1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %% ws-procs975x65.tex : 10 October 2003
3: %% Text file to use with ws-procs975x65.cls written in Latex2E.
4: %% The content, structure, format and layout of this style file is the
5: %% property of World Scientific Publishing Co. Pte. Ltd.
6: %% Copyright 1995, 2002 by World Scientific Publishing Co.
7: %% All rights are reserved.
8: %%
9: %% Proceedings Trim Size: 9.75in x 6.5in
10: %% Text Area: 8in (include runningheads) x 5in
11: %% Main Text is 10/13pt
12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: %%
14:
15: %\documentclass[draft]{ws-procs975x65}
16:
17:
18: \documentclass{ws-procs975x65}
19: %\usepackage{rotating}
20: \begin{document}
21: \newcommand{\be}{\begin{equation}}
22: \newcommand{\ee}{\end{equation}}
23: \newcommand{\bea}{\begin{eqnarray}}
24: \newcommand{\eea}{\end{eqnarray}}
25:
26:
27: \title{Finite Casimir Energies in Renormalizable Quantum Field
28: Theory\footnote{\uppercase{I}nvited Talk given at \uppercase{M}arcel
29: \uppercase{G}rossmann
30: \uppercase{X}, \uppercase{R}io de \uppercase{J}aneiro, \uppercase{B}razil,
31: \uppercase{J}uly 20-26, 2003}}
32:
33: \author{Kimball A. Milton}
34:
35: \address{Department of Physics and Astronomy,\\ University of Oklahoma,\\
36: Norman, OK 73019-2061 USA\\
37: E-mail: milton@nhn.ou.edu}
38:
39:
40:
41: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
42: % You may repeat \author \address as often as necessary %
43: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
44:
45: \maketitle
46:
47: \abstracts{Quantum vacuum energy has been known to have observable consequences
48: since 1948 when Casimir calculated the force of attraction
49: between parallel uncharged
50: plates, a phenomenon confirmed experimentally with ever increasing precision.
51: Casimir himself suggested that a similar attractive self-stress existed for
52: a conducting spherical shell, but Boyer obtained a repulsive stress.
53: Other geometries and higher dimensions have been considered over the years.
54: Local effects, and divergences associated with surfaces and edges have been
55: investigated by several authors. Quite recently, Graham et
56: al.\ have re-examined
57: such calculations, using conventional techniques of perturbative quantum
58: field theory to remove divergences, and have suggested that previous
59: self-stress results may be suspect.
60: Here we show that most of the examples
61: considered in their work are misleading; in particular, it is well-known
62: that in two dimensions a circular boundary has a divergence in the Casimir
63: energy for massless fields, while for general dimension $D$ not equal to an
64: even integer the corresponding Casimir energy arising from massless
65: fields interior and exterior to a hyperspherical shell is finite.
66: It has also long been recognized that the Casimir energy for massive
67: fields is divergent for curved boundaries. These conclusions are reinforced by
68: a calculation of the relevant leading Feynman diagram in $D$ dimensions.
69: Divergences do occur in third order, as has been recognized for many years,
70: but this logarithmic divergence is of questionable relevance to real shells.
71: }
72:
73: \section{Introduction}
74: \label{Sec1}
75: The Casimir effect
76: remains one of the least intuitive consequences of
77: quantum field theory, and stands rather outside the usual development
78: of renormalization theory. This is because it is inherently nonperturbative,
79: in that macroscopic boundary conditions or backgrounds cannot be
80: easily mimicked by perturbative interactions. Its origins go back to
81: the very beginnings of quantum mechanics, because it can be thought of
82: as the change in the zero-point energy when the background is introduced.
83:
84: After examining the van der Waals interaction between two
85: molecules and between a molecule and a conducting plate,\cite{casimir48}
86: Casimir was challenged by Bohr to
87: interpret this interaction in terms of zero-point energy,\cite{casimir49}
88: and then to
89: recognize that the zero-point fluctuations of the electromagnetic field
90: implied a force between two such plates.\cite{casimir48a}
91: The attractive nature of this
92: force was obviously consistent with the action-at-a-distance interpretation
93: of it as due to the attraction between fluctuating dipoles making up the
94: material of the plates. But intuition flew out the window when Boyer
95: discovered that the energy, and hence the self-stress, on a perfectly
96: conducting spherical shell of zero thickness was positive or
97: repulsive.\cite{boyer68}
98: Later, it was found that a cylinder was intermediate, giving rise to a
99: small but attractive force.\cite{deraad81}
100:
101: Dimensional dependence was also dramatic. Sen examined a circular boundary
102: in two dimensions and found that the energy was infinite.\cite{sen81}
103: This was later found to be part of a pattern: For a hyperspherical shell
104: in $D$ spatial dimensions, the Casimir energy of a massless scalar field
105: subject to Dirichlet boundary conditions
106: was finite except when $D$ was a positive even integer, where the
107: energy or stress exhibits a simple pole, as seen in Fig.~\ref{fig1}.\cite{bender94}
108: (For $D\le0$, branch points
109: occur at the integers.) An intuitive explanation of this,
110: and of the corresponding sign changes, is still lacking.
111: \begin{figure}[ht]
112: \begin{center}
113: \begin{turn}{270}
114: \epsfxsize=8cm %width of figure - will enlarge/reduce the figures
115: \epsfbox{teplustmnew.ps}
116: \end{turn}
117: \end{center}
118: \caption{A plot of the TM and TE Casimir stress for $-2<D<4$ on a
119: perfectly conducting spherical
120: shell. For $D<2$ ($D<0$) the stress $\mathcal{S}^{\rm TM}$
121: ($\mathcal{S}^{\rm TE}$)
122: is complex and we have plotted the real part. The TM calculation was given
123: in Ref.~\protect\refcite{Milton:1997ri}.}
124: \label{fig1}
125: \end{figure}
126:
127: Deutsch and Candelas were the first to examine the local effects of
128: fluctuating fields,\cite{deutsch79} for other than the geometry of
129: parallel planes, which was considered by Brown and Maclay a decade
130: earlier.\cite{brown69}
131: Typically, surface divergences occur near boundaries,
132: although for flat boundaries with conformally-coupled fields, those divergences
133: disappear.
134: The reason the global Casimir energy of a (hyper)sphere is
135: finite is that there is a perfect cancellation between the interior
136: and exterior divergences. This perfect cancellation is spoiled if the
137: shell has finite thickness, or if the speed of light is different on the
138: two sides of the boundary.\cite{milton80}
139: Giving the fluctuating field a mass also
140: yields an unremovable divergence\cite{bordag97}
141: except for the case of plane boundaries.
142:
143:
144: Recently, Graham et al.\cite{graham01}
145: have questioned these findings. They have
146: developed an approach in which idealized boundary conditions are replaced
147: with interactions with an external (nondynamical) field. Potentially
148: divergent terms are subtracted and replaced by perturbatively calculable
149: Feynman diagrams. After renormalization of these diagrams, the limiting
150: case when the external field becomes a delta function is taken. In this way
151: the results for $D=1$ are reproduced; but the authors find those finite
152: results rather unsatisfactory, so they discuss how their limiting procedure
153: gives rise to a different energy, corresponding,
154: however, to the conventional force.
155:
156: Then they turn to $D=2$ and find that it is divergent; the implication is
157: that this is a general feature, so that all calculations of Casimir self-stress
158: are called into question.
159: However, as we remarked above, $D=2$ is a singular point. What is called
160: for is a calculation for general $D$. That is the purpose of this talk.
161: For simplicity, our attention will be restricted to scalar fields.
162: We will first, in Sec.~\ref{Sec2}, re-examine the $D=1$ calculation,
163: and show that the force is completely finite, while the energy density,
164: or more generally, the stress tensor, has a constant divergent part which
165: would be present if the boundaries were not present, and is therefore quite
166: without observable consequence. For parallel plates in $D$
167: dimensions, unphysical surface
168: divergences appear in the stress tensor (unphysical because they do not
169: contribute to the stress on the plates), which, for zero mass, vanish
170: if the conformal stress tensor is used.
171: Then, in Sec.~\ref{Sec3}, we re-examine
172: the self-stress on a sphere in three dimensions,
173: \normalcolor using time-splitting
174: to regulate the divergences. The result
175: is, once again, unambiguously finite.
176: The critical calculation is given in Sec.~\ref{Sec4}, where we
177: review and simplify the diagrammatic subtraction method, and explicitly
178: compute the graph in which two external fields are inserted, in $D$ spatial
179: dimensions.
180: As expected, the result is divergent at $D=2, 4, 6, \dots$,
181: but is otherwise finite for $D>3/2$. (A divergence, however, will
182: occur in the graph with three insertions.\cite{bkv,delta04})
183: Concluding remarks are offered in Sec.~\ref{Sec5}. This talk is
184: based largely on Ref.~\refcite{prd03}.
185:
186:
187: \section{Casimir Effect for Dirichlet Plates}
188: \label{Sec2}
189: \subsection{Massless Scalar in 1+1 Dimensions}
190:
191: We begin by reconsidering the Casimir effect for a massive scalar field
192: which vanishes on two parallel plates (Dirichlet boundary conditions).
193: Although these considerations are familiar, we will concentrate on the local
194: effect in 1+1 dimensions in order to make the divergence structure
195: manifest and make contact with the work of Graham et al.\cite{graham01}
196:
197: For a massless scalar field $\phi$, the stress tensor is
198: \be
199: T^{\mu\nu}=\partial^\mu\phi\partial^\nu\phi-\frac12g^{\mu\nu}\partial_\lambda
200: \phi\partial^\lambda\phi.\label{stresstensor}
201: \ee
202: It will be noted that for one spatial dimension, this canonical tensor
203: coincides with the conformal one,
204: \be
205: T^\mu{}_\mu=0.
206: \ee
207: The scalar field satisfies the free equation
208: \be
209: -\partial^2\phi=0,
210: \ee
211: but is subject to the Dirichlet boundary conditions on the plates
212: at $x=0$ and $x=a$:
213: \be
214: \phi(x=0)=\phi(x=a)=0.
215: \ee
216: The corresponding Green's function satisfies
217: \be
218: -\partial^2G(x,t;x',t')=\delta(x-x')\delta(t-t'),
219: \ee
220: and
221: \be
222: G(0,t;x',t')=G(a,t;x',t')=0.
223: \ee
224: Since the Green's function is translationally invariant in time, it
225: is natural to introduce a corresponding Fourier transform,
226: \be
227: G(x,x';t-t')=\int_{-\infty}^\infty\frac{d\omega}{2\pi}e^{-i\omega(t-t')}
228: g(x,x';\omega);
229: \ee
230: the reduced Green's function satisfies the ordinary differential equation
231: \be
232: -\left(\omega^2+\frac{d^2}{dx^2}\right)g(x,x';\omega)=\delta(x-x').
233: \ee
234: We only need the solutions of this equation in two regions:
235: \bea
236: g(x,x';\omega)&=& -\frac{\sin \omega x_<\sin\omega(x_>-a)}
237: {\omega\sin\omega a},\quad 0\le x,x'\le a,\label{ingf}\\
238: g(x,x';\omega)&=&
239: \frac1\omega\sin\omega(x_<-a)e^{i|\omega|(x_>-a)},\quad a\le x,x'.\label{outgf}
240: \eea
241: Here $x_>$ ($x_<$) is the greater (lesser) of $x$ and $x'$.
242: These are to be compared to the free Green's function, when no plates
243: are present:
244: \be
245: g_0(x,x';\omega)=\frac{i}{2|\omega|}e^{i|\omega||x-x'|}.
246: \label{freegreen}
247: \ee
248: When we recognize that the Green's function is the time-ordered product
249: of the fields,
250: \be
251: \langle \phi(x,t)\phi(x',t')\rangle=\frac1iG(x,t;x',t'),
252: \ee
253: we see that the vacuum expectation value of the stress tensor may be
254: obtained by applying a differential operator to the Green's function,
255: and then taking the spacetime points to be coincident.
256: For the $00$ component, that is, the energy, that differential
257: operator is
258: \be
259: \partial_0\partial'_0+\frac 12\partial^\lambda\partial'_\lambda
260: =\frac12\partial_0\partial_0'+\frac12\partial_x\partial'_x,
261: \ee
262: and so we obtain between the plates
263: \bea
264: \langle T^{00}\rangle&=&\int\frac{d\omega}{2\pi}\frac1{2i}(\omega^2+\partial_x
265: \partial'_x)g(x,x';\omega)\bigg|_{x=x'}\nonumber\\
266: &=&\int\frac{d\omega}{2\pi}\frac{\omega^2}2\frac{i}{\omega\sin\omega a}
267: [\sin\omega x\sin \omega(x-a)+\cos\omega x\cos\omega(x-a)]\nonumber\\
268: &=&\int\frac{d\omega}{2\pi}\frac{i\omega}2\cot\omega a
269: \to-\frac1{4\pi}\int_{-\infty}^\infty d\zeta\,\zeta\coth\zeta a.
270: \label{unsubt00}
271: \eea
272: Here, in the last step we have made the complex frequency rotation,
273: \be
274: \omega\to i\zeta.
275: \ee
276: We notice that this last integral in Eq.~(\ref{unsubt00})
277: does not exist. This is because for
278: large $\zeta$ the hyperbolic cotangent approaches unity. If we subtract
279: off this limiting value we obtain a finite result:
280: \bea
281: \langle T^{00}\rangle&\to&-\frac1{2\pi}\int_0^\infty d\zeta\,\zeta(\coth\zeta a
282: -1)
283: =-\frac1\pi\int_0^\infty \zeta\,d\zeta \frac1{e^{2\zeta a}-1}
284: =-\frac\pi{24 a^2}.
285: \eea
286: The energy is obtained from this by multiplying by the distance between
287: the plates:
288: \be
289: E=-\frac\pi{24 a},
290: \ee
291: which is the well-known L\"uscher potential.\cite{luscher80}
292:
293: In the same way we can calculate the vacuum expectation value of the
294: $xx$ component of the stress. The relevant differential operator
295: \be
296: \partial_x\partial'_{x}-\frac12\partial_\lambda\partial^{\prime\lambda}
297: \to\frac12(\omega^2+\partial_x\partial'_{x})
298: \ee
299: is unchanged, so we obtain the same result as for $\langle T^{00}\rangle$.
300: The off-diagonal terms $\langle T^{0x}\rangle$ result from the application
301: of the symmetric differential operator
302: \be
303: \frac12(\partial^0\partial^{\prime x}+\partial^x\partial^{\prime0}),
304: \ee
305: so are necessarily zero. Keeping the divergent term we subtracted off,
306: the result for the stress tensor between the plates, $0\le x,x'\le a$,
307: is
308: \be
309: \langle T^{\mu\nu}\rangle=\left[u_{\rm vac}-\frac\pi{24 a^2}\right]
310: \left(\begin{array}{cc}
311: 1&0\\
312: 0&1\end{array}\right),
313: \ee
314: where the divergent term is
315: \be
316: u_{\rm vac}=-\frac1{2\pi}\int_0^\infty d\zeta\,\zeta.
317: \ee
318: $\langle T^{\mu\nu}\rangle$ is traceless, as required by conformal symmetry:
319: \be
320: \langle T^\mu{}_\mu\rangle =0,
321: \ee
322:
323: If we follow the same operations to find the stress tensor outside
324: the plates from Eq.~(\ref{outgf}) we obtain
325: \bea
326: \langle T^{00}\rangle&=&\langle T_{xx}\rangle=\frac1{4\pi i}\int\frac{d\omega}
327: {\omega}\bigg[i\omega|\omega|\cos\omega(x-a)e^{i|\omega|(x-a)}
328: +\omega^2\sin\omega(x-a)e^{i|\omega|(x-a)}\bigg]\nonumber\\
329: &=&\frac1{4\pi}\int_{-\infty}^\infty d\omega\,|\omega|=-\frac1{2\pi}
330: \int_0^\infty d\zeta\,\zeta=u_{\rm vac}.
331: \eea
332: That is, in the two regions outside the plates, $x<0$ or $x>a$,
333: \be
334: \langle T^{\mu\nu}\rangle =u_{\rm vac}
335: \left(\begin{array}{cc}
336: 1&0\\
337: 0&1\end{array}\right).
338: \ee
339: This is exactly the stress tensor that would be found everywhere
340: if the free Green's function $g_0$ in Eq.~(\ref{freegreen}) were used.
341: This means that the force on one of the plates is completely finite
342: and unambiguous, because it is given by the discontinuity of the $xx$
343: component of the stress tensor across the plate (which follows immediately
344: from the physical meaning of the stress tensor in terms of the flux of
345: momentum):
346: \be
347: F=\langle T_{xx}\rangle\bigg|_{x=a-}-\langle T_{xx}\rangle\bigg|_{x=a+}
348: =-\frac\pi{24a^2}.
349: \ee
350: Since energies are undefined up to a constant, without any loss of generality
351: we may take the stress tensor to be completely finite:
352: \be
353: \langle T^{\mu\nu}(x)\rangle\to\left\{\begin{array}{cc}
354: -\frac\pi{24a^2}\left(\begin{array}{cc}
355: 1&0\\
356: 0&1\end{array}\right),&0\le x\le a,\\
357: 0,&x<0\quad\mbox{or}\quad x>a.
358: \end{array}\right.
359: \ee
360: \subsection{Massless Scalar in 3+1 Dimensions}
361: In higher dimensions, surface divergences appear. These were discussed
362: in detail in Ref.\cite{milton01}.
363:
364: In three space dimensions,
365: the use of the canonical stress tensor (\ref{stresstensor})
366: leads to the following expression for the vacuum
367: expectation value of the energy density,
368: \be
369: \langle T^{00}\rangle=\int\frac{d\omega}{2\pi}\frac{d^2k}{(2\pi)^2}\langle
370: t^{00}\rangle,\ee
371: where we have Fourier transformed both in frequency and transverse momentum.
372: If we take the plates to be located at $z=0$ and at $z=a$, we obtain
373: $\langle t^{00}\rangle$ by applying
374: the differential operator
375: \be
376: \frac12(\partial_0\partial_0'+\partial_x\partial_x'+\partial_y\partial_y'
377: +\partial_z\partial_z')\to\frac12(\omega^2+k^2+\partial_z\partial_z')
378: \ee
379: to the Green's function (\ref{ingf}) with $\omega\to\lambda\equiv
380: \sqrt{\omega^2-k^2}$,
381: \be
382: 0\le x,x'\le a:\quad g(x,x';\lambda)=-\frac{\sin\lambda x_<\sin\lambda(x_>-a)}
383: {\lambda\sin\lambda a}.
384: \label{ingf2}
385: \ee
386: The result,
387: \begin{equation}
388: \langle t^{00}\rangle =- \frac{1}{2i\lambda\sin\lambda a}[\omega^2\cos\lambda a
389: -k^2\cos\lambda(2z-a)],
390: \end{equation}
391: is evaluated by making a Euclidean rotation,
392: \begin{equation}
393: \omega\to i\zeta,\quad \lambda\to i\kappa,
394: \end{equation}
395: and introducing polar coordinates in the $\zeta$, $k$
396: plane,
397: \begin{equation}
398: \zeta=\kappa\cos\theta,\quad k=\kappa\sin\theta,
399: \label{polarcoord}
400: \end{equation}
401: so
402: \begin{eqnarray}
403: \langle T^{00}\rangle(z)
404: &=&-\frac{1}{4\pi^2}\int_0^\infty \kappa\,d\kappa\int_0^{\pi/2}d\theta\,\kappa^2
405: \frac{\sin\theta}{\sinh\kappa a}[\cos^2\theta\cosh\kappa a+\sin^2\theta\cosh\kappa
406: (2z-a)]\nonumber\\
407: &=&-\frac{1}{12\pi^2}\int_0^\infty d\kappa\,\kappa^3\frac{1}{\sinh\kappa a}
408: [\cosh\kappa a+2\cosh\kappa(2z-a)]\nonumber\\
409: &=&-\frac{1}{6\pi^2}\int_0^\infty d\kappa\,\kappa^3\left(\frac{1}
410: {e^{2\kappa a}-1}+
411: \frac{1}{2}+\frac{e^{2\kappa z}+e^{2\kappa(a-z)}}{e^{2\kappa a}-1}
412: \right).\nonumber\\
413: \end{eqnarray}
414:
415: Notice that the second term in the last integrand here corresponds to a constant
416: energy density, independent of $a$, so as before
417: it may be discarded as irrelevant. If we integrate the third term over $z$,
418: \begin{equation}
419: \int_0^a dz\left[e^{2\kappa z}+e^{2\kappa(a-z)}\right]=\frac{1}{\kappa}
420: \left[e^{2\kappa a}-1\right],
421: \end{equation}
422: we obtain another (divergent) constant term,
423: so the only part of the vacuum
424: energy corresponding to an observable force is that coming from the first term:
425: \begin{equation}
426: \int_0^a dz\,\langle T^{00}\rangle(z)
427: =-\frac{a}{6\pi^2}\int_0^\infty d\kappa\frac{\kappa^3}{ e^{2\kappa
428: a}-1}=-\frac{\pi^2}{1440 a^3},
429: \label{localcasenergy}
430: \end{equation}
431: which is the well-known Casimir energy/area for a massless scalar field subject
432: to Dirichlet boundary conditions, one-half that for an electromagnetic
433: field.
434:
435: In general, we have
436: \begin{eqnarray}
437: \langle T^{00}\rangle (z)&=&u+g(z),\label{localed1}
438: \end{eqnarray}
439: where
440: \begin{eqnarray}
441: u=-\frac{\pi^2}{1440 a^4},\quad
442: g(z)=-\frac{1}{6\pi^2}\frac{1}{16a^4}\int_0^\infty dy\,y^3\frac{e^{yz/a}
443: +e^{y(1-z/a)}}{ e^y-1}.\label{localed}
444: \end{eqnarray}
445: If we expand the denominator in a geometric series,
446: \begin{equation}
447: \frac{1}{e^y-1}=\frac{e^{-y}}{1-e^{-y}}=\sum_{n=1}^\infty e^{-ny},
448: \end{equation}
449: we can express $g$ in terms of the generalized or Hurwitz zeta
450: function,
451: \begin{equation}
452: \zeta(s,a)\equiv \sum_{n=0}^\infty \frac{1}{(n+a)^s}, \quad a\ne \mbox{ a
453: negative integer},
454: \end{equation}
455: as follows:
456: \begin{equation}
457: g(z)=-\frac{1}{16\pi^2 a^4}[\zeta(4,z/a)+\zeta(4,1-z/a)].
458: \label{gzzeta}
459: \end{equation}
460: This function is plotted in Fig.~\ref{fig2}, where it will be
461: observed that it diverges quartically as $z\to 0$, $a$. (Its $z$ integral
462: over the region between the plates diverges cubically.) As we have seen,
463: this badly behaved function does not contribute to the force on the
464: plates.
465:
466: \begin{figure}[ht]
467: \begin{center}
468: \begin{turn}{270}
469: \epsfxsize=8cm %width of figure - will enlarge/reduce the figures
470: \epsfbox{gofz.ps}
471: \end{turn}
472: \end{center}
473: \caption{Local Casimir energy density between parallel plates.}
474: \label{fig2}
475: \end{figure}
476:
477:
478: Next, we turn to $\langle T_{zz}\rangle$. According to the stress tensor
479: (\ref{stresstensor}) and the Green's function (\ref{ingf2}),
480: that is given by
481: \begin{eqnarray}
482: &&\langle T_{zz}\rangle=\frac{1}{2i}(\partial_z\partial'_{z}-\partial_x
483: \partial'_{x}-\partial_y\partial'_{y}+\partial_0\partial'_{0})G(x,x')
484: \nonumber\\
485: &=&-\frac{1}{2i}\int\frac{d\omega\, d^2k}{(2\pi)^3}(\partial_z\partial'_{z}
486: +\lambda^2)\frac{1}{\lambda\sin\lambda a}\sin\lambda z_<\sin\lambda
487: (z_>-a)\nonumber\\
488: &=&-\frac{1}{2i}\int\frac{d\omega \,d^2k}{(2\pi)^3}\frac{\lambda}{\sin\lambda a}
489: [\cos\lambda z\cos\lambda(z-a)+\sin\lambda z\sin\lambda(z-a)]\nonumber\\
490: &=&\int\frac{d\omega \,d^2k}{(2\pi)^3}\frac{i\lambda}{2}\cot\lambda a,
491: \end{eqnarray}
492: which is independent of $z$; that is, the
493: normal-normal component of the
494: expectation value of the stress tensor between the plates is constant.
495: If once again, the irrelevant $a$-independent part is removed,
496: what is left
497: is just three times the constant part of the energy density (\ref{localed}),
498: \begin{equation}
499: \langle T_{zz}\rangle =-3\times \frac{\pi^2}{1440 a^4}.
500: \end{equation}
501:
502: The remaining nonzero components of the stress tensor are
503: \begin{eqnarray}
504: \langle T_{xx}\rangle&=&\langle T_{yy}\rangle=\frac{1}{2i}[
505: \partial_x\partial'_{x} -\partial_y\partial'_{y}-\partial_z\partial'_{z}+
506: \partial_0\partial'_{0}]G(x,x')\nonumber\\
507: &=&-\frac{1}{2i}\int\frac{d\omega\,d^2k}{(2\pi)^3}\frac{1}{\lambda\sin\lambda a}
508: [\omega^2\sin\lambda z\sin\lambda(z-a)-\lambda^2\cos\lambda z\cos\lambda(z-a)]
509: \nonumber\\
510: &=&-u-g(z),
511: \label{localtxx}
512: \end{eqnarray}
513: where we have again introduced polar coordinates in the frequency-wavenumber
514: plane, and again dropped the infinite ($a$-independent) constant in $u$.
515: Thus the tensor structure of stress tensor is
516: \begin{equation}
517: \langle T^{\mu\nu}\rangle(z)=u\left(\begin{array}{cccc}
518: 1&0&0&0\\
519: 0&-1&0&0\\
520: 0&0&-1&0\\
521: 0&0&0&3
522: \end{array}\right)+g(z)\left(\begin{array}{cccc}
523: 1&0&0&0\\
524: 0&-1&0&0\\
525: 0&0&-1&0\\
526: 0&0&0&0
527: \end{array}\right),
528: \label{tmunucan}
529: \end{equation}
530: where $u$ is given by (\ref{localed}) and $g$ by (\ref{gzzeta}).
531: Because $u$ is constant, this vacuum expectation value is divergenceless,
532: since $g(z)$ does not contribute to $\langle T^{zz}\rangle$:
533: \begin{equation}
534: \partial_\mu \langle T^{\mu\nu}\rangle=\partial_z\langle T^{zz}\rangle=0.
535: \end{equation}
536: The second term in (\ref{tmunucan}) diverges at the boundaries, $z=0$, $a$,
537: and has a integral over the volume which diverges; yet as we have seen, it is
538: physically irrelevant because its integral is independent of
539: $a$, and it has no normal
540: component. Is there a natural way in which it simply does not appear in the
541: local formulation?
542:
543: The affirmative answer hinges on the ambiguity in defining the stress
544: tensor.
545: This ambiguity is without effect as far as the
546: total stress or the total energy was concerned.
547: Now, however, we see the
548: virtue of the conformal stress tensor\cite{callan70}:
549: \begin{equation}
550: \tilde T^{\mu\nu}=\partial^\mu\phi\partial^\nu\phi-\frac{1}{2}g^{\mu\nu}
551: \partial_\lambda\phi\partial^\lambda\phi-\frac{1}{6}(\partial^\mu\partial^\nu
552: -g^{\mu\nu}\partial^2)\phi^2,
553: \end{equation}
554: which, because of the equation of motion $\partial^2\phi=0$, has a vanishing
555: trace,
556: \begin{equation}
557: \tilde T^\mu{}_\mu=0.
558: \end{equation}
559: If we use this stress tensor rather than the canonical one, we merely need
560: supplement the above computations by that of the vacuum expectation value
561: of the extra term. Thus to obtain $\langle \tilde T^{xx}\rangle$ we add to
562: (\ref{localtxx})
563: \begin{eqnarray}
564: \frac{1}{6i}(\partial_y^2+\partial_z^2-\partial_0^2)G(x,x)
565: &=&\frac{1}{6i}\int\frac{d\omega\,d^2k}{(2\pi)^3}\partial_z^2\left[-
566: \frac{1}{\lambda\sin\lambda a}\sin\lambda z\sin\lambda(z-a)\right]\nonumber\\
567: &=&-\frac{1}{6i}\int\frac{d\omega\,d^2k}{(2\pi)^3}\frac{2\lambda}{\sin\lambda z}
568: \cos\lambda(2z-a)=g(z),
569: \label{extraconformal}
570: \end{eqnarray}
571: which just cancels the extra term in (\ref{localtxx}). Again, because $G(x,x)$
572: only depends on $z$, there is no extra contribution to $\langle T_{zz}\rangle$:
573: \begin{equation}
574: -\frac{1}{6}(\partial_z^2-g_{zz}\partial^2)\langle\phi^2\rangle
575: =\frac{1}{6i}(\partial_x^2+\partial_y^2-\partial_0^2)G(x,x)=0.
576: \end{equation}
577: The extra term for $\langle T_{00}\rangle$ is just the negative of that
578: in (\ref{extraconformal}),
579: \begin{equation}
580: -\frac{1}{6i}\partial_z^2G(x,x)=-g(z),
581: \end{equation}
582: which cancels the second term in (\ref{localed1}).
583: Thus, the conformal
584: stress tensor has the following vacuum expectation value for the
585: region between the parallel plates:
586: \begin{equation}
587: \langle \tilde T^{\mu\nu}\rangle =u\left(\begin{array}{cccc}
588: 1&0&0&0\\
589: 0&-1&0&0\\
590: 0&0&-1&0\\
591: 0&0&0&3
592: \end{array}\right)
593: \label{st-pp}
594: \end{equation}
595: which is traceless, thereby respecting the conformal invariance of the massless
596: theory. This is just the result found by Brown and Maclay by general
597: considerations\cite{brown69} who argued that
598: \begin{equation}
599: \langle \tilde T^{\mu\nu}\rangle=u[4\hat z^\mu\hat z^\nu-g^{\mu\nu}],
600: \end{equation}
601: where $\hat z^\mu$ is the unit vector in the $z$ direction.
602:
603: \subsection{Massive Scalar in $D$ Spatial Dimensions}
604: It is instructive to repeat the above calculation for a massive scalar
605: field where the plates have $D-1$ transverse dimensions. We will use
606: the conformal stress tensor,
607: \bea
608: T^{\mu\nu}&=&\partial^\mu\phi\partial^\nu\phi-\frac12g^{\mu\nu}
609: (\partial_\lambda\phi
610: \partial^\lambda\phi+\mu^2\phi^2)-\alpha(\partial^\mu\partial^\nu-g^{\mu\nu}
611: \partial^2)\phi^2.
612: \eea
613: Here $\alpha$ has to be chosen to be $(D-1)/(4D)$ in order that the trace
614: vanish (by virtue of the field equations) in the massless limit:
615: \be
616: \alpha=\frac{D-1}{4D}:\quad T^\mu{}_\mu=-\mu^2\phi^2.
617: \ee
618: The calculation proceeds very similarly to that given above. The only
619: new element is writing the momentum integral in polar coordinates:
620: \be
621: d^{D-1}k=\frac{2\pi^{(D-1)/2}}{\Gamma\left(\frac{D-1}{2}\right)}k^{D-2}\,dk,
622: \ee
623: and then introducing polar coordinates as in Eq.~(\ref{polarcoord}).
624: We encounter the integrals
625: \be
626: \int_0^{\pi/2} d\theta\,(\sin\theta)^{D-2}=2^{D-3}\frac{\Gamma\left(
627: \frac{D-1}2\right)^2}{\Gamma(D-1)},
628: \ee
629: relative to which
630: \be
631: \langle\sin^2\theta\rangle=\frac{D-1}D,\quad\langle\cos^2\theta\rangle=\frac1D.
632: \ee
633: The result for the various nonzero components of the stress tensor are
634: ($\kappa^2=\rho^2+\mu^2$)
635: \bea
636: \langle T^{00}\rangle&=&-\frac{2^{-D}\pi^{-D/2}}{D\Gamma(D/2)}
637: \int_0^\infty d\rho\,\rho^{D-1}\frac1{\kappa\sinh\kappa a}
638: [\rho^2\cosh\kappa a
639: +\mu^2\cosh\kappa(2z-a)],\\
640: \langle T^{zz}\rangle&=&-\frac{2^{-D}\pi^{-D/2}}{\Gamma(D/2)}
641: \int_0^\infty d\rho\,\rho^{D-1}\kappa\coth\kappa a,
642: \\
643: \langle T^{xx}\rangle&=&\langle T^{yy}\rangle=\dots=-\langle T^{00}\rangle.
644: \eea
645: Surface divergent terms, which do not contribute to the observable force,
646: appear proportional to the square of the mass.
647: Now, the trace of the expectation value of the stress tensor is
648: nonzero because of the mass:
649: \bea
650: \langle T^\mu{}_\mu\rangle&=&\langle T^{zz}\rangle-D\langle T^{00}\rangle
651: =-\mu^2\langle\phi^2\rangle\nonumber\\
652: &=&-\mu^2\frac{2^{-D}\pi^{-D/2}}{\Gamma(D/2)}
653: \int_0^\infty d\rho\,\rho^{D-1}\frac1{\kappa\sinh\kappa a}
654: [\cosh\kappa a-\cosh\kappa(2z-a)].
655: \eea
656: Of course, the infinite $a$-independent stress which would be present
657: if the boundary were not present ($\coth\kappa a\to1$) is to be
658: removed. The well known\cite{ambjorn83} expressions for the force and the
659: energy between parallel plates may be easily recovered.
660: We do not see the necessity for the additional terms found by Graham
661: et al.\cite{graham01} in the energy to make the energy finite at zero
662: separation (the fact that the Casimir energy diverges at $a=0$ reflects the
663: infinite amount of energy released when the plates are pushed into
664: coincidence) nor the requirement that the energy should be infinite
665: at zero mass, when the observable force is finite there.
666:
667:
668: \section{Scalar Casimir Effect for a Dirichlet Sphere}
669: \label{Sec3}
670:
671: The calculation given in the first part of this talk
672: was that for a sphere in one
673: spatial dimension. Now we consider a massless scalar in three space
674: dimensions, with a spherical boundary on which the field vanishes.
675: This corresponds to the TE modes for the electrodynamic situation first
676: solved by Boyer.\cite{boyer68} The general calculation in $D$
677: dimensions was given in Bender and Milton\cite{bender94}; the force per unit area is
678: given by the formula
679: \bea
680: \mathcal{F}&=&-\sum_{l=0}^\infty\frac{(2l+D-2)\Gamma(l+D-2)}{l!2^D\pi^{(D+1)/2}
681: \Gamma(\frac{D-1}2)a^{D+1}}\int_0^\infty dx\,x\frac{d}{dx}\ln\left[I_\nu(x)
682: K_\nu(x)x^{2-D}\right].\eea
683: Here $\nu=l-1+D/2$. For $D=3$ this expression reduces to
684: \bea
685: \mathcal{F}&=&-\frac1{8\pi^2a^4}\sum_{l=0}^\infty(2l+1)\int_0^\infty dx\,x
686: \frac{d}{dx}\ln\left[I_{l+1/2}(x)K_{l+1/2}(x)/x\right].
687: \label{fsphere}
688: \eea
689: In Ref.~\refcite{bender94} we evaluated this expression by continuing in $D$ from
690: a region where both the sum and integrals existed. In that way, a completely
691: finite result was found for all positive $D$ not equal to an even integer.
692: Here we will adopt a perhaps more physical approach, that of allowing the
693: time coordinates in the underlying Green's function to approach each other,
694: as described in Ref.~\refcite{milton78}.
695: That is, we recognize that the $x$
696: integration above is actually a (dimensionless) frequency integral, and
697: therefore we should replace
698: \be
699: \int_0^\infty dx\,f(x)=\frac12\int_{-\infty}^\infty dy\,e^{iy\delta}f(|y|),
700: \ee
701: where at the end we are to take $\delta\to0$. Immediately, we can
702: replace the $x^{-1}$ inside the logarithm in Eq.~(\ref{fsphere})
703: by $x$, which makes the integrals
704: converge, because the difference is proportional to a delta function in
705: the time separation, a contact term without physical significance.
706:
707: To proceed, we use the uniform asymptotic expansions for the modified
708: Bessel functions. This
709: is an expansion in inverse powers of $\nu=l+1/2$, low terms in which turn
710: out to be remarkably accurate even for modest $l$. The leading terms
711: in this expansion are
712: \be
713: \ln\left[x I_{l+1/2}(x)K_{l+1/2}(x)\right]
714: \sim\ln\frac{zt}2+\frac1{\nu^2}g(t)+\frac1{\nu^4}
715: h(t)+\dots,
716: \label{uae}
717: \ee
718: where $x=\nu z$ and $t=(1+z^2)^{-1/2}$.
719: Here
720: \bea
721: g(t)&=&\frac18(t^2-6t^4+5t^6),\\
722: h(t)&=&\frac1{64}(13t^4-284t^6+1062t^8-1356t^{10}+565t^{12}).\nonumber
723: \eea
724: The leading term in the force/area is therefore
725: \bea
726: \mathcal{F}_0&=&
727: -\frac1{8\pi^2a^4}\sum_{l=0}^\infty(2l+1)\nu\int_0^\infty dz\, t^2
728: =-\frac1{8\pi a^4}\sum_{l=0}^\infty\nu^2=\frac3{32\pi a^4}\zeta(-2)=0,
729: \eea
730: where in the last step we have used a formal zeta function
731: evaluation.\footnote{Note that the corresponding TE contribution for
732: the electromagnetic Casimir effect would not be zero, for there the sum
733: starts from $l=1$.}
734: Here the rigorous way to argue is to recall the
735: presence of the point-splitting factor $e^{i\nu z\delta}$ and to carry out
736: the sum on $l$ using
737: \be
738: \sum_{l=0}^\infty e^{i\nu z\delta}=-\frac1{2i}\frac1{\sin z\delta/2},
739: \ee
740: so
741: \bea
742: \sum_{l=0}^\infty \nu^2e^{i\nu z\delta}&=&-\frac{d^2}{d(z\delta)^2}\frac{i}
743: {2\sin z\delta/2}
744: =\frac{i}8\left(-\frac2{\sin^3z\delta/2}+\frac1{\sin z\delta/2}\right).
745: \eea
746: Then $\mathcal{F}_0$ is given by the divergent expression
747: \be
748: \mathcal{F}_0=\frac{i}{\pi^2 a^4\delta^3}\int_{-\infty}^\infty \frac{dz}{z^3}
749: \frac1{1+z^2},
750: \ee
751: which we argue is zero because the integrand is odd.
752:
753: The next term in the uniform asymptotic expansion (\ref{uae}),
754: that involving $g$, is
755: likewise zero, as intimated by the formal zeta function identity,
756: \be
757: \sum_{l=0}^\infty \nu^s=(2^{-s}-1)\zeta(-s),
758: \ee
759: which vanishes at $s=0$. The same conclusion follows from point splitting,
760: as we can see through use of the Euler-Maclaurin sum formula,
761: \be
762: \sum_{l=0}^\infty f(l)=\int_0^\infty dl\,f(l)+\frac12f(0)-\sum_{k=1}^\infty
763: \frac{B_k}{(2k)!}f^{(2k-1)}(0).
764: \ee
765: Here we have
766: \be
767: \int_0^\infty dl\,e^{i\nu z\delta}=-\frac{e^{iz\delta/2}}{iz\delta}
768: =-\frac1{i z\delta}-\frac12+\mathcal{O}(\delta).
769: \ee
770: We argue again that the first term here gives no contribution to the integral
771: over $z$ because it is odd,
772: and then the first two terms in the Euler-Maclaurin formula give
773: \bea
774: \mathcal{F}_1=-\frac1{8\pi^2a^4}\bigg[-\frac12\int_{-\infty}^\infty dz\,z\frac{d}{dz}
775: g(t)+\frac12\int_{-\infty}^\infty dz\,z\frac{d}{dz}g(t)\bigg]=0.
776: \eea
777: Derivatives of $e^{i\nu z\delta}$ with respect to $l$ all vanish at $\delta=0$.
778: Again, this cancellation does not occur in the electromagnetic case
779: because there the sum starts at $l=1$.
780: So here the leading term which survives is that of order $\nu^{-4}$
781: in Eq.~(\ref{uae}),
782: namely
783: \be
784: \mathcal{F}_1=\frac1{4\pi^2a^4}\sum_{l=0}^\infty \frac1{\nu^2}\int_0^\infty
785: dz \,h(t),
786: \ee
787: where we have now dropped the point splitting factor because this expression
788: is completely convergent. The integral over $z$ is
789: \be
790: \int_0^\infty dz \, h(t)=\frac{35\pi}{32768}
791: \ee and the sum over $l$ is $3\zeta(2)=\pi^2/2$, so the leading contribution
792: to the stress on the sphere is
793: \be
794: \mathcal{S}_2=4\pi a^2\mathcal{F}_2=\frac{35\pi^2}{65536a^2}=\frac{0.00527094}{a^2}.
795: \ee
796: Numerically this is a terrible approximation.
797:
798:
799:
800: What we must do now is return to the full expression and add and subtract
801: the leading asymptotic terms. This gives
802: \be
803: \mathcal{S}=\mathcal{S}_2-\frac1{2\pi a^2}\sum_{l=0}^\infty(2l+1)R_l,
804: \ee
805: where
806: \be
807: R_l=Q_l+\int_0^\infty dx\left[\ln zt+\frac1{\nu^2}g(t)+\frac1{\nu^4}h(t)\right],
808: \label{remainder}
809: \ee
810: where the integral
811: \be
812: Q_l=-\int_0^\infty dx\ln[2xI_\nu(x)K_{\nu}(x)]
813: \ee
814: has the asymptotic form ($l\gg1$)
815: \bea
816: Q_l&\sim&\frac{\nu\pi}2+\frac\pi{128\nu}-\frac{35\pi}{32768\nu^3}
817: +\frac{565\pi}{1048577\nu^5}\label{ql}-\frac{1208767\pi}{2147483648\nu^7}
818: +\frac{138008357\pi}{137438953472\nu^9}.\nonumber\\
819: \eea
820: The first two terms in Eq.~(\ref{ql}) cancel the second and third terms in
821: Eq.~(\ref{remainder}), of course.
822: The third term in Eq.~(\ref{ql}) corresponds to $h(t)$,
823: so the last three terms
824: displayed in Eq.~(\ref{ql}) give the asymptotic behavior of the remainder,
825: which we call $w(\nu)$. Then we have, approximately,
826: \be
827: \mathcal{S}\approx \mathcal{S}_2-\frac1{\pi a^2}\sum_{l=0}^n\nu R_l-\frac1{\pi a^2}
828: \sum_{l=n+1}^\infty \nu w(\nu).
829: \ee
830: For $n=1$ this gives $\mathcal{S}\approx0.00285278/a^2$, and for larger $n$
831: this rapidly approaches the value first given in Bender and
832: Milton\cite{bender94}:
833: \be
834: \mathcal{S}=0.002817/a^2,
835: \ee
836: a value much smaller than the famous electromagnetic
837: result\cite{boyer68,milton78}
838: \be
839: \mathcal{S}^{\rm EM}=\frac{0.04618}{a^2},
840: \ee
841: because of the cancellation of the leading terms noted above.
842:
843: \newcommand{\Tr}{\mbox{Tr}\,}
844: \section{Diagrammatic Divergence Structure}
845: \label{Sec4}
846: So far we have come to rather different conclusions from
847: those of Graham et al.\cite{graham01} For the case of parallel plates,
848: we found:
849: \begin{itemize}
850: \item The massless theory is perfectly well defined (no infrared divergences), and
851: surface divergences, which in any case have no physical consequences, do not
852: appear if the conformal stress tensor is used.
853: \item The vacuum expectation value of the stress tensor for the case of a massive
854: scalar does have surface divergences, which are proportional to the mass squared, but
855: which do not contribute to the force and are therefore physically irrelevant.
856: \end{itemize}
857: For a massless scalar with a spherical boundary in three dimensions, the formal
858: expressions for the force/area and the energy are formally divergent, yet if they
859: are regulated, say by point-splitting, the divergences cancel and the energy and
860: self-stress on the sphere are completely finite and unambiguous.
861:
862: Graham et al.\cite{graham01} dispute this. However, their disagreement
863: with us on the $D=1$ case seems entirely semantic, and without observable consequence.
864: Their substantial argument hinged on their $D=2$ calculation. However,
865: it is well known
866: that the Casimir effect for a circle is divergent,
867: so it is hard to draw general
868: inferences from an examination of that situation.\footnote{However,
869: in their most recent papers, Graham et al.\cite{graham03}
870: consider a three-dimensional $\delta$-shell potential, and show a divergence
871: occurs there not only
872: in second order, which we do not find, but in third order,
873: with which we concur.\cite{delta04}} Here, we will re-examine
874: some of their
875: general arguments for a hypersphere in $D$ space dimensions.
876:
877: The general analysis for that case was given in Bender and
878: Milton\cite{bender94}; it is clear that
879: the point-splitting method given in the previous section could be applied in that analysis.
880: Instead, we will here focus on the issue of the second-order Feynman graph which
881: supposedly is the signal for the divergence of the theory in any number of
882: space dimensions. (It is the oversubtracted graph which leaves the
883: mode sum more convergent.)
884: We will adopt a somewhat simpler formalism than that given by Graham
885: et al.,\cite{graham01} based
886: on the ``trace-log'' formula for the energy,
887: \be
888: E=\frac{i}{2T}\Tr \ln G,
889: \label{trln}
890: \ee
891: where for a ``polarization'' operator $\Pi$
892: \be
893: G=G_0(1+\Pi G)=G_0(1+\Pi G_0+\Pi G_0\Pi G_0+\dots).
894: \label{gpig}
895: \ee
896:
897: The highly sensible approach of Graham et al.\cite{graham01} is to replace
898: ideal boundary conditions by an interaction with an external field $\sigma$.
899: The Lagrangian for the scalar field is thus taken to be
900: \be
901: \mathcal{L}=-\frac12(\partial_\mu\phi\partial^\mu\phi+m^2\phi^2+\sigma(r)\phi^2),
902: \ee
903: where, anticipating spherical symmetry, we have taken the external field to depend
904: only on the spatial radial coordinate. In the end, we may take $\sigma$ to
905: be a delta function,
906: \be
907: \sigma(r)=\frac{g}{a}\delta(r-a),
908: \ee
909: where $g$ is dimensionless
910: and the formal $g\to\infty$ limit corresponds to the situation of a
911: Dirichlet
912: spherical shell. We can now evaluate the one-loop vacuum energy by the replacement
913: $\Pi\to\sigma$ in Eqs.~(\ref{trln}), (\ref{gpig}). It is the second-order graph
914: that is supposed to signal nonrenormalizability.
915:
916: We carry out the calculation in $D$ dimensions.\footnote{Equation (\ref{ed})
917: incorporates a net $1/(4\pi)$ correction relative to the formula (4.4) in
918: Ref.~\refcite{prd03}.}
919: \bea
920: E&=&\frac12\frac{i}{2T}\Tr\sigma G_0\sigma G_0\nonumber\\
921: &=&\frac{i}{4T}\int d^{D+1}x\,d^{D+1}y\,
922: \sigma(x)G_0(x-y)\sigma(y)G_0(y-x)\nonumber\\ &=&\frac{i}{4}\int d^Dx \,d^Dy\,
923: \sigma(|x|)\sigma(|y|)\int\frac{d\omega}{2\pi}
924: \int\frac{d^Dp}{(2\pi)^D}
925: \frac{d^Dq}{(2\pi)^D}\frac{e^{i{\bf (p-q)\cdot(x-y)}}}{(p^2+m^2)(q^2+m^2)},
926: \label{ed}
927: \eea
928: where in the last line we have carried out the integral on $t$ and $t'$, and as a result
929: $p^0=q^0=\omega$. Now we introduce polar coordinates, so in terms of the
930: last angle
931: \be
932: d^D x=A_{D-1}x^{D-1}dx \sin^{D-2}\theta\, d\theta,
933: \ee
934: where $A_n=2\pi^{n/2}/\Gamma(n/2)$ is the surface
935: area of a sphere in $n$ dimensions. Then we encounter a Bessel function
936: \bea
937: \int_0^\pi d\theta\sin^{D-2}\theta\, e^{i|{\bf p-q}|x\cos\theta}
938: =\left(\frac2{|{\bf p-q}|x}\right)^{D/2-1}\sqrt{\pi}\,
939: \Gamma\left(\frac{D-1}2\right)
940: J_{D/2-1}(|{\bf p-q}|x).\nonumber\\
941: \eea
942: Thus the Fourier transform of the field $\sigma(|x|)$ is defined by
943: \bea
944: \tilde\sigma(k)=\int d^D x \,e^{i{\bf k\cdot x}}\,\sigma(x)
945: =k\left(\frac{2\pi}k\right)^{D/2}\int_0^\infty dx\,x^{D/2}J_{D/2-1}(kx)
946: \sigma(x).\label{4.7}
947: \eea
948: (This agrees with the expression in Graham et al.\ for $D=2$.)
949: The expression for the energy reduces to
950: \be
951: E=\frac{i}4\int\frac{d\omega}{2\pi}\int\frac{d^Dq \, d^Dp}{(2\pi)^{2D}}
952: \frac{\tilde\sigma(|{\bf p-q}|)^2}{(p^2+m^2)
953: (q^2+m^2)}
954: \bigg|_{p^0=q^0=\omega}.
955: \ee
956: We carry out the momentum integrations by first using the proper-time
957: representation to combine the denominators:
958: \bea
959: \frac1{p^2+m^2}\frac1{q^2+m^2}&=&-\int_0^\infty ds\int_0^\infty ds'
960: e^{-is(p^2+m^2)-is'(q^2+m^2)}\nonumber\\
961: &=&-\int_0^\infty ds\,s\int_0^1 du\,e^{-ism^2-is(1-u)p^2-isuq^2},
962: \eea
963: where in the second line we replace $s\to s(1-u)$, $s'\to su$. In terms of
964: $\bf k=p-q$, we complete the square in the exponent by writing
965: \be
966: s(1-u){\bf p}^2+su{\bf q}^2=s[({\bf p}-u{\bf k})^2+{\bf k}^2u(1-u)],
967: \ee
968: while the corresponding $0$ components combine to give $-s\omega^2$. Now the
969: frequency and ${\bf p}$ integrals are just Gaussian:
970: \be
971: \int d\omega \,e^{is\omega^2}=e^{i\pi/4}\sqrt{\frac\pi s},\quad
972: \int d^D(p-uk)\,e^{-is({\bf p}-u{\bf k})^2}=e^{-i\pi D/4}
973: \left(\frac\pi s\right)^{D/2}.
974: \ee
975: Finally, we introduce polar coordinates for the $\bf k$ integration, with the result
976: \bea
977: E&=&-\frac{\pi^{-D-1/2}}{2^{2+2D}}\frac{\Gamma\left(\frac{3-D}{2}\right)}
978: {\Gamma\left(\frac{D}{2}\right)}\int_0^\infty dk\,k^{D-1}\tilde
979: \sigma(k)^2\int_0^1du\,[m^2+u(1-u)k^2]^{D/2-3/2},\nonumber\\
980: \label{4.12}
981: \eea
982: which yields the $D=2$ result given by Graham et al.\cite{graham01}
983:
984:
985: If we choose a delta-function potential,
986: \be
987: \sigma(x)=\frac{g}{a}\delta(x-a)
988: \ee
989: we obtain
990: \bea
991: E&=&-\frac{\pi^{-1/2}}{2^{2+D}}\frac{\Gamma\left(\frac{3-D}2\right)}{\Gamma\left(\frac{D}2
992: \right)}\frac{g^2}{a}\int_0^\infty d\xi\,\xi\,J_{D/2-1}^2(\xi)
993: \int_0^1du\,[m^2a^2
994: +\xi^2u(1-u)]^{(D-3)/2}.\nonumber\\
995: \eea
996: This appears to converge for $0<D<2$
997: except for the exceptional case $m=0$. In that
998: case the $u$ integral is simply
999: \be
1000: \frac{\Gamma\left(\frac{D-1}2\right)^2}{\Gamma(D-1)}
1001: =2^{2-D}\pi^{1/2}\frac{\Gamma\left(\frac{D-1}2\right)}{\Gamma(\frac{D}2)},
1002: \ee
1003: and the integral over the Bessel functions is
1004: \bea
1005: \int_0^\infty d\xi\, \xi^{D-2}J^2_{D/2-1}(\xi)&=&2^{D-2}\frac{\Gamma(2-D)
1006: \Gamma(D-3/2)}{\Gamma\left(\frac{3-D}2\right)^2\Gamma(\frac12)}\nonumber\\
1007: &=&\frac{\Gamma(1-D/2)\Gamma(D-3/2)}{2\pi\Gamma\left(\frac{3-D}2\right)},
1008: \eea
1009: which is valid in the region
1010: $\frac32<D<2$.
1011: Thus the energy for a massless scalar is
1012: \be
1013: E=-2^{-1-2D}\frac{g^2}{\pi a}\frac{\Gamma\left(\frac{D-1}2\right)\Gamma(D-3/2)
1014: \Gamma(1-D/2)}{\Gamma\left(\frac{D}2\right)^2},
1015: \label{result}
1016: \ee
1017: which we take to be the appropriate analytic continuation for all $D$.
1018: This exhibits poles at $D=2, 4, 6, \dots$, in congruence with the known
1019: divergence structure of the Casimir effect. There are also poles occurring
1020: at $D=1, -1, -3, \dots$, and at $D=3/2, 1/2, -1/2, \dots$. These latter
1021: two sequences of divergent dimensions correspond to infrared divergences
1022: that have no counterpart in the Casimir calculations, unlike the ultraviolet,
1023: even-integer poles. For space dimension between 2 and 4 the Casimir energy
1024: is completely finite, in concert with this diagnostic. The divergence at
1025: $D=2$, even putting aside the question of mass, is seen not to be generic.
1026:
1027:
1028: Instead of dimensional continuation, one can work directly in $D=3$.
1029: Let us regulate the theory by inserting a lower limit $s_0\to 0$ in the
1030: proper-time integration, so that for $m=0$ the energy (\ref{4.12}) becomes
1031: \bea
1032: E&=&\frac{1}{2^7\pi^4}\int_0^\infty dk\,k^2\tilde\sigma(k)^2\int_0^1 du
1033: \ln[s_0k^2u(1-u)]\nonumber\\
1034: &=&
1035: \frac{1}{2^7\pi^4}\int_0^\infty dk\,k^2\tilde\sigma(k)^2\frac{d}{d\alpha}
1036: \int_0^1 du[s_0k^2u(1-u)]^\alpha\bigg|_{\alpha=0}.
1037: \eea
1038: If the derivative acts on anything but $k^{2\alpha}$ we have
1039: \be
1040: \int_0^\infty dk\,k^2\tilde\sigma(k)^2=(2\pi)^3\int_0^\infty dx\,x^2\sigma(x)^2.
1041: \ee
1042: This diverges as $\sigma(x)\to(g/a)\delta(x-a)$;
1043: but if we regulate the divergence by point-splitting
1044: \be
1045: \sigma(x)^2\to\lim_{\xi\to\infty}\sigma(x-\xi)\sigma(x+\xi),
1046: \ee
1047: we have
1048: \be
1049: \int_0^\infty dk\,k^2\tilde\sigma(k)^2=(2\pi)^3g^2\delta(2\xi),
1050: \ee
1051: which is seen to be a contact term, independent of $a$.
1052: We are left with
1053: \bea
1054: E%&=&\frac{1}{2^5\pi^3}\frac{d}{d\alpha}\int_0^\infty dk\,k^2\tilde\sigma(k)^2
1055: %k^{2\alpha}\\
1056: =\frac1{8\pi^2}\int_0^\infty dx\,x\,\sigma(x)\int_0^\infty dy \,y\,
1057: \sigma(y)
1058: \frac{d}{d\alpha}\int_0^\infty dk\,k^{2\alpha}\sin kx\sin ky
1059: \bigg|_{\alpha=0}.
1060: \eea
1061: Here we have used the Fourier transformation expression (\ref{4.7}), but
1062: replaced Bessel functions of order 1/2 by the corresponding trigonometric
1063: functions. The $k$ integral is now obtained from ($-1<\alpha<0$)
1064: \bea
1065: \int_0^\infty dx\,x^\alpha\cos\beta x=\frac{\Gamma(\alpha+1)\cos(\alpha+1)
1066: \frac\pi2}{\beta^{\alpha+1}},\nonumber
1067: \eea
1068: which gives for the energy
1069: \bea
1070: E=\frac1{16\pi}\int_0^\infty dx\,x\sigma(x)\int_0^\infty dy \,y\sigma(y)
1071: \left(\frac1{x+y}-\frac1{|x-y|}\right)
1072: \to\frac{g^2}{32\pi a},
1073: \label{ogresult}
1074: \eea
1075: where we have omitted another infinite term that is independent of $a$.
1076: The result is exactly the $D=3$ value of Eq.~(\ref{result})!
1077: The justification for omitting (infinite) constant terms
1078: in the energy is that they are unobservable, not corresponding to a
1079: self-stress on the sphere.
1080:
1081: \subsection{Third-Order Divergence}
1082:
1083: In their most recent papers, Graham et al.\cite{graham03} explicitly
1084: consider a three-dimensional spherical shell, and show that
1085: with a $\delta$-function potential a divergence
1086: occurs not only in second order (which they acknowledge might be redefined
1087: away) but also in third order. The former is, as we have seen, illusory,
1088: while the latter is a real, logarithmic divergence, first discovered some
1089: years ago by Bordag et al.\cite{bkv} I have recently carried out a
1090: complete reconsideration of the Casimir self-stress due $\delta$-function
1091: potentials, reproduced the $\mathcal{O}(g^2)$ energy in Eq.~(\ref{ogresult}),
1092: and identified the $\mathcal{O}(g^3)$ divergence
1093: for the spherical shell configuration.
1094: It is possible that this divergence could be cancelled in the electromagnetic
1095: case between TE and TM modes. For further details see
1096: Ref.~\refcite{delta04}.
1097:
1098: \section{Conclusions}
1099: \label{Sec5}
1100: The challenge set forth by Graham et al.\cite{graham01,graham03}
1101: is physically appropriate and timely given the
1102: development of our understanding of the Casimir effect. Certainly
1103: those authors are justified in objecting to the loose use of
1104: the term ``renormalization'' in connection with various dubious processes
1105: for removing divergences in boundary-value Casimir problems. However,
1106: it is important to separate the wheat from the chaff. The Casimir
1107: force between parallel plates, the self-stress (or the force per unit
1108: area) on a perfect (Dirichlet or Neumann) spherical or cylindrical
1109: shell due to massless fields, the energy of fields confined to a curved
1110: manifold (a hypersphere or torus for example)
1111: are examples where the Casimir
1112: energy is unambiguous and finite, except for exceptional numbers of spatial
1113: dimensions.
1114: Of course, these are special cases, and
1115: generically Casimir energies are infinite.
1116: This is true if fields bounded by a spherical shell have
1117: mass, if the shell has finite thickness, or if
1118: the speed of light inside and outside the shell has different values.
1119: The latter case is the interesting one of a dielectric ball.
1120: The stress or the energy in that
1121: case is quartically divergent. I argued, very tentatively in
1122: 1980,\cite{milton80}
1123: and others argued more forcefully later,\cite{bordag97} that the divergent
1124: terms could be reabsorbed into the definition of physical properties
1125: of the material medium, the mass density, surface tension, and the like.
1126: This ``renormalization'' was in the spirit of the first use of renormalization
1127: in physics.\cite{poisson32}
1128: Obviously, this was not an altogether convincing
1129: argument, but it was argued to be on a par with
1130: perturbative renormalization of
1131: a quantum field theory.\cite{bordag02}.
1132: However, fairly recently, the discovery by
1133: several groups\cite{brevik99}
1134: that the finite part of the Casimir energy for a dilute
1135: dielectric sphere was unique, and coincided with that obtained by a
1136: regulated (dimensionally continued) calculation of the van der Waals
1137: energy,\cite{milton97}
1138: did provide some evidence that the divergences could be removed
1139: unambiguously,\footnote{This divergence of the Casimir energy in
1140: order $(\epsilon-1)^3$ for a dilute dielectric ball is analogous
1141: to the divergence of the Casimir energy of a $\delta$-function sphere
1142: at $\mathcal{O}(g^3)$, as noted in Ref.~\refcite{bkv}.}
1143: and had the practical consequence of destroying the hope
1144: of explaining sonoluminescence on the basis of quantum vacuum
1145: energy.\cite{brenner02}
1146:
1147: Obviously we are still at the early stages of understanding quantum
1148: field theory. The nature of divergences in vacuum energy calculations
1149: is still not understood. However, there are a few established peaks
1150: that rise above the murky clouds of ignorance, and we should not abandon
1151: them lightly because the rest is obscure.
1152:
1153:
1154: \section*{Acknowledgments}
1155: I am grateful to the US Department of Energy for
1156: partial support of this research. I thank Bob Jaffe,
1157: Steve Fulling, and Michael Bordag for helpful conversations.
1158: I thank Vladimir Mostepanenko for inviting me to participate in MGX.
1159:
1160:
1161:
1162:
1163:
1164:
1165:
1166:
1167:
1168:
1169: \begin{thebibliography}{99}
1170: \bibitem{casimir48} H. B. G.
1171: Casimir and D. Polder, Phys.\ Rev.\ {\bf 73}, 360 (1948).
1172:
1173: \bibitem{casimir49} H. B. G. Casimir, J. Chim.\ Phys.\ {\bf 46}, 407 (1949).
1174:
1175: \bibitem{casimir48a} H. B. G. Casimir,
1176: Proc.\ Kon.\ Ned.\ Akad.\ Wetensch.\ {\bf 51}, 793 (1948).
1177:
1178: \bibitem{boyer68} T. H. Boyer, Phys.\ Rev.\ {\bf 174}, 1764 (1968).
1179:
1180: \bibitem{deraad81} L. L. DeRaad, Jr., and K. A. Milton,
1181: Ann.\ Phys.\ (N.Y.) {\bf 136}, 229 (1981).
1182:
1183: \bibitem{sen81} S. Sen, Phys.\ Rev. D {\bf 24}, 869 (1981); J. Math.\ Phys.\
1184: {\bf 22}, 869 (1981).
1185:
1186: \bibitem{bender94} C. M. Bender and K. A. Milton, Phys.\ Rev.\ D {\bf 50}, 6547
1187: (1994).
1188:
1189: \bibitem{Milton:1997ri}
1190: K. A. Milton, Phys.\ Rev.\ D {\bf 55}, 4940 (1997).
1191:
1192:
1193:
1194:
1195: \bibitem{deutsch79} D. Deutsch and P. Candelas, Phys.\ Rev.\ D {\bf 20}, 3063
1196: (1979).
1197:
1198: \bibitem{brown69} L. S. Brown and G. J. Maclay, Phys.\ Rev.\ {\bf 184}, 1272
1199: (1969).
1200:
1201: \bibitem{milton80} K. A. Milton, Ann.\ Phys.\ (N.Y.) {\bf 127}, 49 (1980).
1202:
1203: \bibitem{bordag97} M. Bordag, E. Elizalde, K. Kirsten,
1204: and S. Leseduarte, Phys.\ Rev.\ D {\bf 56}, 4896 (1997);
1205: M. Scandurra, J. Phys.\ A {\bf 33}, 5707 (2000);
1206: S. K. Blau, M. Visser, and A. Wipf, Nucl.\ Phys.\ B {\bf 310}. 163 (1988).
1207:
1208: \bibitem{graham01}
1209: N. Graham, R. L. Jaffe, M. Quandt, and
1210: H. Weigel, Phys.\ Rev.\ Lett.\ {\bf 87}, 131601 (2001);
1211: N. Graham, R. L. Jaffe, and H. Weigel,
1212: Int.\ J. Mod.\ Phys.\ A {\bf 17}, 846 (2002);
1213: N. Graham, R. L. Jaffe, V. Khemani, M. Quandt, M. Scandurra, and
1214: H. Weigel, Nucl.\ Phys.\
1215: B {\bf 645}, 49 (2002); Phys.\ Lett.\ B {\bf 572}, 196 (2003).
1216:
1217: \bibitem{graham03}
1218: N. Graham, R. L. Jaffe, V. Khemani, M. Quandt, O. Schroeder, and H. Weigel,
1219: Nucl.\ Phys.\ {\bf B677}, 379 (2004); H. Weigel, hep-th/0310301.
1220:
1221: \bibitem{bkv} M. Bordag, K. Kirsten, and D. Vassilevich, Phys.\ Rev.\
1222: D {\bf 59}, 085011 (1999).
1223:
1224: \bibitem{delta04} K. A. Milton, hep-th/0401090.
1225:
1226: \bibitem{prd03} K. A. Milton, Phys.\ Rev.\ D {\bf 68}, 065020 (2003).
1227:
1228:
1229: \bibitem{luscher80} M. L\"uscher,
1230: K. Symanzik, and P. Weisz, Nucl.\ Phys.\ B {\bf 173}, 365 (1980).
1231:
1232: \bibitem{milton01} K. A. Milton, {\it The Casimir Effect:
1233: Physical Manifestations of Zero-Point Energy\/} (World Scientific,
1234: Singapore, 2001).
1235:
1236: \bibitem{callan70} C. G. Callan, Jr.,
1237: S. Coleman, and R. Jackiw, Ann.\ Phys.\ (N.Y.) {\bf 59}, 42 (1970).
1238:
1239:
1240: \bibitem{ambjorn83} J. Ambj\o rn and S. Wolfram, Ann.\ Phys.\ (N.Y.) {\bf 147},
1241: 1 (1983).
1242:
1243: \bibitem{milton78} K. A. Milton, L. L. DeRaad, Jr., and J.
1244: Schwinger, Ann.\ Phys.\ (N.Y.) {\bf 115}, 388 (1978).
1245:
1246: \bibitem{bordag 97} M. Bordag, E. Elizalde, K. Kirsten, and S. Leseduarte,
1247: Phys.\ Rev.\ D {\bf 56}, 4896 (1997).
1248:
1249: \bibitem{poisson32} S. D. Poisson, Mem.\ Acad.\ Sci. {\bf xl}, 521
1250: (1832), quoted in H. Lamb, {\it Hydrodynamics\/} (Dover, New York, 1945).
1251:
1252: \bibitem{bordag02} M. Bordag, A. S. Goldhaber,
1253: P. van Nieuwenhuizen, and D. Vassilevich,
1254: Phys.\ Rev.\ D {\bf 66} 125014 (2002).
1255:
1256: \bibitem{brevik99} I. Brevik, V. N. Marachevsky, and K. A. Milton,
1257: Phys.\ Rev.\ Lett.\ {\bf 82}, 3948 (1999); G. Barton, J. Phys.\ A {\bf 32},
1258: 525 (1999); M. Bordag, K. Kirsten, and D. Vassilevich, Phys.\ Rev.\ D
1259: {\bf 59}, 085011 (1999); J. S. H\o ye and I. Brevik, J. Stat.\ Phys.\
1260: {\bf 100}, 223 (2000); G. Lambiase, G. Scarpetta, and V. V. Nesterenko,
1261: Mod.\ Phys.\ Lett.\ A {\bf 16}, 1893 (2001).
1262:
1263: \bibitem{milton97} K. A. Milton and Y. J. Ng, Phys.\ Rev.\ E {\bf 55}, 4207
1264: (1997).
1265:
1266: \bibitem{brenner02} M. P. Brenner, S. Hilgenfeldt, and D. Lohse, Rev.\ Mod.\
1267: Phys.\ {\bf 74}, 425 (2002).
1268:
1269: \end{thebibliography}
1270: \end{document}
1271: