PlanetMath: multi-index notation

(more info)

Math for the people, by the people.

Main Menu

sections Encyclopædia
Papers
Books
Expositions

meta Requests (210)
Orphanage
Unclass'd (1)
Unproven (472)
Corrections (37)
Classification

talkback Polls
Forums
Feedback
Bug Reports

downloads Snapshots
PM Book

information News
Docs
Wiki
ChangeLog
TODO List
Legalese
About

Entry average rating: No information on entry rating

multi-index notation

(Definition)

Multi-indices form a powerful notational device for keeping track of multiple derivatives or multiple powers. In many respects these resemble natural numbers. For example, one can define the factorial, binomial coefficients, and derivatives for multi-indices. Using these one can state traditional results such as the multinomial theorem, Leibniz' rule, Taylor's formula, etc. very concisely. In fact, the multi-dimensional results are more or less obtained simply by replacing usual indices in $\mathbb{N}$ with multi-indices. See below for examples.

Definition A multi-index is an -tuple $\alpha=(\alpha_1,\ldots, \alpha_n)$ of non-negative integers $\alpha_1,\ldots, \alpha_n$ . In other words, $\alpha \in \mathbb{N}^n$ . Usually, is the dimension of the underlying space. Therefore, when dealing with multi-indices, is usually assumed clear from the context.

Operations on multi-indices

For a multi-index $\alpha$ , we define the length (or order) as

$\displaystyle \vert\alpha\vert = \alpha_1+\cdots + \alpha_n,$

and the factorial as

$\displaystyle \alpha! = \prod_{k=1}^n \alpha_k!.$

If $\alpha=(\alpha_1,\ldots, \alpha_n)$ and $\beta=(\beta_1,\ldots, \beta_n)$ are two multi-indices, their sum and difference is defined component-wise as

$\displaystyle \alpha+\beta$	$\displaystyle =$	$\displaystyle (\alpha_1+\beta_1, \ldots, \alpha_n+\beta_n),$
$\displaystyle \alpha-\beta$	$\displaystyle =$	$\displaystyle (\alpha_1-\beta_1, \ldots, \alpha_n-\beta_n).$

Thus $\vert\alpha\pm \beta\vert = \vert\alpha\vert\pm \vert\beta\vert$ . Also, if $\beta_k\le \alpha_k$ for all $k=1,\ldots, n$ , then we write $\beta\le \alpha$ . For multi-indices $\alpha,\beta$ , with $\beta\le \alpha$ , we define

$\displaystyle {\alpha\choose \beta} = \frac{\alpha!}{(\alpha-\beta)! \beta!}.$

For a point $x=(x_1,\ldots, x_n)$ in $\mathbb{R}^n$ (with standard coordinates) we define

$\displaystyle x^\alpha= \prod_{k=1}^n x_k^{\alpha_k}.$

Also, if $f\colon \mathbb{R}^n \to \mathbb{R}$ is a smooth function, and $\alpha=(\alpha_1,\ldots, \alpha_n)$ is a multi-index, we define

$\displaystyle \partial^\alpha f = \frac{\partial^{\vert\alpha\vert}}{\partial^{\alpha_1}e_1 \cdots \partial^{\alpha_n}e_n} f,$

where $e_1,\ldots, e_n$ are the standard unit vectors of $\mathbb{R}^n$ . Since

is sufficiently smooth, the order in which the derivations are performed is irrelevant. For multi-indices $\alpha$ and $\beta$ , we thus have

$\displaystyle \partial^\alpha \partial^\beta = \partial^{\alpha+\beta} = \partial^{\beta+\alpha} = \partial^\beta \partial^\alpha.$

Examples

If is a positive integer, and $x_1,\ldots, x_k$ are complex numbers, the multinomial expansion states that
$\displaystyle (x_1+\cdots + x_k)^n = n! \sum_{\vert\alpha\vert=n} \frac{x^\alpha}{\alpha!},$
where $x=(x_1,\ldots, x_k)$ and $\alpha$ is a multi-index. (proof)
Leibniz' rule: If $f,g\colon\mathbb{R}^n \to \mathbb{R}$ are smooth functions, and $\beta$ is a multi-index, then
$\displaystyle \partial^\beta(fg) = \sum_{\alpha\le \beta} {\beta \choose \alpha} \partial^\alpha(f)\, \partial^{\beta-\alpha}(g),$
where $\alpha$ is a multi-index.