0710.2309/analysis2.tex
1: \chapter{Analysis II:  Including $\sigma \rightarrow \sigma_\mathrm{eff} $ } \label{chap:analysis2}
2: 
3: One problem for the Isgur-Paton model is understanding what happens as $r \rightarrow 0$. The string picture of a flux tube is only appropriate on scales where the radius is much larger than the thickness of the flux tube $\sim {1\over \sqrt{\sigma}}$.  With the radius $\sim {1\over \sqrt{\sigma}}$, the flux tube would look more like an annulus in two spatial dimensions or a torus in three dimensions, and the $M/r$ potential for the phonon contribution seems unreasonable.  Indeed, the fluctuations might not even be able to be described by a phonon potential.  A detailed theory of membrane dynamics might address the problem, but such is beyond the scope of this work.
4: 
5: \begin{figure}[h]
6: \centering
7: \includegraphics[width=\textwidth]{sigeff.eps}
8: \caption{$\sigma_\mathrm{eff}$ vs r}
9: \label{fig:sigeff}
10: \end{figure}
11: 
12: What if as $r \rightarrow 0$, the string itself ``dissipated''?  A study of the dual Ginzburg-Landau theory of type-II superconductors~\cite{Koma:1999sm} determined a function $\sigma_\mathrm{eff}(r)$, Figure~\ref{fig:sigeff}, defined to be the total energy of a flux ring solution of radius r divided by the circumference.  The total energy was calculated from the microscopic differential equations for the fields.  The non-abelian gauge theory's flux tube is often thought to arise from a dual Meissner effect for a type-II superconducting picture of the QCD vacuum~\cite{Mandelstam:1973,'tHooft:1981ht}.  As $\sigma$ is defined from the interquark potential as $r \rightarrow \infty$, its behavior as $r \rightarrow 0$ is uncertain, and we might apply $\sigma \rightarrow \sigma_\mathrm{eff}(r)$ to the Isgur-Paton model.  Transcribing the function from Figure~\ref{fig:sigeff} and converting scales, we can fit the function with the form
13: \beq \sigma_\mathrm{eff} = 1 - e^{-f r} \label{eqn:sigeff} \eeq
14: for f=1.7177, best in a least squares sense.  Knowing the functional form of $\sigma_\mathrm{eff}(r)$, we can investigate its effect on the Isgur-Paton model.
15: 
16: When we quantized the model earlier, the difficult term was the kinetic energy, so that is where we will begin.  Abusing notation slightly,
17: \bea T &=& \frac{P_r^2}{2 \mu} \\
18:  &\rightarrow& \frac{-1}{4 \pi} \left( \frac{\frac{\pa}{\pa r}}{\sqrt{\sigma r}}\right)^2 , \eea
19: where $\sigma$ is now included in the r dependence of the kinetic energy.  If $\sigma$ were still a constant, we would get (as before)
20: \beq T \rightarrow \frac{-1}{4 \pi \sigma_\mathrm{const}} {9\over 4} \frac{\pa^2}{\pa \rho^2} , \eeq
21: since $d\tilde{\rho} = r^{\half} dr$, $\tilde{\rho} = {2\over 3} r^{3\over 2}$.  Now we must include $\sigma_\mathrm{eff}$ in our differential equation relating r and $\rho$.  For some canonical variable $\rho$,
22: \beq \frac{\frac{d\rho}{dr}\frac{d}{d\rho}}{\sqrt{\sigma_\mathrm{eff} r}} = \frac{d}{d\rho} , \eeq
23: so
24: \beq d\rho = dr \sigma_\mathrm{eff}^\half r^\half \eeq
25: with boundary conditions
26: \beq \rho(0)=0 , \; \rho(1) = 1 . \eeq
27: Substituting our function for $\sigma_\mathrm{eff}$, we find
28: \beq \rho = \int dr r^\half \left( 1 - e^{-f r} \right)^\half . \eeq
29: This integral cannot be put into closed form, yet we need the antiderivative explicitly to substitute into the potential $V(r(\rho))$.  While it would be nice to include explicitly the function from Equation~\ref{eqn:sigeff}, any similar function which captures the salient features should suffice to test its influence on the spectrum.  We need a function which goes to zero as $r \rightarrow 0$ and which approaches unity asymptotically, and which is integrable with respect to the measure $r^\half dr$.  After looking at several functional forms, the closest one meeting these criteria is
30: \beq \sigma_\mathrm{eff} = \left( 1 - e^{-f' r} \right)^2 , \eeq
31: where f' is now given by 2.696.  Performing the integral (and dropping the prime)
32: \beq \int dr r^\half \left( 1 - e^{-f r} \right) = \frac{\sqrt{r}}{f} \left(e^{-f r} + {2\over 3} f r \right) - \frac{\sqrt{\pi}}{2 f^{3\over 2}} \mathrm{erf}\sqrt{f r} , \eeq
33: thus
34: \beq \tilde{\rho} = {2\over 3} r^{3\over 2} + \frac{\sqrt{r}}{f}e^{-f r} - \frac{\pi}{2 f^{3\over 2}} \mathrm{erf}\sqrt{f r} . \eeq
35: Applying the condition $\rho(r=1) = 1$,
36: \beq \tilde{\rho}(1) = c = 0.495544 , \eeq
37: so
38: \beq \rho = {1\over c}\tilde{\rho} = \frac{\sqrt{r}}{f c} \left( e^{-f r} + {2\over 3} f r \right) - \frac{\sqrt{\pi}}{2 c f^{3\over 2}} \mathrm{erf}\sqrt{f r} . \label{eqn:rvrho} \eeq
39: We cannot find the functional inverse explicitly, but as ultimately we are working with a discrete set $\{\rho_j\}$ related to $\{t_j\}$, we can solve the equation numerically to give $r_j$,
40: \beq \rho_j = \frac{n+1}{j} \Rightarrow r_j , \eeq
41: which we can put into the potential
42: \beq V = V(r(\rho_j)) . \eeq
43: The solution of Equation~(\ref{eqn:rvrho}) is shown in Figure~\ref{fig:rvsrho}.
44: \begin{figure}[t]
45: \centering
46: \includegraphics[width=.5\textwidth]{rvsrho.eps}
47: \caption{r vs $\rho$}
48: \label{fig:rvsrho}
49: \end{figure}
50: Returning to the kinetic energy,
51: \beq T \rightarrow \frac{-1}{4 \pi} {1\over c^2} \frac{\pa^2}{\pa \rho^2} , \eeq
52: so our Hamiltonian goes to
53: \bea T + V = E &\Rightarrow& \\
54: \frac{-\pa^2}{\pa \rho^2} + 4 \pi c^2 V(r(\rho)) &=& 4 \pi c^2 E = \lambda , \eea
55: which we transform as before to the matrix equation 
56: \beq S_b + S_d = \lambda \eeq
57: and solve.
58: 
59: So, how does this model perform?  Looking at Table~\ref{table:sigchi2}, we see the values of $\chi^2$ for various N at $\beta =0$.  For the direct mechanism, $\chi^2$ decreases from 117 at N=3 down to 15.3 for N=$\infty$.  This N dependence of $\chi^2$ seems a little odd.  For small N, $\chi^2$ is too large to believe the model including the effective string tension, but as $N\rightarrow \infty$, $\chi^2$ drops to a value which is nearly believable, even with the tight errors.  As the details of the conjugate splitting might depend on N, we are seeing evidence that the Isgur-Paton model is most applicable to the $N\rightarrow \infty$ limit.
60: 
61: Looking now at the indirect mechanisms, we see the same type of behavior or $\chi^2$.  For the adjoint mechanism, $\chi^2$ ranges from 135 for N=3 down to 26.5 for N=$\infty$.  Apparently, having $\sigma_\mathrm{eff} \rightarrow 0$ works reasonably well for $N\rightarrow \infty$, but for small N including such behavior is detrimental to the model.  For vertex mixing, $\chi^2$ simply remains unbelievably large, $\gtrsim 100$.  As this mechanism has an extra flux tube for its diameter, imposing $\sigma_\mathrm{eff} \rightarrow 0$ is a major complication to the model and is not supported by the evidence.
62: 
63: \begin{table}
64: \centering
65: \begin{tabular}{|l|ccccc|} \hline
66: N & 2 & 3 & 4 & 5 & $\infty$ \\\hline
67: direct & & 117 & 79.3 & 36.5 & {\bf{15.3}} \\\hline
68: adjoint & 13.2 & 135 & 101 & 81.1 & 26.5 \\\hline
69: vertex & & 257 & 130 & 96.4 & \\\hline
70: \end{tabular}
71: \caption{$\chi^2$ for $\sigma_\mathrm{eff}$}
72: \label{table:sigchi2} 
73: \end{table}
74: 
75: Comparing the values of $\chi^2$ at $\beta =0$ for both $\sigma \equiv 1$ and $\sigma_\mathrm{eff} \rightarrow 0$, we can say that the lattice data best supports the direct mixing mechanism in the limit as $N\rightarrow \infty$.  Were we to believe in using the incomplete gamma function~\cite{Press:1992} to assess this model's goodness-of-fit, we would find 
76: \bea Q &=& {{\Gamma({\nu\over 2},{\chi^2\over 2})}\over {\Gamma({\nu\over 2})}}\\
77:  &=& .88 \eea
78: 
79: \begin{table}[t]
80: \centering
81: \begin{tabular}{|r|c|c|} \hline
82: N & $\gamma$ & $\alpha$ \\\hline
83: direct & & \\
84: 3 & .716 & 1.92  \\
85: 4 & .602 & 1.89  \\
86: 5 &-.007 & .973  \\
87: $\infty$ & -.069 & 1.06 \\
88: $(1/N^2)|_\infty$ &      &  \\\hline
89: adjoint & & \\
90: 2 &-.104  & .897  \\
91: 3 & .706  & 3.35  \\
92: 4 & .594  & 3.26  \\
93: 5 &-.292  & 1.22  \\
94: $\infty$ &-.324 & 1.29 \\
95: $(1/N^2)|_\infty$ &      & \\\hline
96: vertex & & \\
97: 3 & .512  & 1.52  \\
98: 4 & .472  & 2.12  \\
99: 5 &-.223  & 1.28  \\
100: $(1/N^2)|_\infty$ &      & \\\hline
101: \end{tabular}
102: \caption{Parameter values for $\sigma_\mathrm{eff}$ for $\beta$ = 0.}
103: \label{table:sigParametersB0}
104: \end{table}
105: 
106: Turning now to the parameter values for the direct and adjoint mechanisms, Table~\ref{table:sigParametersB0}, we find some interesting behavior.  The $(1/N^2)$ dependence is not readily apparent for $\alpha$ and is dubious for $\gamma$.  Overall, the results seem to split into two sectors across mechanisms, one for N=3 and 4, the other for N=5 and $\infty$.  The smaller N return a $\gamma$ which is positive $\sim$ .6 or .7, while for higher N a negative $\gamma$ is supported $\sim$ -.07 or -.3.  As $\gamma$ shifts discontinuously, $\alpha$ also makes a jump.  In the direct scenario, $\alpha$ jumps from $\sim$ 1.9 to $\sim$ 1.0 as $\gamma$ goes from positive to negative.  For adjoint mixing, $\alpha$ goes from $\sim$ 3.3 down to $\sim$ 1.2.  No longer seeing our $(1/N^2)$ dependence, we need to explore the posteriors in more detail.
107: 
108: \newpage
109: 
110: \begin{figure}[!t]
111: \centering
112: \includegraphics[width=\textwidth]{b0wcdirN3.eps}
113: \caption{The direct mixing posterior and spectrum for N=3.}
114: \label{fig:b0wcdirN3}
115: \end{figure}
116: 
117: \begin{figure}[!t]
118: \centering
119: \includegraphics[width=\textwidth]{b0wcdirN4.eps}
120: \caption{The direct mixing posterior and spectrum for N=4.}
121: \label{fig:b0wcdirN4}
122: \end{figure}
123: 
124: \begin{figure}[!t]
125: \centering
126: \includegraphics[width=\textwidth]{b0wcdirN5.eps}
127: \caption{The direct mixing posterior and spectrum for N=5.}
128: \label{fig:b0wcdirN5}
129: \end{figure}
130: 
131: \begin{figure}[!t]
132: \centering
133: \includegraphics[width=\textwidth]{b0wcdirN6.eps}
134: \caption{The direct mixing posterior and spectrum for N=$\infty$.}
135: \label{fig:b0wcdirN6}
136: \end{figure}
137: 
138: 
139: Looking at the posterior for direct mixing first,
140: Figures~\ref{fig:b0wcdirN3} through \ref{fig:b0wcdirN6}, we see that the peaks for N=3 and 4 are likely to be related, while starting at N=5 the peaks shifts dramatically.  If we turn our attention to comparing the spectra, we can see what is driving the different peaks.  The prime difficulty is getting the J=0 sector to fall into place.  For N=4, Figure~\ref{fig:b0wcdirN4}, we see that the model is choosing to fit the conjugate splitting between the $0^{++}$ and the $0^{--}$ at the expense of the magnitude of the radial excitations.  (Such is also true for N=3.)  When N=5, the model shifts to favoring matching the radial excitations at the expense of the low lying $0^{--}$.  Now the excited states are within $\sim 1 \sigma$, while the $0^{--}$ is about $2\sim3 \sigma$ away.  This behavior explains the shift in the peak of the posterior.  What we would like is a model which can incorporate both the conjugate splitting as well as the radial excitations.  The J=2 states seem to be little affected by the shift of the peak -- so far they are compatible with either emphasis.
141: 
142: \begin{figure}[!t]
143: \centering
144: \includegraphics[width=\textwidth]{b0wcadjN3.eps}
145: \caption{The adjoint mixing posterior and spectrum for N=3.}
146: \label{fig:b0wcadjN3}
147: \end{figure}
148: 
149: \begin{figure}[!t]
150: \centering
151: \includegraphics[width=\textwidth]{b0wcadjN4.eps}
152: \caption{The adjoint mixing posterior and spectrum for N=4.}
153: \label{fig:b0wcadjN4}
154: \end{figure}
155: 
156: \begin{figure}[!t]
157: \centering
158: \includegraphics[width=\textwidth]{b0wcadjN5.eps}
159: \caption{The adjoint mixing posterior and spectrum for N=5.}
160: \label{fig:b0wcadjN5}
161: \end{figure}
162: 
163: \begin{figure}[!t]
164: \centering
165: \includegraphics[width=\textwidth]{b0wcadjN6.eps}
166: \caption{The adjoint mixing posterior and spectrum for N=$\infty$.}
167: \label{fig:b0wcadjN6}
168: \end{figure}
169: 
170: \begin{figure}[!t]
171: \centering
172: \includegraphics[width=\textwidth]{b0wcverN3.eps}
173: \caption{The vertex mixing posterior and spectrum for N=3.}
174: \label{fig:b0wcverN3}
175: \end{figure}
176: 
177: \begin{figure}[!t]
178: \centering
179: \includegraphics[width=\textwidth]{b0wcverN4.eps}
180: \caption{The vertex mixing posterior and spectrum for N=4.}
181: \label{fig:b0wcverN4}
182: \end{figure}
183: 
184: \begin{figure}[!t]
185: \centering
186: \includegraphics[width=\textwidth]{b0wcverN5.eps}
187: \caption{The vertex mixing posterior and spectrum for N=5.}
188: \label{fig:b0wcverN5}
189: \end{figure}
190: 
191: 
192: Turning to the indirect mechanisms, we see much the same behavior.  For N=3 and 4, we see one peak corresponding to fitting the conjugate splitting, while for N=5 and $\infty$ we see the other peak corresponding to fitting the radial excitations.  Fitting the radial excitations gives a lower $\gamma$ than fitting the splitting, but the coefficient of elasticity is still positive, as the signal for zero elasticity is $\gamma \sim -1$, from which the peak is still some distance away.
193: 
194: Looking in detail at the spectrum resulting from the overall best fit (to the statistical errors), the $0^{--}$ notwithstanding, we see that the higher states are agreeing quite well with the hypothesis of spin misassignment.  The M=4 states match the J=0 states with mixed quantum numbers quite well.  When M=5, the $\pm+$ state agrees with the J=$1^{\pm+}$, but the $\pm-$ state is predicted to be too heavy, yet the lattice data would agree will with the M=3.  With the free predictions of the model giving such a signal, we should take seriously the issue of the spin assignments of the lattice operators.
195: 
196: 
197: \begin{table}[h]
198: \centering
199: \begin{tabular}{|lc|cc|cc|cc|cc|cc|} \hline
200: N & direct & $0^{++}$ & $0^{--}$ & $2^{\pm+}$ & $2^{\pm-}$ &
201: $3^{\pm+}$ & $3^{\pm-}$ & $4^{\pm+}$ & $4^{\pm-}$ & $1^{\pm+}$ &
202: $1^{\pm-}$ \\\hline
203: 
204: 3 & $m_{J^{PC}}$ &
205: 4.308  &  6.228  &  6.838  &  8.758  &  8.005  &  9.924  &  9.108  &
206: 11.03  &  10.15  &  12.07 \\
207:   & $m_{J^{PC}}^\star$ & 
208: 7.083  &  9.002  &  9.047  &  10.97  &  9.986  &  11.91  &   10.9  &
209: 12.82  &  11.77  &  13.69 \\\hline
210: 
211: 4 & & 
212: 4.183  &  6.076  &  6.727  &   8.62  &  7.901  &  9.794  &  9.012  &
213: 10.9  &  10.06  &  11.95 \\ & & 
214: 6.994  &  8.887  &  8.965  &  10.86  &  9.907  &   11.8  &  10.82  &
215: 12.71  &   11.7  &   13.6 \\\hline
216: 
217: 5 & & 
218: 4.276  &  5.249   &   6.9  &  7.873  &  8.114  &  9.087  &  9.263  &
219: 10.24  &  10.35  &  11.32 \\ & & 
220: 6.223  &  8.265  &  8.847  &  10.27  &  10.06  &  11.23  &  11.19  &
221: 12.16  &  12.09  &  13.06 \\\hline
222: 
223: $\infty$ & & 
224: 4.105  &  5.164  &  6.737  &  7.796  &  7.955  &  9.014  &  9.108  &
225: 10.17  &   10.2  &  11.26 \\ & & 
226: 6.223  &  8.201  &  8.855  &  10.21  &  10.07  &  11.17  &  11.04  &
227: 12.1  &  11.95  &  13.01  \\\hline
228: 
229: \end{tabular}
230: \caption{The spectrum for direct mixing with $\sigma_{eff} \rightarrow
231: 0$ at various N.  For each N, the
232: first mass is the lowest state in that channel, and the second mass is
233: the radial excitation.}
234: \label{table:b0wcdir}
235: \end{table}
236: 
237: 
238: 
239: \begin{table}[h]
240: \centering
241: \begin{tabular}{|lc|cc|cc|cc|cc|cc|} \hline
242: N & adjoint & $0^{++}$ & $0^{--}$ & $2^{\pm+}$ & $2^{\pm-}$ &
243: $3^{\pm+}$ & $3^{\pm-}$ & $4^{\pm+}$ & $4^{\pm-}$ & $1^{\pm+}$ &
244: $1^{\pm-}$ \\\hline
245: 
246: 3 & $m_{J^{PC}}$ &
247: 4.25  &  6.215  &  6.963  &  8.746  &  8.214  &  9.913  &  9.396  &
248: 11.02  &  10.51  &  12.06 \\
249:   & $m_{J^{PC}}^\star$ & 
250: 7.227  &  8.992  &  9.357  &  10.96  &  10.36   &  11.9  &  11.34  &
251: 12.81  &  12.27  &  13.69 \\\hline
252: 
253: 4 & & 
254: 4.124  &  6.066  &   6.86  &  8.611  &  8.121  &  9.786  &  9.313  &
255: 10.9  &  10.44  &  11.95 \\ & & 
256: 7.134  &  8.879  &  9.281  &  10.85  &  10.29  &  11.79  &  11.27  &
257: 12.71  &  12.21  &  13.59  \\\hline
258: 
259: 5 & & 
260: 4.178  &  4.855  &  7.021  &  7.516  &  8.317  &  8.748  &  9.534  &
261: 9.916  &  10.68  &  11.02 \\ & & 
262: 6.373  &  7.971  &   9.28  &  9.991  &  10.46  &  10.96  &   11.5  &
263: 11.9  &  12.47  &  12.81  \\\hline
264: 
265: $\infty$ & & 
266: 4.06   &  4.81  &  6.925  &  7.475   & 8.229   & 8.709  &  9.455   &
267: 9.879   & 10.61  &  10.98 \\ & & 
268: 6.298  &  7.937  &  9.191  &  9.959  &  10.37  &  10.93  &  11.43  &
269: 11.87  &   12.4  &  12.78 \\\hline
270: 
271: \end{tabular}
272: \caption{The spectrum for adjoint mixing with $\sigma_{eff} \rightarrow
273: 0$ at various N.  For each N, the
274: first mass is the lowest state in that channel, and the second mass is
275: the radial excitation.}
276: \label{table:b0wcadj}
277: \end{table}
278: 
279: 
280: \begin{table}[h]
281: \centering
282: \begin{tabular}{|lc|cc|cc|cc|cc|cc|} \hline
283: N & vertex & $0^{++}$ & $0^{--}$ & $2^{\pm+}$ & $2^{\pm-}$ &
284: $3^{\pm+}$ & $3^{\pm-}$ & $4^{\pm+}$ & $4^{\pm-}$ & $1^{\pm+}$ &
285: $1^{\pm-}$ \\\hline
286: 
287: 3 & $m_{J^{PC}}$ &
288: 4.55  &  5.955  &  6.287  &  8.511  &  9.691  &   7.36  &  8.459  &
289: 10.81  &  11.86  &  6.091 \\
290:   & $m_{J^{PC}}^\star$ & 
291: 6.973  &  8.796  &  8.565 &   10.77 &   11.72  &    9.5 &   10.43 &
292: 12.63  &  13.52  &  8.793 \\\hline
293: 
294: 4 & & 
295: 4.33  &  5.901  &  6.301  &  8.463  &  9.645  &  7.439 &   8.592 &
296: 10.76  &  11.82  &  6.612 \\ & & 
297: 7.123  &  8.755  &  8.738   & 10.73  &  11.68  &  9.719  &   10.7  &
298: 12.6  &  13.48  &  9.546  \\\hline
299: 
300: 4 & & 
301: 4.328  &  4.951  &    6.6 &   7.603  &  8.831  &  7.839  &  9.075 &
302: 9.994  &  11.09  &  6.659 \\ & & 
303: 6.782  &  8.042  &   8.52  &  10.06  &  11.03  &  9.725  &  10.91 &
304: 11.96  &  12.87  &  10.22 \\\hline
305: 
306: \end{tabular}
307: \caption{The spectrum for vertex mixing with $\sigma_{eff} \rightarrow
308: 0$ at various N.  For each N, the
309: first mass is the lowest state in that channel, and the second mass is
310: the radial excitation.}
311: \label{table:b0wcver}
312: \end{table}
313: