1: \subsection{Beliaev-Popov Approximation}
2: In Sec.~\ref{pf}, we discussed in some detail the Beliaev approximation
3: at zero temperature. We recall that this approximation is defined
4: by all one-loop diagrams contributing to the self-energies.
5: It was shown that the resulting dispersion relation is gapless.
6: Popov~\cite{popov} generalized the Beliaev approximation
7: to finite temperature,
8: and hence it is
9: often called the Beliaev-Popov (B--P) approximation~\cite{grif}.
10: He calculated the one-loop diagrams in the limit of zero external
11: energy and momentum.
12: However, one has to be careful because the limits
13: $\omega\rightarrow0,\;{\bf p}\rightarrow0$
14: and ${\bf p}\rightarrow0,\;\omega\rightarrow0$ do not commute at finite
15: temperature.
16: Later,
17: Shi and Griffin have given a detailed discussion of the B--P approximation
18: and explicit formal expressions for the normal and anomalous self-energies
19: at finite temperature and arbitrary external energy $\omega$
20: and momentum ${\bf p}$.
21: An analysis of the B--P approximation similar to the one
22: presented at zero temperature in Sec.~\ref{t0} would probably be
23: significantly simpler. For instance it follows immediately from
24: the finite temperature version of Eq.~(\ref{eqs})
25: that the chemical potential
26: is free of the infrared
27: divergences that plague separately the expressions
28: for $\Sigma_{11}(0,0)$
29: and $\Sigma_{12}(0,0)$.
30: The expression for the chemical potential is
31: \bqa\nonumber
32: \mu&=&gv^2+{1\over2}g\sumint_{P}{4p^2-4\mu+6gv^2\over\omega_n^2+\epsilon^2(p)}
33: \\ &=&\nonumber
34: gv^2+{1\over4}g\left[
35: 3I_{1,1}+I_{-1,-1}\right]
36: \\ &&
37: +{1\over2\pi^2}g\int_0^{\infty}
38: dp\;{p^2\left(2p^2-2\mu+3gv^2\right)\over\epsilon(p)}n(\epsilon(p))\;.
39: \eqa
40: In the high-temperature limit, where $\beta gv^2\ll1$, one finds
41: \bqa
42: \mu&=&gv^2\left[1-{3\over8\pi^2}{T\over gv^2}\sqrt{2g^3v^2}\right]\;.
43: \label{htpt}
44: \eqa
45: Thus at high temperature, the dimensionless expansion parameter
46: has an extra factor of $T/gv^2$ compared to the zero-temperature case,
47: where it is $\sqrt{g^3v^2}$.
48:
49: So far our description of the dilute Bose gas has been in terms of
50: spontaneous breaking of the $U(1)$ symmetry and the use of a chemical
51: potential to ensure that the mean number of particles is constant.
52: There exist alternative approaches that are number-conserving.
53: Such an approach was first introduced by Giradeau and Arnowitt
54: as early as in 1959~\cite{arno}. This approach is variational in nature
55: and, like the HBF approximation,
56: it exhibits a gap in the dispersion relation.
57: However, later it was shown by Takano~\cite{taka} that the gap is removed if
58: one takes into account cubic terms in the Hamiltonian in a consistent manner.
59: More recently, Morgan~\cite{sam} has discussed in detail
60: number-conserving approaches and generalized them to trapped Bose gases.
61: The starting point is the Hamiltonian including interaction terms
62: describing binary collisions written in terms of pair operators
63: that conserve particle number. These operators are defined by
64: \bqa
65: \alpha_{i}&=&\beta_0^{\dagger}a_i\;,
66: \eqa
67: where $a_i$ are the standard annihilation operators and
68: $\beta_0=(1+a_0^{\dagger}a_0)^{-1/2}a_0$.
69: The Hamiltonian can then be divided into
70: terms containing zero, one, two, three, and four creation/annihilation
71: operators.
72: The terms that contain up to two operators are then diagonalized, while
73: the cubic and quartic terms are treated as perturbations using
74: first and second order perturbation theory.
75: The validity of perturbation theory is the usual requirement that
76: $\sqrt{na^3}\ll 1$ at $T=0$. At high temperature, it follows
77: from~(\ref{htpt}) that the requirement be
78: $T\sqrt{g^3v^2}/gv^2\ll 1$.
79: In the zero temperature limit, the calculations reproduce several
80: known results presented in Sec.~\ref{t0}. The Hamiltonian
81: can be used to calculate the ground state energy density
82: ${\cal E}$
83: (rather than the free
84: energy density ${\cal F}$) and a calculation involving the quadratic terms
85: yields the result of Bogoliubov given in Eq.~(\ref{ed}).
86: A second-order calculation
87: reproduces the $na^3\log(na^3)$-term of Wu~\cite{wu}, but one is faced with
88: a logarithmic ultraviolet divergent term which is proportional
89: to $na^3$. As we argued in Sec.~\ref{noneff},
90: the correct way to treat this divergence is to absorb
91: it in the coefficient of the operator $(\psi^*\psi)^3$, which represents
92: $3\rightarrow3$ scattering. The approach also reproduces
93: Beliaev's results for the phonon
94: spectrum~(\ref{pre}) and the damping rate~(\ref{imw}).
95: Finally, we mention that the modified Popov
96: approximation~\cite{ext1,ext2} has also been examined within the framework
97: of number-conserving approaches~\cite{sam}.
98:
99:
100:
101: