Categories
Math

Sine Angle Addition Functional Equation

In this post, we investigate further the sine angle addition functional equation that I introduced in a previous post, Trigonometric Identities as Functional Equations. (See the work done there before reading this post.) Specifically, the question is to find all functions f such that

\displaystyle f(a + b) = f(a)f\left( \frac{\pi}{2} - b \right) + f\left( \frac{\pi}{2} - a \right)f(b).

If f is of the form f(x) = c\sin x, what must c be?

\displaystyle c\sin(a + b) = c^{2}\sin(a)\sin\left( \frac{\pi}{2} - b \right) + c^{2}\sin\left( \frac{\pi}{2} - a \right)\sin(b)

\displaystyle c = c^{2}

Thus, c = 0 or c = 1.

Actually, f(x) = 0 works. So let’s assume that f(x) is nonzero, then see all the possible solutions.

If f is of the form f(x) = r\sin(sx + t) + u, what must r,s,t,u be? (Assuming f is nonzero.)

\displaystyle r\sin\left( s(a + b) + t \right) + u = \left( r\sin(sa + t) + u \right)\left( r\sin\left( s\left( \frac{\pi}{2} - b \right) + t \right) + u \right) + \left( r\sin(sb + t) + u \right)\left( r\sin\left( s\left( \frac{\pi}{2} - a \right) + t \right) + u \right),

for any a,b.

First, say s = 0. Then, f(x) = c (constant.)

Are there any nonzero constant solutions?

\displaystyle c = c^{2} + c^{2} \rightarrow c = 2c^{2} \rightarrow c = \frac{1}{2}

Also, we have that f(x) = \frac{1}{2} works, so this is a solution.

Are there any nonconstant linear solutions, f(x) = cx + d with c \neq 0?

\displaystyle c(a + b) + d = (ca + d)\left( c\left( \frac{\pi}{2} - b \right) + d \right) + \left( c\left( \frac{\pi}{2} - a \right) + d \right)(cb + d)

Let a = b = 0:

\displaystyle d = 2d\left( \frac{\pi}{2}c + d \right)

Let a = b = \frac{\pi}{2}:

\displaystyle \pi c + d = 2d\left( \frac{\pi}{2}c + d \right)

Thus, d = \pi c + d \rightarrow c = 0. Therefore, there are no nonconstant linear solutions.

Let’s say f(x) is nonzero, what can we say about f?

We have

\displaystyle f(a) = f(a)f\left( \frac{\pi}{2} \right) + f\left( \frac{\pi}{2} - a \right)f(0)

\displaystyle \rightarrow f(a)\left( f\left( \frac{\pi}{2} \right) - 1 \right) + f\left( \frac{\pi}{2} - a \right)f(0) = 0

Letting c = f\left( \frac{\pi}{2} \right) - 1 and d = f(0), substituting a \rightarrow \frac{\pi}{2} - a,

\displaystyle cf\left( \frac{\pi}{2} - a \right) + df(a) = 0

\displaystyle cf(a) + df\left( \frac{\pi}{2} - a \right) = 0

\displaystyle \rightarrow cdf\left( \frac{\pi}{2} - a \right) + d^{2}f(a) = 0

\displaystyle c^{2}f(a) + dcf\left( \frac{\pi}{2} - a \right) = 0

\displaystyle \left( d^{2} - c^{2} \right)f(a) = 0

Thus, for f(x) to be nonzero, we must have d^{2} = c^{2}. Hence,

\displaystyle f(0)^{2} = \left( f\left( \frac{\pi}{2} \right) - 1 \right)^{2}.

But from before we also had that either

\displaystyle f(0) = \pm \frac{1}{2}\ and\ f\left( \frac{\pi}{2} \right) = \frac{1}{2},

or

\displaystyle f(0) = 0\ and\ f\left( \frac{\pi}{2} \right) \in \left\{ 0,1 \right\}.

From our new relation, we can eliminate the possibility that f(0) = 0,f\left( \frac{\pi}{2} \right) = 0. However, all the others still work, so we can have

\displaystyle f(0) = 0\ and\ f\left( \frac{\pi}{2} \right) = 1,

or

\displaystyle f(0) = \pm \frac{1}{2}\ and\ f\left( \frac{\pi}{2} \right) = \frac{1}{2}.

From the above, we have that if f(0) = f\left( \frac{\pi}{2} \right) = 0 then f must be zero. We still have the possibility f(0) = f\left( \frac{\pi}{2} \right) = \frac{1}{2}. We do know that f(x) = \frac{1}{2} is a solution; is it true that this is the only solution for f(0) = f\left( \frac{\pi}{2} \right) = \frac{1}{2}? Equivalently, can we show that if f(0) = f\left( \frac{\pi}{2} \right) then f must be constant? Then, to find nonconstant solutions we would only need to consider either (f(0) = 0 and f\left( \frac{\pi}{2} \right) = 1) or (f(0) = - \frac{1}{2} and f\left( \frac{\pi}{2} \right) = \frac{1}{2}.)

Thus, our conjecture is: if f(0) = f\left( \frac{\pi}{2} \right) then f is constant. Let’s see how we can show this. We equivalently can have f(0) = f\left( \frac{\pi}{2} \right) = \frac{1}{2} and show that f(x) = \frac{1}{2}.

Can we identify any other solutions quickly? So far we have f(x) = 0,f(x) = \frac{1}{2},f(x) = \sin x.

Well, we have \sin 0 = 0,\sin\frac{\pi}{2} = 1. We suspect that finding the sinusoidal function so that f(0) = - \frac{1}{2},f\left( \frac{\pi}{2} \right) = \frac{1}{2} may yield another solution.

So say f(x) = r\sin(sx + t) + u, and

\displaystyle f(0) = r\sin t + u = - \frac{1}{2},

\displaystyle f\left( \frac{\pi}{2} \right) = r\sin\left( \frac{\pi}{2}s + t \right) + u = \frac{1}{2}.

We know that f(x) = \sin x - \frac{1}{2} is a sinusoidal function that works. Is this a solution?

\displaystyle \sin(a + b) - \frac{1}{2} = \left( \sin a - \frac{1}{2} \right)\left( \sin\left( \frac{\pi}{2} - b \right) - \frac{1}{2} \right) + \left( \sin\left( \frac{\pi}{2} - a \right) - \frac{1}{2} \right)\left( \sin b - \frac{1}{2} \right)

\displaystyle = \left( \sin a - \frac{1}{2} \right)\left( \cos b - \frac{1}{2} \right) + \left( \cos a - \frac{1}{2} \right)\left( \sin b - \frac{1}{2} \right)

\displaystyle = \sin a\cos b - \frac{1}{2}\cos b - \frac{1}{2}\sin a + \frac{1}{4} + \cos a\sin b - \frac{1}{2}\sin b - \frac{1}{2}\cos a + \frac{1}{4}

\displaystyle = \sin(a + b) + \frac{1}{4} - \left( \frac{1}{2}\cos b + \frac{1}{2}\sin a + \frac{1}{2}\sin b + \frac{1}{2}\sin a \right)

\displaystyle \rightarrow \frac{1}{4} + \frac{1}{2} = \frac{1}{2}\cos b + \frac{1}{2}\sin a + \frac{1}{2}\sin b + \frac{1}{2}\sin a

\displaystyle \rightarrow \frac{3}{2} = \sin a + \cos a + \sin b + \cos b

which is clearly false. (Just pick a = b = 0.) Thus, this is not a solution.

Let’s say f,g are two solutions. What can we say about their difference, h = f - g?

\displaystyle f(a + b) = f(a)f\left( \frac{\pi}{2} - b \right) + f\left( \frac{\pi}{2} - a \right)f(b)

\displaystyle g(a + b) = g(a)g\left( \frac{\pi}{2} - b \right) + g\left( \frac{\pi}{2} - a \right)g(b)

\displaystyle h(a + b) + g(a + b) = \left( h(a) + g(a) \right)\left( h\left( \frac{\pi}{2} - b \right) + g\left( \frac{\pi}{2} - b \right) \right) + \left( h\left( \frac{\pi}{2} - a \right) + g\left( \frac{\pi}{2} - a \right) \right)\left( h(b) + g(b) \right)

Expand the RHS:

\displaystyle h(a)h\left( \frac{\pi}{2} - b \right) + h(a)g\left( \frac{\pi}{2} - b \right) + g(a)h\left( \frac{\pi}{2} - b \right) + g(a)g\left( \frac{\pi}{2} - b \right) + h\left( \frac{\pi}{2} - a \right)h(b) + h\left( \frac{\pi}{2} - a \right)g(b) + g\left( \frac{\pi}{2} - a \right)h(b) + g\left( \frac{\pi}{2} - a \right)g(b)

\displaystyle = \left( g(a)g\left( \frac{\pi}{2} - b \right) + g\left( \frac{\pi}{2} - a \right)g(b) \right) + \left( h(a)h\left( \frac{\pi}{2} - b \right) + h\left( \frac{\pi}{2} - a \right)h(b) \right) + \left( h(a)g\left( \frac{\pi}{2} - b \right) + g(a)h\left( \frac{\pi}{2} - b \right) + h\left( \frac{\pi}{2} - a \right)g(b) + g\left( \frac{\pi}{2} - a \right)h(b) \right)

\displaystyle \rightarrow h(a + b) - \left( h(a)h\left( \frac{\pi}{2} - b \right) + h\left( \frac{\pi}{2} - a \right)h(b) \right) = h(a)g\left( \frac{\pi}{2} - b \right) + g(a)h\left( \frac{\pi}{2} - b \right) + h\left( \frac{\pi}{2} - a \right)g(b) + g\left( \frac{\pi}{2} - a \right)h(b)

Unfortunately it doesn’t look like there’s a good or easy way to simplify the RHS to derive something about h

Now let’s go back to finding all (nonconstant) sinusoidal solutions. This means s \neq 0 and also r \neq 0. The hope was that information about the difference between solutions could help us say something about u, but oh well … in fact, this might be easier with a more specific sinusoidal form.

We have

\displaystyle r\sin\left( s(a + b) + t \right) + u = \left( r\sin(sa + t) + u \right)\left( r\sin\left( s\left( \frac{\pi}{2} - b \right) + t \right) + u \right) + \left( r\sin(sb + t) + u \right)\left( r\sin\left( s\left( \frac{\pi}{2} - a \right) + t \right) + u \right).

Let a = b = 0:

\displaystyle r\sin t + u = \left( r\sin t + u \right)\left( r\sin\left( \frac{\pi}{2}s + t \right) + u \right)

Let a = b = \frac{\pi}{2}:

\displaystyle r\sin(\pi s + t) + u = \left( r\sin t + u \right)\left( r\sin\left( \frac{\pi}{2}s + t \right) + u \right)

Thus,

\displaystyle r\sin t = r\sin(\pi s + t),

and since r \neq 0,

\displaystyle \sin t = \sin(\pi s + t).

Also, either r\sin t + u = 0 or

\displaystyle r\sin\left( \frac{\pi}{2}s + t \right) + u = 1.

Take the first case, r\sin t + u = 0. If \sin t = 0, then u = 0 and \sin(\pi s + t) = 0. We have t = n\pi for some n\mathbb{\in Z}, so \sin\left( \pi(s + n) \right) = 0 and s\mathbb{\in Z}. Hence, the form is

\displaystyle f(x) = r\sin(sx + n\pi),

and the offset provided by n\pi can be absorbed into r, yielding

\displaystyle f(x) = r\sin(sx),s\mathbb{\in Z.}

What solutions of this form are there?

It remains to investigate this.

We continue discussing this equation in a follow-up post.

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.