11
$\begingroup$

Let $f:[0,1]\to\mathbb{C}$ be a given "nice" function and let $a,b:[0,1]\to\mathbb{R}$ be two unknown functions to be solved for (with some initial conditions $a(0),a'(0),b(0),b'(0))$.

How can one solve the following equation: $$ f(x) = \exp\left(-\mathrm{i}\int_0^x\cos(a(y))b'(y)\mathrm{d}y\right)\left(a'(x)+\mathrm{i}\sin(a(x))b'(x)\right)\qquad(x\in[0,1])\,? \tag{E}$$

I'm a bit stumped by the mis-match of the $\sin$ and $\cos$. For example, had the equation been $$ f(x) = \exp\left(-\mathrm{i}\int_0^x\cos(a(y))b'(y)\mathrm{d}y\right)\left(a'(x)-\mathrm{i}\cos(a(x))b'(x)\right)\qquad(x\in[0,1]) $$we could have re-written the RHS as $$ \left(a \exp\left(-\mathrm{i}\int_0^\cdot\cos(a)b'\right)\right)' $$ so that $$ \int_0^x f = a(x) \exp\left(-\mathrm{i}\int_0^x\cos(a(y))b'(y)\mathrm{d}y\right) $$ and we could have found $a,b$ by looking at the absolute value or respectively the phase of the LHS.

However, due to the mismatch between the cosine and sine in (E) I am not sure how to proceed, though I hope it does not mean that there isn't a closed form solution.

$\endgroup$
5
  • $\begingroup$ The $\displaystyle\exp\left(-\mathrm i\int_0^x\cos(a(y))b'(y)\,\mathrm dy\right)$ term seems quite conspicuous. Where is Equation E derived from? $\endgroup$ Commented May 18, 2020 at 12:53
  • $\begingroup$ Also, even for $f = 0$ (being “nice” enough) with the initial conditions $a(0)=a'(0)=0$ and arbitrarily given $b(0)$ and $b'(0)$, there're infinitely many solutions:$$\begin{cases}a(x) = 0\\b(x)=b(0)+b'(0)x+cx^2\end{cases},$$where $c\in\mathbb R$ is any constant. $\endgroup$ Commented May 18, 2020 at 12:56
  • $\begingroup$ @Saad, thanks for your comments. The equation comes from doing parallel transport on the sphere ($a,b$ being the spherical coordinates). I think the case you point out is quite degenerate, perhaps one should insist on initial conditions where the derivatives are not zero. $\endgroup$ Commented May 18, 2020 at 15:28
  • $\begingroup$ Did you try to look at the absolute value and the phase on the first equation? $\endgroup$ Commented May 21, 2020 at 15:04
  • $\begingroup$ @jvc, thanks, yes, I tried. The absolute value equation is easy, but I don't know what to do with the phase equation then. $\endgroup$ Commented May 21, 2020 at 23:11

1 Answer 1

5
+500
$\begingroup$

REVISED.

Assume the equation $$ f(x)=\exp\left(-i\int^{x}_{0}\cos(a(y))b'(y)dy\right)[a'(x)+i\sin(a(x))b'(x)],\tag 1 $$ where $f:\textbf{R}\rightarrow\textbf{C}$ and $a(x),b(x)$ are $\textbf{R}\rightarrow\textbf{R}$.

Seting $a(y)=t$ in (1) we get $$ f(x)=\exp\left(-i\int^{a(x)}_{a(0)}\cos(t) b'\left(a^{(-1)}(t)\right)a^{(-1)}{'}(t)dt\right)[a'(x)+i\sin(a(x))b'(x)]\Leftrightarrow $$ $$ f\left(a^{(-1)}(x)\right)=\exp\left(-i\int^{x}_{a(0)}\cos(t)\left(b\left(a^{(-1)}(t)\right)\right)'dt\right)\times $$ $$ \times [\frac{1}{a^{(-1)}{'}(x)}+i\sin(x)b'\left(a^{(-1)}(x)\right)]\Leftrightarrow $$ $$ a^{(-1)}{'}(x)f\left(a^{(-1)}(x)\right)=\exp\left(-i\int^{x}_{c}\cos(t)g'(t)dt\right)[1+i\sin(x)g'(x)],\tag 2 $$ where $c_0=a(0)$ and $g(x)=b\left(a^{(-1)}(x)\right)$. Hence (2) becomes $$ \frac{d}{dx}\int^{a^{(-1)}(x)}_{c_1}f(t)dt=e^{-iY}-e^{-iY}(-i\cos(x)g'(x))\tan(x),\tag 3 $$ where $$ Y=Y(x):=\int^{x}_{c_0}\cos(t)g'(t)dt. $$ Hence if we also set $$ H(x):=\int^{a^{(-1)}(x)}_{c_1}f(t)dt\tag 4 $$ and $f(t)=f_1(t)+if_2(t)$, with $f_1,f_2$ real, then $$ H_{j}(x)=\int^{a^{(-1)}(x)}_{c_0}f_j(t)dt\in\textbf{R}\textrm{, } j=1,2.\tag{4.1} $$ Hence $$ H(x)=H_1(x)+iH_2(x). $$ But then $$ H'(x)=e^{-iY}-\frac{d\left(e^{-iY}\right)}{dx}\tan(x)\tag 5 $$ Setting lastly $e^{-iY}=w(x)$, we get $$ H'(x)=w(x)-w'(x)\tan(x)\Leftrightarrow $$ $$ w(x)=-\sin(x)\int^{x}_{c_2}\frac{\cos(x)}{\sin^2(x)}H'(x)dx\Leftrightarrow $$ $$ \exp\left(-i\int^{x}_{c_0}\cos(t)g'(t)dt\right)=-\sin(x)\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H'(t)dt.\tag 6 $$ Hence $$ \cos\left(\int^{x}_{c_0}\cos(t)g'(t)dt\right)=-\sin(x)\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_1'(t)dt\tag 7 $$ and $$ \sin\left(\int^{x}_{c_0}\cos(t)g'(t)dt\right)=\sin(x)\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_2'(t)dt\tag 8 $$

Solving (8) with respect to $g'(x)$, we get $$ \int^{x}_{c_0}\cos(t)g'(t)dt=\sin^{(-1)}\left(\sin(x)\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_2'(t)dt\right)\Rightarrow $$ $$ g(x)=b\left(a^{(-1)}(x)\right)= $$ $$ =\int^{x}_{c_3}\frac{1}{\cos(u)}\frac{d}{du}\left[\sin^{(-1)}\left(\sin(u)\int^{u}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_2'(t)dt\right)\right]du. $$ Hence $$ b(x)=\int^{a(x)}_{c_3}\frac{\csc^2(u)H_2'(u)}{\sqrt{1-\left(\int^{u}_{c_2}\cot(t)\csc(t)H_2'(t)dt\right)^2}}du,\tag 9 $$ Now from (7) and (8), we have $$ \left(\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_1'(t)dt\right)^2+\left(\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_2'(t)dt\right)^2=\frac{1}{\sin^2(x)} $$ Hence we can assume that exists real function $\phi(x)$ with inverse $\theta(x)$ such that $$ \sin(x)\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_1'(t)dt=\cos(\phi(x))\tag{10} $$ and $$ \sin(x)\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_2'(t)dt=\sin(\phi(x))\tag{11} $$ Hence $$ \int^{\theta(x)}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_1'(t)dt=\frac{\cos(x)}{\sin(\theta(x))} $$ and $$ \int^{\theta(x)}_{c_2}\frac{\cos(t)}{\sin^2(t)}H_2'(t)dt=\frac{\sin(x)}{\sin(\theta(x))} $$ Differentiating these formulas we get $$ \sin(x)+\cot(\theta(x))[\cos(x)+H_1'(\theta(x))\theta'(x)]=0\tag{12} $$ and $$ \cos(x)-\cot(\theta(x))[\sin(x)+H_2'(\theta(x))\theta'(x)]=0\tag{13} $$ Solving (12) with respect to $\theta'(x)$, we get $$ \theta'(x)=-\frac{\sin(x)\tan(\theta(x))}{\cos(x)+H_1'(\theta(x))}.\tag{14} $$ Setting this value of $\theta'(x)$ in (13) we get $$ H_2'\left(\theta(x)\right)=-\csc(x)-\cot(x)H_1'(\theta(x)).\tag{15} $$ Now using (4),(4.1) we get $$ a^{(-1)}{'}\left(\theta(x)\right)=\frac{-1}{\cos(x)f_1\left(a^{(-1)}\left(\theta(x)\right)\right)+\sin(x)f_2\left(a^{(-1)}\left(\theta(x)\right)\right)}\tag{15.1} $$ and $$ \theta'(x)=\tan(\theta(x))\frac{\cos(x)f_1\left(a^{(-1)}\left(\theta(x)\right)\right)+\sin(x)f_2\left(a^{(-1)}\left(\theta(x)\right)\right)}{\sin(x)f_1\left(a^{(-1)}\left(\theta(x)\right)\right)-\cos(x)f_2\left(a^{(-1)}\left(\theta(x)\right)\right)}.\tag{15.2} $$ Notes.

  1. $f'(g(x))$ is not $(f(g(x)))'=f'(g(x))g'(x)$.

  2. In general if $h(x)$ is the inverse of $f(x)$, then holds $f(h(x))=h(f(x))=x$ and $f'(h(x))=\frac{1}{h'(x)}$ and $h'(f(x))=\frac{1}{f'(x)}$.

Hence equivalently if $A(x)$ denotes the inverse of $a(x)$ and $\phi(x)$ the inverse of $\theta(x)$: $$ A'(x)=\frac{-1}{\cos(\phi(x))f_1\left(A\left(x\right)\right)+\sin(\phi(x))f_2\left(A\left(x\right)\right)}\tag{16} $$ $$ \phi'(x)=\cot(x)\frac{\sin(\phi(x))f_1\left(A\left(x\right)\right)-\cos(\phi(x))f_2\left(A\left(x\right)\right)}{\cos(\phi(x))f_1\left(A\left(x\right)\right)+\sin(\phi(x))f_2\left(A\left(x\right)\right)}\tag{17} $$

For given $f=f_1+if_2$ the equations (9),(16),(17) are determine the problem in the case of $a(x),b(x)$ real functions. The DE are not linear and quite involved. Hence very hard to solve.

REMARK.

If we assume that $b(x)$ not necesary real, then (with notation independent of that above):

Assume the equation $$ f(x)=\exp\left(-i\int^{x}_{0}\cos(a(y))b'(y)dy\right)[a'(x)+i\sin(a(x))b'(x)].\tag{R1} $$ Seting $a(y)=t$, we get $$ f(x)=\exp\left(-i\int^{a(x)}_{a(0)}\cos(t) b'\left(a^{(-1)}(t)\right)a^{(-1)}{'}(t)dt\right)[a'(x)+i\sin(a(x))b'(x)]\Leftrightarrow $$ $$ f\left(a^{(-1)}(x)\right)=\exp\left(-i\int^{x}_{a(0)}\cos(t)\left(b\left(a^{(-1)}(t)\right)\right)'dt\right)\times $$ $$ \times[\frac{1}{a^{(-1)}{'}(x)}+i\sin(x)b'\left(a^{(-1)}(x)\right)]\Leftrightarrow $$ $$ a^{(-1)}{'}(x)f\left(a^{(-1)}(x)\right)=\exp\left(-i\int^{x}_{c}\cos(t)g'(t)dt\right)[1+i\sin(x)g'(x)],\tag{R2} $$ where $c_0=a(0)$ and $g(x)=b\left(a^{(-1)}(x)\right)$. Hence (R2) becomes $$ \frac{d}{dx}\int^{a^{(-1)}(x)}_{c_1}f(t)dt=e^{-iY}-e^{-iY}(-i\cos(x)g'(x))\tan(x),\tag{R3} $$ where $$ Y=Y(x):=\int^{x}_{c_0}\cos(t)g'(t)dt. $$ Hence if we also set $$ H(x):=\int^{a^{(-1)}(x)}_{c_1}f(t)dt,\tag{R4} $$ then from (R3): $$ H'(x)=e^{-iY}-\frac{d\left(e^{-iY}\right)}{dx}\tan(x) $$ Setting lastly $e^{-iY}=w(x)$, we get $$ H'(x)=w(x)-w'(x)\tan(x)\Leftrightarrow $$ $$ w(x)=\left(C_1-\int \frac{\cos(x)}{\sin^2(x)}H'(x)dx\right)\sin(x)\Leftrightarrow $$ $$ \exp\left(-i\int^{x}_{c_0}\cos(t)g'(t)dt\right)=\sin(x)\left(C_1-\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H'(t)dt\right),\tag{R5} $$ where $H(x)=\int^{a^{(-1)}(x)}_{c_0}f(t)dt$. Solving (R4) with respect to $g'(x)$, we get $$ -i\int^{x}_{c_0}\cos(t)g'(t)dt=\log\left(C_1\sin(x)-\sin(x)\int^{x}_{c_2}\frac{\cos(t)}{\sin^2(t)}H'(t)dt\right)\Rightarrow $$ $$ g(x)=b\left(a^{(-1)}(x)\right)= $$ $$ =i\int^{x}_{c}\frac{1}{\cos(u)}\frac{d}{du}\left[\log\left(\sin(u)\left(C_1-\int^{u}_{c_2}\frac{\cos(t)}{\sin^2(t)}H'(t)dt\right)\right)\right]du. $$ Hence $$ b\left(a^{(-1)}(x)\right)=i\log\left(\tan\left(\frac{x}{2}\right)\right)-i\int^{x}_{c_3}\frac{\csc^2(u)H'(u)}{C_1-\int^{u}_{c_2}\cot(t)\csc(t)H'(t)dt}du.\tag{R6} $$ Hence if $a(x)$ is any (real invertible and smooth) function, then $b(x)$ is given from (R6).

$\endgroup$
6
  • $\begingroup$ Thanks! So just to clarify, this only finds solutions where $a$ is invertible and sufficiently regular, but is otherwise unconstrained? $\endgroup$ Commented May 22, 2020 at 22:47
  • $\begingroup$ Another question: the solution should be real which is not manifestly obvious from (6). Do you have a clue about that? $\endgroup$ Commented May 22, 2020 at 23:25
  • $\begingroup$ So after you revision, $a,b$ are manifestly real and $a$ is not arbitrary. $\endgroup$ Commented May 25, 2020 at 17:00
  • 1
    $\begingroup$ @PPR. Yes if $a,b$ are real, then they depend on $Re(f)$ and $Im(f)$ and thus are unique. Equations (15),(16) show this better. However if $a,b$ are arbitray (complex numbers etc), then $a(x)$ becomes a parameter and $b(x)$ is evaluated from $a(x)$. $\endgroup$ Commented May 25, 2020 at 17:21
  • 1
    $\begingroup$ Im asking sorry. I mean equations (16),(17). $\endgroup$ Commented May 25, 2020 at 17:28

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.