What is an ideal lattice?

1k Views Asked by At

Can anyone describe in layman's terms what an ideal lattice is? I've seen them mentioned in many places, but haven't found a good definition of what exactly they are, nor any good terms to know where to start my search to find more information.

I'm not exactly sure what a non ideal lattice might be either :P

1

There are 1 best solutions below

2
On

I don't know much about ideal lattices outside the definition, but I can start you off with what a lattice is. If $n \geq 1$, a lattice in $V = \mathbb{R}^n$ is an additive subgroup of the form $$\Gamma = \mathbb{Z}v_1 + \cdots + \mathbb{Z}v_t$$ where $v_1, ... , v_t$ are linearly independent over $\mathbb{R}$. If $t = n$, then $\Gamma$ is called a full lattice (many authors mean full lattice when they write lattice). In other words, a lattice is a free abelian group (that is, an abelian group with a $\mathbb{Z}$-basis), with a basis which is also linearly independent over $\mathbb{R}$.

For example, if $V = \mathbb{R}$, then $\Gamma = \mathbb{Z} + \mathbb{Z}\pi$ is not a lattice, because the $\mathbb{Z}$-basis $1, \pi$ for $\Gamma$ is not linearly independent over $\mathbb{R}$.

But if $V = \mathbb{R}^3$, then $\Gamma = \mathbb{Z}(1,0,0) + \mathbb{Z}(1,0,1)$ is a lattice (of rank two).

Actually, there are several equivalent conditions for an additive subgroup $\Gamma$ of $V$ to be a lattice:

(i) $\Gamma$ is a lattice.

(ii) $\Gamma$ is a discrete subgroup (at every point $x$ in $\Gamma$, you can find an open set in $\mathbb{R}^n$ about $x$ which contains no other point of $\Gamma$)

(iii) Every bounded subset of $\mathbb{R}^n$ intersects $\Gamma$ at most at finitely many points.

You can also talk about quotient groups. If $\Gamma$ is a lattice in $V = \mathbb{R}^n$, you can talk about the quotient group $V/\Gamma$. For example, if we take $V = \mathbb{R}$ and the lattice $\Gamma = \mathbb{Z}$, then $V/ \Gamma = \mathbb{R}/\mathbb{Z}$ is really just the circle. As a topological group, the circle is compact. More generally, if $\Gamma = \mathbb{Z}v_1 + \cdots + \mathbb{Z}v_t$ for $v_1, ... v_t$ linearly independent over $\mathbb{R}$, you can extend $v_1, ... , v_t$ to a basis $v_{t+1}, ... , v_n$ for $V$. Then the quotient group $V/\Gamma$ is $$(\mathbb{R}v_1 + \cdots + \mathbb{R}v_n)/(\mathbb{Z}v_1 + \cdots + \mathbb{Z}v_t) \cong \frac{\prod\limits_{i=1}^n \mathbb{R}}{\prod\limits_{i=1}^t \mathbb{Z} \times \prod\limits_{i=t+1}^n \{0\}} $$ $$ \cong \prod\limits_{i=1}^t \mathbb{R}/\mathbb{Z} \times \prod\limits_{i=t+1}^n \mathbb{R} $$ Thus $\Gamma$ is a full lattice if and only if $t = n$, if and only if $V/\Gamma$ is compact. For detailed proofs, see J.S. Milne's excellent notes on algebraic number theory.