Alexander Bass

Math Notes Fall 25

Proof by multiplicative inverse

We have functions f(x),g(x),h(x) .

Given h(x)f(x)=1 . (they are multiplicative inverses)

We would like to determine whether f(x)=g(x) .

Naturally, h(x)g(x)=1  f(x)=g(x)

Application

I discovered this trick when searching for a proof to Euler’s Formula:

eit=cost+isint

We define

f(x)=eitg(x)=(cost+isint)

Naturally, eith(x)=1 , so

h(x)=eit

So, if we can prove that h(x)g(x)=1 , then it is clear that f(x)=g(x) . Introduce j(x) for brevity.

h(x)g(x)=j(x)=eit(cost+isint)

Differentiate j(x)

j(x)=eit(sint+icost)ieit(cost+isint)=eit(sint+icost)+eit(icost+sint)=eit(icostsint+sinticost)=0

That is, j(x)=0 for all x , which means that j(x) is constant for all x .

Find value of j(x)

j(0)=e0(cos0+isin0)=1

So, j(x)=1 for all x , which shows that g(x) is the multiplicative inverse of h(x) . Given that g(x) is also the multiplicative inverse of h(x) :

f(x)=g(x)

Source: https://math.stackexchange.com/a/8612

Complex Exponentials and Trigonometric functions

The value of the exponential function with an imaginary input is of interest.

eix=?

Using the well known series expansion of the exponential function:

ex=n=0xnn!

Substitute in imaginary constant

eix=n=0(i)nxnn!

There is no direct way to simplify this expression as (i)n doesn’t directly work out to anything nice. This expression is equivalent to a few other, more-useful ones.

eix=n=0(i)nxnn!=n=0(i)2nx2n(2n)!+(i)2n+1x2n+1(2n+1)!=n=0(i)4nx4n(4n)!+(i)4n+1x4n+1(4n+1)!+(i)4n+2x4n+2(4n+2)!+(i)4n+3x4n+3(4n+3)!

Now the powers of i can be simplified

i4n=i4n=1i4n+1=i4ni=ii4n+2=i4ni2=1i4n+3=i4ni3=i

Back-substituting these into the above sum

eix=n=0x4n(4n)!+ix4n+1(4n+1)!x4n+2(4n+2)!ix4n+3(4n+3)!

Remember the power series for sine and cosine

sin(x)=n=0x2n+1(1)n(2n+1)!cos(x)=n=0x2n(1)n(2n)!

Reorder series

eix=n=0x4n(4n)!x4n+2(4n+2)!+ix4n+1(4n+1)!ix4n+3(4n+3)!

Split sums based on evens, and odds

eix=[n=0x4n(4n)!x4n+2(4n+2)!]+[in=0x4n+1(4n+1)!x4n+3(4n+3)!]

Notice the first sum is simply iterating over the even numbers, and the second is iterating over the odds. Simplify

eix=[n=0x2n(1)n(2n)!]+[in=0x2n+1(1)n(2n+1)!]

Notice

eix=cosx+isinx

With Negatives

Recall these two properties of sine and cosine

sin(x)=sin(x)cos(x)=cos(x)

With that, use the exponential function with a negative imaginary number

eix=cosxisinx

We can now solve for cosine and sine using the exponential function.

eix+eix=2cosx

cosx=eix+eix2 eixeix=2sinx

sinx=eixeix2

Clever proof of solutions to linear diophantine equation

A simple result in elementary number theory relates to the linear diophantine equation

ax+by=c

a,x,b,y,c

A property of this equation (which is not proven here) is that it has infinitely many (x,y) solutions if c is a multiple of gcd(a,b) , otherwise it has none.

The clever part is, after finding the first solution we can find the rest. Suppose there exists a solution (x,y) , and that (x,y) is another solution:

c=ax+by=ax+by

Rearrange equation

a(xx)=b(yy)

Let d=gcd(a,b) . Naturally, a and b are both divisible by d . Divide by d

ad(xx)=bd(yy)

Now, a/d and b/d share no common factors, therefor a/d does not divide b/d and b/d does not divide a/b . Note however that b/d does divide the right side of the equation, and therefor must divide the left side of the equation. Because we already determined that b/d does not divide a/d , b/d must divide (xx) , thus (xx) must be a multiple of b/d (likewise (yy) is a multiple of a/d ).

State that (xx) is a multiple of b/d in a new equation:

bnd=xx

Likewise of y

and=yy

Thus, given the initial solutions of x,y , the other solutions are found as:

bnd+x=x

and+y=y

This proof is not complete but highlights the clever trick of extracting an equation based on divisibility. It comes down to the (often subtle) fact that when x divides y , y=xq .