Download presentation
Presentation is loading. Please wait.
1
Mixed order phase transitions
David Mukamel Amir Bar, DM (PRL, 122, (2014))
2
Phase transitions of mixed order
(a) diverging length πββ as in second order transitions π~ π‘ βπ or π~ π 1/ π‘ π (b) discontinuous order parameter as in first order transitions
3
1d Ising model with long range interactions π½(π)~1/ π 2
Examples: 1d Ising model with long range interactions π½(π)~1/ π 2 non-soluble but many of its properties are known 2. Poland-Scheraga (PS) model of DNA denaturation π~ π 1/ π‘ π π~ π‘ βπ 3. Jamming transition in kinetically constrained models π~ π 1/ π‘ π Toninelli, Biroli, Fisher (2006) 4. βExtraordinary transitionβ in network rewiring Liu, Schmittmann, Zia (2012)
4
IDSI : Inverse Distance Square Ising model
π»=β π>π π½ πβπ π π π π π π =Β±1 π½(πβπ)~ π½ πβπ πΌ πΌ=2 For 1<πΌβ€2 the model has an ordering transition at finite T A simple argument: L π πΈ π =β 2π½ πΌβ1 π πΏ 1βπΌ + 2π½ πΌβ1 2βπΌ ( π 2βπΌ β1) Anderson et al (1969, 1971); Dyson (1969, 1971); Thouless (1969); Aizenman et al (1988)β¦
5
πΌ=2 model is special The magnetization m is discontinuous at π π (βThouless effectβ) Thouless (1969), Aizenman et al (1988) π~exp 1 πβ π π KT type transition, Cardy (1981) Phase diagram H T IDSI Fisher, Berker (1982)
6
Dyson hierarchical version of the model (1971)
1/ 2 6 1/ 2 4 1/ 2 2 Mean field interaction within each block The Dyson model is exactly soluble demonstrating the Thouless effect
7
Exactly soluble modification of the IDSI model
microscopic configuration: π»=β π½ ππ π π π π π+1 β π,π π€ππ‘βππ πππβ ππππππ π½ πβπ π π π π π½(πβπ)~ 1 πβπ 2 π½ ππ , π½ πβπ >0 The interaction is in fact not binary but rather many body.
8
Summery of the results diverging correlation length at π π π~ πβ π π βπ with nonuniversal π Extreme Thouless effect with π=0βπ= Β±1 Phase diagram H T The model is closely related to the PS model of DNA denaturation
9
The energy of a domain of length π
πΈ π =β π>π πβπ 2 =β π=1 πβ1 πβπ π ~βππ+ π ln π+ππππ π‘. Interacting charges representation: π» π π , π =βπ π=1 π π π + π π=1 π ln π π +Ξπ a, π , Ξ >0 Charges of alternating sign (attractive) on a line Attractive long-range nearest-neighbor interaction Chemical potential --suitable representation for RG analysis --similar to the PS model
10
Analysis of the model Grand partition sum
π πΏ,π = π π π π π½ π π 1 π½ π π΄ π π½ π π 2 π½ π β¦..π΄ π π½ π π π π π½ π π 1 + β¦+ π π =L Grand partition sum π π§ = πΏ π§ πΏ π(πΏ, π) π π§ =π΄ π 2 π§ 1+ π΄ 2 π 2 + π΄ 4 π 4 +β¦ Polylog function π π½ π π§ = π=1 β π§ π π π½π π π½ π β‘ Ξ¦ π½ π (π§ π π½ ) π π§ = π΄ π 2 (π§) 1β π΄ 2 π 2 (π§)
11
π§ β is the closest pole to the origin
π π½ π π§ = π=1 β (π§ π π½ ) π π π½π Μ
β‘ Ξ¦ π½ π (π§ π π½ ) π π§ = π΄ π 2 (π§) 1β π΄ 2 π 2 (π§) Polylog function π§ β is the closest pole to the origin π(π§)~ π§ βπΏ π βπ½πΉ πΏ,π ~ π§ β π βπ½π π πΏ π§ β = π π½π(π) π π§ β ,π =1/π΄ π΄<1, ferromagnetic coupling
12
Μ Μ π π½ π π§ π π½ = π=1 β (π§ π π½ ) π π π½ π β‘ Ξ¦ π½ π (π§ π π½ )
1< 1/π΄ π(π§ π π½ ) π§ π π½ 1 π½ββ π½β0 Μ π½ π π π½ π π§ π π½ = π=1 β (π§ π π½ ) π π π½ π β‘ Ξ¦ π½ π (π§ π π½ ) Μ π§ π π½ 1 π§ β e π½ Phase transition: (π§ β π π½ π )=1 , π½ π π β‘π Unlike the PS model the parameter c is not universal
13
Nature of the transition
π(π)= π(πΏβπ) π(πΏ) π΄ 1 π π½ π ~ (π§ π π½ ) π π π½ π Domain length distribution π π ~ π βππΏπ§ π π = π βπ/π π π πβ‘ π½ π π Close to π π : π§=1βπΏπ§ characteristic length π= 1 πΏπ§ πββ πππ πβ π π Ξ¦ π 1βπΏπ§ β Ξ¦ π 1 βπΏ π§ πβ <πβ€2 Ξ¦ π 1 βπΏπ§ π>2 π‘β‘πβ π π ~ πΏ π§ πβ <πβ€2 πΏπ§ π>2 π~ π‘ βπ π= 1 πβ <πβ€ π>2
14
Two order parameters 1. π order parameter π= π πΏ number of domains
π= π πΏ magnetization 1. π order parameter π= 1 π π ~ 1 β π 1 π π ππ Where at π π for 1<cβ€ π ββ for π> π is finite π is continuous <πβ€2 in both cases πββ π is discontinuous π>2
15
2. π order parameter π πΏ,π πΏ= π π π= β1 π π π
π πΏ,π πΏ= π π π= β1 π π π π π§,π = πΏ,π π§ πΏ π πΏ + βπΏ β π(πΏ,π= πΏ + β πΏ β ) π π§,π = π΄π π§π π( π§ π ) 1β π΄ 2 π π§π π( π§ π ) π=β π ln π§ β (π) π ln π (π=1) ββ‘ ln π is the magnetic field π> π π π= (πβ 1 π symmetry) π< π π either π§π=1 or π§ π =1 π=Β±1 Extreme Thouless effect
16
Phase diagram I n is continuous II and III n is discontinuous
17
β,π phase diagram β1<π<1 π=1 π=β1 β π
18
Free energy πΉ πΏ,π,π =πΏπ(π,π)
Canonical analysis Free energy πΉ πΏ,π,π =πΏπ(π,π) π π§,π = πΏ,π π πΏ,π π§ πΏ π π π π§,π = π΄π π 2 (π§) 1β π΄ 2 π 2 π 2 (π§) π πΏ,π = 1 2ππ ππ§ππ π(π§,π) π§ πΏ+1 π π+1 π πΏ,π = π βπ½πΏπ(π,π) π πΏ,π = ππ§ 4ππ π΄π(π§) πΏπ+2 π΄ π§ πΏ+1 β ππ§ 4ππ π βπΏ π 1 (π,π§)
19
π π> π π =ππ π§ β βπππ π΄π( π§ β )
π πΏ,π = ππ§ 4ππ π΄π(π§) πΏπ+2 π΄ π§ πΏ+1 β ππ§ 4ππ π βπΏ π 1 (π,π§) saddle point: π= π( π§ β ) π§ β πβ²( π§ β ) π π> π π =ππ π§ β βπππ π΄π( π§ β ) π π< π π =ππ π§ π βπππ π΄π( π§ π )
20
c=2.5
21
Finite L correction: π=2.5 πΏ=1000
22
c=2.5
23
Finite L corrections c=2.5 L=1000
24
Renormalization group - charges representation
π π¦,π = π=0 β π¦ π π=1 π π π π π π π+1 β π π π βπ Ζ( π π+1 β π π βπ) y - fugacity a - short distance cutoff Length rescaling πβπ π π
This can be compensated by y rescaling π¦β π¦ π π
(1βπ)
25
Ζ π π+1 β π π βπ π π
βΖ π π+1 β π π βπ βaπ
πΏ( π π+1 β π π βπ)
π πβπ π β π¦ 2 ππ
π πβ2π ππ 1 π π πβπ βπ π The integral scales like 1/ π π hence it does not renormalize c . Rather it renormalizes y.
26
Renormalization group equations
ππ¦ ππ
=π₯π¦+ π¦ 2 ππ₯ ππ
=0 π₯β‘1βπ compared with the Kosterlitz-Thouless model: ππ¦ ππ
=π₯π¦ ππ₯ ππ
= π¦ 2
27
Contribution of the dipole to the renormalized partition sum:
In the KT case: π πβπ π Contribution of the dipole to the renormalized partition sum: π¦ π π 1+ π¦ 2 ππ
π πβπ ππ π π +π π πβπ βπ πβπ π 1β π π β π πβπ π¦ π π 1+ π¦ 2 π
πππ renormalizes c. (Cardy 1981)
28
π~ π 1/ πβ π π ππ¦ ππ
=π₯π¦+ π¦ 2 ππ¦ ππ
=π₯π¦ ππ₯ ππ
=0 ππ₯ ππ
= π¦ 2 π₯β‘1βπ
Line of fixed points π¦ β =β π₯ β π~ πβ π π βπ π= 1 πβ1 π~ π 1/ πβ π π
29
Coarsening dynamics Particles with n-n logarithmic interactions
Biased diffusion, annihilation and pair creation π π 1 , π 2 + π π 3 , π 4 β π π 5 , π 6 + π 1 π 2 π 3 π 4 π 5 π 6 π π,π Β± = ππ πΒ±1 πβ1 βπ β1 π π,π Β± = 1+ π¦ β2 ππ π+π+1 Β±π β1
30
ππ(π,π‘) ππ‘ = π 2 π(π,π‘) π π 2 + π ππ ( π π π π,π‘ )
Coarsening dynamics π=ββπ< π π The coarsening is controlled by the T=0 (y=0) fixed point ππ(π,π‘) ππ‘ = π 2 π(π,π‘) π π 2 + π ππ ( π π π π,π‘ ) π π,π‘ = 1 π‘ π( π π‘ ) Like the dynamics of the T=0 Ising model
32
π= 1 <π> - number of domains
Coarsening dynamics π=ββπ= π π π π,π‘ = 1 π π π( π π‘ 1/π§ ) Expected scaling form π π₯βͺ1 = π(π₯β«1)~ π βπ₯ π= 1 <π> - number of domains π~ π‘ βπ(π) with π π = 2βπ π§
33
L=5000 c=1.5 π= π π π π,π‘ π‘ 1.5 z=2 z=1.5 π/ π‘ 0.5 π/ π‘ 0.66 π π,π‘ ~ 1 π π π π π‘ 1/π§
34
- Voter model (y=0, fixed c) π~ π‘ βπ(π) with π π = 2βπ π§
35
Summary Some models exhibiting mixed order transitions are discussed.
A variant of the inverse distance square Ising model is studied and shown to have an extreme Thouless effect, even in the presence of a magnetic field Relation to the IDSI model is studies by comparing the renormalization group transformation of the two models. The model exhibits interesting coarsening dynamics at criticality.
36
Domain representation of the 1 π 2 Ising model
(Fortuin-Kasteleyn representation) H=β π½ π,π ( π π ,π π ) π π = π π½ π½ π ( π π , π π ) β1 π πΏ = π π,π (1+ π πβπ ) = π πΈ π,π π πβπ πΈ π,π πΈ π,π =0,1 defines a graph on the vertices 1,β¦,πΏ The sum is over all graphs E
37
A graph can be represented as composed of sub-graphs
separated by βbreaking pointsβ π π = π π½ π½ π ( π π , π π ) β1 One has to calculate π π - the probability that the distance between adjacent breaking points is π.
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.