Physlib

PhyslibAlpha.ProbabilisticTheory.LocalNet.Lattice

Lattice systems

Finite sets of sites as regions, a group acting on sites, disjointness as separation, and neighbors.

i. Overview

For a lattice system, such as a crystal or a lattice of spins, the regions are finite sets of sites ordered by inclusion. A group acting on the sites, for instance the translations of the lattice, moves whole regions and preserves which lies inside which. Two regions are separated when they have no site in common. Nearby sites are described by a neighbor structure, from which enlargement of a region by a radius is built. For a periodic pattern one takes the sites to be `Λ × B` with translations of `Λ` acting on the first factor, `B` being the contents of a unit cell; nothing here requires this form, so non-rectangular lattices are included.

Restricting to a finite box and imposing periodic identification of its boundary are different constructions with different physics. Only the first is treated in this file.

ii. Key results

- `Lattice.regionAction` : translating a finite set of sites. - `Lattice.disjointSeparation` : regions with no common site are separated, and translation preserves this. - `Lattice.Neighborhoods` : which sites lie within a given radius of a site. - `Lattice.Neighborhoods.hasEnlargement` : the enlargement of a finite region by a radius. - `Lattice.Neighborhoods.enlargementEquivariant` : translating commutes with enlarging.

iii. Table of contents

  • A. Regions and the group action
  • B. Separation
  • C. Neighbor data and enlargement

iv. References

* N. Ashcroft & N. D. Mermin, Solid State Physics, Chapters 4 and 8. * E. Lieb & D. Robinson, The finite group velocity of quantum spin systems.

A. Regions and the group action

B. Separation

C. Neighbor data and enlargement

8 declarations

instance

Group action of GG on finite lattice regions of SS

Given a group GG acting on a set of sites SS, there exists an induced group action on the collection of finite regions of the lattice (represented as finite subsets of SS, denoted Finset S\text{Finset } S). For any group element g∈Gg \in G and any finite region Λ⊆S\Lambda \subseteq S, the translated region is defined as g⋅Λ={g⋅s∣s∈Λ}g \cdot \Lambda = \{g \cdot s \mid s \in \Lambda\}.

instance

Separation of regions via disjointness (X∩Y=∅X \cap Y = \emptyset)

For a set of sites SS, the separation relation on the collection of finite regions Finset S\text{Finset } S is defined such that two regions XX and YY are separated if and only if they are disjoint, denoted X∩Y=∅X \cap Y = \emptyset. This relation satisfies the properties of symmetry (X∩Y=∅  ⟹  Y∩X=∅X \cap Y = \emptyset \implies Y \cap X = \emptyset) and monotonicity (X′⊆X,Y′⊆Y,X∩Y=∅  ⟹  X′∩Y′=∅X' \subseteq X, Y' \subseteq Y, X \cap Y = \emptyset \implies X' \cap Y' = \emptyset).

instance

Group action preserves and reflects region disjointness (X∩Y=∅X \cap Y = \emptyset)

Let GG be a group acting on a set of sites SS. For any finite regions X,Y∈Finset SX, Y \in \text{Finset } S, let the separation relation be defined by disjointness, X∩Y=∅X \cap Y = \emptyset. Then the group action of GG on the regions preserves and reflects this separation: for any g∈Gg \in G, X∩Y=∅X \cap Y = \emptyset if and only if g⋅X∩g⋅Y=∅g \cdot X \cap g \cdot Y = \emptyset.

definition

Enlargement of region XX by radius rr as ⋃a∈XNr(a)\bigcup_{a \in X} N_r(a)

Given a finite region X⊂SX \subset S and a non-negative radius r∈R≥0r \in \mathbb{R}_{\ge 0}, the enlargement of XX by rr is the finite set of sites defined by the union ⋃a∈XNr(a)\bigcup_{a \in X} N_r(a), where Nr(a)N_r(a) denotes the neighborhood of site aa with radius rr. This set contains all sites c∈Sc \in S that are within radius rr of at least one site in XX.

theorem

c∈enlarge(r,X)  ⟺  ∃a∈X,c∈Nr(a)c \in \text{enlarge}(r, X) \iff \exists a \in X, c \in N_r(a)

For a non-negative radius r∈R≥0r \in \mathbb{R}_{\ge 0} and a finite region of sites X⊂SX \subset S, a site c∈Sc \in S is an element of the enlargement of XX by rr if and only if there exists a site a∈Xa \in X such that cc is in the neighborhood of aa with radius rr, denoted by Nr(a)N_r(a).

abbrev

Enlargement structure for finite regions Finset S\text{Finset } S

This definition equips the collection of finite regions Finset S\text{Finset } S (finite sets of sites) with an enlargement structure derived from the neighbor data nbnb. For a given radius r∈R≥0r \in \mathbb{R}_{\ge 0} and a finite region X⊂SX \subset S, the enlargement E(r,X)E(r, X) is defined as the union of neighborhoods of all sites in XX: E(r,X)=⋃a∈XNr(a) E(r, X) = \bigcup_{a \in X} N_r(a) where Nr(a)N_r(a) is the neighborhood of site aa with radius rr. This enlargement operation satisfies the following properties: - **Zero radius**: E(0,X)=XE(0, X) = X, meaning enlargement by zero radius returns the original region. - **Region monotonicity**: If X⊆YX \subseteq Y, then E(r,X)⊆E(r,Y)E(r, X) \subseteq E(r, Y). - **Radius monotonicity**: If r≤sr \le s, then E(r,X)⊆E(s,X)E(r, X) \subseteq E(s, X). - **Radius subadditivity**: The composition of two enlargements is contained within the enlargement of the sum of their radii, i.e., E(s,E(r,X))⊆E(r+s,X)E(s, E(r, X)) \subseteq E(r+s, X).

definition

Equivariance of neighborhood data nbnb under group action GG

Let SS be a set of sites and GG a group acting on SS. A neighbor structure nbnb is said to be equivariant if for every radius r∈R≥0r \in \mathbb{R}_{\ge 0}, group element g∈Gg \in G, and site s∈Ss \in S, the neighborhood of the transformed site g⋅sg \cdot s at radius rr is equal to the neighborhood of ss transformed by gg. That is, nb(r,g⋅s)=g⋅nb(r,s)nb(r, g \cdot s) = g \cdot nb(r, s), where the group action on the right-hand side is the lifted action of GG on the set of sites.

theorem

GG-Equivariance of Neighborhoods implies E(r,g⋅X)=g⋅E(r,X)E(r, g \cdot X) = g \cdot E(r, X)

Let SS be a set of sites and GG a group acting on SS. Let Nr(s)⊆SN_r(s) \subseteq S denote the neighborhood of a site s∈Ss \in S with radius r∈R≥0r \in \mathbb{R}_{\ge 0}. If the neighbor structure is equivariant, such that Nr(g⋅s)=g⋅Nr(s)N_r(g \cdot s) = g \cdot N_r(s) for all g∈Gg \in G, s∈Ss \in S, and r∈R≥0r \in \mathbb{R}_{\ge 0}, then the enlargement of any finite region X⊆SX \subseteq S (defined as E(r,X)=⋃a∈XNr(a)E(r, X) = \bigcup_{a \in X} N_r(a)) is also equivariant. That is, for any group element g∈Gg \in G, radius r∈R≥0r \in \mathbb{R}_{\ge 0}, and finite region XX, the following holds: E(r,g⋅X)=g⋅E(r,X) E(r, g \cdot X) = g \cdot E(r, X)