hep-ph0602217/ope.tex
1: \documentclass[prd,aps,preprintnumbers,eqsecnum,amsmath,amssymb,nofootinbib]{revtex4}
2: %\documentclass[prd,aps,twocolumn,preprintnumbers,eqsecnum,amsmath,amssymb,nofootinbib]{revtex4}
3: \usepackage{bm} 
4: \usepackage{epsfig}
5: %\renewcommand{\thesection}{\arabic{section}}
6: %\renewcommand{\thesubsection}{\arabic{subsection}}
7: % HEPHY-PUB 815/06
8: \renewcommand{\thetable}{\arabic{table}}
9: \renewcommand{\thefigure}{\arabic{figure}}
10: \renewcommand{\tablename}{Table}
11: \renewcommand{\figurename}{Fig.}
12: \begin{document}
13: \title{Quark--hadron duality and hadron properties \\
14: from correlators of pseudoscalar and axial currents} 
15: \author{W. Lucha$^{a}$ and D. Melikhov$^{a,b}$} 
16: \affiliation{
17: $^a$ Institute for High Energy Physics, Austrian Academy of Sciences, Nikolsdorfergasse 18, A-1050, Vienna, Austria\\
18: $^b$ Institute of Nuclear Physics, Moscow State University, 119992, Moscow, Russia}
19: \date{\today}
20: \begin{abstract}
21: We study the operator product expansion (OPE) 
22: and quark--hadron duality for 2- and 3-point correlators of the axial ($A$) 
23: and pseudoscalar ($P$) 
24: currents of the light quarks. In the chiral limit these correlators are often dominated by 
25: nonperturbative power corrections leading to subtleties of quark--hadron duality relations and of the extraction of 
26: properties of light pseudoscalars. 
27: For the 2-point correlators, we show the sum rule for $\langle PP\rangle$ to be sensitive to the 
28: excited light pseudoscalar. 
29: For the 3-point correlators, we derive the Ward identites which provide the normalization of the pion electromagnetic form factor 
30: at zero momentum transfer. 
31: For large momentum transfer, we demonstrate the way the correct behaviour of the pion 
32: form factor in agreement with perturbative QCD emerges from condensate terms in the OPE for the 
33: $\langle PVP \rangle$ and $\langle AVP\rangle$ correlators. 
34: The local-duality sum rule for $\langle AVA \rangle$ is shown to lead to 
35: the pion form factor with the required properties for all values of the momentum transfer. 
36: \end{abstract}
37: %\keywords{Quark-hadron duality, OPE, chiral symmetry}
38: %\pacs{12.Aw, 11.30.Rd, 12.38. Lg}
39: \maketitle
40: \section{Introduction}
41: Correlators of the axial and pseudoscalar currents are the basic objects for studying  
42: properties of light pseudoscalars within QCD sum rules \cite{svz,ioffe,radyushkin}, 
43: bound-state equations \cite{bound-state}, and lattice QCD \cite{lattice}. 
44: The axial current $j^5_\alpha(x)=\bar u(x) \gamma_\alpha \gamma_5 d(x)$ and the pseudoscalar current 
45: $j^5(x)=i\bar u(x)\gamma_5 d(x)$ satisfy the equation  
46: \begin{eqnarray}
47: \label{div}
48: \partial^\alpha j^5_\alpha(x)=(m_u+m_d)j^5(x). 
49: \end{eqnarray}
50: The axial current is conserved in the limit of massless quarks (the chiral limit). 
51: Therefore, the correlators of the axial current have the following property: 
52: perturbative contributions to the longitudinal structures of 
53: these correlators are suppressed by the light quark mass, 
54: and the operator product expansion (OPE) \cite{wilson} for the longitudinal structures 
55: is dominated by nonperturbative power corrections. 
56: This leads to specific features of the quark--hadron duality \cite{duality} in these cases. 
57: 
58: We shall be interested in extracting contributions of light pseudoscalars from the correlators of axial
59: and pseudoscalar currents of light quarks. The coupling of pseudoscalar mesons $P$ to these currents is governed 
60: by the decay constants $f_P$ defined according to equations 
61: \begin{eqnarray}
62: \langle 0|\bar u\gamma_\alpha\gamma_5 d|P(q)\rangle=if_P q_\alpha, \qquad 
63: \langle P(q)|\bar d\gamma_\alpha\gamma_5 u|0\rangle=-if_P q_\alpha, 
64: \end{eqnarray}
65: and 
66: \begin{eqnarray}
67: \langle 0|i\bar u\gamma_5 d|P(q)\rangle=\frac{f_P m_P^2}{m_u+m_d}, \qquad 
68: \langle P(q)|i\bar d\gamma_5 u|0\rangle=\frac{f_P m_P^2}{m_u+m_d}. 
69: \end{eqnarray}
70: The divergence equation requires 
71: \begin{eqnarray}
72: \label{relation}
73: f_Pm_P^2\propto m,  
74: \end{eqnarray}
75: implying that at least one of the quantities on the l.h.s.\ vanishes in the chiral limit. 
76: If chiral symmetry is spontaneously broken (as in QCD at zero temperature), 
77: Eq.~(\ref{relation}) leads to the following alternatives \cite{svz}:\footnote{
78: If no spontaneous breaking of chiral symmetry occurs, then in the chiral limit the pion 
79: behaves the same way as the excited pseudoscalars: it stays massive and its decay constant 
80: vanishes \cite{krassnigg}.}
81: \begin{eqnarray}
82: \label{constraints}
83: &&m^2_\pi=O(m),\quad f_\pi=O(1),\quad \mbox{ground-state pion} \nonumber\\  
84: &&m^2_{P}=O(1),\quad f_{P}=O(m),\quad \mbox{excited pseudoscalars}.  
85: \end{eqnarray}
86: Therefore, the vanishing of the decay constants of the excited pseudoscalars in the limit of massless quarks  
87: is a direct consequence and signature of chiral symmetry. 
88: (Hereafter, the subscript $P$ denotes the excited massive pseudoscalar, and $\pi$ stands for the pion). 
89: An important relation between the pion observables and the quark condensate is given by the 
90: Gell-Mann--Oakes--Renner (GMOR) formula \cite{gmor} 
91: \begin{eqnarray}
92: \label{gmor}
93: f_\pi^2 m^2_\pi =-(m_u+m_d)\langle \bar uu+ \bar dd\rangle+O(m_\pi^4).   
94: \end{eqnarray}
95: %Corrections to this relation proportional to higer powers emerge in ChPT. 
96: 
97: In this paper we study the consequences of spontaneously broken chiral symmetry for the 2- and 3-point
98: correlators of the axial and pseudoscalar currents, and for the extraction of properties of 
99: pseudoscalar mesons from these correlators. 
100: 
101: 
102: In Sec.~\ref{ii} we discuss OPE and duality for the 
103: $\langle PP\rangle$, $\langle AP\rangle$, and $\langle AA\rangle$ correlators, and the possibility to extract from these 
104: correlators the decay constants of the excited pseudoscalars. (For simplicity, we denote the vacuum average of the 
105: $T$-product of axial and pseudoscalar currents as $\langle AP\rangle$, etc.) 
106: 
107: In Sec.~\ref{iii} we consider the 3-point correlators $\langle PVP\rangle$, 
108: $\langle AVP\rangle$, and $\langle AVA\rangle$, $V$ being the electromagnetic current. 
109: We derive Ward identites for
110: these correlators and show the way the normalization of the pion form factor at zero momentum transfer 
111: arises due to these identites. We then analyse the region of large momentum transfers and study the way 
112: quark--hadron duality works for this case. We discuss a recent statement in the literature that 
113: the pion form factor as extracted from the $\langle PVP\rangle$ correlator does not have the right 
114: asymptotics according to perturbative QCD (pQCD). We show that the OPE for $\langle PVP\rangle$ and $\langle AVP\rangle$ at large $q^2$ is
115: dominated by nonperturbative corrections. In particular, the four-quark condensate 
116: $\langle\bar\psi\psi\bar\psi\psi\rangle$ in the case of $\langle PVP\rangle$ and the mixed condensate 
117: $\langle\bar \psi\sigma^{\mu\nu} G_{\mu\nu}\psi\rangle\equiv 
118: \langle\bar \psi\sigma G\psi\rangle$ in the case of $\langle AVP\rangle$ are demonstrated to be crucial for extracting the 
119: correct large-$q^2$ behavior of the pion form factor. 
120: We then discuss $\langle AVA\rangle$ and point out that the recent
121: calculation of radiative corrections to this correlator opens the possibility to 
122: obtain the pion electromagnetic form factor for all spacelike momentum transfers 
123: making use of the local duality sum rule. 
124: 
125: Section \ref{iv} summarizes our results. 
126:  
127: %\newpage 
128: 
129: 
130: \section{\label{ii}Two-point functions, $f_\pi$ and $f_{\pi'}$}
131: In this section, we consider all possible two-point correlators of the axial current 
132: $j^5_\alpha=\bar u \gamma_\alpha \gamma_5 d$, 
133: $j^{5\dagger}_\alpha=\bar d \gamma_\alpha \gamma_5 u$ and the pseudoscalar current
134: $j^5 =i\bar u \gamma_5 d$, $j^{5\dagger} =i\bar d\gamma_5 u$.  
135: 
136: \subsection{\boldmath $\langle PP\rangle$}
137: %{\boldmath $\langle T j^5 j^5\rangle$}
138: We start with the correlator of two pseudoscalar currents 
139: \begin{eqnarray}
140: \label{pi5}
141: \Pi^{5}(q)\equiv i\int d^4x\, e^{iqx}\langle T (j^5(x) j^{5\dagger}(0))\rangle =\Pi_5(q^2),  
142: \end{eqnarray}
143: where $\langle \cdot\rangle$ stands for $\langle 0|\cdot|0\rangle$. 
144: At small $q^2$, the correlator is dominated by the Goldstone boson, leading in the chiral limit to the expression 
145: \cite{chpt}   
146: \begin{eqnarray}
147: \Pi^{5}(q)= -\frac{\langle\bar uu+\bar dd\rangle^2}{f_\pi^2}\frac1{q^2}.
148: \end{eqnarray}
149: At large $q^2$, the OPE for this correlator reads \cite{svz} 
150: \begin{eqnarray}
151: \Pi_{5}(q^2)=\Pi^{\rm pert}_{5}(q^2)+\frac{(m_u+m_d)\langle\bar uu+\bar dd\rangle}{4q^2} 
152: -\frac{\langle \alpha_s GG \rangle}{8\pi q^2}+O\left(\frac{1}{q^4}\right), 
153: \end{eqnarray}
154: %$GG=G_{\mu\nu}G^{\mu\nu}$ 
155: where ${\rm Im}\,\Pi^{\rm pert}_5(s)={3s}/{8\pi}$ (for radiative corrections and higher 
156: condensates, see Ref.~\cite{khodjamirian} and references therein). 
157: 
158: \subsection{\boldmath $\langle AP\rangle$}
159: %{\boldmath $\langle T j^5_\alpha j^5\rangle$}
160: The correlator of a pseudoscalar and an axial current has the form 
161: \begin{eqnarray}
162: \Pi^{5}_{\alpha}(q)\equiv 
163: i\int d^4x\, e^{iqx}\langle T (j^5_\alpha(x)j^{5\dagger}(0))\rangle =iq_\alpha \Pi(q^2).  
164: \end{eqnarray}
165: To determine $\Pi(q^2)$, we calculate 
166: \begin{eqnarray}
167: q^\alpha \Pi^5_\alpha(q)&=&-\int d^4x\, e^{iqx}\frac{\partial}{\partial x_\alpha}  
168: \langle T( j^5_\alpha(x)j^{5 \dagger}(0))\rangle 
169: \nonumber\\&=&
170: -\int d^3 xe^{-i\vec q\vec x}\langle[j_0^5(0,\vec x), j^{5 \dagger}(0)]\rangle 
171: -\int d^4x\, e^{iqx}\langle T( \partial^\alpha j^5_\alpha(x)j^{5 \dagger}(0))\rangle.  
172: \end{eqnarray}
173: Making use of the commutation relation 
174: \begin{eqnarray}
175: \label{com1}
176: [\bar u\gamma_0\gamma_5d(0,\vec x),\bar d\gamma_5 u(0)]=-(\bar u u+\bar dd)\delta(\vec x), 
177: \end{eqnarray}
178: and the divergence equation (\ref{div}), we obtain 
179: \begin{eqnarray}
180: \label{pi}
181: \Pi(q^2)=\frac{\langle\bar u u+\bar dd \rangle}{q^2}+(m_u+m_d)\frac{\Pi_5(q^2)}{q^2},  
182: \end{eqnarray}
183: with $\Pi_5(q^2)$ given by Eq.~(\ref{pi5}). 
184: In the chiral limit $m_u=m_d=0$, the OPE for $\Pi^5_\alpha$ contains only one 
185: operator and takes the simple form 
186: \begin{eqnarray}
187: \label{pi5alpha}
188: \Pi^5_\alpha(q)=iq_\alpha\frac{\langle\bar u u+\bar dd \rangle}{q^2}.   
189: \end{eqnarray}
190: Let us now saturate the correlator $\Pi^{5}_{\alpha}$ with intermediate hadron states. 
191: Here we have contributions of single-state pseudoscalars (ground-state $\pi$ and excitations $P$) 
192: and the hadron continuum (e.g., multipion states with the relevant quantum numbers). 
193: The contribution of a single pseudoscalar state $P$ to the invariant amplitude $\Pi(q^2)$ has the form 
194: \begin{eqnarray}
195: \frac{f^2_P m_P^2}{m_u+m_d}\frac{1}{(m_P^2-q^2)}.  
196: \end{eqnarray}
197: In the chiral limit, neither excited pseudoscalars (since $f_P\propto m$) nor continuum states contribute to the correlator 
198: and it is fully saturated by the intermediate Goldstone boson for all values of $q^2$, 
199: \begin{eqnarray}
200: \label{2.10}
201: \Pi^5_\alpha(q)=-i\frac{q_\alpha}{q^2} \frac{f^2_\pi m_\pi^2}{m_u+m_d},    
202: \end{eqnarray}
203: and both representations (\ref{pi5alpha}) and (\ref{2.10}) are equal to each other by virtue of the GMOR relation.  
204: (Notice that the analysis of this correlator also provides a derivation of the GMOR relation.) 
205: It is interesting to note that in the chiral limit the duality between the OPE and the Goldstone 
206: contribution is locally fulfilled for all values of $q^2$. 
207: 
208: %\newpage
209: 
210: \subsection{\boldmath $\langle AA\rangle$}
211: %{ \boldmath $\langle T j^5_\alpha j^5_\beta\rangle$}
212: The correlator of two axial currents contains two Lorentz structures:  
213: \begin{eqnarray}
214: \label{piT}
215: \Pi^{5}_{\alpha\beta}(q)\equiv
216: i\int d^4x\, e^{iqx}\langle T (j^5_\alpha(x)j^{5\dagger}_\beta(0))\rangle 
217: =\left(g_{\alpha\beta}-\frac{q_\alpha q_{\beta}}{q^2}\right)\Pi_T(q^2)
218: +\frac{q_\alpha q_{\beta}}{q^2} \Pi_L(q^2). 
219: \end{eqnarray}
220: To determine the longitudinal part, let us calculate $q^\alpha\Pi^{5}_{\alpha\beta}$:
221: \begin{eqnarray}
222: \label{2.12}
223: q^\alpha \Pi^{5}_{\alpha\beta}(q)&=&
224: -\int d^4x\, e^{iqx}\frac{\partial}{\partial x_\alpha} 
225: \langle T (j^5_\alpha(x)j^{5\dagger}_\beta(0))\rangle 
226: \nonumber\\
227: &=&  
228: -\int d^3x e^{-i\vec q\vec x}\langle [j_0^5(0,\vec x), j^{5\dagger}_\beta(0)] \rangle 
229: -\int d^4x\, e^{iqx}\langle T (\partial^\alpha j^{5}_\alpha(x)\,j^{5\dagger}_\beta(0))\rangle 
230: \nonumber\\
231: &=& 
232: -(m_u+m_d)\int d^4x\, e^{iqx}\langle T (j^5(x)j^{5\dagger}_\beta(0))\rangle 
233: =i(m_u+m_d)\Pi^5_\beta(-q)=(m_u+m_d)q_\beta\Pi(q^2).
234: \end{eqnarray}
235: The contribution of the commutator term in Eq.~(\ref{2.12}) vanishes since 
236: \begin{eqnarray}
237: [j_0^5(0,\vec x), j^{5\dagger}_\beta(0)]=(\bar u\gamma_\beta u(0)-\bar d\gamma_\beta d(0))
238: \delta(\vec x)  
239: \end{eqnarray}
240: and $\langle \bar u\gamma_\beta u \rangle=\langle\bar d\gamma_\beta d\rangle=0$ 
241: because of the Lorentz invariance of the vacuum. 
242: Making use of Eq.~(\ref{pi}) we arrive at   
243: \begin{eqnarray}
244: \label{piab}
245: \Pi^{5}_{\alpha\beta}(q)
246: =\left(g_{\alpha\beta}-\frac{q_\alpha q_{\beta}}{q^2}\right)\Pi_T(q^2)
247: +\frac{q_\alpha q_{\beta}}{q^4}
248: \left((m_u+m_d)\langle\bar uu+\bar dd\rangle +(m_u+m_d)^2\Pi_5(q^2)\right). 
249: \end{eqnarray}
250: The OPE for the transverse function $\Pi_T(q^2)$ is known from Ref.~\cite{svz}. 
251: 
252: In the chiral limit, the axial current is conserved  
253: and the correlator is transverse. At small $q^2$, the correlator is dominated by the Goldstone 
254: pole 
255: \begin{eqnarray}
256: \Pi^{5}_{\alpha\beta}(q)=\left(g_{\alpha\beta}-\frac{q_\alpha q_\beta}{q^2}\right)
257: \left[f_\pi^2+O(q^2)\right]. 
258: \end{eqnarray}
259: If we switch on a small quark mass, the pion pole is shifted to its physical value and we have 
260: \cite{chpt}
261: \begin{eqnarray}
262: \Pi^{5}_{\alpha\beta}(q)=
263: \left(g_{\alpha\beta}+\frac{q_\alpha q_\beta}{m_\pi^2-q^2}\right)\left[f_\pi^2+O(q^2)\right]+O(m^2). 
264: \end{eqnarray}
265: The shift of the pion pole gives the only effect linear in the quark masses, all other
266: corrections are quadratic in $m$. The correlator may be splitted into transverse and longitudinal
267: structures as follows: 
268: \begin{eqnarray}
269: \label{2.17}
270: \Pi^{5}_{\alpha\beta}(q)=
271: \left(g_{\alpha\beta}-\frac{q_\alpha q_\beta}{q^2}\right)\left[f_\pi^2+O(q^2)+O(m^2)\right]
272: +q_\alpha q_\beta\left[\frac{m_\pi^2f_\pi^2}{q^2(m_\pi^2-q^2)}+O(m^2)\right].
273: \end{eqnarray}
274: As soon as we are not in the precise chiral limit, the pion pole appears in the 
275: longitudinal structure of the correlator, corresponding to $J^P=0^-$. Nevertheless, 
276: the transverse part of the correlator ``remembers'' the Goldstone nature of the pion and 
277: contains $f_\pi^2$. Therefore, $f_\pi^2$ can be related to the OPE for the 
278: transverse function $\Pi_T(q^2)$ by a ``transverse'' sum rule, as done in Ref.~\cite{svz}.\footnote{We note 
279: that the excited pseudoscalars do not contribute to the transverse structure, and therefore decay 
280: constants of excited pseudoscalars have no relationship to $\Pi_T$.} 
281: 
282: Notice that the pion contribution to the longitudinal part has a different structure than 
283: the contribution of the excited pseudoscalar states, which has the form 
284: \begin{eqnarray}
285: q_\alpha q_\beta\frac{f_P^2}{m_P^2-q^2}. 
286: \end{eqnarray}
287: In the longitudinal structure of the expression (\ref{2.17}), the pion contribution is the only hadron contribution of order $O(m)$, 
288: and, by virtue of the GMOR relation, it is locally dual to the quark condensate term of the OPE (\ref{piab}) \cite{svz}. 
289: %\newpage
290: \subsection{QCD sum rules for two-point correlators}
291: Let us now summarize the OPE results for the two-point correlators discussed above: 
292: \begin{eqnarray}
293: \label{2.19}
294: \Pi^5(q)&=&\Pi_5(q^2),
295: \nonumber\\ 
296: \Pi^5_\alpha(q)&=&\frac{iq_\alpha}{q^2}
297: \left(\langle\bar u u+\bar dd \rangle+(m_u+m_d)\Pi_5(q^2)\right),
298: \nonumber\\  
299: \Pi^{5}_{\alpha\beta,L}(q)&=&
300: \frac{q_\alpha q_{\beta}}{q^4}
301: \left(
302: (m_u+m_d)\langle\bar uu+\bar dd\rangle +(m_u+m_d)^2\Pi_5(q^2)\right), 
303: \end{eqnarray}
304: where $\Pi^{5}_{\alpha\beta,L}$ is the longitudinal part of the correlator 
305: $\Pi^{5}_{\alpha\beta}$. 
306: The corresponding hadron saturation for these correlators reads 
307: \begin{eqnarray}
308: \label{2.20}
309: \Pi^5(q)&=& \left(\frac{f_\pi m_\pi^2}{m_u+m_d}\right)^2\frac{1}{m_\pi^2-q^2}
310: +\left(\frac{f_P m_P^2}{m_u+m_d}\right)^2\frac{1}{m_P^2-q^2}+\cdots,
311: \nonumber\\ 
312: \Pi^5_\alpha(q)&=&
313: {iq_\alpha}\left(
314: \frac{f^2_\pi m_\pi^2}{m_u+m_d}\frac{1}{m_\pi^2-q^2} 
315: +
316: \frac{f^2_P m_P^2}{m_u+m_d}\frac{1}{m_P^2-q^2}+\cdots 
317: \right),
318: \nonumber\\ 
319: \Pi^{5}_{\alpha\beta,L}(q)&=& {q_\alpha q_\beta}
320: \left(
321: \frac{f^2_\pi m_\pi^2}{q^2(m_\pi^2-q^2)} 
322: +
323: \frac{f^2_P}{m_P^2-q^2}+\cdots
324: \right),  
325: \end{eqnarray}
326: where dots indicate the contribution of the hadron continuum states. 
327: 
328: The representations (\ref{2.19}) and (\ref{2.20}) for $\Pi^5$, $\Pi^5_\alpha$, 
329: and $\Pi^5_{\alpha\beta,L}$ lead to a single relation between the imaginary parts: 
330: \begin{eqnarray}
331: \frac{f^2_\pi m_\pi^4}{(m_u+m_d)^2}\pi\delta(s-m_\pi^2)+
332: \frac{f^2_P m_P^4}{(m_u+m_d)^2}\pi\delta(s-m_P^2)+\cdots %{\rm  continuum}
333: \simeq
334: {\rm Im}\, \Pi_5(s). 
335: \end{eqnarray}
336: The sign $\simeq$ means that both sides are equal to each other after application of an 
337: appropriate smearing, which gives a ``longitudinal'' sum rule. 
338: Taking into account the GMOR relation and performing a Borel transform %\cite{khodjamirian} 
339: of the equation above (with the Borel parameter $M^2$), we find 
340: \begin{eqnarray}
341: \label{sr2}
342: \frac{\langle \bar uu+\bar dd\rangle^2}{f_\pi^2}\exp\left(-\frac{m_\pi^2}{M^2}\right)+
343: \frac{f^2_P m_P^4}{(m_u+m_d)^2}\exp\left(-\frac{m_P^2}{M^2}\right)=
344: \frac{1}{\pi}\int\limits_{(m_u+m_d)^2}^{s_0} ds\,\exp\left(-\frac{s}{M^2}\right){\rm Im}\, \Pi_5(s), 
345: \end{eqnarray}
346: where $s_0$ denotes the continuum subtraction point. 
347: 
348: It is not our goal to perform here a detailed analysis of this sum rule. We would rather like to 
349: point out the sensitivity of this sum rule to the contribution of the excited pseudoscalar state. 
350: %check that reasonable numbers result from this sum rule. 
351: If we keep, on the hadron side, only the pion pole and neglect, on the QCD side, 
352: higher-order (both radiative and power) corrections 
353: and terms which vanish in the chiral limit, we obtain, after Borel transform,  
354: \begin{eqnarray}
355: \frac{4\langle \bar \psi\psi\rangle^2}{f_\pi^2}\exp\left(-\frac{m_\pi^2}{M^2}\right)=
356: \frac{3}{8\pi^2}\int_{0}^{s_0} {ds}\, s\,\exp\left(-\frac{s}{M^2}\right)+\frac{1}{8\pi}
357: {\langle \alpha_s GG\rangle},  
358: \end{eqnarray}
359: where we assume that $\langle \bar uu\rangle=\langle \bar dd\rangle$ and denote this quantity as 
360: $\langle \bar \psi\psi\rangle$. 
361: Taking into account the known values of the condensates and of $f_\pi$, we can
362: neglect the $\langle GG\rangle$ term on the r.h.s. and arrive at the sum rule
363: \begin{eqnarray}
364: \frac{4\langle \bar \psi\psi\rangle^2}{f_\pi^2}=
365: \frac{3}{8\pi^2}\int_{0}^{s_0} {ds}\, s\,\exp\left(\frac{m_\pi^2-s}{M^2}\right), 
366: \end{eqnarray}
367: For $s_0\simeq 1\div 2$ GeV we have an approximate relation:
368: \begin{eqnarray}
369: \frac{4\langle \bar \psi\psi\rangle^2}{f_\pi^2}\simeq
370: \frac{3}{8\pi^2}{M^4}. 
371: \end{eqnarray}
372: To obtain $f_\pi=0.13$ GeV for 
373: $\langle \bar\psi\psi\rangle=-(0.22\div 0.25\, {\rm GeV})^3$ requires $M=0.9\div 1.1$ GeV.
374: % a rather reasonable value for our crude estimates. 
375: On the other hand, the contribution of the excited pseudoscalar meson ${\pi'}(1300)$ to 
376: the sum rule (\ref{sr2}) is sizeable for a Borel mass $M$ around $M\simeq 1$ GeV. Thus the sum rule (\ref{sr2}) provides an  
377: interesting possibility to study the decay constant $f_{\pi'}$ of the excited pseudoscalar meson ${\pi'}(1300)$. 
378: 
379: %\newpage
380: %
381: %
382: \section{\label{iii}Three-point functions and the pion form factor}
383: In this section, we discuss properties and duality relations for three-point correlators,  
384: laying our main emphysis on $\langle PVP\rangle$ and $\langle AVP\rangle$. Nonperturbative power corrections 
385: play a crucial 
386: role in the OPE for these correlators: for $\langle PVP\rangle$ power corrections dominate the correlator 
387: at large momentum transfers, and for $\langle AVP\rangle$ only nonperturbative terms contribute in the
388: chiral limit. This leads to subtleties in extracting the single-hadron contribution, in particular, 
389: the pion form factor, from these correlators. 
390: 
391: 
392: \subsection{\boldmath $\langle PVP\rangle$}
393: %{\boldmath $\langle T j^5 j_\mu j^5\rangle$}
394: Let us start with the correlator 
395: \begin{eqnarray}
396: \Gamma_{\mu}(q, p,p')\equiv i^2\int d^4x\, d^4y\, e^{iqx-ipy}\langle T (j^5(y) j_\mu(x) j^{5\dagger}(0))\rangle,\qquad p'=p-q,  
397: \end{eqnarray}
398: where $j_\mu$ is the electromagnetic current 
399: \begin{eqnarray}
400: j_\mu=\frac23\bar u\gamma_\mu u-\frac13\bar d\gamma_\mu d. 
401: \end{eqnarray}
402: The conservation of the electromagnetic current and the commutation relation  
403: \begin{eqnarray}
404: [j_0(0,\vec x),\bar d \gamma_5 u(0)]&=&-\bar d \gamma_5 u(0)\delta(\vec x) 
405: \end{eqnarray}
406: lead to the Ward identity 
407: \begin{eqnarray}
408: q^\mu\Gamma_\mu(q,p,p')=\Pi^5(p')-\Pi^5(p). 
409: \end{eqnarray}
410: This relation determines the longitudinal structure of $\Gamma_\mu$, so we can write
411: \begin{eqnarray}
412: \Gamma_\mu(q,p,p')=2\left(p_\mu-\frac{qp}{q^2}q_\mu\right)\Gamma(q^2,p^2,p'^2)
413: +\frac{q_\mu}{q^2}\left(\Pi^5(p')-\Pi^5(p)\right). 
414: \end{eqnarray}
415: Since there is no massless hadron state in the vector channel, the function 
416: $\Gamma_\mu$ is regular for $q^2\to 0$. This leads to the relation, valid for all 
417: values of $p^2$ and $p'^2$, 
418: \begin{eqnarray}
419: \label{3.7}
420: \Gamma(0,p^2,p'^2)=-\frac{\Pi^5(p')-\Pi^5(p)}{p'^2-p^2}.
421: \end{eqnarray}
422: %This relation is valid for all values of $p^2$ and $p'^2$. 
423: 
424: Let us set the quark masses equal to zero. 
425: At small $p^2$ and $p'^2$, the Goldstones dominate the correlators, leading to a single pole in $\Pi^5$,   
426: \begin{eqnarray}
427: \Pi^5(p)=-\frac{4\langle \bar \psi\psi\rangle^2}{f_\pi^2 p^2}, 
428: \end{eqnarray}
429: and to a double pole in $\Gamma(q^2,p^2,p'^2)$, 
430: \begin{eqnarray}
431: \Gamma(0,p^2,p'^2)=-\frac{F_\pi(0)}{p^2p'^2}\frac{4\langle \bar \psi\psi\rangle^2}{f_\pi^2},
432: \end{eqnarray}
433: with the pion electromagnetic form factor $F_\pi(q^2)$ defined by 
434: \begin{eqnarray}
435: \langle \pi(p')|j_\mu(0)|\pi(p)\rangle=(p+p')_\mu{F_\pi(q^2)}.
436: \end{eqnarray}
437: The Ward identity (\ref{3.7}) then implies the normalization of the pion form factor at $q^2=0$: 
438: \begin{eqnarray}
439: {F_\pi(q^2=0)}=1.
440: \end{eqnarray}
441: At large values of $p^2$, $p'^2$, and $q^2$, we have the following two 
442: representations for $\Gamma(q^2,p^2,p'^2)$, namely, a hadronic one and a QCD one: 
443: \begin{eqnarray}
444: \label{pvp}
445: \Gamma^{\rm hadr}(q^2,p^2,p'^2)&=&-\frac{4\langle \bar \psi\psi\rangle^2}{f_\pi^2}\frac1{p^2p'^2}F_\pi(q^2)
446: +\cdots \nonumber\\
447: \Gamma^{\rm QCD}(q^2,p^2,p'^2)&=&
448: \Gamma_{\rm pert}(q^2,p^2,p'^2)+\Gamma_{\rm inst}(q^2,p^2,p'^2)+\Gamma_{\rm cond}(q^2,p^2,p'^2). 
449: \end{eqnarray}
450: where the dots in $\Gamma^{\rm hadr}$ represent double-pole contributions of pseudoscalars 
451: ($\pi$-$P$ and $P$-$P$), mixed pole--continuum and double--continuum contributions. 
452: On the QCD side, one has an instanton contribution $\Gamma_{\rm inst}$ \cite{shuryak} 
453: in addition to the perturbative part $\Gamma_{\rm pert}$ and the condensate part $\Gamma_{\rm cond}$. 
454: Quark--hadron duality states that $\Gamma^{\rm hadr}$ and $\Gamma^{\rm QCD}$ should be equal to each 
455: other after a proper smearing is applied. 
456: For large values of $q^2$ the pion double pole behaves like  
457: \begin{eqnarray}
458: \label{3.12}
459: -\frac{4\langle \bar \psi\psi\rangle^2}{f_\pi^2}\frac{F_\pi(q^2)}{p^2p'^2}\propto 
460: \frac{\alpha_s\langle \bar \psi\psi\rangle^2}{q^2p^2p'^2}. 
461: \end{eqnarray}
462: On the QCD side the double spectral density of $\Gamma_{\rm pert}$, including $\alpha_s$ corrections,  
463: was found to have the leading behavior $1/q^4$ at large $q^2$ \cite{braguta}. 
464: The absence of the term ${\alpha_s}/{q^2}$ in the perturbative diagrams 
465: led the authors of Ref.~\cite{braguta} to the conclusion that the behavior of the pion 
466: form factor as extracted from the $\langle PVP \rangle$ correlator 
467: should not reproduce the correct pQCD asymptotics because of the ``wrong'' twist of the 
468: pseudoscalar current. However, this argument cannot be correct: the correlator $\langle PVP \rangle$ 
469: has a nonzero overlap with the pion double pole, and 
470: the residue in this double pole should reproduce the full pion form factor, behaving according to 
471: pQCD at large $q^2$. As we shall demonstrate, the solution to this puzzle is simple: 
472: the condensates dominate the OPE for $\Gamma^{\rm QCD}(q^2,p^2,p'^2)$ at large $q^2$ and provide the 
473: necesssary $\alpha_s/q^2$ terms.  
474: %\footnote{Notice that even if the term behaving as $\alpha_s/q^2$ had been absent in OPE, it would not
475: %have meant that the pion form factor at large $q^2$ as extracted from $\langle PVP\rangle$ has a behaviour different
476: %from pQCD; it would have just meant that there is a conspiracy between different hadronic states 
477: %leading to the compltete cancellation of the $\alpha_s/q^2$ terms in the correlator.}
478: The odd-dimensional operators $\bar \psi\psi$ and $\bar \psi\sigma G\psi$ 
479: are irrelevant for the effect under discussion as their contribution vanishes in the chiral limit:  
480: \begin{eqnarray}
481: \Gamma^{\langle \bar \psi\psi\rangle}_{\rm cond} \propto {m\langle \bar \psi\psi\rangle},\qquad 
482: \Gamma^{\langle  \bar\psi\sigma G \psi\rangle}_{\rm cond} 
483: \propto m \langle\bar \psi\sigma G \psi\rangle.
484: \end{eqnarray}
485: The contribution which is essential in the chiral limit comes from the even-dimensional operators 
486: $GG$ and $\bar\psi \psi \bar\psi \psi$. Since the pion double pole (\ref{3.12}) contains the factor 
487: $\langle\bar \psi\psi\rangle^2$, one can expect the operator 
488: $\bar\psi \psi \bar\psi \psi$ to provide the contribution of interest. For dimensional reasons, its contribution 
489: may contain the structures
490: \begin{eqnarray}
491: \Gamma_{\rm cond}^{\langle\bar \psi\psi\rangle^2}\propto
492: \alpha_s\langle\bar \psi\psi\rangle^2
493: \left\{
494: \frac{1}{p^2p'^2q^2},\frac{1}{p^2q^4}+\frac{1}{p'^2q^4},
495: \frac{1}{p^4p'^2}+\frac{1}{p'^4p^2}, \frac{q^2}{p^4p'^4}, \cdots\right\}.
496: \end{eqnarray}
497: The explicit calculation of all coefficients is a cumbersome task \cite{ioffe}. 
498: Of particular interest for us is, however, the term ${1}/{p^2p'^2q^2}$. We find 
499: that only the two diagrams with hard gluon exchange shown in Fig.~\ref{fig:2.2} give rise to this structure. 
500: \begin{figure}[t]
501: \begin{center}
502: \epsfig{file=fig2_pion.eps,height=3.0cm}
503: \caption{\label{fig:2.2} 
504: $\Gamma^{\langle\bar\psi\psi\rangle^2}_{\rm cond}$: diagrams leading to 
505: the structure $\propto \alpha_s\langle\psi\psi\rangle^2/(q^2p^2p'^2)$.}
506: \end{center}
507: \end{figure}
508: Using the factorization formula for the four-quark condensate \cite{khodjamirian}, this contribution 
509: is easily calculable, leading to 
510: \begin{eqnarray}
511: \label{qqq}
512: \Gamma_{\rm cond}^{\langle\bar \psi\psi\rangle^2}=\frac{8}{9}
513: \alpha_s\langle\bar \psi\psi\rangle^2\frac{1}{p^2p'^2q^2}.
514: \end{eqnarray}
515: This result is sufficient for our argument and we do not need the explicit $GG$ contribution: 
516: because of its dimension, the operator $GG$ leads to a constant term, whereas we are interested 
517: in $1/q^2$ terms. 
518:  
519: Now, notice that the perturbative term is power suppressed at large $q^2$ compared to the 
520: nonperturbative contribution, which dominates the correlator. In order to use the sum rule for the 
521: analysis of the form factor at large $q^2$, precisely following Ref.~\cite{radyushkin} we doubly Borelize the
522: function $\Gamma(q^2,p^2,p'^2)$ in the variables $p^2$ and $p'^2$, choose the same Borel parameter 
523: $M^2$ for both channels, and send $M^2\to \infty$. 
524: 
525: The contribution of order $1/q^2$ which survives on the QCD side is the one given by Eq.~(\ref{qqq}).  
526: Other structures in $\Gamma^{\langle\bar\psi\psi\rangle^2}_{\rm cond}$ are either 
527: killed by the double Borel transform or, similar to $\Gamma_{\rm pert}$, have a $1/q^4$ behavior.   
528: 
529: At order $1/q^2$ the sum rule takes the form
530: \begin{eqnarray}
531: \label{ass}
532: -4\frac{\langle\bar \psi\psi\rangle^2}{f_\pi^2} F_\pi(q^2)+\cdots 
533: =\frac{8}{9}
534: \pi\alpha_s\langle\bar \psi\psi\rangle^2\frac{1}{q^2}.
535: \end{eqnarray}
536: Clearly, we have the necessary analytic structure on the QCD side, 
537: but the coefficient is different from the known pQCD result \cite{pqcd}
538: \begin{eqnarray}
539: \label{pqcd}
540: F_\pi(q^2)=8\pi \alpha_s f_\pi^2/(-q^2). 
541: \end{eqnarray}
542: This does not seem to signal an error in our calculation: on the hadronic side there 
543: are contributions coming from other states which are not suppressed 
544: (since we have set $M^2\to\infty$). 
545: So we do not see a possibility to isolate the pion contribution and thus Eq.~(\ref{ass}) 
546: is a genuine sum rule which relates the sum of several hadron contributions to a single term of the OPE.  
547: 
548: To summarize, we do not find any disagreement between the large-$q^2$ behavior of the pion form factor 
549: and the OPE for $\langle PVP\rangle$. Of interest for us is that, unlike the case of the two-point correlators, 
550: in the chiral limit there is no local duality between the pion double pole and separate 
551: double-pole terms in the OPE. 
552: 
553: We would like to add one more comment on the interesting finding of Ref.~\cite{braguta} on the absence of the
554: $\alpha_s/q^2$ term in the double spectral density of the quark triangle diagram. This result may 
555: pose difficulties for the analysis of the large-$q^2$ behavior of the pion form factor within the constituent
556: quark model \cite{anis} since this model makes use of the $\bar q \gamma_5 q$ structure for the description of the
557: pseudoscalar meson.
558: %\footnote{Likewise, the observed cancellation of the double logarithmic terms 
559: %poses potential problem for the so-called Sudakov suppression of the soft form factor. 
560: %This interpretation may originate from improper resummation of the radiative corrections to 
561: %the three-point functions.} 
562: 
563: 
564: 
565: 
566: 
567: 
568: 
569: 
570: 
571: 
572: 
573: 
574: 
575: 
576: \subsection{\boldmath $\langle AVP\rangle$}
577: %{\boldmath $\langle T j^5_\alpha j_\mu j^5\rangle$}
578: Next, we discuss the correlator 
579: \begin{eqnarray}
580: \Gamma_{\mu\alpha}(q,p,p')\equiv i^2\int d^4x\, d^4y\, e^{iqx-ipy}\langle T( j^5_\alpha(y) j_\mu(x) j^{5\dagger}(0))\rangle, 
581: \qquad p'=p-q.   
582: \end{eqnarray} 
583: %$j_\mu$ is the conserved electromagnetic current
584: %\begin{eqnarray}
585: %\label{em}
586: %j_\mu=\frac23\bar u\gamma_\mu u-\frac13\bar d\gamma_\mu d,  \qquad \partial_\mu j_\mu=0. 
587: %\end{eqnarray}
588: Let us calculate 
589: $q^\mu\Gamma^5_{\mu\alpha}(q,p,p')$ and $p^\alpha\Gamma^5_{\mu\alpha}(q,p,p')$.
590: We make use of the divergence equations for $j^5_\alpha$ and $j_\mu$, and the commutation 
591: relations  
592: \begin{eqnarray}
593: [j_0(0,\vec x),\bar u\gamma_\alpha \gamma_5 d(0)]&=&\bar u\gamma_\alpha \gamma_5 d(0)\delta(\vec x),
594: %\end{eqnarray}
595: \nonumber\\
596: %\begin{eqnarray}
597: {}[j_0(0,\vec x),\bar d \gamma_5 u(0)]&=&-\bar d \gamma_5 u(0)\delta(\vec x),
598: \nonumber\\
599: {}[\bar u\gamma_0\gamma_5d(0,\vec x),\bar d\gamma_5 u(0)]&=&-(\bar u u+\bar dd)\delta(\vec x). 
600: \end{eqnarray}
601: Taking into account the relation 
602: \begin{eqnarray}
603: \int d^4x\, e^{iqx}\langle T(j_\mu(x)(\bar uu+\bar dd)\rangle=0 
604: \end{eqnarray}
605: in the evaluation of $p^\alpha\Gamma^5_{\mu\alpha}$, 
606: we find the following Ward identities: 
607: \begin{eqnarray}
608: q^\mu\Gamma_{\mu\alpha}(q,p,p')&=&\Pi_\alpha^5(p')-\Pi_\alpha^5(p),
609: \nonumber\\
610: p^\alpha\Gamma_{\mu\alpha}(q,p,p')&=&\Pi_\mu^5(p')+(m_u+m_d)\Gamma_{\mu}(p,p',q). 
611: \end{eqnarray}
612: In the chiral limit, by virtue of Eq.~(\ref{pi5alpha}), these Ward identites take the form 
613: \begin{eqnarray}
614: \label{wi_avp}
615: q^\mu\Gamma_{\mu\alpha}(q,p,p')&=&2i\langle \bar\psi\psi\rangle 
616: \left( \frac{p'_\alpha}{p'^2}- \frac{p_\alpha}{p^2} \right), 
617: \nonumber\\
618: p^\alpha\Gamma_{\mu\alpha}(q,p,p')&=&2i\langle \bar\psi\psi\rangle\frac{p'_\mu}{p'^2}. 
619: \end{eqnarray}
620: We may split $\Gamma_{\mu\alpha}$ into transverse and longitudinal parts, 
621: \begin{eqnarray}
622: \label{3.21}
623: \Gamma_{\mu\alpha}(q,p,p')=\Gamma^\perp_{\mu\alpha}(q,p,p')+\Gamma^L_{\mu\alpha}(q,p,p'), 
624: \end{eqnarray}
625: where $q^\mu \Gamma^\perp_{\mu\alpha}=0$, $p^\alpha \Gamma^\perp_{\mu\alpha}=0$. Making use of 
626: the Ward identities (\ref{wi_avp}), we determine the longitudinal part $\Gamma^L_{\mu\alpha}$ 
627: in the form 
628: \begin{eqnarray}
629: \label{3.24}
630: \Gamma^L_{\mu\alpha}(q,p,p')=2i\langle \bar\psi\psi\rangle 
631: \left[\frac{q_\mu}{q^2}\left(\frac{p'_\alpha}{p'^2}- \frac{p_\alpha}{p^2} \right)
632: +\left(g_{\mu\alpha}-\frac{q_\mu q_\alpha}{q^2}\right)\frac{1}{p'^2}
633: \right].
634: \end{eqnarray}
635: The transverse part $\Gamma^\perp_{\mu\alpha}$ may be parametrized by two invariant amplitudes: 
636: \begin{eqnarray}
637: \label{3.25}
638: \Gamma^\perp_{\mu\alpha}(q,p,p')=\left(g_{\mu\alpha}-\frac{q_\mu p_\alpha}{qp}\right)i F_1(q^2,p^2,p'^2)
639: +\left(p_{\mu}-\frac{qp}{q^2}q_\mu\right)\left(q_{\alpha}-\frac{qp}{p^2}p_\alpha\right)i F_2(q^2,p^2,p'^2).
640: \end{eqnarray}
641: The terms in $\Gamma^5_{\mu\alpha}$ singular at $q^2=0$ should cancel each other since there is no massless 
642: hadron state in the $q^2$-channel. This leads to the relation 
643: \begin{eqnarray}
644: \label{wi22}
645: F_2(0,p^2,p'^2)=-\frac{8\langle \bar\psi\psi\rangle}{p'^2(p^2-p'^2)},  
646: \end{eqnarray}
647: valid for all values of $p^2$ and $p'^2$. 
648: 
649: Let us look at the structure $p_\mu p_\alpha$ in $\Gamma_{\mu\alpha}$ at $q^2=0$. 
650: Due to the Ward identity (\ref{wi22}), Eqs. (\ref{3.24}) and (\ref{3.25}) lead to the 
651: following expression for this structure: 
652: \begin{eqnarray}
653: \label{3.27}
654: \frac{4\langle \bar\psi\psi\rangle}{p^2p'^2}p_\mu p_\alpha. 
655: \end{eqnarray}
656: At small values of $p^2$ and $p'^2$, and for all values of $q^2$, the Goldstone pole dominates 
657: the correlator, leading to the following $p_\mu p_\alpha$ term:
658: \begin{eqnarray}
659: \label{3.28}
660: \frac{4\langle \bar\psi\psi\rangle}{p^2p'^2}F_\pi(q^2)p_\mu p_\alpha. 
661: \end{eqnarray}
662: Comparing Eq.~(\ref{3.28}) at $q^2=0$ with Eq.~(\ref{3.27}), we see that the Ward identity guarantees the correct normalization of the 
663: pion form factor $F_\pi(0)=1$. 
664: 
665: The other region where rigorous relations for the correlator may be obtained is the region of large 
666: $q^2$, $p^2$, $p'^2$. Here the OPE may be applied. Perturbative diagrams, being $O(m)$, do not contribute 
667: to the correlator in the chiral limit. 
668: So the leading contribution to the OPE series for $\Gamma_{\mu\alpha}$ comes from the 
669: $\bar \psi\psi$ operator. 
670: This contribution is given by the triangle diagrams with quark-condensate insertions 
671: in one of the quark lines in the loop. The result reads 
672: \begin{eqnarray}
673: \Gamma^{\langle \bar\psi\psi\rangle}_{\mu\alpha}(q,p,p')=
674: i{\langle \bar\psi\psi\rangle}
675: \left(
676: \frac{p_\mu p'_\alpha+p'_\mu p_\alpha-g_{\alpha\mu}pp'}{p^2 p'^2}
677: +\frac{q_\mu p'_\alpha+p'_\mu q_\alpha-g_{\alpha\mu}qp'}{q^2 p'^2}
678: +\frac{p_\mu q_\alpha-q_\mu p_\alpha-g_{\alpha\mu}qp}{p^2 q^2}\right).
679: \end{eqnarray}
680: This expression fully saturates the chiral Ward identites (\ref{wi_avp}). 
681: 
682: Let us now study which operator in the OPE corresponds to the pion contribution at large $q^2$. 
683: To this end, we look at the Lorentz structure $p_\mu p_\alpha$, perform the double Borel 
684: transform of $\Gamma_{\mu\alpha}$, and take the limit $M^2\to\infty$. After this procedure, 
685: the pion double pole takes the form 
686: \begin{eqnarray}
687: 4F_\pi(q^2)\langle\bar\psi\psi\rangle p_\mu p_\alpha.
688: \end{eqnarray}
689: Taking into account the $\alpha_s/q^2$ behavior of the pion form factor, 
690: we now look for the term on the OPE side which, after the transformations described above,  
691: behaves as $1/q^2$. Only the dimension-5 operator $\bar \psi\sigma G\psi$ can lead to such a 
692: structure; the operator $\bar\psi\psi$ leads to the constant term 
693: \begin{eqnarray}
694: 2\langle\bar\psi\psi\rangle p_\mu p_\alpha, 
695: \end{eqnarray} 
696: which is relevant for continuum states, and not for the pion. 
697: The contribution of the $\bar \psi\sigma G\psi$ operator to the correlator has been found in Ref.~\cite{moussallam}: 
698: \begin{eqnarray}
699: \frac13 \langle\bar \psi\sigma G\psi \rangle \left(\frac{1}{q^2p'^4}-\frac{q^2}{p^4p'^4}\right)p_\mu p_\alpha.
700: \end{eqnarray}
701: This structure does not contribute after performing the double Borel transform and sending $M^2\to\infty$. 
702: The radiative corrections to the Wilson coefficient have not been calculated. Nevertheless, the term in the OPE series  
703: \begin{eqnarray}
704: \frac{C\alpha_s}{q^2 p^2 p'^2}\langle\bar \psi\sigma G\psi \rangle p_\mu p_\alpha
705: \end{eqnarray}
706: leads to 
707: \begin{eqnarray}
708: \frac{C\alpha_s}{q^2}\langle\bar \psi\sigma G\psi \rangle p_\mu p_\alpha. 
709: \end{eqnarray}
710: All other possible kinematic structures of the relevant dimension vanish in the limit $M^2\to\infty$. Therefore, 
711: the duality relation takes the form  
712: \begin{eqnarray}
713: 4F_\pi(q^2)\langle\bar\psi\psi\rangle+\cdots=
714: C \frac{\alpha_s}{q^2}\langle\bar \psi\sigma G\psi \rangle, 
715: \end{eqnarray}
716: with the dots standing for contributions of other hadronic states (such as 
717: pseudoscalar--axial meson transition form factors, etc.). We do not see a possibility to isolate 
718: the pion term. Nevertheless, it is interesting that, while at small values of $q^2$ the pion 
719: pole is due to Ward identites ``dual'' to $\langle\bar \psi\psi \rangle$, at large values of 
720: $q^2$ it contributes to the sum rule for a different operator, viz., 
721: $\langle\bar \psi\sigma G\psi \rangle$. 
722: 
723: 
724: 
725: 
726: 
727: 
728: 
729: 
730: 
731: 
732: 
733: \subsection{\boldmath $\langle AVA\rangle$}
734: The extraction of the pion form factor from the correlator 
735: \begin{eqnarray}
736: \Gamma_{\mu\alpha\beta}(q,p,p')\equiv 
737: i^2\int d^4x\, d^4y\, e^{ip'z-ipy}\langle T (j^5_\alpha(y) j_\mu(0) j^{5\dagger}_\beta(z))\rangle, \qquad q=p-p', 
738: \end{eqnarray}
739: is known quite well \cite{ioffe,radyushkin,bakulev}. 
740: We would only like to point out the following features of this case. 
741: %1. In the chiral limit, the excited pseudoscalars decouple from the axial current 
742: %and do not appear in the set of intermediate states; 
743: To consider the pion form factor at large-$q^2$ one has to proceed along the lines of 
744: Ref.~\cite{radyushkin} which we followed above: namely, to perform the double Borel transform and to take $M^2\to\infty$. 
745: Then the condensate contributions vanish and one obtains the ``local duality'' representation for the form factor 
746: \cite{radyushkin}
747: \begin{eqnarray}
748: \label{ld_ff}
749: f_\pi^2 F_\pi(q^2)=\int_0^{s_0} ds_1\int_0^{s_0} ds_2 \;\Delta_{\rm pert}(s_1,s_2,q^2), 
750: \end{eqnarray}
751: where $s_0$ is the pion duality interval and $\Delta_{\rm pert}$ is the double spectral density of the 
752: $p_\mu p_\alpha p_\beta$ structure of $\Gamma_{\mu\alpha\beta}$. Similarly, one has the local duality representation 
753: for the decay constant 
754: \begin{eqnarray}
755: \label{fpi1}
756: f_\pi^2=\int_0^{s_0} ds \;\rho_{\rm pert}(s), 
757: \end{eqnarray}
758: where $\rho_{\rm pert}(s)=(1+\alpha_s/\pi)/{4\pi^2}$ is the spectral density of the transverse invariant
759: amplitude $\Pi_T(q^2)$ (\ref{piT}). Eq. (\ref{fpi1}) yields $s_0=4\pi^2f_\pi^2/(1+\alpha_s/\pi)$. 
760: Due to the vector Ward identity for $\Gamma_{\mu\alpha\beta}$, one has the relation \cite{ich}
761: \begin{eqnarray}
762: \lim_{q^2\to 0}\Delta_{\rm pert}(s_1,s_2,q^2)=\delta(s_1-s_2)\rho_{\rm pert}(s).  
763: \end{eqnarray}
764: Clearly, in $\Delta_{\rm pert}$ and $\rho_{\rm pert}$ radiative corrections up to the same order should be included. 
765: Recently, the calculation of the $O(\alpha_s)$ contributions to $\Delta_{\rm pert}$ has been reported \cite{braguta2}. 
766: For large negative $q^2$ and at fixed $s_1$ and $s_2$, the spectral density was found to behave like  
767: \begin{eqnarray}
768: \Delta_{\rm pert}(s_1,s_2,q^2)\to -\frac{\alpha_s}{2\pi^3 q^2}+O(1/q^4).  
769: \end{eqnarray}
770: Interestingly, substituting this expression into the representation (\ref{ld_ff}) one precisely reproduces 
771: the asymptotic pQCD result (\ref{pqcd}). 
772: 
773: Thus the local duality representation for the form factor has several attractive features:
774: (i) It is applicable for all spacelike momentum transfers. 
775: (ii) At $q^2=0$ the form factor is properly normalized due to the Ward identity. 
776: (iii) At large $q^2$ the form factor behaves in accordance with pQCD. 
777: As a result, the local duality representation with the spectral density which includes $O(\alpha_s)$ corrections 
778: \cite{braguta2} is expected to give reliable predictions for the form factor for all values of $q^2$. 
779: 
780: To summarize, in the $\langle VAV\rangle$ case it is possible to relate the pion form factor for all spacelike $q$ 
781: to a definite part of the perturbative contribution. 
782: 
783: 
784: 
785: %\newpage
786: \section{\label{iv}Summary and results}
787: We analysed 2- and 3-point correlators of the pseudoscalar and axial currents. 
788: Because of the partial conservation of the axial current, the OPE for correlators 
789: of this currents exhibits a rather specific feature: namely, in the chiral limit the OPE 
790: is dominated in many cases by {nonperturbative} power corrections. 
791: 
792: Our results are as follows: 
793: \begin{itemize}
794: \item 
795: A detailed study of the  OPE and duality for the $\langle PP\rangle$, $\langle AA\rangle$, and 
796: $\langle AP\rangle$ correlators was presented. 
797: We pointed out that quark--hadron duality for $\langle AP\rangle$, similar to the longitudinal structure 
798: of $\langle AA\rangle$, has an interesting feature: in the chiral limit the pion contribution turns out 
799: to be dual to a single $\bar \psi\psi$ term in the OPE. 
800: We discussed sum rules for $\langle PP\rangle$, $\langle AA\rangle$, and 
801: $\langle AP\rangle$, which are governed by 
802: the quark condensates $\langle \bar u u\rangle$ and 
803: $\langle \bar dd\rangle$, and the two independent functions 
804: ${\rm Im}\; \Pi_T(q^2)$ and ${\rm Im}\;\Pi_5(q^2)$. 
805: Accordingly, one can obtain two independent QCD sum rules: 
806: The first sum rule, for the transverse structure of the $\langle AA \rangle$ correlator, involves 
807: ${\rm Im}\; \Pi_T(q^2)$. On the hadronic side, it includes the contributions of axial-meson states, 
808: but contains also $f^2_\pi$ as the reminder of the Goldstone nature of the pion. 
809: This sum rule was used for the extraction of $f^2_\pi$ in \cite{svz}. 
810: The second sum rule, for the $\langle PP\rangle$ correlator, involves ${\rm Im}\; \Pi_5(q^2)$. 
811: On the hadronic side, it contains contributions of the ground-state and excited light pseudoscalars. 
812: According to our estimates, this sum rule receives a sizeable contribution from the excited 
813: $\pi'(1300)$ and provides a promising possibility to extract its decay constant $f_{\pi'}$. 
814: 
815: \item
816: We studied properties of the 3-point correlators $\langle PVP\rangle$ and $\langle AVP\rangle$. 
817: We derived the Ward identites for 
818: these correlators and demonstrated the way the normalization of the pion form factor at zero momentum transfer  
819: arises due to these Ward identites. We then analysed the region of large momentum transfers and  
820: the way quark--hadron duality works in these cases.  
821: We proved that the OPE for $\langle PVP\rangle$ and $\langle AVP\rangle$ at large $q^2$ is
822: dominated by nonperturbative corrections, and identified the operators responsible for 
823: providing the correct large-$q^2$ asymptotics of the pion form factor in accordance with pQCD: 
824: the four-quark condensate $\langle\bar\psi\psi\bar\psi\psi\rangle$ in the case of $\langle PVP\rangle$, 
825: and the mixed condensate 
826: $\langle\bar\psi\sigma G\psi\rangle$ in the case of $\langle AVP\rangle$. 
827: We have thus disproved the recent statement in the literature that 
828: the pion form factor as extracted from the $\langle PVP\rangle$ correlator does not have 
829: the right asymptotics required by pQCD. 
830: 
831: \item
832: For the $\langle AVA \rangle$ correlator, we pointed out that the  
833: local duality representation for the pion form factor with the double spectral density, which
834: includes the radiative corrections, is applicable for all spacelike momentum transfers and has the following 
835: interesting features: At $q^2=0$ the form factor is properly normalized due to the vector Ward identity, 
836: and at large $q^2<0$ it reproduces the pQCD asymptotic behavior. Therefore, this parameter-free representation 
837: is expected to give reliable predictions for all spacelike momentum transfers. 
838: \end{itemize}
839: %\newpage
840: \acknowledgments
841: We are grateful to G.~Ecker and H.~Neufeld 
842: for clarification of some aspects of the chiral expansion, and to A.~Bakulev, R.~Bertlmann, 
843: V.~Lubicz, M.~Neubert, O.~Nachtmann, S.~Simula, and B.~Stech for interesting discussions and remarks. 
844: D.~M. was supported by the Austrian Science Fund (FWF) under project P17692. 
845: 
846: \begin{thebibliography}{99} 
847: \bibitem{svz}M.~A.~Shifman, A.~I.~Vainshtein, V.~I.~Zakharov, Nucl.~Phys.~B {\bf 147}, 385 (1979).  
848: \bibitem{ioffe}B.~L.~Ioffe, A.~V.~Smilga, Phys.~Lett.~B {\bf 114}, 353 (1982). 
849: \bibitem{radyushkin} V.~A.~Nesterenko, A.~V.~Radyushkin, Phys.~Lett.~B {\bf 115}, 410 (1982). 
850: \bibitem{bound-state}W.~Lucha, F.~F.~Schoberl, D.~Gromes, Phys.~Rep. {\bf 200}, 127 (1991); 
851: M.~A.~Ivanov, Yu.~L.~Kalinovsky, C.~D.~Roberts, Phys.~Rev.~D {\bf 60}, 034018 (1999).  
852: \bibitem{lattice}C.~R.~Allton et al., Phys.~Rev.~D {\bf 49}, 474 (1994). 
853: \bibitem{wilson} K.~G.~Wilson, Phys.~Rev. {\bf 179}, 1499 (1969). 
854: \bibitem{duality}M.~Shifman, {\it Quark hadron duality}, hep-ph/0009131 and references therein. 
855: \bibitem{krassnigg}A.~Holl, A.~Krassnigg, C.~D.~Roberts, Phys.~Rev.~C {\bf 70}, 042203 (2004). 
856: %
857: \bibitem{gmor}M.~Gell-Mann, R.~J.~Oakes, B.~Renner, Phys.~Rev. {\bf 175}, 2195 (1968). 
858: \bibitem{chpt}J.~Gasser, H.~Leutwyler, Ann.~Phys.~(NY) {\bf 158}, 142 (1984). 
859: \bibitem{khodjamirian}P.~Colangelo, A.~Khodjamirian, 
860: {\it QCD sum rules, a modern perspective}, hep-ph/0010175. 
861: \bibitem{moussallam}B.~Moussallam, Nucl.~Phys.~B {\bf 504}, 381 (1997). 
862: \bibitem{shuryak}P.~Faccioli, A.~Schwenk, E.~V.~Shuryak, Phys.~Rev.~D {\bf 67}, 113009 (2003). 
863: \bibitem{braguta}V.~V.~Braguta, A.~I.~Onishchenko, Phys.~Lett.~B {\bf 591}, 255 (2004). 
864: \bibitem{pqcd}G.~P.~Lepage, S.~J.~Brodsky, Phys.~Lett.~B {\bf 87}, 359 (1979);  
865: A.~V.~Efremov, A.~V.~Radyushkin, Phys.~Lett.~B {\bf 94}, 245 (1980). 
866: \bibitem{anis} V.~V.~Anisovich, D.~I.~Melikhov, V.~A.~Nikonov, Phys.~Rev.~D {\bf 52}, 5295 (1995);
867: Phys.~Rev.~D {\bf 55}, 2918 (1997). 
868: \bibitem{bakulev}A.~P.~Bakulev, A.~V.~Radyushkin, Phys.~Lett.~B {\bf 271}, 223 (1991).  
869: \bibitem{braguta2}V.~V.~Braguta, A.~I.~Onishchenko, Phys.~Lett.~B {\bf 591}, 267 (2004). 
870: \bibitem{ich}V.~V.~Anisovich et al., Nucl.~Phys.~A {\bf 544}, 747 (1992); 
871: D.~Melikhov, S.~Simula, Eur.~Phys.~J.~C {\bf 37}, 437 (2004). 
872: \end{thebibliography}
873: \end{document}
874: 
875: 
876: 
877: 
878: 
879: 
880: 
881: 
882: 
883: 
884: 
885: 
886: 
887: 
888: 
889: 
890: 
891: 
892: 
893: 
894: 
895: 
896: 
897: 
898: 
899: 
900: 
901: 
902: 
903: 
904: 
905: 
906: 
907: 
908: 
909: 
910: 
911: 
912: 
913: 
914: 
915: 
916: 
917: 
918: