1: \subsection{Higher orders of the uniform asymptotic expansion of the \jf\\ and the \hk coefficient $a_{5/2}$}
2: \label{a52}
3: %
4: The magnetic background considered in Sec.(\ref{hamiltonians})
5: has singular surfaces where the magnetic field jumps. Here we represent the
6: calculations have been done in \cite{Drozdov:2002um}. The heat kernel expansion for the case of
7: singularities concentrated at surfaces has been considered in
8: [\cite{borvass}, \cite{moss}, \cite{gkv}, \cite{gkv2}]. Although the
9: general analysis of \cite{gkv} is valid for our background, an
10: explicit expression for $a_{5/2}$ has not been calculated before.
11:
12: We can use our obtained Jost function (\ref{f_v_of_ik+}) to calculate
13: the coefficient $a_{5/2}$ in the heat kernel expansion.
14:
15: Suppose we have obtained the value of $E_0(s)$ (\ref{e0_of_lnf})in the
16: point $s=-1/2$. It follows from (\ref{zeta_d_expansion})
17: that
18: \begin{equation}
19: E_0(s)\sim - \fr{\mu^{2s}}{2(4\pi)^{3/2} } \fr{m^{4-2s}}{\Gamma(s-\half)}
20: \sum\limits_{n=0, \half, 1, . . . }\fr{a_n}{m^{2n}} \Gamma(s-2+n)
21: \end{equation}
22: and at the limit of $s \rightarrow -\half$ only the term containing
23: $a_{5/2}$ remains to be nonzero in the sum.
24: \begin{equation}
25: E_0(\half)= - \fr{\mu^{2s}}{2(4\pi)^{3/2} } a_{5/2}
26: \end{equation}
27:
28: From the other hand we have (\ref{e0_of_lnf})
29: \begin{equation}
30: E_0(s)= C_s \sum\limits_{\nu=1/2, 3/2, . . . } \int\limits_m^\infty
31: dk(k^2-m^2)^{1-s} \fr{\pd}{\pd k} f_\nu(ik)=C_s h(s).
32: \end{equation}
33: Here we substitute the exact Jost function by its uniform asymptotic
34: represented in the form
35: \begin{equation}
36: f^{ua}_\nu(ik)=\sum\limits_{n=1,3,4,5...} \fr{h_n(t_1, t_2)}{\nu^n},
37: \label{unif_as_f}
38: \end{equation}
39: where the coefficients $h_n$ are functions of
40: $t_1=(1+(\fr{kR_1}{\nu})^2)^{-1/2}$,\\
41: $t_2=(1+(\fr{kR_2}{\nu})^2)^{-1/2}$, the power $\nu^{-2}$ is absent,
42: as noticed above (Sec.\ref{lippsw}),
43: \begin{equation}
44: C_s= -\fr{\mu^{2s}\Gamma(s-1)}{4 \sqrt{\pi}\Gamma(s-1/2)}\fr{-4 \sin(\pi s)}{\pi}
45: \end{equation}
46: At the limit $s \rightarrow -\half$ it yields
47: \begin{equation}
48: E_0(-\half)=\fr{4}{3\pi} Res_{s \rightarrow -\half} h(s)
49: \end{equation}
50: and therefore we obtain for $a_{5/2}$
51: \begin{equation}
52: a_{5/2}=-\fr{64\sqrt{\pi}}{3} Res_{s \rightarrow -\half} h(s).
53: \end{equation}
54:
55: To obtain the explicit form of $Res_{s \rightarrow -\half} h(s)$ we
56: use the uniform asymptotic expansion of the Jost function
57: (\ref{f_v_of_ik+}).
58: The terms $h_n(t_1,t_2)$ in uniform asymptotic expansion can
59: be obtained either by iterations of Lippmann-Schwinger equation (see
60: \cite{borkir2}, \cite{borkir1}) or by using the explicit form of the
61: Jost function as well. All the further terms up from $h_4$ are
62: produced from the explicit form of the Jost function
63: (\ref{f_v_of_ik+}) because of complication of the first way for higher
64: orders $n$ (see the remark to \ref{X_nj_spinor} in Sec.\ref{lippsw}). Namely, we
65: obtain several higher orders $1/\nu$ of uniform asymptotic expansion
66: for special functions $I_{\nu},K_{\nu},_{1}F_1$ which the exact Jost
67: function (\ref{f_v_of_ik+}, \ref{f_v_of_ik-}) consists of (it can be done starting with
68: the explicit form for two first orders and executing the recursive
69: algorithm several times\cite{abrsteg}), then after substitution of
70: each of functions $I ,K, _{1}F_1$ by its corresponding uniform
71: asymptotic expansion and separation of powers of $\nu$ we arrive at
72: the form (\ref{unif_as_h}). The coefficients $h_n(t_1, t_2),
73: n=1,...,4$ are given in the Appendix (\ref{unif_as_h}).
74:
75: If the function $h_n(t_1, t_2)$ is a polynomial over $t_1, t_2$, so
76: we can consider some term $t^j , (t=t_i, i=1, 2 ) $ of it. Notice,
77: that for $h_1(t_1,t_2),h_3(t_1,t_2)$ it is not the case, but we can
78: treat the terms of kind $\fr{1}{1+t}$ and $\fr{1}{(1+t)^k}$,
79: $k$-integer, as an infinite sum of powers $t$.
80:
81: Apart from the construction of $t= t_1, t_2$ (\ref{unif_as_f}) they
82: are strictly positive and less than 1, therefore each of the series in $
83: \fr{1}{1+t}=\sum\limits_{i=0}^{\infty}t^{2i} - \sum\limits_{i=0}^{\infty} t^{2i+1}$ converges regular and
84: uniform and respecting that we have only finite integer powers $k$ so
85: that $\left( \sum\limits_{i'=0}^{\infty} t^{i'} \right)^{k'}$ converges as well
86: (the $k',i'$ are integer). The sum over $i'$ can be interchanged with the one over $\nu$
87: (Princeheim's Theorem) thus the following procedure for some integer power
88: $t^j$ is valid.
89:
90: Performing the sum
91: \begin{equation}
92: h(s)=\sum\limits_{\nu=1/2, 3/2,\dots} \int\limits_m^\infty
93: dk(k^2-m^2)^{1-s} \fr{\pd}{\pd k} \sum\limits_{n=1, 3, 4, 5} h_{n,j}(R_1,R_2) \fr{t^j}{\nu^n}
94: \end{equation}
95: by meaning of (\ref{abelplan}) we obtain the sum of two parts
96: \begin{eqnarray}
97: h(s)&=&\int\limits_0^\infty d\nu\int\limits_m^\infty dk(k^2-m^2)^{1-s}\fr{\pd}{\pd
98: k}\sum\limits_n h_{n,j}(R_1,R_2)\fr{t^j}{\nu^n}+ \\\nn
99: &&\int\limits_0^\infty \fr{d\nu}{e^{2\pi\nu}+1}\fr{1}{i}\left[ \int\limits_m^\infty
100: dk(k^2-m^2)^{1-s}\fr{\pd}{\pd k}\sum\limits_n h_{n,j}(R_1,R_2)\fr{t^j}{\nu^n}
101: \right]_{\nu=i\nu}^{\nu=-i\nu},
102: \label{two_parts}
103: \end{eqnarray}
104: where $t=t_{1, 2}$. The first summand of (\ref{two_parts}) gives for each power $j$ of $t=\{ t_1, t_2 \}$
105: (using the (\ref{ident1}))
106: \begin{equation}
107: -\fr{m^3}{2}
108: \fr{\Gamma(2-s)\Gamma(\fr{1+j-n}{2})\Gamma(s+\fr{n-3}{2})}{(Rm)^{n-1}\Gamma(j/2)}
109: \label{first_part}
110: \end{equation}
111: (where R denotes $R_1$ and $R_2$ respectively)
112: at the limit $s \rightarrow -\half$ only the terms corresponding to $n=2$
113: and $n=4$ (which will be calculated below) have a pole.
114:
115: For the second part of the Abel-Plana formula (\ref{two_parts}) we have
116: \begin{eqnarray}
117: &&\fr{1}{i}\int\limits_0^\infty\fr{d\nu}{1+e^{2\pi\nu}}
118: \fr{\Gamma(2-s)\Gamma(s+j/2-1)}{jR^{4-2s}\Gamma(j/2)} \\
119: &\times&\left[(i\nu)^{2+j-n}[(mR)^2+(i\nu)^2]^{1-j/2-s}-(-i\nu)^{2+j-n}[(mR)^2+(-i\nu)^2]^{1-j/2-s}\right]\nn
120: \label{2_part}
121: \end{eqnarray}
122:
123: It can be seen using the Laurent expansions at $s \rightarrow -\half$
124: of Gamma functions entering the (\ref{2_part}), that only the terms
125: containing -1, 1, and 3 powers of $t_i$ can contribute to the residuum
126: at $s \rightarrow -\half$. But the expression in square brackets in
127: (\ref{2_part}) can be performed as
128: \begin{equation}
129: -\nu^2(\nu^2-(mr)^2)[(i\nu)^{j-n}e^{i\pi (1-j/2-s)}-(-i\nu)^{j-n}e^{-i\pi (1-j/2-s)}]
130: \end{equation}
131: thus for $j-n$ even the expression in square brackets produces (dropping the
132: non sufficient coefficient $\pm 2$ or $\pm 2i$):\\
133: \hspace*{2cm}for $j=-1, 1 :\ \ \ \ \sin \pi(1-j/2-s)=\cos \pi s $\\
134: \hspace*{2cm}for $j=3 :\ \ \ \ \sin \pi(1-j/2-s)= -\cos \pi s $\\
135: \\
136: and for $j-n$ odd respectively\\
137: \hspace*{2cm}for $j=\pm 1 :\ \ \ \ \cos \pi(1-j/2-s)=\pm\sin \pi s $\\
138: \hspace*{2cm}for $j=3 :\ \ \ \ \cos \pi(1-j/2-s)= -\sin \pi s $\\
139: \\
140: But for $s \rightarrow -\half$ these functions behave as\\
141: \hspace*{2cm}$\sin \pi s \sim -1+\fr{\pi^2}{2}(\half+s)^2, \\ \cos \pi
142: s \sim \pi(\half+s)\\ $ and it means that only the contribution of
143: terms with odd powers $j-n$ of $i\nu$ could survive and these
144: corresponds for the possible values of $j$ to $\nu^{-2}, \nu^{-4},
145: \nu^{-6}, . . . $. In fact the coefficient $h_2(s)$ at $\nu^{-2}$ is
146: zero, and all other possible terms up from $\nu^{-4}$ does not contain
147: any powers of $t_i$ lower than 4; one can see it for example in the
148: explicit form of the uniform asymptotic expansion of special
149: functions entering the $\ln f_v(ik)$ (\ref{unif_as_h}, Sec.\ref{high_ord}).
150: Therefore the
151: second summand of (\ref{two_parts}) does not produce any contribution to
152: the $Res_{s \rightarrow -\half} h(s)$. Thus we have that only the
153: contribution from the first summand of (\ref{two_parts}) remains, and
154: since the term of $n=2$ is zero the searched residuum resulting from
155: the term of $n=4$ is:
156: \begin{equation}
157: Res_{s \rightarrow -\half} h(s)=-\half.\fr{3 \sqrt{\pi}}{4}
158: \fr{\Omega^2}{4(R_1^2-R_2 ^2)^2} (R_1+R_2)\sum\limits_{j=4, 6}\fr{\Gamma(\fr{1+j-n}{2})}{j/2},
159: \end{equation}
160: where the (\ref{first_part}) and the explicit form of $ h_4(t_1, t_2)=\fr{\Omega^2}{4(R_1^2-R_2 ^2)^2}[R_1^4(t_1^4-t_1^6)+R_2^4(t_2^4-t_2^6)]$ have been used.
161: Finally we have
162: \begin{equation}
163: Res_{s \rightarrow -\half} h(s)=\fr{15\pi}{128}\fr{\Omega^2(R_1+R_2)}{(R_1^2-R_2^2)^2}
164: \end{equation}
165: and therefore
166: \begin{equation}
167: a_{5/2}= \fr{5\pi^{3/2}}{2} \fr{\Omega^2(R_1+R_2)}{(R_1^2-R_2^2)^2}
168: \label{coeff_a52}
169: \end{equation}
170: %
171: This is the heat kernel coefficient $a_{5/2}$ for the configuration of
172: the magnetic background field as given by
173: Eqs.(\ref{potential_A}-\ref{h_r1}).
174:
175: We can calculate the heat kernel coefficient $a_{5/2}$ in a more
176: general situation when the magnetic field jumps on an {\it arbitrary}
177: surface $\Sigma$. The coefficients $a_n$ for $n=1/2,...,2$ can be read
178: off rather general expressions of the paper \cite{gkv}. Let $B^\pm$
179: be values of the magnetic field on two sides of $\Sigma$. According
180: the analysis of \cite{gkv}, the coefficient $a_{5/2}$ must be an
181: integral over $\Sigma$ of a local invariant of canonical mass
182: dimension 4, which is symmetric under the exchange of $B^+$ and $B^-$
183: and which vanishes if $B^+=B^-$ (i.e. when the singularity
184: disappears). There is only one such invariant which gives rise to the
185: following expression:
186: %
187: \begin{equation}
188: a_{5/2}= \xi \int\limits_{\Sigma} (\vec{B}^+ - \vec{B}^-)^2 d\mu(\Sigma),
189: \label{general_a52}
190: \end{equation}
191: where the integration goes over the surface $\Sigma$ and
192: $\vec{B}^{\pm}$ are the values of the magnetic field on both sides of
193: $\Sigma$ in the given point. The yet undefined constant $\xi $ can be
194: found using Eq.(\ref{coeff_a52}) which constitutes a special case of
195: (\ref{general_a52}). Here the surface $\Sigma$ consists of two circles
196: in the $ (\vec{X},\vec{Y})$-plane so that
197: \begin{equation}
198: \int\limits_{\Sigma} (\vec{B}^+ - \vec{B}^-)^2 d\mu(\Sigma)=2\pi(R_1+R_2)\vec{B}^2,
199: \end{equation}
200: where the jump $(\vec{B}^+ - \vec{B}^-)$ is just the value of
201: $\vec{B}$ at $r \in [R_1,R_2]$. With this, Eq.(\ref{general_a52})
202: takes the form
203: \begin{equation}
204: a_{5/2}=2\pi e^2 \xi\vec{B}^2(R_1+R_2).
205: \label{agen}
206: \end{equation}
207: On the other hand, from Eq.(\ref{h_r1}, \ref{delta}) we have
208: \begin{equation}
209: \Phi= B\int\limits_{r \in[R_1,R_2]}B d^2 x=\pi B(R_2^2-R_1^2)
210: \label{phiofb}
211: \end{equation}
212: and with Eq.(\ref{delta}) from Eq.(\ref{coeff_a52}) it follows that
213: \begin{equation}
214: a_{5/2}= \fr{5}{8\sqrt{\pi}} e^2 \Phi^2 \fr{R_1+R_2}{(R_2^2-R_1^2)^2}.
215: \label{asp}
216: \end{equation}
217: comparing (\ref{agen}) with (\ref{asp}) we get
218: \begin{equation}
219: \xi = \fr{5}{16\pi^{3/2}}.
220: \label{ksi}
221: \end{equation}
222:
223: %%% Local Variables:
224: %%% mode: latex
225: %%% TeX-master: t
226: %%% End:
227: