) for the distribution of \(Y_t\) in B for all \(t \ge 0\). Then, as in Lemma 3.1, we obtain
Together with (21), this implies there exist constants \(c,C,{{\mathrm{{\lambda }}}}> 0\) such that for all \(r \ge c\) we have \(\Pr [\,X_{T}(e) \ne Y_{T}(e) \,] \le C \exp (-\lambda r^{1/4})\), as desired.
Rights and permissions
About this article
Cite this article
Blanca, A., Sinclair, A. Random-cluster dynamics in \({{\mathrm{\mathbb {Z}}}}^2\) . Probab. Theory Relat. Fields 168, 821–847 (2017). https://doi.org/10.1007/s00440-016-0725-1
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00440-016-0725-1
Keywords
- Random-cluster model
- Glauber dynamics
- Markov chains
- Spatial mixing
- Statistical physics
Mathematics Subject Classification
- 60J10
- 60K35
- 82B20