TrigReduce applied to certain terms? Finding coefficients of trig terms

I have an expression that involves trig functions of t and y. I would like to find the coefficients of Cos[t], Sin[t], Cos[2 t], Sin[2 t] ... and the constant term independent of t. I can get the terms in t and y separated by using TrigExpand however this then leaves me with trig expressions like Cos[t]^2. To reduce these to terms in Cos[n t] etc. I have to apply TrigReduce to them but NOT to the terms involving y. TrigReduce does not have a second parameter to tell it to just reduce terms in say t. Can we make one?

My workaround is to use FourierTrigSeries guessing the highest trig multiple I need (here 5 but I don't want to guess) and then to use Coefficient to find the individual terms and the constant term. This is a bit long. Are there other ways?

Thanks

Here is a simple example. I will need to do more complex examples.

e1 = -(Cos[t] - E^(-(y/Sqrt)) Cos[t - y/Sqrt])^2 +
1/2 ((E^(-(y/Sqrt)) Cos[t - y/Sqrt])/Sqrt - (
E^(-(y/Sqrt)) Sin[t - y/Sqrt])/Sqrt[
2]) (-(Sqrt - 2 y) Cos[t] +
Sqrt E^(-(y/Sqrt[
2])) (Cos[t - y/Sqrt] - E^(y/Sqrt) Sin[t] +
Sin[t - y/Sqrt])) ;

e2 = e1 // TrigExpand;
e3 = FourierTrigSeries[e2, t, 5];
trigs = Flatten@Table[{Cos[n t], Sin[n t]}, {n, 5}];
coffs = Coefficient[e3, #] & /@ trigs;
const = e3 - coffs.trigs;

The coefficients are

coffs // TableForm and the constant terms is

const A sanity check gives

Simplify[e1 == const + coffs.trigs]

(* True  *)

Try the following: These are the transformation rules:

rules = {Cos[t]^n_Integer :> TrigReduce[Cos[t]^n],
Sin[t]^n_Integer :> TrigReduce[Sin[t]^n],
Times[a_, Cos[t] , Sin[t], c_] :> a*c*TrigReduce[Cos[t] Sin[t]]};

With these rules

Collect[e2 /. rules, {Cos[2 t], Sin[2 t]}]

yields You might want to simplify the coefficients of the sin(2t) and cos(2t) as follows:

MapAt[Simplify,
Collect[e2 /. rules, {Cos[2 t], Sin[2 t]}], {{7, 2}, {8, 2}}]

giving the following Have fun!

• Good idea, write my own rules. Thanks. – Hugh Nov 9 '17 at 14:19

One idea is to use FourierCoefficient to compute the general term:

f[n_] = FourierCoefficient[e1, t, n];
f[n] //TeXForm

$\begin{cases} \frac{1}{8} \left(-2+e^{-\frac{(1-i) y}{\sqrt{2}}} \left(2+(1-i) \sqrt{2} y\right)\right) & n=-2 \\ \frac{1}{8} \left(-2+e^{-\frac{(1+i) y}{\sqrt{2}}} \left(2+(1+i) \sqrt{2} y\right)\right) & n=2 \\ \frac{1}{8} \left(e^{-\frac{(1+i) y}{\sqrt{2}}} \left((1+i) \sqrt{2} y+(1-i) e^{i \sqrt{2} y} \left(\sqrt{2} y+(1+3 i)\right)+(4-2 i)\right)-4 e^{-\sqrt{2} y}-4\right) & n=0 \end{cases}$

We can then recover the Sin and Cos coefficients as follows:

sin = Simplify @* ComplexExpand /@ Simplify[I(f[n]-f[-n]), n>0];
sin //TeXForm

$\begin{cases} \frac{1}{4} e^{-\frac{y}{\sqrt{2}}} \left(\left(\sqrt{2} y+2\right) \sin \left(\frac{y}{\sqrt{2}}\right)-\sqrt{2} y \cos \left(\frac{y}{\sqrt{2}}\right)\right) & n=2 \\ 0 & \operatorname{True} \end{cases}$

cos = Simplify @* ComplexExpand /@ Simplify[f[n] + f[-n], n>0];
cos //TeXForm

$\begin{cases} \frac{1}{4} e^{-\frac{y}{\sqrt{2}}} \left(-2 e^{\frac{y}{\sqrt{2}}}+\sqrt{2} y \sin \left(\frac{y}{\sqrt{2}}\right)+\left(\sqrt{2} y+2\right) \cos \left(\frac{y}{\sqrt{2}}\right)\right) & n=2 \\ 0 & \operatorname{True} \end{cases}$

zero = Simplify @* ComplexExpand @ f;
zero //TeXForm

$\frac{1}{4} e^{-\sqrt{2} y} \left(-2 \left(e^{\sqrt{2} y}+1\right)+e^{\frac{y}{\sqrt{2}}} \left(\sqrt{2} y-2\right) \sin \left(\frac{y}{\sqrt{2}}\right)+e^{\frac{y}{\sqrt{2}}} \left(\sqrt{2} y+4\right) \cos \left(\frac{y}{\sqrt{2}}\right)\right)$

Using FourierSinCoefficient/FourierCosCoefficient to obtain the general term doesn't work quite as well, but we can use them to check the above results:

(sin /. n->2) == FourierSinCoefficient[e1, t, 2]
(cos /. n->2) == FourierCosCoefficient[e1, t, 2]

True

True

• This is a good way of getting a general term. One problem is how to find how many terms there are and FourierCoefficient is certainly a good way of doing this. Although you have to recombine terms this may be the best method. I am not sure what happens if there is no general term. It seems a shame that one has to go to the effort of taking a Fourier transform when all that is needed is to do algebra to find the trig terms. Thanks for your help. – Hugh Nov 9 '17 at 19:12

You can try also with this:

e1 = -(Cos[t] - E^(-(y/Sqrt)) Cos[t - y/Sqrt])^2 + 1/2 ((E^(-(y/Sqrt)) Cos[t - y/Sqrt])/
Sqrt - (E^(-(y/Sqrt)) Sin[t - y/Sqrt])/
Sqrt) (-(Sqrt - 2 y) Cos[t] +
Sqrt E^(-(y/Sqrt)) (Cos[t - y/Sqrt] -
E^(y/Sqrt) Sin[t] + Sin[t - y/Sqrt]));
CoefficientList[FourierTrigSeries[TrigExpand@e1, t, 5] /. {Cos[a_*t] :>  x^a, Sin[a_*t] :>  z^a}, {x, z}] After obtaining the Fourier series you can transform the trig polynomial to a usual polynomials and then extract the coefficients.

I think you can save steps.

• Interesting idea. Are there dangers in replacing a trig with an exponent? I will try this out. Thanks – Hugh Nov 9 '17 at 15:41