2
$\begingroup$

I'm reading some works on the hierarchical model in statistical mechanics and I came across an strange definition, which I need to clarify. Consider a finite set $\Lambda \subset \mathbb{Z}^{d}$. The set of all functions $\varphi : \Lambda \to \mathbb{R}$ is isomorphic to $\mathbb{R}^{|\Lambda|}$, so that these functions are represented as vectors $\varphi = (\varphi_{x})_{x\in \Lambda} \in \mathbb{R}^{|\Lambda|}$, and are called fields. Given $\xi_{1},...,\xi_{N} \in \mathbb{R}^{|\Lambda|}$, we set $\varphi_{j} := \sum_{k> j}\xi_{k}$ for each $j=0,...,N-1$ and $\varphi_{N} = 0$.

Now the author states the following: "Given $X\subset \Lambda$, let $\mathcal{N}_{j}(X)$ be the algebra of functions measurable with respect to the $\sigma$-algebra generated by $\{\varphi_{j}(x), \hspace{0.1cm} x \in X\}$. In more down to earth terms, an element of $\mathcal{N}_{j}(X)$ is a function only of fields at points $x \in X$."

I don't understand what is this $\sigma$-algebra. It seems confusing to me. What does it mean?

Edit: The text can be found in math.ubc.ca/~db5d/Seminars/PCMILectures/lectures.pdf

$\endgroup$
2
  • 3
    $\begingroup$ As a rule for this kind of questions it is good to indicate exactly the reference you are reading. I guess this is the book by Bauerschmidt, Brydges and Slade, correct? $\endgroup$ Jan 16, 2020 at 21:44
  • $\begingroup$ Actually is the Brydges lecture notes on RG. I will add it to the post. $\endgroup$
    – IamWill
    Jan 16, 2020 at 22:14

2 Answers 2

1
$\begingroup$

$\newcommand\vpi{\varphi}$ $\newcommand\La{\Lambda}$ $\newcommand\R{\mathbb R}$ You misunderstood the notes: For each $j\in\{0,\dots,N\}$ and each $x\in\La$, $\vpi_j(x)$ is (not a real number but) a real-valued random variable (r.v.).

Indeed, (i) formula (2.17) on page 25 in the notes implies that $\vpi_j=\sum_{k=j+1}^N\zeta_j$ (with $\vpi_N=0$), (ii) line 2 on page 22 there tells us that $\zeta_j\sim N(C_j)$, and (iii) by formula (1.20) on page 10, $\zeta_j\sim N(C_j)$ means that $\zeta_j=(\zeta_j(x)\colon x\in\La)$ is a random zero-mean Gaussian vector (with values in $\R^\La$ and) with the covariance matrix $C_j=(C_j(x,y))_{(x,y)\in\La^2}$.

So, for each $j\in\{0,\dots,N\}$ and each $X\subseteq\La$, the $\sigma$-algebra in question is the $\sigma$-algebra generated by the set $\{\varphi_j(x)\colon x\in X\}$ of real-valued r.v.'s, that is, the smallest $\sigma$-algebra with respect to which all the r.v.'s $\varphi_j(x)$ with $x\in X$ are measurable.

$\endgroup$
6
  • $\begingroup$ Thanks for your answer! Now I can spot my misunderstanding. But then, for each $j $, $\varphi_{j}(x) $ is a function, right? What is that function? I don't seem to understand how this emerge from $\varphi_{j}=\sum_{k>j} \xi_{k}$. $\endgroup$
    – IamWill
    Jan 17, 2020 at 3:24
  • $\begingroup$ @Willy.K : For each $j$ and each $x$, $\zeta_j(x)$ is a real-valued random variable defined on a probability space $(\Omega,\mathscr F,\mathbb P)$. That is, $\zeta_j(x)$ is an $\mathscr F$-measurable function from $\Omega$ to $\mathbb R$: $\Omega\ni\omega\mapsto\zeta_j(x)(\omega)\in\mathbb R$. Therefore, for each $j$ and each $x$, $\varphi_j(x)$ is an $\mathscr F$-measurable function from $\Omega$ to $\mathbb R$. (Just as in these notes, oftentimes the probability space $(\Omega,\mathscr F,\mathbb P)$ remains in the background and is not mentioned explicitly.) $\endgroup$ Jan 17, 2020 at 14:48
  • $\begingroup$ It's getting clearer to me. There is just one thing that is bothering me: he seems to be using variables $\varphi$ interchangeably as a vector in $\mathbb{R}^{|\Lambda|}$ and as a random vector. For example, in page 10 he writes that, for a positive-definite matrix $A$ in $\mathbb{R}^{|\Lambda|}$, $\mu := \mbox{const} e^{-\frac{1}{2}\langle \varphi, A^{-1}\varphi\rangle}$ is a Gaussian measure. Thus $\varphi \in \mathbb{R}^{|\Lambda|}$ as I understand. Then, he writes $\varphi \sim N(C)$, where $C=A^{-1}$ and here $\varphi$ is a random vector $\varphi: \Omega \to \mathbb{R}^{\Lambda}$. $\endgroup$
    – IamWill
    Jan 17, 2020 at 15:26
  • $\begingroup$ Is this just abuse of notation or both things are essentially the same? $\endgroup$
    – IamWill
    Jan 17, 2020 at 15:26
  • $\begingroup$ @Willy.K : I also noticed that. It is indeed abuse of notation, which may confuse readers. $\endgroup$ Jan 17, 2020 at 15:39
1
$\begingroup$

Disclaimer: maybe this would be more appropriate as a comment since I do not know the cited text, but I am posting this as an answer because I do cannot yet comment:

My understanding would be that it is meaning the $\sigma$-algebra generated by the restriction $\phi_j|_{X}$, that is the $\sigma$-algebra generated by all the sets $C \subseteq \Lambda$ such that $C=\{x\,: \, \phi_j|_X(x) \in C\}$.

By a standard result, a real function $g$ is measurable with respect to a $\sigma$-algebra generated by another real function $f$ if and only if there exist a measurable function $h:\mathbb{R} \to \mathbb{R}$ such that $g=h \circ f$.

So in this case, if the definition is the one I conjectured, then $f \in \mathcal{N}_j$ if and only if there is a measurable $h$ such that $f=h \circ \phi_j|_X$, that is, $f$ would depend only on the values of $\phi_j$ on the set $X$, as the text says. Cannot help with the physics though :)

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.