advertisement

Three Lectures on Soft Modes and Scale Invariance in Metals Quantum Ferromagnets as an Example of Universal Low-Energy Physics Soft Modes and Scale Invariance in Metals Quantum Ferromagnets as an Example of Universal Low-Energy Physics Dietrich Belitz, University of Oregon with T.R. Kirkpatrick and T. Vojta Reference: Rev. Mod. Phys. 77, 579, (2005) Part I: Phase Transitions, Critical Phenomena, and Scaling Part II: Soft Modes, and Generic Scale Invariance Part III: Soft Modes in Metals, and the Ferromagnetic Quantum Phase T Transition Part 1: Phase Transitions I. Preliminaries: First-Order vs Second-Order Transitions Example 1: The Liquid-Gas Transition Schematic phase diagram of H2O February 4-5, 2013 Singapore Winter School 3 Part 1: Phase Transitions I. Preliminaries: First-Order vs Second-Order Transitions Example 1: The Liquid-Gas Transition Schematic phase diagram of H2O February 4-5, 2013 Singapore Winter School 4 Part 1: Phase Transitions I. Preliminaries: First-Order vs Second-Order Transitions Example 1: The Liquid-Gas Transition T < Tc: 1st order transition (latent heat) Schematic phase diagram of H2O February 4-5, 2013 Singapore Winter School T > Tc: No transition T = Tc: Critical point, special behavior 5 Example 2: The Paramagnet - Ferromagnet Transition H = 0: Transition is 2nd order T > Tc: Disordered phase, m = 0 T < Tc: Ordered phase, m ≠ 0 T -> Tc: m -> 0 continuously m is called order parameter Examples: •Ni Tc = 630K •Fe Tc = 1,043K •ZrZn2 Tc = 28.5K •UGe2 Tc = 53K Demonstration of the FM critical point February 4-5, 2013 Singapore Winter School 6 2nd order transitions, a.k.a. critical points, are special! • The OP goes to zero continuously: and is a nonanalytic function of H: • The OP susceptibility diverges m(H=0) ~ (Tc - T) β m(T=Tc ) ~ H 1/δ χ ~ |T - T | -γ c • The specific heat shows an anomaly C ~ |T – Tc| -α Underlying reason: Strong OP fluctuations lead to a diverging length scale scale (correlation length ξ): ξ ~ |T – Tc| -ν Examples: • Critical opalescence in a classical fluid • Simulation of the 2D Ising model Consequences: • Universality • Scale invariance Explanation: February 4-5, 2013 • Homogeneity laws (a.k.a. scaling laws) Singapore Winter School 7 Mohan et al 1998 February 4-5, 2013 Singapore Winter School 8 2nd order transitions, a.k.a. critical points, are special! • The OP goes to zero continuously: and is a nonanalytic function of H: • The OP susceptibility diverges m(H=0) ~ (Tc - T) β m(T=Tc ) ~ H 1/δ χ ~ |T - T | -γ c • The specific heat shows an anomaly C ~ |T – Tc| -α Underlying reason: Strong OP fluctuations lead to a diverging length scale scale (correlation length ξ): ξ ~ |T – Tc| -ν Examples: • Critical opalescence in a classical fluid • Simulation of the 2D Ising model Consequences: • Universality • Scale invariance Explanation: February 4-5, 2013 • Homogeneity laws (a.k.a. scaling laws) Singapore Winter School 9 Mohan et al 1998 February 4-5, 2013 Singapore Winter School 10 2nd order transitions, a.k.a. critical points, are special! • The OP goes to zero continuously: and is a nonanalytic function of H: • The OP susceptibility diverges m(H=0) ~ (Tc - T) β m(T=Tc ) ~ H 1/δ χ ~ |T - T | -γ c • The specific heat shows an anomaly C ~ |T – Tc| -α Underlying reason: Strong OP fluctuations lead to a diverging length scale scale (correlation length ξ): ξ ~ |T – Tc| -ν Examples: • Critical opalescence in a classical fluid • Simulation of the 2D Ising model Consequences: • Universality • Scale invariance Explanation: February 4-5, 2013 • Homogeneity laws (a.k.a. scaling laws) Singapore Winter School 11 Source: Scientific American 2nd order transitions, a.k.a. critical points, are special! • The OP goes to zero continuously: and is a nonanalytic function of H: • The OP susceptibility diverges m(H=0) ~ (Tc - T) β m(T=Tc ) ~ H 1/δ χ ~ |T - T | -γ c • The specific heat shows an anomaly C ~ |T – Tc| -α Underlying reason: Strong OP fluctuations lead to a diverging length scale scale (correlation length ξ): ξ ~ |T – Tc| -ν Examples: • Critical opalescence in a classical fluid • Simulation of the 2D Ising model Consequences: • Universality • Scale invariance Explanation: February 4-5, 2013 • Homogeneity laws (a.k.a. scaling laws) Singapore Winter School 13 T << Tc T > Tc Source: Ch. Bruder T ≈ Tc 2nd order transitions, a.k.a. critical points, are special! • The OP goes to zero continuously: and is a nonanalytic function of H: • The OP susceptibility diverges m(H=0) ~ (Tc - T) β m(T=Tc ) ~ H 1/δ χ ~ |T - T | -γ c • The specific heat shows an anomaly C ~ |T – Tc| -α Underlying reason: Strong OP fluctuations lead to a diverging length scale scale (correlation length ξ): ξ ~ |T – Tc| -ν Examples: • Critical opalescence in a classical fluid • Simulation of the 2D Ising model Consequences: • Universality • Scale invariance Explanation: February 4-5, 2013 • Homogeneity laws (a.k.a. scaling laws) Singapore Winter School 15 Universality: All classical fluids share the same critical exponents: α = 0.113 ± 0.003; β = 0.321 ± 0.006; γ = 1.24 ± 0.01; ν = 0.625 ± 0.01 The exponent values are the same within the experimental error bars, even though the critical pressures, densities, and temperatures are very different for different fluids! Even more remarkably, a class of uniaxial ferromagnets also shares these exponents! This phenomenon is called universality. We also see that the exponents do not appear to be simple numbers. However, all critical points do NOT share the same exponents. For instance, in isotropic ferromagnets, the critical exponent for the order parameter is β = 0.358 ± 0.003 which is distinct from the value observed in fluids. All systems that share the same critical exponents are said to belong to the same universality class. Experimentally, the universality classes depend on the system’s • dimensionality d • symmetry properties For sufficiently large d, the critical behavior of most systems becomes rather simple. Example: FMs in d ≥ 4 have β = 1/2, γ = 1, ν = 1/2 (“mean-field exponents”). 2nd order transitions, a.k.a. critical points, are special! • The OP goes to zero continuously: and is a nonanalytic function of H: • The OP susceptibility diverges m(H=0) ~ (Tc - T) β m(T=Tc ) ~ H 1/δ χ ~ |T - T | -γ c • The specific heat shows an anomaly C ~ |T – Tc| -α Underlying reason: Strong OP fluctuations lead to a diverging length scale scale (correlation length ξ): ξ ~ |T – Tc| -ν Examples: • Critical opalescence in a classical fluid • Simulation of the 2D Ising model Consequences: • Universality • Scale invariance Explanation: February 4-5, 2013 • Homogeneity laws (a.k.a. scaling laws) Singapore Winter School 17 Scale invariance: Measure the magnetization M of a FM as a function of the magnetic field H at a fixed temperature T very close to Tc . The result looks like this: Now scale the axes, and plot h = H / |T – Tc| x Versus m = M/ |T – Tc| y If we choose y = β, and x = βδ, then the all of the curves collapse onto two branches, one for T > Tc, and one for T < Tc ! Note how remarkable this is! It works just as well for other magnets. It reflects the fact that at criticality the system looks the same at all length scales (“self-similarity”), as demonstrated in this simulation of a 2-D Ising model. Mohan et al 1998 Source: J. V. Sengers Measure the magnetization M of a FM as a function of the magnetic field H at a fixed temperature T very close to Tc . The result looks like this: Now scale the axes, and plot h = H / |T – Tc| x Versus m = M/ |T – Tc| y If we choose y = β, and x = βδ, then the all of the curves collapse onto two branches, one for T > Tc, and one for T < Tc ! Note how remarkable this is! It works just as well for other magnets. It reflects the fact that at criticality the system looks the same at all length scales (“self-similarity”), as demonstrated in this simulation of a 2-D Ising model. Mohan et al 1998 2nd order transitions, a.k.a. critical points, are special! • The OP goes to zero continuously: and is a nonanalytic function of H: • The OP susceptibility diverges m(H=0) ~ (Tc - T) β m(T=Tc ) ~ H 1/δ χ ~ |T - T | -γ c • The specific heat shows an anomaly C ~ |T – Tc| -α Underlying reason: Strong OP fluctuations lead to a diverging length scale scale (correlation length ξ): ξ ~ |T – Tc| -ν Examples: • Critical opalescence in a classical fluid • Simulation of the 2D Ising model Consequences: • Universality • Scale invariance Explanation: February 4-5, 2013 • Homogeneity laws (a.k.a. scaling laws) Singapore Winter School 21 Homogeneity laws: Consider the magnetization M as a function of ξ and H. Suppose we scale lengths by a factor b, so ξ -> ξ / b. Suppose M at scale b = 1 is related to M at scale b by a generalized homogeneity law x M(ξ ,H) = b –β/ν M(ξ / b, H δβ/ν) x But ξ ~ t –ν => ξ / b = (t b 1/ν) –ν (This was initially postulated as the “scaling hypothesis (Widom, Kadanoff), and later derived by means of the renormalization group (Wilson) ) where t = |T – Tc| / Tc and therefore M(t, H) = b –β/ν M(t b 1/ν, H b δβ/ν) But b is an arbitrary scale factor, so we can choose in particular b = t –ν. Then M(t ,H) = t β M(1, H / t δβ) And in particular M(t, H=0) ~ t β and M(t=0, H) ~ H 1/δ No big surprise here, we’ve chosen the exponents such that this works out! But, it follows that M(t, H)/t β = F(H / t βδ) , with F(x) = M(t=1, x) an unknown scaling function. This explains the experimental observations! 2nd order transitions, a.k.a. critical points, are special! • The OP goes to zero continuously: and is a nonanalytic function of H: • The OP susceptibility diverges m(H=0) ~ (Tc - T) β m(T=Tc ) ~ H 1/δ χ ~ |T - T | -γ c • The specific heat shows an anomaly C ~ |T – Tc| -α Underlying reason: Strong OP fluctuations lead to a diverging length scale scale (correlation length ξ): ξ ~ |T – Tc| -ν Examples: • Critical opalescence in a classical fluid • Simulation of the 2D Ising model Consequences: • Universality • Scale invariance Explanation: February 4-5, 2013 • Homogeneity laws (a.k.a. scaling laws) Singapore Winter School 23 II. Classical vs. Quantum Phase Transitions Critical behavior at 2nd order transitions is caused by thermal fluctuations. Question: What happens if Tc is suppressed to zero, which kills the thermal fluctuations? This can be achieved in many low-Tc FMs, e.g., UGe2: Answer: Quantum fluctuations take over. There still is a transition, but the universality class changes. Question: How can this happen in a continuous way? Saxena et al 2000 February 4-5, 2013 Singapore Winter School 24 II. Classical vs. Quantum Phase Transitions Critical behavior at 2nd order transitions is caused by thermal fluctuations. Question: What happens if Tc is suppressed to zero, which kills the thermal fluctuations? This can be achieved in many low-Tc FMs, e.g., UGe2: Answer: Quantum fluctuations take over. There still is a transition, but the universality class changes. Question: How can this happen in a continuous way? Answer: By means of a crossover. February 4-5, 2013 Singapore Winter School 25 Crucial difference between quantum and classical phase transitions: Coupling of statics and dynamics Consider the partition function Z, which determines the free energy F = -T log Z Classical system: (β = 1/T) Hkin and Hpot commute Hpot determines the thermodynamic behavior, independent of the dynamics => In classical equilibrium statistical mechanics, the statics and the dynamics are independent of one another February 4-5, 2013 Singapore Winter School 26 Quantum system: H = H(a+, a) in second quantization Hkin and Hpot do NOT commute => statics and dynamics couple, and need to be considered together! Technical solution: Divide [0,β] into infinitesimal sections parameterized by 0 ≤ τ ≤ β (“imaginary time”), making use of BCH, and represent Z as a functional integral over auxiliary fields (Trotter, Suzuki) with S an “action” that depends on the auxiliary fields: The fields commute for bosons, and anticommute for fermions. February 4-5, 2013 Singapore Winter School 27 T = 0 corresponds to β = ∞ => Quantum mechanically, the statics and the dynamics couple! => A d-dimensional quantum system at T = 0 resembles a (d+1)-dimensional classical system! Caveat: τ may act akin to z spatial dimensions, with z ≠ 1, and z not eve even integer Example: In a simple theory of quantum FMs, z = 3 (Hertz) Quantum FMs in d ≥ 1 act like classical FMs in d ≥ 4 February 4-5, 2013 Singapore Winter School 28 T = 0 corresponds to β = ∞ => Quantum mechanically, the statics and the dynamics couple! => A d-dimensional quantum system at T = 0 resembles a (d+1)-dimensional classical system! Caveat: τ may act akin to z spatial dimensions, with z ≠ 1, and z not eve even integer Example: In a simple theory of quantum FMs, z = 3 (Hertz) Quantum FMs in d ≥ 1 act like classical FMs in d ≥ 4 Prediction: The quantum FM transition is 2nd order with mean-field exponents (Hertz 1976) February 4-5, 2013 Singapore Winter School 29 III. The Quantum Ferromagnetic Transition Problem: The prediction does not agree with experiment ! When Tc is suppressed far enough, the transition (almost *) invariably becomes 1st order! Example: UGe2 Many other examples Generic phase diagram February 4-5, 2013 Taufour et al 2010 * Some exceptions: • • • Singapore Winter School Strong disorder Quasi-1D systems Other types of order interfere 30 URhGe Huxley et al 2007 February 4-5, 2013 Singapore Winter School 31 III. The Quantum Ferromagnetic Transition Problem: The prediction does not agree with experiment ! When Tc is suppressed far enough, the transition (almost *) invariably becomes 1st order! Example: UGe2 Many other examples Generic phase diagram February 4-5, 2013 Taufour et al 2010 * Some exceptions: • • • Singapore Winter School Strong disorder Quasi-1D systems Other types of order interfere 33 Questions: What went wrong with the prediction? Hint: It’s a long way from the basic Trotter formula to a theory of quantum FMs. What is causing the wings? Hint: Wings are known in classical systems that show a TCP. Why is the observed phase diagram so universal? Hint: It must be independent of the microscopic details, and only depend on features that ALL metallic magnets have in common. February 4-5, 2013 Singapore Winter School 35 Part 2: Soft Modes I. Critical soft modes Landau theory for a classical FM: t>0 t=0 t<0 FL(m) = t m2 + u m4 + O(m6) Assumptions: • m is small • The coefficients are finite • t ~ T – Tc • Landau theory replaces the fluctuating OP by its average (“mean-field approx.”) • FL can in principle be derived from a microscopic partition function • Describes a 2nd order transition at t = 0. • NB: No m3 term for symmetry reasons => 2nd order transition ! In In a classical fluid there is a v m3 term that vanishes at the critical point February 4-5, 2013 Singapore Winter School 36 How about fluctuations? Write M(x) = m + δM(x) and consider contributions to Z or F by δM(x): Landau-Ginzburg-Wilson (LGW) For small δM(x), expand to second order => integral can be done => Ornstein-Zernike result for the susceptibility: How good is the Gaussian approximation? • Qualitatively okay for d > 4 (Ginzburg) • Qualitatively wrong for d < 4. In general, February 4-5, 2013 Singapore Winter School 37 Discuss the Ornstein-Zernike result: Obeys scaling with γ = 1 and ν = 1/2. Holds for both t > 0 and t < 0. For |t| ≠ 0, correlations are ranged (exponential decay) short For t = 0, correlations are long ranged (power-law decay) ! No characteristic length scale => scale invariance The homogeneous susceptibility and the correlation length diverge for t = 0 • No resistance against formation of m ≠ 0 • m rises faster than linear with H • The OP fluctuations are a soft (or massless) mode (or excitation) These critical soft modes are soft only at a special point in the phase diagram, viz., the critical point => There is scale invariance only at the critical point. February 4-5, 2013 Singapore Winter School 38 II. Generic soft modes, Mechanism 1: Goldstone modes So far we have been thinking of Ising magnets Consider a classical planar magnet instead: Spins in a plane; OP m is a vector Disordered phase: Random orientation of spins m = <m> = 0 February 4-5, 2013 Singapore Winter School 39 Ordered phase at T << Tc: Near-perfect alignment of spins, m ≠ 0 NB: The direction of the spins is arbitrary! February 4-5, 2013 Singapore Winter School 40 Suppose we rotate all spins by a fixed angle: •This costs no energy, since all spin directions are equivalent! •The free energy depends only on the magnitude of m, not on its direction. •Another way to say it: There is no restoring force for co-rotations of the spins. February 4-5, 2013 Singapore Winter School 41 Suppose we rotate the spins by a slightly position dependent angle: This will cost very little energy! February 4-5, 2013 Singapore Winter School 42 Conclusions: There is a soft mode (spin wave) consisting of transverse (azimuthal) fluctuations of the magnetization. The free energy has the shape of a Mexican hat. The transverse susceptibility diverges everywhere in the ordered phase The longitudinal (radial) fluctuations do cost energy; they are massive. The spin rotational symmetry is spontaneously broken (as opposed to explicitly broken by an external field): The Hamiltonian is still invariant under rotations of the spin, but the lowest-free-energy state is not. However, the free energy of the resulting state is still invariant under co-rotations of the spins. Works analogously for Heisenberg magnets: 2 soft modes rather than 1 There is a simple mechanical analog of this phenomenon. February 4-5, 2013 Singapore Winter School 43 massive massive massive massive massive soft This is an example of Goldstone’s Theorem: A spontaneously broken continuous symmetry in general leads to the existence of soft modes (“Goldstone modes”). More precisely: If a continuous symmetry described by a group G is spontaneously broken such that a subgroup H (“little group” or “stabilizer group”) remains unbroken, then there are n Goldstone modes, where n = dim (G/H). Example: •Heisenberg magnet: G = SO(3) (rotational symmetry of the 3-D spin) H H = SO(2) (rotational symmetry in the plane perpendiperpendicular cular to the spontaneous magnetizaton) n = dim(SO(3)/SO(2)) = 3 – 1 = 2 (2 transverse magnons) N n The transverse susceptibility diverges as No characteristic length scale => scale invariance! Note: The Goldstone modes are soft everywhere in the ordered phase, not just at the critical point! This is an example of “generic scale invariance” February 4-5, 2013 Singapore Winter School 45 III. Generic soft modes, Mechanism 2: Gauge invariance Electrodynamics => For charged systems, gauge invariance is important. For the study of, e.g., superconductors, we need to build in this concept! Consider again the LGW action, but with a vector OP a complex scalar OP : , or, equivalently, (“ “) Now postulate that the theory must be invariant under local gauge transformations, i.e., under with an arbitrary real field Λ(x). A does not fulfill this requirement because of the gradient term Modify the gradient term February 4-5, 2013 Singapore Winter School 46 The simplest modification that does the trick is where the gauge field A(x) transforms as and is the field tensor. q (“charge”) and μ are coupling constants. February 4-5, 2013 Singapore Winter School 47 Notes: • This is the LGW version of a model Ginzburg and Landau proposed as a model for superconductivity. • GL solved the action in a mean-field approximation that replaced both ϕ and the magnetic field by their expectation values. This theory was later shown by Gorkov to be equivalent to BCS theory. • The LGW theory is much more general: A describes the fluctuating electromagnetic field that is nonzero even if there is applied magnetic field (i.e., if the mean value of B is zero.) February 4-5, 2013 Singapore Winter School 48 Now consider the soft modes in Gaussian approximation. Disordered phase: < ϕ > = 0 . • A appears with gradients only => A is soft. In Coulomb gauge ( ) one finds two soft modes “transverse photon”: 2 soft modes (“transverse photon”) • ϕ is massive with mass t > 0: 2 massive modes • Conclusion: • Two massless and two massive modes • Photon is a generic soft mode (result of gauge invariance) • Photon has only two degrees of freedom • Any phase transition will take place on the background of the generic scale invariance provided by the photon ! February 4-5, 2013 Singapore Winter School 49 Ordered phase: < ϕ > ≠ 0 . • Write • Expand to second order in ϕ1, ϕ2, and A: February 4-5, 2013 Singapore Winter School 50 Ordered phase: < ϕ > ≠ 0 . • Write • Expand to second order in ϕ1, ϕ2, and A: • A acquires a mass ~ v2 • ϕ2 couples to the massive A, can be eliminated by shifting A: February 4-5, 2013 Singapore Winter School 51 Ordered phase: < ϕ > ≠ 0 . • Write • Expand to second order in ϕ1, ϕ2, and A: • A acquires a mass ~ v2 • ϕ2 couples to the massive A, can be eliminated by shifting A: • This yields 3 massive modes (“transverse + longitudinal photons”) with m2 ~ q2 v2, and 1 massive mode February 4-5, 2013 Singapore Winter School 52 • Conclusion: • No soft modes! • The Goldstone mode candidate ϕ2 gets eaten by the gauge field, which becomes massive in the process (“Anderson-Higgs mechanism”) • Photon now has three degrees of freedom, all of them massive • Physical manifestation of the massive modes: Meissner effect • Note: The same principle applies to more complicated gauge groups Example: Electroweak symmetry breaking SU(2)xU(1) gets broken to U(1) => o One massless gauge boson (photon) o 4 – 1 = 3 Goldstone bosons that become massive via Anderson-Higgs => W±, Z vector bosons o Physical manifestation: Short-ranged weak interaction February 4-5, 2013 Singapore Winter School 53 IV. Digression. Generic soft modes, Mechanism 3: C Conservation laws • There is a third mechanism leading to generic scale invaviance: Conservation laws • They can lead to time-correlation functions in classical systems to decay algebraically rather than exponentially => temporal long-range correlations • Through mode-mode-coupling effects, this can happen even to time correlation functions of modes that are not themselves conserved. Example: Transverse-velocity correlations in a classical fluid. • As result, transport coefficients (viscosities) are nonanalytic functions of the frequency, and hydrodynamics break down in d = 2. • For classical systems in equilibrium, this affects the dynamics only. • For quantum systems, and for classical nonequilibrium systems, the statics and dynamics couple, and the thermodynamic behavior is affected as well ! February 4-5, 2013 Singapore Winter School 54 V. Fluctuation-induced 1st order transitions Consider the fluctuating GL theory again: • A appears only quadratically => A can be integrated out exactly! • Still replace ϕ -> <ϕ> => “generalized mean-field approximation”, or “renormalized mean-field theory” • The difference between this an GL theory is that it takes into account the fluctuating electromagnetic field. • A couples to ϕ => The resulting action contains ϕ to all orders. • <AA> is soft for ϕ = 0, and massive for ϕ ≠ 0 => S[ϕ] cannot be an analytic function of ϕ ! February 4-5, 2013 Singapore Winter School 55 The resulting generalized MFT for 3-D systems is t>t1 t=t1 t<t1 • This describes a 1st order transition at some t1 > 0 ! • The fluctuating A - field changes the nature of the phase transition! • This is called a fluctuation-induced 1st order transition (Halperin, Lubensky, Ma 1974) • NB: This is a classical transition! • There is an analogous mechanism in particle physics (Coleman-Weinberg). • This is a consequence of generic scale invariance, with the generic soft modes coupling to the OP. • An essentially identical theory applies to the transition from the nematic phase to the smectic-A phase in liquid crystals, with the nematic Goldstone modes providing the GSI. February 4-5, 2013 Singapore Winter School 56 • In superconductors, the effect is too small to be observed • In liquid crystals, the transition is 1st order in some systems, but 2nd order in others. This is believed to be due to the fluctuations of the OP, which are neglected in the generalized MFT. • For 4-D superconductors, or liquid crystals, the result is with v > 0 Why do we care? See below. February 4-5, 2013 Singapore Winter School 57 Part 3: Soft Modes in Metals, and the Ferrom magnetic Quantum Phase Transition I. Soft modes in metals There are (at least) two types of soft modes in clean metals at T=0: Single-particle excitations Described by Green’s function Soft at k = kF, iωn = 0 => The leading properties of a Fermi liquid follow via scaling (Nayak & Wilczek, Shankar) February 4-5, 2013 Singapore Winter School 58 Two-particle excitations +T+m+H kF • Soft mode, mixes retarded and advanced degrees of freedom; result of a spontaneously broken (unobvious) continuous symmetry (F Wegner). Weight given by DOS. • These are Goldstone modes, i.e., they represent generic scale invariance in a Fermi liquid. • Soft only at T = 0. • Appear in both spin-singlet and spin-triplet channels, the latter couples to the magnetization. • A nonzero magnetization gives the triplet propagator a mass. So does a magnetic field. • A ferromagnetic phase transition at T=0 will take place on the background of these generic soft modes. (Cf. the classical superconductor/liquid Xtal!) February 4-5, 2013 Singapore Winter School 59 II. The ferromagnetic quantum phase transition Idea: Construct a renormalized mean-field theory in analogy to HLM, with the magnetization m as the OP and the two-particle electron excitations as the generic soft modes. Result: The free energy maps onto that of the superconductor/liquid Xtal problem in D = 4: Discussion: • First-order transition at T = 0 (always!), and at low T • Second-order transition at higher T => tricritical point February 4-5, 2013 Singapore Winter School 60 • First-order transition at T = 0 (always!), and at low T • Second-order transition at higher T => tricritical point February 4-5, 2013 Singapore Winter School 61 • First-order transition at T = 0 (always!), and at low T • Second-order transition at higher T => tricritical point • First-order transition at low H, ends in a quantum critical point February 4-5, 2013 Singapore Winter School 62 • First-order transition at T = 0 (always!), and at low T • Second-order transition at higher T => tricritical point • First-order transition at low H, ends in a quantum critical point • T – t - H phase diagram displays surfaces of first-order transitions (“tricritical wings”) • This explains the experimental observations! February 4-5, 2013 Singapore Winter School 63 • First-order transition at T = 0 (always!), and at low T • Second-order transition at higher T => tricritical point • First-order transition at low H, ends in a quantum critical point • T – t - H phase diagram displays surfaces of first-order transitions (“tricritical wings”) • This explains the experimental observations! • An open question: Why are the OP fluctuations so inefficient? February 4-5, 2013 Singapore Winter School 64 Summary of Crucial Points All relevant soft modes are important to determine the physics at long length and time scales. Generic soft modes, and the resulting scale invariance, are quite common, and there are various mechanism for producing them. Analogies between seemingly unrelated topics can be very useful ! February 4-5, 2013 Singapore Winter School 65