THE BEHAVIOR OF THIN-FILM SUPERCONDUCTING-PROXIMITY-EFFECT SANDWICHES IN HIGH MAGNETIC FIELDS by J. WILLIAM GALLAGHER University Creighton B.S., 1974) SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY August, Q Signature 1978 Massachusetts Institute of Technology of 1979 Author Department of /PhysicFs, August 11, 1978 Certified by TheIis Accepted Supervisor by rhairman, Departmental Committee on Graduate Students 2 1973 NOV 2 1973 OF LIBRARIES 2 THE BEHAVIOR OF THIN-FILM SUPERCONDUCTING-PROXIMITY-EFFECT SANDWICHES IN HIGH MAGNETIC FIELDS by GALLAGHER WILLIAM J. on August Submitted to the Department of Physics 11, 1978, in partial fulfillment of the requirements for Degree the of Doctor of Philosophy ABSTRACT study the behavior tunneling Hamiltonian model to We use a two-metal proximity-effect superconducting thin of the develop We fields. sandwiches in high parallel magnetic formalism in manner a valid for all temperatures and field the zero-temperature at explicitly calculate the potential, pair the states, of dependence of the density free energy magnetization, the density, and the The the sandwich. of sides on both their in splitting field-induced a spin-susceptibility display sandwiches The of states. spin-densities metal superconducting stronger effect from proximity sandwich keeps in the the the spin-split its after even superconducting metal weaker The level. Fermi the crossed have of states densities by accompanied is states of spin-densities crossing of the pair the of degradation field-dependent a of onset the magnetic nonzero a of onset the by and potentials superconducting the values, field some At susceptibility. may even exceed the state susceptibility in some sandwiches Pauli susceptibility state. normal the of Preliminary indication of the crossing tunneling experiments do give an We Fermi level. at the states of the spin-densities of comment on these experiments and indicate several directions optimize the experimental pursued in order to which can be In a final chapter we show how to properly characteristics. estimate the spin-orbit scattering times in thin reasonable agreement between our superconductors and find a values. experimental the and estimates Thesis supervisor: Brian B. Schwartz Head, Theory Group, MIT Francis Bitter National Magnet Laboratory Dean, School of Science Brooklyn College of CUNY 3 ACKNOWLEDGEMENTS I B. am pleased to for giving me Schwartz aspects have the an of superconductivity, I am for me for for many suggesting the Brian subtle problem patiently supervising the his openness with grateful for also especially thank Dr. appreciation described in this thesis, and work. to opportunity and for his interests and abilitiy in aiding my education in matters beyond those related to this thesis. many aspects I of his influence will from benefitted remain with me. interesting physicists conversations with particularly Drs. many I hope that and at the Magnet instructive including Lab, Sonia Frota-Pessoa, Robert Meservey, Paul Tedrow, and Demetris Paraskevopoulos. For other useful and interesting discussions I am particularly grateful to two of and Ronald Andre Tremblay student colleagues, my graduate Pannatoni. For instruction programs Ann Carol the IBM Thomas J. of MIT Watson Research for noting in using text Carol H. Center Paul Wang of this various editing document, I thank Mrs. Mrs. I also thank the figures with this Hohl and particularly some of Bostock and help for the preparation of Heights, New York. with in thesis Thompson of in Yorktown MIT for help and Dr. Judith gramatical errors and 4 For aid thesis. statements in a draft version of this vague assistance of Don in using MIT's computers I acknowledge the Nelson of MIT's Magnet Lab. the support of the financial gratefully acknowledge I National Science Foundation through a Predoctoral Fellowship of my graduate studies and through for the first three years the Core Grant to the Francis my am grateful I also final year. Bitter National Magnet Lab for Center for the preparation of this In back looking culminating with the encouragement Dr. J. Sam Nebraska, and School also this thesis, of two of Cipolla of I Omaha. Wang who, in a few days, are physics instructors, in Omaha, Creighton Preparatory am Finally I that to acknowledge University Creighton the encouragement and study of my early continual encouragement of my parents, months, for Research Watson would like Fr. Willard Dressel of in text its document. years over the International use of J. Thomas the at facilities processing the for Machines Corportation Business to the grateful for and, over the last support of will be my wife. the 13 Martha Liwen 5 CONTENTS Abstract . . . . . . . . . . . . . . . . . . . . . . . . . 2 Acknowledgements . . . . . . . . . . . . . . . . . . . . . 3 List of figures. . . . . . . . . . . . . . . . . . . . . . 9 . . . . . . . . . . . . . . . . .12 Introduction . . . . . . . . . . . . . . . . .13 . .13 Concepts in the Theory of Superconductivity. .21 List of tables Chapter I: . . . . . An Overview of Superconductivity and the Magnetic A. Properties of Superconductors. . . . . . . . . . B. Basic C. Magnetic Properties of Thin Film Superconductors in Parallel Magnetic Fields The Proximity Effect D. Chapter II: . . . . . . . . . . . . . . . .30 . . . . . . . . . . . . . .49 for the Proximity Effect .54 Theoretical Models . A. Introduction B. Theories Based . . . . . . . . . . . . . . . . . . .54 on the Gor'kov and the Ginzburg-Landau Equations. . . . . . . . . . . . . . . . . . .57 . . . . . . . . . . . . .63 C. The Cooper Limit D. Bogoliubov-de Gennes Equation Approach to the Proximity Effect . . . . . . . . . . . . . . . . . .66 . . . . . . . . . . . . . . . . . . .66 . . . . . . . . . . . . . . . . . . .67 1. Introduction 2. Review . . . 6 McMillan Tunneling Model of Thin Proximity-Effect E. . Sandwiches . . . . . . . . . . . . . . Tunneling Model of Paramagnetically Limited Chapter III: Formalism Proximity-Effect Sandwiches: . . . . . . . . . . . . . . . . . .. . . A. Introduction B. The C. Green's D. Iteration of the Green's Function Equations. E. McMillan Solution to the Hamiltonian. . . .80 . . . .80 .82 .87 Motion. Tunneling Model for .92 . . . the -98 . . . . . . . . . . . . . . . .. Comparison to the Diagrammatic Expansion of the Function Green's 2. of their Equations and . -. . . . . . . . . . . . . . .. Functions Proximity-Effect 1. .74 . . .. . . . . . . . . . . . . .. .. Equations for the Renormalization Functions. Chapter IV: -98 . . . 101 the Tunneling Model Predictions for Properties of Paramagnetically Limited Proximity-Effect Sandwiches. Introduction: A. . . . . . 107 Calculational Procedures. . . . . . 107 . . . . 109 . - 111 . . 120 . . . . . . . . . . . . . . . .. Renormalization Functions. . . . . . . . . . . .. 1. The 2. The Density 3. The Magnetization, Susceptibility, and Free Energy 4. . . of States. . . . . . . . . . . . . . Finite Temperature Calculations . . . . . . . . . . 7 B. A Normal-Superconducting Sandwich. C. A Two-Superconductor Chapter V: Sandwiches . . Sandwich...... . . . . . 124 . . . . . 14 1 on Thin Proximity-Effect Experiments in High Magnetic Fields. . . . . . . 157 . . . 157 . . . 159 A. Introduction . . . . . . . - - - - . B. Tunneling Experiments. . . . . . . . . . . Tunneling. . . . . . . . . . . . . . . 159 Measurements . . . . . . . . . . . . . 165 . . Theory 2. Tunneling 3. Suggested Future Tunneling Work. . . . . . . . . . 176 Measurements of the Magnetic Susceptibility. Summary Chapter VI: . . . . . . . . . . . . --.. . . A. Introduction B. Background . . . . . . . Superconductors. Spin-Orbit . . . . . . . . 181 183 .. ..... . . . . . . . . . . . . . .. on the . in Spin-Orbit Scattering Times Superconductors . . . . - 185 . 185 Interaction in . . . . . . . . -- 188 . Matrix Elements from Spin-Orbit and Regular C. Impurity D. . 1. C. D. of . . . . . . . . 195 Scattering Hamiltonian. . . . . . . 195 . . . 198 . - - - 203 Scattering. Impurity . . . . . . . 1. The 2. Estimation of Scattering Matrix Ion Core Screening Delocalization and Metallic . Elements Electron . . . . . . . . . . . .. . 8 E. Comparison of Estimated Spin-Orbit Scattering Times 1. to Experiment. Comparison of the . . . . . . . . . . . Ratio to . . 207 the of the Scattering Potential. 207 Estimating the Contribution of Surface Scatterers 3. . Scattering Time Ratio of the Square 2. . . . . . . . . . . . . . . . . . . . . 211 Conclusion and Suggested Further Experimental Work . . . . . . . . . . . . . . . . . . . . . . 214 Chapter VII: Summary and Conclusion . . . . . . . . . . 217 Some Results from the Bogoliubov-de Appendix A: Gennes-Equation Approach to the Proximity Effect. . . 220 Calculation of the Sandwich Superconducting Appendix B: Transition Temperature as a Function of Field Biographical Note. References . . . . . . . . . 229 . . . . . . . . . . . . . . . . . . . 233 . . . . . . . . . . . 235 9 List of Figures I.1 BCS density of states I.2 Film . . . . . . . . . . . . .25 and sandwich geometry. . . . . . . . . . . . .32 I.3 Meissner diamagnetic effect . . . . . . . . . . . .33 I.4 Partial Meissner effect in thin films . . . . . . .34 1.5 Spin-split . . . . . . .37 I.6 Pauli Paramagnetic limit in thin films . . . . . . .39 1.7 Pair potential as a function of field = 0. . . 40 1.8 Free energy in the paramagnetic limit at T = 0. . .42 1.9 Free energy in the paramagnetic limit for T 1 0 . .44 1.10 Paramagnetic limit phase diagram. I.11 Metamagnet T-H-Hst density of . . states. . . . . T at . . . . . . .45 . . . . . . .47 II.1 Density of states from the tunneling model. . . . .78 IV.1 Free energy, Magnetization, and Susceptibility phase diagram. of an isolated film at T =0 IV.2 Density of states for Density of states for . . . . . . . . . . . . 116 . . . . . . . . . . 127 . . . . . . . . 128 Pair potential as a function of field for a strongly coupled sandwich at T IV.5 . a strongly coupled sandwich at T = 0 . IV.4 . a strongly coupled sandwich at T = 0 . IV.3 . . = 0. . . 130 . Total density of states for the 'n-side of a strongly coupled sandwich at T = 0 . . 132 10 IV.6 Total the s-side density of states for of a strongly coupled sandwich at T = 0 IV.7 . . . . . . . . . . . . . . . . . . 135 . . . . . . . . . . IV.14 . . . . . . . . . . 137 . . . . 139 . . . 145 . . . . . . 146 Pair potential as a function of field for a weakly coupled sandwich at T = 0. . . . . 148 Density of states for . . . . . . . a weakly coupled . . . . . . . . . . Magnetization of a weakly coupled sandwich . . . . . Total density of states . . . .. 149 . . . . . . . . for the n-side . . . . 150 . . . . 151 . . . 153 Total density of states for the s-side Susceptibility of a weakly coupled sandwich at T = 0. IV.18 . . . of a weakly coupled sandwich at T = 0 IV.17 . . . . . . . . . . . . . . . 143 of a weakly coupled sandwich at T = 0 IV.16 . . at T = 0. IV.15 . . sandwich at T = 0 IV.13 . . Density of states for a weakly coupled sandwich at T = 0 IV.12 . . Density of states for a weakly coupled sandwich at T = 0 IV.11 . . . . Free energy of a strongly coupled sandwich at T = 0. IV.10 133 Susceptibility of a strongly coupled sandwich at T = 0. IV.9 . Magnetization of a strongly coupled sandwich at T = 0. IV.8 . . . . . . . . . . . . . . . . . Free energy of a weakly coupled sandwich at T = 0. . . . . . . . .. . -. . . . . . . 154 11 V.1 Thermally smeared n-side density of states for a strongly coupled sandwich V.2 . . . . 163 sandwich . . . . . . . 164 . . . . . . . . . . . . . . . . 171 Conductance for tunneling into 37 A Al backed by VI.1 . Conductance for tunneling into 25 A Mg backed by 40 A of Al V.4 . Thermally smeared s-side density of states for a strongly coupled V.3 . 15 1 of Cu . . . . . . . . . . . . . . . . 175 Scattering time ratio atomic number as a function of . . . . . . . . . . . . . . . . 208 12 List of Tables . . . . . . * .93 IV.1 Proximity effect sandwich parameters. . . . . . * 123 VI.1 Parameters data. . . . . . . 205 VI.2 Measured . . . 213 III.1 Green's function dictionary times . . . . . . . spin-orbit scattering predicted . . spectral from atomic and . . . . . . . . . . . . 13 INTRODUCTION CHAPTER I: A. AN OVERVIEW PROPERTIES OF OF the potential and magnetic fields magnetic properties of Kamerlingh interest Onnes's and Meissner destroy and the perfect diamagnets, there was no have it was determined hysteresis This implied that London 4 electrodynamics beyond afterwards the formulated for superconductors. flux applied to were in fact cooling. Thus below its cooled a magnetic field and Gorter and Casimir theory in a in 1933, reversible thermodynamics two-fluid thermodynamic soon that Later, upon metal was after hundred gauss was that all flux when a be applied to superconductors out a realized superconductors expelling of constant Shortly superconductivity. superconducting transition temperature and applied. been a bewilderment. (B =0), by H. variety implied by Maxwell's equations conductors perfect fascinating field of a few Ochsenfeld 2 change expulsion the discovery, 1911 uses of superconductors to of superconductors relatively small magnetic enough to MAGNETIC SUPERCONDUCTORS Kamerlingh Onnesi, subject THE discovery of superconductivity in Since the generate AND SUPERCONDUCTIVITY 1934. a Later, F. 3 could worked and H. macroscopic in 1950, H. 14 Londons and the other perfect diamagnetism function was what led to wave of the a "stiffness" that somehow suggested electromagnetic properties of superconductors. By experiments 6 . superconductors came in when Cooper 8 to state a bound-pair Finally free electron gas. in phonons, at the Fermi the to the surface and of the ground state 50 nearly 1957, such as led by of the normal an instability demonstrated that interaction, mediated interaction formation of thus 1956 attractive electron-electron effective spectrum of The key insight into what was happening in superconductors. an giving were the excitation gap in an energy evidence for measurements 7 heat Specific shift isotope from in coming were superconductivity phonons in to the role of early 1950's clues as the years after the discovery of superconductivity, a microscopic theory was put forth by Bardeen, Cooper, and Schrieffer'. The of genesis end signaled the properties Abrikosov 1 0 properties. Ginzburg-Landauli transition discovery of more of the theory of to predict superconductor superconductor including superconductors, of in had earlier the second the existence which magnetic in a theory the microscopic regular array by no means fascinating new new (1956) magnetic used the order superconducting of a flux would second type of penetrate the of quantized vortices. 15 Abrikosov's startling prediction went when, stimulated by Goodman's until the early sixties years publicizing 1 2 Abrikosov's was it work, of Abrikosov's realized that theory could clear up several anomalies that had properties magnetic the in observed been unnoticed for several of superconductors. is It now that there well known two classes are of superconductors distinguished most clearly by their magnetic properties. The first type is characterized by the perfect Ochsenfeld. diamagnetism first observed by Meissner and second is flux can characterized by an quantized vortices. superconductor the penetrate Type II in 600 kilogauss' some europium doped 3 for Pb 1 ternary detailed microscopic theory Type II superconductors an in which array of superconductors, in contrast to type I superconductors, can have very large up to state intermediate The MO 1 critical fields, S(06 and even molybdenum sulfides1 4 . of the upper critical has been work out higher for A field of by Maki's and 16 Werthamer, Hohenberg, and Halpern though certain anomalies between theory and experiment have become evident in recent years*. Also in *See the early sixties, Chapter VI for references. Brian Josephson 1 7 was led by 16 some considerations of broken symmetry and by meticulous use some separated superconductors. function and junctions Nowadays Josephson fields. as potential are commonly and in elements storage and switching placed in and galvanometers sensitive magnetometers employed in intimately connected, features when show dramatic the wave phase of Because the two in functions wave potential are the vector Josephson junctions magnetic a thin result of the difference in macroscopic the between phases tunneling two superconductors separated by These effects are a insulator. the in effects dc and ac startling characteristics of show to predict newly introduced tunneling Hamiltonian's of the computers. was Historically there This understand. of shift lowered many towards proportional to to was the Knight the finally cleared difficult very As was shift up by most expected to was to be susceptibility of of the combined problems refinements shift debate is of the Knight *The history the spin-orbit Chapter VI, where fully in is analyzed superconductors given. references are in Knight temperature the the vanishing Pauli spin superconducting electron pairs*. were be anomalously nonvanishing superconductors. zero magnetic effects of class proved that superconductors one systematically, here of discussed more interaction in and detailed 17 experiment and theory. done on aluminum showed vanishing combined Knight as the cause of the early experiments that aluminum shift. realization superconductors finite Repetition in Theoretically that (1) the temperature did is there spin-orbit Pauli superconductor approaches was susceptibility lowered and have a the impurities (2) to the orbital susceptibility does not vanish as the the fact in remain Van Vleck temperature of zero. While presently the Knight shift problem is thought to be understood, quantitative systematic, theoretical analyses of the spin-orbit superconductors have not been done. importance of spin-orbit This in scattering and experimental scattering in is in spite of the allowing for very ultrathin films high-critical-field superconducting materials. In in recent years high magnetic tunneling fields have in shown directly experiments allow scattering times a direct the reduced from density of stateszo. However, scattering in weaker dependence on the the theoryz1. determination spin of experiments atomic number Zeeman the These spin-orbit splitting of the magnitude these the states 1 9 . the quasiparticle density of splitting of spin-orbit experiments the of the observed shows a much Z than that given by 18 of between competition a showing Materials research. of field fascinating a remain themselves properties the superconductors, of properties magnetic the in the understanding remaining difficulties Besides these superconductivity and long range magnetic ordering have been practical, Moreover, recentlyzz. discovered are mandatory high-critical-field superconducting materials for being now technologies energy new the of many pursuedz 3 . ultrathin namely superconductor or to a and normal each other over high parallel proximity to another Superconducting metal normal metal. thousands of proximity effect. In of tens up to distances contact, influence placed in metal films, when Angstroms via an this in placed regime, in superconducting films fields magnetic unexplored an in superconductors of properties magnetic the of study the with deals thesis This effect known as the proximity effect is extended to thesis a theory of the apply to thin films in parallel magnetic fields and dramatic changes in the structure superconductor-normal Preliminary stimulation experiments, systematic metal however, enough to predictions show are are sandwiches undertaken experiments by these density of the for not verify the yet as a states predicted. result new features. clean detailed of of The enough and structure and 19 indeed can eliminate totally) scattering this on the spin effects of a parallel to a superconductor, magnetic field applied present so in of the nature superconductors, which features of Instead recently have superconductors, are considered intuitive ideas about of the spin-orbit interaction are use a simple from emerged from single in We find in detail. in that the atomic number Z dependence simple, interaction in is not scattering experiments tunneling at perplexing certain the spin-orbit spin-polarized the conventional analysis theory to the spin-orbit scattering here. detail a systematic comparison compared quantitatively the complication of considered proximity The proximity effect experiments to experiment can be made. yet be effects of the of thin-film the properties sandwiches must be considered before cannot (and drastically alters scattering Because spin-orbit thesis. in this systematic behavior calculated variance with vastly at of the magnitude of the show further We these experiments. spin-orbit how to assumption about the contribution of scattering displaced estimate of the surface atoms to yield spin-orbit scattering time, a and quantitative we compare these estimates to experimental values. Following an introduction in this of superconductivity which chapter of the concepts are important in this work, 20 Chapter II presents an in depth review of theoretical models for the proximity effect. of one of the models, thin sandwiches In tunneling Hamiltonian model, to the a in high parallel detailed a IV Chapter Chapter III provides an extension fields of of picture zero-temperature behavior of these interest here. expected the sandwiches is presented. In Chapter V we give a comparison of preliminary experiments with theory. experiments We also indicate directions the future should take in attempting to test critically for the properties discussed in Chapter IV. improved quantitative superconducting films. conclusion and summary. picture of Finally Chapter VI gives spin-orbit scattering Chapter VII provides an in a 21 IN THE BASIC CONCEPTS B. baffling as to Schafrothz 6 Hamiltonian showed remained 25 based on Bardeen , self-energy, Furthermore derived be effect cannot Frohlich the from electron-phonon the features of myriad the starting that, which described Meissner the role superconductivity. of characteristic this the electron resembling nothing and FrohlichZ4 modifications of phonon induced led by theories play they How superconductivity. for explanation the in role crucial a play recognized that phonons would fifties it was In the early SUPERCONDUCTIVITY OF THEORY interaction, of order in any perturbation theory. Also specific in the early measurements 7 heat an energy superconductors. excitation spectrum of demonstrated 27 Pippard's that for nonlocal from evidence there was fifties, in gap the further Bardeen electrodynamics 28 would likely follow from a model containing an energy gap. Leon Cooper that, supplied the in the presence of interaction (such as electrons in the ground state that key missing He showed any type of effectively attractive mediated by vicinity of the Fermi of the concept8 . electron gas phonons) surface, is unstable between the the normal against the 22 formation of consists of states states so and spin attractive scattering enjoy the as to of opposite momentum pairs scattering coherently into other electrons and spin momentum opposite bound-pair The electrons. pairs of bound interaction. binding energy pair 2.A where problem are pairs in the Cooper The 4 IwD given 2A e WAL~D is the width of surface where there is by a characterized by: N(EF:)V(1) the Fermi the energy region above assumed to be an attractive effective electron-electron potential of strength V, 2N(E and ) is the density of electron states at the Fermi level when the metal is in the normal This result state*. function of the potential V and an analytic is not it explained the failure of the earlier perturbative approaches. this concept of With Cooper, and Schrieffer bound pairs of (BCS) 9 were able electrons Bardeen, to write down a new ground state wave function consisting of many bound pairs electrons and to describe superconductors in terms that N(E ) *Note orientation ony. is the of the observed this ground density properties state and of states for one of of single spin 23 electron and pair excitations found an excitation spectrum clap N(E,)e V VE where Fermi They state. with a zero-temperature energy by: given 2A 0 ground this above N(E)V ) is now the width 2p), surface which in effectively the is there the near the region of the energy The attractive electron-electron interaction of strength V. quantitity A , which is nonzero ordered phase, referred to often is only in as the low-temperature the order parameter or pair potential of the superconductor*. Bardeen, Cooper, and Schrieffer temperature of a superconductor was given by I .~ S . where k. -t e u. and is Boltzman's constant zero-temperature energy gap related is _ _ I.3) N(F)V is Euler's constant. ' seen that in the weak coupling It can be _ 1.13 '(w#e N(F,)V transition that the found limit to the (V- O) the transition temperature by A0 IC (1.4) G. *Although the order parameter equals half the energy gap in the not is this superconductor, BCS a proximity-effect the In superconductors. here, considered well as as in magnetic impurities, the density of the BCS form displayed shape from energy gap, different which we from twice the will order denote all for case superconductors superconductors with states has a different and the in Figure I.1, by 21)., parameter, 2A. is generally 24 this is generally useful for NCE )V less than 0.25. and plot of the density of states Figure I.1a is a schematic states is BCS N theory the gap. In the the Fermi level, A measurement of the below and above the energy singular near of states density density of that the It is evident "semiconductor model." the as known is what in superconductor BCS a of is given by: (E), IEI.5) Ns(E) = N(E,) Re I.1b. is plotted in Figure and this junction, metal-insulator-superconductor normal a of conductance Giaever 2 9 , by as pioneered the superconductor's measurement of direct provides a density of states and therefore its energy gap. The is lower superconducting state normal state ground condensation G energy temperature T = 0) of the - G61 N free per than the in energy by gas electron unit volume a (at given by: 2 S(T= 0) One can get FigureI.1 energies a - "feel" by noting that lowered by G(T o) for - this 2N(E aMounts N rN(E,) condensation energy from occupied states have their ) of order /\ A type I 25 BCS DENSITY OF STATES (0) 0 -EF (b) -A 0 ENERGY Figure 1.1 A 26 the of lines thermodynamic critical field Hcb (T =0) simply related the Fermi level. was done here, is chapter where is critical field density of and the order parameter states at this given by: the thermodynamic to the a -(1.7) -rT=0) seen that can be It to leads this and field magnetic the flux field expels in a magnetic superconductor placed Balancing free energies, such as section of the next done frequently in of thin BCS superconductors the behavior in high magnetic fields is discussed. The original BCS theory was a gauge particular wave function many-body Once the number. assimilated, more the using theory were independently, 31 transformation techniques; aspects Green's were Anderson useful for of superconductivity; function techniques few lines. We will employ Bogoliubov-Valatin techniques 3 apply 2 developed seeing the Gor'kov 33 and, canonical a pseudospin powerful utilized results in a 3 both a generalization 4 and were broken symmetry 1 7 to derive the BCS transformation in this thesis. to quick 30 Bogoliubov forth: put a of formulating concise ways elegant and theory of the ideas on particle not conserve did which quickly approach variational a fundamental Valatin formalism which is rather clumsily formulated in Green's of the function 27 superconductivity, scales importance in of diamagnetism characteristic of type fact perfect from follows equations, length A samples. As electromagnetic constituative London4 the I superconductors is in oriented bulk only for suitably "perfect" The superconductors. length two are that there knows one of theories macroscopic earlier the From diamagnetic screening occurs exponentially over a called the clean In depth. London penetration superconductors this is given by: mC where mcz 100 Typically 2 is such as superconducting metals, in clean prototype Ginzburg-Landau theory scale near theories, Ginzburg-Landau phase-transition of length order parameter T the London electrons in of superconducting phase transition, contain the the electron, and nS is the to 500 R. All the %-rvi~e(I.8) is the rest energy of number density theory. A the of the superconducting over which of the spacial variations For temperatures are energetically allowed. transition that gives a coherence length temperature T , length this -- characteristically superconductors temperatures, this where diverges as length also it can be (1 has - T/T meaning thought of ) In . at as low the 28 At pairs. Cooper the of radius root-mean-square zero-temperature it is simply related to the energy gap: (1.9) tr A, as electrodynamics over when lengths in Impurities the are clean superconductors length 3 coherence 5 the decrease impurities temperatures the coherence the order length $ Coherence is given 104 of approximately decreases and ( and t by: I The penetration low At in terms of the pure superconductor coherence length the mean free path f. depth and leng th. coherence in reasonable, intuitively seems As . on show can parameter London penetration the both modify the be important chapter. next the in proximity-effect sandwiches order the of variations consider we parameter This interpretation will spacial variations. of the minimum distance order superconducting the degree the temperatures, all be thought of as length can which characterizing at Physically, non-locality. coherence length the non-local Pippard's in appears also length This depth, on the at low 0 other hand, increases temperatures is 10) 2(I. given by: as ., 29 Near the transition temperature this length also diverges as (1 - T/Tc ' 30 IN SUPERCONDUCTORS FILM OF THIN PROPERTIES MAGNETIC C. PARALLEL MAGNETIC FIELDS depth penetration the by characterized superconductors, diamagnetism of of the The interplay A , the and superconducting wavefunction, which varies on a length scale coherence length given by the properties of magnetic without other hand, <2 A in , type In array type ignored of II and flux penetrates the regime considerations some depth 3 6 . of the On result when the the diamagnetic the respectively, superconductor can be almost the Zeeman energy associated with In high fields is flux to intermediate and mixed states superconductors, type spin alignment of I in which a flux penetrates parallel fields, I and the type quantized vortices. ultrathin films in uniformly. which II superconductors, effects which give rise to the in not allow superconductivity. display a mixed state a periodic are samples to the destroying Superconductors ( than less in bulk superconductors and enter superconductors. of depth penetration with results in a rich variety , electrons interest important in this here becomes and regime are important. the free This energy now described in 31 of thickness d located in Consider an infinite flat film superconductor illustrated an has planes between the According to in Figure I.2a. film a such is located internal field *C-dok % The H. of strength field external parallel applied a x=0 and x=d as Londons' equations with spacial a distribution given by: H (x) = This distribution thin film the bulk. The (.12) ( for The excluded field for respectively. be much reduced from can be seen to Gibbs free energy per (average) I.4a Figures I.3a and is plotted in thick and thin films, the H that in unit volume G, (T,H) of a superconducting film in a field is given by: H(TT H =0 H = S I (1.13) the zero field free where G (T,H=0) is By energy density. performing the integration we find: GS(r, H) The &(Tr) G H= ) Ho (I. T 87r corresponding Gibbs free energy (1c14) IA density for the film in the normal state is: G,(T The ) zero-temperature GT,0 ) -1.15) ( 81r free energies are plotted in Figures 32 Y (a) x =d b. X Ha Z Y ( b) + ~X X =d n Z Figure 1.2 MEISSNER DIAMAGNETI C EFFECT Hcb Free Energy i k 2 8 7r T hick Film H Figure 1.3 Hcb =4r N(EF)A 0 MEISSNER EFFECT Hcb,' Free H Energy 2 Hcb 8-fr / / / / H Thin Film He > Hcb Figure 1.4 35 One notes that the thinner the film is, I.3b and I.4b*. superconducting state (d << A), state the normal crossing of by the field determined critical the the higher is the excluded field and is less and thin films For very free energies. the 6 H1 b/d which can be quite the critical field is large indeed. far So critical fields on the films. order of on In a magnetic field bulk of the the order not those of thin critical field a temperature a normal metal has susceptibility Pauli paramagnetic independent states. of elemental superconductors but the high parallel spin the consider superconducting and fields permissible for is to neglected the normal in paramagnetism This have we 37 4, given by: (1.16) X1P'KEF),1 where A49 is the Bohr other hand, have a ground electrons separated thus have a temperature *For clarity Superconductors, magnetont. state consisting by an energy gap vanishing approaches zero we actually plot of spin-paired from excited states and spin susceptibility Pauli as was GA tWe adopt the convention that IA (T,H) on the first pointed out + H 2 /(8r). is positive. as the by 36 Yosida 38 The . susceptibility of BCS superconductors vanishing spin can be understood quite simply by looking at what happens to superconducting limit.) 2uasiparticle states up-spin electrons and raised by down-spin lowered by split A H. as shown Ferrel 3" Tedrow, and Fulde1 7 aligned with Figure 1.5. and it possibility The was spin first pointed out by by Meservey, was first observed occurs in filled states below Fermi level can shift and lower their energy. remain locked in Cooper pairs shows no in because, as the are no empty down-spin states into up-spin quasiparticles increases. of . field is increases, there change the field. spin densities of states are The vanishing spin susceptibility which energy have their holes splitting in the density of states Fulde and paramagnetic and up-spin holes have their energies The resulting in (Such consisting of superpositions Similarly quasiparticle states of down-spin electrons the formed from a superposition of their spin is ,mH when that neglected. the Pauli be in said to superconductors are so enough be can diamagnetism Meissner thin is that film to a field is applied a parallel of states a the density the The particles and the superconducting state paramagnetic One can however anticipate energy as the field that there should be - a a. ........ ~- :1 :1 1% I- .' I I -- *00 37 38 some drastic behavior when the at their singular of states would cross the spin densities where field reaches H =,61/4 points. normal the corresponding and Chandrasekhar-Clogston4 0 or field, Pauli the plot with limiting paramagnetic The Pauli included. paramagnetism state free paramagnetic limit superconductor in the Pauli energy of a Gibbs of the resulting plot the gives 1.6 Figure is field critical determined by the crossing of the normal and superconducting condensation N(E, )A,/2, energy, ^eA /2, energy, equating by i.e. energies, free superconducting the paramagnetic the with and is: -4 - For a (1.17) "7 zero-temperature energy related by gap is transition temperature TC by limit, the weak coupling in the BCS superconductor Eq. 1.4 to the Lo = 1.76k 8 TC' and this gives a zero-temperature critical field of 18.4 kilogauss per degree Kelvin of the transition temperature. Sarma4 1 and later Maki and Tsuneto4 2 generalized the BCS gap equation to include spin paramagnetism and Figure 1.7 is a plot function of their of field spin-paired nature results at the for zero temperature. of the ground as order parameter state, the Due to a the applied field PAULI PARAMAGNETIC LIMIT Gn Free Energy N(EF 2 H Very Thin Film Figure 1.6 N(EF) A Xn 0 AO =2Bo 40 Order Parameter as a Function of Field Paramagnetically Limited, T = 0 1.0A(H) As U-1 0.5 z U 00 D.5 0.707 MAGNETIC FIELD Figure 1.7 1BH AS 1.0 41 A,/'A, 6 0 /4 8 , at Right H,= A./f2, the normal are equal energies field supercooling solution unphysical at and The superheating and supercooling points. also is there connecting equation the gap the is Similarly there /ASH,,= &,/2 of free state transition to order occurred. have state will normal and superconducting a first and reached, specifically at field is state saw As we field. superheating the is above however, before this P, with the a catastrophic decay into the normal state is This inevitable. are pairs single Cooper and aligning both spins unstable against breaking field and thus For fields above a superconducting solution of the there is no longer equation. gap pH <4. when gap equation BCS generalized from their obtained parameter the order effect on has no free energies a an the of unphysical solutions of the superheating, supercooling, and the gap equation are given along with the physical curves in 1.8. Figure At finite First change. temperature two the energy difference between energy states. Second, in superconductor features of gap decreases the the are the superconductor acquires magnetic susceptibility pairs thermally now normal in the excited to the magnetic field. quasiparticles and these can respond Thus ground picture the free as does superconducting and addition to there the above a and its free temperature-dependent energy is lowered by 0 0.5 0.707 Fiaure 1.8 1.0 P-BH 43 the application of the magnetic field. The features. normal zero reflecting the function of a independence temperature of the below the (for temperatures much susceptibility Pauli spin two temperature is raised above field remains unaffected as the absolute as free energy state plot these illustrates and temperature finite at field is a function of energy as a superconducting state free of the 1.9 Figure Fermi temperature.) As temperature the normal the eventually increases, and superconducting state free energies meet with the same slope are there and T = 0. 56 T supercooling and superheating longer the temperature At fields. no , the order of phase transition changes from first to second order. I.10 is these sketch of the field-temperature a films superconducting and H-T to the for Frota-Pessoa an phase example, was as spins to order is of second order into a first order line. diagram and Schwartz" applied field, there for the and as well as the first phase diagram, displays a line transition points changing FeCl12 diagram of order transition lines. and second identical Figure superheating the supercooling field lines are indicated This phase the 4 . a found in recently This is some metamagnets4 out pointed When a metamagnet is placed competition between in an antiferromagnetic a 3 , by in tendency manner and T >0 8=.88 %~ t=68 t=.044 t=.22 t0 00.5 1.0 Figure 1.9 HBH A0 45 Paramagnetic Limit Phase Diagram 1.0PBH Hs AO 0.707- Ist He 0.5 - enormal B CS 0.5 T 0 t=Tc Figure I, 1 0 2nd Hc 1.0 46 spins of alignment transitions which antiferromagnetic strong changes into a first order phase as the model with neighbor nearest next equally and coupling stronger* ferromagnetic or This order line Ising neighbor nearest phase. second of An lowered. is temperature paramagnetic line a in results competition a into the which aides magnetic field an applied effect of the and thus the coupling, reproduces the phase diagram of FeCl phase diagram in Figure 1.10. The order parameter in the metamagnet is the magnitude of between the the difference easily pictured and one It is imagine applying alternates This a staggered tendency to order field magnetic which Ht antiferromagnetic staggered field is conjugate to the enhances the in principle, can, at least the direction with in magnetizations. two sublattice spins. order parameter and antiferromagnetically. One then has the more complicated H-Hat-T phase diagram which is given in Figure in the Hst two "wings" The first order I.11. = 0 plane can be seen to be the of first order each terminate in a line The point line of transitions phase of second in the Hs* = 0 plane boundaries. The "wings" order phase transitions. where details of for reference 43 *See strong. how of precise definition meeting line of the second the theory order line and for a Metamagne t T-H - Hst Phase Diagram H 2nd Paramagnetic Tricritical Point \2nd Antiferromagnetic T Hst Figure 1.11 48 first order line is changes into a point named this order line) (and the first order lines also Critical phenomena 3 He- 4 He mixtures. particular have been the fruitful focus flexible system work in with an enormous value. As of a fair will be points in amount of years recent experimentally controllable field conjugate to the of occurs in many tricritical in general and and experimental Griffiths4 5 meet. a tricritical point and it metamagnets as well as in liquid theoretical where three second and accessible a and order parameter would be described in the next section, and developed at length in the succeeding chapters, the proximity field conjugate these effect affords a to the flexible way of superconducting order applying a parameter in Pauli-paramagnetically-limited superconductors display a tricritical point in their phase diagrams. which 49 D. THE EFFECT PROXIMITY coworkers 4 6 , effect was late until the Misener and 1950's that the D. superconductors. thin film on first done in Misener et al.'s experiments clearly identified. among the were not it was A. by mid-1930's the in observed fact was superconductors in proximity effect Although the They observed the disappearance of superconductivity in lead and films tin top deposited on of normal-metal the films were thinner than certain (~104 This effect the 1). thin film "critical" thicknesses characteristic of was thought to be to not and superconductors when wires be related to the substrate metal. Later by experiments others4 7 done with , to down of earlier experiments of thickness Misener et al. films that tin remained deposited on insulating substrates, showed superconducting thin about were then 50 A. The thought to be in error4 8 . in the late Only observed and 1950 was identified experiments 4 9 , H. carry a such. Meissner showed that wires when coated with thin would as the proximity supercurrent (<, 104 A) between In two effect finally a series of superconducting normal metal coatings them when they were 50 supercurrent only when they support the the than thinner demonstrated deposited that magnetic normal metals He observed placed in contact. on coatings. non-magnetic superconducting a when that above a it were superconducting unless corroborating the earlier experiments A few years a later, induced superconductivity proximity effect was given demonstrated the Moreover he coating is not coating would the metal, a normal were substantially certain thickness, of Misener et al. of the via the definitive fingerprint a in be metal normal by tunneling experimentsso which energy gap in existence of an the normal side of a superconductor-normal metal sandwich structure. An effects intuitive understanding concepts of superconductivity introduced basic concept utilized, and other the condense. state into which the above. that most characteristic of that of the superconductivity, is of effects and observed in proximity effect structures can be given using the The of these macroscopic wavefunction Cooper pairs of electrons Interference effects involving this wave function are responsible for the and characteristics of Josephson effects 1 this 7 . wave function also The existence explain the origin of the proximity effect. The starting point of all quantum mechanics, and indeed 51 of most is physics, every point in space equation at every point that it be satisfied at The requirement time. space and This equation. the electron wave functions is satisfied by in the Schrodinger boundary conditions that leads to the result in the appearance of energ y and momentum quantization in sandwich metal (on discontinuous* Furthermore, significant spacial is occur proximity effect is mechanics spacial of the over which the order parameter can the coherence length. The of thus the natural result a macroscopic wave variations A variations of order the on the distances mentioned above, as that is interface. the through extend must superconductivity means f unction wave macroscopic continuous Angstroms). few a of scale the phonons by induced interaction electron-electron effective the though even continuous be satisfying that the wave is interface at the the Schrodinger equation function prerequisite for a sandwich), (s/n superconductor/normal of a case the In systems. many on which can function scale the of of the quantum only show superconducting the coherence length. The leaking of superconductivity the *To consequent be sure the influence the of derivatives into of the the potential in the continuous when is discontinuous. a normal metal and normal wave metal function Schrodinger on the are not equation 52 result obvious is the enhancement or depression of in density the thicknesses on the order of geometrical resonances 5 4~5 from quasiparticle the coherence length, there are 3 and reflections normal metals been used to alloys 5 8 Additionally, behavior of these of promise 7 in superconducting proximity induce superconductivity in Kondo as the above, mentioned fields shows nature of been brought to of the elucidation the study of in high magnetic sandwiches aiding in structure5S-S Recently the density of states of n/s sandwiches. effect has pair proximity effect the induced in via structure resulting discontinuous the off studying phonon for is emerging as a tool states 5 4 bound Moreover the the interface. potential at with sandwiches clean In states. of the transition characteristic structures appear Additionally, temperature. Most phenomena. of a variety in superconductor tricritical points. The theoretical techniques which have bear on proximity effect sandwich of above features the is differ scrutiny. under chapter a review of the theoretical models chapter following models in we detail then begin in proximity-effect sandwiches reader interested in is referred to Ref. other 59 and order to In the next of 60 where the the In the of these understand thin one magnetic fields. in high aspects which is given. explore to to according proximity The effect proximity effect has 53 been reviewed through 1976. 54 II: CHAPTER A. INTRODUCTION I Chapter films The were discussed, including density properties these characterizes proximity-effect well quasiparticles. This of chapter become clear, is which theory a theory to superconducting of the the the of aspects generalize such reviews be other and a use zero-temperature and to of proximity-effect these to will we thesis tunneling theory developed by context a superconducting thin proximity various the aim of introducing effect theories with and putting in McMillan 6 1 , which, theory for the most appropriate purposes. Actually, as of this the spin-paramagnetism account for as will need will We sandwiches. of spin-splitting this on zero-temperature parallel magnetic the Zeeman spin-splitting In states. of concentrating our properties magnetic behavior of discussion of the sandwiches in high thin proximity-effect fields. superconductivity of concepts the introduced be needed for our that will the EFFECT PROXIMITY MODELS FOR THE THEORETICAL effects mentioned in Chapter I, which fall under the there category are of a number proximity 55 appropriate to use depends on the Gor'kov equations, principle rigorous, but the solution of theory is been used near with Ginzburg-Landau type theories. only been calculate used to Even in this the sandwich A second proximity-effect sandwiches. Cooper and discussed in Section C in thin films. Gennes equations the Bogoliubov-de layer, is reasonable quite potential. to The are observed complicated than that but does imposed in most (Section for when the encompass due to a simple of the Gennes-equation theories. for D) discussing the has those sandwich interface effective electron-electron attraction, is based on the solution (Section E) McMillan theory phenomenological origin, which proximity effect self-consistency for achieve of properties a the superconductors, but it is geometrical resonance effects in difficult transition of this chapter, gives an theory, Still another regime, it theory, developed by intuitive and reasonable description of the double only to make a connection detailed the not and temperature, of has the sandwich transition temperature, where it is possible to linearize the theory and has of the the equations date this theory To quite formidable. in are B), Section first (in discussed be will which on which effect one is interested The theories based in isolating. is effect proximity the of theory Which effects. BCS pair rather a effects is more discontinuity in the a restriction which Cooper-limit and Bogoliubov-de It is this tunneling theory which 56 will emerge as the low sandwiches. the most suitable for temperature properties simply characterizing of proximity-effect 57 BASED ON THEORIES B. THE GOR'KOV AND THE GINZBURG-LANDAU EQUATIONS As is particularly and in physics the case always in solid state physics, one must begin by developing a model which is to be simple enough still the contains of proceeds One discussing. basis the effects then where necessary in such a way so the which interpretable and solvable and by is one interested in approximations making as to treat the essence of problem correctly. that introduced interaction it (and the Coulomb overcomes) are electron-electron presumed their to be momentum surrounding the vectors Fermi and repulsion generalization of Schrieffer only within surface. when a of the . an effective certain BCS narrow case, is have both electrons In the All electrons interaction This V. 9 electron-phonon into lumped interaction attractive a retarded the of a slight Cooper, by Bardeen, complications which which is with start effect the proximity Hamiltonian simplified the of theories All region V was 58 presumed to have no spatial dependence*. proximity-effect sandwich it is In the presumed case of a that the potential has a position dependence reflecting the attractive interactions bilayer. Thus the in the Hamiltonian two metals 1'J for a BCS different which form the proximity-effect sandwich takes the following form: where U(x) is the one-electron potential and V(x) is given by VS -dGS < X <0 (II.2) VV, where the s-n interface illustrated in Figure of the phonons is I.2b. local and interaction changes Realistically the phonon atomic layers. is chosen to It is assumed be at at interaction as that the effect that the nature of abruptly x = 0 the changes the phonon interface. over a few This is a small distance on the scale of the momentum space as being a *Since V is actually defined in nonzero constant only near the Fermi surface, its Fourier transform into real space technically results in some oscillatory behavior. (See Ref. 62 for a precise definition of the spatial form of the BCS potential.) Here we will ignore this fine point and speak as if the potential in real space is a constant. 59 step-function change superconducting coherence length so the indicated spatial above is in variation of effective the choice for reasonable fact a the electron-electron interaction. The subscripts n and s above refer to a normal metal and a superconductor, respectively. The n-metal is not a normal a "weaker" superconductor than the but rather metal per se, s-metal. Thus the itself, superconductor n-metal may be a or it may not be. most direct The properties Eq. II.1 proceeding to and write Green's (x' ;t')> F4 (Hxx';v-')e Gor'kov 3 follow to is regular = indicate thermal the next chapter. function GI the averages ) anomalous (X',')>. and T.is the which is Green's The for = (,x';z-*') function brackets here Matsubara imaginary function is related local, self-consistent pair potential basic equation of the of motion defined more precisely in The anomalous Green's F itself is an integral function of the the finite to (generalized equations <Trtt(x, 3 3 down the and time ordering operator 6 is calculate of a system described by the Hamiltonian given in temperature) to the of way A (x) theory, is A&(x) by so Eq. II.3, which in fact a nonlinear 60 integral equation. extremely difficult. addressed by situation which might be The simplest the only situation considered, and is formulation, this date been which has to to look nontrivial for transition solutions of this equation in the vicinity of the temperature. = AJ) K(xx') (II.4) K metals of the properties depends only on clean In be linearized to AX X state. normal the cX) the kernel where 11.3 can In this case Eq. is 11.3 Eq. of solution self-consistent fully A simply is the 6 electron-hole pair propagator 4 given by c' (W . .., .. &. IV 111.16 in is a Matsubara frequency and is defined by Eq. chapter), the next because rigorously treat clean limit is though the 6 limit this in s difficult to the spatial dependence of the kernel is difficult to handle. a calculation Even proximity-effect transition temperature of the sandwich is difficult in many different approximations have of these successful a diffusion Guyon6 6 . is the Recently Silvert'" has calculation been used by de a and The most been employed. dirty-limit approximation has this model of in which Gennes and numerically solved 11.3 in 61 various to compare the have some standard with which an attempt to of none that finds He approximations. the approximations seem to be reliable. of results one of the type We have given a brief account gets from trying to solve directly the Gor'kov equations for a For our proximity-effect sandwich. there sandwiches, is to little transition temperature of the film yields. this from gained just the which is all this rigor of a direct is clear that the It be the proximity thin of more detail than are interested in We approach. properties the of characterization purposes here, theory attack on the full Gor'kov equations will have to be sacrificed before of the detailed properties any account of proximity-effect sandwiches can be given. It should be mentioned that Ginzburg-Landau theories of course, in also the temperature 6 7 . In regime in more Gor'kov approach boundary conditions which the Ginzburg-Landau wave function must are interested transition the near fact, results from the are useful in determining the we proximity effect useful for describing the temperature are, than satisfy. Again however, this limited temperature regime. Actually even the calculation of the transition 62 for thin film sandwiches temperature is This indicates. discussion above the is more difficult than because the superconducting properties of a number of metals change when thin films. they are in the form of much increased transition temperature is observed to have a other 0 K vs. 1.17 degradation a displays hand, (2.5 film is a thin when it Aluminum, for example, 0 transition its of temperature when it is made into a thin film, on the Tin, K). and this could be caused by a number of effects 68 . In section of the next opposite extreme here. We shall offered by Cooper lead treatment of this of this the indicated of a very thin proximity-effect see that sandwich. we describe complicated calculation of the We examine the limit understanding this chapter, to a limit. simple considerations qualitative and The limit, however, first quantitative simplicity of Cooper's precludes anything but rather unexciting properties for proximity-effect sandwiches described by this theory. 63 C. THE COOPER LIMIT In superconductivity bilayer 6 9 . in pictured the He that rate length of to the both superconductor, states and the = 0) having the (V metal of a normal simple case metals [ where s-slabs d and dS are the respectively. This the energy gap as well as sandwich using the simple of density argued that the is given by IN(EF )V 4J~ -s N(1E0)V, ~EF ( the to a in proximity same effective mass, Cooper effective BCS coupling parameter for Thus same the proportion in side. in each time spent of average two sides the electron-electron interaction of at such a two metals an only feel they structure to such a in electrons be scuttling back and forth between the rapid metal superconductor-normal thin a theory of microscopic Cooper presented a simple 1961 (11.6) the thicknesses of formula can be n- used to and the calculate the transition temperature of the BCS expressions, Eqns. 1.2 and 1.3. Others 7 0 have generalized this obvious manner nonvanishing Fermi-level to apply BCS for two interaction averaging procedure in an metals and both of which which (normal-state) densities of a differing have states, have N (EF ), i 64 The result is: n or s. [N (E"V",,1 F NhEF) V + N JA averaging with the simple however, some problems There are, (EFM s procedure indicated here. Derivations of this result, as well as generalizations to temperatures Debye differing Gor'kov formalism described possible to use fact, In derive Cooper's next section to in the above. Bogoliubov-de Gennes the given have been for the using the cases (in both , only) temperature transition 71 states and densities of differing Fermi-level metals with it is formalism outlined Eq. 11.6*. result, We discuss this more in the next section of this chapter in Appendix that the this A of are We thesis. simple averaging procedure to able properties all the when there is a difference in two metals. The reason when originally of Eq. indicated in these thin and demonstrate for the effective potential in proximity sandwiches characterize also 11.7 does not sandwiches the Fermi wave vectors of the for this, discussing his hypothesized as Cooper formula, Eq. 11.6, is that for the have recently done so *Entin-Wolman and Bar-Sagi 7 ' metals to 11.6 Eq. and generalized transition temperature with finite mean free satisfy all do not however, suspect. are results their Appendix A. Their paths. the See Green's functions, so boundary conditions and section next the 65 differing Fermi wave will vectors in result partial transmission and partial reflection of electrons incident on the The interface. electrons on attraction of each side will be describe different all feel and properties a thin oxide transmission the entire also and reflection the extent feel sample. cannot (Other at the interface, such would and formula simple Cooper's barrier, extent to a greater complications which might be present as lesser the the properties on each side Thus of a to the other side and that of "their own" side. means that transmission reduced result affect in partial the averaging in similar deleterious ways.) The Cooper of limit is useful for but we proximity-effect theories, such a check may be misguided. not sound and the result, Eq. appropriate, is too checking certain results The 11.6, trivial, for have also basis of the theory is in the limit where it is the Cooper limit much more use to us in this thesis. indicated to be of 66 D. GENNES BOGOLIUBOV-DE PROXIMITY THE TO EQUATION APPROACH EFFECT 1. Introduction one based on This superconductor. nonhomogeneous a for equations (BdG) Gennes 34 Bogoliubov-de the solving above, is approach outlined the Gor'kov closely related to quite actually proximity effect, the to An approach approach has been fruitful for elucidating the nature of the geometrical resonances full self-consistency in the potential with a to this solve simple manner familiar Equivalently, a functions allows functions for Green's these from simple It is only matching function This pair allows separate a wave the finite-layer Green's metals and then so as double layer 57 to use to electrostatics, functions in mechanics. of dependence spatial Green's function of the complete techniques quantum elementary Green's trial a equations for a nonhomogeneous familiar from separate layer for dependence. one to construct the each of the theorem, equation spatial explicit solution of the BdG superconductor using wave achieve gap function, one would have to of the Gor'kov equations. solve the equivalent feasible however To structures. in proximity to get 17 2 . match the 67 simple closed form expressions for the If one is to have of the pair potential Green's functions, the initial choice must a one of remain the achievement practice prohibits is the potential, which the pair approach. Nevertheless successful in elucidating interference Furthermore sandwiches. main deficiency of the reflections there are simple quite geometrical destructive and constructive of this been has nature quasiparticle of in self-consistency for of approach the by caused resonances the This functional form. simple clean in geometries, as the thin-film Cooper-limit geometry, where such it is possible to achieve self-consistency. 2. Review of this The use Gennes and Saint-James 5 4 states in a normal 1963 when de total density of a superconductor by back to approach dates calculated the metal backed by solving the wave matching problem: (E7~ E~VQ.(0E)VE) EFt. + AxVX) A4~)3 (11.8) 68 electron-like and v(x) are respectively the where u(x) and the hole-like amplitudes of the quasiparticle wave function, A where is (x) A, taken as 0<x<d,, and where the wave at the x= functions are required to vanish states of They found d, surface. for as zero -ds<x<O and for energy less than A, localized or "bound" in the n-side of the barrier as a consequence at the interface. not clearly observed until about These bound states were years later 7 ten 3 resonant state had by a the McMillan unbacked superconductors 53 energy gap when been had tunneling into a 51 'sz into superconductors resonance, Another superconductorsz. involving two reflections this one the energy been observed by Tomasch above weaker of another type long after much enhanced, by tunneling and then, backed This was . by tunneling into gap, first above pair potential barrier the of reflections off off interface, n-s the observed by Rowell normal metal backed and by a superconductor. McMillan other in the and Anderson7 4 oscillations theory of the for the Later McMillan 7 Tomasch effect 5 these a discontinuity based on scattering off pair potential. complete theory developed developed using an similar to that of de Gennes and Saint-James. a more approach 69 density of states a calculated the local as the thickness of the This barrier. result from one description of distance d, taken well the Tomasch reflection off the n-s not capable of providing a but the calculation is interface and superconducting slab, away from the calculation characterized oscillations which metals semi-infinite two considered McMillan first of (the reflections the higher-order which McMillan had discussed in the earlier calculation with Anderson 7 "). Recently solved the Arnold 57 Bogoliubov-de geometry assuming a and the s-side Bar-Sagi 7 2 and Entin-Wohlman and equations Gennes both the n- constant pair potential in of a sandwich structure. finite a in have Entin-Wohlman and Bar-Sagi discuss the limit where both n and s are thin which is a case in which full self-consistency of the be constant achieved with metal*. backed geometry claims one a pair potentials discusses the case Arnold by gap semi-infinite used in some recent can achieve BCS pair potentials solution can of a thin superconductor, experiments 7 6 . self-consistency with in each side of the metal. for each normal layer which He is a likewise constant gap This claim as the in the calculation to be some errors *There appear conditions boundary the does not satisfy Green's function that it is stated to satisfy. 70 on based is superconductor over s on calculating interferences over to coherence the length, parameter need to be been done on potential looks is interested variations taken into account. forms), to treat has more (though only of parameter though calculations for the normal-superconductor 7 Bar-Sagi 7 possible it is order the of the of the order spatial realistic variations of the order prescribed functional transition sandwich of the resonances, which are lengths demonstrated that recently when neglected be However, if one from the interface. in studying the precise nature due Spatial variations on what the pair temperature which depends like far semi-infinite the as such quantities assumed indeed can scale length this is perturbed, layer. thickness of the superconducting the potential the to compared small is which the pair into distance the that belief a have not sandwiches described here. As above, Entin-Wohlman eluded to used a theory based from solutions on of normal-superconductor potential in s). expression, Eq. films which are difference in construction of the sandwich (with a in all electron-electron a for constant pair the validity of Cooper's for the transition temperature of identical have Green's function equations BdG the They demonstrate 11.6, and Bar-Sagi 7 z thin respects except for interactions in the a two 71 metals. function at with an author shows Appendix A this In potential. average pair the density of states, a BCS behavior shows zero-temperature similarly the Green's that Cooper limit Thus in the by the example, is characterized for same average interaction as is the transition temperature. result is rather uninteresting. This average interaction It despite the that is probed in interaction is layer in really what is is in Eq. Hamiltonian indeed be that from interaction surface may layer this the is of a tunneling experiment7 8 If bulk. measured, It different a have unimportant. 11.8 This is the local density of states fact that it is a surface layer and tunneling experiments. observed in are routinely states densities of bulk total characteristics reflecting tunneling good why explain however, may, an BCS average then this different appears that incapable of giving any the more complicated structure than that which can be described by an average BCS interaction thin in proximity-effect sandwiches. In Appendix BdG derivation A we sketch a further of the Cooper limit generalization of the for the case also differing in their Fermi wave vectors. this case one does obtain density of states than of metals We show that in more complicated structure in the that due to one average BCS pair 72 physical mechanism The potential. original in his fact indicated by Cooper occurrence was in this for responsible 6 discussion of the "Cooper limit" 9 vector leads to reflections and reduced transmission of the that the retain density two metals will to The characteristics. their own in states of the naive pictured in straight transmission each metal complicated than that which would be because other coupled. considered this of added in deriving Eq. the Cooper limit reflection to metals differing in and its be then more BCS the structure properties of still intimately this chapter, the not was mechanism 11.7, the naive generalization of Fermi-level densities of states. this effect are metals above in Section C of As noted occurrence the two the to an average present in each metal will be reflected in the the extent structure in would expect however, that One interaction. the Cooper limit. a greater should due all instead of the interface, electrons encountering This means A differing Fermi wave . their normal-state In fact, the occurrence of significance have not been generally appreciated. Structure similar to structure this the same normal properties but barrier separating the two metals. by a been noted by Bar-Sagi discontinuous Fermi wave vector, has and Entin-Wohlman7 9 for a proximity caused sandwich of metals with with a regular potential Physically this is not a 73 surprising result since would also lead similar to vector. to a potential Experimentally the interface cannot and reduced reflections that attributed between the to a transmission difference in characteristic be very well two metals Fermi wave parameters of the indeed, controlled and, are even difficult to characterize. Nevertheless it is certainly such a potential plausible that but the most perfect metal-metal barrier is present at all interfaces. The effects of differing Fermi wave vectors in the metals comprising effects of calculated a potential barrier using The calculations model the BdG theory are very long which captured the between the metals, outlined in of the essence for McMillan 61 proximity-effect about the next two chapters for quite useful. Appendix A. and Such a given by is introduced chapter and developed in discussing sandwiches in high parallel magnetic fields. *See was This model briefly in the next section of this a simpler reflection sandwiches ten years ago. can be this section. and tedious* and reduced transmission effects would be model as well as the a thin proximity-effect sandwich, proximity-effect 74 OF MODEL TUNNELING MCMILLAN E. PROXIMITY-EFFECT THIN SANDWICHES out that any non-BCS-like vectors effect). between the of barrier at the 6 approximately by McMillan 1, which The of such a potential barrier. for simple model a solved and introduced incorporates the effects and its results are model A detailed discussion is only sketched here. similar a in result outline first sandwiches, Fermi wave to differing which section we In this proximity-effect metals, two the (or two metals reduced and the reflections sandwich with a potential transmission of a interface of thin proximity-effect features be traced to could structures was pointed this chapter it the previous section of In given in the next chapter where the model is generalized to apply to thin proximity-effect bilayers Pauli in high paramagnetic effects McMillan that noted proximity-effect sandwich by a potential of This if consists was one metal a more 8 Falicov, and Phillipsi the important. of two to another familiar that imagines one barrier, the situation is tunneling from barrier. are parallel fields where the metals a separated identical to that through an problem and oxide Cohen, had introduced a Hamiltonian, called tunneling Hamiltonian, which had proven to be quite 75 useful in the calculation junctions. also McMillan approximately tunneling which can In the and of wave each i n . higher order in coupling of the metals in tunneling problem, with considered by electrode or s) is Expressed in annihilation first-order vector k and Cohen, Falikov, (signified described terms of operators a here by its the own BCS creation operators for electrons z spin component by in states o=+ localized in side these Hamiltonians are given by: Ej ~~ where E; - ~ZVaO'O ~ ~ Q1 ~&L 0 is the energy relative to the the attractive HT the of the higher orders adequately tunneling problem Hamiltonian H i, a if theory. superscript a sandwiches Better treatment treated Phillips1 8 , and would compared to the regular be perturbation this Hamiltonian treated to to the more intimate problem as such describe proximity perturbation theory. this characteristics of proposed that Hamiltonian were is needed due of the of the single-particle Fermi energy and V BCS interaction. which couples the two sides The gives state 1k x> the magnitude of tunneling Hamiltonian is defined to be* *The meaning of the matrix elements appearing equation is discussed in the next chapter. in this 76 ..... For St VA known in result the that oxide to proportional is current the well obtained Phillips and Cohen, Falicov, Hr. real a of calculate the current to first barrier, it is sufficient to order (11. 10) characteristics of the calculation the ai t a convolution of the densities of states of the two metals. In the sandwich, other. It is therefore reflecting metal. functions for the conditions which potentials in n be thin enough sides, two be must satisfied The n- and s. that the of this the Green's self-consistency by and s-films in BCS the pair are presumed to spatial dependence solution the other processes into there are is the other metal, potentials within each metal can be neglected. nature *The chapter. order the sum on depends equations coupling addition to the In metal Each function of virtual tunneling the first beyond go function which states of the Green's over all to solution*. perturbation a Green's described by necessary the McMillan postulated a "self-consistent perturbation theory. second-order" has on effects each in the but rather just two metals comprising current between the interested in the not is one proximity effect, the treating of these This discussed in BCS is true the next 77 for films in which the total thickness is much less than the coherence lengths. Because next chapter, 8= .8Aand the a the n-side has one of is a gap side density of states because of potential of 0.11AS(in units A,= bulk a of gap energy There results densities of the pair self-consistent* has superconductor). in the sides of a proximity sandwich. of both zero-temperature the of sketch only I1.1 is a plot Figure of electronic states s-side give a we here McMillan obtained. The in great depth this theory is discussed induced in s-metal the normal in the normal side electrons can tunnel back and forth to the s-side where they enjoy the interaction. Note of states is s-side. the This is also the electron-electron attractive effectively s-side's that the energy gap as it is same for the n-side a reflection of the fact coupling between the two metals, in the density the that, the value of the potential, the states have s-side's because of no states are longer localized on either side of the barrier. below for the At energies self-consistent pair a higher density on the n-side of specified various quantities of the precise meaning *The reader the now For chapter. next the in here is clarified "standard" some mean to this interpret just should The metal. a normal coupled to superconducting junction calculating the McMillan parameters used in values of the 0.5 and T1/As= 0.2. are: 1/As curves in Figure II.1 Nn(E) Nn(EF) Ns(E) Ns(EF) -2 2 0 -4 ENERGY Figure 11.1 eV) \As 79 s-side's the of energy the Above sandwich. the self-consistent pair potential the states on the s-side have of states on the n-side appreciable density and the density rapidly diminishes to its 1 transmission due on the barrier to lead Physically, These tunneling to the mention should be made of to introduce Zuckermann 8 0 back and forth freely moving electrons must rates for decay to the McMillan model, some extensions of the model. attractive simplicity, and study and Machida 8 added phonon i have Because been possible it has various complications. impurities in the n-metal and have which can other metal. In closing this introduction of the model's scuttling the characteristic below energies limit. Cooper the which contribute to be freely between the two metals. have of modifications really be considered his the potential presence of to the only electrons the within anticipated above, that restrictions He finds, as . the bulk normal state. Cooper limit the discussed McMillan also model value in considered Kaiser and magnetic Chaikin, Arnold, and HansmaS6 self-energy effects to the model. As mentioned above, in the next chapter we generalize the model to include the quasiparticle spin-paramagnetism. 80 CHAPTER INTRODUCTION constant except for a proven quite useful approach has in of density the approximations states the scattering off for a to leads in reasonable of but rather interesting features also makes Bogoliubov-de and trivial more realistic wave in the the Gennes at vector regular potential barrier at the It namely and from the simplest only that the is in the however, uninteresting more slightly such as in the Hamiltonian, Fermi the Employing the superconductivity, for structure interference resonant to the two metals in the bilayer responsible complicated details change studying for a thin-film bilayer, Inclusion result. due This step discontinuity. discontinuity. difference between the interaction the solving Gennes equations for an assumed pair-potential Bogoliubov-de which is on is based approach One effect. proximity to understand been used which have theoretical approaches various the reviewed we chapter preceding the In the PARAMAGNETICALLY LIMITED FORMALISM PROXIMITY-EFFECT SANDWICHES: A. OF TUNNELING MODEL III: an abrupt interface or a interface, results in more properties of theoretical formalism quite these sandwiches. calculations formidable. the in It is 81 there apparent that presence of Such a the two metals. potential barrier between a regular in the a difference or from the normal properties of the two metals which simpler model a effects resulting from the reproduces for a need is for studying all model is preferable in the thin-film regime but the most detailed properties. It which was detailed discussion generalizing it paramagnetism of applied chapter In this to of this the results the model while include of this model. The next of calculations of some next, we of at the of give we a same time the quasiparticles in chapter In this simple A final section effects the the superconducting magnetic field. basic formalism in the and a model introduced features. these was given the model II. McMillan'i exactly incorporates overview of Chapter out that pointed develop spin an the chapter presents interesting properties of paramagnetically limited proximity-effect sandwiches THE B. 82 HAMILTONIAN in the As sketched effects complicated the occurring at interface with a tunneling Hamiltonian, H. so that the sandwich's the , given by: 111.1 -. S H, =( the sandwich, HTOT' total Hamiltonian of the into account all takes is that it McMillan tunneling model essence of chapter, the last can be written as the sum of three separate components: The matrix from a element T- Ik'T> state n-side. HsV H = HTO H gives (111.2) < the amplitude to a on the s-side dictates Time reversal symmetry amplitude to the amplitude for this 1-kl> on the n-side to the definition of HT superconductors n sufficiently state incorporates and s will dirty that there state for tunneling the equivalence of tunneling from the state tunneling this eventually be presumed carry we will place no such restriction the analysis as far to be is enough momentum scattering the other side. wave vector dependence of these The equivalence. into all states on moment, however, The |-k'l> on the s-side. so that all states on each side have equal amplitudes for on the fkt> T/ sT For the on the matrix elements and we will as possible without imposing this 83 restriction. The potential barrier interface. The transmission of the electrons Above number of physical mechanism. Fermi wave metal-metal potentials at of interfaces well defined very not bilayers are the speaking, Generally in the mechanism Another presence of impurity scattering interface. a out that one difference a metals. the two vectors of might be the the we pointed would do this is other mechanism which from arise interface, could on the are incident rise to, barrier gives effects which this a the at encounter electrons which namely reflection and reduced which directly represents most Hamiltonian tunneling or well characterized. We shall in the calculations of this rates f matrix element appears see that this tunneling =fV(2) theory in the form of scattering for scattering from side i to other side. These are given by - ( ZSN .1-T where A is the which have 12A 2't'4 h thicknesses N (III.3) 5(F) (III-4) the interface between the two metals area of densities of states s p d, (E ) and normal-state These scattering Fermi-level rates can be 84 related to the barrier transmission coefficient C0 by 4 B; A; where vF. is the Fermi velocity of metal i and B. is a dimensionless function of the ratio of the mean free path in metal i to If we the thickness of metal i and is take the parameters r ratio of we to of order unity. find that these two are related by: -1-.NE ar. (111.6) Ns,:) leaving only one free parameter in the theory. have the dimensions of energy energy units in of the bulk, and we will Each of scale of as constant. the coherence length that its BCS 11.3, can be (See magnetic properties Chapter I.) Thus Figure 1.4 taken as be taken applied When field. as spatially the scale discussion of superconductors given field constant magnetic also of significant Meissner and the magnetic superconductor can on the field effects, we is no of thin-film the 4S pair potential two metals are thin enough on penetration depth that there diamagnetism. gap to be thin enough For treating the magnetic assume that the the Eq. energy (the s-metal). the metals is assumed defined by usually measure zero-temperature of the stronger superconductor F; The rates flux within and equal penetrates the in each to the the 85 superconductor it the field on the magnetic moment the Bohr is where from the three quasiparticle is of the the order of s-side BCS Hamiltonians The n- generalized from Eq. II.9 +MIS and are: A H-)O (111.7) ii \/ in this equation The symbols formed vector a fields because it can be on pairing energy. H(e is z direction as illustrate in Figure field alignment energy The of the electron, = -A, (r effects of We take the magnetic field Pauli matrices. important in high the 1 magneton and H to be directed in the I.2b. consider the is necessary to have been defined in Chapter II. In all but the lowest atomic number include the necessary to which moderate acts to magnetic field on include spin-orbit since at to allow effect. effects of the superconductors it is spin-orbit scattering of the We do not paramagnetic effects the superconducting state. scattering in the theory developed here the present time experiments are not refined enough extraction of the In Chapter VI, any information however, we about include this some consideration of the magnitude of the spin-orbit interaction in thin dominant. metals where Spin-orbit surface spin-orbit effects have been scattering is observed in 86 paramagnetically should be limited included for a thin film complete superconductors description proximity-effect sandwiches in parallel fields. of and thin 87 C. GREEN'S FUNCTIONS AND THEIR EQUATIONS OF MOTION We use matrix matrix space. identity and i = 0 to 3, Green's functions This space is spanned the Pauli with the matrices written as, 3 The matrix Green's and space , or i , two-by-two = 0 to 3. 0 / - 6 = -; 0 0 functions will down-spin electron anomalous by the products of the Green's off-diagonal thus in These products include both the up- functions Green's important in the presence of penetrates the superconductor. the The Green's function is defined given by: kTr C r) as well as functions The explicit inclusion (t) , 0~ 0 o o 0 0C0 nonzero elements. operators spin space for example, , 1 a four - by-four in identity and the Pauli matrices the two-by-two Nambus 2 are in G among of the magnetic the its spins field is which in terms of vector field 88 where T Matsubara, 6 2 ' 6 3 i.e., formalism of finite temperature in the imaginary time is the evolution of the "time" the field operators is described by The -Vr-r I - ;tVro 7 1 ( II. 8) finite temperature Green's function is then defined as: )r imaginary time (Operators with a more positive ordered further to the left with denote brackets proximity For states. calculating the Hamiltonian or coupled by Josephson 3 Eq. the energy the so Eq. current tunneling flowing the barrier to useful are with for tunneling the between Hamiltonian, not (q,q';t:), and G of is defines new 111.9 functions associated occupied it sandwiches, from one side Green's Such other. i', this ordering.) thermally over G S(R,k';T ) Green's functions, which describe propagation the effect i be equal to necessary that types of an average argument are a sign alteration for each interchange of operators requi-red to produce The (111.9) operator. ordering time imaginary the is TI where c metals the such as the current at finite temperatures. 111.8 implies that the imaginary time derivatives of 89 the field operators are given by: T2 7(III.10) 1U We use this to differentiate the Green's functions to obtain For the s side Green's function, their equations of motion. we find 3 A8s + (III.11t) 'I" where the matrix T{.. is defined as -0 O %i:~o 0 -T- .. 0<m 0 0 a 0-T. of the the standard BCS-Gor'kov factorization We have used two particle Green's function resulting from the interaction term in the Hamiltonian using the definition LZ7 t -0, 'te) As should be obvious but is worth emphasizing, in 111.13 is of sandwich evolving full the field according to H of the H t. self-consistent pair Thus the also depends on the properties of coupling between the two sides. the averaging operators for and not the uncoupled metals, the operators (III.13) the average of which evolve potential on the n-side n-s with the s-side and on the 90 define We imaginary time dependence the of representation series the Fourier in the standard way by: t Got (3(III.14) where the sum is over all j and the frequencies (),' integers are given by (= |4/' (3=1/k with T (k is 16) 'TL(III. temperature) guaranteeing have the in Fermi statistics. the is and T will Green's functions that the periodicity 1/(PAi) constant Boltzman's ? appropriate which is for Using these we get v3( &a.~ 3 o3) .06 Li~r Z'r Z(A4.ivJ A ,-> t (III.17a) P The quantity in the square brackets in the first term on left hand side function of superconductor, GS (k,iw)]. s-side inverse of the an unperturbed, bulk, of this equation is just the superconducting Green's the -/ equation as SS, 55t.p.I)1 -a)17b) We can rewrite this 91 In G a similar manner one which finds describes propagation for the Green's function from the s-side into the n-side: /(III.18) WIr./ There are 111.18 G also analogous the for expressions Green's (k,k';76v) as well as to G functions to Eq. Eqns. 111.13 for 111.17 and (k,k';/W) and the pair potential on the n-side. A The equations A sAfs G ,and G for the and the pair potentials, functions, G 31 - A K60 G , two self-consistency equations for the A rhand tunneling theory of four Green's Ar, are the basic equations of the the proximity effect. In the next two sections of this chapter we discuss the analytic solution of these equations as this discuss the equations. in so direct In far iteration the McMillan 6 " in the case The next chapter the section, We Green's we first function discuss an equations first introduced by where there then gives method of solution. of following approximate solution of these is possible. numerical is no magnetic field. results of McMillan's 92 ITERATION OF D. we introduce section on a describes expression for form a closed propagation of the perspective The Green's function and H' respectively, are i (i=n with superscript subscript The referring referring to respectively, to or i=s, respectively) and each side, the of Finally propagation from are final and excitation metal denoted by '), wave functions which to the other, directed double and a )', vectors, by described are with again two subscripts, k the Green's one lines lines III.1.) (k,k';/ G H with a (See Table momentum. double and s, directed by single the give Hamiltonians denoted initial the BCS the by but this time with propagator. Gas k,k';;w-), metals n directed by superscript i, for the isolated functions for full Green's represented solution. described G (k;iW) and G (k;iW.) system. the of approximate to McMillan's to function which Green's the this section, however, The considerations presented in some are unable We normal modes the of diagrammatic equations, and re-sum the resulting infinite series. find In G. representation diagrammatic We iterate these these equations. function Green's the which itself depends on 11.18, described by Eq. this depends for Gss Equation 111.17 Gs FUNCTION EQUATIONS THE GREEN'S this describe such as lines, again 93 G0 (k Ito* J 1% S 7w It G (k,k';oI/) S.7t Table III.1: with two Green's function diagrammatic dictionary. momentum subscripts, superscripts referring to the initial and final electrons. Eq. 111.12 all Table this sides The of time also the with two sandwich of the The transfer matrix defined by is indicated explicitly summations. Green's but diagrams in the introduced functions are summarized in diagrams for the as are various the dictionary given in III.1 Multiplying Eq. 111.17 on the left by Go(k;-a)), we get 94 for the full s-side propagator: S S (III.19) 111.18 for the Similarly, Eq. the n-side can be rewritten: kk r propagator from the "f k (111.20) find Plugging this into the preceding equation, we S s-side to Se A %I (1II.21) which is an s-side of the integral* equation for the Green's function on the bare n- the sandwich and depends only on and s-propagators. manipulating Eq. 111.21 By form involving a sum over over full s-side an equivalent full n-side propagators propagators. Setting p =k', instead of multiplying . * A , by Tt, we can obtain and summing over k', we get an integral equation for the summation over the s-side propagator in Eqns. and A 111.21 I_ A 111.20 5 *In our calculation of values. k k is allowed to range r over a continuum 95 Putting this equation back into itself, we find S ^ 4- P - A S el *'~i ~Js can One this continue .b,,Jr (111.23) scheme indefinitely. 111.21 for the s-side Green's iteration Plugging this result into Eq. function, we get AS S. 4Ac: ?1 ± VI.' 4. Te Tx ~'I do j /..& - - k e jA~' This -I -+Pt r A% A ~ may be re-summed to give S A S (111.25) hII which is the desired rendering of involves sums over all the Eq. 111.21. intermediate This equation n-side propagators 96 an physical interpretation for these equations in terms obvious of There is propagators. (explicitly) over s-side and none contributions from of crossings more and more the interface. is a complicated Equation 111.25 can little headway be made some approximations. and 111.24 one might bear strong in The infinite series in resemblances to drastic simplifying rather of the momentum dependence consider this will not tunneling means of illustrative purposes below) series possible matrix solution only placed on are and closed form obtain a assumptions 111.23 Eqns. geometric This is these series. expression for introducing without it solving is possible to imagine it equation, and integral elements. when the We than for (other or the exact solution of this equation. McMillan61 gave an approximate solution for the tunneling He did model. not proceed along the lines developed here, but we shall comment in the next section on the relationship proposed of his to Eqns. We 111.21 solution to and approximation contributions to the above series, in particular 111.25. give an expression for drastic the that series the Green's all function under the off-diagonal 111.24 can be (in momentum) neglected. This 97 will in the sandwich because the coupling between the two metals series for the terms in the we find that Eq. that such with phases By neglecting averages to zero. S in order the higher to sums contributing come contributions be expected that Furthermore, it might T. zeroeth and vanishes in the off-diagonal terms first orders in for weak approximation be a reasonable contributions might off-diagonal the Neglecting solution. McMillan the compared to are equations when these be useful their sum diagonal terms, these off 111.25 becomes ~ 9 7 5 When iterated it becomes (II. 111.25 can be seen to be and fourth that the wave vectors in the second propagators in the that this importance tunneling Eq. tends to approximation of the back and the higher order forth between deficiency. right-hand at the same time, summed independently in McMillan solution compensates for of this term on the third side of Eq. 111.27 are summed over these are that in Eq. series and between this The biggest difference .27) 111.25. wheread This means underestimate contributions the two of metals. the the The the most significant part 98 E. THE TO SOLUTION MCMILLAN MODEL FOR THE Expansion of the TUNNELING PROXIMITY-EFFECT to Comparison 1. Diagrammatic the Green's Function the Green's McMillan proposed an approximate solution for function of the to permits of by Ansatz was given the pair a coupled 61 . Dyson's equations propagators of down. form which in a section. and the n- wave vector arguments written "solution" diagrammatic expression given in McMillan assumed that his Anzatz solution, In This be rendered but can comparison with the preceding form of solutions Hamiltonian 111.2 in the s-sides done done when as was However, diagonal were McMillan replaced Eq. the full in their 111.26 was the final bare propagator in this equation by the full propagator giving T+ A0 S1 111.28S) Tez(III.29) < trk These can each indeed be seen to be in the form of a Dyson's 99 equation. (p,iO.) Defining a matrix self-energy - S I & we such that -J ) (j (111.30) see that* A (111. 31) k m.2, Eop A diagrammatic expansion of yields: + +) + 5 (terms of order T6 ) + 14 (T8 ) + 42 (III.32) + In comparing 10 ) CTi with Eqns. this 111.24 and 111.27, there are two things to note: (1) There are 111.32 than more higher many 111.27. (2) In some of these higher in the term last *McMillan's original equations T) T = T't whereas correct expression is T are however correct. order terms 1 TI given by in Eq. . 111.12 111.32, the explicitly, are such that (with from Eqns. = given Eq. terms in "diagonal" approximation in the Eq. notably order and III. Tg= 31, the McMillan's final results 100 what is similar to in the roughly approximate the s-propagator these diagrams, however, In is 111.24 and exact Eq. done in the different from what is done 111.27. This independently. summed are n-propagators intervening between two n-side sums is restricted to be bare the the the original The electron. not itself summed over as includes the propagator not does representing Specifically exact many a in It order) The 111.24. Although this it should be. most important contribution, it there versus n-1 intermediate those from contributions. of terms is a plethora sums. restricted are there T, n/2 only one for but in 111.27 also higher in summations in propagators the these which one would expect Thus this solution should not approximation. McMillan Ansatz is best justified between the two metals is weak, in which is clear that the coupling case it is not important to the higher order terms. identical is in in each when the propagator order n-th the (as bad intervening of restricted summations are be as combinations 111.32 the greatest wave vector the everything. in summations term include higher orders in T, Going to same s-propagator with to order Ansatz is in T 2 .) know the exact contributions of (Indeed Eqns. 111.24 The ultimate test the comparison of the results and 111.32 are for the McMillan derived from it 101 to experiment. sandwiches There are (specifically, classes thin, 9 Hamiltonian interface, . This, does are the McMillan Ansatz to the thin 2. Equations The bases for solution side of the for the These important a in that the of the physics generalization the magnetic field 111.28 this of the effects for work. and 111.29 can be expressed of two renormalization matrices ZI(iW ) for the ) for the pair on each mass renormalization potential renormalization are defined in terms of the self-energies by: $ a H' WZ where we now belief Renormalization Functions of Eqns. (iW and effect coupled sandwiches) the using describe the barrier, matrix. states with sandwiches considered compactly in terms matrix along contain weakly proximity gives a reasonably quantitative for which the McMillan model description of (1II.33) assume a constant tunneling contributing to the sum, amplitude for all namely T-.., rk renormalization matrices have the forms = T. The A( 0) 0 0 7/M. 7A.- = the superconductor, electron i A+ coupling the down-spin H. 0 0 As is equations Putting the find that the 34b) the same cases is BCS a up-spin the couple electron and the up-spin as those hole. just + ^H. The If renormalization define the of one scalar function eachO. self-energy given inverse Green's function given in we (III. for case the are hole this shift, we can matrices in terms 34a) 0 which paramagnetic energy shift for these we allow for (III. 0 the down-spin and 102 Id ) 0 0 tO-) where .2 C) 0 A 0 in Eq. Eq. 111.17 Green's function itself is renormalization functions 111.33 into the and inverting, related to the by the and up-spin the of uncoupling "This occur when does not renormalization equations scattering is included. down-spin spin-orbit 103 O0 ~: ~tz 0 O 6)-aj ( Zi" 0 0 zir/)_4__ evaluated explicitly [i'11- in Eq. diagram the a sum by performing states on this Green's function. We find, Ti-60 (7I7 over the is n-side for example*, (111.36) alp t Z. Lz rate 111.31 Z S. where we - (III. 35) given by self-energy 0 0 -iwfi' X->f- 0 (z decay 0 # z</ The ^'' - 5 have used i. Eq. v&'I ) III.3 Finally we can iwo (III.37) 1- 0 which defines use Eq. the tunneling 111.33 to relate the contour in the were defined by closing a *The square roots to have positive so they must be chosen upper half plane, imaginary parts. 104 renormalization functions s-side on the n-side to those (111.38) Y_ equations From the n-side Green's the (111.39) / ( corresponding to and 111.30 111.31 find identical function, we for equations with the sub- and superscripts n and s interchanged. Equations must 111.38 and 111.39 satisfied be necessary that and their n-side counterparts the BCS pair potentials equations satisfy self-consistently their 111.13. Using the 111.35, we VL functions in /~)' Z &/(kjk=) 0+. This type of ~~ k} jt handled in the after converting Manner in Ref. a contour .40) (111. 41) sum occurs frequently described the sum to (III - finite-temperature Green's function formalism Eq. become find that the self-consistency equations e? where in these used definition in Eq. renormalized Green's VZo4' ,L1 ',(H) h6, is it addition In simultaneously. and 84. in the it may be We find, integral, deforming 105 the contour, and then integrating, that ApH) =Vi N5(EF Re hhf where is : quti o.42) the solution ofE the implicit equation Z. -- (111.43) as and where we have defined the renormalized gap (111.44) In the case where : to find it convenient to make this In the next section we of variables. change form allows present The gap equation to the superconductor the a degradation At temperatures. curve field Pauli this equation factors in to in of paramagnetic the calculated the (for the Fermi The limit. curly bracket) at pair potential zero temperature BCS isolated an terms in the (the of who first drawn be to direct connection a of Sarma4 1 , versus parameter order lead from integration variable changing the equation by E~s. McMillan's self-consistency a form identical to be cast in 111.42, can Eq. H = 0, the gap equation, BCS finite case) factor in the bracket gives unity unless the field energy greater allowed than the pair potential. to be smaller that the the is When the pair potential is gap energy we get the 106 unphysical solution connecting the at = ,H proximity effect sandwiches, the physical curves influence at energy starts term in the It chapter. the next in see The causing the first the gap to exceed energy bracket to suddenly go from one to zero. Eq. 111.42, reflected in because renormalization for final equations in these acquired a field sandwich the term in brackets influences the field non-trivial way. be In these Figure 1.7. in the zero-temperature always begins when is only potentials, and at zero-temperature as well as at finite we shall as temperature Os /2 equation gap aH =,&, at point superheating supercooling point dependent field the of that the BCS potentials dependence and this field of functions the two of implicit the on BCS pair magnetic field enters in a The self-consistent the properties the the have dependence will sides of the of the dependence self-consistent pair analogs, and Eqns. potentials. Equations 111.38, 111.39, 111.42 are their n-side the fundamental equations of the tunneling model. Their solution, which must be the most trivial cases, done numerically in is discussed all but in the next chapter. 107 OF A. PROXIMITY-EFFECT SANDWICHES PARAMAGNETICALLY LIMITED INTRODUCTION: superconductor of solution of the of two pair field. a the Green's the sandwich can be potentials, equation, functions found we functions on the two sides are of this that for the Given the model Hamiltonian absence of original McMillan's of for functions expressed in Zi(E) renormalization functions, Green's the Dyson's BCS comprising each metal When the effect in proximity terms tunneling the limited paramagnetically lines the along a solution of the presented we Pauli a for model tunneling values CALCULATIONAL PROCEDURES preceding chapter, In the PROPERTIES FOR THE PREDICTIONS TUNNELING MODEL CHAPTER IV: form were and put (E). into renormalization the interrelated by the following equations: Y z - ( (IV.1) (IV. 2) 108 E) 1zt -= il(IV. Zn 3) 4t) (IV. 4) The BCS pair potentials in the two metals themselves must satisfy the self-consistency equations 4z(E) E 0 0 tE)Z (IV.5) and are thus functions of In both field and temperature. this chapter we present these equations and discuss the a results of number proximity effect sandwiches which can solutions. We calculate density of states of the three main potentials in self-consistent pair n and (3) of these be derived from these quantities: s; (2) the calculational procedures. this In the field-dependent of Eqns. IV.1 the magnetization, susceptibility, and In (1) be obtained potentials which satisfy Eq. sandwiches. of to IV.4 for given values of the which depend implicitly on solutions and properties sandwiches which can from solutions of Eqns. IV.1 pair of solutions to section we IV.5 and to IV.4; free energy outline Sections B and our C of this 109 a sandwich and normal-superconducting for a calculations zero-temperature of results the present we chapter two - superconductor The numerical method of self-consistently calculating the coupled) proximity-effect (weakly sandwich. Renormalization Functions The 1. using Eqns. renormalization functions Af and A The which are properties: be below ftL Eqns. IV.1 (5 and E\j). solutions to the energy in gap renormalization of the , as well can be well above as energies solved by At t(A E) Initial values of values are used in IV.4 to calculate The iterated are chosen to are chosen to be the right-hand the next _a straightforward a A'and initial values of Z1 (E) These unity. the the Above this energy they are complex. iteration procedure. be equal to BSC pair potentials associated with to IV.4 IV.1 Eqns. following which equations are real. energies have the A!(and assumed spectrum) excitation IV.4 to which is out to turns IV.1 to IV.4. At energies below a certain energy fl, of the a function IV.1 in solving Eqns. used Eqns. solutions to pair potentials initial values for the We assume follows: as IV.5 is to IV.1 sides of iterate values of values of these quantities again put into the right hand side of these equations are giving 110 To get solutions in iteration procedure does not converge. with the of E initial value out turns In practice it down tol.f. energy region where that the differential to be obtained from the which is large enough for solutions iteration procedure are differentiated resulting integrated from an E are in equations to IV.4 The energy. to respect IV.1 Eqns. regime, this energy this slightly above -01, For energies converges. until it The procedure is repeated A%)and ZjE). of estimate a better procedure does iteration not converge is very narrow and it is possible to accurately interpolate above far enough those solutions -f2t to have interpolation scheme when we calculate interpolate interpolation potentials We however, could, (We regime. in scheme use the of the pair as well.) that require self-consistent. It the values above is necessary functions, when put into Eqns assumed whole energy justifiably calculation the not do we over the solve the equations but the In calculating however, potentials, pair this simple the density of states and the magnetization as outlined below. self-consistent a converging We use iteration procedure. solution from the and just below between solutions at energies for the quadratically convergent pair solution fully renormalization that the IV.5, reproduce potentials be the initially Anh(H). We Newton-like method use a to numerically 111 solve 2. these nonlinear self-consistency equations. The Density of States well known that the analytic It is temperature Green's are such that the functions electronic states* N(E), of finite properties density of defined by 8 s N'CE) = 3 (where the by E- ) sum is over all states (k o> we function spin-up are and we electrons electrons respectively. HN(E) with from For example, matrix a four-by-four can separately NI(E) related to The usual relationship is function. dealing with energies denoted representation which is has a spectral that of the Green's Here civ 6) and G the density calculate the (k,kjiW) Green's density and of spin-down of G (k,kii) for N(E) we have each side of states for defining a "bulk" density *We are of states in equal to the local density which is actually on the films are so thin This is because the each side. superconductors lengths in characteristic of the scales vary not do states of densities local their that appreciably across the films. 112 7-- of terms in expressed functions Green's The (IV. 8) the renormalization functions given in Eq. 111.35 can be used in the right hand side of this equation. We Et 2(IV.9) N (E)VEt)Ree -- _(__ N'(EF) - <tE) .[ZtE)E± before, MT(E where, as find of one of states is the density ) spin orientation of side i at the Fermi energy when it is in the normal state. 3. The Magnetization, Susceptibility, and Free Energy magnetization The M on each side of the sandwich is defined by The thermal Green's average here may be functions manipulated to and expressed in terms resulting the be an integral expression over the density of the can be of states. The result is MC, VO E Thus once we have calculated [N(EHEI the + (IV.10b) density of states from the 113 Green's function we can easily calculate the magnetization. the magnetization with respect to From the derivative of the magnetic we field, get the spin paramagnetic susceptibility M (IV. 11) In performing account has pair the differentiation to be taken of potential for a superconductor, the implicit dependence on H as well as the explicit of the dependence of the magnetization on H. If Gibbs we work at constant temperature, the change in the free energy density as we vary the field is given by Ot H _ where B is -- (IV.12) - the magnetic induction of side i. The magnetic induction is related to the magnetization by 13 (IV.13) - Thus the magnetization is related to the free energy density by SH If one is o MH(IV.14) 'rr interested in the free between the normal and the energy density difference superconducting states, the first 114 term in B cancels and we have (IV. _____ the magnetization, it would Since we can easily calculate be integrate to able be to convenient At H integration. the sandwiches, discuss the corresponding results Pauli paramagnetically we can point of effect constant for example). susceptibility magnetization, and energy, proximity coupling conventional the integration' 2 , We first for these 0 = one this in itself would be a formidable calculation doing (when so is doing states at the starting superconducting normal and and difference between the know the free energy to first needs difficulty with The magnetization. IV.15 Eq. using our knowledge of calculate the free energy difference the 15) use knowledge of an limited superconductor. of the form of for the free We isolated see how the proximity-effect phase diagram gleaned from this simpler example to avoid the difficult initial the between calculation of normal and the free energy difference superconducting states of the sandwich. In isolated Chapter I Pauli we discussed paramagnetically the phase diagram limited of an superconductor. 115 (See Fig. 1.7.) we see field, From the that and supercooling at free energy as low temperatures of the end point supercooling curve. In Figure first and second derivatives These IV.1 we are as an that of the display again the these curves and of derivatives the give well curve to superheating zero-temperature free energies include the there of gap equation which connected unphysical solution of the BCS the as curves superheating a function of we these magnetization also curves. and the susceptibility, respectively. The susceptibility of the normal state is the well known constant Pauli susceptibility 84 and this results linear normal in the state magnetization and magnetization reflecting a vanishing susceptibility the fact that the up- and down-spin electrons in the Cooper are locked together and the paramagnetic pairs great enough to break state susceptibility the <IV.16) A superconductor at zero-temperature shown in Figure IV.1. has N'(E,) solid horizontal indicated those state lines. curves The these pairs. superconducting and magnetization are thus lines in Figure portions of the normal that can be vertical The energy is not IV.1. in the We have and superconducting physically realized by jump plotted as solid line the solid for the 116 Free Energy Gs 0 0.5 0.707 1.0 BH /PBM X Pauli A normal 0 0.5 0- / N / ) / / / / -N (b) \ / -"supercond. 0 0 I I 0.5 0.707 0 I 1.0 L BH A k x X Pauli 4- -Q normal I 0* - - 0 supercond. 0.5 0.707 Figure IV. 1 (c) 1.0 *p H B 117 magnetization and the susceptibility occurs at free energies the of the normal and superconducting states lines the extensions of these We indicate by dashed cross. beyond the solutions for the superconducting field H for the normal curve. of the unphysical where the energy gap a has the to zero. function of as the The is greatest and It supercooling point field for the an isolated energy and gap all the Pauli Fig. 1.7) of magnetization to zero as the energy spin aligning of the (See the plot superconductor given in point superheating 2 H net magnetization. value at the energy gap goes unphysical state goes large has a This solution negative slope. the normal state as a connects the less than the spin-alignment energy paramagnetically limited At solution which have broken pairs and a magnetization equal to curve to the supercooling supercooling points. energy gap which is so it does superheating dashed line the susceptibility and We also indicate by a superheating and to the field critical field H free energy is the field at which This critical field. the first order /,r2, AH this is then as electrons become spin paired. Many features of the the susceptibility paramagnetically limited free energy, the magnetization, and curves of an isolated superconductor will be seen Pauli to be 118 for paramagnetically the corresponding curves reflected in there In particular effect sandwiches. limited proximity will be a supercooling field at which the pair potentials of the solution goes Thus same energy. these two solutions have the starting point for use thus can We the magnetization Then, by the by density When we free field, the be equal to that of the integrating the difference energy of the unphysical state will superconducting solution between the integration, and superheating to the superconducting state. as the unphysical state. between the normal state and integrated out this point in free energy using IV.15 calculate the difference have unphysical state solution. into the normal continuously at the supercooling point free point the this At zero. to sandwich go and the normal state solution, we can calculate the difference in the free energy density between the state. This proceeding, but minimum of is normal state a albeit it gives the additional and the rather superconducting calculations of what we know beyond from these solutions and of with a free energy densities self-consistent solution for the Green's function. full use way roundabout the It makes what we 119 the qualitative shape of the free energy curve*. know about This calculational procedure yields the difference in the density between the normal free energy solutions on each side transitions are determined We shall sandwich. free energy density is equal to normal-state the by the total free for each side gi free energy of the normalized calculating a actually be the sandwich which of divided density energy phase The sandwich. the of and superconducting by the Fermi-level density of states, (IV.17) N (E,) The total free energy of the sandwich is then + Sd N d )S pK Thus proper weighting of the determined parameters ratio. the next by the that same can be (See Eq. III.6.) section it ratio the free energy of the determined from For CIV.18) densities is proximity coupling the specific heat the sandwiches we consider in turns out that the critical field is of the an explicit calculation also have avoided *Note we and normal the between difference energy free the with associated is that states superconducting tunneling Hamiltonian. 120 almost entirely determined by the free energy differences of the s-side. pair This is because near the transitions potentials, differences, and therefore are small in the n-side comparison the n-side free with those energy of the s-side. 4. Finite Temperature Calculations All of the temperature above sandwiches. numerical results temperature formulas We more zero-temperature ones. point of isolated We finite present Finite difficult than They would be the most phase diagrams the region near the tricritical paramagnetically limited superconductors. choose not concentrate on for sandwiches. certain details of the of these sandwiches such as sandwiches. no valid however, only for zero-temperature useful for exploring Here we shall, calculations are corresponding are to explore these the detailed properties shall see that these features but instead of zero-temperature properties themselves are quite interesting. Finite temperature calculations vicinity of a second-order phase parameter, zero. in this case the In this regime, become simpler in transition where BCS pair the the order potential, goes to the renormalization equations become 121 possible to make some headway analytically. it is linear and We discuss some analytic results in Appendix B. procedures we will We have now outlined the calculational sandwiches results of order parameters, states, the of calculations numerical the present sections we two next In the . an proximity-effect limited paramagnetically Pauli isolated results for corresponding indicated the and have use densities the of magnetizations, free and the and susceptibilities of two representative energy densities, First, in Section paramagnetically limited superconductors. B, we consider a sandwich consisting of a superconductor and a much superconducting weaker so the of superconductivity n-side consider superconductor sandwich a coupled these films remains (n-side) and this n-metal film). take the the weaker The parameters of (i.e. an the strong but still proximity loss of the superconductor transition can be compared to weaker superconductor a case, there In this to be weak. superconductivity of isolated thin We transition associated with the a phase "inherent" of weaker, another superconductor. moderately strong, coupling of to In Section C we consists which its by swamped are proximity-effect-induced superconductivity. then "inherent" the of vestiges all that strong be rather strong, these two films to proximity coupling of the take We "n-metal." that of an isolated thin two sandwiches we 122 study are given in Table IV.1. 123 Table IV.1: Proximity-effect sandwich parameters Sandwich in: As Section F X (E )v Section 1 1 0.0011 0.32 152. N'(E )V B C 6.46 0.08 0.27 0.175 0.39 0.46 0.02 0.184 0.04 'i- is the energy gap of metal 1 in the absence of the proximity effect. 124 A NORMAL-SUPERCONDUCTING SANDWICH B. next we this section and the In proximity-effect sandwiches zero-temperature. In magnetization, the parallel in we consider this section spin the thin of susceptibility the and results of for model pair potential, the energy, free the states, of densities tunneling the using calculations present the at fields a sandwich The which is effectively a normal-superconducting sandwich. much weaker superconductor "n-metal" used is actually a superconductivity proximity-effect-induced the at look the two-superconductor a of properties overwhelms In the next section we its weak inherent superconductivity. will and sandwich. The are parameters of the sandwich we discuss in given first column the superconductor, when isolated and not zero-temperature, zero-field energy gap the unit of calling a The n-metal, energy. "normal superconductor which zero-field energy gap metal" of widths associated sufficiently strong 4 that the in proximity, has a AS which a bulk, =0.0011,65. with s-side is metal, would have The IV.1. Table of the in this section which we take as we are loosely weaker a actually zero-temperature, The tunneling tunneling model proximity-effect decay are induced 125 overwhelms its weak inherent superconductivity in the n-side superconductivity. In Figure II.1 for prediction plotted the tunneling model we the zero-field density of states for both the n- and later the curves in this figure figures indicates which semiconductor model. in the considered occupied states are in the n-side also the energy gap in the s-side density half of the amplitude in energy states longer localized are no the sandwich but sandwich, be seen s-side density of states finite that the lowest n-side. in the are primarily localized in either a have they rather It can each side. of states, which that in the n-side because the must in fact be identical to in This is density of states. of 2 Jl0=2(0.34) states energy gap a substantial proximity-effect induced There is the shading under The a proximity-effect sandwich. s-sides of and value is greatest near the The of its self-consistent pair potential which, in the presence of the proximity effect, is reduced to 0.81 As. this energy reflecting the of states decreases rapidly near fact states there is that suddenly on the states s-side. resembles, except Abrikosov-Gor'kov8 6 by Fulde and Maki0 for density depaired superconductor. 7 that a much larger amplitude rounded shape The The n-side density the of of the density of two hump states (Indeed, it has for feature, of been proximity effects are a the partially pointed out in certain 126 to the to identical limits depairing other of effects perturbations.) As a parallel field is and down-spin quasiparticle states are top the of portion of states H = plot, this energy gap £Q0 and is not great The number and this field strength cross pair the Fermi level. state Cooper field case. of field is There zero-temperature. and As pairs a result, potentials have not been altered by a of application (half) enough to cause the up- from the zero the self-consistent pair As the zero-field the nature of the ground remains unchanged these The applied field for states to cross at down-spin densities of the than less In corresponding densities electrons. 0.2gry is In down-spin n-side and the s-side. we give the the for the up-spin the give we figure densities of states for both the the bottom portion The shifted by ±dAf. split as shown in Figure IV.2. spin density of states is the the up- applied the energies of also no this magnitude net magnetization and the spin susceptibility at at is zero at field values. the field at increases enough the Fermi states have quasiparticles with with the field. level (Fig. energies to cause IV.3), some of exceeding both of their magnetic In this the states that to the ground of two moments aligned case, as the shading in Figure IV.3 N(E) N (E F) 2 sn N1(E) N(E F 0 0 -2 ENERG Figure IV. 2 2 2 N (E) N(EF) 2 Nt(E) N(EF) 0 ENERGY Figure IV. 3 2 ( eV -) 129 field exceeds _1l./,us, start zero-temperature superconductor at in Fig. IV.4) (indicated by the dashed line the critical field from zero through as the potential pair a constant would have s-metal limited paramagnetically isolated an of behavior The reduced. be to IV.4) in Figure plotted lines self-consistent in the is reflected (the solid pair potentials breaking of pairs once the This the spin alignment energy. due to at lower energy down-spin which are now and occupy their spins electrons flip of the up-spin some indicates, field varies up to the- all the way superheating field. As the field decrease. continue to sandwich when Pauli the free energy of the and the potentials pair the and increases depairing resulting the splitting increases further, of the the critical field We reach energy of the sandwich in the equals the sandwich normal state. The proximity-effect sandwich then undergoes a sudden transition into the reflection normal state. of the zero-temperature is first The transition of the first-order nature transition of an paramagnetically limited superconductor. with an isolated there is, paramagnetically limited associated with the proximity effect equations for the sandwich, a order, a field-driven isolated Pauli In direct analogy superconductor, first order transition of the superheating solution pair potentials as well of the as an unphysical 130 1.0 ph P ~A(H) S. / As LU In~ 0.5 n/ -/ 0 ph 0~r 0 0 hsc hc hsh MAGNETIC FIELD Figure IV. 4 1.0 h = P-BH 131 normal state supercooling extension of the Figures IV.5 and In The total respectively. these density the sandwich, the peaks in the another once cross appears once the which of states the a superconductor's zero-bias n-side. states to cross states of These Fermi level. peaks region in The states density of a to cause large enough are the field-induced splitting is at The most peak spin densities in the energy gap measured by curves is the large zero-bias large peak results from This Fermi level. identical to at zero-temperature. and down-spin densities of the up- of of states is induced splitting is magnetic field them to density of s-sides density sides dramatic feature of these the n-side plot the total characteristics which would be the conductance tunneling into IV.6 we and n- the for states the does but it 1 for the isolated film at finite temperatures" . occur in This corresponding temperature, zero superconductor at limited superheating paramagnetically Pauli isolated an of parameter order dotted for the not occur evolution did the by the unphysical curve. solution goes continuously into continuous These be seen that the It can in Figure IV.4. lines solution. indicated are gap equations the of solutions the to point superheating the connecting solution the addition of just above the added to one sufficient for zero-energy and, at the region of correspondingly normal-oxide-superconductor 132 Tunneling to Normal Side 4 2 =0.0 U) O (I) N(E)0 >.N(EFI - /.BH As z LU 0.2 O 4 _J 0 2 -=0.4 OL -2 ENERG Y rigure IV.5 2 0 \ eV s I Tunneling to Superconducting Side 4 133 2 H s 0- (fO =0.0 L- N(E) 2 0~ N(EF) = 0.2 -2- H 0-2 Figure IV. 6 0 ENERGY eV $ ' =0.4 2 134 tunneling experiment, are usually The dramatic peak structure in should be readily featureless flat regions. the n-side density of states observable. In Figure IV.7 we plot the magnetization on both sides of this h = sandwich MOHg/Ag. as We a function of indicate a dimensionless with solid superconducting state magnetizations of the sandwich and also magnetization. the corresponding Dotted lines indicate superheating comparison the and the two sides of the state the magnetizations solutions, magnetization s-metal paramagnetically limited lines linear normal the unphysical the corresponding field of and a of for isolated thin film is given as the dashed line. As we indicated above, magnetization reflecting of the of states both sides cross is an onset of the superconducting breaking of electrons can align Once the field there of the Cooper at pairs -Wco the point at the Fermi level, sandwich starts the h so their magnetic moments with exceeds at some state a nonzero =/tv that the the field. which the density magnetization to grow. The rate on of growth of the n-side magnetization is very rapid and exceeds that of the linear Pauli The growth magnetization in the normal state. of the magnetization increased further until continues as the the critical field of field is the sandwich 135 a- m 0.5 z 0 N ZJ Or 0 a0o hhsc hChhsh hc As MAGNETIC FIELD Figure IV.7 h=A 'BH I 136 is reached at which point the sandwich makes a first order transition into the normal state. The susceptibility, the is plotted in physically Figure derivative of the magnetization, IV.8. The solid are the realizable curves. The magnetizations of the lines superheating solution, the supercooling solution, and the unphysical connecting superheating and solution supercooling points are indicated fields the which allow susceptibility, the n- the with dashed breaking of pairs and susceptibilities s-side similar in shape to the respective states above identical if decreasing as the energy gap. n- and (These the self-consistent unusual the field increased.) In the next chapter behavior in the and a For nonzero are s-side density would in fact pair potentials Note that field dependent susceptibility actually normal state. lines. of be were not the n-side exceeds that in the we comment on susceptibility how this might be experimentally verified. A zero-temperature superconductor case in is which this has partially depaired In the an nonzero susceptibility unusual occurrence. been predicted occur for The only to occur is state postulated by Fulde Fulde-Ferrel state, momentum states the depairing localized in a pure other in the and Ferrel 3 9 . is predicted in certain to regions of 137 xS 2 I F0 0 -1 D co -2L 0 hsc he h sh M AGNETIC FIELD h=A Figure IV.S I Ss 138 As one might guess even a small amount of the Fermi sphere. in defined regions well prohibit the scattering would impurity momentum been Fulde-Ferrel state has never the sandwiches, the depaired states are and indeed In the such the case of proximity-effect these localized in real space, the possibility of observing reasonable. For example, primarily on the n-side, such a state appears to be quite the n-side Knight shift should space, and observed. in state depaired partially existence of show a sharp growth as we go into the partially depaired state*. curves We integrate the magnetization to get the Figure densities which are Gibbs free energy IV.9. The realizable solutions lines indicate and the normal state the unphysical the indicate lines solid as described above plotted in physically solution, dotted extrusions of these curves. The dashed curve is the corresponding free energy density of a isolated paramagnetically limited for comparison. The n- and s-side the superconducting state cause the Thereafter free energies are flat in until the field is great enough to up- and down-spin the free s-metal and is included densities of states energies densities start to to cross. decrease reflecting the response of the increasing number of depaired next chapter we *In the systematically. discuss possible experiments more 139 O normal state n s LU LL LUJ LL 0 I I 1 S2 h h h A sc MAGNETIC FIELD Figure IV. 9 C sh h =PBH As 140 electrons sandwich free and s-side free energy densities*. IV.18) of the n- sandwich, it of density susceptibility occur level. most interesting that the field needed exceed the spin-split density of states to We indicated in Chapter cross we have seen that the superconductors. is still at the Fermi transitions In effect proximity zero-temperature transition with increasing phase transition to I that the proximity effect should also influence the nature of the phase paramagnetically limited n-side These effects of the n-side of the sandwich. values which features larger-than-normal-state the and states in the which occurs zero-bias peak for field cause the this proximity the predicted behavior of is clear the huge were the total weighted sum given by the energies which is In summarizing of by the crossing determined entire sandwich is (Eq. The critical field of the to the magnetic field. first order. in this example rounds field, In fact but the the the most interesting features of the phase transitions occur not when the entire sandwich superconductor loses goes normal but rather its inherent when the weaker superconductivity but induced the proximity-effect weakness of of the *Because very occurs crossing the the n-side, superconductivity of free superconducting and the n-side normal close to where the in difference the Indeed cross. densities energy discerned be cannot s-sides and nthe crossing fields of on a plot of this scale. 141 The inherent superconductivity of effect. the proximity of superconducting because remains the sandwich discussed in this section was too weak compared to the proximity coupling of its own superconductivity. strength to show any vestiges In the next section we study a sandwich which consists of comparably two strong are which superconductors weakly coupled by the proximity effect. C. A TWO-SUPERCONDUCTOR SANDWICH In this in interested we are section transition and the influence zero-temperature phase the studying of the proximity effect from a stronger superconductor on the phase transition with superconductor. Chapter I and I.10 expect, We the that increase the critical field loses its "inherent" this section moderately strong s-metal with only we the of a weaker analogy and a metamagnet in drawn (see Figs. act effect will proximity to where the weaker superconductor superconductivity superconducting only because In from between this system I.11), field increasing and left is of the proximity effect. consider superconductor. is n-metal which an We couple a weak proximity effect. it to a the The parameters 142 given in column 2 of Table for the sandwich we consider are IV.1. Again we choose an s-metal have a zero-temperature, zero-field energy table 1 widths coupling decay 0.32,A, gap of in the are given magnitude smaller than these pair and are an order of potentials. we The n-metal when isolated would take as our unit of energy. The proximity which 6& potential pair zero-field zero-temperature, which when isolated has a This makes the proximity coupling rather weak. plot the zero-field, zero-temperature In Figure IV.10 we density of states for this As should be expected sandwich. for the case of weak coupling, each metal has states which is only slightly a density of its perturbed from BCS-like The most significant alteration from a BCS shape bulk form. which now has occurs in the stronger superconductor a small but non-zero density of states below its bulk gap energy all the the to down way the of gap energy weaker This is again a reflection of the fact that superconductor. the states are no longer completely localized in either side superconductor, but this plot. anomaly in a slight states near the gap energy n-side density of this is also There sandwich. of the of the s-side cannot be discerned on the scale of and zero-temperature At the field, the self-consistent pair potential of the s-side is reduced from its bulk value of increased from A, 0.32&S to 0.96 AS, to 0.344s. and that of The the n-side zero-field (half) 143 Nn (E) N(E) N, (E) 4 Ns(EF -2 0 ENERGY Figure IV.10 ( As) eV 144 0.364A3 sides, 14, is on both energy gap just is and this slightly greater than the pair potential on the n-side. As the the splitting of Zeeman parallel or antiparallel alignment occurs due to the either There should be spin with the field. of a quasiparticle's again states densities of spin the (Figure IV.11), is applied field magnetic no reduction in the pair potentials until we reach the field ASH =1J6, where the first-order critical field However such a field exceeds an of superconductor. its isolated form by proximity effect, the weak "inherent" superconductivity the lost. The n-side is then its proximity perturbed from is only the n-metal H /, h' first-order critical field reach a of Since n-metal limited paramagnetically isolated states cross. densities of the spin-split of we first =0.33 where n-side metal the is left superconducting only because still to the superconducting strongly s-metal. In Figure IV.12 we plot the density of states at a field above this first order transition the zero-field density Fermi level. It can of states would have be seen that single-spin density of states has through this self-consistent phase but below the field where transition. n-side pair the crossed at the energy gap in the been much reduced on going is This potential as true well. of This the is 145 N (E) 4 N(EF) 0 Nt(E) 4 N(EF) o0 -2 0 ENERGY Figure IV.11 1 AS 2 N (E) N (EF) Nt (E) N (EF) 0 -2 0 E NERGY Figure IV.12 1 ( eV ) 147 plotted in a discontinuity shows IV.12), the phase transition (Fig. have gone through there quasiparticles. of spin-aligned number a substantial at hG. plot after we of states under the density From the shading are and Figure IV.13 in the magnetization on the There is also a sudden increase n-side which is plotted in Figure IV.14. IV.14 and Figures IV.13 the also show corresponding behavior of the s-side pair potential and magnetization near h discontinuities at h'C . This is a reflection The jumps the of are rather that fact show also these that seen be can It . the jump small though. states with pair potential energy of energies below the superconducting the s-side are predominantly localized on the n-side because of the weak coupling. are plots of the total density of Figures IV.15 and IV.16 states of the field where n-side n-side and the s-side values below and the n-side loses total spin-splitting. just its respectively for these above the transition at inherent superconductivity. h The clearly the This splitting takes a rather curious shape density of states shows e near zero-energy The s-side, at fields above on the other hand, the transition field he. also shows the spin-splitting clearly but the features near zero-bias are small. 148 1.0 -J ------------------h ph A(S(H) . A s LUJ -0.5 ph 0 s nP r:1 0 hscIhcIhshI hsc2 MAGNETIC FIELD Figure IV.13 hc2 hsh 2 h= I BH U) < L 0.5 0 normal state IA NLU 0 0 ----------I I I hschc hsh h2sc MAGNETIC FIELD Figure IV.114 h2 c h = /-BH AS Tunneling to Normal Side 150 8 4 Cl) LU H LL 0 H O 8 N(E) N(EF) 4 -- 4 0 -2 0 0 ENERGY Figure IV.15 2 2 ( eV 151 Tunneling to Superconducting Side 4 LU F0 Co I- 0 CO z 4 N(E) LU N(EF O 4 0'L -2 0 ENERGY Figure IV.16 2 eV) \As 152 Pauli paramagnetically superconductor limiting field labeled as he in the order B above. This stronger a transition parameter and magnetization is to the IV.14, is similar the normal-superconductor transition of the transition, which occurs at a in Figures IV.13 and plots given of then have a flat density Both sides This second phase of states. field and the whole sandwich undergoes into the normal state. Section we encounter the field increases still further When the in film discussed because the "inherent" superconductivity of the weaker superconductor, the n-metal, is destroyed by the field and the is only n-side left superconducting because of the proximity effect. Finally in respectively. densities, realizable portions of s-metals. behavior the free In the dashed lines fact that both the above, the dotted solution, unphysical corresponding As of these curves portions the free energy curve the Figure and susceptibility paramagnetic of and the we the curves of the of the to get free energy plot the line gives the physical physical lines. and the unphysical unphysical n- we indicate solutions. is of the curve of isolated curve In the behavior dashed transitions are first order the existence IV.13 with solid energy densities susceptibility differentiate and we magnetizations given in integrate the the and IV.18 Figures IV.17 and with The reflected in superheating, supercooling, and 153 (n 0- 2 n statfe .-~2 is 0 n ss s - =- n 0_ LLJ -2 0 hChc hsh MAGNETIC FIELD I h2 h/-BH A Figure IV, 17 s I H0 0 LUi z LUJ LUJ LUJ O0 2 hschC hhhhshh sc M AGNETIC FIELD Figure IV.18 c2 h sh = BH 155 unphysical excursion he are There . a to the of number h. various of the crossings and in divergences the unphysical curves and these susceptibilities of the traced energy near both of the free nearly can be singular density of state curves. weakly coupled existence of two remnants the of order when the second In effect. the proximity by transition the normal state. is loss of leaving only induced superconductivity the field from the first transition results sandwich goes into first as superconductivity of the n-metal the inherent proximity predicts sandwich transitions phase such a tunneling model for proximity-effect The increased. the that the we have seen Thus weak a the entire transitions are Both coupling of the films is weak. for evidence is experimental There state the as raised"8 . is I, order. of (instead it would As mentioned be in the interesting experimental and theoretical exploration and field dependence of the order the sandwich into the normal the In this temperature case, however, second Chapter temperature of and then the phase inherent superconductivity of the n-side transition of the proximity effect the loss field) is the transition introduction to have a in detailed of the temperature parameter of the n-side in 156 transition should change from first terms in interpreted order, all*. and finally to no transition at of an exploration this temperature increasing With superconductivity. inherent its loses n-side the where region the to second order, This change of the could be "wings" of the tricritical phase diagram given in Figure I.11. We survey of given a have now expected for Pauli-paramagnetically we describe shall some tunneling undertaken in an attempt to in the spin-split density experiments were suggest some experimental experiments besides experiments were that see these conclusive, and we shall of states. We shall tunneling also shall lead which might modifications We In the next chapter observe the structure predicted promising but not results. limited superconductors model. based on the McMillan tunneling of properties the type to discuss some experiments clearer other which might be attempted. *In fact in the sandwich considered in Section B illustrates considered in this section the no-transition case and that Both of these illustrates the first-order-transition case. things very did not vary zero temperature and we were at Nevertheless it is clear one could use the systematically. model presented here and to do such a systematic study, and get results to indicate one would these calculations seem the winged in terms of indeed be interpreted which could phase diagram. 157 CHAPTER V: IN A. EXPERIMENTS ON THIN PROXIMITY-EFFECT SANDWICHES HIGH MAGNETIC FIELDS INTRODUCTION In the McMillan's the preceding tunneling behavior of chapter model for we the used an proximity thin proximity-effect study sandwiches in high We saw that such predicted and striking have unusual their densities of states and magnetic susceptibilities. IV of a sandwich coupled to in their sandwiches are characteristics in magnetizations and Furthermore the study in Chapter consisting another of effect to parallel magnetic fields. to extension weaker of a strong superconductor superconductor as well as preliminary considerations given in Chapter I indicated that these some two-superconductor sandwiches interesting phase transitions as transition of states and into the in the at the normal state. reason for undertaking more detail sandwiches would thus be to look both at these interesting in order have the field is increased. Experiments on proximity-effect densities should nature There such experiments. next chapter, features in the of is the one As we spin-polarized phase further discuss in tunneling 158 from isolated superconductors spin-orbit scattering effect it may be possible By mixing spin Furthermore it spin-orbit scattering times to measure in proximity extend the measurable range of to metals which are not superconducting of with moderate using the spin-orbit interaction. the strength of the would be possible superconductors interactions. strength spin-orbit The superconductors. times in these limited however to method is affords a way of measuring the these in by analyzing the degree metals when have they a proximity-effect-induced superconductivity. In this chapter preliminary tunneling Drs. their R. Meservey, D. unique we describe experiments tunneling National Magnet Francis Bitter comment which were Paraskevopoulos, high-field and undertaken by Tedrow using and P. facilities Laboratory. some on We at MIT's shall also present suggestions for future tunneling experiments and experiments to measure the field-dependent susceptibility of proximity-effect sandwiches. for magnetic 159 EXPERIMENTS B. TUNNELING 1. Theory of Tunneling most the of One across an oxide the is superconductivity metal below, directly is of the quasiparticle states tunneling between The we density the to shall of superconductor, and Josephson (which we shall not two superconductors reflects the coherent nature of the superconducting discuss) ground normal structure, as related another and a of of conductance the of superconductor conductance metal-oxide-superconducting outline a probes experimental measurement placed between The metal. detailed state. current-voltage characteristics between two metals tunneling are easily Hamiltonian we of interpreted1 8 have been oxide an barrier using the same to describe using 160 two between the coupling metals enough that a potential difference weak to be supposed is the case however, this In sandwiches*. proximity-effect between the metals can be maintained and that the tunneling Hamiltonian can be treated order perturbation from metal 1 to metal 2 is then given by A E IT NJE) f(E)N2(E+e\/)I-f(E *eV)J I where current flowing The theory. in lowest difference applied that is V N(E) is the density of states in metal i, and A function, is a constant probability of (proportional to states in metal electrons ITI2) metal energy in metal 2), N (E+eV) [ + eV - f(E) is 1, N,(E)f(E), from the number of 2 at energy E metals, the Fermi This product of the number tunneling times resulting of proportionality. equation results from integrating the of occupied initial states in the the two of chemical potential in the is eV voltage, (V.1) these . states available empty (relative to f(E+eV) times the the Fermi Summing this and the reverse current, we find AI4 ITIl N,(E) NI(E t eV)[f (E) -4f2(E +e V)] (V.2) a nonequilibrium through the barrier is *A current flowing treats theory Hamiltonian tunneling This situation. nor the perturbation neither the nonequilibrium statistics has Feuchtwang 78 Recently rigorously. case this in of tunneling for barriers developed a more rigorous theory perturbation nonequilibrium a using metals normal in to theory this extended has Arnold 8 9 and theory, in theories are these of The results superconductors. accord with the results of the arguments reproduced here. 161 that Assuming the considering the matrix case where element metal is T constant and metal, this normal 1 is a becomes 7JT12 N\)(E,,) {'4 ( E eV) Or where is G the corresponding are in the both metals of Eq. (V.3) V.3 with we find voltage and to respect given by the derivative is The conductance (JE~eV)LF&)-A 4-eV)3 f(E eV)1 (V.4) when constant conductance zero temperature At normal state. just this is , VI( N,(e Ivl)I (V.5) NZ(EF superconducting density Eq. the derivative smearing results from This temperature average of the density of V.3 gives a thermally smeared states. At finite of states. the yields directly zero-temperature conductance Thus the of the Fermi function which is a bell shaped function with width of order k T. We have not solved IV.5, for rough idea the self-consistency equations, finite temperature. of temperature looks what the like by We can, conductance performing the Eqns. however, obtain at finite a (low) thermal smearing 162 calculations carried out resulted from the zero-temperature Chapter in IV. considered there of Table IV.1, V.2 for For the normal-superconducting results given in Figures V.1 we find the at bias zero spin splitting of the density of conductance. This is the conductance at low bias is would be obtained from a As the field (times the states is reflected in the visible an the the thermal in the glancing angle from readily noted In in the splitting curves from a s-side.) the a the field is applied the to mask below, the splitting can be more curves of result from density of states but sufficient (By viewing would than is clearly splitting thermally smeared smearing here field and sides are predicted to have As smeared BCS density of states. n-side. 1 the normal and superconducting sides respectively. conductance s-side sandwich and having the parameters given in column In zero magnetic field, both more which states density of the over V.3 Eq. in indicated in the applied field, again larger than that lower the which spin-split BCS density of states*. Bohr magneton) increases beyond the more show curves conductance spin-split *Generally, the by explained be than can zero bias conductance near spin-orbit and depairing orbital both of inclusion It is not implausible to suppose that there is scattering. some proximity-like perturbation near the oxidized surfaces bias low enhanced this causes which metals the of tunneling the that recall should One conductance. conductance actually reflects the local densities of states near the oxide surfaces. n -side C 0.48 < 0.45 00.3 0.2 0 BH s C -o O 02 -2 Voltage Figure V. 1 eV2 s- side 0( :D 0.4 0 00. - 0 00 .3 0.2 - -B 0 Voltage Figure V.2 -- H I 2 eV 165 oero-field induced densities of energy gap, Jl0, states cross at the which resulted in indeed reflected in the normal the up- and down-spin Fermi level and metal density the conductance for the peak of states tunneling is into the n-side of the sandwich. At peak fields great in the enough to n-side density smaller peak in the s-side IV.5 and however, peak. IV.6.) In the thermal For of the large zero-energy states, there was also a density of states. (See Figures this finite-temperature conductance, smearing is At lower temperatures this s-side cause enough to obscure this there would indeed be a peak in conductance at zero bias. suitably chosen proximity coupling parameters, the s-side peak may be pronounced enough to show up even after a significant amount of account. The always be energy thermal smearing has low bias peak on greater than that on states which been the n-side, the are involved taken into however, will s-side because have larger the low probability densities on the n-side. 2. Tunneling Measurements A number of considerations restrict the choice of 166 materials which boundary between the metals must the low solubilities The metals chosen can have in order that the superconductor to low atomic beryllium90 , alloys 9 1 , and gallium' 2 . in which spin splitting of case (In the only limits the choice of vanadium-titanium or These are the only superconductors been clearly the densities has V-Ti alloys, the and of V the number metal atoms such as vanadium aluminumzo, splitting has this In practice states. densities of splitting of spin an observable be there still in which does preparation amount of spin-orbit scattering only a small observed. junctions, be sharp and well defined. of sample and a method not allow much atomic diffusion. the in must choose metals with This means one one another behavior discussed make good proximity effect To thesis. proximity-effect construct exhibit the that should sandwiches this used to be may the To avoid very recently been observed.) the depairing which accompanies the Meissner diamagnetism in thicker films, If the thin as possible. may form disconnected electrically it always necessary to use it is is, to use islands of metal Finally, metals temperatures as possible so that be carried out with are as rather than it is desirable, as high an as transition the experiments can easily at a reduced temperature smearing effects are minimized. which are too thin, however, they films continuous film. films T/T. where thermal 167 in V and its The spin splitting large enough to make the effects hard to scattering in Ga is The initial candidates for experiments were Al and observe. films Be the warming Upon a transition Be has substrate, 0 K. their lose and transition high relatively The superconductivity. 8.7 around temperature anneal glass cooled helium liquid onto a deposited When Be. spin-orbit the because observed and been recently very only Ti has alloys with temperature of thin film Be makes it a desirable material to use, designed for low-temperature tunneling film in an apparatus experiments in high fields and the magnetic make it a difficult material with which temperature transition film), has the on Aluminum, (2.4 0 K used experiments, extensively and the proved were for undertaken contained in this thesis. as to choose a metal with a for the of the result For the n-metal, it low atomic number little spin-orbit scattering the up- and down-spin states in the n-side also. and copper were chosen for these It tunneling choice simplest low thin a as spin-polarized calculations there will be relatively to deposit and handle. which so that a when deposited experiments is also necessary toxicity of Be to work. has other hand, but it is relatively easy been such a cryogenically depositing the necessity of but to mix Magnesium initial experiments. 168 The films used were made by vacuum peak in the the predicted zero-bias in the were made with Al on the normal side of had shown that 50 thickness for a single electrically continuous f first the was tunnel sandwich. junctions Experience about the lower Al film if it film. sandwiches were made with both the n- limit of is still to be the Therefore Since is largest conductance sandwich, the n-side of the evaporation. an initial and the s-films having thickness on this order. First a 500 % Al film was evaporated onto Next the Mg was substrate and allowed to oxidize. as a cross thicknesses those top strip of of the Al and the Mg this was would subsequently thickness of the Al superconducting deposited thickness. itself has in the form of bulk Al). The proximity effect this transition substrate by could only be cooled to so Al. to (versus much about 0.4 25 R and of the the actual than the of 1.175 2.5*K 0 K for metal lowered as 0 The Some temperature from the Mg as deposited film is less a thin film temperature f X. oxidize a transition when it is by 40 50 A to 37 layer Aluminum covered layer varied from top Al layer from a sapphire K 0.8 0 K. (using The 3 He at low pressure), so there was considerable thermal smearing of the density of states characteristics. in the conductance 169 The 40 a 1.7 0 K. the evident, but states was of density of the splitting spin field the parallel in a placed When of transition temperature zero field and had of Mg about 50 R of Al deposited on initial sandwich consisted of sandwich was too thick to neglect the orbital effects due to the state long before into the normal for enough spin-split peaks the sample returned The field. of the magnetic screening splitting was this conductance the in large to cross. The transition caused and becomes more and more of states with increasing field significant in thicker films. of a series of reflecting boundaries. taken to spheres equal to the eigenvalue of the eigenvalues those of the spin of this operator operator, both the densities of states will be spread out. the occupied states have larger the angular a magnetic field will be raised or lowered the energies of these eigenstates Since specularly sphere can be z component of In the presence of momentum operator La. by an amount have which The eigenstates of the be eigenstates of the To imagine that the film explain the broadening qualitatively, consists apparently This effect causes largely by the orbital depairing. a broadening of the density was normal state the into the operator AL*. are unrelated upFor and to down-spin larger spheres, orbital quaptum numbers and 170 but the effect of the spheres, films are really not Thin the orbital depairing increases. diamagnetic screening will be similar to that in this simple sphere picture. Paraskevopoulos evaporated thinner Meservey, Tedrow, and this deleterious films to try to mitigate The characteristics they observed for side of a 25 normal state from the state from 2.50K to into the superconducting increases further and there is an appreciable 3.2 Tesla there is a be must In a field about 2.30K. reflection of the bias. at zero crossing actual At bias which at zero a substantial peak As is evident more spin-splitting conductance of 2.0 is evident. spin splitting of the conductance Tesla, the the bias in Figure V.3. a rather broad transition a midpoint at 2.20K with tunneling into the Mg Al sandwich are shown t Mg-40A junction displayed This thickness effect. of the spin-split densities of state at the Fermi energy. When the the sandwich the conductance curve is field returns increased is to 3.4 further to state and the normal Tesla; flat. Due to densities temperature conductance the significant of states in orbital this sample, smearing, there is not characteristics to of the as to the broadening as well sufficientt detail in the justify quantitatively fit the characteristics. an attempt to However there are a 171 Tunneling into 25A Mg on 40A Al U, :D 0 0 : 0 0 0 -l 0 Voltage (mV) Figure V.3 I 172 things to couple of Figure V.3 to the n-side characteristic of self-consistent the fields, lower qualitative calculation values. peaks at for the field included field at these strengths This is a result of There is a greater as compared zero-temperature field dependence of the pair potential. must include also the caused by the potentials It is clear that the degradation dependent theor'etical reproduce the the lack calculation meaningful. does We pair the proper inclusion of the temperature of is the pair in the experimental curves. data of to orbital depairing. of a sample which experimental degradation added and self-consistent pair temperature finite at potentials included. increasing field of the decrease with In the theoretical curves show of most which should be the appropriate decreases to about 0.3 mV at H = 3.2 T. two effects in the experimental peak remains ^ +( A5 + AH) whereas 0.4 mV At have apparently At higher field values, nearly at Figure V.1. the theoretical curve. we is smearing than more thermal smeared density evident clearly splitting is experimental curve though not in our in given pair potential experimental zero-temperature the from calculated states of comparing the note when potentials right At this direction For a quantitative such a the to time however, gives distinct features not make in in the comparison too investigation of the theory 173 a more distinctive achieved be be could with thinner diamagnetism, or with lower showing less films This orbital and temperature measurements, sandwiches measurements on systematic is required. higher-transition-temperature with superconductors, more curve experimental with varying did attempt proximity-effect coupling parameters. diamagnetism which went to They curves. film. orbital the 25 A Mg These sandwiches with 1.50K, a again splitting The conductance. the broadening a 35 i Al transition to a broad the transition ranging from 2.30K about midpoint at was is films backed by again showed the superconducting state, to an attempt to even thinner film in such measurements on an lower Tedrow and Paraskevopoulos, Meservey, clearly evident diamagnetism was also very lowered 1.850K. The spin in the apparent. Although these sandwiches had a lower transition temperature than A the 25 given in Mg-40 R Al Fig. V.3, film they had characteristics whose a much higher critical field. Unfortunately, at 0.40K, the critical field was the 3.8 Tesla maximum field in which performed. the measurements states cross. greater than of the superconducting solenoid on this At this maximum field, reached the point were where the up- and particular sample the were sandwich had not yet down-spin densities of 174 Paraskevopoulos, across conductance % to the that illustrates transition differ sandwiches proximity-effect This K. is Mg-Al films sharp The n-sides. thicker had which above discussed 0 for the broad transitions the into transition a temperature of 2.19 superconducting state at in contrast These films displayed of Cu. zero-field sharp rather a % Al film backed with 15 the this for a They did s-side of a proximity-effect sandwich. 37 on constructed oxide barrier an field-dependent at the Tedrow looked and Meservey, measurements, n-side to these addition In these thinner greatly from Cu the Mg sandwiches discussed earlier. H = 0.22 Figure in displayed T does not side of the s-side zero-bias reached. this s-side maximum If before just this were sandwich both of this conductance whereas one sandwich which is sandwich its One the fields and the appearance before the field driven shape of of the also displays critical tunneling into the large transition into at a bump on There is tunneling Hamiltonian model. well characterized by the notes the curve s-side density of look like any of the in Chapter IV. are junction conductance the low energy side for a is expected on also The V.4. states curves displayed the high energy this obtained for conductance curves The field the normal the normal is side of conductance at zero-bias a low maximum state Tunneling into 37A Al on 15 A Cu 4- _0 C: -o0 Ll C) 0 -1 Figure V.4 0 Voltage (mV) 176 explain model can for s-side tunneling. is indeed doubtful that however, that it the McMillan so thin on an in this film is The copper even doubtful that the pair potential is Cu film is that it is not possible to draw any validity of the the assumption of It is justified. one. a continuous of this metallurgical characterization atomic scale, step-function shape tunneling model about the of the self-consistent to the the tunneling into is features the tunneling these are not s-side and this However understood. could be junction is The so poor conclusions with respect tunneling model from the conductance characteristics for tunneling into this sandwich. In summarizing these results, measurements show did tunneling model. we can say that the n-side agreement reasonable It would have been with very interesting the if s-side measurements had been made on sandwiches identical to ones used the section we for the discuss other n-side measurements. In the next suggestions for future experimental work. 3. Suggested Future The results Tunneling Work presented here of initial experiments 177 and suggestions result of the calculations undertaken as a in this thesis are promising in some respects and perplexing in clear It is some others. In required. subsection this and additional to theory, quantitative comparison of experiment experiments are a systematic that for we indicate promising for future experimental what directions look most work. thermal smearing of the density of states which occurs in a the either decreasing the conductance temperature at which by increasing the transition is measured or the proximity effect sandwich. orbital diamagnetic broadening the films is desirable to make each them can be done by This conductance versus voltage measurement. have to minimize the it would be advantageous Most obviously to reduce the of the density of states it as thin as possible and still continuous be superconductors well pair potentials. constant by spatially characterized Additionally, temperature of comment on each of these three suggestions We in turn. Lower Temperatures a By using 3 dilution refrigerator He-4He samples, instead of cooling with liquid it would be possible temperature by a factor of temperature tunneling to 100 easily 3 the He at low pressure, reduce or more. to cool the sandwich The high-field low facility at the Magnet Laboratory is 178 going to ultra low temperature would or 0 K) 0.026 0 K) bulk (with a Be One is restricted to condensed Beryllium, temperatures, was, 0.84 0 K) temperature transition of Superconductors spin until very advantage for an additional Low temperature sandwich transition 1 ow deposition temperatures has proximity effect experiments. in proximity-effect two layers to a sharper interface between the by most of the a sharp interf ace is assumed The high transition temperature another advantage only candidate interdiffusion of This leads and such He deposition mi nimizes the the between liquid at recently, the the observed. be splitting can maintained and Going to such temperature. theories. using materials with low spin-orbit increasing for material sandwiches. the so that scattering films of about might be the weaker superconductor. Higher Transition Temperature atoms stronger superconductor (with a bulk transition temperature of and Zn annealed be the thin film transition temperature (with a isolated 2.5 would Aluminum Chapter IV. of Section C in discussed geometry superconductor the allow superconductor-stronger weaker the of realization experiment, the would temperature of decades two additional a number be useful for For this this one. besides of experiments capability of the additional facility, and already a unique for its use. Aluminum of Be gives can be used still for the 179 n-metal and Be the s-metal, realization of resulting in a the realization of the two-superconductor sandwich discussed in Chapter IV Section C. spin-splitting recently Very states has been very clearly observed density of some advantages and offer have 40K respectively 30K and of about transition temperatures in Vanadium These films films. alloy in Vanadium-Titanium and superconducting the of Additionally, over the use of Al. V forms compounds with other low atomic-number elements such Al, as Ga, and and Si, and temperatures. critical fields spin splitting of the comparisons as well as have very in these compounds quantitative for high observe the Attempts to density of states useful very be would compounds these single film for possible uses in proximity-effect structures. Thinner Films Going to thinner films the advantages of acts to from from 40 A to 35 to an increase of 3.4 T to occurred reducing the smear the density of thickness Mg leads in the proximity sandwiches offers states. the diamagnetic which A reduction of the Al A in an Al sandwich with 25 A of the 0.40K sandwich critical field well above 3.8 T. This despite temperature in orbital fact that the thinner film the higher critical field sandwich was lowered from transition 2.40K to 180 1.85 0 proximity-effect for thickness this the n- gained from studying both to be sandwiches. and the s-sides of (or else identical) sandwiches. the same In minimizing to make it is London the because proportional to the n-side the s-side thinner higher The superconducting electrons. s-metal results in a smaller of root is the simply inversely is depth penetration square This thinner. orbital of the effects the deleterious diamagnetism it is more important to make the than just of information be a significant amount Further there would work continuity and for electrical minimum thickness determine the to do experiments be worthwhile to it would above films appears possible, and Preparing even thinner K. density density in of the London penetration depth on the s-side. Summarizing these suggestions we can say that there are improve a for number future tunneling work, of directions to take to the tunneling characteristics which were obtained on the first experimental samples. Better resolution in the tunneling characteristics and more systematic investigations of the influence of should make possible various parameters a quantitative of the sandwiches comparison between 181 theory and experiment. OF THE MAGNETIC SUSCEPTIBILITY MEASUREMENTS C. of were also predicted to have some proximity effect sandwiches Both sides unusual features. strikingly magnetic properties that the Chapter IV saw in We strongly have a field dependent susceptibility and the n-side susceptibility which is several times larger actually can approach a value the corresponding than should sandwich S2UID a with The spin magnetometer. on the information obtained nearly as sensitive very normal susceptibility of each side separately with the separate in for the susceptibility as that would be very in the n-side One sandwich susceptibility nuclei detailed experiment*, but it susceptibility or measured be can measurements not the total be able to measure directly state susceptibility. normal given Knight films. metal of the shift The measurement is by interesting a tunneling to observe a of a proximity-effect sandwich potentials is small dependence of the pair *When the field at a susceptibility the temperature, zero at are and we the to proportional just is IV.11 Egn. by field H as given measurement a Thus pH. of energy an at states density of point in the at a field H gives one of the susceptibility at measurement tunneling a whereas curve, states of density curve. whole the yields strength field the same 182 the in sandwich is the when actually greater is which as the normal state superconducting state than when it is in predicted in Figure IV.8. The measurement a response proximity-effect with the this layers of proximity effect difficulty, one can deposit many sandwiches overcome To itself. sandwich of response the overwhelms which a proximity sandwich itself, may much greater volume than the give having support, This films. for the thin the to have a is necessary sandwich may be difficult because it rigid support of susceptibility total of the insulating nonmagnetic, intervening thin layers. arise when measuring the Knight This difficulty does not shift. In a by the a sample nuclei in when they angular momentum eigenstate to another. levels involved are given by the values character near the nuclear cores) frequency. the and When the nucleus is in this of the turn by one magnetic The (with s-like affects the local field at influences electrons are polarized, changed the The energies of the conduction electrons spin paramagnetism of the nuclei go from the nuclear magnetic moments. field at the nuclei times the measures of the resonant absorption of frequency in a magnetic field energy one experiment Knight shift an amount the resonant the field at proportional to the 183 separately measure the shifts of the the nnot and s-side nuclei, obstruct the has resonant frequencies of the presence of the substrate does side of the sandwich, one each side on the magnetization of a handle resonant measuring the By measurement. the nuclei on each frequency of separately. to field the magnetization with respect By differentiating can one Because magnetization. the to proportional is also resonant frequency The shift of the magnetization. for the electrons one can obtain the susceptibility on each side of the proximity-effect sandwich. Both of be interesting see to if the Knight shift beyond the normal As the for feasible these measurements appear choice optimal considerations apply for Knight predicted and it would increase of the state value can be observed. materials, of same the as were shift experiments given above for tunneling experiments. D. SUMMARY In summary it appears and susceptibility sandwiches are that both measurements on feasible and tunneling measurements these proximity interesting. effect Preliminary tunneling experiments were encouraging though it appears sample can be optimized considerably by suitable the choice of 184 the materials and the thicknesses of the films the measurements at which temperature and by lowering were taken. A systematic study of the conductance into both the n-side and the s-side of otherwise identical valuable. It interface between the sandwiches would be quite two the oxidize slightly possible to should be metals and in exert this way additional control over the strength of the proximity-effect coupling the in quantitatively sandwich. investigate Chapters III and IV. The basic features these If better theory needed to given in was resolved and more systematic experimental data becomes available, it will be necessary to add the spin-orbit effects of interaction the of orbital the and diamagnetism electrons for a precise quantitative comparison between theory and experiment. can be done in a straight forward manner. the This 185 A. SUPERCONDUCTORS IN TIMES SPIN-ORBIT SCATTERING CHAPTER VI: INTRODUCTION limited paramagnetically sufficient to extent the In properties. sandwiches to proximity-effect a predict we zero-bias properties, the predicted conductance for tunneling into the a sandwich. We suggested a number experimental situation so to as peak in these higher field weaker superconductor usefulness increase the and we the about indeed produce do characteristics parallel fields, some it addition effects thin states. the films in sandwiches theory. One is present to a effect is high of the small extent in to broaden the the density in necessary to consider the effect of spin-orbit spin-splitting of cause a mixing the and which acts Another is these will no doubt be orbital diamagnetism which these of If these information detailed more of also suggested a some other experiments which might be attempted. experiments of of ways of modifying the theory, the results for comparison to the is of for one from tunneling experiments some evidence that there saw a interesting of number chapter preceding for model tunneling the developed have We density of scattering on states. This will of the up- down-spin states and a decrease in 186 the extent of the splitting. Adding these effects to the theory so far developed is an which the experiments are we of such an on the theory here, to understood poorly some Pauli isolated in spin-orbit scattering the of aspects pursued and the success concentrate instead will extent to add these effects Rather than effort. depends on the a usefulness which exercise with paramagnetically limited superconductors. We simply estimating the magnitude develop arguments for of the spin-orbit scattering time from impurities, including interpretation of these experiments, the in factors With elements. and factors agreement becomes a trace this many even so, and, neglected of between We then are scattering matrix elements, estimate numerical fortuitous experiment93 estimation of to the of as We how spin-reverse the spin-orbit and a simple hypothesis many displaced surface atoms act matrix magnitude. orders of contributing use our the of theory and an incorrect discrepancy to numerical scattering apparently an discrepancy of several surface atoms scattering. inclusion the screening, recently noted this screening has been spin-orbit proper an conventional the some missing there are estimated the In electrons. core its by impurity charge of the nuclear screening of of the the effect "impurities," of how to compare 187 theoretically predicted experimentally spin-orbit magnitude agreement which is as expected in the theory order of arguments can be this chapter (Section B) aim of putting this interaction into context. C and D, we give arguments which proper numerical factors in the estimate of the in Sections Following this, lead to the we review spin-orbit interaction of superconductivity with the the current understanding of spin-orbit scattering times arising from impurity scattering We further include the effects of the screening of centers. of the nuclear charge electrons. scattering the Tedrow to our the data data to an improper estimate of contributing to results of the number in reasonable of and the We trace the tabulated the number of surface the spin-orbit scattering counting and experiment. theory and our improved (Section E) by Meservey improved theoretical estimate. discrepancy between potentials tabulated core by its center In the final section of this chapter then compare proper an good as our the inclusion of the the history of we to to give. In the next section of the times find We ones. determined scattering and show surface atoms that a scattering agreement between theory 188 IN INTERACTION SPIN-ORBIT THE ON BACKGROUND B. SUPERCONDUCTORS role of the spin-orbit Understanding the interaction in superconductors has proven to be a long and evasive process. developed, Yosida 3 8 calculated and found it to be reflecting conducting ground Sn,9' which was and A19 scattering all out impurities Ferrell"8 Anderson and a theory excitation in Knight shift spin spin-orbit that and Abrikosov found indeed and 99 but Hg, the Al a random non-vanishing by a spin-flip lifetime, spin Ta,. non-vanishing Knight shift in results remained a mystery. Al results might be explained suggested that the surfaces and Gor'kovI 0 0 scattering from of spin-orbit by spin-flip scattering from at the super- nonvanishing a suggested This presumably explained the Matthias 1 01 a nonvanishing reflect susceptibility characterized and the at surfaces was responsible for this nonvanishing susceptibility Sn the in gap of function of Knight shift experiments' 4 showed to thought susceptibility. worked and the state 7 pair nature bound the On the other hand, Hg,9 5 spin susceptibility an exponentially vanishing temperature spectrum. Pauli the was superconductivity of theory BCS the after Soon paramagnetic of the Al samples, later proved to be unnecessary. oxygen impurities though this explanation 189 Ferrell 1 0 2 to the Van Vleck37 non-vanishing susceptibility might be due be present in addition to orbital paramagnetism which should the is and states Appel 1 superconducting. Sn, normal Hg, and It . in this contribution 4 estimated order needed the of small too much was but 3 becomes metal a when Sn data, bands 1 0 non-s for Al and found it was and the Hg to explain 0 to high spin) electrons important unaffected remain should paramagnetism arises orbital (both of transitions from virtual energy This paramagnetism. spin measured the that suggestion the made to that of Al. explain Later the Knight shift Al repeated 1 experiments were 106 the be vanishing as was found to Os the and temperature approached zero in agreement with the original expectations. standing long The Anderson 1 07 resolved by finally it, was has called Knockabout," Shift Knight "Great as this combination of refined theory and experiment. Spin-orbit scattering superconductors. interaction, the In in high also important is absence the difference between the of the superconducting state susceptibilities results field where the magnetic superconducting and normal energy states difference equals the field spin-orbit normal and in a critical between the superconducting 190 18.4 kilogauss paramagnetically limiting field of or Pauli per 0 is the Chandrasekhar-Clogston4 (This condensation energy. the spin into the superconducting from impurities is only spin-orbit scattering paramagnetism, which resolved the The orbital metals, in transition discrepancies the in superconductors. of high field theory of the critical field the spin scattering impurity spin-orbit and paramagnetism Knight shift in not important is it which effects only include Thus one need this susceptibility. Furthermore, critical field. paramagnetically limiting the alone determines it state, transition upon the susceptibility changes only Since transition temperature.) of the degree Kelvin determining the critical field. Hohenberg 16 lifetime ls, spin-orbit . Maki 1" by superconductors characterizes in scattering is spin-orbit scattering dramatic, as in Pb 0 Mo5 this smaller, but even a where of say a the effect of 10 percent 120 kilogauss also. the effects limiting field is exceeded by over less extreme cases, case II and spin-orbit of the When the effects the of all large, S6 , the parameter, of type Helfand, Werthamer, and single A theory in the included have been impurities from scattering spin-orbit and paramagnetism Pauli quite can be the Pauli paramagnetically a factor spin-orbit change material, is of three 13 . scattering In is in the critical field extremely significant 191 we In this paper means of effecting such a change. lead to a the reexamine carefully will an would spin-orbit scattering role of of the understanding that hope might One applications. practical for of theory impurity spin-orbit with this in mind. Gor'kov1 0 0 provides the Abrikosov and including As their microscopic theory of superconductivity. was in the effects of spin-orbit scatterers origin interaction. spin-orbit Z the on commented parenthetically structure and constant (m Z) 2 , where of the of the the fine is c. of number atomic is the Z the in Since impurity. as scale to ratio the merely to regular sc attering spin-orbit scattering matrix elements matrix elements no t in the and dependence expected They the interest and Gor'kov Abrikosov scattering, of this in impurities of way of most natural from scattering spin-orbit technique Green's-function impurity-averaging The impurity-averaging the t echnique relevant scattering diagrams involve two scattering from the same impurity ratio of potential, transport scattering time, Experimenters indeed the first field data from TC5, scattering time, that as scales set out to thin 'r,,, 08 films , to spin-orbit (<C.Z)". measure this experiments1 into a translates comment their Z dependence and using parallel agreed well critical with the 192 result. this a much slower dependence on Z than T, to normalized is T, when of Z4 corroborate did not These experimenters found that experiments by and critical field Tedrow 2 o and Meservey Later constant of unity. dependence with a proportionality spin-polarized tunneling (04Z)4 an in fact finding prediction, Abrikosov-Gor'kov the transport scattering time. Very recently Meservey and Tedrow data assuming time. impurity scattering scale, log-log of role plays the time thin films that in many over 93 have the boundary scattering (i.e. transport) the regular agreement, on now find They reanalyzed their Here of magnitude. orders a we consider reconsider the calculation of the scattering matrix elements which enter into the Abrikosov-Gor'kov theory. drastic screening of the inclusion of reductions ruins and potentials Meservey and Tedrow. surface scatterers. contribution this brings cores results ion apparent agreement the noted effective concentration reestimate data into by agreement to an We trace the lack of Our in scattering the of strength the of the improper estimation of the Our of the of the surface order-of-magnitude agreement with the corrected theory. In this chapter theory which have we examine a number of points never been clearly stated: in the Exactly what 193 is predominantly from actual impurities which have a different the Given the the scattering correct identity of spin-orbit scattering to experiment? future the impurity What directions should take? experiments Because what center, theory of compare the should one How potentials? the scattering matrix elements are reasonable estimates for involved? lattice the host as number same atomic centers with resulting in scattering imperfections from Is it atomic number than the ions of the host lattice? lattice Is it impurity spin-orbit scattering? the origin of the behind arguments Abrikosov-Gor'kov the atomic-number dependence prediction have never been formally an the constant. dependence Z the Z 4 effect electrons atomic to give not proportionality the also but matrix prediction attributed The arguments are readily extended to include the effect of core electron screening the of magnitude Furthermore the arguments Gor'kov. Abrikosov and only to leads elements, which the on based argument, includes Section C, given, the next section of this chapter, of delocalized the compared to their of the nuclear charge and nature metallic the counterparts. atomic effects are considered in Section D of These where they are shown to lead to a much more erratic and overall slower dependence of T / TsO than simply Z 4 . prediction with In the Section E we confront the new the experimental data. We show that when 194 the correction is screening and experiment between theory and Tedrow becomes magnitude. We normalize the argue normalizing boundary a that spin-orbit scattering time. of properly We the scattering which discrepancy Meservey scattering was noted of many and time and by Meservey orders of Tedrow incorrectly time to then describe our new spin-orbit agreement the included, scattering demonstrate the and boundary simpler way times an magnitude agreement between theory and experiment. to order the of 195 ELEMENTS MATRIX C. IMPURITY SCATTERING clarify exactly 1. The what Schrodinger the are. "impurities" The Impurity Scattering Hamiltonian nonrelativistic limit of electron in We also matrix elements. Hamiltonian and then estimate its attempt to impurity scattering down the we write this section In AND REGULAR FROM SPIN-ORBIT an external scalar equation with an an the Dirac equation for an potential V results in the potential U given equivalent by 1 09 V The the first term,V(r), is second the from resulting Eq. IV.1 interaction magnetic of field one the gets limit Other terms the Dirac of spin-independent corrections Except of importance to us. is what interaction. nonrelativistic in small equation result are not the regular scattering potential and the spin-orbit is VI. 1) A for a factor classically electron's resulting from magnetic a Lorentz by of two, considering moment and with transformation the the of 196 the electric field through which the electron is moving. seen by a single electron moving The total potential V through an array of potentials like Eq. lattice ions VI.1 centered on each of the ion sites: - R) In a a sum of is then given by (VI.2) may V potential real solid the total be separated into a periodic part and an impurity part: V d VostbI Vto. The first over sum is R .v-R)(VI.3) + over all lattice all impurity sites. a periodic host lattice impurity This equation defines the potential as the difference between potential. second the vectors and the total potential and This difference clearly contains negative host ion potentials at vacancy sites, host ion potentials potentials potentials. destroy at at interstitials, and dislocations (We are the impurity assumption to cause better justification host ion "real" impurity as well assuming that the dislocations the long range order so periodic potential from as a series of that the subtraction of the eliminates most potential. We of the do total potential not procedure can be this point we also implicitly assume that the same can account for surfaces and grain boundaries.) this expect serious difficulties and expect for this do not that a given. At procedure The reason 197 for separating periodic the potential potential in and impurity the different physical effects. Bloch states, the manner is that potential The impurity and effective the lead The periodic potential effective masses, usual way. this to leads to g-factors potential causes in scattering between Bloch states. With this picture in mind What is the it dominant source predominantly from that "real" one can answer the question: of the impurity scattering? impurities of the host metal? with potentials Or is it with the same potential as the host? be that unless apparent impurity doping, it there is the that impurity are lattice. In same as the amount host of a series potentials particular the "impurity" given by that of the host lattice. in of the of lattice "impurity" scattering. Hamiltonian consists the from from It should a substantial is scattering potentials which dominates the different predominantly "impurities" Is Thus potentials periodic atomic number Z is 198 impurity treatment of takes conventionally uses these the scattering spin-orbit regular the and in metals one If one states. impurity matrix the it is clear that impurity scattering VI.1, the potential in Eq. elements from the Green's In electron basis free matrix potential states. states and calculates basis Elements Matrix impurity basis must chose our elements, we function the can estimate Before we both of Scattering Estimation 2. matrix elements will be proportional to the atomic number of the center. scattering down by a factor that is element spin-orbit matrix a to This leads -5 (V the number. atomic ratio transport ten orders and, failure cores, time, where in as find and can be of independent This traced to the spin-orbit that of magnitude result the the is fact the spin-orbit longer than the atomic host the type impurity of fact, independent well. estimate result to this square times scattering concentration The We of scattering scattering time is number independent of regular scattering matrix element, from the and nonsense. clearly that near the ion strongest, the interaction is free electron wave functions are a poor approximation to the real wave basis functions in underestimates the solid. the Clearly electron a free density electron near the 199 positively charged core and gives too rg . large a value of wave function accurately it is To estimate this part of the more reasonable to go to the opposite extreme and start from a tight-binding basis. r In ket) wave tight-binding our normalized take We functions as: k* J re where n of number atoms function, and the sum over R is taking the we are that all integrals wave functions from q) , the solid in index, Ow, is a spin band index, is the is an the is atomic wave over all ion core sites. limit, we extreme tigh t-binding except thos e the same N involving atomic assume and potentials site vanish. As Using these basis states the second qu antized form of the impurity Hamiltonian in Eq. VI.3 is: Y%'k**7- (V C where The 1K o (VI.4) -t- er'Kh spin-orbit part of this matrix element can be written so terms of the corresponding atomic matrix elements La. 4 t'as: in 200 I ' so Li EquationVI.4 (VI.5a) 14,40C, of exactly is the point of the Abrikosov-Gor'kov theory the basis and atomic matrix elements, use wave their of 1/N should of a unit atomic cell so that: / 50 S "(VI.5b) is identical in to form Abrikosov-Gor'kov theory, we of the defined Abrikosov and splittings. normalization be replaced by the volume v VI.4 are in a well functions, in which case the normalization factor Since Eq. the starting point their results can use directly with our interpretation of the matrix elements. obtain transport and spin-flip scattering times Ir I; N(Yj) NI M F where n density represent is in the same as those that appear in the continuum a are the matrix elements calculation of atomic spin-orbit Gor'kov free electrons for Here however we presence of random impurities. tight-binding the starting same form as the "impurity" of states averages at of the level, appropriate given by (VI.6) A(VI.7) concentration, the Fermi We N(E ) and the scattering is the brackets matrix 201 time can be measured in resistivity experiments whereas the be measured in spin-polarized spin-flip scattering time can tunnelingz 0 measurements of the determined from theoretical estimates our of these can be temperature dependence in thin films. field parallel critical of the In addition both times experiments. 92 transport scattering The the Fermi surface. elements over times We compare experimental to values in Section E*. We leads conclude this section by showing how a simple argument to the dependence ((Z)4 to Abrikosov and Gor'kov. We of 't/'Cso p electron relevant ones. charge Z, the VI.6 and VI.7. and spin-orbit matrix elements are the p electrons For attributed use hydrogenic wave functions to estimate the matrix elements in Eqns. note that the which is (radial part of the) in an atom with nuclear spin-orbit matrix element is l I L (VI.8) 3 the principle quantum number and ao where n is radius. e& On the other hand, the is the Bohr of the matrix element *The experiments are done on thin films and, as Meservey and determined resistivity the point out, correctly Tedrow unknown the eliminate not properly time does scattering the normalization of The proper impurity concentration. regime scattering surface in the "impurity" concentration is discussed more in Section E. 202 regular Coulomb potential is: Using these and taking - -(VI.9) a.o v1 the ratio of Eq. VI.7 to VI.6 we find: --- where n is the in question. principle quantum number This result, (3~ (VI.10) 340M %, 5S of the with the coefficient i. effectively set to unity has to be the been interpreted Abrikosov-Gor'kov prediction for the Z dependence This result modifications is simplistic at section. best. of core electron screening charge and of delocalization of next p electrons We of 't/T consider the out of the nuclear conduction electrons in the 203 D. ELECTRON CORE SCREENING AND METALLIC ION DELOCALIZATION this In section we ion of effect their atomic counterparts. of "pulling out" the interested this make in order complication. for splittings neglected and (In and primarily are we we will fact a later approximation numbers atomic energy levels acts ignore will from spin-orbit this to cancel delocalization effect.) has Yafet effects of magnitude less in spin-orbit the paper we In this effective quantum the resulting where core the ion strongest. interaction is be seen to result from This can the electron orbitals density near electron % from 25 to 40 about reduces spin-orbit matrix elements by typically solids in delocalization electron The charge. nuclear of the screening out core the by taking into account naive Abrikosov-Gor'kov Z dependence the correct data to use atomic will quoted most of the simple metals. state valence atomic One A5 0 for splittings can fit the expressions for each row of the periodic table with ) where ,,P is the orbital M atomic number . 11) (here . = 1), o is the 204 The values number. we simply take this paper for p effective quantum number The the approximations 2 and consistent with is always near in are White 1 1 2 by those given given in are here quantum numbers Rb. Na, and in Li, effective quantum and neff so determined agreement with reasonable electrons r effective The VI.1. Table of the nefi is and core screening parameter, be its this to in Thus we value. take: 4 y_=_.__. (VI. matrix element in Eq. for the value of the spin-orbit have dropped (Note we the which enters into the IV.11 in the (I + final ( + to 1/2) in Eq. factor (j + 3/2) IV.6. splitting into the calculation case but does not enter atomic 1/2) 12) in the solid.) The regular scattering matrix elements by atomic hydrogenic matrix elements from are approximated screened nuclei. In the tight-binding approximation we get: (V. 'e~(~-O) (If we had used scattering part, different from Eq. a free electron our numerical VI.13.) basis for the results would We then regular not be obtain for our 13) much ratio of transport to spin-orbit scattering time: I TSO 36 ot'(2- tr) soY Ve; '(VI.14) 205 Spectral Data Parameters from Atomic Table VI.1: Lb) (C .) Z ________________ n ne~f 'I I-U I nef f 3 7 1. 92 2 2 84 1 6 11 15 7. 69 3 2 76 2 29 33 23. 45 4 2 05 47 51 41. 25 5 1 68 79 83 68. 8 6 1 93 from spin-orbit splitting (a) Estimated (b) Taken from Ref. 112. 12 2.29 tabulated in Ref. 111. 206 This result chapter. is discussed in the next section of this 207 E. COMPARISON OF TO TIMES ESTIMATED EXPERIMENT 1. Comparison of the the Square A plot of the of Eq. Meservey and field Bhatnagar, 1 0 8 92 Time Scattering (H ) given in Figure VI.1 data own their Knight shift conduction electron normal metals). 1 1 5 of 1 7 93 of Crow, Ratio of data, 1 1 3 (KS) Meservey 11 along with the parallel Strongin, spin-polarized and (SPT) tunneling 4 and data from (CESR) experiments spin resonance -1 the Potential Scattering VI.14 is Ratio to Tedrow's recent compilation critical data,ZO SCATTERING SPIN-ORBIT and Tedrow have (in normalized the spin-orbit scattering time to the surface collision time T given by L. L (VI.15) F where L is velocity. the thickness The data has a of the sample and v, scatter of orders of magnitude, but falls which has been interpreted to that ranges roughly on be the is the over Fermi a couple the line (K Z)4 Abrikosov-Gor'kov prediction. Our result, Eq. VI.14, can be seen to display a much more 208 0 10 -2 10 -4 10 T so -6 10 -8 10 -\0 10 2 Figure VI.1 5 10 20 50 Atomic Number z 100 209 erratic behavior with increasing atomic number. This reflection of the closing of atomic shells and the is a resulting abrupt jumps in the spin-orbit scattering potentials seen by outer electrons. The striking feature however is not this erratic behavior of this figure but rather the size of the discrepancy between theory and experiment when screening is included It and the is unusual theoretical magnitude numerical coefficients an experimental to call prediction which extends are estimated. verification of seven orders a of an error in favor of a theory which disagrees with the data by as be emphasized, magnitude. It should the considerations presented much as six orders of however, that above are quite simple and in accord with what is well known about the spin-orbit interaction in other branches of atomic and solid state physics. If we accept the validity of the theoretical estimates of the scattering times, the other in the normalization of the the surface scattering possible source of error is spin-orbit scattering times. Indeed, it is times to here where we find the source of the disagreement between our more careful theoretical estimates and the data in Figure VI.1. When the manner of experimental data is Figure VI.1, compared to theory in the it is assumed that all of the regular and the spin-orbit scattering is governed by the scattering 210 off the ion core potentials is further assumed results in a surface that these proportional and inversely of impurity scatterers that correct is it is spin-orbit scattering scattering. regular potential assume to governed However, we believe squares of VI.14. We as given in Eq. the corresponding matrix elements This is independent of ratio of the to the scatter bulk metals. scattering time ratio which the presumably unknown concentration argue impurities impurities in strength as with the same It of displaced surface atoms. that the surface spin-orbit by potential it is incorrect to say that scattering has anything to do with the boundary scattering time. The difference can be understood as follows. electron incident on an interface. Independent of what the nature of the scattering potentials is at electron backwards. a has unit Spin-orbit occur, however. The same contribution as probability scattering Consider an of does therinterface, the reflected being not necessarily displaced surface atoms will make the always to spin-orbit scattering whereas the boundary scattering occurs independent the impurity scattering potentials. of the nature of 211 2. Estimating the Contribution of Surface Scatterers what now Consider potential is scattering "impurity" estimate the to used regular the when doing is one Equation concentration when surface scattering is dominant. VI.15 gives the surface scattering time as the time it takes an electron to travel at the Fermi velocity from one side of hits the surface the other. the metal to be reflected in some it will certainly we say When an electron that the scattering such a probability for the necessary amount necessarily overestimate of the number of When this overestimate is then Eq. used in of VI.9, we will VI.6 potentials of surface the number Eq. we use Then if much get something concentration of scattering calculate the would give and we integrate the atoms scattering, we unit probability. less than only to surface is due at the scattering off displaced surface However if manner. to which scattering, we surface scatterers. surface scatterers spin-orbit estimate a scattering time which is much smaller than it should be. The correct way of estimating the effective concentration of impurity scattering centers when the scattering displaced surface atoms of encounters relative to a is merely electron will the times to estimate with have it encounters an is due to the fraction surface ion core atoms from the 212 lattice*. In tunneling experiments the tunneling electrons move primarily perpendicular to the plane of oxide barrier, and the fractional number of encounters will have with displaced merely given by sample surface scattering units a film of which is will the lattice 50 atomic tunneling electrons "impurity" scattering potential about 2 * proper be under "concentration" of impurity then given by constant. layers the the fractional (0.02 in the example). Table we theoretically. We calculated of a The scattering potentials is compare spin-orbit scattering rates with the of the time. per unit volume number of encounters with VI.2 For thick, influence the number of host ion cores times In potentials is 1 atom (say) divided by the thickness of the measured in example, in a tunneling electron the experimentally surface measured scattering rates predicted the theoretical rates by assuming the fractional concentration of scatters was indeed given by measured 1 in over the lattice matrix elements to spin-orbit experimentally given film constants. took our be those splittings. We tabulated by (This was the thickness spin-orbit Yafet from atomic source of our *Note this method of estimating the surface contribution to the "impurity" spin-orbit scattering time gives an estimate which is independent of the magnitude of the regular scattering potential of displaced surface ion cores. w w Table VI.2: Element Atomic No. Measured and Predicted Spin-Orbit Scattering Times Method Size Measured (meV) Al 11 13 Ref. so (1) Na Predicted (meV) CESR 6000 (sphere) 4.6 -10-6 1.2 -10-6 115 CESR 700 (sphere) 3.4-10-5 1.1-10-5 115 SPT 50 (film) 0.063 0.010 H 50 (film) 0.15 0.010. 108 20 CESR 16500 (foil) 2 -10-4 3 -10-5 116 CESR 8000 (foil) 5.2 -10-4 6 -10-5 116 100 (film) 0.9 0.126 92 Ga 31 SPT Sn 50 H 38 (film) 100. 26. 108 H 63 (film) 50. 16. 108 5.8 113 7.0 3.3 113 (elpsd) 5.8 2.5 113 (elpsd) 3.6 1.7 113 KS 170 (elpsd) 12.4 KS 300 (elpsd) KS 400 KS 580 214 18 plotted by the data only points obtained from Tedrow in Figure VI.1 Meservey and Fermi level get the to selected from We states. density of Kittel data from specific heat used We above.) given constants screening empirical thin samples in which surface scattering is dominant. Furthermore spin-orbit predicted our one It can be estimate that atom effectively contributes as an impurity scatterer on reasonable displaced atoms to that assume have elements which an it is probably In fact there is are five or so surface scattering to assume that there is only atom which contributes. Conclusion and Suggested Further We the electrons which contribute to the on each transversal than it is one displaced by factor of five. underestimate by a more sample of the transversal each 3. that our to mean quantitatively interpreted rates scattering encouraging result. actually an This is values. measured of five below the measured consistently fall about a factor values. the with of magnitude order in agree we estimate all of the times first of all that We note estimated enter spin-orbit scattering the into Experimental Work spin-orbit the scattering Abrikosov-Gor kov in superconductors. We matrix theory of included all 215 numerical coefficients as of the have We displaced surface atoms from in superconductors and have achieved experimental agreement with good estimate to scattering spin-orbit contribution of the how indicated further properly reasonably previously been interaction than had spin-orbit scattering supposed. a much smaller These considerations lead to the ion core. screening of inner shells on by the electrons in the charge the nuclear well as the determinations of spin-orbit scattering times. is the The recognition that surface spin-orbit scattering spin-orbit scattering times dominant contribution to the spin-polarized tunneling experiments was by and Meservey or whether not mechanism is to thickness, but scattering this is indeed do experiments to spin-polarized scale with vary the tunneling scattering series of on a films. The would be demonstrate a factor of possible, and because two variation of the size dependence. would It significantly sizes experiments this surface thickness. requirement that the orbital diamagnetism be small. appear that different if the spin-orbit see the sample film verifying for dominant the identical otherwise only given recently test crucial mechanism will be verified scattering times difficult A Tedrow. in be of is in the It does sample thickness sufficient to 216 One should predicting samples be able spin-orbit that are to scattering intentionally concentration of impurity. all of the information spin-orbit scattering Preliminary doping Tedrow, and Bruno by doped working with a of with known case one knows directly that is needed to time according estimate to Eq. VI.7. were never Such a direct quantitative scattering times the by Meservey, a few years ago but the results predicted spin-orbit interesting. In this times our method experiments were attempted analyzed quantitatively. the verify directly would be test of quite 217 We fields. magnetic in richer than the properties which properties either metal of each other via the proximity effect include states the spin-density.of films influence the splitting of the to the extent that peaks in the sandwich side of states on the densities of the spin New film alone. two the only appear when coupled considerably are limit Pauli paramagnetic the films two of properties the that saw parallel in sandwiches proximity-effect of thin behavior to predict the thesis we used a tunneling.model this In CONCLUSION AND SUMMARY CHAPTER VII: which is the weaker superconductor can actually cross at the Fermi level. enough great states to Once the spin splitting caused by the field is the for cross, of energies the range, coupling between there will be density of states peaks For this strength This split densities can result a certain peak in the total the addition of of coupling, the rapidly of states to cross in the separated the down-spin densities of states. superconducting sandwich grows the of falls within a sharp zero-bias in the up-spin and sides a net magnetization. acquires the films reflecting down-spin BCS pair potentials start the sandwich, the self-consistent If and depairing in both there is some to decrease, and the sandwich up- some a magnetization of once at the field the causes the Fermi level. superconducting-state spin 218 of the susceptibility which exceeds the Pauli susceptibility normal state. the results of the Preliminary experiments stimulated by showed a clear peak in calculations described in this thesis just below the sandwich the zero-bias conductance at fields the normal metal side of critical fields for tunneling into a was not experiments the conductance structure in a experiments might we Furthermore measurements would can be it will complications caused by severely We smearing. tunneling the be modified to increase the resolution. suggested that some susceptibility be quite interesting. experiments If these ways data of the observed characteristics was feasible of number amount of orbital diamagnetic limited by thermal and by suggested the resolution limited and taken was quite the possible because the of analysis quantitative A sandwich. proximity be are as fruitful as it appears they to the theory the add necessary to of orbital spin-orbit deparing and of the off scattering spin-mixing impurities and surfaces. Although we did not include our theory of the this spin-orbit scattering magnetic field proximity-effect sandwiches, we gave behavior some estimates of in the for the 219 impurities. from had surface atoms we found experiment. tested films the number of td the measured spin-orbit When we corrected for both of these errors reasonable of for surface matrix scattering by elements can of our measuring the The be measured Quantitative these lines would be quite interesting. and times can be correctness scattering. of this doped with impurities. scattering the the theory between agreement Our explanation of the thickness dependence the spin-orbit The analysis rates. also previously misestimated experimentally normalization nuclear of an impurity ion drastically which contribute scattering time. the screening of predicted spin-scattering modifies the data that saw We core electrons charge by the of the matrix elements dependence of the scattering atomic number strength of directly experiments in along 220 APPENDIX A: SOME RESULTS FROM THE GENES-EQUATION APPROACH TO THE In Chapter methods of treating out that it Gennes 3 7 is In the proximity the to in each half for a appendix we equations various sandwich with the pair a by others 5 7, of the sandwich be equal. case where equal, we of the show that the with an energy gaps of metals which only of a sandwich.) has a We Our 72 Fermi wave is (2) the metals on the things: of the is an (1) For the are two metals of states states are average of derivations7 1, discussed vector is then use the derived (Other In result because we do The densities of the two metals. have sharp function for zero-temperature densities energy gap limit Cooper temperature the wave vectors two metals are equal. BCS-like the to demonstrate two the Fermi a potential thin sandwich. not require that the Fermi wave vectors of Green's function pointed in very thin sandwiches construct the Green's describing We sandwich. a generalization of that given two sides theoretical solve the Bogoliubov-de thin assume that of the EFFECT effect sandwiches. This is possible because this BdG discussed equations justified constant PROXIMITY is possible to exactly (BdG) interface. it II we BOGOLIUBOV-DE the the 72 of transition the limit where one of which very much exceeds 221 equivalent to sandwich are of the outside boundaries at the Green's functions evaluated the other, the that of Thus Green's functions of the corresponding isolated films. in of average no show and differently sides two the limit, this behave sandwich the thus We properties. the demonstrate that the difference in the Fermi wave vector acts to invalidate of averaging the otherwise occur in properties would superconducting the the Cooper limit. In Chapter II, this effect of a difference in Fermi wave vector was in justify to used Hamiltonian model. the part use of tunneling the in the In the tunneling model, the films of their own properties when sandwich each also retain some the coupling via the tunneling Hamiltonian is weak. the some It should be mentioned that .results derived here, function for a sandwich comprised of metals Fermi wave vectors, are quite published. Here we results fully. the double as such of the intermediate by no means evaluate Green's with differing general and have In particular, we potentials are the layer never been and exploit these assume that the given pair self-consistent ones and do not discuss the achievement of this self-consistency. As in Chapter II, superconducting x = -ds and metal with x = 0, and a sandwich consisting we study a pair a weaker potential AS of a between superconductor with pair 222 potential A, x = d,. between x = 0 and We rewrite the BdG equations given in Eq. II.8 as E-4xx] 3A(,X) 1)*4) (A. 1) where )~2Wx We k (A.2) the Fourier transformed have =(O,k ,k 1AX i The Green's C ( , A.1 )= x E, )X(X ') are appropriate for conditions which x = -ds and to the satisfies function also satisfies the at to be Green's function corresponding wave functions given by Eq. - dependence -(A.3) The zero-temperature insulator Y to ) and have defined and z y a free (A.4) following boundary an interface surface with an at x= dn, 0G( (A.5) respectively kx) X',E) ---G,(C -, = xx, E X=-Ss = I=$ E)= Following and Green's theorem Feuchtwang? 8 to construct the and (A.6) Arnold Green's 57 , we function can use for the 223 entire sandwich from the Green 's sandwich the Green's the of and s-sides n- of absence side. the other These v l i x,x',E) and G0 ( denoted G" (k it ,x,x',E) functions, the for in the functions for each side of respectively, sandwich, satisfy -E (x)] along with )= 0k X Xj )= ) - x, x', ( the boundary condition A.5 x,x <O X), c > (A.8) G:, for (A.7) A.6 for , G and additionally, at x = 0, 0 II Cx X x=o x'450 x':0 X1o G"(kt)x;e (A. 0x 10) single films, functions for the of these Green's In terms (A.9) 78 the Green's function for the sandwich is given by G(x,x',E) (A .11) G(0(X)xX'E) we k -rP Y. ,- E) = X, X'E) - G X)X' < 0 G(0oX.[GE),(A.)1 XoE) (0,0, E) + G(S C whoyk It isolated is possible to construct 0, xxx', V1 (, 1 A(.12 o the Green's function for the metals from the eigenfunctions of the BdG equations 224 for each isolated film. 5 (XX, Er) CT The results are 5 7 -ts I E c~rK" J "IO ft +n mGe(K; Xy)(#-$(K" (yg+d) Als Kv3 4+ (A. 13 ) E f. AS~ Vet. ± A V,l * (A. 14) where K (v. .,L (A. 15) A,it) % (A. 16 ) a 'IL = Putting these find for G I Ouf~I (-ds , -d [S(! - <,(x: A 3+ 1A,) -2( I - X1) I ( is G for and G" into Eq. A. 11 (1X) (Es. -S - X SAo(Kdt,-- 'K ) ( -I CA1(KV-k-)os 4) + cod<kf4, -d (KAs+ 104) - XKi LA(K- 4A.) -x cZ(K- KN'd )3 ~,- Es S&) £L*1(i(2' kf)dwJ (A. 17) where 'D -x') (EKEs-,SS we , E) -(E04IX)4 5 + I) -Ads I +2 expressions J- 3o/ (Ks - K!)sc. - (K'-g.")k -; 2 f.O*(k!s As,+ K+"oA k)- X coq(0d,- j<4"dm)j [eOD( k x2 _IC, + .. K4. S + E ,~ J A sJ ">ts j $ jL~~) S sC 225 now evaluate We identical differing Making the this for Fermi wave vectors and with two metals limits of metals with vastly A.17 in the two Fermi wave vectors. of Eq. a partial fraction expansion limit where the two films comprising the sandwich are thin and have the same Fermi wave vector, we find k AS /s +A18) where tege*+B . ± + Tr ___ The local density imaginary part of of states __ (___ at (A.20) _( -d x given by is this when integrated over k BCS density of states with an average (A.19) . the We find a gap given by Ais+ If we interchange the n- and for the n-side density s-sides we get the same result of states as this result for the 226 s-side density of where thin sandwich and s-side vectors, the n- the same BCS shape with the the case take We now densities of states both display same average gap. We first assume X- -1. Then A.17 becomes V )- GAs 9 ( +0)C4Pt Ods which is equivalent to the s-side film, not greatly metals have the where different Fermi wave vectors. Eq. Fermi wave the same metals have both limit for a Thus in the Cooper states*. be different Eq. A.13. Fermi wave Green's function for an isolated at X =-d . evaluated surprising because ( A .21) cAt(K.E )I -) in vectors, the limit of there is This should such almost vastly total states directly n-side density of not evaluate the *We do in an identical at X = d from the local density of states the s-side density of states. manner to our calculation of A.6 imposed on the boundary condition Eq. This is because leads to a vanishing density the Green's function at X = d 8 indicated how one is to has Feuchtwang7 of states there. more any avoid have we but case, this in proceed calculations by this interchange trick. 227 reflection at the interface*. In this same limit of vastly differing Fermi wave vectors, we evaluate the equivalent Green's function for the n-side by X -+ 1. interchanging n- and s- and taking the limit as We find K%(t-X) si,(-A , E=) (6,+1)-1 C-t-f(1)-to"" k+'d() k (A.22) This results in a density of states equivalent to that of an which can be isolated n-metal all the s-side counterparts. obtained from Eq. quantities The replaced difference by between A.13 with their the n-side and tangent cotangent functions can be traced to an effective difference the interface between a medium in the boundary condition at in which waves propagate with a high velocity (v,= fkF/m) and one in which they propagate with a low velocity. This difference easily pictured which occurs is physically of the partial reflection at an interface between light rope held in tension. rope are reflected inverted, condition requiring the to one equivalent more and transmission a heavy rope and a Waves incident from the lighter as if there were amplitude of the wave a physically illuminating give *Below we explains the near total reflection. a boundary to vanish at analogy which 228 the interface are reflected uninverted, heavier rope of the wave with were a as if there vanishing derivative for the boundary condition requiring a amplitude from the Waves incident between the ropes. at to position respect the interface. We have now demonstrated that one effect of a difference in Fermi wave vectors between metals in the sandwich is that each side in Each side identity. when will the tunneling qualitatively then, results which in differing obvious Hamiltonian manner) tunneling of more used. perfect interface This is, of own its own identity of least At Hamiltonian should the tunneling their superconducting the is its of more retain also retains some describe a Fermi wave vectors. description some sense give between metals properties their and course, in addition to the Hamiltonian gives complicated containing real tunneling barriers. (in interfaces a more perhaps 229 CALCULATION OF THE SANDWICH SUPERCONDUCTING APPENDIX B: TRANSITION TEMPERATURE 1.9), these 4 (E) order transition field to the r +4+ ) A when made A form given in 0 IV.1 111.41 B.1 and solve values of second we find &rift(B. to the (a discussing 1) of case a simplification the transition by setting (B.2) = 2N from Eq. Eq. these to IV.4, in the McMillan model'is 80) = Now putting IV.1 sandwich normal-superconducting temperature order. temperature. simplified have ,pk , relating the an equation to derive conventionally is second and use A;(E), for the pair potential we sandwich into parameter order of the Proceeding to linearize Eqns. where field and Chapter I (see it is possible to expand equations in powers in IV.5 FIELD proximity with increasing This being the case, through IV.4 the transition of the normal state the A FUNCTION OF high temperatures At sufficiently Figure AS into the gap equation performing the in the sum over states, 230 we find + , SQ p) * (B. 3) .J Making use of the relationship (2 e I(. N(E,) V, where ' Tc, (I is Euler's constant ( temperature of bulk transition is the .57722) and Tc ' the superconductor, and of the relationship IT 2eW) (B.5) 04 W;4C t where we are the frequencies appropriate critical temperature, T (3, c T 4s , Eq. W for the sandwich B.3 can be rendered /~(+ ce 6' e (B.6) These sums can be expressed in terms of the +(X)= by using the property 11 9 - - - ++ Finally we digamma function get ) - (B.7) 231 T ~ ~±~ ) .!-AA4 (B.8) 2,10 - cancel at where we have made the approximation that the sums their upper limits. function of temperature. field as a critical the that temperature We might order critical field" which we have all at increasing with and we have no experiments with which to compare a quantitative estimate. exists know qualitatively We decreases field critical of the calculations no numerical We present temperatures just calculated actually the whether of regardless first order the normal state is sandwich's transition into "second mention that the or second. In the case of a first order transition, such as illustrated in Figs. calculated is the parameter IV.20 goes equation to the Eq. T = 0 plots calculate H5 makes the difficult. where the supercooling comparison of We Eq. In this back can go field function formalism. I order the singularity of appropriate transformations zero-temperature Green's we find: field actually have what we Actually the zero, to IV.13, make the directly and IV.4, supercooling to zero. goes as T 1.7 to there, and using the this case I E-) /= He 2. + 'S ( B .+9 C WD (B.9) 232 is the where &s T=O, H=O order parameter of one of weak coupling. absence of 6s/(2A.) so a proximity we can see As isolated in writing the last line s-metal and the approximation made is the discussed in Chapter I, effect, the that the expected and reduces this field. in the supercooling field proximity effect acts is as 233 BIOGRAPHICAL NOTE I was There I grew Grade School, Creighton the school's chapter of the to summers I 1972 was 1974 employed to work accelerator to excite of physics and while Dr. Sam Department May a on Sciences and X-rays was Senior as I was J. for of a corecipient support the to courses instrumentation for and I of 1976 superconductivity I spent electron utilizing MIT's 400 Mev Bates summer Physics small a In analysis. cum laude from Creighton as a as a graduate student in physics. took During the using element the at Creighton, in at aimed of Students. Cipolla trace served President studying the College started under Dr. three years first graduate studies, and in September of I I and of Arts Predoctoral Fellowship from the National Science Foundation MIT, School, Award. I was awarded a at High Physics project 1974 I graduated summa major year of Catholic Lourdes University, Creighton the Society by of Our Lady school Omaha, Nebraska. in Preparatory At 1972-1973 1953, 24, attending up, University. Creighton during on September born of my 1974 I enrolled at MIT During my first two years a few months working scattering experiments Electron Accelorator. working Brian B. in the Schwartz on In the theory at of MIT's 234 Francis Bitter National Magnet Laboratory and eventually this became my thesis work. In February of position Public as assistant of Affairs 1978-79 Chairman to of 1978 I IBM's Thomas completed, I theoretical assistant to American J. Watson will commence a one half-time of Panel Society. Physical Research Richard L. Center in graduate studies second half-time as a postdoctoral superconductivity. in a Chairman whom I report is Dr. In September, my Heights, NY. IBM, this the started working researcher on The Garwin Yorktown now being position at in the area of 235 REFERENCES 124c 122b, 120b, Leiden Comm. 1. Kamerlingh Onnes, H. (1911). 2. W. Meissner and R. 787 (1933). 3. C.J. Gorter and H.B.G. Casimir, Z. (1934); 35, 963 Z. Phys. (1934). 4. Proc. Roy. Soc. A149, (1935); F. London and H. London, following the also see (1935); 341 2, Physica reference. 5. F. London, Superfluids, 6. E. Maxwell, Ochsenfeld, Naturwissenschaften 21, 78, B. Serin, W.H. Wright, and 487. 7. W.S. Wexler, 8. L.N. 9. J. 10. Rev. Phys. Phys. Cooper, Bardeen, 104, Rev. and C.B. Satterthwaite, (1954). 1189 and A. (1956). Phys. J.R.Schrieffer, Rev. (1957). 108, 1175 A.A. Abrikosov, V.L. 1442 Cooper, L.N. Ginzburg Zh. i Teor. Eksp. in Soviet Phys. English trans. 11. 96, 1950. C.A. Reynolds, 477 (1950); L.B.Nesbitt, Phys. Rev. 78, Goodman, B.B. Corak, New York, Wiley, Vol. I, Rev. Phys. Physica 1, 306 (1934); 539 15, techn. Phys. and Fiz. JETP Landau, Zh. L.D. 5, 32, 1442 (1957); 1174 (1957). i Eksp. Teor. Fiz. 20, 1064 (1950), English translation in Men of Physics: 1965, p. I, Oxford, Pergamon Press, L.D. Landau, Vol. 138. 12. B. also (1962); see 63 S2UIDS Applications: S. Research and Goodman, IBM Journal of Foner, eds., New York, 13. McNiff, S. Foner, E.J. 49A, 269 (1974). 14. Decroux, Fisher, M. 0. C: Phys. J. Sergent, (1975). 15. K. Maki, Phys. Rev. Development 6, Superconductor Pippard in A.B. Schwartz and and Machines, B.B. 148, 1977, Plenum, and E.J. p. Alexander, 1. Phys. Lett M. Chevrel, and Roth, R. S. L474 8, Phys. State Solid 362 (1966). 236 16. Hohenberg, Phys. E. Helfand, and P.C. N.R. Werthamer, Werthammer, p. R. N. also see (1966); 295 Rev. 147, 817, in p. Hohenberg, P.C. and Fetter A.L. and 321, Dekker, York, New ed., Parks, R.D. Superconductivity, 1969. 17. B.D. Josephson, Physics B.D. Josephson, Lett. Rev. of Mod. 1, 251 M.H. Cohen, L.M. Falicov, Lett. 8, 316 (1932). 19. R. Meservey, P.M. Tedrow, and P. 25, 1270 (1970). 20. P.M. Tedrow, R. Meservey, 4224 (1975). B11, 21. Meservey and R. (1976). 22. M.B. Maple, therein; Appl. Phys. Rev. Lett Phys. Rev. R.C. Bruno, and (1976), references and W.A. Fertig, D.C. Johnston, L.E. DeLong, McCallum, M.B. Matthias, B.T. and Maple, (1977); F. Acker, B.T. Lett. 38, 987 Rinderer, J. Low. Temp. Phys. 30, 133 Phys. Physics Today 24. H. Frohlich, Phys. 25. J. Bardeen, Phys. 26. Helv. Phys. R. Schafroth, M. Nuovo Cimento 9, 291 (1951). 27. J. Bardeen, in Handbuch der Physik, Vol. XV, ed., Berlin, Springer, 1956, p. 274. 28. Pippard, Proc. A.B. (1953). 29. I. Giaever, 30. N.N. Bogoliubov, Nuovo Cimento 7, 794 (1958); English trans. 58 (1958); Fiz. 34, i Teo. Phys. JETP 7, 41 (1958). 31. J.G. Valatin, Nuovo Cimento 7, 32. P.W. Anderson, Phys. Rev. Phys. Rev. 845 587 80, Royal Rev. Lett. 112, Rev. 7, p. (1950). (1950). Acta Soc. 5, No. 30, B.B. Schwartz and S. Forter, 34, (July, 1977). 79, R.W. Matthias, and L. (1978). 23. Rev. 131 58A, Lett. Phys. 179 9, also Rev. Fulde, Tedrow, Phys. see (1962); 251 (1974). Phillips, Phys. and J.C. 18. P.M. 46, Phys. 147 843 1900 (1951); 241, 645, S. Flugge, (London) A2 16, 547, (1960). (1958). (1958). Zh. Eksp. in Soviet 237 33. 34. Zh. Exsp. i Teo. Gor'kov, English trans. in Soviet Physics de P.G. Gennes, For for a Superconductivity, 67ff, 36. For 37. Metals and Alloys, discussion see, and quantitative to Tinkham, Introduction New York, review of high McGraw-Hill, 1975, superconductivity field in Phys. 22, see P. Fulde, Adv. pp. 667 in thin (1973). Principles of the Theory of Cambridge University Press, example J.M. Ziman, 2nd ed., Cambridge, See for Solids, 1972, (1958). 111ff. a films of (1958); 735 505 chapter 5. 1966, more detailed example, M. 34, JETP 7, Superconductivity New York, Benjamin, 35. Fiz. L.P. p.330 ff. 38. Y. Yosida, Phys. 39. P. Fulde Rev. 110, Ferrel, and R.A. 769 (1958). Phys. Rev. 135, A550 (1964); see also A.I. Larkin and Y.H. Orchinnikov, Zh. Eksp. i Teor. Fiz. 47, 1136 (1964); Soviet Physics JETP 20, 762 (1965). 40. B. S. Chandrasekhar, Appl. Clogston, Phys. Rev. Lett. Sarma, J. Phys. Phys. Lett 9, 266 1, 7 of Solids, 24, Chem. (1962); A.M. (1962). (1963). 1029 41. G. 42. K. Maki and T. 945 (1964). 43. transitions in recent reviews of the phase For metamagnets see J.M. Kincaid and E.G.D. Cohen, Physics N. and E. Stryjewski and Reports 22, 57 (1975); Tsuneto, Prog. Giordano, Advances 44. S. Frota-Pessoa 20, 505 45. R.B. 46. E. F. Roy. in Theor. Phys. Physics, and B.B. 26, 487 Schwartz, (Kyoto), 31, (1977). Solid State Comm. (1976). Griffiths, Phys. Rev. Lett. 24, Burton, J.O. Wilhelm, Soc. Can. 28, 65 715 and A.D. (1934); A.D. (1970). Misener, Trans. Misener and J.0. Wilhelm, 29, 13 (1935); A.D. Misener, H. Gayson Smith, (1935); A.D. Misener, Can. J. and J.0. Wilhelm, 29, 1, of Research 14, 25 (1936) 47. A. 48. L.A. Shal'nikov, Nature 142, 74 (1938). Feigin and A.I. Shal'nikov, Sov. Phys. Doklady 1, 238 377 (1957); D. Shoenberg, Superconductivity, Cambridge, Cambridge University Press, 1952, p. 166. Rev. (1958); Phys. 109, 686 Phys. Rev. (1959); Phys. Rev. 117, 672, (1960). 49. H. Meissner, Lett. 2, 458 50. Shapiro, J.L. Miles, P. Smith, S. Rev. Lett. 6, 686 (1961). 51. W.J. 52. 16, 16 (1966); W. W.J. Tomash, Phys. Rev. Lett. and T. Wolfram, Phys. Rev. Lett. 16, 352 (1966). 53. J.M. Rowell and W.L. McMillan, Phys. (1966). 54. P.G. de Gennes (1966). 55. P.M. 1552, 62A, Phys. Phys. 56. P.M. Chaikin, G. Arnold, and Phys. 26, 229 (1977). 57. G. 58. L. Dumoulin, E. 1086, 1977. 59. A. Gilabert, Ann. Phys. 60. P. 61. Equations 6 W.L. McMillan, Phys. Rev. 175, 537 (1968). Tr; Eq. 15 of missing factors and 10 in this paper are Eq. 39 in Q and 2; place of the a 4 in should have should be Tomash, Phys. Rev. and D. Lett. (1965); 672 15, Phys. J. Nicol, and Rev. Saint-James, Phys. Lett. Tomash 16, 453 4, 151 Lett. 36, Rev. Lett. Phys. P.K. Hansma, Chaikin and (1976); E.L. Wolf and J. Zasadzinsky, Phys. Lett. Finnemore, and D.K. Toplicar J.R. 165 (1977); C.K. So, and Bermon (1977); S. B16, 2072 Rev. Rev. B17, 4256, (1978). P.K. Hansma, J. Low Temp. Arnold, to be published. Guyon, and P. Nedellec, Nedellec, Ann. Phys. 203 (1977). 2, 253 (1977). 2, Phys. Rev. B16, 7. 62. Dzyaloshinski, and I.E. L.P. Gorkov, A.A. Abrikosov, Methods of Quantum Field Theory in Statistical Physics, p. Dover, 1975, New York, R.A. Silverman, trans. by 283-4. 63. T. Matsubara, Prog. Theor. 64. 36, 1918 (1959); Exsp. i Teor. Fiz. L.P. Gor'kov, Zh. (1959). 1364 9, Engl. trans. in Soviet Phys. JETP Phys. 14, 351 (1955). 239 W. 65. references gives Silvert (1975); proximity the on work earlier to 439 20, Silvert, J. Low Temp. Phys. effect using the Gor'kov equation approach. P.G. 66. Gennes de E. and 3, Lett. Phys. Guyon, 168 (1963). P. 67. W. Fulde and (1967); A. Kondens. Morrman, Phys. Migliori and Mater. 6, 403 Rev. BS, Phys. Ginzberg, D.M. N.L. Smith, and Blackburn, H.J.T. (1973); J.A. 5063 M. H.J. Fink, (1975); B11, 1053 Phys. Rev. Rowell, J.P. Laheurte, J.P. Gilabert, A. Sheikholeslam, B14, 1052 E. Guyon, Phys. Rev. Romagnan, J.C, Nioray, Phys. F and A.A. Babaei-Brojeuy, J. (1976); N. Rivier 7, L 111 (1977). Yu, M. 68. M. A. Strongin, Rev Phys. Paskin, 996 B14, (1976). Cooper, L.N. Jorunal 6, 75 70. P.G. 71. G.V. Minnigerode, Z. 72. . de Gennes, Rev. Mod. 0. Entin-Wohlman Phys. Rev. B. Phys. fur Phys. and (1961); 689 6, Rev. Lett. Phys. (1962). 69. 36, 225 (1964). 192, 379 (1966). J. Bar-Sagi, IBM published in to be 73. D. Belin, M. Leger, A. Klein, J. Bellanger, D. J.M. Rowell, 459 (1973); Lett 42A, Defourneau, Phys. Phys. Rev. Lett. 30, 167 (1973); J. Vac. Sci. Tech. 10, 702 (1973); S.L. Colucci, W.J. Tomasch, H.J. Lee, Phys. Rev. Lett. 32, 590 (1974). 74. W.L. McMillan and P.W. (1966). 75. W.L. McMillan, Phys. Rev. 76. S. 77. J. Bar-Sagi, 1669 (1976). 78. T.E. Feuchtwang, Phys. Rev. B10, 4121, 3979 (1975); B13, 517 (1976); B13, 517 79. J. (1974); 15, 729 Comm. State Bar-Sagi, Solid J. Bar-Sagi and 0. Entin- Wohlman, Solid State Comm. 22 29 (1977). Bermon and C.K. So, Z. Physik Anderson, Phys. Rev. 559 175, Phys. 269, 16, 85 (1968). Rev. B17, 7 Lett (1974); 4256 J. (1978). Phys. F 6, 4135(1974); B12, (1976). 240 Rev. B1, Phys. Zuckermann, and M.J. 229 80. A.B. Kaiser (1970). 81. K. 82. Y. Nambu, Phys. 83. V. Ambegaokar and A. Baratoff, Phys. Rev. (1963); Erratum 11, 104 (1963). 84. of Theory 2uantum Walecka, J.D. and Fetter A.L. pp. 1971, McGraw-Hill, York, New Systems, Many-Particle 248-50. 85. R. 86. 0. S. Skalski, A1500 (1964). 87. P. 88. Romagnan, and E. A. Gialbert, J.P. Comm. 9, 1295 (1971). 89. G.B. Arnold, Phys. 90. Tedrow and P.M. (1976). 91. P.M. Tedrow and R. Meservey, to be published. 92. 57 51A, Lett. Phys. R. Meservey, and Tedrow P.M. (1975); R. Meservey, P.M. Tedrow, and R.C. Bruno, Phys. Rev. B17, 2915 (1978). 93. R. Meservey and P.M. Tedrow, to be published. 94. D. E. MacLaughlin, Solid State Physics 31, 1 (1976) has in experiments magnetic resonance reviewed recently superconductors, including Knight shift experiments. 95. F. 96. Androes and G. M. (1961). 97. R. H. Hammond (1964). 98. R. 99. P. W. Machida, J. Rev. Weiss, B17, 3576 106, W. and G. M. Ferrell, Phys. Anderson, Phys. 208 Rev. Phys. 136, (1965). 675 18, Solid State Lett. 58A, 237 (1957). Phys. Rev. Kelly, Rev. Mod. Rev. Lett. Rev. 468 (1978). Knight, D. 10, Lett. Guyon, Phys. Meservey, R. (1977). (1960). Rev. Lett. Maki, Phys. Rev. 27 737 Appendix B. 1972, MIT, Betbeder, P.R. Fulde and K. A. 648 117, Rev., Boro, Thesis, Reif, Phys. Phys. Low Temp. 3, 262 Lett. 3, 121, 779 36, 185 Phy. (1959). 325, (1959). 241 Zh. Eksperim. Gorkov, Phys. JETP 15, Soviet and L. P. A. A. Abrikosov 100. 42, 1088, (1962); i. Teo. Fiz. 752 (1962). Matthias, unpublished; see Ref. 14. 101. B.T. 102. R. A. Farrell, University of Maryland Technical Report No. 329 (unpublished). 103. R. Kubo (1956). 104. J. Appel, Phys. 105. R. H. Hammond Phys. Obata, J. and Y. Rev 139, G. and 547 11, (1965). A1536 M. Japan Soc. Rev. Lett 18, Kelly, P. and M. Strongin, Phys. 156 (1967). 106. H. L. Fine, M. Lipsicas, 29A, 366 (1969). 107. P. W. Anderson New York, 108. J. E. Rev. 109. 110. in Superconductivity, Dekker, 1969, R. D. I. 1352. p. and A. K. Crow, M. Strongin, B9, 135 (1974). Schiff McGraw-Hill, ed., Parks, Bhatnagar, R. and Bethe A. H. example for See ed. Mechanics, second Intermediate 2uantum MA, Benjamin, 1968), p. 377. L. Lett. Mechanics, Q 2uantum 1968, p. 438. 3rd ed., 14, 1 (1963). Yafet, Solid State Physics Phys. Jackiw, , Reading, New York, 111. Y. 112. H. E. White, Introduction to Atomic Spectra, New York, McGraw-Hill, 1934, p. 109, 112. 113. F. W.A. Wright, Lett. 18, 115 Hine s, and W.D. Knight, Phys. Rev. (1967). 114. F. Wright, Phys. Rev. 163, W.D. Knight, Phys. Rev. B4, Hines and 115. V.I. Petinov and A. Yu. Ardashev, Fis. Tverd. Tela 3 (1969); Sov. Phys. Solid State 11, 1 (1969). 11, 116. Sambles, J.R. Stesmans, and J. (1977). 117. W.M. Walsh, G. 420 893 Sharp-Dent, Witters, Phys. L.W. Rupp, and (1967); (1971). W.A. Cousins, J.E. Stat. Sol. (b) 79, P.H. Schmidt, Phys. A. 645 Rev. 242 Schultz and G. Latham, Phys. Lett. 16, 181 (1966); S. T. Kato, S. Hiramatsu, and Rev. Lett. 13, 148 (1965); (1978); D. Japan 44, 449 3. Phys. Soc. H. Hirakawa, 1104 Lett. 36, Rev. Schultz, Phys. and S. Lubzens Phys. Ref. R.T. Schumacher, K. Wang and (1976); S. Phys. Rev. correction in (1973); important B8, 411 9 Zh. Petinov and V.I. Ya. Gen B10, 21 29 (1974); M. Ekxp. T eor Fiz. 48, 29 (1965); Sov. Phys. JETP 2 1, 19 S tatus Cousins, Phys. and J.E. A.J. Watts (1965); Chem. Phys. Taupin, J. C. (1968); Solidi 30 105 Solids 28, 41 (1967); K. Saiki, T. Fujita, Y. Shi mizu, 32 , 447 Soc. Japan Wada, J. Phys. Sakoh, and W. S. (1972). 118. Physics, Solid State Introduction to C. Kittel, ed., New York, John Wiley & Sons, 1976, p. 167. 5th 119. Functions, Vol. Higher Trancendental A. Erdelyi, New York, McGraw-Hill, 1953, p. 19. I,