Suppose we are seeking the primitive $$\int R(\cos x, \sin x) dx$$ where $R(u,v)=\frac {P(u,v)}{ Q(u,v)}$ is a two-variable rational function. Show that

(a) if $R(−u,v)= R(u,v)$,then $R(u,v)$ has the form $R_1(u^2,v)$;

(b) if $R(−u, v) = −R(u, v)$, then $R(u, v) = u \cdot R_2(u^2, v)$ and the substitution $t = \sin x$ rationalizes the integral above;

(c) If $R(−u, −v) = R(u, v)$, then $R(u, v) = R_3( u , v^2)$, and the substitution $t =\tan x$ rationalizes the integral above.

My attempt: Suppose $$P(x)=\sum_{i,j\geq0, i+j\leq n}{a_{ij}x^iy^j} =\sum_{i,j\geq0, 2i+j\leq n}{a_{2i,j}x^{2i}y^{j}}+\sum_{i,j\geq0, 2i+j\leq n}{a_{2i+1,j}x^{2i+1}y^{j}}=P_1(x^2,y)+xP_2(x^2,y)$$ Similarly, $$Q(x)=Q_1(x^2,y)+xQ_2(x^2,y)$$ Then $$P(x,y)=\frac{P_1(x^2,y)+xP_2(x^2,y)}{Q_1(x^2,y)+xQ_2(x^2,y)}=\frac{[P_1(x^2,y)+xP_2(x^2,y)][Q_1(x^2,y)-xQ_2(x^2,y)]}{[Q_1(x^2,y)]^2-[xQ_2(x^2,y)]^2}$$ $$=\frac{P_3(x^2,y)+xP_4(x^2,y)}{Q_3(x^2,y)}$$ where $$P_3(x^2,y)=P_1(x^2,y)Q_1(x^2,y)-x^2P_2(x,y^2)Q_2(x^2,y)$$ $$P_4(x^2,y)=-P_1(x^2,y)Q_2(x^2,y)+P_2(x^2,y)Q_1(x^2,y)$$ $$Q_3(x^2,y)=[Q_1(x^2,y)]^2+[xQ_2(x^2,y)]^2$$

For part (a), since $R(−u,v)= R(u,v)$, then $$R(-u,v)=\frac{P_3(u^2,v)-uP_4(u^2,v)}{Q_3(u^2,v)}$$ How could I eliminate $uP_4(u^2,v)$, or in the other way, this is alike an even function for single variable function, to show $R(u,v)$ in this form has no odd degree for $u$? If part (a) is done, then part (b) can be brought down using similar way as part (a).

Update: Since $R(−u,v)= R(u,v)$, then $$R(-u,v)=\frac{P_3(u^2,v)-uP_4(u^2,v)}{Q_3(u^2,v)}=\frac{P_3(u^2,v)+uP_4(u^2,v)}{Q_3(u^2,v)}=R(u,v)$$ implies $uP_4(u^2,v)=0$ for all $u$ and $v$. (Not necessary $u=0$). Then it is proven and follows part (a)? I am confused but this idea suddenly come into my mind.


This topic was considered in a Russian book “Differential and Integral Calculus” by Grigorii Fichtenholz (v. II, 7-th edition, M.: Nauka, 1970). I copied the relevant pages: 74, 75, and 76. According to Wikipedia “book was translated, among others, into German, Chinese, and Persian however translation to English language has not been done still”. Below I translated the key moments relevant to your question.

(a) The conclusion $uP_4(u^2,v)=0$ looks OK. Indeed, even taking into account possible zeroes of the denominator $Q_3(u^2,v)$, we have a that a polynomial $S(u,v)= uP_4(u^2,v)Q_3(u^2,v)$ is zero for each $u,v$. Then $S(u,v)$ is the zero polynomial. Indeed, assume to the contrary that $S(u,v)=s_0(u)+s_1(u)v+\dots+s_m(u)v^m$ for some polynomials $s_0,\dots, s_m$ such that $s_m(u)$ is not the zero polynomial. Since $s_m(u)$ has only finitely many roots, we can pick $u_0$ such that $s_m(u_0)\ne 0$. Then $S(u_0,v)$ is not the zero polynomial on $v$, so there exists $v_0$ such that $S(u_0,v_0)\ne 0$, a contradiction.

(b) Fichtenholz says that the first part follows from (a) applid to the function $\frac {R(u,v)}u$. The second part is I at at p. 75 which states

$$R(\sin x,\cos x)dx=R_0(\sin^2 x,\cos x)\sin x dx=-R_0(1-\cos^2 x,\cos x) d\cos x.$$

(c) I translated III from pp. 75-76:

Substtuting $u$ by $\frac uvv$, we have

$$R(u,v)=R\left(\frac uvv,v\right)=R^*\left(\frac uv,v\right).$$

By the property of the function $R$ with the respect to change of signs of $u$ and $v$ (which does not change the fraction $\frac uv$),

$$R^*\left(\frac uv,-v\right)= R^*\left(\frac uv,v\right),$$

then, as we know,

$$R^*\left(\frac uv,v\right)=R_1^*\left(\frac uv,v^2\right).$$


$$R(\sin x,\cos x)=R^*_1(\tan x,\cos^2 x)= R^*_1\left(\tan x,\frac 1{1+\tan^2 x}\right),$$

that is

$$R(\sin x,\cos x)=\tilde R(\tan x).$$

Here a substitution $t=\tan x$ $\left(-\frac\pi{2}<x<\frac\pi{2} \right)$ reaches the goal, because $$R(\sin x,\cos x)dx=\tilde R(t)\frac{dt}{1+t^2}$$


For part a), note that if $$R(u,v)=R(-u,v)$$then $$P(-u,v)Q(u,v)=P(u,v)Q(-u,v)$$Define $S(u,v)=P(-u,v)Q(u,v)$. Then $S(u,v)$ is a polynomial function of $u,v$ since it is the product of two other polynomials and therefore can be expressed as $$S(u,v)=\sum_{i=0}^{m}\sum_{j=0}^{n}a_{ij}u^iv^j$$from $S(u,v)=S(-u,v)$ we obtain$$\sum_{i=0}^{m}\sum_{j=0}^{n}a_{ij}u^iv^j=\sum_{i=0}^{m}\sum_{j=0}^{n}a_{ij}(-u)^iv^j$$which yields to $$\sum_{i=0\\i\text{ is even}}^{m}\sum_{j=0}^{n}a_{ij}u^iv^j+\sum_{i=0\\i\text{ is odd}}^{m}\sum_{j=0}^{n}a_{ij}u^iv^j=\sum_{i=0\\i\text{ is even}}^{m}\sum_{j=0}^{n}a_{ij}(-u)^iv^j+\sum_{i=0\\i\text{ is odd}}^{m}\sum_{j=0}^{n}a_{ij}(-u)^iv^j$$from which we obtain $$\sum_{i=0\\i\text{ is odd}}^{m}\sum_{j=0}^{n}a_{ij}u^iv^j=0$$for any $u,v$. Therefore we can write $$S(u,v){=\sum_{i=0}^{m}\sum_{j=0}^{n}a_{ij}u^iv^j\\=\sum_{i=0\\i\text{ is even}}^{m}\sum_{j=0}^{n}a_{ij}u^iv^j\\=\sum_{i'=0}^{m'}\sum_{j=0}^{n}a_{i'j}u^{2i'}v^j\\=S_1(u^2,v)}$$This means that $S(u,v)$ is a polynomial of $u^2$ and $v$ so is $P(-u,v)Q(u,v)$. Then both $P(-u,v)$ and $Q(u,v)$ must be polynomials of $u^2$ and $v$ (otherwise at least one term as $u^{2k+1}v^l$ would be appeared in $S(u,v)$) and by dividing $P(u,v)$ on $Q(u,v)$ we conclude that $$R(u,v)=R_1(u^2,v)$$The other parts can be proved easily $\blacksquare$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.