1: \section{The 3-point function}
2: \label{sec:3point}
3: For the 3-point function, already introduced
4: in Ch. 2, we now consider a pseudo-scalar source
5: at $t = 0$, a sink at $t_f = 11$ and a coupling to the
6: photon at $t$ with $0 < t < t_f$.
7: Barad \textit{et al.}~\cite{Barad:1984px} pointed out
8: that current conservation provides an important
9: numerical test which relates the 2- and 3-point
10: functions. Translated into momentum space and for
11: our periodic boundary conditions, this relation reads
12: \begin{equation}
13: G_3(t_f, t; \bm{p}, \bm{p}) - G_3(t_f, t';
14: \bm{p}, \bm{p}) = G_2 (t_f,\bm{p}) \; ,
15: \label{eq:chargecons}
16: \end{equation}
17: where, in the second term on the left hand side,
18: $ t_f < t' < N_{\tau}$. It amounts to considering the
19: total charge that leaves the source in the forward
20: and backward direction in time and guarantees that
21: $F(Q^2) = 1$ at $Q^2 = 0$.
22: This relation holds for each background gauge field configuration separately
23: and thus also for configuration averages. We have verified
24: that our results for the conserved current
25: satisfy this relation
26: to an accuracy better than ${\cal O}(10^{-5})$.
27:
28: In order to obtain $F(0)$ for the
29: renormalized local and the improved current,
30: we again exploited Eq.~\ref{eq:chargecons}.
31: The lhs of this equation was averaged over pairs of values $t$ and $t'$
32: symmetric around $t_f$ and normalized by the 2-point function.
33: Utilizing the $Z_V$ factor
34: from~\cite{Bhattacharya:2000pn} gives $F^I(0) = F^{RL}(0) = 1$ within
35: a jackknife error of 1 \%.
36: Alternatively, we could have applied this method to
37: independently extract $Z_V$ as done {\it e.g.} in~\cite{Bakeyev:2003ff}.
38: However, for consistency we used the entire set
39: of improvement parameters from~\cite{Bhattacharya:2000pn}.
40:
41: While the above method allows one to determine $F(0)$,
42: we now describe how we extract the form factor for\\
43: $Q^2 > 0$.
44: As in the 2-point function we allow two states to
45: contribute and parametrize the 3-point function, Eq.~\ref{eq:3p}, as
46: \begin{eqnarray}
47: G_{3,\mu}(t_f,t;\bm{p}_f,\bm{ p}_i) =
48: \sum_{m=0}^{1}\sum_{n=0}^{1}
49: \sqrt {Z_R^m (\bm{p}_f) Z_0^n (\bm{p}_i)}\nonumber \\
50: \times\;\;\left<m,\bm{p}_f| j_\mu (0) |n,\bm{p}_i \right>
51: e^{-\, E^m_{\bm{p}_f} \, (t_f - t) -\, E^n_{\bm{p}_i} \, t}\;,
52: \end{eqnarray}\\
53: where $(m,n)\neq(1,1)$.
54: %\begin{widetext}
55: %\begin{eqnarray}
56: %G_{3,\mu}(t_f,t;\bm{p}_f,\bm{ p}_i) =
57: %\sqrt {Z_R^0 (\bm{p}_f)Z_0^0 (\bm{p}_i) }\; \left<0,\bm{p}_f|j_\mu(0)|0,
58: %\bm{p}_i \right> e^{- E^0_{\bm{p}_f }\; (t_f - t)\, - E^0_{\bm{p}_i}\; t} \nonumber \\
59: %+\left\{\sqrt {Z_R^1 (\bm{p}_f) Z_0^0 (\bm{p}_i)}\left<1,\bm{p}_f|
60: %j_\mu (0) |0,\bm{p}_i \right>
61: %e^{-\, E^1_{\bm{p}_f} \, (t_f - t) -\,
62: %E^0_{\bm{p}_i} \, t} + (1 \leftrightarrow 0) \right\}\;.
63: %\end{eqnarray}
64: %\end{widetext}
65: Contributions from, for example, the production of pion pairs, as well
66: as 'wrap around effects' due to the propagation of states beyond $t_f$
67: are exponentially suppressed ($ < \mathcal{O}$$(e^{-5})$); similarly,
68: an elastic contribution from the excited state was estimated to be of
69: the order of $1\%$ or less. All these effects are not reflected in
70: the chosen parametrization. The inelastic transitions $0
71: \leftrightarrow 1$ are included to better describe the data. However,
72: it should be understood that the state labelled $1$ parametrizes
73: contributions from all possible excited states. Therefore we do not
74: interpret our parameters as the energy or the transition form factors
75: corresponding to a single genuine excited state.
76:
77: Since for a given momentum the pion is the state with the
78: lowest energy, one gets with Eqs.~\ref{eq:lorentz_dec}
79: and~\ref{eq:normalisation}
80: \begin{widetext}
81: \begin{eqnarray}
82: G_{3,\mu}(t_f,t;\bm{p}_f,\bm{ p}_i) = F(Q^2)\;
83: \frac{(p_f+p_i)_{\mu}}{2\sqrt{E^0_{{\bf p}_f} E^1_{{\bf p}_i}}}\sqrt {Z_R^0
84: (\bm{p}_f)Z_0^0 (\bm{p}_i) }\; e^{- E^0_{\bm{p}_f }\; (t_f - t)\, - E^0_{\bm{
85: p}_i}\; t }~~~~~~~~~~~~~~~~~~~~~~ \nonumber\\
86: +\left\{\sqrt {Z_R^1 (\bm{p}_f) Z_0^0 (\bm{p}_i)}\left<1,\bm{p}_f|
87: j_\mu (0) |0,\bm{p}_i \right>e^{-\, E^1_{\bm{p}_f} \, (t_f - t) -\,
88: E^0_{\bm{p}_i} \, t} \;+ (1 \leftrightarrow 0) \right\}\;.
89: \label{eq:param}
90: \end{eqnarray}
91: \end{widetext}
92: In the simulations, we took $|\bm{p}_f|=|\bm{p}_i|=\sqrt{2}|\bm{p}_{min}|$, where
93: \begin{equation}
94: |\bm{p}_{min}| = \frac{2 \pi}{N_{\sigma} a}
95: \end{equation}
96: is the minimal momentum for a lattice with $N_{\sigma}$ lattice
97: points in the spatial extension. In our case, it amounts to
98: \begin{equation}
99: \bm{p}_i^2 = \bm{p}_f^2 = 0.48 \text{ GeV}^2 \; .
100: \end{equation}
101: For our analysis we use the fourth component of the current, $\mu = 4$.
102: With our choice of momenta the kinematical factors in
103: the fit function, Eq.~\ref{eq:param}, therefore simplify considerably,
104: see Eq.~\ref{eq:prefac}; note also that the $t$-dependence of the first term in Eq.~\ref{eq:param}
105: vanishes.
106: As a result, the form factor is more easily extracted without
107: restricting the simulation too much. Different momentum
108: transfers are obtained by varying the relative orientation
109: of $\bm{p}_i$ and $\bm{p}_f$.
110:
111: \subsection{Extraction of parameters}
112: \label{Ext_par}
113: We begin by averaging the 3-point correlation
114: functions which have the same four-mo\-men\-tum transfer
115: squared and then again combine the configurations in
116: jackknife averages. Typical jackknife averages of the
117: 3-point function are shown in Fig.~\ref{fig:3point_all_Q2}
118: \begin{figure}
119: \subfigure %[$m_{\pi}=780$ MeV.]
120: {
121: \includegraphics[height=85mm,angle=270]{3point_all_Q2.eps}
122: \label{fig:3point_all_Q2}
123: }
124: \subfigure %[$Q^2=0.97$ GeV$^2$]
125: {
126: \includegraphics[height=85mm,angle=270]{3point_all_k.eps}
127: \label{fig:3point_all_k}
128: }
129: \caption{Improved 3-point function for (a) different $Q^2$ at
130: $m_{\pi}=780$ MeV and (b) different $m_{\pi}$
131: at $Q^2=0.97$ GeV$^2$.}
132: \end{figure}
133: for the one but highest pion mass for various momentum
134: transfers and for different masses at fixed momentum
135: transfer in Fig.~\ref{fig:3point_all_k}. If only the
136: pion ground state would contribute, there would be
137: no $t$-dependence in this quantity. The data however
138: clearly indicate the admixture of an excited state.
139: This is especially seen at high $Q^2$ and for low pion
140: masses, where there is no time-slice where it is safe to
141: assume that only the pion ground state is present.
142:
143: We therefore chose to proceed by simultaneously fitting
144: the parameters in the 2-point and 3-point function,
145: Eq.~\ref{eq:2point_para} and Eq.~\ref{eq:param}, respectively.
146: We hereby exploit the fact
147: that certain parameters appear in both Green's functions.
148: In the case of the 2-point function,
149: we fitted the energies, $E^0$ and $E^1$ and the $Z$ factors
150: over the complete time interval, $1 \leq t \leq N_\tau-1$.
151: In the 3-point function, we fit the same energies and $Z$-factors,
152: and in addition the form factor $F(Q^2)$ and the transition
153: matrix elements over the interval $t_i < t < t_f$.
154: We assume that the energies and $Z$ factors only depend on the
155: magnitude of the three-momenta and use the fact that we
156: chose $|{\bf p}_f |=|{\bf p}_i |$.
157: The fits are done for each value of $Q^2$ separately. The values for
158: $\chi^2/dof$ lie between 0.15 and 0.40, depending on mass and momentum
159: transfer. The energies and $Z$ factors we obtain from our fits at
160: different $Q^2$ agree to high accuracy because they are largely
161: determined by the 2-point function.
162:
163: To compare with earlier work
164: \cite{Martinelli:1988bh,Draper:1989bp}
165: we also extracted an estimate for the form factor
166: from the ratio of 3- and 2-point functions.
167: However, the assumption of just a single state contributing
168: is at the basis of this method. Correspondingly,
169: we found differences ranging from $5 - 10 \%$ for
170: $F(Q^2)$ between the ratio method and our combined fit
171: method, where inclusion of an excited state clearly
172: improved the fit quality. The size of the difference
173: depends on the pion mass and the momentum transfer, which
174: influences the flatness of the 3-point function in the middle
175: between source and sink. All our results in the next
176: section are therefore based on the fit method.
177: