First Elements of Thermal Neutron Scattering Theory (II) Daniele Colognesi Istituto dei Sistemi Complessi, Consiglio Nazionale delle Ricerche, Sesto Fiorentino (FI) - Italy Talk outlines 0) Introduction. 1) Neutron scattering from nuclei. 2) Time-correlation functions. 3) Inelastic scattering from crystals. 4) Inelastic scattering from fluids (intro). 5) Vibrational spectroscopy from molecules. 6) Incoherent inelastic scattering from molecular crystals. 7) Some applications to soft matter. 4) Inelastic scattering from fluids (intro) Disordered systems (gasses, liquids, glasses, amorphous solids etc.): atomic order only at short range (if existing). For simplicity’s sake only monatomic fluid systems are considered here. key quantities: density, , constant, and pair correlation function, g(r) 1 g (r ) (r ri rj ) N i j connected to the static structure factor, S(Q), via a 3D spatial Fourier transform: 1 S (Q) N i j exp i Q (ri rj ) 1 dr ( g (r ) 1) exp(i Q r ) where both S(Q) and g(r) exhibit some special values at their extremes: S (0) T kBT ; S () 1 g (0) 0; g () 1 Since S(Q)=I(Q,t=0), it is possible to generalize g(r) by introducing the time-dependent pair correlation function, G(r,t): G (r , t ) 1 8 3 dQ exp(i Q r) I (Q, t ) s r (0) s r r (t ) ds 1 N i j i, j and the time-dependent self pair correlation function, Gself(r,t): Gself (r , t ) 1 N 1 8 3 dQ exp(i Q r) I self s r (0) s r r (t ) ds i i i (Q, t ) where the t=0 values of G(r,t) and Gself(r,t) are: G(r ,0) (r) g (r ) Gself (r ,0) (r) No elastic scattering,(), in fluids! the elastic components in S(Q,) and Sself(Q,) come from the asymptotic values of I(Q,t) and Iself(Q,t): I (Q, ) ( ) S (Q, ) I (Q, ) ( ) elastic inelastic dtdr (r , ) G (r , t ) G (r , ) exp(it iQ r ) G 2 asymptotic approaching 0 for t Due to the asymptotic loss of time correlation, and making use of =i(r-ri), one writes: i, j 1 2 s ri (0) s r rj (t ) ds V ; N 1 Gself (r , ) N 1 s ri (0) s r ri (t ) ds 0 N 1 G ( r , ) N i so, finally: I (Q, ) 8 3 (Q) (norealscattering: 0o ) I self (Q, ) 0 Gas of non-interacting distinguishable particles: a useful “toy model”. No particle correlation: S(Q,)Sself(Q,). Starting from the definitions: 2 N Sself (Q, ) N 1 pn m expiQ ri n Em En i 1 m,n 2 ki2 with E n ; i 1 2 M N n N i 1 expik i ri V one writes: 2k 2 2 Mk BT V e Sself (Q, ) 3 d k 8 2MkBT 2 2 k '2 2 k 2 2M 2M 3/ 2 dr d k ' V expiQ k 'k r After some simple algebra: Sself (Q, ) d k 2 M k T 2 3 / 2 B 2k 2 2Q 2 2k Q exp 2M M 2MkBT 2 2 Q 2M 1 exp 2 2 1 2 Q k BT 2 Q M k BT M recoil Doppler broadening Very important for epithermal neutron scattering! Coherent inelastic scattering from liquids a.k.a. “Neutron Brillouin Scattering”: the acoustic phonons become pseudo-phonons (damped, dispersed). A new undispersed excitation appears too. Very complex, not discussed here. Brillouin triplet(in theQ 0 hydrodynamic limit and for not toolarge) : QQ 2 S (Q) exp(Q q - τ ) ( r (q)) p 2 4 2 2 ( cQQ) Q Q 0 S (Q)( 1) Ds, QQ 2 Ds, Q Q 2 2 4 S Liquid Al g(r) Liquid Ni S(Q) Incoherent inelastic scattering from liquids: the elastic component becomes quasi-elastic (diffusive motions), not discussed here in great detail. for ω and Q not toobig : DQ2 ( ) 2 D 2Q 4 1 On the contrary, the inelastic component is not too dissimilar from the crystal case (pseudophononic excitations). Starting from the well-known: d 2 INC k ' dt exp i t ddE' INC 4 N k 2 N exp iQ R n (0)exp iQ R n (t ) n 1 it is possible to show (Rahman, 1962) that: d 2 INC k ' dt L 2L exp i t exp ( ) Q L (t ) L 1 ddE' INC 4 k 2 INC k ' dt 2 exp i t exp Q 1 (t ) 4 k 2 where we made use of the Gaussian approximation in Q. The t-dependent factor has apparently a tough aspect: 1 (t ) 2M n 0 d 1 cost i sin t f liq coth 2 k BT but it is actually equal to Q-2[B(Q,0)-B(Q,t)]. Then fliq() has to be analogous to g() in solids… Surprising! Let’s study it, starting from the velocity selfcorrelation function of an atom in a crystal: cvv(t). Expanding in normal modes through the Bloch theorem, one gets (in the isotropic case): 1 cvv (t ) N N n 1 v n (0) v n (t ) 3 v1, z (0)v1, z (t ) 3 cos t i sin t d g coth 2M n 0 2 k BT 3 d 2 B(Q, t ) 2 Q dt 2 t 0 It applies to fliq() too. Using the fluctuationdissipation theorem, linking Re[cvv(t)] with Im[cvv(t)], one writes: 4M d tanh (2kBT ) f liq ( ) 3 1 Rec vv (t )cos t dt 0 However, there is a property distinguishing fliq() from g(): dt Rec 0 vv (t ) 3D 3 lim t 2M d D R (t ) t f liq (0) kBT 2 1 where D is the self-diffusion coefficient, while g(0)=0. from CMD, T=14.7 -1 fliq(E) (meV ) 0.09 0.06 0.03 0.00 0 5 10 15 20 25 30 E (meV) -1 Sself (Q,E) (meV ) 0.03 0.02 0.01 from CMD, T=14.7 via Gaussian approx. exp., T=14.3 K 0.00 -10 0 10 20 30 E (meV) 40 50 60 Example: liquid parahydrogen, measured on TOSCA at T=14.3 K (Celli et al. 2002) and simulated through Centroid Monte Carlo Dynamics (Kinugawa, 1998). 5) Vibrational spectroscopy from molecules chemical-physical spectroscopy: studying the forces that: - bind the atoms in a molecule [covalent bond: E400 KJ/mol]. - keep the functional groups close to one another [hydrogen bond: E20 KJ/mol]. - place the molecules according to a certain order in a crystalline lattice [molec. crystals: E2 KJ/mol]. Wide range of energies! Here only intra-molecular modes (vibrational spectroscopy). Cross-section summary 2 tot 4 bˆ ; coh 4 bˆ 2 inc 2 2 4 bˆ bˆ H case (ideal incoherent scatterer): inc=80.27 b, coh=1.76 b Proton selection rule D case (quite different): inc=2.05 b, coh=5.59 b Then only incoherent scattering will be considered in the rest of this talk! Comparing various spectroscopies (neutron)10-28 m2/molec. (Raman)10-32 m2/molec. (IR)10-22 m2/molec. Why neutron spectroscopy ? 1. In Raman polarizability generally grows along with Z: possible problems in detecting H. 2. In IR (sensitive to the electric dipole) the H-bond gives rise to a large signal, but it is distorted by the so-called electric anharmonicity (not vibrational). 3. Molecules with elevate symmetry: many modes are optically inactive (e.g. in C60 up to 70%!). 4. Direct relationship between neutron spectra and vibrational eigenvectors. Conclusions Neutron spectroscopy is complementary to optical spectroscopies (Raman and IR) and is often essential for studying proton dynamics! Example: nadic anhydride (C9H8O3) on TOSCA Molecular vibrations and normal modes Polyatomic Molecules: N atoms instantaneously in the positions {rα}, vibrating around their equilibrium positions {rα0}: rα= rα0+uα Normal modes 3 traslations 3 rotations (2 if linear) 3N-6 vibrations (3N-5 if linear) Translations elimination (center-of-mass fixed): αmαrα= αmαrα0 =R αmαuα=0 Rotations elimination (small oscillations): αmαrαvα= J=0 αmα rα0tuα αmα rα0uα=cost.0 The normal modes of a molecule can be classified according to the character of the atomic motions, starting from the symmetry of the equilibrium configuration of the molecule (group theory). General Theory 1 s T M ijui u j (rigorous) of normal modes 2 i, j with s d.o.f. qi: 1 s U K ijui u j (harmonicapprox.) ui=qi-qi0 2 i, j One gets s Lagrange equations: Oscillating test solutions: Characteristic equation : (in general one has s real and positive roots: 1,… s) Eigenvectors aj(s): s M u K u ij j ij j 0 j u j aj exp(it ) K ij 2 M ij 0 K s j (s) 2 (s) M a ij ij j 0 General solution: u j a (j ) C ( ) exp(i ( )t ) α (m j ) 1/2 a (j )Q ( ) α where theQ ( ) - th normalmode is given by : ( ) ( ) 2 ( ) Q Q 0 Example: normal modes in H2O a. Symmetric stretching b. Bending c. Anti-symmetric stretching Normal mode quantization 1 1 ( ) 2 H cl Qi ( ) 2 Qi( ) 2 2 i,α 2 α i fα 2 1 2 ( ) 2 ( ) 2 ˆ H Qi ( ) 2 2 α i 1 Qi fα ( ) ( ni( ) 1 / 2) α i 1 Mean square amplitude: quantized ( ) 2 i Q ( ) 1 ( ) ni Qi( ) 2 2 0 2 ( ) Diffusion from a harmonic oscillator The mono-dimensional harmonic oscillator is then the simplified prototype of the true intra-molecular vibrations: ~1000 cm-1 <0<4400 cm-1 (H-H): 2 Q 1 1 S (Q, E ) exp 2 E exp cot h 2 0 20 Q 2 ( E n0 ) I n 1 2 sinh( n 0 0 2 0 ) Typical experiment : T=20 K (i.e. 14 cm-1<< 0) then: 1 2 0 70 cot h(12 0 ) 1, sinh(12 0 ) 1 n Q 2 1 x 1 I n ( x) , 1 2 0 sinh(2 0 ) n! 2 from which: S (Q, E ) exp Q u 2 Q 2 0 n 0 2 u 2 n! 0 n ( E n0 ), where u20 is the mean square displacement (at T=0). 0.16759 u (nm ) . 1 0 2 0 (amu) (cm ) 2 2 Again on the harmonic oscillator Mass problem: what is μ in a molecule? It depends on all the atomic masses, but MH obviously plays a primary role! However, in general, μMH . Elastic Line: there is no exchange of energy between oscillator and neutron, then n=0. It is intense, but it decreases rapidly with Q. Then it will be neglected: S el (Q, E ) exp[ Q 2 u 2 ] ( E ) 0 Fundamental: for n=1 there is a peak centered at 0, while in Q one gets a competition between the Debye-Waller factor and the term Q2u20 : S n 1 (Q, E ) exp[ Q 2 u 2 ] Q 2 u 2 0 0 (E ) 0 The maximum of Sn=1(Q,E) appears at Q2=u20. So, the ideal measurement conditions for H are: k1<<k0 k0Q for any value of E. Namely: 2mn 2 k1 2 E k12 u 2 0 (cm1 )mn (amu) 0.16759 0.16759 1 1 (amu) (cm ) Overtones: excitations from the ground state (n=0) to states higher than the first (i.e. n=2,3…): S n 1 (Q, E ) exp Q u 2 2 0 Q 2 u 2 0 n! n ( E n0 ). considering that: 2mn k Q 2 E, 2 1 2 one obtains: u 2 0 20 , n 1 1 n n S n 1 (Q, E ) exp ( E n0 ). n! The relative intensity of the overtones (with respect to n=1) quickly decreases along with μ. It is important to separate the high-frequency fundamental excitations from the overtones. 1.2 n=1 ZrH2, T=20 K from J. Tomkinson (ISIS) 1.0 S(Q,E) (arb. units) Example: fundamental and overtones in ZrH2, almost a harmonic oscillator (threedimensional). 0.8 0.6 n=2 0.4 n=3 0.2 0.0 0 1000 2000 3000 -1 E (cm ) 4000 Anharmonicy Ideal vibrational model: set of decoupled harmonic oscillators (normal modes). Anharmonicity: breaking of the harmonic approximation, implying inseparability and mixing of normal modes. In practice overtones are not simple multiples of the fundamental frequency any more, i.e. there is an anharmonicity constant, . One often has that >0 (e.g. in the Morse potential). 0,(0n) ne 1 (n 1) n0,(01) In practice, in real molecules one uses a pseudoharmonic approach in which the structure factor for a single atomic species is approximated by: Qi u 2 (tot) ij 2 S (Q, E ) exp Qi u Qj ij n! ij k n 0 (k) ij ( E 0(k)n ) where n labels the sum over the overtones and k the multi-convolution in E over the normal modes, from which: u 2 (tot) ij u k 2 (k) ij Qj n 6) Incoherent inelastic scattering from molecular crystals External molecular modes So far only isolated molecules have been dealt with, having a fixed center-of-mass (no recoil). In reality, at low temperature, one observes molecular crystals kept together by intermolecular interations: weak (van der Waals), medium (H bond), or strong (covalent). External modes (pk, lattice vibrations and undistorted librations): in general (but not always…) softer than the internal ones (e.g. lattice v. ~150 cm-1). Similarly to what seen for the internal modes, an external structure factor for the molecular lattice can be defined: S ext (Q, E ) exp Qi U 2 ij ( E npk ), Qi U 2 (tot) ij Qj ij n! k n 0 making implicitly use of the decoupling hypothesis between internal and external modes: S (Q, E) Sext (Q, E) Sint (Q, E) (k) ij Qj n using the distributive property of the convolution one gets: S (Q, E ) exp Qi U 2 ij 2 (k') Qi U Q j ij ij k, k' n '! n, n' n' (tot) ij Q j exp Qi u 2 ij Qi u 2 ij n! (tot) ij Qj n (k) ij Q j ( E n' p n ), k' k then for each internal mode k there is also a shifted replica of all the external spectrum {pk’} (phononic branch), but with a strong intensity reduction due to the external Debye-Waller factor: S (Q, E ) S orig (Q, E ) S branc (Q, E ) S orig (Q, E ) exp Qi U 2 ij (tot) ij Q j Sint (Q, E ) At low Q, Sorig(Q,E) is intense and Sbran(Q,E) has a shape similar to that of Sext(Q,E) (but translated). At high Q, Sbran(Q,E) is dominated by the multiphonon terms (difficult to be simulated). Comparison to the mean square displacements worked out by diffraction: 1 Tr u 2 3 (tot) ij 1 Tr U 2 3 (tot) ij of Bragg peaks): Tr U Biso 8 2 2 (tot) ij (cont rolling t heint ensit y Tr u 2 (tot) ij Discrepancies between Biso and the inelastic mean square displacements: static disorder Example: hexamethylenetetramine (C6H12N4) on TOSCA C6H12N4, T=20 K from J. Tomkinson (ISIS) S(Q,E) (arb. units) 5 <--- Lattice modes 4 Pure vibration/libration | | Phonon wing | | | | 3 2 1 0 0 200 400 600 -1 E (cm ) 800 Anisotropy and spherical mean We have seen that, owing to the presence of various normal modes, scattering depends on the orientation of Q with respect to the molecule (anisotropy). Toy-model: 1-D harmonic oscillators with frequency x , all oriented along the x axis(e.g. parallel diatomic molecules and one lattice site only): Sn 1 (Q, E ) exp Q u where : Q u 2 2 Qu Qu cos 2 2 E x QT u x Q QT B x Q Sn=1(Q,E) is maximum for φ=0 (Q||x) and zero for φ=90o (Qx). Similar to E in IR. It is also defined a displacement tensor Bij: Bx u uT In practice the powder spectrum will be a spherical average containing various modes i: S n 1 (Q, E ) exp Q T AQ Q i where : A B i u i u iT i i T B i Q E i θ, One can prove that a good approximation of the spherical mean is given, for the fundamental, by: Q2 S n 1 (Q, E ) exp iQ Tr Bi E i 3 i where: 2 Tr Bi A 1 i Tr A 2 5 Tr Bi This expression is formally identical to the isotropic harmonic oscillator one: all the vibrations are visible, but wakened by a factor 1/3. Example of the anisotropy importance in highly-oriented (>90%) polyethylene –––––––––– c ––––––– Example: lattice modes in highly-oriented polyethylene simulated for TOSCA Qc (calc. by Lynch et al.) Q||c (calc. by Lynch et al.) 7) Some applications to soft matter What is soft matter? Soft matter: it is often macroscopically and mechanically soft, either as a melt or in solution. On a short scale: there is a mesoscopic order together with weak intermolecular force constants [v/(3kBT)1]. It is in between solids and liquids (both for its structure and for its dynamics). It is not yet rigorously defined. Main classes (after Hamley, 1999): polymers, colloids, amphiphiles and liquid crystals. Good picture, but there is still some overlap! What is spectroscopy? •A microscopic dynamical technique: spectral analysis (k,) of a probe, before and after its interaction with a sample. •Absorption (0) or scattering (k, ). •Basic idea: 02/t; |k|2/|r| and 2/t. •Differences: i) probe [e.m. waves: =c|k|, neutrons: =|k|2/(2mn)]; ii) interaction [e.m. waves: Aj, neutrons: (22/mn) b (r)]. E = Ei – Ef Q = ki – kf Main spectroscopic techniques for soft matter i) Nuclear Magnetic Resonance (NMR). ii) Infrared absorption and Raman scattering (IR and Raman). iii) Dielectric Spectroscopy iv) Visible and ultraviolet optical spectroscopy v) Inelastic neutron scattering (INS). Why INS for soft matter? • Limitations of IR and Raman: selection rules (from f|D|i and f|P|i). Group theory. • General problems with optical techniques: i) dispersion and acoustic modes; ii) selection rules; iii) proton visibility; iv) spectral interpretation. INS is always complementary and often essential i) Dispersion and acoustic modes collective modes dispersion: =j(q), con 0<|q|<2/a20 nm-1. What |q| can be obtained through e.m. waves? Green light (E=2.41 eV): |q|=0.0122 nm-10… X-rays are needed (E>1 KeV): IXS. Acoustic modes: ac(|q|0)=cs|q|0. Thermal neutrons: (E=25.85 meV): |q|=35.2 nm-1. ii-iii) Selection rules and proton visibility High symmetry: many modes are optically inactive (C60: 70%!). Neutrons: pseudo-selection rule for H (H=81.67 barn >>x1-8 barn). Isotopic substitution: HD (D=7.63 barn). Proton visibility in Raman: Tr (P) grows along with Z. Proton visibility in IR: strong signal for H-bonds (e.g. O-H), but there is also the electric anharmonicity (distortions). iv) Spectral Interpretation Direct interpretation of the spectral line intensities: vibrational eigenvectors (IR and Raman: f|D|i, f|P|i). Example: one-dimensional harmonic oscillator (at T=0): S (Q, E ) exp Q u 2 Q 2 0 n 0 2 u 2 n! 0 n ( E n0 ), + - + - K [C6H4(COOD)COO] K [C6H4(COOH)COO] 5 da D. Colognesi, TOSCA (OH) S(Q,E) (a.u.) ¦ 4 (OH) ¦ 3 2 1 0 25 50 75 100 125 150 175 200 E (meV) Example: isotopic substitution in potassium hydrogen phthalate. Two hydrogen-bond modes are clearly pointed out. Would you like to know more? (from easy to difficult) “Introduction to the Theory of Thermal Neutron Scattering” by G. L. Squires (1978). “Vibrational Spectroscopy with Neutrons” by P. C. H. Mitchell et al. (2005). “Molecular Spectroscopy with Neutrons” by H. Boutin and S. Yip (1968). “Neutron Scattering in Condensed Matter Physics” by A. Furrer, J. Mesot and T. Straessle (2009). “Slow Neutrons” by V. F. Turchin (1965). “Theory of Neutron Scattering from Condensed Matter I” by S. W. Lovesey (1984). Acknowledgements Many thanks to: Dr. R. Senesi (Univ. Roma II) for the kind invitation to talk. The audience for its attention and interest.