1: \chapter{Efimov's Function}
2: \label{ch:efimov}
3:
4: In the 1970's, Vitaly Efimov investigated the behavior of three-body systems whose particles interacted via pair-wise ``resonant'' forces. A resonant force is one that supports a shallow two-body bound state or a virtual state close to the two-body threshold, resulting in a scattering length whose absolute value is much larger than the range of the interaction. This disparity in length scales allowed him to discover properties that were independent of the exact form of the two-body interaction. One of the most useful results is the relation of the entire three-body spectrum to a single universal function.
5:
6: This chapter begins with a brief introduction to Efimov's work. We outline the arguments used to derive Efimov's universal function and exhibit the discrete scaling symmetry it possesses. Some uses of his work in the fields of nuclear and atomic physics are discussed. Since our three-body problem satisfies the constraints used by Efimov, we use the leading order integral equations to numerically compute Efimov's function to high accuracy. We also analytically derive the relationship between Efimov's function and the function $\tilde{\theta}$ defined in Eq.~(\ref{eqn:theta_tilde}).
7:
8:
9: %%%%%%%%%%%%%%%%%%
10: %% BACKGROUND %%
11: %%%%%%%%%%%%%%%%%%
12: \section{Background}
13:
14: Efimov's work hinges upon the use of a resonant two-body force. Assuming a positive scattering length, this simply means that the potential associated with this force has a shallow two-body bound state. Since $B_2 = \hbar/ma^2$ to leading order, this is the same as saying the potential has a large scattering length. Such a situation may arise either naturally (e.g., the deuteron) or artificially (e.g., alkali atoms by fine-tuning an external magnetic field). In any case, it will be assumed that the scattering length is much larger than the range of the interaction, which we call $r_0$. Efimov also assumed that there are no deeply-bound states in the two-body spectrum.
15:
16: We begin with the adiabatic hyperspherical representation of the three-body equation \cite{Nielsen:review,Federov:1993}. The hyperspherical radius $R$ is defined by
17: %
18: \begin{equation}
19: R^2 = \frac{1}{3} \left( r_{12}^2 + r_{13}^2 + r_{23}^2 \right) ,
20: \end{equation}
21: %
22: \noindent where $r_{ij} = |\vec{r_i} - \vec{r_j}|$. In the region $r_0 \ll R \ll a$, the three-body Schr\"odinger equation for a zero angular momentum state takes the form
23: %
24: \begin{equation}
25: - \frac{\hbar^2}{2 m} \left[ \frac{\partial^2}{\partial R^2} + \frac{s_0^2 + 1/4}{R^2} \right] f(R) = E \, f(R) \label{eqn:hamil}.
26: \end{equation}
27: %
28: \noindent Here, $f(R)$ is the radial wavefunction and $s_0 \approx 1.00624$ is the same constant seen in previous chapters. The goal is to construct solutions for the bound state energies $E = -B_3$.
29:
30: Since we are interested only in low-energy states ($|E| \sim 1/ma^2$), the requirement $R \ll a$ implies that the $1/R^2$ potential is much larger than $E$. Therefore, we can ignore the right-hand side of Eq. (\ref{eqn:hamil}). The resulting equation has a general solution of the form
31: %
32: \begin{equation}
33: f(R) = \sqrt{H R} \left[ A\,\me^{\mi s_0 \ln(H R)} + B\,\me^{-\mi s_0 \ln(H R)} \right] ,
34: \end{equation}
35: %
36: \noindent where $H$ has dimensions of momentum, and $A$ and $B$ are dimensionless. For bound state solutions, these quantities must be functions of the bound-state energy $B_3$ and the two-body scattering length $a$. Let us define
37: %
38: \begin{eqnarray}
39: H & = & \sqrt{\frac{m B_3}{\hbar^2} + \frac{1}{a^2}} \label{eqn:Hdef},
40: \\
41: \xi & = & - \arctan(a \sqrt{m B_3}/\hbar) \label{eqn:xidef}.
42: \end{eqnarray}
43: %
44: \noindent This definition of $\xi$ is valid only for positive scattering lengths. For $a < 0$, we must use the definition $\xi = - \arctan(a \sqrt{m B_3}/\hbar) - \pi$. In either case though, the dimensionless quantities $A$ and $B$ can depend upon $\xi$.
45:
46: The radial wavefunction in the region $r_0 \ll R \ll a$ is a sum of incoming and outgoing hyperspherical waves. At short distances ($R \sim r_0$) and long distances ($R \sim a$) the wavefunction is more complicated, and it is the boundary conditions at these points that determine the actual bound-state energies.
47:
48: We first consider the constraints from unitarity in the short-distance region $R \sim r_0$. Because there are no deeply-bound two-body states, all the probability associated with the incoming hyperspherical wave must be reflected into the outgoing wave. This implies that $A$ and $B$ can only differ by a phase: $B = -A \, \me^{\mi \theta}$. The wavefunction in the region $r_0 \ll R \ll a$ is now
49: %
50: \begin{equation}
51: f(R) \propto \sqrt{H R} \, \sin\left( s_0 \, \ln(H R) - \theta/2 \right),
52: \end{equation}
53: %
54: \noindent so we may express the phase as
55: %
56: \begin{equation}
57: - \frac{\theta}{2} = -s_0 \, \ln(HR) + \mathrm{arccot}\left[ \frac{1}{s_0}\left(R \frac{f'(R)}{f(R)} - \frac{1}{2}\right) \right] .
58: \end{equation}
59: %
60: \noindent The phase is determined by specifying the derivative $R_0 f'(R_0)/f(R_0)$ at any point $R_0$ satisfying $r_0 \ll R_0 \ll a$. We will simplify the form by writing it as
61: %
62: \begin{equation}
63: \theta = 2 s_0 \ln\left(H / \Lambda_*\right) .
64: \end{equation}
65: %
66: \noindent The quantity $\Lambda_*$ is a complicated function of $R_0$ and $R_0 f'(R_0)/f(R_0)$, the details of which are unimportant. What is important is that the derivative at $R_0$ is ultimately determined by the short-distance behavior. Essentially, $\Lambda_*$ parameterizes the effects of all short-distance interactions without knowledge of the detailed forms.
67:
68: We now consider the constraints from unitarity in the long-distance region $R \sim a$. Because we are considering three-body bound states, all of the probability associated with the outgoing hyperspherical wave must be reflected into the incoming wave. This implies that $A$ and $B$ differ only by a phase: $B = -A \, \me^{\mi \Delta}$. The phase $\Delta$ is determined by the long-range parameters $B_3$ and $a$. We will write this phase as $\Delta(\xi)$ to explicitly show its dependence upon $B_3$ and $a$ through the variable $\xi$.
69:
70: The two phases $\theta$ and $\Delta$ must match to within an additive multiple of $2 \pi$, so
71: %
72: \begin{equation}
73: 2 s_0 \ln\left(\frac{H}{\Lambda_*}\right) = \Delta(\xi) + 2 \pi n ,
74: \end{equation}
75: %
76: \noindent where $n$ is an integer. Using Eq.~(\ref{eqn:Hdef}), we rewrite the relation in the form
77: %
78: \begin{equation}
79: B_3 + \frac{\hbar^2}{m a^2} = \frac{\hbar^2 \Lambda_*^2}{m} \me^{2 \pi n/s_0} \exp[ \Delta(\xi) / s_0 ] \label{eqn:efunc}.
80: \end{equation}
81: %
82: \noindent The values for $n$ correspond to different 3-body bound states. Once the function $\Delta(\xi)$ is known, we are able to calculate the entire spectrum by solving Eq.~(\ref{eqn:efunc}) for various $n$. $\Delta(\xi)$ is what we have referred to as Efimov's universal function.
83:
84: The variables $H$ and $\xi$ in our equations can be treated like a ``radius'' and an ``angle'' respectively. Suppose we plot a 3-body bound-state energy on a graph, choosing the x-axis to be $1/a$ and the y-axis to be $- \sqrt{m B_3/\hbar^2}$ (See Fig.~\ref{fig:HvsXi}). Then $H$ is the distance of a line from that bound-state point to the origin, and $\xi$ is the angle that line makes with the x-axis. Figure \ref{fig:EfimovStates} shows two Efimov states as a function of the 2-body binding energy which is proportional to $1/ma^2$ to leading order. In this figure, we have chosen to label the axes differently so that the behavior of both states can be better seen.
85:
86: Using these parameters, we can demonstrate that the solutions to Efimov's equation have a discrete scaling symmetry. If there exists a bound state with binding energy $B_3$ for the parameters $a$ and $\Lambda_*$, then there will also be a bound state with binding energy $\lambda^2 B_3$ for the parameters $\lambda^{-1} a$ and $\Lambda_*$ provided that $\lambda$ is of the form $\exp[n' \pi / s_0]$ for some integer $n'$. This is equivalent to saying that for any given value of $\xi$, the values of $H$ for any successive bound states differ by a multiplicative factor of $\exp(\pi/s_0) \simeq 22.7$. Since we are considering only the case $a > 0$ with three-body binding energies that satisfy $B_3 > B_2$, the variable $\xi$ is restricted to the range $-\pi/2 \le \xi \le -\pi/4$. This periodic behavior can be easily seen by examining the value of the phase in $f_{d0}$ along a path of constant $\xi$. Figure \ref{fig:H-1} shows the value of $\theta$ for bound states with $B_2/B_3 = 0.5$ as $H$ increases. Successive Efimov states must have the same phase which is periodic in $\ln(H)$.
87:
88: It should be noted that Eq.~(\ref{eqn:efunc}) is an approximation based upon the limit $r_0/a = 0$, and as such is exact only for zero-range theories. Any calculated energies can be expected to have errors of $\mathcal{O}(r_0/a)$. Some first order corrections due to a non-zero effective range have been calculated by Efimov \cite{efimov:3,efimov:7}.
89:
90:
91: %%%%%%%%%%%%%%%%%%%%
92: %% APPLICATIONS %%
93: %%%%%%%%%%%%%%%%%%%%
94: \subsection{Applications}
95:
96: Efimov's general framework may be applied whenever there is a resonant two-body interaction. Several instances where it is applicable:
97:
98: \begin{itemize}
99: \item The two-nucleon system has a shallow bound state, the deuteron, in the spin-triplet channel and a large negative scattering length in the spin-singlet channel. This led Efimov to suggest that the few-nucleon system could be described using zero-range potentials with the effective range treated as a perturbation. The three-nucleon system includes two bound states: the triton (a $pnn$ bound state) and ${}^3\mathrm{He}$ (a $ppn$ bound state). The equations are more complicated since the nucleons possess spin and isospin, and the $ppn$ state is further complicated by the Coulomb interaction between the two protons. A leading order analysis with the Coulomb interaction neglected has been carried out by Efimov \cite{efimov:2,efimov:5} and revisited in the EFT framework by Bedaque, Hammer, and van Kolck \cite{hammer:orig}. An analysis at NLO in the effective range was carried out by Efimov \cite{efimov:3} and repeated by Hammer and Mehen using EFT \cite{Mehen:r0}.
100:
101: \item ${}^4\mathrm{He}$ atoms have a large two-body scattering length and a shallow two-body bound state. The ground state of the ${}^4\mathrm{He}$ trimer has been observed, but its binding energy has not been measured. The Schr\"odinger equation for the three-body bound states has been solved accurately for potential models of the interaction between ${}^4\mathrm{He}$ atoms. In addition to the ground state trimer, there is an excited state this is shallower by a factor of about 50 to 70. It has been found that Efimov's function can be used to predict the binding energy of one of the two 3-body bound states using the other as input \cite{hammer:orig,Braaten:HeAtoms,Frederico:HeAtoms}.
102:
103: \item For alkali atoms, the atom-atom scattering length can be made large by tuning and external magnetic field to a Feshbach resonance \cite{Verhaar:1,Verhaar:2}. One complication in this case is that the alkali atoms have many deeply-bound two-body states, so an Efimov state can decay into an atom and a deep two-body bound state. As a consequence, the Efimov states are resonances with a binding energy $B_3$ and a width $\Gamma_3$. The complex energies $-(B_3 + \mi \Gamma_3)$ still satisfy Efimov's equation, but it requires the analytic continuation of the function $\Delta(\xi)$ to complex values of the angle $\xi$ \cite{Braaten:uefes}.
104:
105: \end{itemize}
106:
107:
108: %%%%%%%%%%%%%%%%%%%
109: %% COMPUTATION %%
110: %%%%%%%%%%%%%%%%%%%
111: \section{Computation of Efimov's Function}
112:
113: Although Efimov explicitly considered only resonant two-body interactions, his conclusions result from applying a boundary condition on the three-body wavefunction at short distances. This encompasses any short-range interactions, including the three-body contact interaction used in our work. The $g_3$ potential in our equations only acts when all three particles are very close together. Using our previous notation, this would correspond to the region $R \sim r_0$, which is inside the radius where the boundary condition matching occurs. Its influence is combined with that of the two-body interaction, and the total effect is seen only through the parameter $\Lambda_*$.
114:
115:
116: %%%%%%%%%%%%%%
117: %% B2 = 0 %%
118: %%%%%%%%%%%%%%
119: \subsection{$B_2 = 0$}
120: As a verification of this argument, let us consider the case $B_2 = 0$. According to Eq.~(\ref{eqn:efunc}), the bound-state energies are given by
121: %
122: \begin{equation}
123: B_3 = \frac{\hbar^2 \Lambda_*^2}{m} \me^{2 \pi n/s_0} \exp[ \Delta(-\pi/2) / s_0 ] ,
124: \end{equation}
125: %
126: \noindent showing that the ratio of energies for adjacent states is $\exp(2 \pi/s_0)$. To 13 digits, this ratio is 515.0350013848. In Table \ref{tab:EfimovSpacing}, we have computed several of the energies using $B_3 = 1.0$ as a reference point. The table also shows the ratios between adjacent states, all of which equal the predicted value to 11 digits.
127:
128: \begin{table}
129: \begin{center}
130: \begin{tabular}{|c|l|l|}
131: \hline
132: Level & Energy & Ratio \\
133: \hline
134: \hline
135: $+3$ & 1.366187266197138e+08 & 515.0350013845557 \\
136: $+2$ & 2.652610526516549e+05 & 515.0350013850982 \\
137: $+1$ & 5.150350013849171e+02 & 515.0350013849171 \\
138: $\hspace{8pt}0$ & 1.000000000000000 & 515.0350013849775 \\
139: $-1$ & 1.941615613134847e-03 & 515.0350013851568 \\
140: $-2$ & 3.769871189167697e-06 & 515.0350013846162 \\
141: $-3$ & 7.319640760400369e-09 & - \\
142: \hline
143: \end{tabular}
144: \end{center}
145: \caption[Efimov state energies for $B_2 = 0$.]{\label{tab:EfimovSpacing}Efimov state energies for $B_2 = 0$. The value 1.0 is fixed as a starting point. Notice that the ratio of energies for adjacent states equals $\exp(2\pi/s_0)$ to 11 digits.}
146: \end{table}
147:
148: We can prove that the ratio of adjacent binding energies is $\exp(2 \pi/s_0)$ using Eq.~(\ref{eqn:theta_tilde}). Two adjacent bound states, $B_3$ and $B_3'$, must have phases that differ by $\pi$. Therefore,
149: %
150: \begin{eqnarray}
151: && s_0 \ln\left( \Lambda/\sqrt{B_3}\right) = s_0 \ln\left( \Lambda/\sqrt{B_3'}\right) + \pi \nonumber
152: \\
153: & \Longrightarrow & s_0 \ln\left( \sqrt{B_3'}/\sqrt{B_3}\right) = \pi \nonumber
154: \\
155: & \Longrightarrow & \frac{B_3'}{B_3} = \me^{2 \pi/s_0} .
156: \end{eqnarray}
157:
158:
159: %%%%%%%%%%%%%%%%%%%%%%%%%%
160: %% Phase to Delta(xi) %%
161: %%%%%%%%%%%%%%%%%%%%%%%%%%
162: \subsection{From $\theta$ to $\Delta$}
163:
164: Before we can compute Efimov's function, we must find a way to relate the parameter $\Lambda_*$ to the parameters in our equations. We will do this by considering the behavior of the phase $\theta$ from the perspective of both the high- and low-momentum equations.
165:
166: First consider the high-momentum perspective. We have seen in Sec.~\ref{sec:analytic} that it is possible to view $\theta$ as a function of $\delta_0$, but it can just as easily be viewed as a function of $G_3$. For a given value of $G_3$, we write the phase generated by this coupling as $\theta_h(G_3)$.
167:
168: From the low-momentum perspective, the phase is determined only by the parameters $B_2$, $B_3$, and $\Lambda$. We also know explicitly the cutoff dependence of this phase:
169: %
170: \begin{equation}
171: \theta_l(B_2, B_3, \Lambda) = s_0 \ln\left(\Lambda/\sqrt{B_3}\right) + \tilde{\theta}(B_2/B_3).
172: \end{equation}
173: %
174: \noindent Because these phases must match to within a multiple of $\pi$,
175: %
176: \begin{equation}
177: \theta_h(G_3) = \theta_l(B_2, B_3, \Lambda) + n \pi = s_0 \ln\left(\Lambda/\sqrt{B_3}\right) + \tilde{\theta}(B_2/B_3) + n \pi \label{eqn:ThetaHL}.
178: \end{equation}
179: %
180: \noindent The purpose of $G_3$ is to make some value of $B_3$ cutoff-independent. As the cutoff is changed, $G_3$ is adjusted to ensure that $B_3$ remains the same. Removing $\Lambda$ from Eq.~(\ref{eqn:ThetaHL}) can only be done if $\theta_h(G_3)$ contains $\Lambda$ dependence of the form $s_0 \ln(\Lambda)$ implicitly through $G_3$. We use this fact to define a new parameter $\Lambda_*$ via the equation
181: %
182: \begin{equation}
183: \theta_h(G_3) = s_0 \ln\left(\Lambda/\Lambda_*\right) .
184: \end{equation}
185: %
186: \noindent Eq.~(\ref{eqn:ThetaHL}) can then be written as
187: %
188: \begin{equation}
189: s_0 \ln\left(\Lambda/\Lambda_*\right) = s_0 \ln\left(\Lambda/\sqrt{B_3}\right) + \tilde{\theta}(B_2/B_3) + n \pi ,
190: \end{equation}
191: %
192: \noindent which implies
193: %
194: \begin{equation}
195: B_3 = \Lambda_*^2 \me^{2 \pi n/s_0} \me^{2 \tilde{\theta}/s_0} \label{eqn:B3Delta}.
196: \end{equation}
197: %
198: \noindent The ratio $B_2/B_3$ can be written in terms of $\xi$:
199: %
200: \begin{equation}
201: B_2/B_3 = 1/\tan^2(-\xi) \label{eqn:ratioxi}.
202: \end{equation}
203: %
204: \noindent Since $\tilde{\theta}$ is a function of $B_2/B_3$, it is also a function of $\xi$, and from now on we will write $\tilde{\theta}(\xi)$ to emphasize this. Multiplying both sides of Eq.~(\ref{eqn:B3Delta}) by $(1 + B_2/B_3)$ yields
205: %
206: \begin{equation}
207: B_3 + B_2 = \Lambda_*^2 \me^{2 \pi n/s_0} \exp\left[ \tilde{\theta}(\xi) - 2\ln(\sin(-\xi))\right] \label{eqn:myefunc}.
208: \end{equation}
209: %
210: \noindent We see that we can match Efimov's relation if we choose
211: %
212: \begin{eqnarray}
213: \Lambda_* & = & \Lambda \me^{-\theta/s_0} \label{eqn:LamStarDef},
214: \\
215: \Delta(\xi) & = & \tilde{\theta}(\xi) - 2\ln(\sin(-\xi)) \label{eqn:DeltaDef}.
216: \end{eqnarray}
217: %
218: \noindent We should note that this choice for $\Lambda_*$ is proportional to the quantity $\Lambda_*$ defined in Ref.~\cite{hammer:orig} and will result in a function $\Delta(\xi)$ that differs from the one found in \cite{Braaten:uefes} by an inconsequential additive constant.
219:
220:
221: %%%%%%%%%%%%%%%%%%%%%%%%
222: %% NUMERICAL VALUES %%
223: %%%%%%%%%%%%%%%%%%%%%%%%
224: \subsection{Numerical Values}
225:
226: Now that we have relations (\ref{eqn:LamStarDef}) and (\ref{eqn:DeltaDef}), we can numerically compute values for $\Delta(\xi)$. One way to do this is to follow a state with constant $\Lambda_*$ as $B_2$ is changed. For a fixed cutoff, constant $\Lambda_*$ implies a constant $\theta$. The two Efimov states in Fig.~\ref{fig:EfimovStates} are for constant phase, so they may be used to calculate $\Delta(\xi)$. The results are shown in Fig.~\ref{fig:ConstPhaseDelta}, where we have plotted $\Delta(\xi) - \Delta(-\pi/2)$ for both states. The constant shift in the function does not matter.
227:
228: A second alternative is to keep $B_3$ constant and follow any changes in the phase as $B_2$ changes. The function computed from this approach is shown in Fig.~\ref{fig:ConstEnergyDelta} for the case of $B_3 = 1.0$. This method makes it easier to compute $\Delta(\xi)$ for any given $\xi$ since we do not need to search for the value of the three-body binding energy.
229:
230: Since the energy and phase values are accurate to about 12 digits, we expect similar accuracy in our calculation of $\Delta(\xi)$. Figure \ref{fig:DeltaDiff-1} shows the difference between the Efimov function values generated from the Efimov state in Fig.~\ref{fig:EfimovStates} ($B_2 = 0.0, B_3 = 1.0$) and the constant energy $B_3 = 1.0$ state. The errors support the statement that our calculation of $\Delta(\xi)$ is accurate to almost 12 digits, which is much higher than previous calculations \cite{Braaten:uefes}. Table~\ref{tab:DeltaCompare} compares those previous values, labelled $\Delta(\xi)_{\mathrm{BHK}}$, to the ones obtained here. The definition of these functions differ by an additive constant: $\Delta(\xi) = \Delta(\xi)_{\mathrm{BHK}} + C$. We have fixed the constant $C$ by demanding that $\Delta(\xi) = \Delta(\xi)_{\mathrm{BHK}}$ when $\xi = -1.502$. The discrepancies are less than 0.02 with the exception of the final point at $\xi = -0.787$ where the discrepancy increases to about 0.06.
231:
232:
233: %%%%%%%%%%%%%%%%%%%%%%%%
234: %% NUMBER OF STATES %%
235: %%%%%%%%%%%%%%%%%%%%%%%%
236: \section{Number of Efimov States}
237:
238: For any given value of $B_2$, there will be a certain number of Efimov states that lie within the range of validity for this approximation. Efimov's estimate for the number of states $N$ is related to the two-body scattering length $a$ and the range of the interaction $r_0$ by the simple formula
239: %
240: \begin{equation}
241: N = \frac{s_0}{\pi} \, \ln\left( a/r_0 \right) .
242: \end{equation}
243: %
244: \noindent This is easily derived from Eq.~(\ref{eqn:myefunc}). Assume a non-zero value for $B_2$, which is related to the scattering length by $B_2 \propto a^{-2}$. The deepest bound-state energy $B_3$ can only be of order $\Lambda^2$ before the conditions of our approximation are violated. This state will correspond to $n = N$, and we will write the energy as $B_3^{(N)} \sim \Lambda^2$. The shallowest bound-state energy, corresponding to $n = 0$, must be greater than or equal to $B_2$ and should be of the same order. We write this as $B_3^{(0)} \sim B_2$. Therefore, we have the relations
245: %
246: \begin{eqnarray}
247: B_3^{(N)} + B_2 & = & \Lambda_*^2 \me^{2 \pi N/s_0} \exp(\Delta(\xi_N)/s_0) \label{eqn:Nstate1},
248: \\
249: B_3^{(0)} + B_2 & = & \Lambda_*^2 \, \exp(\Delta(\xi_0)/s_0) \label{eqn:0state1},
250: \end{eqnarray}
251: %
252: \noindent where $\xi_N = -\arctan\left(\sqrt{B_3^{(N)}/B_2}\right)$ and $\xi_0 = -\arctan\left(\sqrt{B_3^{(0)}/B_2}\right)$. Since these states are part of the same spectrum, the value of $\Lambda_*$ in both equations must be identical.
253:
254: Because we are assuming that $B_3^{(0)} \sim B_2$, we shall simplify Eq.~(\ref{eqn:0state1}) by making the substitution $B_3^{(0)} + B_2 \simeq 2 B_2$. Upon taking the ratio of (\ref{eqn:Nstate1}) and (\ref{eqn:0state1}), we obtain
255: %
256: \begin{equation}
257: \frac{B_3^{(N)}}{2 B_2} \simeq \me^{2 \pi N/s_0} \exp\left[\left(\Delta(\xi_N) - \Delta(\xi_0)\right)/s_0\right] .
258: \end{equation}
259: %
260: \noindent The exponential will be $\mathcal{O}(1)$, so we may write
261: %
262: \begin{equation}
263: N \simeq \frac{s_0}{2 \pi} \ln\left(\frac{B_3^{(N)}}{2 B_2}\right) .
264: \end{equation}
265: %
266: \noindent The $2 B_2$ term in the logarithm can be approximated by $a^{-2}$. The cutoff $\Lambda$ is related to the interaction range by $\Lambda \sim r_0^{-1}$, which allows us to estimate $B_3^{(N)} \sim r_0^{-2}$. This makes our estimate of the number of states
267: %
268: \begin{equation}
269: N \simeq \frac{s_0}{2 \pi} \ln\left(\frac{a^2}{r_0^2}\right) = \frac{s_0}{\pi} \ln\left(a/r_0\right)
270: \end{equation}
271: %
272: \noindent which matches the estimate given by Efimov.
273:
274: The source of this estimate can also be seen graphically. Recall Figs.~\ref{fig:FL0-Spectrum2} and \ref{fig:FL0-Spectrum1} where it is shown that successive bound states match onto the cosine behavior at successive peaks. The x-axis in these figures is labelled by $x = s_0 \ln(p/\Lambda)$ so that the cosine behavior of $f_{l0}$ as it approaches $f_{d0}$ can easily be seen. Each bound state joins the cosine curve at roughly $x = s_0 \ln(\eta_3/\Lambda)$. If $\eta_3^{(0)}$ represents the shallowest state, then it joins the peak at $ x_0 \simeq s_0 \ln(\eta_3^{(0)}/\Lambda)$. From Fig.~\ref{fig:FH0-2}, we see that the function $f_{h0}$ decays exponentially when $p \sim \Lambda$. The last cosine peak is around $x_N \simeq s_0 \ln(\Lambda/\Lambda) = 0$. The number of peaks between $x_0$ and $x_N$ then gives the number of Efimov states that can exist:
275: %
276: \begin{equation}
277: N = \frac{1}{\pi} (x_N - x_0) = - \frac{s_0}{\pi} \ln(\eta_3^{(0)}/\Lambda) .
278: \end{equation}
279: %
280: \noindent Using $\Lambda \sim r_0^{-1}$ and $\eta_3^{(0)} \sim a^{-1}$ leads to the the same estimate $N \simeq (s_0/\pi)\ln(a/r_0)$.
281:
282:
283:
284: %%%%%%%%%%%%%%%
285: %% FIGURES %%
286: %%%%%%%%%%%%%%%
287: \begin{table}
288: \begin{center}
289: \begin{tabular}{|c|c|c|}
290: \hline
291: $\xi$ & $\Delta(\xi) + C$ & $\Delta(\xi)_{\mathrm{BHK}}$ \\
292: \hline
293: \hline
294: -0.787 & -2.59833945225 & -2.539 \\
295: -0.791 & -2.91699158253 & -2.897 \\
296: -0.797 & -3.21158697182 & -3.194 \\
297: -0.804 & -3.46384093607 & -3.448 \\
298: -0.820 & -3.88981806008 & -3.864 \\
299: -0.836 & -4.21442979183 & -4.196 \\
300: -0.852 & -4.48377281174 & -4.469 \\
301: -0.868 & -4.71685784858 & -4.701 \\
302: -0.899 & -5.09933733420 & -5.076 \\
303: -0.933 & -5.44804603324 & -5.434 \\
304: -0.965 & -5.72971959109 & -5.712 \\
305: -1.019 & -6.13316218831 & -6.123 \\
306: -1.065 & -6.42475682784 & -6.415 \\
307: -1.104 & -6.64354741727 & -6.634 \\
308: -1.166 & -6.94919952087 & -6.943 \\
309: -1.214 & -7.15708283017 & -7.151 \\
310: -1.296 & -7.46636472602 & -7.461 \\
311: -1.347 & -7.63469982709 & -7.632 \\
312: -1.408 & -7.81593232651 & -7.814 \\
313: -1.443 & -7.91119231675 & -7.910 \\
314: -1.482 & -8.01059689781 & -8.009 \\
315: -1.502 & -8.05900000000 & -8.059 \\
316: \hline
317: \end{tabular}
318: \end{center}
319: \caption[Comparison of Efimov function $\Delta(\xi)$ to previously computed values.]{\label{tab:DeltaCompare}Comparison of Efimov function $\Delta(\xi)$ to previously computed values. The values $\Delta(\xi)_{\mathrm{BHK}}$ are those computed by Braaten, Hammer, and Kusunoki \cite{Braaten:uefes}. The additive constant $C$ was chosen so that the two functions agree at $\xi = -1.502$.}
320: \end{table}
321:
322: \begin{figure}
323: \begin{center}
324: \setlength{\unitlength}{1.0in}
325: \begin{picture}(4,4)
326: \thicklines
327: \put(0.5,4.0){\vector(0,-1){4.0}}
328: \put(0.0,3.5){\vector(1,0){4.0}}
329: \put(2.15,1.0){\circle*{0.1}}
330: \put(0.5,3.5){\vector(2,-3){1.6}}
331: \qbezier(1.0,3.5)(1.0,3.22)(0.73,3.2)
332: \put(3.5,3.3){$a^{-1}$}
333: \put(0.7,0.2){$-\sqrt{m B_3/\hbar^2}$}
334: \put(1.8,1.8){$H = \sqrt{m B_3/\hbar^2 + a^{-2}}$}
335: \put(1.1,3.1){$\xi$}
336: \put(0.0,3.7){$(0,0)$}
337: \end{picture}
338: \end{center}
339: \caption[Relation between bound-state energy and the variables $H$ and $\xi$.]{\label{fig:HvsXi}Relation between bound-state energy and the variables $H$ and $\xi$.}
340: \end{figure}
341:
342: \begin{figure}
343: \begin{center}
344: \includegraphics[height=4.0in]{figures/EfimovStates.ps}
345: \end{center}
346: \caption[Binding energies of Efimov states as functions of the two-body binding energy.]{\label{fig:EfimovStates}Binding energies of Efimov states as functions of the two-body binding energy. The dotted line is the scattering threshold.}
347: \end{figure}
348:
349: \begin{figure}
350: \begin{center}
351: \includegraphics[height=4.0in]{figures/H-1.ps}
352: \end{center}
353: \caption[Phase $\theta$ for three-body bound states as a function of $H$ with $B_2/B_3$ = 0.5.]{\label{fig:H-1}Phase $\theta$ for three-body bound states as a function of $H$ with $B_2/B_3$ = 0.5. The periodic behavior as $H$ increases is a direct result of the discrete scaling symmetry in Efimov's equation. For a constant ratio $B_2/B_3$, the phase is linear in $\ln(H)$.}
354: \end{figure}
355:
356: \begin{figure}
357: \begin{center}
358: \includegraphics[height=4.0in]{figures/ConstPhaseDelta.ps}
359: \end{center}
360: \caption[The Efimov function $\Delta(\xi)$ as a function of $\xi$ computed using three-body bound states with constant phase.]{\label{fig:ConstPhaseDelta}The Efimov function $\Delta(\xi)$ as a function of $\xi$ computed using three-body bound states with constant phase. The difference $\Delta(\xi) - \Delta(-\pi/2)$ is plotted since a constant shift in the function is inconsequential.}
361: \end{figure}
362:
363: \begin{figure}
364: \begin{center}
365: \includegraphics[height=4.0in]{figures/ConstEnergyDelta.ps}
366: \end{center}
367: \caption{\label{fig:ConstEnergyDelta}The Efimov function $\Delta(\xi)$ as a function of $\xi$ computed using a three-body bound state with constant binding energy $B_3 = 1$.}
368: \end{figure}
369:
370: \begin{figure}
371: \begin{center}
372: \includegraphics[height=4.0in]{figures/DeltaDiff-1.ps}
373: \end{center}
374: \caption[Difference $\Delta_{err}$ between the Efimov functions calculated using constant phase and constant energy as a function of $\xi$.]{\label{fig:DeltaDiff-1}Difference $\Delta_{err}$ between the Efimov functions calculated using constant phase and constant energy as a function of $\xi$. The constant phase Efimov state with $B_3 = 1.0$ at $B_2 = 0.0$ is compared to the constant energy state $B_3 = 1.0$. The difference in the values suggests that our calculation is accurate to almost 12 digits.}
375: \end{figure}