1: \section{Manifolds with singularities}\label{sec6}
2: All results on the heat kernel expansion formulated in the
3: previous sections are valid on
4: smooth manifolds only. If there are boundaries, they also
5: have to be smooth. As well, any singularities in the potential
6: term or in the field strength are strictly speaking forbidden.
7: However, many physical models deal with singular backgrounds.
8: Even if such backgrounds may be represented through certain limiting
9: procedures from smooth configurations, the heat kernel coefficients
10: are not given by limits of their ``smooth'' values. The most visible
11: manifestation of failure of the smooth field approximation is that $a_k$
12: with sufficiently large $k$ are divergent. Usually, the presence of
13: singularities changes even the structure of the heat kernel
14: kernel expansion as compared to the smooth case.
15:
16: %%%%%%%%%%
17: \subsection{Non-integrable potentials}\label{s6sinpot}
18: According to (\ref{divWLam}) divergences in the effective action
19: are defined by {\it integrated} heat kernel coefficients.
20: Although the formulae (\ref{a0nobou}) - (\ref{a6nobou}) for
21: the {\it localised} heat kernel coefficients are valid on
22: non-compact manifolds (provided the smearing function $f$
23: falls off sufficiently fast), transition to the integrated
24: heat kernel is not that straightforward. Already the coefficient
25: $a_0(D)=a_0(1,D)$, which is proportional to the volume,
26: is divergent. This divergence is usually removed by replacing
27: $\det (D)$ in (\ref{pathdet}) by
28: \begin{equation}
29: \det (D)/\det (D_0) \,,\label{referdet}
30: \end{equation}
31: where the operator $D_0=-\partial^2+m^2$ describes a free particle
32: propagation in an ``empty'' space. It is argued that since $D_0$ does
33: not depend on ``essential'' variables division by $\det (D_0)$ does not
34: change physical predictions of the theory. In all subsequent formulae
35: the heat kernel $K(t;x,y;D)$ is then replaced by the subtracted
36: heat kernel
37: \begin{equation}
38: K_{\rm sub} (t;x,y)=K(t;x,y;D)-K(t;x,y;D_0) \,.
39: \label{Ksub}
40: \end{equation}
41: In flat space the coefficient $a_0$ corresponding to $K_{\rm sub}$
42: is identically zero\footnote{On a curved manifold the subtraction
43: procedure is more subtle. On has to define a reference metric which
44: differs from the physical one on a compact submanifold.}. If the field
45: strength $\Omega_{\mu\nu}$ and the (subtracted) potential $E+m^2$
46: have a compact support or decay sufficiently fast at the infinity,
47: the small $t$ asymptotic expansion of $K_{\rm sub}(t;x,x)$ is
48: integrable on the whole $M$. If not, the very structure of the
49: global heat kernel may be changed.
50:
51: As an example of non-integrable potentials consider the harmonic
52: oscillator in one dimension. The Schr\"{o}dinger operator reads
53: \begin{equation}
54: D=-\partial_x^2 +\nu^2 x^2 \,.\label{harmos}
55: \end{equation}
56: If we consider the problem on the whole real axis, $M=\mathbb{R}$,
57: the potential term is not integrable. Already the expression
58: (\ref{a2nobou}) for $a_2(1,D)$ diverges. Therefore, analytic expressions
59: of sec.\ \ref{sec4} cannot be used in this case. However, the
60: (integrated) heat kernel can be easily calculated. Eigenvalues of the
61: operator (\ref{harmos}) are contained in almost any textbook on quantum
62: mechanics:
63: \begin{equation}
64: \lambda_j=\nu (2j+1),\qquad j=0,1,2,\dots \label{harmspec}
65: \end{equation}
66:
67: The integrated heat kernel reads
68: \begin{equation}
69: K(t;D)=\sum_{j=0}^\infty e^{-t\nu (2j+1)} =
70: \frac 12 [\sinh (\nu t)]^{-1} \,.\label{harmheat}
71: \end{equation}
72: As $t\to 0$ it behaves like $1/t$ while for smooth rapidly decaying
73: potentials in one dimension the leading singularity in the heat kernel
74: is $1/\sqrt{t}$. This statement may be generalised to higher dimensions.
75: If $D=-\partial^2 +P_{\mu\nu}x^\mu x^\nu$ with a non-degenerate
76: matrix $P_{\mu\nu}$ on $M=\mathbb{R}^n$, the leading
77: term in $K(t;D)$ is $(2t)^{-n} (\det P)^{-1/2}$ \cite{Avramidi:1995ik}.
78:
79: \subsection{Conical singularities}\label{s6consin}
80: Conical space is defined as $M=[0,1]\times N$ where $N$ is an
81: $n-1$-dimensional manifold called the base. The metric of the
82: cone has the form
83: \begin{equation}
84: (ds)^2=dr^2 +r^2 d\Omega^2 \,,\label{conmetr}
85: \end{equation}
86: where $r\in [0,1]$ and $d\Omega^2$ is the line element on the
87: base $N$. This metric is, in general, singular at $r=0$. However, if we
88: take the unit sphere $S^{n-1}=N$ with standard round metric,
89: the singularity disappears and we obtain the $n$-dimensional unit ball
90: (\ref{ballmetric}). If a manifold has singular points where
91: the metric can be approximated by (\ref{conmetr}) we say that this
92: manifold has conical singularities.
93:
94: Conical singularities appear in many physical applications.
95: First of all, with $N=S^1$ the metric (\ref{conmetr}) is the Euclidean
96: version of the Rindler metric. Conical singularities appear in
97: classical solutions of the Einstein equations
98: \cite{Sokolov:1977,Frolov:1987dz,Barriola:1989hx}
99: and in the supermembrane theory
100: \cite{Vassilevich:1991vy}. Gravitational field of a point mass in
101: three dimensional gravity is a conical space
102: \cite{Deser:1984tn,Deser:1988qn}. There are evidences \cite{Schleich:1993bs}
103: that ``conifolds'' dominate the path integral for quantum gravity in
104: topological sectors.
105:
106: Sommerfeld \cite{Sommerfeld:1894}
107: was probably the first to consider the heat kernel in the presence
108: of conical singularities. The mathematical theory of the heat kernel
109: asymptotics with conical singularities was developed almost 100 years
110: later \cite{Cheeger:1983,Bruening:1984,Bruening:1987}. There two peculiar
111: features of these asymptotics. First, the heat kernel expansion
112: contains in general both integer and half-integer powers of $t$ even
113: without boundaries. Second, a non-standard $\ln t$ term may be contained in
114: the asymptotic series\footnote{The present author is not aware of
115: any simple example where the $\ln t$ terms actually appear.}.
116:
117: On a manifold with conical singularities no closed analytical expression
118: for the heat kernel coefficients is available. However, usually it is
119: possible to disentangle contributions of the singularities from the
120: smooth part. For example, if $N=S^1$, $d\Omega^2=d\varphi^2$ with
121: $\varphi \in [0,\alpha ]$ only $a_2$ receives a contribution from the
122: tip of the cone:
123: \begin{equation}
124: a_2({\rm tip})=\frac{4\pi^2 -\alpha^2}{24\pi \alpha} \,.\label{a2tip}
125: \end{equation}
126:
127: In many particular cases of conical singularities
128: a very detailed analysis of the heat kernel
129: expansion has been performed
130: \cite{Dowker:1977zj,Dowker:1989pp,Chang:1993fu,Dowker:1994jv,
131: Fursaev:1994qk,Fursaev:1994in,Cognola:1994qg,Cognola:1994ps,
132: Fursaev:1995ef,Bordag:1996fw,DeNardo:1997kp,Shtykov:1995fe,
133: Dowker:1998sc,BezerradeMello:2000mb}. One-loop computations on
134: general orbifolds were considered recently in
135: \cite{GrootNibbelink:2003gd}.
136: %%%%%%%
137: \subsection{Domain walls and brane world}\label{s6dom}
138: Delta function is an example of an extremely sharp background potential.
139: Let us consider a manifold $M$ and a submanifold $\Sigma$ of the
140: dimension $n-1$. Let
141: \begin{equation}
142: D[\dvxi ]=D+\dvxi \delta_\Sigma \,.\label{singop}
143: \end{equation}
144: $D$ is an operator of Laplace type (\ref{laplaceb}).
145: Let $h$ be
146: the determinant of the induced metric on $\Sigma$. Then $\delta_\Sigma$ is a
147: delta function defined such that
148: \begin{equation}
149: \int_M dx\sqrt{g} \delta_\Sigma f(x) =\int_\Sigma dx \sqrt{h}
150: f(x) \,.\label{deltaSig}
151: \end{equation}
152:
153: The spectral problem
154: for $D[\dvxi ]$ on $M$ as it stands is ill-defined
155: owing to the discontinuities (or
156: singularities) on $\Sigma$. It should be replaced by a pair of spectral
157: problems on the two sides
158: $M^\pm$ of
159: $\Sigma$ together with suitable matching conditions on $\Sigma$. In order to
160: find such matching conditions, we
161: consider an eigenfunction $\phi_\lambda$ of the operator
162: (\ref{singop}):
163: \begin{equation}
164: D[v]\phi_\lambda =\lambda \phi_\lambda \,.
165: \label{eigen}
166: \end{equation}
167: Let us choose the coordinate system such that $e_n$ is a unit normal
168: to $\Sigma$ and $x^n=0$ on $\Sigma$.
169: It is clear that $\phi_\lambda$ must be continuous on $\Sigma$:
170: \begin{equation}
171: \phi\vert_{x^n=+0}=\phi\vert_{x^n=-0} \,.\label{mc1}
172: \end{equation}
173: Otherwise, the second normal derivative of $\phi_\lambda$ would create
174: a $\delta'$ singularity on $\Sigma$ which is absent on the right
175: hand side of (\ref{eigen}). Let us integrate (\ref{eigen}) over
176: a small cylinder $\mathcal{C}=C^{n-1}\times [-\epsilon ,+\epsilon]$
177: \begin{equation}
178: \int_{\mathcal{C}} d^nx\sqrt{g} \left( -\nabla_n^2 \phi_\lambda
179: -\left[ \nabla_a^2 \phi_\lambda +(E+\lambda)\phi_\lambda \right]\right)
180: +\int_C d^{n-1}x\sqrt h \dvxi\phi_\lambda =0\,. \label{intcyl}
181: \end{equation}
182: We now take the limit as $\epsilon\to 0$. Since the expression in the
183: square brackets in (\ref{intcyl}) is bounded, the contribution
184: that this term makes vanishes in the limit. We obtain
185: \begin{equation}
186: 0=\int_C d^{n-1}x\sqrt h \left( -\nabla_n \phi_\lambda\vert_{x^n=+0}
187: +\nabla_n \phi_\lambda\vert_{x^n=-0} +\dvxi\phi_\lambda \right)\,.
188: \label{int2cyl}
189: \end{equation}
190: Since $C$ and $\lambda$ are arbitrary, we conclude that a proper
191: matching condition for the normal derivatives is
192: \begin{equation}
193: -\nabla_n \phi \vert_{x^n=+0}
194: +\nabla_n \phi \vert_{x^n=-0} +\dvxi\phi =0 \,.\label{dermatch}
195: \end{equation}
196:
197: Physically this problem corresponds to two domains separated by a
198: penetrable membrane $\Sigma$ (a domain wall). In many cases penetrable
199: membranes are better models of physical boundaries then just boundary
200: conditions which are imposed on each side of $\Sigma$ independently
201: and thus exclude any interaction between the domains
202: \cite{Jaekel:1992bn,Actor:1995vc}. $\delta$-potentials
203: are being used in quantum mechanics \cite{Albeverio:1988}
204: where one studies the Schr\"{o}dinger equation (which is
205: nothing else than the imaginary time heat equation).
206: The Casimir energy calculations have been performed e.g. in
207: \cite{Scandurra:1998xa}.
208: In the formal limit
209: $\dvxi\to\infty$ one obtains Dirichlet boundary conditions
210: on $\Sigma$, although the heat kernel coefficients are divergent in this
211: limit (see below).
212:
213: Further generalisations are suggested by the brane world scenario
214: \cite{Randall:1999ee,Randall:1999vf} which assumes that our world
215: is a four dimensional membrane in a five dimensional space\footnote{
216: A similar scenario was proposed earlier in \cite{Rubakov:1983bb},
217: see \cite{Rubakov:2001kp} for a review.}. According to the Israel
218: junction condition \cite{Israel:1966rt} the metric in such models
219: cannot be smooth on $\Sigma$. Typical form of the metric near
220: $\Sigma$ is
221: \begin{equation}
222: (ds)^2=(dx^n)^2 + e^{-\alpha |x^n|} (ds_{n-1})^2, \label{branemetr}
223: \end{equation}
224: where $\alpha$ is a constant and where $(ds_{n-1})$ is a line element
225: on the $(n-1)$-dimensional hypersurface $\Sigma$. Due to the presence of the
226: absolute value of the $n$-th coordinate in (\ref{branemetr}), the normal
227: derivative of the metric jumps on $\Sigma$.
228: One can think of two smooth manifolds
229: $M^+$ and $M^-$ glued together along their common boundary $\Sigma$.
230: Neither Riemann tensor, nor matrix potential $E$ must be continuous
231: on $\Sigma$. Also, the extrinsic curvatures $L_{ab}^+$ and $L_{ab}^-$
232: of $\Sigma$ considered as a submanifold in $M^+$ and in $M^-$
233: respectively are, in general,
234: different. All geometric quantities referring to $M^-$ (respectively,
235: $M^-$) and their limiting values on $\Sigma$ will be supplied by
236: a superscript ``$+$'' (respectively, ``$-$'').
237:
238: For the case at hand there is still an asymptotic expansion (\ref{asymptotex})
239: for the heat kernel. The heat kernel coefficients can be decomposed as
240: \begin{equation}
241: a_k(f,D[\dvxi ])=a_k^+(f,D) + a_k^-(f,D)^- +a_k^\Sigma (f,D,\dvxi )\,,
242: \label{apmSigma}
243: \end{equation}
244: where $a_k^\pm (f,D)$ are known volume contributions corresponding
245: to $M^\pm$ (cf. (\ref{a0nobou}) - (\ref{a6nobou})). The coefficients
246: $a_k^\Sigma$ are given by integrals over $\Sigma$ of some local
247: invariants. Note, that
248: \begin{equation}
249: \tilde\omega_a=\nabla_a^+-\nabla_a^- \label{tiloma}
250: \end{equation}
251: being a difference of two connection is a (pseudo-) vector with respect to
252: all space-time and gauge symmetries. Therefore, $\tilde\omega_a$ can be
253: used for constructing the surface invariants.
254:
255: To make the formulae more symmetric we introduce {\it two} inward pointing
256: unit normals $\nu^+$ and $\nu^-$ to $\Sigma$ in $M^+$ and $M^-$ respectively.
257: We do not suppose that the smearing function $f$ is smooth on $\Sigma$
258: (therefore, there is no relation between $f_{;\nu^+}$ and $f_{;\nu^-}$),
259: but we assume continuity of $f$: $f^+=f^-=f$ on $\Sigma$.
260:
261: The surface invariants can be constructed from $L_{ab}^\pm$,
262: $R_{ijkl}^\pm$, $E^\pm$, $\dvxi$, $\tilde\omega_a$ and their
263: derivatives. This gives much more invariants than we usually have
264: for a boundary value problem. There are, however, some properties
265: of $a_k^\Sigma$ which simplify the calculations considerably.
266: First of all, $a_k^\Sigma$ must be invariant with respect to
267: interchanging the roles of $M^+$ and $M^-$. Also, $a_k^\Sigma$
268: must vanish when the singularity disappears. The first property
269: excludes, for example, the term $f(E^+-E^-)$ which changes sign
270: under $M^+\leftrightarrow M^-$. The second requirement excludes
271: the invariant $f(E^++E^-)$ because it survives even if there is no
272: singularity on $\Sigma$. These simple arguments show that $a_3^\Sigma$
273: does not contain $E^\pm$ even though such terms are allowed on dimensional
274: grounds. It is also very helpful that in some particular cases the
275: problem in question can be reduces to a sum of Dirichlet and Robin
276: boundary value problems \cite{Bordag:1999ed,Gilkey:2001mj}.
277:
278: The coefficients $a_k^\Sigma$, $k=0,1,2,3$, read
279: \begin{eqnarray}
280: &&a_0^\Sigma(f,D,\dvxi)=0.\nonumber\\
281: &&a_1^\Sigma(f,D,\dvxi)=0.\nonumber\\
282: &&a_2^\Sigma(f,D,\dvxi)=
283: (4\pi)^{-n/2}\frac16\int_\Sigma d^{n-1}x\sqrt{h}
284: \ptr \{2f(L_{aa}^++L_{aa}^-)-6f\dvxi\}.\label{asig03}\\
285: &&a_3^\Sigma(f,D,\dvxi)=(4\pi)^{(1-n)/2}\frac1{384}
286: \int_\Sigma d^{n-1}x\sqrt{h}
287: \ptr\left\{\frac32f(L_{aa}^+L_{bb}^+
288: +L_{aa}^-L_{bb}^-\right. \nonumber\\
289: &&\qquad +2L_{aa}^+L_{bb}^-)
290: +3f(L_{ab}^+L_{ab}^++L_{ab}^-L_{ab}^-+2L_{ab}^+L_{ab}^-)
291: \nonumber\\
292: &&\qquad +9(L_{aa}^++L_{aa}^-)(f_{;\nu^+}^++f_{;\nu^-}^-)
293: +48f\dvxi^2+24f\tilde\omega_a\tilde\omega_a \nonumber\\
294: &&\qquad \left. -24f(L_{aa}^++L_{aa}^-)\dvxi
295: -24(f_{;\nu^+}^++f_{;\nu^-}^-)\dvxi\right\}.\nonumber
296: \end{eqnarray}
297: The coefficients $a_4^\Sigma$ and $a_5^\Sigma$ are too long to be
298: presented here in full generality. Therefore, we restrict ourselves to
299: the case of smooth geometry ($R_{ijkl}^+=R^-_{ijkl}$, $L_{ab}^+=-L_{ab}^-$),
300: smooth connection ($\tilde\omega_a=0$), and smooth smearing function
301: ($f_{;\nu^+}=-f_{;\nu^-}=f_{;n}$). In other words, the only singularity comes
302: from the surface potential $\dvxi$.
303: \begin{eqnarray}
304: &&a_4^\Sigma (f,D,\dvxi)=(4\pi )^{-n/2}
305: \int_\Sigma d^{n-1}x\sqrt{h}\ptr\left\{ -\frac 16 f\dvxi^3 -\frac 1{6}
306: fR \dvxi -fE\dvxi \right. \nonumber \\
307: &&\qquad \left. -\frac 16 f\dvxi_{:aa} +\frac 16 f_{;n}\dvxi L_{aa}
308: -\frac 16 f_{;nn}\dvxi
309: \right\} \nonumber \\
310: &&a_5^\Sigma (f,D)= (4\pi )^{-(n-1)/2} \int_\Sigma d^{n-1}x\sqrt{h}\ptr
311: \left\{\frac 1{64} f\dvxi^4 +\frac 1{48} fR \dvxi^2
312: \right. \label{asig45}\\
313: &&\qquad +\frac 1{192}R_{nn}\dvxi^2 +\frac 18 f\dvxi^2 E
314: -\frac 1{256} f\dvxi^2 L_{aa}L_{bb}
315: +\frac 1{128}f\dvxi^2 L_{ab}L_{ab} \nonumber \\
316: &&\qquad \left. +\frac 1{24}f\dvxi_{:aa}\dvxi
317: +\frac 5{192} f\dvxi_{:a}\dvxi_{:a} -\frac 5{384}f_{;n} \dvxi^2L_{aa}
318: +\frac 1{64}f_{;nn}\dvxi^2 \right\}
319: \nonumber
320: \end{eqnarray}
321: The heat kernel coefficients for $\Sigma =S^{n-1}\subset\mathbb{R}^n$
322: and $\dvxi =const.$ were calculated in \cite{Bordag:1998vs}.
323: Generic $\Sigma$ with arbitrary $\dvxi$ was considered in
324: \cite{Bordag:1999ed}. Moss \cite{Moss:2000gv} added a non-smooth
325: connection (see also \cite{Drozdov:2002um}).
326: Calculations on a particular brane-world background can be found
327: in \cite{Nojiri:2000bz}.
328: The heat kernel coefficients in the general setting
329: described here were calculated in \cite{Gilkey:2001mj}. This latter
330: paper also considered renormalization of the brane-world scenario
331: and predicted a non-standard Higgs potential on the brane.
332: Related calculations in
333: wormhole models were done in
334: \cite{Khusnutdinov:2002qb,Khusnutdinov:2003ii}. The $\zeta$
335: function for brane-world geometries with matching conditions
336: (\ref{mc1}), (\ref{dermatch}) was considered recently in
337: \cite{Elizalde:2002dd,Moss:2003zk}
338: (where one can find some further references).
339:
340: It is very well known \cite{Albeverio:1988}
341: that the conditions (\ref{mc1}),
342: (\ref{dermatch}) (which we call {\it transmittal}) are not the most
343: general matching conditions which can be defined on a surface.
344: In general, boundary values of a function and of its' normal
345: derivatives are related by a $2\times 2$ {\it transfer} matrix:
346: \begin{equation}
347: 0=
348: \left(\begin{array}{cc}
349: \nabla_{\nu^+}^++S^{++},\qquad&S^{+-}\\
350: S^{-+},\qquad&\nabla_{\nu^-}^-+S^{--}\end{array}\right)
351: \left(\begin{array}{l}\phi^+\\\phi^-\end{array}\right)
352: \bigg|_\Sigma\,.\label{BBT}
353: \end{equation}
354: Note, that the transfer conditions (\ref{BBT}) do not assume
355: identification of $\phi^+$ and $\phi^-$ on $\Sigma$. In other
356: words, there is no ad hoc relation between the restrictions
357: of the vector bundles $V^+|_\Sigma$ and $V^-|_\Sigma$. We can even
358: consider the situation when we have ${\rm dim}V^+\ne{\rm dim}V^-$,
359: i.e. the fields on $M^-$ and $M^-$ can have different structures
360: with respect to space-time and internal symmetries. $S^{\pm\pm}$
361: are some matrix valued functions on $\Sigma$ (one can even consider
362: the case when they are differential operators). It is clear from
363: the notations on which spaces they act. For example,
364: $S^{+-}: V^-|_\Sigma\to V^+|_\Sigma$.
365:
366: The matching conditions (\ref{BBT}) arise in heat transfer problems
367: \cite{Carslaw:1986}, some problems of quantum mechanics
368: \cite{Grosche:1994uv}, and in conformal field theory
369: \cite{Bachas:2001vj}. In a formal limiting case
370: $S^{++}-S^{-+}=S^{--}-S^{+-} \to \infty$ while
371: $\dvxi =2(S^{++}+S^{+-})$ is kept finite one arrives at transmittal
372: boundary conditions (\ref{mc1}), (\ref{dermatch}). The heat kernel
373: coefficients are divergent in this limit.
374:
375: In a particular case of spherical $\Sigma$ the heat kernel
376: expansion with transfer boundary conditions was evaluated in
377: \cite{Bordag:2001ta}. General expressions for $a_k$ with
378: $k=0,1,2,3,4$ were obtained in \cite{Gilkey:2002nv}.
379: Somewhat surprisingly, the calculations for (\ref{BBT})
380: are easier than for the singular particular case (\ref{mc1}),
381: (\ref{dermatch}).
382:
383: We should note that not all singular limiting cases of
384: (\ref{BBT}) are described by the transmittal conditions
385: (\ref{mc1}), (\ref{dermatch}). The heat kernel expansion
386: for a generalisation of transmittal condition is known
387: in the spherically symmetric case only \cite{Bordag:2001ta}.
388: Very little is known about the heat kernel if the transfer
389: matrix contains differential operators on $\Sigma$ (conformal
390: walls of ref.\ \cite{Bachas:2001vj} belong to this class
391: of problems).
392:
393: The case when the singular potential is located on a surface
394: of co-dimension larger than one (i.e. when
395: ${\rm dim}\Sigma <n-1$) is rather complicated. Even a careful
396: translation of this problem to the operator language
397: was done only in 1960's \cite{Berezin:1961}. Direct calculations
398: show \cite{Albeverio:1994,Solodukhin:1997xn} that
399: the heat kernel asymptotics may contain very unusual $(\ln t)^{-1}$
400: terms. More references can be found in \cite{Camblong:2001um}.
401: %%%%%%
402: \subsection{Non-smooth boundaries}\label{s6nsbou}
403: Rectangular region in a plane is probably the simplest
404: manifold with boundaries as far as eigenvalues of the Laplacian
405: are concerned. However, the formulae (\ref{a0bou}) -
406: (\ref{a4bou}) are not valid for this case because of the presence
407: of corners. The heat kernel expansion on manifolds
408: with piecewise smooth boundaries was considered by Kac in his
409: famous paper \cite{Kac:1966}. He demonstrated that for a
410: region in $\mathbb{R}^2$ each corner with the inside
411: facing angle $\alpha$ contributes
412: \begin{equation}
413: a_2({\rm corner})= \frac {\pi^2 -\alpha^2}{24\pi\alpha}
414: \label{a2Kac}
415: \end{equation}
416: to the coefficient $a_2$ while $a_0$ and $a_1$ are still defined
417: by their ``smooth'' expressions.
418:
419: The formula (\ref{a2Kac}) looks similar to the contribution of a
420: conical singularity (\ref{a2tip}). The reason for this similarity
421: is that the cone can be obtained from the wedge by gluing the sides
422: together and imposing the periodicity conditions.
423:
424: The study of boundary discontinuities was continued by Apps and
425: Dowker who calculated the coefficients $a_3$ \cite{Dowker:1995sp}
426: and $a_4$ \cite{Apps:1998zr} for piecewise smooth boundaries.
427: We also refer to \cite{Aurell:1994} where functional determinants
428: on simplicial complexes were analysed, and to
429: \cite{Dowker:2000fs} where divergences in the Casimir energy
430: found in \cite{Nesterenko:2000eb} were attributed to non-smoothness
431: of the boundaries. A recent study \cite{Nesterenko:2002ng}
432: should also be mentioned.
433:
434: When the angle $\alpha$ goes to $0$ we obtain a cusp. In this limit
435: (\ref{a2Kac}) is divergent. Presence of the cusp is an essential
436: singularity which modifies {\it powers} of the proper
437: time $t$ which appear in the asymptotic expansion of the heat
438: kernel \cite{Stewartson:1971}.
439: %%%%%%%%%
440: \subsection{Dielectric bodies}\label{s6diel}
441: Calculations of the Casimir energy of a dielectric body
442: have attracted much attention and created many controversial
443: results. A (rather large) literature on this subject is reviewed
444: in \cite{Bordag:2001qi,Milton:2001yy}. Quantum field theory formulation
445: of this problem is known for a long time already (cf.
446: \cite{Schwinger:1979pa,Milton:1980yx,Candelas:1982qw}). However, the
447: heat kernel analysis of divergences in the Casimir energy in
448: dielectric is a relatively new subject.
449:
450: Wave propagation with variable speed of light $c(x)$
451: is described by the operator $D=-c(x)^2 \nabla^2 +\dots$,
452: where we neglected the lower order terms. In a dielectric medium
453: $c$ is expressed in terms of the dielectric permittivity $\varepsilon$
454: and of the magnetic permeability $\mu$:
455: $c(x)^2=1/(\varepsilon (x)\mu (x))$.
456: For a smooth distribution of $\varepsilon$ and $\mu$
457: the operator $D$
458: is a particular case of curved space Laplacian with
459: an effective metric defined by $c(x)$. The heat kernel
460: coefficients can be calculated in the standard way \cite{Bordag:1998fh}.
461:
462: Consider now a dielectric body bounded by $\Sigma$. Typically,
463: $c(x)$ (and the effective metric!) jumps on $\Sigma$.
464: This singularity is much stronger than the one considered
465: in sec.\ \ref{s6dom}. Thus the geometric interpretation of this
466: problem is very difficult. Very little is known about the heat kernel
467: expansion in a dielectric body of an arbitrary shape.
468: There are calculations for a dielectric ball
469: \cite{Bordag:1998vs} and for a dielectric cylinder \cite{Bordag:2001zj}.
470: These calculations exhibit a puzzling property of the heat kernel
471: expansion in dielectrics in dilute approximation ($\varepsilon \sim 1$):
472: for a dielectric body the Casimir energy
473: in the ultra violet limit behaves
474: better than for the ``smooth'' case.
475: The heat kernel expansion for a frequency dependent $\varepsilon$
476: was considered in refs.\ \cite{Falomir:2001uv,Bordag:2001fq}.
477: