Fermi Edge Singularities in the Mesoscopic X-Ray Edge Problem Martina Hentschel, Denis Ullmo, and Harold U. Baranger NIRT Duke University program Metals Mesoscopic Systems Expected: • finite number of electrons on discrete level A( • coherent, chaotic geometry EF th • fluctuations Observed: h A( th Absorption EC {} {} A( ? core level bare th Martina Hentschel: The ‘classical’ X-Ray Edge Problem Yu und Cardona Fund.of SC, p.477 Singularities at the Fermi edge threshold in X-Ray Emission or Absorption Spectra of, e.g., metals e.g. Peaked Edge A( I( L2,3-edge simple metals like Al, Mg, (Na) from K. Othaka, Y. Tanabe, RMP 62 2929 (1990): GaAs-AlxGa1-xAs Quantum well, Lee et al. (1987) What happens when a core electron is excited? ~1023 cond. electrons respond Kondo Problem + 1 of 1023 electrons Many-body ground state |F> made of single particle wf. |ji> although <yi | ji> ~ 1: <Y|F> 0 (as N Sudden perturbation V Anderson Orthogonality Catastrophe Anderson Orthogonality Catastrophe (AOC) P. W. Anderson, Phys. Rev. Lett. 18 1049 (1967) ground state under perturbation V Perturbation can be small there is NO ADIABATICITY in those systems! Y F ~ N ~0 ground state initially Or any state entirely described in terms of plane waves Orthogonality Block Important in: • Fermi edge singularities of x-ray and photoluminescence spectra • Kondo physics • Tunneling (e.g. in double quantum dots) CHECK zero-bias anomaly (in • Similar phenomenon in particle physics dosordered systems) Peaked or rounded edge ? Mesoscopic effects Many-body effect “Mahan’s enhancement” Orthogonality block due to AOC finite N finite N Competition screening dipole selection rules chaotic geometry acts universal relative strength ? Sample-to-sample fluctuations Peaked or rounded edge ? A ( ) ( th ) P hotoabsor ption 2 wit h lo l counteracting (Mahan) manybody process (lo only) where Citrin, PRB (1979) Tanabe and Othaka (1990) 2 ( 2 l 1) l 2 Anderson orthogonality catastrophe (all l ) l .... phase shift for ang. mom. l , t aken at Fermi energy E F lo .... optically excited channel l 0 Friedel' s sum rule : Z l 2 ( 2 l 1) l Outline of talk I. Introduction II. Mesoscopic Anderson Orthogonality Catastrophe • Model, numerical method, results III. X-Ray Photoabsorption Spectra: Mesoscopic vs. Bulk-like • Fermi golden rule approach, role of dipole matrix elements IV. Conclusion, Experimental Realizations II. Anderson Orthogonality Catastrophe in Mesoscopic Systems AOC for a rank-1 perturbation V Tanabe and Othaka, RMP (1990) Aleiner and Matveev, PRL (1998) unpert urbe d : Hˆ perturbed : { k , k } ... F Hˆ Vˆ { , y } ... Y e.g. core hole left behind at r0 N V r0 r0 overlap between perturbed and unperturbed ground states: 2 Y F 2 M i0 filled N j M 1 empty ( j i )( j i ) ( j i )( j i ) = f (eigenvalues only) Example for a rank-1 perturbation • unperturbed level k: equidistant (“picket fence”, “bulk-like”) • perturbed level : Schrödinger equation 1 : k 50 1 V k d 6 level, attractive 25 pertubation V Martina y ~ Hentschel: 0 0 1 V big V small Check this – der ist gar nicht constant!!! -25 0 -50 -4 -2 phase shift 0 1 2 (N ) : b d 4 6 8 ( k k ) arctan V d Rank-1 perturbation in the mesoscopic case Fluctuations: k, k(ro) :: | k ( ro ) | k • Assumptions: {k} {|k(r0)|2} k 2 1 N V GOE / GUE distribution Porter-Thomas distribution Motivation: Random matrix theory chaotic systems: quantum dots, nanoparticles • Joint probability distribution P ({ i }, { i }) i>j (Aleiner/Matveev, PRL 1998) ( i j )( i j ) i, j i j N : i = const. = V-1 1 / 2 exp 2 (i i ) i i 1 ( for GOE (GUE) Boundary effects 50 d y 25 0 0 1 1 V big -25 0 -50 • -4 -2 run-away level 0 2 6 N d 0 e • 4 d /V 1 “pressure” from far away level level-dependent potential and phase shift N i ln i V d i 1 1 1 8 Workhorse: Metropolis algorithm on the circle M of N level filled • Start: picket fence (N+1 level , N+1 level , mean level spacing d/N, shift b) • Random number in (0, 2N+1) level i or i shifted within interval given by neighboring levels Every third step: move pair (i, i) Memory lost after ~ N steps d i+2 i+2 i+1 i i+1 Circle: constant DOS i i-1 i-1 N N 0 0 • • • Metropolis step: accept / reject change with PM=min(1, P({i},{i} generate many ensembles [{k},{k} || ] distribution of overlaps || Results: 1. Ground state overlap distribution P(||) Onset of AOC a) as perturbation V ~ vc increases b) as particle number N increases 25 5 N=1000 v c /d = -0.1 4 N=50 v c /d = -10 2 |Vc| 10 N=250 N=100 P ( | | ) 2 P ( || ) v c /d = -1 15 N N=500 v c /d = -0.25 20 5 3 N=10 2 1 0 0.2 0.4 0.6 overlap || 0.8 0 1 0 2 0.2 0.4 0.6 overlap || 2 b N | b | 2 2 1 b 2 e 0.8 2 2 bulk values 1 1. P(||) cont. Scaling and role of phase shift F at Fermi energy 5 | b | F / 2 2 P(| | ) 4 GOE: N , M , v c /d 250, 78, -1 3 100, 50, -5 2 100, 31, -1 1 1 0 0 , 1 4 , - 0 .5 0 50, 16, -1 0 0.2 0.4 || 0.6 0.8 1 1.2 0.8 2 P(| | ) 1 1 0.6 0.4 GU E 0.5 0.2 0 0 0 1 2 || / | b | 2 2 3 0 1 2 P(|| determined by phase shift F at Fermi energy (as in metallic x-ray edge problem) 3 Results: 2. Origin of Fluctuations in P(||) Reminder : | Reference |b - evaluate | starting at the Fermi edge EF : ( r 1) 2 | ( M 1 M ) ( M 1 M ) d d b range 1 N i0 filled j M 1 empty ( j i )( j i ) 2 M+1 EF M M+2 M+1 M M-1 filled i bulk case {} ( j i )( j i ) bulk case M+1 M b M range 2 bulk case empty j 2 bulk case {} M EF 2. Fluctuations in P(||) cont. GOE GUE 6 6 v c /d = -0.25 4 2 P (| | ) 2 P (| | ) v c /d = -0.25 2 0 0 0.2 0.4 0.6 0.8 4 2 0 1 6 0 2 P (| | ) 2 P (| | ) 2 0 0.2 0.4 0.6 0.8 1 0.4 0.6 0.8 1 2 0 0.2 6 v c /d = -10 v c /d = -10 4 2 P (| | ) 2 0.8 4 0 1 6 P (| | ) 0.6 v c /d = -1 4 2 0 0.4 6 v c /d = -1 0 0.2 0 0.2 0.4 0.6 2 overlap || 0.8 1 N=100, M=50 overlap || 4 2 range 1 range 2 2 0 anal. range 1 0 0.2 0.4 0.6 2 overlap || 0.8 1 2. Fluctuations in P(||) cont. 6 v c /d = -0.25 2 analytically understanding of overlap fluctuations: P (| | ) GUE • consider two level i = 0,1 around EF in the mean field of other level 4 2 0 0.2 0.4 0.6 0.8 1 0.4 0.6 0.8 1 6 v c /d = -1 2 s = (10) Wigner surmise |u0|2 , |u1|2 Porter-Thomas i 2: i , |ui|2, i one random variable P (| | ) • • • 0 deviation of range-n result from P(||): 4 2 0 0 0.2 6 v c /d = -10 -10 0 10 20 30 range n 2 -0.25 P (| | ) 0.1 D KS v c /d RMT justified ! N=100, M=50 overlap || 4 2 range 1 range 2 2 0 anal. range 1 0 0.2 0.4 0.6 2 overlap || 0.8 1 Summary part II AOC in mesoscopic systems bulk-like mesoscopic chaotic • {} equidistant {},{} fix • {},{} fluctuating (GOE/GUE) • single value || |b| • RMT treatment justified | • bulk: N , |b 0 • broad distribution P(||) • fluctuations dominated by levels around EF • analytic treatment of range-1 approximation AOC in disordered systems: Gefen et al. PRB 2002 AOC in parametric random matrices: Vallejos et al. PRB 2002 III. Mesoscopic X-ray Edge Problem Approaching the Mesoscopic X-Ray Edge Problem Fermi edge singularities in x-ray spectra of metals A( Absorption EF h ___ Absorption I( Emission EC {} cor e bare Misses many-body effects of core hole potential on cond. e- : AOC and Mahan’s enhancement Emission bare A( A( I( I( • diagrammatic perturbation theory (Mahan, Nozieres,…) • Fermi golden rule approach (Tanabe/Othaka) Model: Fermi golden rule approach A ( ) 2 Y F Wˆ F c 2 ( EY EF F C F Tanabe and Othaka, RMP 1990 Dipole matrix Y F Wˆ F c element unperturbe d : c k k ( k 1,..., N ) ( 1,..., N ) ~ ~ ~ Y 0 c M c M 1 ... c 0 0 N Wˆ c cc 0 ~ c y dipole operator : core elect ron : c c c perturbed : F 0 c M c M 1 ... c 0 0 k 1 w kc [ c k c c h.c. ] Model: Fermi golden rule approach A ( ) 2 Y F Wˆ F c 2 ( EY EF F C F g j EF F M m i 0 {} {} h ___ core o direct process ~ |wjc|2 || replacement shake-up ~Sm |wmc|2 |mg| repl. direct Dipole matrix element wjc w jc c y j r x ray E ~(ro y j nonloc bulk-like e ik j r il J l (k r ) e w jc φ c u jc = s-like: jc = p-like: mesosc. jc = s-like: jc = p-like: k j ',| k j '| | k j | l orbital channel, partial wave decomp. bulk-like s-like mesoscopic l V= (r-r0) l=0: s-like cond. el. u (r ) nk j ~J0’(ro) = 0 =0 a kj' e ik j 'r l not conserved in chaotic systems φc u =0 ~J0(ro) 0 wjc =0 at K-edge rounded wjc 0 at L-edge peaked wjo ~ ’(r0) ~ yj` peaked or rounded K-edge (yj`, yj indep.) wjo ~ (r0) ~ yj stronger correlations at L-edge 0 5 10 15 20 0 Results:300 0 0 50 0 vc10 = -10 d, 15 K-edge 20 N = 100, M = 50, GOE 5 200 0 {} 150 V = 10 A () 1 A () 10 A () A () a) Contributions from the various processes 100 ( 5 V= 1 1. Average Photoabsorption K-edge 100 200 1.5 10 150 V= 1 2 5 100 V = 10 2 ~|w jc| 100 shake-up bare 0.5 50 0 0 0 1 0 0 j 2 53 4 10 5 ( ( ) / d / th 615 7 20 8 0 5 ( M T total (direct+repl.+shake up) i naiv e bare (norm.) direct+repl. 10 T / direct shake up 0 • peaked edge • replacement processes near EF dominate • one-pair shake-up processes dominate {} {} ~|wjc|2 || direct process replacement direct + replacement Results: 1. Average Photoabsorption K-edge EF b) Taking spin into account 2 vc = -10 d, K-edge N = 100, M = 50, GOE ( 1.5 active 1 full spin 0.5 a ctive sp in 0 0 1 2 3 4 ( 5 th ) 6 7 8 /d EF spectator width of |F0 in basis of perturbed final states |YF spectator spin Comparison with bulk-like case L-edge bulk-like ( 3 1 K-edge mesoscopic 0.3 0.1 vc = -10 d, GOE K-edge bulk-like 0 1 2 3 4 ( 5 th ) 6 7 8 /d Rounded edge goes into a (slightly) peaked edge as the system becomes coherent M.H., D.Ullmo, H.U. Baranger, cond-mat/0402207, subm. to PRL Dependence on the number of electrons 2 N : Anderson wins Mesoscopic K-edge ( 1.5 bare N=24, M=12 N=50, M=25 1 N=100, M=50 N=200, M=100 0.5 0 0 1 2 3 4 5 6 ( th ) / d 7 8 9 10 10 Bulk-like L-edge ( 7.5 N : Mahan wins 5 2.5 0 0 1 2 3 4 5 6 ( th ) / d 7 8 9 10 Results: 2. Average Photoabsorption L-edge Coupling to the wave function: wjo~ yj 50 y 25 bound state 0: 0 0 1 1 • y0 (r0) piles up • screens core hole • s-like V big -25 0 -50 EF -4 -2 0 2 4 6 || 0 8 vc= - 0.1d vc= -10 d 1 0.03 i=0 0.025 0.8 i=1 2 0.015 i=M=N/2 (E F) i=N 0.4 0.01 100 random plane waves, N=100, M=50 0.2 0.005 0 -10 0.6 |y i | |y i | 2 i=2 0.02 -5 0 x 5 10 0 -10 -5 0 x 5 10 Average Photoabsorption L-edge cont. 40 vc = -10 d, GOE, active spin Mesoscopic L-edge < A ()> 30 20 10 0 0 1 2 3 4 5 6 < th >/d 7 8 9 10 40 N=100, M=50 < A ()> Bulk-like L-edge • small differences N=50, M=25 30 mesoscopic vs. bulk-like, 20 and GOE vs. GUE 10 • edge peak with N 0 0 1 2 3 4 5 6 < th >/d 7 8 9 10 Results: GOE, N=40, M=20, V 3. Mesoscopic fluctuations in=-10A() = f ( ||2, repl ,shup ; wjc) c GOE, N=40, M=20, g1V P(A( ) / <A( )>) K-edge 2 c =-10 K-edge g g1 g g g g g0N 1.5 g P.Th. g0N 1 P.Th. 0.5 0 0 1 2 GOE, N=40, M=20, V 3 4 A() / <A()> GOE, N=40, M=20, g1V P(A( ) / <A( )>) L-edge 2 L-edge g g g g0N g P.Th. g0N 1 P.Th. 0.5 0 0 1 2 A() / <A()> 3 c =-10 c =-10 g g1 g 1.5 wjc ~ y’ large Porter-Thomas like fluctuations overwhelm overlap correlations and dominate fluctuations of A() 4 wjc ~ y narrowly distributed Symmetry: replacement through bound state acts like a ground state overlap with F’ = F , results in highly peaked edge Experimental Realizations “Fermi sea of electrons subject to a rank-1 perturbation” • x-ray photoabsorption with metallic nanoparticles: feasible in few years • double quantum dots: constriction Abanin/Levitov, cond-mat/0405383 • photoabsorption via impurity states in semiconductor heterostructures Quantum Dot Array (diam.~100 nm) etched from heterostructure 2DEG and impurities GaAs Control Experiment “bulk-like”: no dots, just 2DEG with impurities - already done? Summary part III Mesoscopic X-ray Edge Problem s-like conduction electrons: 0= /, 1= 0 bulk-like mesoscopic • rounded K-edge • (slightly) peaked K-edge Average A( • peaked L-edge • peaked L-edge Dipole coupling changed because mesoscopic system is - chaotic (loose l as quantum number) - coherent confinement - wave function and derivative independent Mesoscopic fluctuations • individual spectra can even zig-zag IV. 1.2 AOC in Mesoscopic Systems: - broad distribution P(|| - scaling with |b|, F 2 2 P (| | /| b | ) • Conclusions 1 0.8 0.6 0.4 0.2 0 0 1 2 || / | b | - K-edge: A( from rounded to peaked as system becomes coherent, Porter-Thomas fluctuations - L-edge: strongly peaked, same fluctuations as || 3 3 1 0.3 0.1 • 2 Mesoscopic Photoabsorption Spectra and X-Ray Edge Problem: ( • 2 Experimental realizations: - array of quantum dots, impurity level takes role of core electron - nanoparticles, double dots 0 1 2 3 4 ( 5 th ) 6 7 8 /d M. Hentschel, D. Ullmo, H.U. Baranger, cond-mat/0402207