1: %\def\be{\begin{equation}}
2: %\def\ee{\end{equation}}
3: %\def\bea{\begin{eqnarray}}
4: %\def\eea{\end{eqnarray}}
5: \def\slashed{\mskip5.0mu{/}\mskip-13.0mu}
6: %\def\spose#1{\hbox to 0pt{#1\hss}}
7: %\def\ltapprox{\mathrel{\spose{\lower 3pt\hbox{$\mathchar"218$}}
8: % \raise 2.0pt\hbox{$\mathchar"13C$}}}
9:
10: \section{$U(1)_A$ problem and large-$N$ limit}
11: \label{U1}
12:
13: \subsection{The $U(1)_A$ problem}
14: \label{ua1pro}
15: The axial $U(1)_A$ problem is a long standing issue in QCD, which can be
16: traced back to the early 1970's and before; it is intimately related to
17: another, as yet unresolved, puzzle: the ``Strong CP problem''.
18:
19: The QCD Lagrangian describing $N_f$ quark flavors in Euclidean space is
20: \begin{equation}
21: {\cal L}= {1\over 4} F_{\mu\nu}^a(x)F_{\mu\nu}^a(x)
22: + \sum_{f=1}^{N_f} \bar\psi_f\,( D + m_f)\,\psi_f \, .
23: \end{equation}
24: where $D$ is the Dirac operator $D\equiv \gamma_\mu (\partial_\mu + g A_\mu)$.
25: In the chiral limit of massless quarks, assuming $N_f>1$, the Lagrangian is
26: invariant under rotations in flavor space, performed independently for left-
27: and right-handed components, resulting in a global group of vector and axial
28: symmetries:
29: \begin{equation}
30: U(N_f)_L \otimes U(N_f)_R \simeq
31: SU(N_f)_V \otimes SU(N_f)_A \otimes U(1)_V \otimes U(1)_A\,.
32: \end{equation}
33: Let us assume $N_f=3$. In the real world, $U(1)_V$ remains intact,
34: corresponding to baryon number conservation, and $SU(3)_V$ is only softly
35: broken by differences in the quark masses, leading to the approximate symmetry
36: of isospin and strangeness. The axial $SU(3)_A$ is spontaneously broken by the
37: dynamical formation of nonzero quark condensates, leading to the octet of the
38: lightest pseudoscalar mesons which, in the chiral limit, are presumed to be
39: the massless Goldstone bosons corresponding to the eight generators of
40: $SU(3)_A$. As for the $U(1)_A$ symmetry, if it were intact in the chiral
41: limit, one would expect all massless hadrons to have a partner of opposite
42: parity. Since this does not appear to be the case, e.g. there are no scalar
43: pions, one may next assume that $U(1)_A$\, too, is spontaneouly broken. In
44: this case, there should be a corresponding isosinglet pseudoscalar Goldstone
45: boson. However, as originally estimated by Weinberg~\cite{Weinberg-75} using
46: chiral perturbation theory, such a particle, away from the chiral limit,
47: should have a mass less than $\sqrt{3} m_\pi$\,; the closest candidates are
48: the mesons $\eta(549)$ (already assigned to the $0^-$ octet) and $\eta'(985)$,
49: whose masses are clearly beyond the Weinberg bound.
50:
51: The resolution of the $U(1)_A$ puzzle was proposed by
52: 't~Hooft~\cite{Hooft-76,Hooft-76b}, who showed that, as a result of instanton
53: effects in the QCD vacuum, $U(1)_A$ is not a true symmetry of the theory.
54: Under $U(1)_A$ transformations:
55: \begin{equation}
56: \psi \to e^{i\alpha\gamma_5}\,\psi ,
57: \label{chiralpsi}
58: \end{equation}
59: the action of massless fermions is invariant, however the path integral
60: measure over fermion fields is modified~\cite{Fujikawa-79}:
61: \begin{equation}
62: [d\psi][d\bar\psi] \to
63: \exp\left({-i\, \alpha \, g^2\,N_f \over 32\pi^2}\int d^4x \,
64: \epsilon_{\mu\nu\rho\sigma} F_{\mu\nu}^a F_{\rho\sigma}^a\right)\,
65: [d\psi][d\bar\psi].
66: \label{fermionmeasure}
67: \end{equation}
68: This noninvariance of the measure gives rise to the well known axial anomaly
69: equation~\cite{Adler-69,BJ-69,Bardeen-69}, which reads
70: \begin{equation}
71: \partial_\mu j_5^\mu(x)= i 2 N_f q(x)
72: \label{anomaly}
73: \end{equation}
74: in the chiral limit, where $q(x)$ is the topological charge density defined in
75: Eq.~(\ref{topchden}).
76:
77: The integrand in Eq.~(\ref{fermionmeasure}) is a total derivative; nevertheless
78: the existence of nontrivial topological configurations, such as instantons,
79: renders Eq.~(\ref{fermionmeasure}) nontrivial. This is essentially related to
80: the boundary conditions. If the naive boundary conditions $A_\mu=0$ at spatial
81: infinity are used, the integral in Eq.~(\ref{fermionmeasure}) would not
82: contribute and $U(1)_A$ would appear to be a symmetry again. 't Hooft
83: showed~\cite{Hooft-76,Hooft-76b} that the correct choice of the boundary
84: conditions is that $A_\mu$ is a pure gauge field at spatial infinity, i.e.
85: either $A_\mu=0$ or a gauge transformation of it. With these boundary
86: conditions, there are gauge configurations for which $\int d^4 x \,q(x)\ne 0$.
87: Despite being superficially a total divergence, the anomaly term has nonzero
88: matrix elements at zero momentum. Thus, the $U(1)_A$ charge is not conserved,
89: with the result that the theory contains neither a conserved $U(1)_A$ quantum
90: number, nor an extra Goldstone boson.
91:
92: Comparing Eq.~(\ref{fermionmeasure}) with the $\theta$ term in the Lagrangian,
93: $-i\theta q(x)$, we see that a chiral rotation on massless fermions is
94: equivalent to a shift in $\theta$: $\theta \to \theta - 2\alpha N_f$\,.
95: Conversely, an initial $\theta$ term could be rotated away in the presence of
96: massless fermions; on the other hand, if the fermions have nonzero mass, as is
97: the case with all quark species, such a chiral rotation affects their mass
98: matrix. Allowing for complex quark masses, we may write the Lagrangian mass
99: term in the form
100: \begin{equation}
101: {\cal L}_m = {1\over 2} \sum_f m_f \bar\psi_f (1+\gamma_5)\psi_f +
102: {1\over 2} \sum_f m_f^* \bar\psi_f (1-\gamma_5)\psi_f =
103: \sum_{f=1}^{N_f} \bar\psi_f\,({\rm Re} \,m_f +
104: i\, {\rm Im} \,m_f\, \gamma_5)\,\psi_f \,.
105: \end{equation}
106: P and T symmetries are broken for
107: ${\rm Im}\,m_f \ne 0$. The transformation~(\ref{chiralpsi}) modifies $m_f$ as:
108: \begin{equation}
109: m_f \to e^{2 i \alpha}\, m_f \,.
110: \end{equation}
111: Thus, we may rotate $m_f$ to a real value by an appropriate choice of $\alpha$
112: (possibly different for each flavor). The shifted value of $\theta$ after
113: these rotations is still a source of P and T noninvariance; in particular, on
114: dimensional grounds, it leads to an electric dipole moment for the neutron of
115: order~\cite{Weinberg}
116: \begin{equation}
117: d_n \sim \theta {e m_f\over m_n^2} \sim
118: \theta {e m_\pi^2\over m_n^3}
119: \approx 10^{-16} \theta \,e\,{\rm cm}.
120: \label{dno}
121: \end{equation}
122: The experimental bound on $d_n$ is $|d_n| < 2.9 \times 10^{-26}e\,{\rm
123: cm}$~\cite{Baker,Harris-etal-99}. Supplementing this experimental bound
124: with the rough estimate (\ref{dno}), and also with further theoretical
125: calculations indicating $|d_n/\theta| \sim O(10^{-15}$-$10^{-16}) \,e\,{\rm
126: cm}$ (which are reviewed in Sec.~\ref{neutron}), one then gets an
127: exceedingly small value for $\theta$: $|\theta|\ltapprox 10^{-10}$. A similar
128: bound has also been obtained by measuring the electric dipole moment of the
129: $^{199}{\rm Hg}$ atom~\cite{RGJF-01}. We shall return to this issue in
130: Sec.~\ref{neutron}.
131:
132: A number of possible explanations have been proposed for the origin of this
133: small value of $\theta$, such as spontaneous breaking of CP, vanishing of the
134: mass of the lightest $u$-quark, and postulating the axion, which effectively
135: promotes $\theta$ into a dynamical variable (see,
136: e.g.,~\cite{SW-06,Peccei-06,Weinberg}, for some relevant recent reviews). In
137: particular, in the latter theory, originally proposed by Peccei and
138: Quinn~\cite{PQ-77}, the dynamical field replacing the $\theta$ parameter could
139: relax to a minimum of the effective potential where the parity and time
140: reversal symmetries are recovered. This would also require the existence of a
141: new particle~\cite{Weinberg-78,Wilczek-78}, the axion. The various models
142: using this basic mechanism that have been proposed to solve the CP problem
143: have in common the existence of a further chiral symmetry $U(1)_{\rm PQ}$,
144: which is spontaneously broken at high energies, much higher than the QCD
145: scale, and is also broken by the anomaly. This explanation still represents
146: an interesting open possibility, although the parameter regions related to the
147: axion have been much restricted,
148: see Refs.~\cite{Weinberg,Peccei-06,PDG-06} and
149: references therein. The vanishing of the mass $m_u$ of the lightest $u$-quark
150: has also been proposed as solution of the strong CP problem~\cite{KM-86}. This
151: possibility is disfavored by a standard current algebra analysis~\cite{GL-82},
152: which shows that the experimental data are consistent with a nonzero mass.
153: Nonzero quark masses are also strongly supported by calculations within
154: lattice gauge theory, see e.g.
155: Refs.~\cite{Aubin-etal-04,Aoki-etal-03,ALPHA-05,GHIPRSS-06,Becirevic-etal-06}.
156: For a recent review on quark masses see Ref.~\cite{PDG-06}. Moreover, in
157: Refs.~\cite{Creutz-04-2,Creutz-07-3} it has been argued that the vanishing of
158: a single quark mass is renormalization scheme dependent due to nonperturbative
159: effects related to the anomaly, therefore the vanishing of $m_u$ should not be
160: relevant to a physical issue such as violation of the CP symmetry in strong
161: interactions.
162:
163: Through the anomaly equation physically interesting matrix elements of
164: hadron phenomenology are related to the topological properties of the theory.
165: An example is given by the large mass of the $\eta'$, as we shall see
166: later. Another example is related to the so-called proton spin problem, see
167: e.g. Refs.~\cite{Veneziano-89,SV-92-2,NSV-99,Shore-06,Shore-07}.
168:
169:
170: \subsection{The large-$N$ limit}
171:
172: The large-$N$ limit, where $N$ is the number of colors, is a very useful
173: framework to investigate the physics of strong interactions. In the
174: context of particle physics the $1/N$ expansion was introduced by
175: 't~Hooft~\cite{Hooft-74}. He proposed to generalize the $SU(3)$ gauge
176: symmetry of QCD to $SU(N)$ and to expand in powers of $1/N$. In fact, $N$ is
177: the only free parameter in QCD~\cite{Witten-79c}. $N$ is an intrinsically
178: dimensionless parameter; the origin of $N$ dependence is basically
179: group-theoretical, and leads to well-defined field representations for all
180: integer values, hence it is not subject to any kind of renormalization. The
181: 't Hooft large-$N$ limit is given by
182: \begin{equation}
183: N \rightarrow \infty\qquad {\rm keeping} \quad g^2 N, N_f \quad {\rm fixed}.
184: \label{largenlimit}
185: \end{equation}
186: The dominant Feynman graphs in the large-$N$ limit can be classified by simply
187: counting powers of $N$~\cite{Hooft-74,Witten-79c,Coleman-82}. In general,
188: the leading large-$N$ contributions come from planar diagrams, each fermion
189: loop introduces a factor $1/N$, while each nonplanar crossing is suppressed by
190: $1/N^2$. The Feynman diagrams contributing to connected correlation functions
191: of gauge invariant operators constructed using gauge fields are $O(N^2)$. They
192: are planar and do not contain fermion loops. The connected parts of
193: correlation functions of fermionic currents are $O(N)$. The corresponding
194: diagrams are planar, they do not have internal fermion loops, and all current
195: insertions are on a single fermion loop which bounds the graph.
196:
197: The most interesting feature of the large-$N$ expansion is that the
198: phenomenology of QCD in the large-$N$ limit presents remarkable analogies with
199: that of the real world. Assuming that the $N=\infty$ theory confines, so that
200: the physical spectrum consists of color singlets only, one can study
201: properties of hadrons by applying power counting in $N$, and
202: analyzing the intermediate states that contribute to the various $n$-point
203: correlation functions, as discussed e.g. in
204: Refs.~\cite{Witten-79c,Coleman-82}. This leads to a scenario which is
205: qualitatively, and also semi-quantitatively, consistent with those of the real
206: world (see for example Refs.~\cite{Polyakov-87,Das-87,Manohar-98,RCV-98} for
207: reviews).
208:
209: We also mention the Veneziano large-$N$ limit~\cite{Veneziano-76}
210: \begin{equation}
211: N, N_f \rightarrow \infty\qquad {\rm keeping} \quad g^2 N, N_f/N
212: \quad {\rm fixed}.
213: \label{largenlimitv}
214: \end{equation}
215: This limit provides a better explanation of certain aspects of low-energy
216: phenomenology. However, the 't~Hooft limit is simpler, and has been studied in
217: more detail.
218:
219: We should finally note that the large-$N$ solution of 4D $SU(N)$ gauge
220: theories is still unknown, and therefore it is not possible to perform a
221: systematic $1/N$ expansion. The problems in determining the large-$N$
222: saddle point are essentially related to the matrix nature of the theory, see
223: for example Ref.~\cite{Polyakov-87} for a detailed discussion. Nevertheless,
224: the $1/N$ expansion of QCD represents a very useful framework, within which
225: several phenomenological issues can be discussed, and nontrivial predictions can
226: be inferred, see e.g. Refs.~\cite{Polyakov-87,Manohar-98}. We also mention
227: that the $1/N$ expansion has been instead successfully developed in vector
228: models, such as O($N$)-symmetric theories and $CP^{N-1}$ models (see e.g.
229: Refs.~\cite{Polyakov-87,ZJ-book,MZ-03}).
230:
231: The conjectured large-$N$ scenario of QCD (discussed at length in Polyakov's
232: book~\cite{Polyakov-87}) can be investigated by performing lattice
233: calculations at relatively large $N$, studying their convergence for
234: $N\rightarrow\infty$ in the `quenched' case (i.e. with no dynamical fermions,
235: given that contributions of quark loops should be depressed by a factor
236: $1/N$).
237:
238:
239: \subsection{Solution of the $U(1)_A$ problem within the large-$N$ limit:
240: The Witten-Veneziano formula}
241: \label{u1asolln}
242:
243: In general a broken symmetry is best understood by studying the limit leading
244: to a theory where the symmetry is conserved. In the case of $U(1)_A$
245: symmetry, such a limit should be that of a large number of colors:
246: $N\rightarrow \infty$. An explanation of the explicit breaking of the
247: $U(1)_A$ symmetry based on large-$N$ arguments was originally proposed by
248: Witten~\cite{Witten-79} and then refined by Veneziano~\cite{Veneziano-79}.
249: According to Witten's argument the $U(1)_A$ problem should be solved at the
250: lowest nonplanar level, i.e. at the next-to-leading order of its $1/N$
251: expansion. Let us sketch this argument.
252: We introduce the topological susceptibility $\chi$
253: \begin{equation}
254: \chi = \left. {\partial^2 F(\theta)\over \partial \theta^2} \right|_{\theta=0}=
255: \int d^4 x \langle q(x) q(0) \rangle ,
256: \label{chidefu1}
257: \end{equation}
258: as defined in Euclidean space. We recall that in Minkowski space-time the
259: topological susceptibility reads
260: \begin{equation}
261: \chi = -i \int d^4 x \langle 0 | T q(x) q(0)| 0 \rangle
262: \label{chimin}
263: \end{equation}
264: (more details on this definition and its relation with the corresponding
265: Euclidean quantity can be found in
266: Refs.~\cite{Crewther-79,Witten-79,Meggiolaro-98}).
267:
268: While there is no $\theta$ dependence in perturbation theory, there may be a
269: nontrivial dependence at a nonperturbative level, and in particular within the
270: $1/N$ expansion. Assume that the pure gauge theory without quarks is
271: $\theta$-dependent to leading order in $1/N$, and therefore that $\chi\neq 0$
272: at $N=\infty$. This assumption leads to an apparent paradox: on the one hand,
273: when massless quarks are introduced the $\theta$ dependence must disappear and
274: therefore $\chi=0$, on the other hand quark loops give only nonleading
275: contributions of order $1/N$ to the physical processes of the pure gauge
276: theory~\cite{Hooft-74}. This paradox can be solved by the presence of a
277: particle with mass squared $m_s^2$ of order $1/N$, having the same quantum
278: numbers of the topological charge density. Such a particle should be related
279: to the $\eta'$, i.e. the lightest flavor-singlet pseudoscalar in nature.
280:
281: The existence of a particle of mass squared $O(1/N)$ can be justified noting
282: that in the large-$N$ limit the singlet axial current should be conserved.
283: The anomaly is an $O(1/N)$ effect, because it arises from a diagram with one
284: quark loop, which is depressed by a factor $1/N$. Therefore at $N=\infty$
285: there should be an axial singlet Goldstone boson. At $O(1/N)$ the anomaly is
286: recovered and, as a consequence, the $N=\infty$ Goldstone boson gets a mass.
287: At order $1/N$, $m_{s}^2$ should receive an $O(1/N)$ contribution, in that the
288: mass squared of an approximate Goldstone boson is in general linear with
289: respect to the symmetry breaking parameter, which in this context is
290: represented by $1/N$.
291:
292: So as a consequence of a nonzero large-$N$ limit of $\chi$ in the pure gauge
293: theory, we would have $m_s^2\sim 1/N$, while the other Goldstone bosons
294: associated with the nonsinglet axial symmetry remain massless at the chiral
295: limit. A further development of these ideas results in the leading order
296: relationship
297: \begin{equation}
298: \chi={f_s^2 m_s^2\over 4N_f},
299: \label{wf}
300: \end{equation}
301: where $f_s$ is defined by
302: \begin{equation}
303: \langle 0 | \; \partial_\mu j_5^\mu \;|s\rangle=\sqrt{N_f} m_s^2 f_s.
304: \label{mj5}
305: \end{equation}
306: Notice that $f_s$ is of order $\sqrt{N}$ and in the large-$N$ limit $f_s
307: =f_{ns}=f_\pi$ (where $f_\pi$ is defined as in Ref.~\cite{PDG-06}).
308: By performing a more accurate analysis, based on the
309: large-$N$ limit (\ref{largenlimitv}) and using also the anomalous
310: flavor-singlet Ward-Takahashi identities, Veneziano~\cite{Veneziano-79}
311: refined the relationship (\ref{wf}) obtaining
312: \begin{equation}
313: {4N_f\over f_\pi^2} \chi=m_{\eta'}^2 + m_\eta^2 - 2m_K^2.
314: \label{vf}
315: \end{equation}
316:
317: In Eqs.~(\ref{wf}) and (\ref{vf}), due to their large-$N$ based derivation,
318: $\chi$ should be considered that of the pure gauge theory. This fact favors a
319: check by Monte Carlo simulation of the lattice formulation, in that pure gauge
320: simulations are much simpler than full QCD ones. Substituting $N_f=3$ and the
321: experimental values in (\ref{vf}) (i.e.~\cite{PDG-06} $f_\pi\approx 131$ MeV,
322: $m_{\eta'}\approx 958$ MeV, $m_{\eta}\approx 547$ MeV, $m_{K}\approx 494$ MeV)
323: the prediction is $\chi\approx \left( 180 \;{\rm MeV}\right)^4$. This
324: prediction has been substantially verified by lattice computations, as we
325: shall see below.
326:
327: We mention that the Witten-Veneziano formula has been also derived within the
328: lattice formulation of QCD, when the so-called Ginsparg-Wilson lattice
329: fermions are considered~\cite{GRTV-02}, see also Sec.~\ref{ua1latt}.
330: Moreover, it has also been investigated within the AdS/CFT framework in
331: Refs.~\cite{Armoni-04,BHMM-04,HO-99,SS-05,Katz-07}.
332:
333:
334: Models based on instanton semiclassical pictures can hardly explain the
335: nontrivial large-$N$ behavior of the topological susceptibility,
336: $\chi=O(N^0)$, and therefore of the $\eta'$ mass, $m_{\eta'}^2\sim 1/N$. The
337: naive dilute-instanton picture would suggest that the topological
338: susceptibility and $m_{\eta'}$ vanish exponentially as $e^{-c N}$, $c$ being
339: some constant, because the instanton weight behaves as $\exp(-8\pi/g^2)$, and
340: the 't~Hooft large-$N$ limit must be taken keeping $g^2 N$ fixed, thus leading
341: to an apparent exponential suppression. Actually, since a nontrivial
342: $N$-dependent prefactor~\cite{Hooft-76b} appears in the weight of the
343: instanton distribution, this argument strictly applies only to small
344: instantons, leaving open the possibility to get a consistent scenario by
345: appropriately taking into account instanton contributions at the QCD scale.
346: The consistency of the semiclassical instanton calculation and $1/N$ expansion
347: of QCD has been discussed in Refs.~\cite{Schafer-02,Schafer-04}; in particular
348: it was argued that the instanton liquid model, where large (overlapping)
349: instantons get suppressed by ad-hoc short-range repulsive instanton
350: interactions~\cite{Shuryak-94}, is not necessarily incompatible with the
351: large-$N$ scenario. This issue has been also
352: discussed~\cite{Jevicki-79,David-84} within the 2D $CP^{N-1}$ models,
353: see also Sec.~\ref{thetadepcpnln}.
354:
355: