1: \documentclass[prl,amsmath,amssymb,superscriptaddress,twocolumn]{revtex4}
2: \usepackage{graphicx}
3: %\documentstyle[aps]{revtex}
4: %\documentstyle[preprint,aps,epsf,floats,eqsecnum,amsmath,bm]{revtex}
5: \newcommand{\al}{\alpha}
6: \newcommand{\be}{\beta}
7: \newcommand{\ba}{{\bf a}}
8: \newcommand{\br}{{\bf r}}
9: \newcommand{\bx}{{\bf x}}
10: \newcommand{\by}{{\bf y}}
11: \newcommand{\bk}{{\bf k}}
12: \newcommand{\bl}{{\bf l}}
13: \newcommand{\bj}{{\bf j}}
14: \newcommand{\bp}{{\bf p}}
15: \newcommand{\bu}{{\bf u}}
16: \newcommand{\bv}{{\bf v}}
17: \newcommand{\bq}{{\bf q}}
18: \newcommand{\bA}{{\bf A}}
19: \newcommand{\bB}{{\bf B}}
20: \newcommand{\bE}{{\bf E}}
21: \newcommand{\bK}{{\bf K}}
22: \newcommand{\bQ}{{\bf Q}}
23: \newcommand{\bP}{{\bf P}}
24: \newcommand{\bR}{{\bf R}}
25: \newcommand{\es}{{e^{\ast}}}
26: \newcommand{\ms}{{m^{\ast}}}
27: \newcommand{\psis}{{\psi^{\ast}}}
28: \newcommand{\transq}{\left[\delta_{ij}-\frac{q_iq_j}{\bq^2}\right]}
29: \newcommand{\eps}{\epsilon}
30: \newcommand{\tomega}{\tilde{\omega}}
31: \newcommand{\xig}{{\xi^{>}}}
32: \newcommand{\xil}{{\xi^{<}}}
33: \newcommand{\chig}{{\chi^{>}}}
34: \newcommand{\chil}{{\chi^{<}}}
35: \newcommand{\xiv}{\mbox{\boldmath$\xi$}}
36: \newcommand{\psitil}{\tilde{\psi}}
37: \newcommand{\psidagtil}{{\tilde{\psi}}^{\dagger}}
38: \newcommand{\psidag}{\psi^{\dagger}}
39: \newcommand{\psidot}{\dot{\psi}}
40: \newcommand{\psidotdag}{\dot{\psi}^{\dagger}}
41: \DeclareMathAlphabet{\mathpzc}{OT1}{pzc}{m}{it} \pagestyle{empty}
42: \begin{document}
43: \title{Renormalization group approach to 2D Coulomb interacting Dirac fermions with random gauge potential}
44: \author{Oskar Vafek}
45: \author{Matthew J. Case}
46: %\altaffiliation[Also at ]{Department of Physics, FSU}
47: \affiliation{National High Magnetic Field Laboratory and Department
48: of Physics, Florida State University, Tallahassee, Florida 32306,
49: USA}
50: \address{}
51: \date{\today}
52: \begin{abstract}
53: We argue that massless Dirac particles in two spatial dimensions
54: with $1/r$ Coulomb repulsion and quenched random gauge field are
55: described by a manifold of fixed points which can be accessed
56: perturbatively in disorder and interaction strength, thereby
57: confirming and extending the results of arXiv:0707.4171. At small
58: interaction and small randomness, there is an infra-red stable fixed
59: curve which merges with the strongly interacting infra-red unstable
60: line at a critical endpoint, along which the dynamical critical
61: exponent $z=1$.
62: \end{abstract}
63:
64: \maketitle
65:
66: The properties of two dimensional massless Dirac fermions have
67: recently sprung back into focus, largely due to the experimental
68: discovery of the quantum Hall effect in
69: graphene\cite{Novoselov:2005fk,Zhang:2005uq}, the single layer
70: graphite. Moreover, the ability to control the density of carriers
71: by the electrical field effect allows experimental access to the
72: rich physics of the neutrality point, where in the clean
73: non-interacting picture, the conduction and the valence bands touch.
74: It is well known\cite{Gonzales1994} that, at the neutrality point,
75: the exchange self-energy gives a logarithmic enhancement of the
76: Fermi velocity $v_F\rightarrow v_F+(e^2/4\epsilon_d)\ln(\Lambda/k)$
77: where $k$ is a small wavevector near the nodal
78: point\cite{Gonzales1994} and $\eps_d$ is the dielectric constant of
79: the medium. Physically, this effect is due to the lack of screening
80: of the $1/r$ Coulomb interaction, an important consequence of which
81: is the suppression of the single particle density of states ($N(E)$)
82: at low energies. This in turn leads to the suppression of the
83: electronic contribution to the low temperature specific
84: heat\cite{vafek:prl2007}.
85:
86: This suppression of $N(E)$ may lure one into the (incorrect)
87: conclusion that, at $T=0$, the Coulomb interactions turn the clean
88: system into an electrical insulator. However, the vertex corrections
89: contribute an {\it exactly} compensating enhancement of the
90: conductivity\cite{HerbutJuricicVafek2007}, making the system a {\it
91: metal} with its residual conductivity asymptotically equal to the
92: non-interacting value $\sigma_0=(\pi/8)e^2/h$ per node.
93:
94: In this work, we analyze the effects of the unscreened Coulomb
95: interactions and the quenched random gauge disorder beyond leading
96: order in the perturbative renormalization group (RG) of Ref.
97: \cite{HerbutJuricicVafek2007}. Our principle findings, which support
98: and extend those of Ref. \cite{HerbutJuricicVafek2007} are twofold:
99: first, in the clean case, there is an unstable fixed point at finite
100: strength of Coulomb interactions characterized by the dimensionless
101: ratio $\alpha=e^2/(\hbar \eps_dv_F)$ which represents a quantum
102: critical point (QCP) separating the semimetal from an excitonic
103: insulator; and second, the interplay between Coulomb interactions
104: and disorder induces a downward curvature of the fixed
105: line\cite{stauber2005,HerbutJuricicVafek2007}, causing it to end at
106: the clean QCP (see Fig.1).
107:
108: \begin{figure}[h]
109: \begin{center}
110: \begin{tabular}{c}
111: \includegraphics[width=0.325\textwidth]{rgCase1.eps}
112: \end{tabular}
113: \end{center}
114: \caption{The renormalization group flow diagram in the (disorder)
115: $\Delta$ -- (interaction) $\alpha=e^2/(\eps_d v_F)$ plane. There is
116: a line of stable fixed points at small $\Delta$ and small $\alpha$
117: which merges with the line of unstable fixed points at the critical
118: end point. The (clean) unstable fixed point at $\alpha_c$
119: corresponds to a quantum phase transition into an excitonic
120: insulator. Above the critical $\Delta^*$, the disordered but
121: non-interacting fixed line is unstable directly to the insulator.}
122: \label{fig:RGflows}
123: \end{figure}
124:
125:
126: In two dimensions, the Hamiltonian for Coulomb interacting massless
127: Dirac fermions in the presence of a quenched random gauge field can
128: be written as
129: \begin{equation}
130: \label{H} {\cal H}={\cal H}_0+{\cal H}_{dis}+\hat{V}
131: \end{equation}
132: with the free part given by,
133: \begin{equation}
134: \label{H0} {\cal H}_0=\sum_{j=1}^N\int{\rm d}^2{\bf
135: r}v_F\left[\psi_j^\dag({\bf r}){\bf p}\cdot\sigma\psi_j({\bf
136: r})\right]
137: \end{equation}
138: where the operator $\psi_j$ annihilates a two-component Dirac
139: fermion, ${\bf p}=-i\nabla$, $\sigma_i$ is the $i$th Pauli matrix,
140: and we have set $\hbar=k_B=1$ for convenience, unless otherwise
141: stated. $N$ represents the number of fermion species; for single
142: layer graphene, $N=4$. The disorder part of the Hamiltonian is
143: \begin{equation}
144: \label{Hdis} {\cal H}_{dis}=\sum_{j=1}^N\lambda_jv_F\int{\rm
145: d}^2{\bf r}\left[a_{\mu}(\br)\psi_j^\dag({\bf
146: r})\sigma_{\mu}\psi_j({\bf r})\right]
147: \end{equation}
148: where $\mu=1,2$ andthe quenched random gauge field is assumed to be
149: uncorrelated
150: \begin{eqnarray}
151: \langle a_{\mu}(\br)\rangle&=&0;\hspace{1cm}\langle
152: a_{\mu}(\br)a_{\nu}(\br')\rangle=\Delta\delta_{\mu\nu}\delta(\br-\br').
153: \end{eqnarray}
154: The charges $\lambda_j=(-1)^j$ vary in sign from node to node as
155: dictated by the overall time reversal symmetry of the system.
156: The Coulomb interaction between Dirac fermions
157: is given by
158: \begin{equation}
159: \label{V} \hat{V}=\frac{1}{2\eps_d}\int d^2{\bf r}d^2{\bf r}'\left[
160: \delta\hat{n}({\bf r})\frac{e^2}{|{\bf r}-{\bf r}'|}
161: \delta\hat{n}({\bf r}')\right]
162: \end{equation}
163: where $\delta\hat{n}({\bf r})=\sum_j\psi_j^\dag({\bf r})\psi_j({\bf
164: r})-n_0$ and $e$ is the electronic charge.
165: The background charge density $n_0$ ensures
166: overall charge neutrality.
167:
168: This system can be described by three "coupling constants" -- $e^2$,
169: $\Delta$ and $v_F$ -- of which $\Delta$ is dimensionless and, in our
170: units, both $v_F$ and $e^2/\eps_d$ have dimensions of velocity. In
171: what follows, we will set $\eps_d=1$ and restore it in the final
172: results by rescaling the charge. Generically, these coupling
173: constants flow under the renormalization group transformation.
174: However, the charge $e^2$ does not flow because it is a coefficient
175: on a non-analytic term in the action, and as detailed elsewhere
176: $\Delta$ doesn't flow either \cite{HerbutJuricicVafek2007}. The
177: entire flow of the renormalized coupling constants then comes from
178: the scale dependence of the Fermi velocity. The RG beta functions
179: take the form:
180: \begin{eqnarray}
181: \frac{de^2}{d\ln\kappa}&=&0\\
182: \frac{d\Delta}{d\ln\kappa}&=&0\\
183: \frac{dv_F}{d\ln\kappa}&=&v_F\frac{\Delta}{\pi}-\frac{e^2}{4}
184: +\mathcal{A}v_F{\Delta}^2+\mathcal{B}e^2\Delta+\mathcal{C}\frac{e^4}{v_F}+\ldots
185: \label{eq:BetafxnsVF}
186: \end{eqnarray}
187: where the ellipses mean terms of cubic order in the double expansion
188: in small $e^2$ and $\Delta$. The lowest order terms in the expansion
189: come from the exchange diagrams for disorder and interactions (see
190: Fig. 2), respectively\cite{HerbutJuricicVafek2007}. The result of
191: our analysis presented below is the following values of the above
192: coefficients
193: \begin{eqnarray}
194: \mathcal{A}=0, \hspace{0.3cm}
195: \mathcal{B}=\frac{1}{8\pi},\hspace{0.3cm}
196: \mathcal{C}=\frac{N}{12}-\frac{103}{96}+\frac{3}{2}\ln2.
197: \end{eqnarray}
198: The vanishing coefficient $\mathcal{A}$ agrees with the result of
199: Ludwig {\it et.al.}\cite{Ludwig:prb94} that, for $e^2=0$, the
200: dynamical critical exponent $z=1-d\ln v_F/d\ln\kappa=1-\Delta/\pi$
201: holds to all orders in perturbation theory. The values of
202: $\mathcal{B}$ and $\mathcal{C}$ are the new results first reported
203: in this paper.
204:
205: If we rescale $e^2$ by $v_F$, we can define a dimensionless coupling
206: constant $\alpha_F=e^2/(\hbar\eps_d v_F)$ which characterizes the
207: strength of Coulomb interactions. The corresponding flow diagram is
208: shown in Fig.1. In the clean case, small $\alpha_F$ flows to zero
209: due to the growth of the Fermi velocity. At $\alpha_F=\alpha_c$,
210: there is a quantum phase transition into an excitonic insulator,
211: controlled by the strongly interacting fixed point. Within the above
212: approximation, for $N=4$, which is appropriate for the single layer
213: graphene, the unstable fixed point appears at
214: $\alpha_c=1/4\mathcal{C}\approx 0.833$. As discussed in greater
215: detail below, since the expansion of the beta function
216: (\ref{eq:BetafxnsVF}) is carried out in $\alpha_F N\ll 1$, the
217: semimetal-insulator fixed unstable point appears beyond the strict
218: validity of the perturbative RG. Nevertheless, for finite $\Delta$,
219: there is a fixed manifold given by
220: \begin{eqnarray}
221: \Delta=\frac{\pi}{4}\frac{\alpha_F(1-4\mathcal{C}\alpha_F)}{1+\pi\mathcal{B}\alpha_F}
222: \end{eqnarray}
223: which is shown in Fig.1 by the solid (red) curve. Importantly, at
224: small $\alpha_F$ and $\Delta$, this manifold represents the line of
225: stable fixed points \cite{stauber2005,HerbutJuricicVafek2007} which
226: is asymptotically {\it exact}, and which merges with the line of
227: unstable fixed points at the critical endpoint
228: $(\alpha^*,\Delta^*)$.
229:
230: %\section{Perturbative renormalization group and dimensional regularization}
231: %
232: \begin{figure}[t]
233: \begin{center}
234: \begin{tabular}{cc}
235: \includegraphics[width=0.15\textwidth]{self_energy1.eps}&
236: \includegraphics[width=0.15\textwidth]{self_energy2.eps}\\
237: \end{tabular}
238: \end{center}
239: \caption{Diagrams contributing to the self energy to the
240: 1$^{st}$order in $\alpha_F$ and $\Delta_A$. As can be readily seen
241: from the link cluster expansion, the second diagram comes with an
242: overall $-$ sign.}\label{sigma1st}
243: \end{figure}
244: %
245: \begin{figure}[t]
246: \begin{center}
247: \begin{tabular}{cccccc}
248: \includegraphics[width=0.15\textwidth]{double_exchangec.eps}&
249: \includegraphics[width=0.15\textwidth]{self_selfenergyc.eps}&
250: \includegraphics[width=0.15\textwidth]{RPAc.eps}\\
251: \includegraphics[width=0.15\textwidth]{double_exchanged.eps}&
252: \includegraphics[width=0.15\textwidth]{self_selfenergyd.eps}&
253: \includegraphics[width=0.15\textwidth]{RPAd.eps}\end{tabular}
254: \end{center}
255: \caption{Diagrams contributing to the self energy to the order
256: $\alpha^2_F$ (first row) and $\Delta^2$ (second
257: row).}\label{sigma2nd}
258: \end{figure}
259: \begin{figure}[t]
260: \begin{center}
261: \begin{tabular}{cccccc}
262: \includegraphics[width=0.15\textwidth]{double_exchange1.eps}&
263: \includegraphics[width=0.15\textwidth]{double_exchange2.eps}&
264: \includegraphics[width=0.15\textwidth]{self_selfenergy1.eps}\\
265: \includegraphics[width=0.15\textwidth]{self_selfenergy2.eps}&
266: \includegraphics[width=0.15\textwidth]{RPA1.eps}&
267: \includegraphics[width=0.15\textwidth]{RPA2.eps}\\
268: %\includegraphics[width=0.25\textwidth]{double_exchange2.eps}
269: \end{tabular}
270: \end{center}
271: \caption{Diagrams contributing to the self energy to the order
272: $\alpha_F\Delta_A$.}\label{sigma2ndCross}
273: \end{figure}
274: %
275:
276: Next, we detail the perturbative renormalization group calculation
277: which we perform in dimensional regularization scheme by
278: analytically continuing the space integrals to $D=2-\eps$ where
279: $\eps>0$. This formal device serves as a regulator for various
280: divergent integrals. The bare Green's function has the form
281: \cite{Negele:QMP88}
282: $$
283: G_0(i\omega,\bk)=(-i\omega+\sigma\cdot\bk)^{-1}=\frac{i\omega+\sigma\cdot\bk}{\omega^2+\bk^2}.
284: $$
285: The resulting self-energy matrix, which is defined through the two
286: point irreducible vertex $\Gamma=\mathcal{G}^{-1}$ as
287: \begin{eqnarray}\label{eq:bareGamma}
288: \Gamma_{\bk}(i\omega)=-i\omega+v_F\sigma\cdot\bk+\Sigma_{\bk}(i\omega),
289: \end{eqnarray}
290: is calculated at finite external momentum $\bk$ and frequency
291: $i\omega$. To illustrate the procedure, consider the self-energy to
292: first order in the coupling constant $e^2$ (the first diagram in
293: Fig.\ref{sigma1st}):
294: \begin{eqnarray}
295: \Sigma_{\bk}^{exch}(i\omega)&=&2\pi
296: e^2\int\frac{d^D\bq}{(2\pi)^D}\int\frac{d\omega}{2\pi}\frac{1}{|\bk-\bq|}G_0(i\omega,\bq)
297: \nonumber\\
298: &=&e^2\sigma\cdot\bk
299: k^{D-2}\frac{\Gamma[1-\frac{D}{2}]}{(4\pi)^{\frac{D}{2}}}\frac{\Gamma[\frac{D+1}{2}]\Gamma[\frac{D-1}{2}]}{\Gamma[D]}\nonumber\\
300: &\rightarrow&e^2\sigma\cdot\bk
301: k^{-\eps}\left(\frac{1}{4\eps}+\frac{1}{8}\ln(64\pi
302: e^{-\gamma_E})\right).
303: \end{eqnarray}
304: In the last line, we have used the analytic properties of the Euler
305: $\Gamma-$functions and expanded the result near $\eps=0$. The Euler
306: constant $\gamma_E\approx0.577$. The pole at $D=2$ corresponds to a
307: logarithmic divergence when a large momentum cutoff is introduced
308: directly in two dimensions i.e. the Hartree-Fock approximation leads
309: to the logarithmic enhancement of velocity and
310: $\Sigma_{\bk}(i\omega)=\frac{e^2}{4}\sigma\cdot\bk \ln\Lambda/k$ is
311: frequency independent. This logarithmic enhancement of the Dirac
312: velocity persists to second order in the coupling constant
313: expansion, in qualitative agreement with
314: Ref.\onlinecite{mishchenko:prl2007}, as well as the leading order
315: expansion in large $N$ \cite{Gonzalez:prb99}.
316:
317: Similarly, the leading order self energy due to the disorder
318: scattering (second diagram in Fig. \ref{sigma1st}) is
319: \begin{eqnarray}
320: &&\Sigma^{dis}_{\bk}(i\omega)=\Delta
321: \int\frac{d^D\bq}{(2\pi)^D}\sigma_{\mu}G_0(i\omega,\bq)\sigma_{\mu}=-\Delta
322: i\omega
323: |\omega|^{D-2}\nonumber\\
324: &\times&\frac{2\Gamma[1-\frac{D}{2}]}{(4\pi)^{\frac{D}{2}}}
325: \rightarrow-\Delta i\omega
326: |\omega|^{-\eps}\left(\frac{1}{\pi\eps}+\frac{1}{2\pi}\ln\left(4\pi
327: e^{-\gamma_E}\right)\right).
328: \end{eqnarray}
329: Again, the pole at $D=2$ corresponds to a logarithmic divergence
330: of the self-energy.
331:
332: Before addressing the higher order contributions to the self-energy,
333: let us define the renormalization conditions. The standard
334: relationship between the renormalized two point function
335: $\Gamma^{ren}_{\bk}(i\omega)$ and the bare one (\ref{eq:bareGamma})
336: is
337: \begin{eqnarray}\label{eq:Zdef}
338: \Gamma^{ren}_{\bk}(i\omega)&=&Z\Gamma_{\bk}(i\omega)
339: \end{eqnarray}
340: where $Z$ is the wavefunction renormalization
341: \cite{PeskinSchroeder}. The renormalized coupling constants can now
342: be defined through the following renormalization conditions
343: \begin{eqnarray}\label{eq:renConditions1}
344: \frac{i}{2}\frac{\partial{{\mbox Tr}[\Gamma^{ren}_{\bk}(i\omega)]}}{\partial\omega}\bigg|_{\omega=k=\kappa}&=&1\\
345: \frac{1}{4}\frac{\partial{{\mbox
346: Tr}[\sigma_{\mu}\Gamma^{ren}_{\bk}(i\omega)]}}{\partial
347: k_{\mu}}\bigg|_{\omega=k=\kappa}&=&v^R_F. \label{eq:renConditions2}
348: \end{eqnarray}
349: Physically the above equations demand that at the renormalization
350: scale $\kappa$, the renormalized single particle Green's function
351: $G^{ren}_{\bk}(i\omega)$ takes the form
352: $$
353: {G^{ren}_{\bk}}^{-1}(i\omega)=\left[-i\kappa + v^R_F\sigma\cdot
354: \kappa\right]^{-1}.
355: $$
356:
357: \begin{widetext}
358: From the above considerations, it is clear that, to find the RG flows,
359: we need only to find the self energies for $\omega=|\bk|=\kappa$. At
360: the orders $\Delta^2$, $e^2\Delta$, and $e^4$ we find
361: \begin{eqnarray}
362: \Sigma^{(a)}(i\omega,\bk)\bigg|_{\kappa}&=&-i\omega \Delta^2
363: \frac{\kappa^{-2\eps}}{2\pi^2}\left[
364: \frac{1}{\eps^2}+\frac{\ln\left(4\pi e^{-\gamma_E}\right)}{\eps}
365: \right]\\
366: \Sigma^{(b)}(i\omega,\bk)\bigg|_{\kappa}&=&-i\omega\frac{e^2\Delta}{v_F}\frac{\kappa^{-2\eps}}{4\pi\eps}
367: + \sigma\cdot\bk
368: e^2\Delta\kappa^{-2\eps}\left[\frac{1}{8\pi\eps^2}+\frac{2\ln\left(16\pi
369: e^{-\gamma_E}\right)+3}{16\pi\eps}\right]\\
370: \Sigma^{(c)}(i\omega,\bk)\bigg|_{\kappa}&=&
371: \left[-i\omega-v_F\sigma\cdot\bk\right]
372: \frac{e^4}{v_F^2}\frac{N}{4}\frac{\kappa^{-2\eps}}{12\eps} -i\omega
373: \frac{e^4}{v_F^2}\frac{\kappa^{-2\eps}}{4\eps}\left[
374: -\frac{2}{3}+\ln2
375: \right]+\sigma\cdot\bk\frac{e^4}{v_F}\frac{\kappa^{-2\eps}}{32\eps}\left[\frac{71}{6}-16\ln2\right]
376: \end{eqnarray}
377: \end{widetext}
378: Thus, to this order in the perturbative expansion, the self energy
379: matrix in Eq.(\ref{eq:bareGamma}) is
380: $$
381: \Sigma_{\bk}(i\omega)=\Sigma^{exch}_{\bk}(i\omega)+\Sigma^{dis}_{\bk}(i\omega)+\Sigma^{a}_{\bk}(i\omega)+
382: \Sigma^{b}_{\bk}(i\omega)+\Sigma^{c}_{\bk}(i\omega),
383: $$
384: and using Eqs.(\ref{eq:Zdef}-\ref{eq:renConditions2}) we find the RG
385: flow equation for the renormalized Fermi velocity to be
386: \begin{eqnarray}\label{eq:BetavF}
387: \frac{dv_F^{R}}{d\ln\kappa}=-\frac{e^2}{4}+v_F^R\frac{\Delta}{\pi}+\frac{e^2\Delta}{8\pi}
388: +\left(\frac{N}{12}-\frac{103}{96}+\frac{3}{2}\ln2\right)\frac{e^4}{v_F^R}\nonumber.
389: \end{eqnarray}
390: This is the result displayed in Eq.(\ref{eq:BetafxnsVF}). It is
391: apparent from this flow equation that at $\Delta=0$ the Fermi
392: velocity increases logarithmically provided that the dimensionless
393: coupling $\alpha<\alpha_c$. Such logarithmic growth implies
394: suppression of the electronic density of states near the Dirac point
395: and concomitant suppression of the specific
396: heat\cite{vafek:prl2007}.
397:
398: In the case of $\alpha>\alpha_c$, the runaway RG flows can be
399: interpreted as the flow towards an excitonic insulator. The physical
400: nature of this insulator depends on the details of the lattice
401: model. For spinless fermions on a honeycomb lattice, for example, it
402: would correspond to a state with a spontaneous breaking of
403: (plaquette or bond centered) inversion symmetry and unequal
404: population of the two sublattices; for strictly $1/r$ Coulomb
405: repulsion and for the spinfull case it would correspond to an
406: "antiferromagnetic" order (but no unit cell doubling) with unequal
407: spin population of the two sublattices, although details of the
408: short range part of the repulsive interactions could destabilize it
409: towards the inversion symmetry broken state\cite{herbutPRL06}.
410: %At the clean quantum
411: %critical point, the quasiparticle weight $Z$ vanishes and the
412: %Green's function becomes
413: %$$
414: %G_{\bk}(i\omega)=\frac{i\omega+v^c_F\sigma\cdot\bk}{(\omega^2+{v^c_F}^2\bk^2)^{1-\frac{\eta}{2}}}
415: %$$ with an exponent
416: %$$\eta=-\frac{d\ln Z}{d\ln \kappa}=\frac{1}{2}\left(\frac{N}{12}-\frac{2}{3}+\ln2\right)>0,$$
417: % implying that quasiparticles cannot be
418: %adiabatically connected to the electrons; the Fermi velocity at the
419: %QCP is finite and equal to $v_F^c=e^2/(\hbar \eps_d\alpha_c)$.
420:
421: We conclude with a discussion of the validity of the approximations
422: employed here. First, the RG was organized perturbatively in both
423: $\alpha$ and $\Delta$ so the weak coupling portion of the fixed line
424: in Fig.\ref{fig:RGflows} is rigorously justified. Additionally, the
425: downward curvature of this line implied by the sign of the terms in
426: the second order expansion is also rigorous. Diagramatically,
427: integration of these RG equations corresponds to an infinite
428: parquet-like resummation of the two leading logarithms at each order
429: in the pertrubative expansion (see for instance \cite{herbutBook}).
430:
431: On the other hand, the unstable portion of the fixed line and the
432: (unstable) clean fixed point appear at a finite value of the
433: coupling constant at which $\alpha_F N$ is not small. They are,
434: therefore, beyond the reach of the perturbative RG. Nevertheless,
435: the very {\em existence} of the clean unstable fixed point is
436: perhaps on somewhat firmer
437: footing\cite{KhveshchenkoPRL01,GorbarPRB2002,KhveshchenkoLeal04},
438: and due to the negative curvature of the IR stable fixed line in the
439: perturbatively accessible region of the flow diagram, we expect that
440: the gross topological features of the fixed manifold in
441: Fig.\ref{fig:RGflows}, if not its quantitative aspects, are valid.
442:
443: We wish to thank Professors I. Herbut and Z. Tesanovic for useful
444: discussions and for their critical reading of the manuscript.
445:
446:
447: \bibliography{RG}
448: \end{document}
449:
450:
451: We have also performed a different resummation scheme in which the
452: single particle self-energy is evaluated to order $e^2$, but the two
453: particle Green's function is determined within the conserving
454: approximation of Baym and Kadanoff\cite{KadanoffBaym62}.
455: Diagramatically, this corresponds to the summation of the ladder
456: vertex diagrams with the Hartree-Fock single-particle
457: propagator\cite{mishchenko:prl2007}. We were able to find a closed
458: form solution for the resulting integral equation at zero external
459: frequency, which gives a quantum phase transition at $\alpha_c=4$.
460: One of the drawbacks of this method, however, is its insensitivity
461: to $N$ (it finds $\alpha_c=4$ for all $N$.) Further cautionary note
462: is the perturbative RG which is $4.99$. While in both approaches
463: the extrapolation to the intermediate coupling is strictly
464: uncontrolled, it is encouraging that both agree on the gross
465: topological features of the RG flow shown in Fig.(\ref{fig:RGflows})
466: and for $N=1$ are numerically close as well.
467: