1: \documentclass[prl,aps,twocolumn,floats,showpacspsfig]{revtex4}
2: %\documentclass[preprint,floats,aps,showpacs]{revtex4}
3: \usepackage{amssymb}
4:
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: \usepackage{epsfig}
7: \usepackage{graphicx}
8: \usepackage{subfigure}
9:
10: %\documentclass[preprint,floats,aps,showpacs,epsf]{revtex4}
11:
12:
13: \newcommand{\be}{\begin{equation}}
14: \newcommand{\ee}{\end{equation}}
15: \newcommand{\bea}{\begin{eqnarray}}
16: \newcommand{\eea}{\end{eqnarray}}
17: \newcommand{\rD}{\mbox{D}}
18: \newcommand{\reff}{\mbox{eff}}
19: \newcommand{\rR}{{\rm R}}
20: \newcommand{\rL}{{\rm L}}
21: \newcommand{\p}{\partial}
22: \newcommand{\s}{\sigma}
23: \newcommand{\rF}{{\rm F}}
24: \newcommand{\rf}{{\rm f}}
25: \newcommand{\la}{\langle}
26: \newcommand{\ra}{\rangle}
27: \newcommand{\rd}{\mbox{d}}
28: \newcommand{\ri}{\mbox{i}}
29: \newcommand{\re}{\mbox{e}}
30: \newcommand{\rc}{{\rm c}}
31: \newcommand{\rs}{{\rm s}}
32: \newcommand{\rt}{{\rm t}}
33:
34:
35: %\documentclass[prb,aps,amsfonts,showpacs,eqsecnum,superscriptaddress,floatfix,multicol,twocolumn]{revtex4}
36:
37: %\usepackage{epsfig}
38: %\usepackage{amssymb}
39: %\usepackage{psfrag}
40: %\usepackage{color}
41:
42: %\newcommand{\be}{\begin{equation}}
43: %\newcommand{\ee}{\end{equation}}
44: %\newcommand{\bea}{\begin{eqnarray}}
45: %\newcommand{\eea}{\end{eqnarray}}
46: %\newcommand{\up}{\uparrow}
47: %\newcommand{\down}{\downarrow}
48: %\def\nn{\nonumber\\}
49: %\def\tt{\tilde{\theta}}
50: %\def\r#1{(\ref{#1})}
51: %\def\sgn{{\rm sgn}}
52: %\def\eps{\epsilon}
53:
54: \begin{document}
55:
56: \title{Orbital Dependence of Quasiparticle Lifetimes in $Sr_2RuO_4$}
57:
58: \author{ R. M. Konik$^1$ and T. M. Rice$^{1,2}$}
59: \affiliation{$^{1}$Department of Physics, Brookhaven National Laboratory, Upton, NY 11973-5000, USA\\
60: $^{2}$Institute f\"ur Theoretische Physik, ETH-Z\"urich, CH-8093 Z\"urich, Switzerland}
61: \date{\today}
62:
63: \begin{abstract}
64: Using a phenomenological Hamiltonian, we investigate the quasiparticle lifetimes
65: and dispersions in the three low energy bands, $\gamma$, $\beta$, and $\alpha$ of $Sr_2RuO_4$. Couplings
66: in the Hamiltonian are fixed so as to produce the mass renormalization as measured
67: in magneto-oscillation experiments.
68: We thus find reasonable agreement in all bands between our computed lifetimes and those measured in ARPES
69: experiments by Kidd et al. \cite{kidd} and Ingle et al. \cite{ingle}.
70: In comparing computed to measured quasiparticle dispersions, we however find good agreement
71: in the $\alpha$-band alone.
72: \end{abstract}
73: \pacs{71.10Pm, 71.10Ay}
74: \maketitle
75:
76: %%%%%%%%%%%%%%%%%%%%%%%
77: \section{Introduction}
78: %%%%%%%%%%%%%%%%%%%%%%%
79:
80: The ruthenate $Sr_2RuO_4$ has been extensively studied for a number of reasons,
81: chief among them its unconventional superconducting state \cite{mack}. Its electronic
82: structure, reviewed in detail by Bergemann et al. \cite{bergemann},
83: has also been of much interest.
84: Three bands belonging to the $t_{2g}$
85: complex of 4d $Ru$ orbitals cross the Fermi energy. They are divided into two sets.
86: One derived from the $d_{xy}$ orbital has a two-dimensional dispersion with little
87: dispersion along the c-axis due to the layered structure of the material.
88: The second
89: set comprises the $d_{xz}$ and $d_{yz}$ bands which have predominantly one-dimensional
90: dispersion. A key feature is the absence of hybridization between these sets in a
91: single layer due to the opposite parity under reflection about a $RuO_2$ plane.
92: This contrast in their dispersion has been invoked by Kidd and collaborators \cite{kidd}
93: to explain the strong difference in the energy and temperature dependence of the
94: quasiparticle lifetimes between the two sets observed in recent ARPES (Angle
95: Resolved Photoemission Spectroscopy) experiments. In this note we report on
96: some simple calculations to examine this interpretation.
97:
98: \begin{figure}
99: \begin{center}
100: \noindent
101: \epsfysize=.4\textwidth
102: \epsfbox{srobands.eps}
103: \end{center}
104: \caption{A schematic of the three bands on $Sr_2RuO_4$.}
105: \end{figure}
106:
107: The Fermi surface is illustrated in Fig. 1 and consists of three sheets. The almost
108: circular $\gamma$-sheet derives from orbitals with $d_{xy}$ symmetry. The $\alpha$
109: and $\beta$ sheets
110: derive from the approximately linear sections due to the orbitals with $d_{xz}$ and $d_{yz}$
111: symmetry which however hybridize weakly with each other where they cross.
112: Kidd and collaborators \cite{kidd} measured the dispersion and linewidths of the
113: quasiparticles on the $\gamma$- and the $\beta$-
114: sheets along the line $\Gamma$-M and found a clear difference
115: in their behavior as a function of both energy and temperature.
116: Ingle et al. \cite{ingle} measured the same quantities for the $\alpha$-band along
117: the $\Gamma$-X direction finding similar linewidths to those observed
118: by Kidd et al. in the $\beta$-band. At first sight this
119: difference between the $\gamma$-band and the $\alpha$-, $\beta$-bands
120: would seem to be simply a consequence of their differing dispersion and
121: the lack of hybridization between them. While direct scattering of a quasiparticle
122: between these bands is forbidden due to their different parities, quasiparticles in
123: these band will still interact through the Coulomb interactions which leads to
124: modifications of their strictly one and two-dimensional character. To explore this
125: effect we undertake low order calculations of the lifetime using phenomenological
126: interaction strengths. These interaction strengths are chosen so that the mass renormalization
127: observed in magneto-oscillation experiments \cite{mack,bergemann} is reproduced.
128:
129: Doing so we are able to account for the broad features of the observed linewidths. In particular
130: we obtain a good fit to the linewidths in the $\gamma-$band as a function of binding energy and we
131: find, consistent with experiment that the linewidths in the $\beta$- and
132: $\alpha$-bands are far smaller. Our computations do however consistently underestimate to a small degree the
133: linewidths. This points to, perhaps, some other contributory mechanism beyond electron-electron interactions.
134: We also explored the finite temperature behavior of the linewidths. In both the $\gamma$- and $\beta$- bands,
135: our computations match well
136: the observed behaviour seen in \cite{kidd}.
137: Unlike the linewidths, the agreement between the measured and computed dispersion relations is mixed.
138: For the $\alpha$-band,
139: good agreement is found while for the $\beta$ and $\gamma$-bands,
140: the measured mass renormalization is far less than what
141: would be expected from magneto-oscillation experiments. We comment on this further in the discussion and
142: conclusion section.
143:
144:
145:
146: \begin{figure*}
147: \begin{center}
148: \noindent
149: \epsfysize=.4\textwidth
150: \epsfbox{diagrams.xfig.eps}
151: \end{center}
152: \caption{Diagrams contributing to the self energy at first and second order. Here the greek indices $\alpha,\beta$ refer
153: to generic bands.}
154: \end{figure*}
155:
156: \section{Calculations of the Lifetime and Effective Mass}
157:
158: The multi-band Hamiltonian that describes a single layer is as follows
159: \begin{eqnarray}
160: {\cal H} &=& {\cal H}_{0} + {\cal H}_{\rm int};\cr
161: {\cal H}_0 &=& \sum_{k\sigma\mu}\epsilon^{\mu}(k) c^\dagger_{\mu k\sigma}c_{\mu k\sigma};\cr
162: {\cal H}_{\rm int} &=& \sum_{\mu,\nu,\sigma,\sigma',i}V_{\mu\nu}n_{\mu\sigma}(i)n_{\nu\sigma'}(i).
163: \end{eqnarray}
164: Here the greek indices, $\mu,\nu$, are band indices and sum over the three bands, $\alpha,\beta,$ and $\gamma$.
165: The two-dimensional $\gamma$-band, describing
166: orbitals with $d_{xy}$ symmetry, takes the form \cite{morr,ll},
167: \begin{equation}
168: \epsilon^\gamma({\bf k}) = -.88\cos(k_x) - .88\cos(k_y) - .56\cos(k_x)\cos(k_y) - .5.
169: \end{equation}
170: The one-dimensional $\alpha$- and $\beta$-bands arise from weak hybridization between the $d_{xz}$ and $d_{yz}$
171: symmetry orbitals and appear as
172: \begin{eqnarray}
173: \epsilon^{xz}({\bf k}) &=& -.62\cos(a k_x) - .09\cos(a k_y) \cr
174: && \hskip .5in + .04\cos(a k_x)\cos(a k_y) - .24;\cr
175: \epsilon^{yz}({\bf k}) &=& -.09\cos(a k_x) - .62\cos(a k_y) \cr
176: && \hskip .5in + .04\cos(a k_x)\cos(a k_y) - .24;\cr
177: \epsilon^{\pm}({\bf k}) &=& \frac{1}{2}(\epsilon^{xz}\pm\epsilon^{yz});\cr
178: \epsilon^{\alpha/\beta} &=& \epsilon^+ \mp \sqrt{(\epsilon^-)^2+.01}.
179: \end{eqnarray}
180: All energies are in eV's and $a=3.86A^o$ is the lattice spacing.
181: The interaction vertices reduce to four terms describing intraband scattering between
182: electrons in the $\gamma$-band ($V_{\gamma\gamma}$) and in the
183: $\alpha$,$\beta$-bands ($V_{\alpha\alpha}=V_{\beta\beta}$) and interband scattering
184: between $\gamma$ and the $\alpha,\beta$-bands($V_{\gamma\alpha}=V_{\gamma\beta}$)
185: and between the $\alpha$ and $\beta$ bands ($V_{\alpha\beta}$).
186: Simple estimates of the Fermi-Thomas screening length give
187: strong screening due to the large density
188: of states at the Fermi energy so we neglect the wavevector dependence of the interaction
189: vertices. This leads to four independent parameters which we adjust phenomenologically
190: as discussed below.
191: The calculation of the self energy is restricted to the lowest orders in the
192: phenomenological interaction vertices, illustrated in Fig 2. For the lifetime this
193: amounts to calculating the decay processes of a single quasihole into three
194: quasiparticles with renormalized interaction vertices. These processes will dominate
195: the decay rate of low energy quasiholes as multiquasiparticle decays will rise more
196: slowly with the energy of the quasihole measured from the Fermi energy. Our aim
197: is to explore the effects of the differing band dispersions on the decay rates.
198:
199: Our approach to this problem is then in the spirit of Landau Fermi liquid theory -- we use the necessary
200: mass renormalization to
201: fit the coupling constants (akin to the Fermi liquid parameters) and then determine other quantities
202: (the inverse lifetimes)
203: in terms of these same couplings. In this spirit, the largeness of certain parameters
204: (for example. the dimensionless coupling $u_{\gamma\gamma} = K_{F\gamma}V_{\gamma\gamma}/2\pi v_{F\gamma}$),
205: should not be taken as problematic.
206:
207: \renewcommand{\thesubfigure}{}
208: \renewcommand{\subfigcapskip}{-40pt}
209: \begin{figure*}
210: \begin{center}
211: \vskip -.2in
212: \subfigure[]{
213: \epsfysize=1.25\textwidth
214: \includegraphics[height=7.5cm,angle=-90]{lgamma_dhva.eps}}
215: \subfigure[]{
216: \epsfysize=1.25\textwidth
217: \includegraphics[height=7.5cm,angle=-90]{lbeta_dhva.eps}}
218: \vskip .2in
219: \end{center}
220: \centerline{\hbox{}}
221: \begin{center}
222: \subfigure[]{
223: \epsfysize=1.25\textwidth
224: \includegraphics[height=7.5cm,angle=-90]{lalpha_dhva1.eps}}
225: \subfigure[]{
226: \epsfysize=1.25\textwidth
227: \includegraphics[height=7.5cm,angle=-90]{lalpha_dhva.eps}}
228: \vskip .5in
229: \end{center}
230: \caption{Plots of the inverse lifetimes in the three bands, $\gamma$, $\beta$ and $\alpha$.}
231: \end{figure*}
232:
233: The real part of the self energy is calculated employing the same set of diagrams.
234: Any non-zero contribution to the real part at $\omega=0$ at the Fermi surface is
235: absorbed into a redefinition of the chemical potential.
236: The first order diagram, marking the exchange energy, contributes to the real part of
237: the self-energy alone. It is the first in a series of diagrams but is the only member
238: of the series to make a contribution. The remaining diagrams in the series simply renormalize
239: the chemical potential and so are ignored.
240:
241: \begin{table}
242: %\setcaptionmargin{0mm}
243: %\onelinecaptionstrue
244: %\captionstyle{flushleft}
245: \begin{tabular}{|c|c|c|c|}
246: \hline
247: & $\alpha$ & $\beta$ & $\gamma$ \\
248: \hline
249: dHvA ($m_*$) & 3.3 & 7.0 & 16 \\
250: \hline
251: ARPES & 3.3 & 2.3 & 3.0 \\
252: \hline
253: LDA AV ($m_{LDA})$ & 1.2 & 2.3 & 2.3 \\
254: \hline
255: $m_R=m_*/m_{LDA}$ & 2.8 & 3.0 & 7.0\\
256: \hline
257: \end{tabular}
258: \caption{Mass renormalization of the electrons in the three bands of $Sr_2RuO_4$ as given by
259: magneto-oscillation experiments \cite{mack,bergemann} (dHvA),
260: ARPES measurements \cite{kidd,ingle}, and LDA computations .
261: These three mass renormalizations are all given in terms of the bare electron mass.}
262: \end{table}
263:
264: To compute the strengths of phenomenological interaction parameters we insist that
265: the computed self energy reproduce the mass renormalizations expected from
266: magneto-oscillation experiments \cite{mack,bergemann} (see top line of Table 1).
267: The mass renormalizations induced by the interaction terms in Eq. 2.1
268: are given by
269: \begin{equation}
270: m_R = \frac{m_*}{m_{LDA}} = \frac{1-\partial_\omega{\rm Re}\Sigma(w,k=K_F)}{1+v_{LDA}^{-1}\partial_k{\rm Re}\Sigma(w,k=K_F)}
271: \end{equation}
272: where $m_{LDA}$ is the mass renormalization induced by the LDA band structure (also listed in Table I).
273: The LDA values given in Table I are averages around the bands' Fermi surfaces weighted by the local
274: density of states (appropriate as the magneto-oscillations masses are themselves weighted averages).
275: The interaction strengths are then determined by insisting that $m_R$ equal
276: $m_*/m_{LDA}$ (as listed in Table 1) at a single point along the Fermi surface of each band.
277: We do not attempt to average $m_R$ itself around the Fermi surfaces (requiring a computation
278: of the self-energy everywhere) as too computationally costly.
279:
280: In the presence of contact interactions, the $T=0$
281: self-energy for a given band $\mu$ at first order consists of
282: a term independent of both momentum and energy and so may be ignored (at finite $T$, this
283: term however does do more than renormalize the Fermi energy). At second order
284: the contribution takes the form
285: \begin{eqnarray}
286: \Sigma_\mu(\omega, k) &=& \sum_\nu \Sigma_{\mu\nu}(\omega, k) \cr\cr
287: &=& \sum_\nu \frac{V_{\mu\nu}^2}{4\pi^3}\int dq
288: \int dx \frac{{\rm Im}\chi_{\nu}(q,x)}{w+i\delta-\epsilon_\mu(k-q)-x}\cr\cr
289: && \times (\theta(-x)-\theta(\epsilon_\mu(k-q))),
290: \end{eqnarray}
291: where $\chi_\nu$ is the susceptibility of electrons in band $\nu$.
292: We numerically compute the derivatives of the real part of the self energy necessary
293: to computing $m_R$ in order
294: to accurately capture the effect of the dispersion in each of the bands (without
295: recourse to approximating the bands through a quadratic dispersion relation).
296:
297: The values of the interaction parameters so determined are given in Table II.
298: The first
299: number marks the value used in later computations of the dispersion and lifetime.
300: The ranges in brackets mark the range of parameters producing the desired mass
301: renormalization, $m_R$. However other choices within the given ranges yield
302: results for the lifetimes that differ by no more than $10\%$.
303: \begin{table}
304: \begin{tabular}{|c|c|c|c|}
305: \hline
306: $V_{\gamma\gamma}$ & $V_{\gamma\beta}$ & $V_{\alpha\beta}$ & $V_{\beta\beta}$ \\
307: \hline
308: 5.9 (5.1-6.5) & 1.4 (0-2) & 1.3 (.8-1.6) & 1.5 (0-2.1) \\
309: \hline
310: \end{tabular}
311: \caption{Strengths of the interaction couplings (all values in $a^2\cdot eV$ where $a$ is
312: the lattice spacing.}
313: \end{table}
314: The largest coupling by far is $V_{\gamma\gamma}$, dictated by the need to produce a mass
315: renormalization of $m_R \sim 7$.
316:
317:
318: The real part of the self-energy will have terms beyond linear order in $\omega$ that
319: however are not necessarily negligible {\it a priori}. In three dimensions the real
320: part of the self energy would be expected to take the form
321: \begin{equation}
322: {\rm Re}\Sigma (\omega) = a \omega + b\omega^3 + \cdots.
323: \end{equation}
324: By dimensional analysis $b$ would of the form, ${\rm dimensionless~constant}\times E_F^{-2}$,
325: and so for $\omega$ small, the cubic term can be ignored.
326: But in two dimensions, the self-energy develops non-analyticities and will appear as
327: \begin{equation}
328: {\rm Re}\Sigma (\omega) = a \omega + b\omega^2{\rm sgn}(\omega) + \cdots.
329: \end{equation}
330: It thus cannot be immediately neglected.
331:
332:
333:
334:
335: We can estimate this non-analytic term more precisely. Using \cite{chubukov},
336: we know the non-analytic contribution to the self-energy at second order at arbitrary
337: $\omega$ and $k$ is given for bands with quadratic dispersions \cite{chu1} by
338: \begin{eqnarray}
339: {\rm Re}\Sigma^{\rm non-anal.}_{\mu\mu}(\omega,k) &=& \frac{V^2_{\mu\mu}m^2_\mu}{64\pi^2E_{F\mu}}\times\cr
340: &&\hskip -1in \bigg[(\omega^2+\frac{1}{4}(\omega-\epsilon_\mu(k))^2){\rm sign}(\omega-\epsilon_\mu(k))\cr
341: &&\hskip -1in + (\omega^2-\frac{1}{4}(\omega-\epsilon_\mu(k))^2){\rm sign}(\omega+\epsilon_\mu(k))\cr
342: &&\hskip -1in + \omega^2{\rm sign}(\omega) + \omega^2{\rm sign}(\frac{\omega}{\epsilon_\mu(k)});\cr
343: &&\hskip -1in - (\omega+\epsilon_\mu(k))^2{\rm sign}(\frac{\omega+\epsilon_\mu(k)}{\epsilon_\mu(k)})\bigg];\cr\cr
344: {\rm Re}\Sigma^{\rm non-anal.}_{\mu\neq\nu}(\omega,k) &=& \frac{V^2_{\mu\nu}m_\mu m_\nu^2}{32\pi^2K_{F\mu}K_{F\nu}}\times\cr
345: &&\hskip -1in \bigg[(\omega^2+\frac{1}{4}(\omega-\epsilon_\mu(k))^2){\rm sign}(\omega-\epsilon_\mu(k))\cr
346: &&\hskip -1in + (\omega^2-\frac{1}{4}(\omega-\epsilon_\mu(k))^2){\rm sign}(\omega+\epsilon_\mu(k))\bigg],
347: \end{eqnarray}
348: where $E_{F\mu}$ is the effective bandwidth of band $\mu$, and $m_\mu = K^2_{F\mu}/2E_{F\mu}$ with
349: $K_{F\mu}$ the band's corresponding Fermi wavevector.
350: $\Sigma^{\rm non-anal.}_{\mu\mu}(\omega,k)$ has two contributions: one arising
351: from the non-analyticity in the electron polarizability, $\chi_\mu (\omega,q)$, at
352: $q$ near 0 and one from another non-analyticity in $q$ near $2K_{F\mu}$ \cite{chubukov}.
353: In contrast, the interband contribution to the self energy, $\Sigma_{\mu\nu}$,
354: only sees a non-analytic contribution from $\chi_\nu (\omega, q\sim 0)$. To estimate the contribution these
355: terms make to the self-energy we need to estimate the appropriate values for the effective (quadratic)
356: masses appearing in the above. To do so we write down the corresponding expressions for the imaginary
357: part of the self energy:
358: \renewcommand{\thesubfigure}{}
359: \renewcommand{\subfigcapskip}{-40pt}
360: \begin{figure*}
361: \begin{center}
362: \subfigure[]{
363: \epsfysize=1.35\textwidth
364: \includegraphics[height=5.75cm,angle=-90]{gam_dis1.eps}}
365: \subfigure[]{
366: \epsfysize=1.35\textwidth
367: \includegraphics[height=5.75cm,angle=-90]{beta_dis1.eps}}
368: \subfigure[]{
369: \epsfysize=1.35\textwidth
370: \includegraphics[height=5.75cm,angle=-90]{alp_dis.eps}}
371: \end{center}
372: \vskip .2in
373: \caption{Plots of the dispersion in the three bands, $\gamma$, $\beta$ and $\alpha$.}
374: \end{figure*}
375: \begin{eqnarray}
376: {\rm Im}\Sigma_{\mu\mu}(\omega,k) &=& \frac{V^2_{\mu\mu}m^2_\mu}{32\pi^3 E_\mu}\times\cr
377: &&\hskip -1.2in \bigg[(\omega^2+\frac{1}{4}(\omega-\epsilon_\mu(k))^2)\ln(\big|\frac{\omega-\epsilon_\mu(k)}{E_{F\mu}}\big|)\cr
378: &&\hskip -1.2in + (\omega^2-\frac{1}{4}(\omega-\epsilon_\mu(k))^2)\ln(\big|\frac{\omega+\epsilon_\mu(k)}{E_{F\mu}}\big|)\cr
379: &&\hskip -1.2in - \omega^2 + \frac{\omega\epsilon_\mu(k)}{2}\cr
380: &&\hskip -1.2in + \omega^2\ln(\big|\frac{\omega}{E_{F\mu}}\big|) + \omega^2\ln(\big|\frac{\omega}{\epsilon_\mu(k)}\big|);\cr
381: &&\hskip -1.2in - (\omega+\epsilon_\mu(k))^2\ln(\big|\frac{\omega+\epsilon_\mu(k)}{\epsilon_\mu(k)}\big|)\bigg];\cr\cr
382: {\rm Im}\Sigma_{\mu\neq\nu}(\omega,k) &=& \frac{V^2_{\mu\nu}m_\mu m_\nu^2}{16\pi^3 K_{F\mu}K_{F\nu}}\times\cr
383: &&\hskip -1.2in \bigg[(\omega^2+\frac{1}{4}(\omega-\epsilon_\mu(k))^2)\ln(\big|\frac{\omega-\epsilon_\mu(k)}{E_{F\mu}}\big|)\cr
384: &&\hskip -1.2in + (\omega^2-\frac{1}{4}(\omega-\epsilon_\mu(k))^2)\ln(\big|\frac{\omega+\epsilon_\mu(k)}{E_{F\mu}}\big|) \cr
385: &&\hskip -1.2in - \omega^2 + \frac{\omega\epsilon_\mu(k)}{2}\bigg].
386: \end{eqnarray}
387: To determine $E_{F\mu}$, we fit a numerical evaluation of
388: the imaginary part of self-energy via Eqn. 2.5 to the above expression. (The non-analytic portion
389: of ${\rm Im}\Sigma(\omega,k)$ dominates because of the presence of the logs.) Having so extracted
390: effective Fermi energies, $E_{F\mu}$ (see Table III), we then are able to evaluate the non-analytic real part of the self-energy.
391: We find that it makes a significant contribution only for $\omega > .04eV$ and then only for
392: ${\rm Re}\Sigma_{\gamma\gamma}$ -- the dispersion relations for the $\alpha$ and $\beta$-bands are
393: insensitive to this correction.
394:
395: \begin{table}
396: \begin{tabular}{|c|c|c|}
397: \hline
398: $E_{F\gamma}$ & $E_{F\beta}$ & $E_{F\alpha}$ \\
399: \hline
400: 0.6 & 0.5 & 1. \\
401: \hline
402: \end{tabular}
403: \caption{The effective bandwidths for each of the three bands.}
404: \end{table}
405:
406: While the spirit of our computation is Landau Fermi liquid theory,
407: we can nonetheless address the question of what contribution higher order diagrams (beyond those in Figure 2)
408: make to the inverse lifetime. This is a difficult question of course, but for at least an infinite subset of diagrams studied by Chubukov et al. \cite{chubukov}, we can give a definitive answer: not much.
409: We focus on $\Sigma_{\gamma\gamma}$ (i.e. the contribution to the self energy
410: of the $\gamma$-band due to the polarizability of $\gamma$-band electrons) as $V_{\gamma\gamma}$ is the largest coupling.
411: Chubukov et al. \cite{chubukov} have studied the behaviour of higher order forward scattering terms in the
412: $V_{\gamma\gamma}$-perturbative series.
413: They found that the relevant expansion parameter
414: is $s = u_{\gamma\gamma}^2\omega/(\omega-\epsilon_\gamma(k))$.
415: In terms of our lifetime computations, we are far off the mass shell because of the
416: large mass renormalizations involved, i.e. $(\omega-\epsilon_\gamma(k))/\omega \sim 6$, $s$ will be smaller than might first appear,
417: i.e. $s \sim 4$. While the radius of convergence of the series is unclear, if the series needs to be resummed, the effects are
418: not particularly drastic. The terms in ${\rm Im}\Sigma (\omega , k)$ (see Eqn. 2.8) behaving as $\omega^2\ln (\omega)$ are modified to
419: \begin{equation}
420: \omega^2\ln (\frac{\omega}{E_{F\gamma}}) \rightarrow \omega^2 (\frac{1}{2}\ln (\frac{\omega}{E_{F\gamma}}) - \frac{1}{2}|\ln (u^2)|)
421: \end{equation}
422: Given that $u_{\gamma\gamma} \sim 5$ and $E_{F\gamma} \sim 0.5eV$,
423: for very small $\omega$ we would then expect higher order contributions to lead to a reduction in the
424: inverse lifetime while for omega larger than $.02eV$, higher order contributions leave the answer effectively unchanged.
425:
426:
427:
428: \renewcommand{\thesubfigure}{}
429: \renewcommand{\subfigcapskip}{-40pt}
430: \begin{figure*}
431: \begin{center}
432: \vskip -.2in
433: \subfigure[]{
434: \epsfysize=1.25\textwidth
435: \includegraphics[height=6cm,angle=-90]{gamma_Temp.eps}}
436: \subfigure[]{
437: \epsfysize=1.25\textwidth
438: \includegraphics[height=6cm,angle=-90]{beta_Temp.eps}}
439: \end{center}
440: \vskip .2in
441: \caption{Plots of the temperature dependence in the $\gamma$- and $\beta$-bands..}
442: \end{figure*}
443:
444: \section{Comparison to ARPES Results}
445:
446:
447: The ARPES studies of Kidd and collaborators \cite{kidd} reported inverse lifetime results
448: for a hole in the $\gamma$ and $\beta$
449: bands measured along the $\Gamma-M$ direction as a function of energy
450: and of temperature. There is a pronounced difference
451: between the bands with the inverse lifetime for holes in the $\gamma$-band
452: rising much more
453: rapidly with energy and temperature than in the $\beta$-band. The results for the
454: $\beta$-band are mimicked by the linewidths observed in the $\alpha$-band in \cite{ingle}.
455:
456: In the top panel of Figure 3, we plot our computed inverse lifetime as a function of binding
457: energy for the $\gamma$ and $\beta$-bands along $\Gamma-M$. To our computed result of the $\gamma$-band,
458: we have added a static impurity contribution of ${\rm Im}\Sigma_{\rm imp}=.055eV$. This is chosen so that
459: the experimental and theoretical values of the inverse lifetime coincide at $\omega=0$. To arrive
460: at the experimental values of the inverse lifetime, we have taken the MDC widths as measured by
461: Kidd et al. and multiplied them by the corresponding LDA value of the band velocity, $v_{F-LDA}$. For the $\gamma$-band
462: this velocity changes rapidly near the Fermi surface and so we have correspondingly employed a $k$-dependent
463: $v_{F-LDA}$. This yields a slightly
464: larger experimental inverse lifetime than reported in \cite{kidd}.
465: While our computed values undershoot the measured values of the inverse lifetime, the agreement
466: is reasonable for the $\gamma$-band and would be improved significantly if we treated ${\rm Im}\Sigma_{\rm imp}$
467: as a completely free parameter. In particular, we note that
468: the uncertainties in the measured values do not reflect the uncertainty in the correct value of the
469: bare LDA velocity to employ in converting an MDC width to an inverse lifetime.
470:
471: For the $\beta$-band the computed and measured values of the inverse lifetime
472: broadly match. In particular, the predicted and measured linewidths are much smaller
473: than in the $\gamma$-band. Again we have added an impurity contribution
474: to the computed inverse lifetime values so that the measured and computed values agree
475: at zero binding energy. We do note however that the computed inverse lifetime is smaller
476: than that measured. Over a range of binding energies of $.06eV$, the
477: measured inverse lifetime increases by $\sim .05eV$ whereas the computed value
478: increases only by half that. This disagreement may be in part explained by the presence at larger energies
479: of correspondingly large error bars. However, unlike for the $\gamma-$ band, the band velocity
480: in the $\beta$-band is not a strong function of wavevector.
481:
482: In the bottom two panels of Figure 3, we plot our computed lifetime for the $\alpha$-band along
483: the $M-X$ direction vs that measured by Ingle et al. \cite{ingle}.
484: Ingle et al. use samples with freshly cleaved surfaces where the $\alpha$-band
485: can be resolved without employing surface aging.
486: In the left panel, we plot our computations, again with a correcting
487: ${\rm Im}\Sigma_{\rm imp}$, against the values measured by Ingle et al.. In the right
488: panel we plot our computations (with no impurity correction) against the measured electron-electron
489: contribution to the self-energy. Ingle et al. deduce this contribution by subtracting out
490: an estimate of both the phonon contribution and the impurity contribution.
491: Again we see that the computed value of the inverse lifetime in the $\alpha$-band is much
492: less than that of the $\gamma$-band. However like the $\beta$-band, this value is less than that
493: measured.
494: If we look at only the value imputed to the electron-electron contribution, we see a rise
495: in the inverse lifetime of $.05eV$ over a $.06eV$ range of binding energies whereas
496: our computed value rises by a little less than $.02eV$.
497:
498: In Figure 4 we compare the corresponding computed and measured dispersion relations for the three bands.
499: The computed dispersions are essentially straight lines whose slope is determined by
500: the value of the dHvA mass renormalizations (as this is how we determine the coupling strength).
501: We have also plotted the LDA dispersion relations.
502: For the $\gamma$- and $\beta$- bands, there is strong disagreement between measured and computed
503: values. Alternatively, the measured dispersion for the $\beta$- and $\gamma$- bands show
504: a mass renormalization far smaller than seen in magneto-oscillation experiments. For the $\beta$-band,
505: the measured mass renormalization equals its LDA prediction. In the $\alpha$-band,
506: in contrast, the dispersion closely matches the computed prediction, that is, the
507: mass renormalization in the $\alpha$-band measured by ARPES closely matches that
508: measured by dHvA.
509:
510: Finally in Figure 5 we compare the computed temperature dependencies of the inverse
511: lifetimes at zero binding energy in the $\gamma$- and $\beta$-bands to that measured by
512: Kidd et al. \cite{kidd}. The $\beta$-band results are well-matched. Both computed and measured
513: values show only a weak dependence on temperature over a 100K range.
514: Our computations, nonetheless are consistent with a quadratic Fermi-liquid like dependency on the temperature.
515:
516: The $\gamma$-band results also agree well with that reported in \cite{kidd}. We roughly
517: expect the temperature dependency to behave as $A_T T^2\log (E_F/T)$ where $A_T$ is some constant. Similarly
518: we expect the dependency of $\tau^{-1}$ on binding energy to be (at leading order,
519: modulo the complications of Eqn. 9) $A_\omega \omega^2 \log (E_F/\omega)$.
520: From fits of the two computed curves we find that $A_T/A_w \sim 10$. The value of this ratio compares well
521: with the expected value, $\pi^2$, from a single quadratic band \cite{chubukov}.
522:
523:
524: \section{Discussion and Conclusions}
525:
526: A primary conclusion from our analysis to the inverse
527: lifetimes and effective masses is that there is a marked difference between the two sets of
528: bands, $\gamma$, and $\alpha$ and $\beta$ in terms of effective coupling
529: strengths. This difference is not a
530: simple consequence of the band dispersion and its effect
531: on the density
532: of states in intraband scattering processes, but rather it is due to a stronger
533: effective intraband interaction in the $\gamma$-band at low energies. This conclusion
534: is rather surprising since all three bands derive from orbitals belonging to the
535: same $t_{2g}$ manifold and standard multiband Hubbard models assign equal onsite interactions
536: to all three orbitals \cite{eremin,millis}.
537: Our results point to substantially different renormalizations
538: of the low energy effective intraband interactions in the two sets of bands.
539:
540: The different standing of the band $\gamma$ from the bands $\alpha$ and $\beta$ may
541: arise from the latter two band's highly one dimensional character. This character
542: raises the question of whether some form of Luttinger liquid behavior is occurring. The
543: fact that the transport properties show clear three-dimensional Landau-Fermi
544: behavior \cite{bergemann}
545: shows this behavior dominates at low energy. Nonetheless a crossover to Luttinger liquid
546: behavior could appear at higher energies. However the ARPES data reported by both Kidd
547: et al. \cite{kidd} and Ingle et al. \cite{ingle} show no sign of such a crossover up to energies of 60 meV.
548: Both the $\beta$ and $\alpha$ bands appear to have inverse lifetimes that depend
549: quadratically on the binding energy.
550: This perhaps is due to the hybridization and interband scattering between the three bands
551: which act to suppress a strictly one-dimensional character in the $\alpha,\beta$-
552: bands.
553:
554: Our computed inverse lifetimes broadly match the scale of the observed linewidths in all three bands.
555: Nonetheless they also consistently underestimate the corresponding measured values, more
556: pronouncedly in the $\alpha$- and $\beta$- bands than in the
557: the $\gamma$-band. This might suggest that some additional mechanism is making a contribution
558: to the self energy, at least in the $\alpha$- and $\beta$-bands. This might point to a possible role
559: played by some bosonic mechanism such as phonons. However if the shortfall is to be explained by phonons,
560: phonons need to make a greater contribution to the self energy than posited in Ingle et al. \cite{ingle}.
561: As we can see from the bottom right panel of Figure 2, the portion of the self-energy in the $\alpha$-band
562: measured by \cite{ingle} due to electron-electron interactions is double that computed in our
563: effective model.
564:
565: In the $\gamma$- and $\beta$-bands we see that ARPES predicts a mass renormalization far
566: smaller than that found in magneto-oscillation experiments. This might suggest that there is an
567: effective scale in the problem below the sensitivity of typically ARPES measurements (i.e. $\ll 1 meV$).
568: However we also found that the ARPES dispersion in the $\alpha$-band closely matches that predicted
569: by dHvA measurements. This might suggest that the surface aging performed by \cite{kidd}
570: to distinguish the bulk $\gamma$- and $\beta-$ bands from the surface counterparts changes the mass
571: renormalization in some
572: unexpected fashion. However our match to the $\alpha$-band dispersion measured in \cite{ingle}
573: is not without difficultly. Our match to their dispersion is predicated solely on electron-electron
574: interactions. If we were to ascribe a role to phonons, it would mean we have overestimated the coupling
575: strengths, $V_{\mu\nu}$. This in turn would mean we have overestimated the inverse lifetimes.
576: But our inverse lifetimes are already smaller than the measured values. It is unclear whether
577: phonons could then self-consistently make up the difference. Nor is it unproblematic to have phonons
578: be the dominant interaction in a material where it is believed electron-electron interactions
579: are responsible for its superconductivity.
580:
581: Of course, the discrepancies found in regards to
582: the real part of self energy might simply point to a need to go beyond our use of low order
583: diagrams based on
584: an effective Hamiltonian. It would be interesting to attempt a more sophisticated treatment of the set
585: of interacting
586: bands in $Sr_2RuO_4$. One approach may be to adopt a functional RG approach \cite{shankar}. Extensive
587: work of this sort has already been done on one-band two dimensional systems \cite{honer}.
588: It should be possible to extend this work to a multi-band system.
589:
590: We thank both Peter Johnson and Tim Kidd for numerous useful
591: discussions and direct access to their data. We are also grateful
592: to A. Damascelli for providing us with data from Ref. \cite{ingle}.
593: RMK acknowledges support
594: from US DOE under contract number DE-AC02 -98 CH 10886. TMR
595: acknowledges hospitality from the Institute for Strongly
596: Correlated and Complex Systems at BNL.
597:
598: %%%%%%%%%%%%%%%%%%%%
599: \begin{thebibliography}{99}
600: %%%%%%%%%%%%%%%%%%%
601:
602: \bibitem{kidd}
603: T. E. Kidd, T. Valla, A. V. Fedorov, P. D. Johnson, R. J. Cava and M. K. Haas,
604: Phys. Rev. Lett. {\bf 94}, 107003, (2005).
605:
606: \bibitem{ingle}
607: N. J. C. Ingle, K. M. Shen, F. Baumberger, W. Meervasana, D. H. Lu, Z. X. Shen,
608: A. Damascelli, S. Nakatsuji, Z. Q. Mao, Y. Maeno, T. Kimura and Y. Tokura,
609: Phys. Rev. B {\bf 72}, 205114, (2005).
610:
611: \bibitem{mack}
612: A. P. Mackenzie and Y. Maeno, Rev. Mod. Phys. {\bf 75} 657 (2003).
613:
614: \bibitem{bergemann}
615: C. Bergemann, A. P. Mackenzie, S. Julian, D. Fortsythe and E. Ohmichi,
616: Adv. Phys. {\bf 52} 639 (2003).
617:
618: \bibitem{morr}
619: D. Morr, P. Trautman, and M. Graf, Phys. Rev. Lett. 86 (2001) 5978.
620:
621: \bibitem{ll}
622: A. Liebsch and A. Lichtenstein, Phys. Rev. Lett. 84 (2000) 1591.
623:
624: \bibitem{eremin}
625: I. Eremin, D. Manske and K. H. Bennemann, Phys. Rev. B {\bf 65}, 220502(R) (2002).
626:
627: \bibitem{millis}
628: S. Okamoto and A. Millis, cond-mat/0402267.
629:
630: \bibitem{chubukov}
631: A. Chubukov and D. Maslov, Phys. Rev. B {\bf 68}, 155113 (2003);
632: A. Chubukov, D. Maslov, S. Gangadharaiah, L. Glazman, Phys. Rev. B {\bf 71}, 205112 (2005).
633:
634: \bibitem{chu1}
635: Deviations from a quadratically dispersing band can lead to a change in
636: the non-analytical structure of the self-energy (A. Chubukov and A. Millis, cond-mat/0604496)
637: if the curvature of the band at the Fermi surface in the transverse direction vanishes.
638: This is not the case here.
639:
640: \bibitem{shankar}
641: R. Shankar, Rev. Mod. Phys. {\bf 66}, 129 (1994).
642:
643: \bibitem{honer}
644: C. Honerkamp, M. Salmhofer, and T. M. Rice, Euro. Phys. J. B {\bf 27}, 127 (2004),
645: C. Honerkamp, D. Rohe, S. Andergassen, and T. Enss, Phys. Rev. B {\bf 70}, 235115 (2004).
646:
647: \end{thebibliography}
648:
649: \end{document}
650:
651:
652:
653:
654:
655:
656:
657:
658:
659:
660:
661:
662:
663:
664:
665:
666: