Use rescaling or translation to normalize parameters

a repository of mathematical know-how

Use rescaling or translation to normalize parameters

Title: *

Area of mathematics: *

A comma-separated list of areas of mathematics to which this article applies. Use ">" to tag in a subcategory. Example: Analysis > Harmonic analysis, Combinatorics

Keywords:

A comma-separated list of keywords associated with this article. Example: free group

Used in:

A comma-separated list of examples of where this technique is used. Example: Cauchy-Schwarz inequality

Parent articles:		Order

Body:

[QUICK DESCRIPTION]

One can often use rescaling transformations such as $x \mapsto \lambda x$ or translation transformations such as $x \mapsto x-x_0$ to normalize one or more parameters in the integral to equal a particularly simple value, such as 0 or 1.  These changes of variables can create additional factors, but usually such factors can be moved outside of the integral, thus simplifying the remaining integrand.

[PREREQUISITES]

Undergraduate calculus; complex analysis

[EXAMPLE]

In the integral $\int_0^\infty \frac{1}{a^7+x^7}\ dx$ discussed [[divide and conquer|on this page]], one can use the substitution $x = ay$ to change this integral to $a^{-6} \int_0^\infty \frac{dy}{1+y^7}$, which in fact already solves the problem except for the task of computing the numerical quantity $\int_0^\infty \frac{dy}{1+y^7}$ (but if one is willing to lose multiplicative constants, one does not even need to do that, once one checks of course that this quantity is finite).

[EXAMPLE]

'''Problem:''' (Fourier transform of Gaussians) Compute the integral $\int_\R e^{-\pi (x-x_0)^2 / r^2} e^{- 2\pi i x \xi}\ dx$, where $r > 0$ and $x_0, \xi \in \R$ are parameters.

'''Solution:''' We begin by using the translation substitution $x' := x-x_0$ to eliminate the $x_0$ factor:

[math] \int_\R e^{-\pi (x')^2 / r^2} e^{- 2\pi i x' \xi} e^{-2\pi i x_0 \xi}\ dx'.[/math]

The factor $e^{-2\pi i x_0 \xi}$ is independent of $x'$ and can be moved out of the integral.  Next, we use the scaling substitution $x' =: ry$ to rewrite the integral as

[math] r e^{-2\pi i x_0 \xi} \int_\R e^{-\pi y^2} e^{-2\pi i r y \xi}\ dy.[/math]

(One could also have used a similar rescaling to eliminate the $\pi$ factors, but they will turn out to be rather convenient for us, actually, so we will keep them.)  The next trick is to [[complete the square]], writing the above integral as

[math] r e^{-2\pi i x_0 \xi} \int_\R e^{-\pi (y+ir\xi)^2} e^{-\pi r^2 \xi^2}\ dy.[/math]

One of the reasons we complete the square is because we can move the $y$-independent factor out of the integral:

[math] r e^{-2\pi i x_0 \xi} e^{-\pi r^2 \xi^2} \int_\R e^{-\pi (y+ir\xi)^2} \ dy.[/math]

Now we would like to normalize away the $ir\xi$ factor by making the substitution $z = y+ir\xi$.  This turns the real integral into a contour integral, but one should not be afraid of this, and obtain

[math] r e^{-2\pi i x_0 \xi} e^{-\pi r^2 \xi^2} \int_{-\infty+ir\xi}^{\infty+ir\xi} e^{-\pi z^2} \ dz.[/math]

But now we can apply the [[Shift the contour of integration|contour shifting]] method and get rid of $ir\xi$ altogether:

[math] r e^{-2\pi i x_0 \xi} e^{-\pi r^2 \xi^2} \int_{-\infty}^{\infty} e^{-\pi z^2} \ dz.[/math]

Now we use the standard integral $\int_\R e^{-\pi x^2}\ dx = 1$ (computed at the [[square and rearrange]] page) and arrive at our final answer of $r e^{-2\pi i x_0 \xi} e^{-\pi r^2 \xi^2}$.
 
[GENERAL DISCUSSION]

It is often a good tactic to perform as many normalizations as one can '''first''', as this can eliminate many parameters which would otherwise clutter up the arguments later.  However, there are exceptions to this rule:

* Sometimes, one wants to cut up an integral into many pieces, and perform a ''different'' normalization for each piece.  In such cases, a premature normalization may be counter-productive.
* In other cases, one wants to do something sneaky with the parameter, e.g. differentiate with respect to it.  Again, in that case it may be counter-productive to normalize the parameter prematurely.

One can also use [[dimensional analysis]] as a substitute for the rescaling normalisation (examples needed).

This is a stub

A stub is an article that is not sufficiently complete to be interesting.

Notifications

File attachments

Changes made to the attachments are not permanent until you save this post. The first "listed" file will be included in RSS feeds.

Attach new file:

Images are larger than 640x480 will be resized. The maximum upload size is 1 MB. Only files with the following extensions may be uploaded: jpg jpeg gif png svg.

Revision information

Log message:

An explanation of the additions or updates being made to help other authors understand your motivations.

Use rescaling or translation to normalize parameters

Recent articles

Active forum topics

Recent comments