|
|
(8 intermediate revisions by 3 users not shown) |
Line 1: |
Line 1: |
| + | {{TEX|done}} |
| A function arising in an attempt to properly determine which singular boundary-value problems are self-adjoint (cf. also [[Self-adjoint differential equation|Self-adjoint differential equation]]). Begin with a formally symmetric differential expression | | A function arising in an attempt to properly determine which singular boundary-value problems are self-adjoint (cf. also [[Self-adjoint differential equation|Self-adjoint differential equation]]). Begin with a formally symmetric differential expression |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t1201202.png" /></td> </tr></table>
| + | $$ |
| + | L y = \frac{-(p y')' + q y}{w} , |
| + | $$ |
| | | |
− | where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t1201203.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t1201204.png" /> are measurable coefficients over <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t1201205.png" />, and which is defined on a domain within <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t1201206.png" />. The Titchmarsh–Weyl <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t1201207.png" />-function is defined as follows: For <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t1201208.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t1201209.png" />, let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012010.png" /> and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012011.png" /> be solutions of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012012.png" /> satisfying | + | where $p\ne 0$, $q,w>0$ are measurable coefficients over $[a,b)$, and which is defined on a domain within $L^2(a,b;w)$. The Titchmarsh–Weyl $m$-function is defined as follows: For $\lambda = \mu + i \nu$, $\nu\ne 0$, let $\phi$ and $\psi$ be solutions of $L y = \lambda y$ satisfying |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012013.png" /></td> </tr></table>
| + | $$ |
| + | \begin{aligned} |
| + | \phi(a,\lambda) &= \sin\alpha, & \psi(a,\lambda) &= \cos\alpha, \\ |
| + | p\phi'(a,\lambda) &= -\cos\alpha, & p\psi'(a,\lambda) &= \sin\alpha . |
| + | \end{aligned} |
| + | $$ |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012014.png" /></td> </tr></table> | + | Now consider a real boundary condition at $b'$, $a<b'<b$, of the form |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012015.png" /></td> </tr></table>
| + | $$\cos\beta\, x(b')+\sin\beta\, px'(b')=0,$$ |
| | | |
− | Now consider a real boundary condition at <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012016.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012017.png" />, of the form
| + | and let $\chi(x,\lambda)=\phi(x,\lambda)+\ell(\lambda)\psi(x,\lambda)$ satisfy it. Then |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012018.png" /></td> </tr></table>
| + | $$\ell(\lambda)=-\frac{\cot\beta\,\phi(b',\lambda)+p\phi'(b',\lambda)}{\cot\beta\,\psi(b',\lambda)+p\psi'(b',\lambda)}.$$ |
| | | |
− | and let <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012019.png" /> satisfy it. Then
| + | If $z=\cot\beta$, $\ell$ is a [[Meromorphic function|meromorphic function]] in the complex $z$-plane; indeed, it is a [[Fractional-linear mapping|fractional-linear transformation]] of the $z$-plane into itself. From the well-known properties of fractional-linear transformations, as $\beta$ varies over real values $0\leq\beta\leq\pi$, $z$ varies over the real $z$-axis, and $\ell$ describes a circle in the $z$-plane. |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012020.png" /></td> </tr></table>
| + | It can be shown that if $b'$ increases, the circles become nested. Hence there is at least one point inside all. For such a point $\ell=m(\lambda)$, |
| | | |
− | If <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012021.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012022.png" /> is a [[Meromorphic function|meromorphic function]] in the complex <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012023.png" />-plane; indeed, it is a bilinear transformation. As <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012024.png" /> varies over real values <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012025.png" />, <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012026.png" /> varies over the real <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012027.png" />-axis, and <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012028.png" /> describes a circle in the <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012029.png" />-plane.
| + | $$\int\limits_a^b|\chi(x,\lambda)|^2w(x)dx<\infty.$$ |
| | | |
− | It can be shown that if <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012030.png" /> increases, the circles become nested. Hence there is at least one point inside all. For such a point <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012031.png" />,
| + | There exists at least one solution of $Ly=\lambda y$, which is square-integrable. |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012032.png" /></td> </tr></table>
| + | If the limit of the circles is a point, then $m(\lambda)$ is unique and only $\chi(x,\lambda)$ is square-integrable. This is the limit-point case. If the limit of the circles is itself a circle, then $m(\lambda)$ is not unique and all solutions of $Ly=\lambda y$ are square-integrable. This is the limit-circle case. |
− | | |
− | There exists at least one solution of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012033.png" />, which is square-integrable.
| |
− | | |
− | If the limit of the circles is a point, then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012034.png" /> is unique and only <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012035.png" /> is square-integrable. This is the limit-point case. If the limit of the circles is itself a circle, then <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012036.png" /> is not unique and all solutions of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012037.png" /> are square-integrable. This is the limit-circle case. | |
| | | |
| Nonetheless, the differential operator | | Nonetheless, the differential operator |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012038.png" /></td> </tr></table>
| + | $$Ly=\frac{-(py')'+qy}{w}$$ |
| | | |
| whose domain satisfies | | whose domain satisfies |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012039.png" /></td> </tr></table>
| + | $$\sin\alpha\, y(a)-\cos\alpha\, py'(a)=0,$$ |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012040.png" /></td> </tr></table>
| + | $$\lim_{x\to b}[p(x)(y(x)\chi'(\lambda,x)-y'(x)\chi(x,\lambda)]=0,$$ |
| | | |
− | where <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012041.png" /> on the limit circle or limit point, is a self-adjoint differential operator (cf. also [[Self-adjoint operator|Self-adjoint operator]]; [[Self-adjoint differential equation|Self-adjoint differential equation]]) on <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012042.png" />. | + | where $\ell=m$ on the limit circle or limit point, is a self-adjoint differential operator (cf. also [[Self-adjoint operator|Self-adjoint operator]]; [[Self-adjoint differential equation|Self-adjoint differential equation]]) on $L^2(a,b;w)$. |
| | | |
− | If the circle limit is a point, the second boundary condition (at <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012043.png" />) is automatic. | + | If the circle limit is a point, the second boundary condition (at $b$) is automatic. |
| | | |
− | The [[Spectral measure|spectral measure]] of <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012044.png" /> is given by | + | The [[Spectral measure|spectral measure]] of $L$ is given by |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012045.png" /></td> </tr></table>
| + | $$\rho(\lambda)-\rho(\mu)=\frac1\pi\lim_{\epsilon\to0}\int\limits_\mu^\lambda\operatorname{Im}(m(\nu+i\epsilon))d\nu.$$ |
| | | |
− | The [[Spectral resolution|spectral resolution]] of arbitrary functions in <img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012046.png" /> is | + | The [[Spectral resolution|spectral resolution]] of arbitrary functions in $L^2(a,b;w)$ is |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012047.png" /></td> </tr></table>
| + | $$f(x) = \lim_{(\mu,\nu) \to (-\infty,\infty)} \int_\mu^\nu g(\lambda) \psi(x,\lambda) d\rho(\lambda) , $$ |
| | | |
| where the limit is in the mean-square sense, and | | where the limit is in the mean-square sense, and |
| | | |
− | <table class="eq" style="width:100%;"> <tr><td valign="top" style="width:94%;text-align:center;"><img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012048.png" /></td> </tr></table>
| + | $$g(\lambda)=\lim_{b'\to b}\int\limits_a^{b'}f(x)\psi(x,\lambda)dx.$$ |
| | | |
| ====References==== | | ====References==== |
− | <table><TR><TD valign="top">[a1]</TD> <TD valign="top"> E.A. Coddington, N. Levinson, "Theory of ordinary differential equations" , McGraw-Hill (1955)</TD></TR><TR><TD valign="top">[a2]</TD> <TD valign="top"> A.M. Krall, "<img align="absmiddle" border="0" src="https://www.encyclopediaofmath.org/legacyimages/t/t120/t120120/t12012049.png" /> theory for singular Hamiltonian systems with one singular point" ''SIAM J. Math. Anal.'' , '''20''' (1989) pp. 644–700</TD></TR></table> | + | <table><TR><TD valign="top">[a1]</TD> <TD valign="top"> E.A. Coddington, N. Levinson, "Theory of ordinary differential equations" , McGraw-Hill (1955)</TD></TR><TR><TD valign="top">[a2]</TD> <TD valign="top"> A.M. Krall, "$M(\lambda)$ theory for singular Hamiltonian systems with one singular point" ''SIAM J. Math. Anal.'' , '''20''' (1989) pp. 644–700</TD></TR></table> |
A function arising in an attempt to properly determine which singular boundary-value problems are self-adjoint (cf. also Self-adjoint differential equation). Begin with a formally symmetric differential expression
$$
L y = \frac{-(p y')' + q y}{w} ,
$$
where $p\ne 0$, $q,w>0$ are measurable coefficients over $[a,b)$, and which is defined on a domain within $L^2(a,b;w)$. The Titchmarsh–Weyl $m$-function is defined as follows: For $\lambda = \mu + i \nu$, $\nu\ne 0$, let $\phi$ and $\psi$ be solutions of $L y = \lambda y$ satisfying
$$
\begin{aligned}
\phi(a,\lambda) &= \sin\alpha, & \psi(a,\lambda) &= \cos\alpha, \\
p\phi'(a,\lambda) &= -\cos\alpha, & p\psi'(a,\lambda) &= \sin\alpha .
\end{aligned}
$$
Now consider a real boundary condition at $b'$, $a<b'<b$, of the form
$$\cos\beta\, x(b')+\sin\beta\, px'(b')=0,$$
and let $\chi(x,\lambda)=\phi(x,\lambda)+\ell(\lambda)\psi(x,\lambda)$ satisfy it. Then
$$\ell(\lambda)=-\frac{\cot\beta\,\phi(b',\lambda)+p\phi'(b',\lambda)}{\cot\beta\,\psi(b',\lambda)+p\psi'(b',\lambda)}.$$
If $z=\cot\beta$, $\ell$ is a meromorphic function in the complex $z$-plane; indeed, it is a fractional-linear transformation of the $z$-plane into itself. From the well-known properties of fractional-linear transformations, as $\beta$ varies over real values $0\leq\beta\leq\pi$, $z$ varies over the real $z$-axis, and $\ell$ describes a circle in the $z$-plane.
It can be shown that if $b'$ increases, the circles become nested. Hence there is at least one point inside all. For such a point $\ell=m(\lambda)$,
$$\int\limits_a^b|\chi(x,\lambda)|^2w(x)dx<\infty.$$
There exists at least one solution of $Ly=\lambda y$, which is square-integrable.
If the limit of the circles is a point, then $m(\lambda)$ is unique and only $\chi(x,\lambda)$ is square-integrable. This is the limit-point case. If the limit of the circles is itself a circle, then $m(\lambda)$ is not unique and all solutions of $Ly=\lambda y$ are square-integrable. This is the limit-circle case.
Nonetheless, the differential operator
$$Ly=\frac{-(py')'+qy}{w}$$
whose domain satisfies
$$\sin\alpha\, y(a)-\cos\alpha\, py'(a)=0,$$
$$\lim_{x\to b}[p(x)(y(x)\chi'(\lambda,x)-y'(x)\chi(x,\lambda)]=0,$$
where $\ell=m$ on the limit circle or limit point, is a self-adjoint differential operator (cf. also Self-adjoint operator; Self-adjoint differential equation) on $L^2(a,b;w)$.
If the circle limit is a point, the second boundary condition (at $b$) is automatic.
The spectral measure of $L$ is given by
$$\rho(\lambda)-\rho(\mu)=\frac1\pi\lim_{\epsilon\to0}\int\limits_\mu^\lambda\operatorname{Im}(m(\nu+i\epsilon))d\nu.$$
The spectral resolution of arbitrary functions in $L^2(a,b;w)$ is
$$f(x) = \lim_{(\mu,\nu) \to (-\infty,\infty)} \int_\mu^\nu g(\lambda) \psi(x,\lambda) d\rho(\lambda) , $$
where the limit is in the mean-square sense, and
$$g(\lambda)=\lim_{b'\to b}\int\limits_a^{b'}f(x)\psi(x,\lambda)dx.$$
References
[a1] | E.A. Coddington, N. Levinson, "Theory of ordinary differential equations" , McGraw-Hill (1955) |
[a2] | A.M. Krall, "$M(\lambda)$ theory for singular Hamiltonian systems with one singular point" SIAM J. Math. Anal. , 20 (1989) pp. 644–700 |