Defining functions in terms of differential equations

1. Jul 2, 2013

Sigma057

I have set myself the task of teaching my Freshman in high school brother Calculus, and today while reviewing some topics I saw something I didn't see before.

To start out, I let y = ln[x] => x = e^y
Obviously, we know that y' = 1/x = e^-y

So, I "discovered" that one can define the natural logarithm as the solution to the IVP

y'[x] = e^-y[x], y[1]=0

I know this is a somewhat trivial example,
But my question is, is there anything to be gained by defining functions in this way in general?

2. Jul 2, 2013

krome

Absolutely! Check out https://www.physicsforums.com/showthread.php?t=105050. It's kind of a long thread about the proof of the orthogonality of Legendre polynomials. I attached a pdf there (mine is the fifth message) showing such a proof which relies solely on the fact that Legendre polynomials satisfy a particular differential equation.

3. Jul 2, 2013

JJacquelin

Hi !
almost all elementary and special functions can be defined in several ways, for example : integral, differential equation, series, inverse of an already defined function, etc.
This allows to prove many relationships. Using a convenient definition often simplifies the calculus or developments.
Comming back to the differential équations, a list of the most common cases is presented in the pdf "Safari in the Contry of Special Functions":
http://www.scribd.com/JJacquelin/documents
The differential equations appear pp.30-36 in Table 1 : usual functions, Table 2 : special polynomials, Tables 6-9 : examples of special functions.

4. Jul 3, 2013

HallsofIvy

Staff Emeritus
You can, for example, define "cos(x)" as "The solution to the differential equation, y''= -y, with initial conditions y(0)= 1, y'(0)= 0". You can similarly define "sin(x)" as "The solution to the differential equation, y''= -y, with initial conditions y(0)= 0, y'(0)= 1.

You can, then, prove all of the properties of those functions from those definitions.

This is better, for example, than just "arbitrarily" extending the usual "right triangle" definitions, which, of course, require that the angle be between 0 and 90 degrees, to any value of x.