Fermionic renormalization group flows. Technique and theory

DOI10.1143/PTP.105.1MaRDI QIDQ2774393zbMATH OpenOpenAlexWikidataFDO

Authors Carsten Honerkamp, Manfred Salmhofer

Publication date 28 April 2003

Published in Progress of Theoretical Physics (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1143/ptp.105.1

Hubbard model one particle irreducible function renormalization group

PDEs in connection with quantum mechanics (35Q40) Renormalization group methods applied to problems in quantum field theory (81T17) Renormalization group methods in equilibrium statistical mechanics (82B28)

Review text

Differential equations for Wilson's renormalization group for the one-Particle Irreducible (1PI) Green function in fermionic systems are derived. The Hamiltonian \(H(a^+,a)\), with \(a^+_{{\mathbf x},\sigma}\), \(a_{{\mathbf x},\sigma}\) being the creation and annihilation operators, respectively, is a polynomial in the fermion operators. The grand canonical trace \(Z= \text{Tr }e^{-\beta(H- \mu N)}\) has a representation NEWLINE\[NEWLINEZ= \int \prod_{\tau,x,\sigma} d\overline\psi_\sigma(\tau,{\mathbf x})d\psi_\sigma(\tau,{\mathbf x}) e^{-{\mathcal A}(\overline\psi,\psi)}.NEWLINE\]NEWLINE For the Hubbard-type Hamiltonian NEWLINE\[NEWLINEH= \sum_{{\mathbf x}, {\mathbf x}',\sigma} T_{{\mathbf x}{\mathbf x}'} a^+_{{\mathbf x},\sigma} a_{{\mathbf x}',\sigma}+ \sum_{{\mathbf x},{\mathbf x}'}: n_{{\mathbf x}} V({\mathbf x},{\mathbf x}') n_{{\mathbf x}'}:,NEWLINE\]NEWLINE \({\mathcal A}\) is a sum of a quadratic and a quartic term: \({\mathcal A}={\mathcal A}_2+{\mathcal A}_4\). Then the generating function for the connected Green function is NEWLINE\[NEWLINEW(\overline\eta, \eta)= -\log\int d\mu_C(\overline\psi, \psi) e^{-{\mathcal V}(\overline\psi, \psi)+ (\overline\eta,\psi)+ (\overline\psi,\eta)},\quad{\mathcal V}={\mathcal A}_4,NEWLINE\]NEWLINE where \(\overline\eta\) and \(\eta\) are Grassmann source terms (Section 2). If \(W\) has a nondegenerate quadratic part, the map \(H\mapsto \Phi_{cl}(H)=\delta W(H)/\delta H(X)\) can be inverted. Denoting the inverse map by \(\phi\mapsto h(\phi)\), the Legendre transform of \(W\) is defined by \(\Gamma(\phi)= W(h(\phi))- (h(\phi),\phi)\). It is expanded as NEWLINE\[NEWLINE\Gamma(s|\phi)= \sum_{m\geq 0} \gamma^{(m)}(s|\phi),\quad \gamma^m(s|\phi)= {1\over m!} \int d^m \underline X\gamma_m(s|\underline X) \phi^m(\underline X),NEWLINE\]NEWLINE where \(\gamma^m(s|X_1,\dots, X_m)\) is the 1PI \(m\)-point function and \(\underline X= (X_1,\dots, X_m)\),NEWLINENEWLINENEWLINE\(\phi^m(\underline X)= \phi(X_1)\cdots \phi(X_m)\). Then the first two equations of the renormalization group differential equations of 1PI functions are derived to be NEWLINE\[NEWLINE\begin{aligned}\dot\gamma_2(s|\underline X) &= \dot{\mathbf Q}_s(X_1, X_2)+ {1\over 2} \int dX_3 dX_4 {\mathbf S}_S(X_4, X_3) \gamma_4(s|X_1, X_2, X_3, X_4),\\ \dot\gamma_4(s|\underline X) &= {1\over 2} \int dY_1 dY_2 \gamma_6(s\underline X, Y_1, Y_2){\mathbf S}_S(Y_2, Y_1)-{1\over 2} \int d^4\underline Y{\mathbf L}(\underline Y){\mathbf B}_S(\underline X,\underline Y),\end{aligned}NEWLINE\]NEWLINE where \({\mathbf Q}_s\) is the \(s\)-dependent quadratic form by which the Gaussian measure is defined, \({\mathbf S}_s\), the single-scale propagator, \({\mathbf L}\) and \({\mathbf B}_s\) are defined by using these operators and \(\gamma_4\) (precise definitions are given in the paper). These equations are not closed. The authors say equation for \(\gamma_m\) involves \(\gamma_{m+2}\). The equation for \(\Gamma\) is also given, but owing to irreducibility, as a nonpolynomial in \(\Gamma\) (Section 3). In Section 4, assuming all continuous symmetries of the action remain unbroken, consequences of symmetries are discussed. In Section 5, the last section, the influence of \(\gamma_6\) on \(\gamma_4\) is estimated. Let NEWLINE\[NEWLINEg(s)= \sup_{0\leq s'\leq s} \sup_{p_1,p_2,p_3} |\gamma_4(s'|p_1, p_2, p_3)|.NEWLINE\]NEWLINE Then the contribution \(I_a\) from a cubic term (single graph) in \(\gamma_4\) arising from the contribution of \(\gamma_6\) to the flow of \(\gamma_4\) is estimated to be NEWLINE\[NEWLINEI_a\leq A^2 Q_{\text{vol}}g(s)^3 \varepsilon_s(1+ s)^2= C(s),\quad \varepsilon_s\sim e^{-s},NEWLINE\]NEWLINE where \(Q_{\text{vol}}\) is a constant that depends on the Fermi surface geometry. By this estimate and the inequality \(\dot g(s)\leq \overline\beta_2 g(s)^2\), which holds where the influence of \(\gamma_6\) is omitted, the flow split into (i) high scales, where \(C(s)\) is not very small, (ii) intermediate scales, where \(C(s)\ll 1\) and (iii) low scales, where \(C(s)\ll 1\) fails. Physical significance of these scales is investigated. Influence of the full \(\gamma_6\) is similarly discussed. If van Hove singularity and a small curvature of the Fermi surface exists, \(\varepsilon_s\) should be replaced by \(\varepsilon_s\log(k^2_c/\varepsilon_s)\), and \(Q_{\text{vol}}\) depends on the curvature which diverges if the curvature vanishes. Influences of these phenomena are discussed in the last subsection of this section. In an appendix, susceptibilities are treated by using Taylor expansions around the Fermi surface, and superconductance, charge and spin susceptibility equations are presented.

Recommendations

Cited in

(37)

This page was built for publication: Fermionic renormalization group flows. Technique and theory

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2774393)