• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

Integral of the Dirac Delta Function

Quasar Chunawala

Quasar Chunawala

Joined
11/5/14
Messages
294
Points
53
Hi guys,

In the introductory chapter in Shreve II, he talks about a sequence of functions [imath](f_n)_{n=0}^{\infty}[/imath] converging to [imath]f[/imath] almost everywhere. He further gives an example of the sequence of functions [imath](f_n)[/imath], where [imath]f_n[/imath] is the density of Gaussian [imath]N(0,\frac{1}{n})[/imath] random variable:

[math] f_n(x) = \sqrt{\frac{n}{2\pi}}e^{-\frac{nx^2}{2}} [/math]
If we fix [imath]x = x_0, x_0 \neq 0[/imath], its easy to see that the limit [imath]\lim_{n \to \infty} \frac{\sqrt{n}}{e^{n{x_0}^{2}/2}}[/imath] by the L'Hopital's rule is equal to [imath]\lim \frac{1}{(x_0^2/2)} \cdot \frac{1}{2\sqrt{n}} \cdot e^{-n\frac{x_0^2}{2}}[/imath]. Since, the exponential term is bounded, the limit approaches zero.

At [imath]x = 0[/imath], [imath]f_n(0)[/imath] is unbounded. So, [imath](f_n)[/imath] converges pointwise to :

[math] f(x) = \begin{cases}0 & \text{ if } x \neq 0 \\ \infty & \text{ if } x = 0\end{cases} [/math]
I recognize that this is the Dirac Delta function. If we were to try and find the Riemann integral of [imath]f[/imath], the lower integral is [imath]0[/imath], but the upper integral is unbounded, regardless of the partition [imath]P[/imath]. But, I read that, [imath]\int \delta(x) dx = 1[/imath], somehow, magically; so how does this come about..

Thanks,
Quasar
 
Last edited:
The above analysis is wrong, in essence. Classical maths breaks down because Dirac is a generalised function aka distribution.

Read this background first and we can talk further :cool:

www.cantorsparadise.com

Distributions: What Exactly is the Dirac Delta “Function”?

The theory of distributions , also known as generalized functions, often remains something of an enigma…
www.cantorsparadise.com www.cantorsparadise.com

L'Hopital's rule
Riemann
no.
 
Last edited:
That was the theory of distributions 101 for me.

So, a generalized function or distribution [imath]T[/imath] is any scattering of unit mass over the real line. And to describe the distribution, we are interested in the density/mass over small intervals, so we use measurement devices aka test functions [imath]\phi \in C^{\infty}_c[/imath] such as the bump function. And we describe the distribution by its action on the test functions; how it maps [imath]\phi[/imath] to a real number [imath]\int_{-\infty}^{\infty} f(x) \phi(x) dx[/imath].

The definition of differentiability of classical functions is too strong, so we define a weak(generalized) derivative, by the rule [imath]T'(\phi) =T(-\phi')[/imath].

Dirac distribution concentrates all mass at a singular point.
 
Last edited:
I didn't realize this, until I actually read it on Wilmott. The delta of a European call with time to expiration [imath]0[/imath] is just the heavyside function, so the gamma of the call is the Dirac distribution.
 
It's nice, inspiring to read Wilmott, before winding up the day; I especially like some of the math brain-teasers, you ask and the general conversation there. I always learn something new.

I am hoping to learn and explore some of the methods, in the thesis paper you shared, during the PDE/FDM course (hope to start in October).
 
Last edited:
Back
Top