High Resolution Spectroscopy with the Neutron Resonance Spin Echo Method
vorgelegt von Diplom-Physiker Felix Groitl aus Erlangen
von der Fakultät II - Mathematik und Naturwissenschaften der Technischen Universität Berlin zur Erlangung des akademischen Grades Doktor der Naturwissenschaften Dr. rer. nat.
genehmigte Dissertation
Promotionsausschuss: Vorsitzender: Prof. Dr. M. Kneissl Gutachter: Prof. Dr. D. A. Tennant Gutachter: Prof. Dr. P. Böni Gutachter: Dr. K. Habicht
Tag der wissenschaftlichen Aussprache: 18.12.2012
Berlin 2013
D 83
Abstract
The first part of this thesis is dedicated to explore new territory for high resolution Neu- tron Resonance Spin Echo (NRSE) spectroscopy beyond measuring lifetimes of elementary excitations. The data analysis of such experiments requires a detailed model for the echo amplitude as a function of correlation time. The model also offers guidance for planning NRSE experiments in terms of a sensible choice of parameters and allows predicting quan- titatively the information content of NRSE spectroscopy for line shape analysis or energy level separation. Major generalizations of the existing formalism, developed in this thesis, allow for violated spin echo conditions, arbitrary local gradient components of the dispersion surface and detuned parameters of the background triple axis spectrometer (TAS) giving rise to important additional depolarizing effects, which have been neglected before. Fur- thermore, the formalism can now be applied to any crystal symmetry class. The model was successfully tested by experiments on phonons in a high quality single crystal of Pb and the results demonstrate the stringent necessity to consider second order depolarization effects. The formalism was subsequently extended to analyze mode doublets. As a major step for- ward, detuning effects for both modes are taken into account here. The model was verified by NRSE measurements on a unique tunable double dispersion setup. The results prove the potential of NRSE spectroscopy to resolve mode doublets with an energy separation smaller than the typical energy resolution of a standard TAS. The second class of NRSE experiments was dedicated to line shape analysis of temperature dependent asymmetric line broadening. Inelastic NRSE spectroscopy was performed on two systems, Cu(NO ) 2.5D O and Sr Cr O . For this purpose high quality single crystals 3 2· 2 3 2 8 of Cu(NO ) 2.5D O were grown in the course of this thesis. As a proof of principle the 3 2· 2 results clearly show that the NRSE method can be used to detect temperature dependent asymmetric line broadening. For the first time this effect was measured with NRSE. The second major part of this thesis was the upgrade of the NRSE option of FLEXX at the BER II neutron source at HZB, Berlin. Redesigned NRSE bootstrap coils allow for a more efficient exploitation of the larger beam cross section, given due to the overall upgrade of FLEXX. Higher accessible coil tilt angles enable measurements on steeper dispersions. The newly designed spectrometer arms result in a more compact instrument, enabling direct beam calibration measurements for the entire accessible wavevector range. In combination with higher coil tilt angles the accessible Q-range in Larmor diffraction geometry is en- hanced. Extensive calibration measurements were performed and the results clearly demon- strate the reliable performance of the new NRSE option, now available for the broader user community at FLEXX.
Zusammenfassung
Der erste Teil dieser Arbeit erkundet Neuland für die hochauflösende Neutronen Resonanz Spin-Echo (NRSE) Spektroskopie über die Messung von Lebensdauern elementarer Anre- gungen hinaus. Die Datenanalyse solcher Experimente erfordert ein detailliertes Modell der Echoamplitude als Funktion der Korrelationszeit. Das Modell bietet zudem eine Hilfestel- lung für die Experimentplanung in Bezug auf die Wahl der Parameter. Des Weiteren erlaubt es eine quantitative Vorhersage des Informationsgehaltes von NRSE Messungen, z.B. im Be- reich der Linienformanalyse oder der Aufspaltung von Anregungsenergien. Wichtige, in die- ser Arbeit entwickelte Verallgemeinerungen des existierenden Formalismus berücksichtigen Depolarisationseffekte durch Spin-Echo-Bedingungen, die nicht exakt erfüllt sind. Lokale Gradienten der Dispersion mit einer Orientierung, die nicht parallel zum Wellenvektor q sein muss, und geringfügige Abweichungen der Parameter des Dreiachsen-Spektrometers (DAS), welche zu zusätzlichen, zuvor vernachlässigten Depolarisationseffekten führen, wer- den jetzt berücksichtigt. Ferner kann der Formalismus nun auf beliebige Symmetrieklassen angewendet werden. Das Modell wurde erfolgreich mit Experimenten an Phononen in einem Pb-Einkristall mit exzellenter Mosaizität überprüft. Die Ergebnisse demonstrieren die Not- wendigkeit, Depolarisationseffekte zweiter Ordnung zu berücksichtigen. Der Formalismus wurde dahingehend erweitert, die Analyse von Anregungsdubletts zu er- möglichen. Dadurch werden nun Dejustageeffekte für beide Anregungen berücksichtigt. Das Modell wurde durch elastische und inelastische NRSE Messungen an einem eigens dafür entwickelten Aufbau, welcher künstlich aufgespaltene Moden realisiert, überprüft. Die Er- gebnisse zeigen das Potenzial der NRSE Spektroskopie, Anregungsdubletts aufzulösen, deren Energieaufspaltung unter der Energieauflösung eines Standard-DAS liegt. Weitere hier durchgeführte NRSE Experimente widmeten sich der Linienformanalyse tem- peraturabhängiger asymmetrischer Linienverbreiterungen. Dafür wurden inelastische NRSE Messungen an Cu(NO ) 2.5D O sowie an Sr Cr O durchgeführt. Hierfür wurden eigens 3 2· 2 3 2 8 hochwertige Cu(NO ) 2.5D O-Einkristalle gezüchtet. Die Ergebnisse zeigen deutlich, dass 3 2· 2 die NRSE Methode in der Lage ist, eine temperaturabhängige asymmetrische Linienverbrei- terung zu bestimmen. Erstmalig wurde dieser Effekt mit NRSE gemessen. Im Zuge dieser Arbeit wurde außerdem die NRSE-Option des kalten Dreiachsen-Spektrome- ters FLEXX an der Neutronenquelle BER II am HZB, Berlin, aufgerüstet. Die dafür neu ge- fertigten NRSE Bootstrap-Spulen erlauben eine effektivere Ausnutzung des größeren Strahl- querschnitts, der durch das FLEXX Upgrade zur Verfügung steht. Höhere erreichbare Spu- lenkippwinkel bieten zusätzlich Zugang zu steileren Dispersionen. Das durch die neu ent- wickelten Spektrometerarme kompaktere Instrument ermöglicht Kalibrationsmessungen im direkten Strahl für den gesamten zugänglichen Wellenvektor-Bereich. In Kombination mit höheren Spulenkippwinkeln wird der zugängliche Q-Bereich in der Larmor Diffraktionsgeo- metrie vergrößert. Umfangreiche Kalibrationsmessungen zeigen deutlich die Zuverlässigkeit und Leistungsfähigkeit der neuen NRSE-Option, die nun einer breiten Nutzerschaft zu Ver- fügung steht. Table of contents
1 Introduction 1
2 NRSE resolution theory 5 2.1 Principle of inelastic neutron spin echo spectroscopy ...... 5 2.2 Principle of neutron resonance spin echo ...... 8 2.2.1 The π-coil...... 8 2.2.2 NRSE instrument with 4 π-coils ...... 10 2.2.3 Spinechophononfocusing...... 12 2.3 ExtendedNRSEresolutionfunction ...... 16 2.3.1 Generalized spin echo phase - violated spin echo conditions . . . . . 17 2.3.2 The τ dependenceofthepolarization...... 24 2.3.3 Quantitative description of depolarization due to sample imperfections 28 2.3.4 Quantitative description of depolarization due to curvature of the dis- persionsurface ...... 31 2.3.5 Quantitative description of depolarization due to sample imperfections andcurvatureofthedispersionsurface ...... 35 2.3.6 Dispersion surface not coinciding with the center of the TAS resolution ellipsoid...... 36 2.3.7 Numericalexamples ...... 40 2.4 Experimentaltest...... 43 2.5 Summary ...... 46
3 NRSE investigations on split modes 47 3.1 Two modes within the TAS resolution ellipsoid - Simplified NRSE model . . 47 3.2 Second dispersion surface within the TAS resolution ellipsoid - General model 51 3.3 Experimental verification ...... 54 3.3.1 Experimentalsetup...... 54 3.3.2 Niobium dispersion models ...... 55
i ii Table of contents
3.3.3 Elastic measurements on split modes ...... 56 3.3.4 Inelastic measurements on split modes ...... 61 3.4 Summary ...... 67
4 NRSE line shape analysis 69 4.1 Asymmetric line shape of excitations in Cu(NO ) 2.5D O...... 70 3 2· 2 4.1.1 Properties of Cu(NO ) 2.5D O...... 70 3 2· 2 4.1.2 Sample deuteration and growth of single crystals ...... 71 4.1.3 InelasticNRSEmeasurements...... 73
4.2 Asymmetric line shape of excitations in Sr3Cr2O8 ...... 82
4.2.1 Properties of Sr3Cr2O8 ...... 82 4.2.2 InelasticNRSEmeasurements...... 84 4.3 Summary ...... 88
5 Upgrade of the NRSE option at FLEXX 89 5.1 Bootstrapcoils ...... 90
5.1.1 B0 coils ...... 91 5.1.2 Cooling circuit ...... 94 5.1.3 RFcoils...... 95 5.2 Spinechoinstrumentarms...... 97 5.2.1 Magnetic shielding ...... 98 5.2.2 Coupling coils ...... 101 5.2.3 Motorsandencoders ...... 102 5.3 Calibration of the new NRSE option at FLEXX ...... 102 5.3.1 Calibration of currents and HF voltage ...... 103 5.3.2 Spinechocurveandechopoint ...... 106 5.3.3 Phasestability ...... 111 5.3.4 Calibration of coil tilt angles ...... 112 5.4 Summary ...... 116
6 Conclusion and perspectives 119
Appendix 123
References 139 Chapter 1
Introduction
To investigate the structure and dynamics of matter on the atomic scale, neutron physics and the field of neutron scattering have produced many distinguished methods and instru- ments. The unique properties of thermal neutrons - the electrical charge neutrality, the magnetic moment, the wavelength being of the order of inter-atomic distances, a high pene- tration power for many materials and an energy of the order of excitations in condensed matter - emphasizes it as an excellent probe for structure and dynamics in matter. Energy resolving neutron scattering techniques have made essential contributions to the un- derstanding of matter, e.g. the development of the triple axis spectrometer (TAS) enabled measuring dispersive excitations. Allowing a reasonable intensity for inelastic experiments, the energy resolution of a conventional TAS is limited to 0.1meV. However, the investigations of linewidths of dispersive excitations and slow dynamics require much higher resolution in the range of µeV. This is only accessible using specially designed, highest resolution spec- trometers. Neutron spin echo (NSE) introduced by Mezei [1, 2] is one of the outstanding methods providing the necessary energy resolution in the µeV-range. The method of neutron resonance spin echo (NRSE), developed by Gähler and Golub [3], in combination with TAS, as suggested by Mezei [5], allows to measure linewidths of dispersive excitations with highest resolution. This is still a young technique getting more and more established. Furthermore, Rekveldt [44] pointed out that the technique can be used to perform high precision mea- surements in Larmor diffraction geometry.
A major part of this thesis focuses on methodological development and experiments, explor- ing new territory of NRSE and pushing the method beyond “standard” linewidth measure- ments. One class of experiments, which demands highest resolution, are those seeking to resolve mode doublets with an energy separation smaller than the typical energy resolution of standard TAS. The data analysis of such experiments requires a model for the echo ampli-
1 2 Introduction tude as a function of the correlation time. A model first developed in [6, 7] expands the spin echo phase to second order and includes second order effects arising from the TAS resolution function, sample imperfections and the curvature of the dispersion surface. However, the model is restricted to cubic systems. In this thesis a more generalized formalism is devel- oped allowing for all crystal symmetry classes. Furthermore, violated spin-echo conditions, arbitrary oriented local gradients of the dispersion surface and a detuning of the instrument parameters are included in the generalized description. The derivation of the model, the resolution function and experimental tests on phonons in Pb are discussed in chapter 2. Subsequently the model is extended to include the interesting case of mode doublets. Exper- imental tests were performed on Nb as a well understood model system. A unique tunable double dispersion setup, allowing to generate artificially split modes, was realized. The model and the experimental verifications are presented in chapter 3. Another class of new NRSE-TAS experiments investigates temperature dependent asymme- tric line broadening. This effect has recently been observed in Cu(NO ) 2.5D O, a model 3 2· 2 system for a 1-D bond alternating Heisenberg chain [53], and Sr3Cr2O8, a 3-D gapped quan- tum spin dimer [65], using time-of-flight (ToF) and triple axis spectrometer (TAS) neutron scattering instruments. In the course of this thesis inelastic NRSE spectroscopy was per- formed on both material systems. For the first time the effect of temperature dependent asymmetric line broadening was measured with NRSE. The particular advantage of the NRSE method is the direct access to the line shape, since there is no convolution of the sig- nal with the resolution function of the spectrometer. The results are discussed in chapter 4.
The instrumental part of this thesis was the upgrade of the NRSE option available at the cold TAS FLEXX at BER II at HZB, Berlin. The upgrade of the NRSE option benefits from the new features available at FLEXX after its major upgrade. In the course of this thesis the NRSE spectrometer arms were redesigned to enable a more compact construction of the overall NRSE option. This gives access to higher scattering angles, allowing a larger Q-range for Larmor diffraction experiments. In combination with the increased instrument area, the direct beam geometry is now accessible for the entire wavelength-range of the NRSE-TAS instrument. Furthermore, new NRSE bootstrap coils were manufactured in collaboration with the Max Planck Institute For Solid State Research, Stuttgart. The new design of the NRSE coils allows for larger beam cross sections exploiting the increased polarized neutron flux. In addition, larger accessible coil tilt angles enable measurements of steeper dispersions. All necessary calibration measurements for the NRSE option were performed with the new instrument components. The individual experimental tests demonstrate the high quality, the high manufacturing accuracy and the excellent performance of the new coils. Calibration Introduction 3 tables for currents and impedance matching, essential for a successful user operation, were measured and implemented successfully. Finally, spin echo measurements were performed, demonstrating high polarization for a wide range of spin echo times virtually independent of coil tilt angle and thus excellent performance of the NRSE option. All measurements are presented in chapter 5. 4 Introduction Chapter 2
NRSE resolution theory
In this chapter the principles of classical neutron spin echo [5], the neutron resonance spin echo technique [3, 4] and the so called “phonon focusing” method [5], used to investigate dispersive excitations, will be introduced. The theory of inelastic neutron spin echo spec- troscopy will be extended to second order including depolarizing effects [6]. In order to develop a more general resolution function, the existent formalism will be extended here by dropping limiting restrictions. The formalism will be subsequently extended to include a detuning of the background TAS spectrometer. At the end of this chapter experimental tests of the model performed at the cold triple axis spectrometer V2/FLEX at BER II at HZB, Berlin, will be discussed. A classical point of view is sufficient to develop the resolution function for inelastic neutron spin echo. Therefore, all principles introduced in this chapter are discussed in a classical frame.
2.1 Principle of inelastic neutron spin echo spectroscopy
In neutron spin echo (NSE), first introduced by Mezei [5], the velocity change of neutrons due to scattering is measured by comparing their Larmor precession angle in magnetic fields before and after the sample. The experimental setup for a classical NSE spectrometer is shown in Fig. 2.1. A polarizer P and an analyzer A are placed before and after the spectrometer, respectively. Propagating in the y-direction the monochromatic neutron beam (∆λ/λ = 10% 20%) is − polarized in the x-direction. Neutrons enter the first precession field region of length L1, with a well defined homogenous magnetic field B1 oriented in z-direction perpendicular to the polarization vector of the beam. In the magnetic field the magnetic moments of the neutrons start to precess with the Larmor frequency ω = γB , where γ = 2π 2.913 103 Hz 1 1 · · G
5 6 2 NRSE resolution theory
Sample B1 B2 polarized A
P beam Detector
x z L1 L2 y 1
0
Polarization -1
Fig. 2.1: Top: Schematic NSE setup. The beam is polarized by a polarizer P. The average polarization decays in the first spectrometer arm. For elastic scattering and equal field integrals the polarization is restored in a spin echo after the second spectrometer arm (black) and analyzed by an analyzer A. For the inelastic case (red) the polarization at the echo point decreases as a function of the spin echo time τ and the phase is shifted. Bottom: The average beam polarization as a function of the instrument length for the elastic case (black) and the inelastic case (red).
is the gyromagnetic ratio of the neutron. A neutron with a velocity v1 needs the time t1 = L1/v1 to pass the first region, where its magnetic moment Larmor precesses by the angle Φ1 = ω1t1. Due to the velocity spread of the incoming beam the distribution of the Larmor precession angles becomes more and more spread out. For a typical configuration of ∆λ B = 300G, L = 1m, λ = 6Å and a wavelength distribution of λ = 10% the spread dΦ is of the order of 102 2π. Since the polarization of the beam is the average of the analyzed spin · component and given by P = cos Φ , it is zero after the first field region. By scattering at h i the sample the neutron velocity changes to v2. The orientation of the second magnetic field
B2 is opposite to B1. Here B2 and B1 denote the magnitudes of the magnetic fields. By passing through the second field region, the magnetic moment of the neutron precesses by an angle of φ = ω t , resulting in a total angle of 2 − 2 1
γB1L1 γB2L2 ΦNSE = (2.1) v1 − v2 after the second spectrometer arm. For elastic scattering (v = v ), a symmetric instrument configuration ( B L = B L ) 1 2 | 1 1| | 2 2| and B~ = B~ , the phase φ will be zero for all velocities after the second field region. 1 − 2 NSE Since all magnetic moments are then oriented in the original direction, this effect is called “spin echo”. As pointed out by Mezei [1, 2] for inelastic scattering processes, the idea of 2.1 Principle of inelastic neutron spin echo spectroscopy 7
neutron spin echo is the use of ΦNSE to determine the energy transfer m ~ω = v2 v2 . (2.2) 2 1 − 2 To get a quantitative relation between ΦNSE and ω, the scattering function S (ω) is consid- ered to be independent of the momentum transfer Q and distributed around a mean energy transfer of ~ω0. The energy and the velocity of the incoming neutron beam then read:
ω = ω0 + dω and v1 =v ¯1 + dv1 (2.3)
Using equation (2.2) v2 can be replaced in equation (2.1):
γB1L1 γB2L2 ΦNSE = (2.4) ~ v¯1 + dv1 − (¯v + dv )2 2 (ω + dω) 1 1 − m 0 q Expanding equation (2.4) to first order in dv1 and dω yields:
γB L γB L v¯ ~ γB L Φ = Φ 1 1 2 2 1 dv 2 2 dω (2.5) NSE NSE,0 − 2 − 3 1 − m 2 (¯v1) (¯v2) (¯v2) where ΦNSE,0 corresponds to the mean velocities v¯1 and v¯2
γB1L1 γB2L2 ΦNSE,0 = (2.6) v¯1 − v¯2 and v¯2 is the mean neutron velocity after the sample
~ 2 v¯2 = (¯v1) ω0. (2.7) r − 2m If the so called spin echo condition
3 B1L1 (¯v1) = 3 (2.8) B2L2 (¯v2) holds, the phase ΦNSE becomes independent of the velocity spread dv1 and dΦNSE is proportional to dω: dΦ = Φ Φ = τ dω (2.9) NSE NSE − NSE,0 − NSE The instrumental constant ~γB1L1 ~γB2L2 τNSE = 3 = 3 (2.10) m (¯v1) m (¯v2) 8 2 NRSE resolution theory is called “spin echo time”. After the second field region the neutron beam passes an analyzer transmitting neutrons polarized in the x-direction. The intensity at the detector is given by:
I I = 0 (1 + P ) (2.11) 2 where I0 is the intensity of the incident neutron beam. The mean beam polarization P at the analyzer, i.e. the echo amplitude AE, is the cosine Fourier transform of the scattering function S (ω)
A = cos Φ = S (dω) cos (Φ dω τ ) d (dω) . (2.12) E h NSEi NSE,0 − · NSE Z As a classical example from quasi elastic scattering let the scattering function S (ω) be a Lorentzian line with a linewidth of Γ:
Γ S (ω) (2.13) ∝ Γ2 + ~2ω2
Equation (2.12) then yields:
S (ω)cos(ωτNSE) dω ΓτNSE A (τ )= = e− ~ (2.14) E NSE S (ω) dω R R 2.2 Principle of neutron resonance spin echo
The method of neutron resonance spin echo (NRSE) was suggested by Gähler and Golub [3, 4] in 1987. NRSE replaces each of the long static fields of the NSE method by two resonance spin flippers with sharp field boundaries placed in a zero field region.
2.2.1 The π-coil
The basic component of NRSE is a resonance flipper called π-coil. Within a resonance spin
flipper two magnetic fields are superimposed: a static field B~ 0 pointing in the z-direction and a rotating field B~ RF in the xy-plane with the frequency ωRF . The superposed magnetic field is described by the vector
B~ 1 (t)=(BRF cos(ωRF t) ,BRF sin(ωRF t) ,B0) . (2.15)
To understand the movement of the magnetic moment of a neutron entering the coil with an initial orientation of the magnetic moment in the xy-plane, it is useful to change the reference 2.2 Principle of neutron resonance spin echo 9
z
y
B1 x
Fig. 2.2: Principle of the π-coil. The incoming neutron polarization vector (green) precesses around the effective magnetic field B1 and leaves the coil πv0 with a new orientation (blue). Setting B1 = γd the polarization vector performs a π-flip and stays in the scattering plane
frame to a system rotating with the frequency ωRF around the z-axis. The effective field will be [10]: ω B~ = B , 0,B RF . (2.16) 1 RF 0 − γ If the frequency of the rotating field is tuned to the Larmor frequency of the static field
(ωRF = γB0), the z-component vanishes and the static field BRF remains. Thus, the magnetic moment of the neutron will precess around BRF with the frequency ωRF = γBRF .
The magnitude of BRF is chosen such that the magnetic moment precesses by an angle π while the neutron passes the coil: πv B = 0 (2.17) RF γd where v0 is the velocity of the neutron and d is the thickness of the coil. Therefore, after leaving the coil the magnetic moment is again in the xy-plane (see Fig 2.2).
In the lab-system the neutron enters the coil at a time t1. At this time the rotating field has a phase of ΦRF = ωRF t1 and while passing the coil, the magnetic moment precesses an angle π around B~ . Simultaneously B~ precesses around the z-axis by an angle of ω d . 1 1 RF v0 Thus, after leaving the coil the phase of the magnetic moment is:
d Φ1 = 2ωRF t1 + ωRF Φ0 (2.18) v0 − where Φ0 is the angle between the magnetic moment and the x-axis. Since the realization of a rotating field setup is challenging, it is favorable to use a linearly oscillating field instead. The oscillating field B1 can be divided into two counter rotating + B1 parts B1 and B1− with an amplitude of 2 . Only the part rotating in the same direction as the magnetic moment of the neutron contributes to the π-flip. The counter rotating part 10 2 NRSE resolution theory can be neglected for B B . This causes a small shift of the resonance frequency, the so 0 ≫ 1 called “Bloch-Siegert-Shift” [11] 2 ∆ω BRF = 2 , (2.19) ω 4B0 which can be neglected for frequencies higher than 50kHz.
The resonance condition given by equation (2.17) is only valid for one v0. The magnetic moments of faster (slower) neutrons will precess less (more) than an angle of π. Therefore, the magnetic moment is oriented slightly out of the xy-plane after the coil and the average polarization decreases. The estimates in [12] show that for small velocity spreads the echo amplitude is: π2 ∆v 2 A = 1 M (2.20) E − 4 v 0 where M is the number of π-flippers used. For a velocity spread of 1.5%, which is typical for a TAS using a crystal monochromator, and 8 flippers, the loss of average polarization will be < 1%.
2.2.2 NRSE instrument with 4 π-coils
Using a NRSE instrument with a setup of 4 π-coils (see Fig. 2.3) provides the same function as a conventional NSE spectrometer with the same length and the static magnetic fields twice as strong.
The two coils before the sample region (A and B) have a static field B~ 1 and are driven with a frequency of ω1 = γB1 while the coils C and D after the sample region have the static
field B~ 2 with a frequency of ω2 = γB2. The two fields B~ 1 and B~ 2 are oriented in opposite directions. Therefore, the precession direction of the magnetic moment of the neutron and the rotating fields is opposite to each other. The regions between the coils are zero field regions and the stray fields of the coils have to be minimized.
A neutron polarized in x-direction (Φ0 = 0 in equation (2.18)) with a velocity v1 enters the
first coil at the time tA. After passing the coil, the phase angle of the magnetic moment is
d ΦA = 2ω1tA + ω1 . (2.21) v1
In the zero field region, between the coils A and B, the magnetic moment of the neutron does not change its phase angle and enters the second coil at the time t = t + L1 . Therefore, B A v1 the phase angle of the magnetic moment after the second coil is:
d L1 ΦB = 2ω1tB + ω1 ΦA = 2ω1 . (2.22) v1 − v1 2.2 Principle of neutron resonance spin echo 11
zero-field regions
Spin
x z y AB CD
L1 L2 d sample region
Fig. 2.3: Schematic drawing of a NRSE setup with 4 π-coils: The orienta- tion of the magnetic moment of the neutron entering each coil is illustrated in green. Within the coil the magnetic moment precesses around the ef- fective magnetic field Beff (red) and has a new orientation (blue) leaving each coil. The flight path between the coils are regions of zero magnetic field.
Φ becomes independent from the time t and only depends on the time of flight L1 between B A v1 the two coils. Compared to a conventional NSE instrument with the same magnitude of the magnetic field and the same length, one obtains twice the phase angle. In other words, the two π-coils give the same Larmor precession angle as a NSE spectrometer arm with length
L1 and magnetic field 2B1.
During the scattering process the neutron velocity changes to v2. Since the sample region is a field free region, the magnetic moment enters coil C with the phase ΦB at time tC and leaves coil C with an angle of
d ΦC = 2ω2tC ω2 ΦB. (2.23) − − v2 −
Hence, after the last coil D the phase is
d ΦD = 2ω2tD ω2 ΦC − − v2 − L1 L2 (2.24) = 2ω1 2ω2 v1 − v2
= ΦNRSE = 2ΦNSE. 12 2 NRSE resolution theory
Therefore, the previous discussion on spin echo applies directly to NRSE if B1 and B2 ω1 ω2 are substituted by 2 γ and 2 γ , respectively. However, in a typical NRSE experiment the minimum accessible spin echo time is determined by a frequency of 50 kHz in the π-coils, since for lower frequencies the effect of the above mentioned Bloch-Siegert-Shift (see equation (2.19)) increases leading to a strong depolarization.
Enhancement of NRSE by the bootstrap technique
The so called bootstrap technique was first introduced in [4]. By replacing each π-coil by a pair of resonance coils with the static fields of each pair oriented in opposite directions, the effective precession angle and hence the resolution is doubled. Within this configuration the first coil within a bootstrap coil provides a return path for the magnetic flux of the second coil and vice versa. This reduces the stray field outside the coils significantly. The phase angle after four bootstrap coils, each containing a series of N resonance coils, is
L L Φ = 2N ω 1 ω 2 = NΦ (2.25) N 1 v − 2 v NRSE 1 2 All existing NRSE spectrometers operate with double bootstrap coils (N = 2). A discussion of the limits of the bootstrap technique can be found in [4].
2.2.3 Spin echo phonon focusing
A triple axis spectrometer (TAS) is a classic instrument to determine dispersion curves in the entire Brillouin zone. However, the instrumental resolution of a TAS is in rare cases sufficient to measure the natural linewidth of excitations and in addition needs to be known very accurately at the inelastic signal. NSE and NRSE spectrometers provide the required high resolution in the µeV range. Mezei [5] suggested to combine TAS and spin echo spectrometers and to use the tilted field technique, which is necessary to measure linewidths of dispersive excitations. For this technique to work, the field boundaries need to be perpendicular to the flight path of the neutrons. Then the phase Φ(ω) is proportional to the energy transfer ω and independent of the momentum transfer q. In the quasielastic case with a mean excitation energy of zero ∆λ and a broad wavelength distribution λ = 10%, a relaxed q resolution does not affect the energy resolution. However, for dispersive excitations where the excitation energy ω is a function of q, ω = ω (q), a finite momentum resolution results in a spread in ω and hence in Φ(ω) even for sharp excitations with zero linewidth. To allow measurements of the intrinsic linewidth of excitations, the spin echo phase needs to be tuned to the slope ~ qω ∇ of the dispersion (see Fig. 2.4). Hence, all scattering events lying within the resolution 2.2 Principle of neutron resonance spin echo 13
ω
ω0
q0 q Fig. 2.4: The resolution function of the TAS selects a certain region in the (q,ω)-space. The lines of constant Larmor phase (Φ = const, pale blue) have to be parallel to the dispersion curve. This can be reached by inclining the precession field boundaries. ellipsoid of the background TAS and on the dispersion surface have the same spin echo phase Φ(ω, q). As pointed out by Mezei [5] the spin echo phase can be tuned to the slope of the dispersion surface by tilting the boundaries of the precession fields relative to the neutron flight path. In such a configuration, for a linear dispersion with zero linewidth, all scattered neutrons have the same total Larmor precession phase. A finite linewidth would cause a spread in Φ. If the lineshape F (ω) within the resolution ellipsoid is independent of q (S (q,ω) = S (q0) F (ω)) and significantly smaller than the width of the resolution ellipsoid, Φ is proportional to the Fourier transform of the line shape as explained above.
Spin echo conditions
In this section the spin echo conditions for the phonon focusing technique are introduced briefly in a classical picture. The discussion of an alternative quantum mechanical framework where Φ is a phase shift between two spin states can be found in [8, 13]. The quantum mechanical approach shows that the spin echo time τNSE is equal to the correlation time in the time dependent van Hove density-density correlation function [9]. Assuming that the variation of the scattering law on the dispersion surface depends on ω (q) only, S (q,ω) can be considered as a function of energy deviation from the dispersion surface: S (q,ω)= S (ω (q) ω (q)) . (2.26) − 0 In order to measure equation (2.26) the Larmor precession angle needs to be proportional to (ω (q) ω (q)): − 0 Φ Φ = τ (ω (q) ω (q)) = τ∆ω. (2.27) − 0 − − 0 − 14 2 NRSE resolution theory
precession field regions i0 B1 j1 j0 l i1 kI sample kI 0
l1 ti nI Θ B2 1 nI si ni L1 kF Q L2
kF nF Θ2 j2
nF l2 sf tf i2 nf Fig. 2.5: Schematic drawing of the tilted precession field regions in an inelastic NSE setup. The defined Cartesian coordinate systems are used throughout.
Assuming a planar dispersion and expanding the dispersion to first order in (q q ) yields: − 0
∆ω = ω (q) ω (q)= ω (q) [ω (q )+(q q ) qω (q0)] . (2.28) − 0 − 0 0 − 0 ·∇
Using momentum and energy conservation with mean wavevectors kI and kF before and after the scattering:
ki = kI +∆ki and kf = kF +∆kf ~2 ~2 (2.29) ω(q)= k2 k2 and ω (q)= k2 k2 , 2m i − f 0 2m I − F ∆ω is expressed as a function of the deviations from the mean wavevectors ∆ki and ∆kf : ~ ∆ω = ω (q) ω (q)= [(k ∆k ) (k ∆k )] . (2.30) − 0 m I · i − F · f Using the relation for the momentum in a perfect crystal
q = k k G and q = k k G (2.31) i − f − 0 0 I − F − 0 where G0 is the reciprocal lattice vector and combining equations (2.28) and (2.30) yields: ~ ~ ∆ω =∆k k qω (q ) ∆k k qω (q ) (2.32) i m I −∇ 0 0 − f m F −∇ 0 0 2.2 Principle of neutron resonance spin echo 15
The Larmor precession angle after two tilted field regions is
m ωL1L1 cosΘ1 ωL2L2 cosΘ2 Φ = Φ0 +∆Φ = ~ ki cosΘ1 − kf cosΘ2 (2.33) m ω L cosΘ ω L cosΘ A A = L1 1 1 L2 2 2 = 1 2 ~ k n − k n k n − k n i · i f · f i · i f · f where ni,f are the unit vectors normal to the precession field boundaries before and after the sample tilted by an angle Θ1,2 w.r.t. the neutron flight path (see Fig. 2.5). The Larmor frequency is ωL1,2 = NγB1,2 with N = 1 for conventional NSE, N = 2 for NRSE with 4 flipper coils and N = 4 for NRSE in bootstrap mode (8 flipper coils). Using
A1 A2 Φ0 = (2.34) k n − k nf I · i F · and expanding the total Larmor precession angle to first order yields
A1 A2 Φ Φ0 = (∆ki ni)+ (∆kf nf ) . (2.35) − −(k n )2 · (k n )2 · I · i F · f For equation (2.27) to hold, the coefficients in equations (2.32) and (2.35) have to be com- pared. The result for the normal vectors of the precession field boundaries is:
~ ~ m kI,F qω0(q0) m kI,F qω0(q0) ni,f = ~ −∇ = −∇ (2.36) k qω (q ) NI,F m I,F −∇ 0 0 and for the spin echo time
A1,2 A1,2 τ = ~ = . (2.37) (k n ) k qω (q ) (kI,F ni,f ) NI,F I,F · i,f m I,F −∇ 0 0 ·
The condition for the adjustment of the field integrals is then:
2 ~ 2 ω L cosΘ2 (kI ni) kI qω0(q0) cosΘ (k n ) N L1 1 = · m −∇ = 2 I · i I . (2.38) ω L 2 ~ 2 L2 2 cosΘ1 (kF nf ) m kF qω0(q0) cosΘ1 (kF nf ) NF · −∇ ·
If the spin echo conditions given by equations (2.36) and (2.38) are satisfied, τ is the same for both spectrometer arms and the total Larmor phase is independent of the momentum transfer in the presence of a finite dispersion. Note that this is only true for a first order approximation of the total Larmor phase. In the general case the echo amplitude, measured by a spin echo spectrometer combined 16 2 NRSE resolution theory with a background TAS, is given by [6]
1 A = S (Q,ω) R (k , k ) eiΦ(ki,kf )d3k d3k + c.c. (2.39) E N T AS i f i f Z where RT AS (ki, kf ) is the transmission function of the background TAS and Φ(ki, kf ) is the sum of Larmor precession angles before and after the sample. Assuming that the linewidth and S (Q) are independent of Q within the TAS resolution ellipsoid, the integral over the momentum components yields
A S (∆ω) R (∆ω) eiΦd∆ω + c.c. (2.40) E ∝ T AS Z with ∆ω given by equation (2.28). The application of spin echo spectroscopy is only rea- sonable if the linewidth S (∆ω) of the excitation is much smaller than the energy resolution
T (∆ω) of the background TAS. Therefore, RT AS (∆ω) is assumed to be constant. Thus, the result for the echo amplitude is again the cosine Fourier transform of the line shape:
A S (∆ω)cos(τ∆ω) d∆ω (2.41) E ∝ Z 2.3 Extended NRSE resolution function
In this section the resolution function for a NRSE spectrometer with a background TAS is extended by expanding the spin echo phase to second order following the approach of [6] and expressing the resolution function in a covariance matrix formalism. This will include second order effects arising from the TAS resolution function, sample imperfections, the curvature of the dispersion surface and a detuning of the instrument parameters. Major differences between the treatment given in [6] and the treatment given in the present thesis are the following generalizations [14, 15]: 1. Allow for violated spin echo conditions
2. The formalism is extended to treat systems with lower crystallographic symmetry than cubic
3. The local gradient of the dispersion surface ω(q ) may have components out of the ∇ c0 scattering plane, i.e. the center of the TAS resolution ellipsoid must not necessarily be
located at a point with high crystallographic symmetry. Here qc0 is the momentum transfer expressed in a Cartesian coordinate system attached to the reciprocal lattice
system. q0 is transformed into qc0 by the so called B-matrix [16, 17]. The UB matrix formalism is explained in Appendix A. 2.3 Extended NRSE resolution function 17
One advantage of the generalizations made here is the applicability of the formalism to a broader range of systems. The allowance of violated spin echo conditions enables the treat- ment of cases where a second excitation is present within the TAS resolution ellipsoid. In such cases the spin echo conditions are at least violated for one excitation. Investigations on split modes using NRSE are treated in chapter 3. Without these generalizations, depolar- ization effects would be neglected. This would result in an overestimation of the linewidth Γ of the excitation. In addition minor errors in [6] in the part discussing lattice imperfections and the curvature of the dispersion surface are corrected.
2.3.1 Generalized spin echo phase - violated spin echo conditions
A simplified situation is discussed by making the following assumptions:
1. A single dispersion surface is located within the TAS resolution ellipsoid.
2. The center of the TAS resolution ellipsoid coincides with the dispersion surface.
3. The precession field boundaries are exactly perpendicular to the scattering plane.
4. The precession field boundaries can only be tilted around axes that are perpendicular to the scattering plane.
5. All instrument components are ideal, i.e. field boundaries are ideal planes, no stray fields are present and RF flippers provide exact π-flips.
Following [6] and expanding equation (2.35) to second order yields:
A1 A2 Φ(ki, kf ) Φ0 = 2 (∆ki ni)+ 2 (∆kf nf ) − − (kI ni) · (kF nf ) · · · (2.42) A1 2 A2 2 + (∆ki ni) (∆kf nf ) . (k n )3 · − (k n )3 · I · i F · f The aim is to express the total Larmor precession angle in terms of the variable vector
J = (∆ω, ∆kin,y1,y2,z1,z2) with
∆ki = x1i1 + y1j1 + z1l1 (2.43)
∆kf = x2i2 + y2j2 + z2l2 (2.44)
∆k =∆k n (2.45) in i · i 18 2 NRSE resolution theory where the variables of J are defined as in [6]. According to Fig. 2.5 i, j and l are the basic vectors of a right handed coordinate system with i1 and i2 pointing along the direction of the mean wavevectors kI and kF , respectively. First, the dispersion relation is expanded to second order
1 T ω (q )=∆ω + ω (q )+∆q ω (q )+ ∆q H q ∆q (2.46) c c0 c0 ·∇ c0 2 c c| c0 c assuming that the line broadening is the same for every qc within the TAS resolution ellipsoid, i.e. ∆ω is independent from qc. Here Hc is the curvature matrix of the dispersion surface at qc0, which is defined in section 2.3.4. In a real crystal lattice imperfections are present and therefore variations of the lattice vector Gc need to be included. The following relation
∆q′ =∆q ∆G =∆k ∆k ∆G (2.47) c c − c i − f − c with the total wavevector transfer
Qc = Gc + qc (2.48) and
Gc = Gc0 +∆Gc (2.49) leads to
ω (qc) =∆ω + ω (qc0)+∆ki ω (qc0) ∆kf ω (qc0) ·∇ − ·∇ . (2.50) 1 T ∆G ω (q )+ ∆q′ H q ∆q′ − c ·∇ c0 2 c c| c0 c Note again that all variables with the index c are expressed in a Cartesian coordinate system attached to the reciprocal lattice. The relation between these quantities and their expression in the basis of the reciprocal lattice are explained in Appendix A. For now the last two terms will be left unchanged, since effects from sample imperfections and curvature of the dispersion surface will be treated separately in sections 2.3.3 and 2.3.4, respectively. The energy conservation in second order reads ~ ω (q ) = k2 k2 (2.51) c 2m i − f ~ ~ = ω (q )+ k ∆k k ∆k c0 m I · i − m F · f ~ ~ + ∆k2 ∆k2 (2.52) 2m i − 2m f 2.3 Extended NRSE resolution function 19 using q = k k (2.53) c0 I − F and ~ ω (q )= k2 k2 . (2.54) c0 2m I − F Combining equations (2.50) and (2.51) yields
~ ~ ∆ω = k ω (q ) ∆k k ω (q ) ∆k m I −∇ c0 · i − m F −∇ c0 · f ~ ~ 2 2 1 T + ∆k ∆k +∆Gc ω (qc0) ∆q′ Hc ∆q′ . (2.55) 2m i − 2m f ·∇ − 2 c |qc0 c Now unit vectors are defined as
~ ~ kI ω (qc0) kI ω (qc0) ǫ = m −∇ = m −∇ (2.56) i ~ N m kI ω (qc0) I ~ −∇ ~ k ω (q ) k ω (q ) ǫ m F c0 m F c0 f = ~ −∇ = −∇ . (2.57) k ω (q ) NF m F −∇ c0
Since these unit vectors, defined by the crystal properties, in general may have components out of the scattering plane, they cannot be assumed to be equal with the normal vectors of the precession field boundaries ni,f , which are defined by the instrument settings. As mentioned above in this subsection kI,F correspond to a point on the dispersion surface coinciding with the center of the TAS resolution ellipsoid. This assumption will be dropped later in section 2.3.6. In the next step the spin echo times in both spectrometer arms τ1,2 given by the instrument parameters are defined:
A1,2 A1,2 τ1,2 = ~ = . (2.58) (k n )2 k ω (q ) (∆k n )2 N I,F · i,f m I,F −∇ c0 I,F · i,f I,F
Here τ1,2 are not longer identical to the correlation time τ in the time-dependent van Hove density-density correlation function. The inclusion of instrument alignment errors in the tilt angle of the precession field boundaries and the ratio of the field integrals means that in general τ1,2 are not equal in the two spectrometer arms. Multiplying equation (2.55) with
τ2 yields: 20 2 NRSE resolution theory
A2 ~ τ2∆ω = kI ω (qc0) ∆ki k n 2 m −∇ · ( F f ) NF · ~ A2 m kF ω (qc0) 2 −∇ ∆kf −(kF nf ) NF · ~ · ~ + τ ∆k2 τ ∆k2 2m 2 i − 2m 2 f 1 T +τ2∆Gc ω (qc0) τ2∆q′ Hc ∆q′ (2.59) ·∇ − 2 c |qc0 c A2 NI ǫ A2 ǫ = 2 i ∆ki 2 f ∆kf (kF nf ) NF · − (kF nf ) · ~· ~ · + τ ∆k2 τ ∆k2 2m 2 i − 2m 2 f 1 T +τ2∆Gc ω (qc0) τ2∆q′ Hc ∆q′ . (2.60) ·∇ − 2 c |qc0 c
In the general case the unit vectors ǫi,f are expressed in a right handed Cartesian coordinate system with the basis vectors ni,f , ti,f and si,f according to Fig. 2.5. The unit vectors ni,f , ti,f and si,f are the same as u1,2, v1,2 and w1,2 in [6].
ǫi,f = ei,f1ni,f + ei,f2ti,f + ei,f3si,f (2.61)
Here ni,f and ti,f are the normal vector and the tangent of the precession field boundaries lying in the scattering plane, respectively. si,f is the unit vector perpendicular to the scattering plane defined by s = n t . Note that the unit vectors s are identical i,f i,f × i,f i,f since they are both perpendicular to the scattering plane. The components ei,f1, ei,f2 and ei,f3 are defined by
e = ǫ n e = ǫ t e = ǫ s (2.62) i1 i · i i2 i · i i3 i · i e = ǫ n e = ǫ t e = ǫ s . (2.63) f1 f · f f2 f · f f3 f · f As a generalization compared to [6], the case where the gradient vector of the dispersion ω (q ) has a component out of the scattering plane is included. Inserting equation (2.61) ∇ c0 in equation (2.59) yields 2.3 Extended NRSE resolution function 21
A2 ∆kf nf = τ2′′∆ω + τ2′′NI ei1∆kin (k n )2 · − F · f 1 +τ ′′N e ∆k tan θ + y 2 I i2 − in 1 1 cos θ 1 +τ ′′N e z τ ′′N e z 2 I i3 1 − 2 F f3 2 1 τ2′′NF ef2 y2 − cos θ2 ~ ~ 2 2 + τ ′′∆k τ ′′∆k 2m 2 i − 2m 2 f 1 T +τ ′′∆Gc ω (qc0) τ ′′∆q′ Hc ∆q′ (2.64) 2 ·∇ − 2 2 c |qc0 c with τ2 τ ′′ = . (2.65) 2 (e e tan θ ) f1 − f2 2 For the calculation of equation (2.64) the following relations for ∆ki,f expressed in the basis of ni,f , ti,f and si,f were used:
s ∆k = z (2.66) i · i 1 s ∆k = z (2.67) f · f 2 n ∆k = ∆k . (2.68) i · i in
∆k = (x cos θ + y sin θ ) n +( x sin θ + y cos θ ) t + z s (2.69) i 1 1 1 1 i − 1 1 1 1 i 1 i 1 = ∆k n + ∆k tan θ + y t + z s (2.70) in i − in 1 1 cos θ i 1 i 1 1 ∆k = (∆k n ) n + (∆k n )tan θ + y t + z s . (2.71) f f · f f − f · f 2 2 cos θ f 2 f 2 1 ti ∆ki = ∆kin tan θ1 + y1 (2.72) · − cos θ1 and 1 tf ∆kf = (∆kf nf )tan θ2 + y2 . (2.73) · − · cos θ2
Substituting equation (2.64) into equation (2.42) introduces the energy deviation from the dispersion surface due to the linewidth broadening ∆ω into the Larmor phase. This is re- quired to provide the Fourier transform of the scattering function S (Q,ω): 22 2 NRSE resolution theory
φ(k , k ) φ = τ N ∆k τ ′′∆ω + τ ′′N e ∆k i f − 0 − 1 I in − 2 2 I i1 in 1 +τ ′′N e ∆k tan θ + y + τ ′′N e z 2 I i2 − in 1 1 cos θ 2 I i3 1 1 1 τ2′′NF ef2 y2 τ2′′NF ef3z2 − cos θ2 − 1 T +τ ′′∆Gc ω (qc0) τ ′′∆q′ Hc ∆q′ 2 ·∇ − 2 2 c |qc0 c ~ ~ 2 2 + τ ′′∆k τ ′′∆k 2m 2 i − 2m 2 f N N +τ I ∆k2 τ F (∆k n )2 . (2.74) 1 k n in − 2 k n f · f I · i F · f The Larmor phase now takes into account second order effects and the above mentioned generalizations. Contributions from components of the gradient vector of the dispersion surface, which are out of the scattering plane are considered by the terms proportional to ei2,3 and ef2,3. If the gradient vector of the dispersion surface lies within the scattering plane, these terms will vanish since then ei2,3 = 0 and ef2,3 = 0. The terms proportional to
∆Gc and Hc account for contributions arising from sample imperfections and the curvature of the dispersion surface, respectively. The term (∆k n )2 is now substituted in equation (2.74) by using equation (2.64). Since f · f only second order effects are considered and higher order terms are neglected, it is sufficient to consider equation (2.64) to first order only:
1 NI Ci ∆kf nf = ∆ω + ∆kin · −Cf NF NF Cf NI 1 1 1 + ei2 y1 ef2 y2 Cf NF cos θ1 − Cf cos θ2 NI 1 1 + ei3z1 ef3z2 + ∆Gc′ ω (qc0) (2.75) Cf NF − Cf Cf NF ·∇ using
A2 τ2 = (2.76) (k n )2 N f · f F C = e e tan θ (2.77) i i1 − i2 1 C = e e tan θ . (2.78) f f1 − f2 2 2.3 Extended NRSE resolution function 23
1 The term + ∆G′ ω (qc0) considers only the first order terms arising from the lattice Cf NF c ·∇ imperfections. Since lattice imperfections have been introduced from the beginning, it is considered here. This term introduces cross terms between the lattice imperfection variables
∆η, ∆ν and ∆Gc and the variables of the 6 component vector J = (∆ω, ∆kin,y1,y2,z1,z2). 2 2 ∆Gc′ is defined in equation (2.99). ∆ki and ∆kf are substituted in equation (2.74) using equations (2.69) and (2.71):
1 2 ∆k2 =∆k2 + ∆k tan θ + y + z2 (2.79) i in − in 1 1 cos θ 1 1 1 2 ∆k2 = (∆k n )2 + (∆k n )tan θ + y + z2. (2.80) f f · f − f · f 2 2 cos θ 2 2 Inserting equations (2.79) and (2.80) into equation (2.74) and using again equation (2.75) to substitute all ∆k n terms allows to express the total Larmor precession angle as a f · f function of squared and cross terms of the six variables (∆ω, ∆kin,y1,y2,z1,z2). The total Larmor phase can conveniently be expressed in matrix notation [13]:
T 1 T φ(k , k ) φ = τ ′′T J τ ′′J ΨJ i f − 0 2 − 2 2 +τ ′′∆G ω (q ) 2 c ·∇ c0 1 T τ ′′∆q′ Hc ∆q′ + X(∆Gc) (2.81) −2 2 c |qc0 c
Here X(∆Gc) denotes all cross terms introduced by ∆η, ∆ν and ∆Gc. Effects from sample imperfections are treated in section 2.3.3. The components of the 6-dimensional column vector T are:
τ N e T = 1 T = N C 1 C T = I i2 (2.82) 1 − 2 I i − τ f 3 cos θ 2 1 NF ef2 T4 = T5 = NI ei3 T6 = NF ef3. (2.83) − cos θ2 −
The elements of the symmetric (6 6) matrix Ψ are given in Appendix B. × For the special case, where the spin echo conditions are satisfied, and the gradient of the dispersion surface lies in the scattering plane:
ei1 = 1 ei2 = ei3 = 0 (2.84)
ef1 = 1 ef2 = ef3 = 0 (2.85)
Ci = Cf = 1 (2.86) 24 2 NRSE resolution theory and thus
τ2′′ = τ2 (2.87)
τ1 = τ2. (2.88)
The matrix Ψ then reduces to the matrix given in [6]. Since only T = 1 remains non-zero, 1 − the terms, which are linear in J, vanish, leaving only the desired term τ∆ω in the spin − echo phase.
2.3.2 The τ dependence of the polarization
In this section the τ dependence of the polarization for violated spin echo conditions for a sample without lattice imperfections (∆G = 0) and a planar dispersion (H q = 0) is c c| c0 derived. With these assumptions equation (2.81) reduces to
T 1 T φ(k , k ) φ = τ ′′T J τ ′′J ΨJ. (2.89) i f − 0 2 − 2 2 The TAS transmission function is derived, following the approach of Popovici [51]. In order to have a consistent nomenclature, this approach is summarized in Appendix C. It reads:
1 R (k , k ) = exp JT L J . (2.90) T AS i f −2 T AS Substituting the equations (2.89) and (2.90) into the fundamental equation (2.39) yields:
1 T 1 T 6 A = S (Q,ω)exp iτ ′′T J exp J L J d J + c.c. (2.91) E N 2 −2 I n Z with
LI = LT AS + iτ2′′Ψ. (2.92)
Terms in ∆ω higher than linear can be neglected, since the integral over the energy coordi- nate will be dominated by the term exp ( iτ∆ω). The only reasonable application of spin − echo is the situation where ∆ω is very small and hence S (∆ω) is very narrow compared to the TAS energy resolution. Therefore, rewriting equation (2.91) yields:
1 A = S (Q,ω)exp iτ ′′∆ω d∆ω E N − 2 Z T 1 T 5 exp iτ ′′T J exp J L J d J (2.93) × 2 −2 I n Z e e e e e 2.3 Extended NRSE resolution function 25
where J = (∆kin,y1,y2,z1,z2) and T are the 5-D sub vectors of J and T without the energy variable ∆ω, respectively, and LI is the corresponding symmetric (5 5) sub matrix of LI . e e × Following [6] and using the general matrix theorem [18] e ∞ 1 exp KT J exp JT MJ dnJ −2 n Z −∞ n/2 (2π) 1 T 1 = exp K M− K (2.94) 1/2 −2 (det M) the resolution function is obtained
det LI (τ2′′ = 0) 1 2 T 1 FI τ2′′ = exp τ2′′ T LI− τ2′′ T . (2.95) s det LI (τ ) −2 2′′ e e e e For strongly violated spin echo conditionse the exponential term will dominate the decay of the polarization. For perfectly satisfied spin echo conditions this term will be unity since T becomes zero. e
Example 1: Numerical calculations for RbMnF3
Fig. 2.6 and 2.7 show numerical examples for depolarization effects arising from a detuning of the tilt angles Θ1,2 and a detuning of the frequency f1,2. Here the echo amplitude is normalized to 1 for the optimum value for each parameter. The numerical calculations were done for RbMnF at the zone boundary excitation Q = [0.5 0.5 1], E = 8.287meV, for 3 − two different spin echo times τ = 12.4ps (Fig. 2.6) and τ = 80ps (Fig. 2.7). Details about the dispersion relation properties of RbMnF3 are listed in Appendix D. At the maximum of the dispersion the slope of the dispersion is equal to zero and hence Θ1,2 = 0◦. Due to the symmetry of the dispersion at the maximum and the first order dependence on the cosine of the coil tilt angles, the depolarizing effects arising from a detuning of Θ1,2 are symmetric with respect to the optimum value. Note that this is different for other values of Q, where the slope of the dispersion is different from zero. The depolarizing effect is slightly stronger for a detuning of Θ1. For the case of Θ1,2 = 0◦ any detuning of the coil tilt angles increases the path length of the neutron within the π-coil. Therefore, a higher Larmor phase is ob- tained and the depolarizing effect due to detuned coil tilt angles increases with increasing frequencies. Since f1 > f2 for the inelastic case, the depolarizing effect is stronger for the first spectrometer arm. In the case of the minimum experimentally accessible τ = 12.4ps a detuning of dΘ = 9 would decrease the echo amplitude by about 3%. 1,2 ± ◦ 26 2 NRSE resolution theory
1 1
0.8 0.8
0.6 0.6
0.4 0.4 echo amplitude echo amplitude
0.2 0.2 a b 0 0 −40 −20 0 20 40 −40 −20 0 20 40 dΘ [deg] dΘ [deg] 1 2
1 1
0.8 0.8
0.6 0.6
0.4 0.4 echo amplitude echo amplitude
0.2 0.2 c d 0 0 −40 −20 0 20 40 −40 −20 0 20 40 df [kHz] df [kHz] 1 2
Fig. 2.6: Depolarization effects due to a detuning of spin echo parameters at τ = 12.4ps corresponding to f2 = 50kHz. The detuned Parameters are: Θ1 (a), Θ2 (b), f1 (c) and f2 (d).
In contrast, the depolarizing effect arising from a frequency detuning is more sensitive to a detuning of the second spectrometer arm. Since f1 >f2 the relative detuning is smaller for higher frequencies and therefore the depolarizing effect decreases with increasing frequency. The calculated frequency ratio of this numerical example is f1 3. For τ = 12.4ps a de- f2 ≈ tuning of df = 25kHz and df = 8kHz would decrease the echo amplitude by about 3%. 1 ± 2 ± The detuning of f1 needs to be a factor of 3 higher compared to a detuning of f2 in order to obtain the same effect. This is in very good agreement with the calculated frequency ratio. The depolarizing effect arising from detuned coil tilt angles increases with increasing fre- quency. Hence, the echo amplitude is more sensitive to a detuning of Θ1,2 for higher spin echo times (see Fig. 2.7 a and b). For τ = 80ps a detuning of dΘ = 1.5 would de- 1,2 ± ◦ crease the echo amplitude by about 3%. In contrast the echo amplitude gets less sensitive for a detuning of the frequencies (see Fig. 2.7 c and d), since the frequency increases with increasing spin echo time τ. For τ = 80ps a detuning of 30kHz for f and 10kHz for f ± 1 ± 2 would decrease the echo amplitude by about 3%. A special case of detuning occurs if no spin-flip scattering occurs in the sample while the second precession field region is tuned to spin-flip-scattering. This case is easily treated if the second frequency f is assumed to be detuned to f . A numerical example for the zone 2 − 2 boundary excitation Q = [0.5 0.5 1], E = 8.287meV, in RbMnF is shown in Fig. 2.8. − 3 2.3 Extended NRSE resolution function 27
1 1
0.8 0.8
0.6 0.6
0.4 0.4 echo amplitude echo amplitude
0.2 0.2 a b 0 0 −40 −20 0 20 40 −40 −20 0 20 40 dΘ [deg] dΘ [deg] 1 2
1 1
0.8 0.8
0.6 0.6
0.4 0.4 echo amplitude echo amplitude
0.2 0.2 c d 0 0 −40 −20 0 20 40 −40 −20 0 20 40 df [kHz] df [kHz] 1 2
Fig. 2.7: Depolarization effects due to a detuning of spin echo parameters at τ = 80ps. The detuned Parameters are: Θ1 (a), Θ2 (b), f1 (c) and f2 (d).
It is obvious that the non-spin-flip signal is already completely depolarized at the lowest accessible spin echo time τ = 12.4ps. Note that this is not the general case and the signal arising from the occurrence of non-spin-flip scattering might be not completely depolarized at the lowest accessible spin echo time for other materials.
1
0.8
0.6
0.4 echo amplitude 0.2
0 0 5 10 15 20 τ [ps]
Fig. 2.8: Depolarization effects for spin-flip scattering without reversing the second field region. The example shown applies to the zone boundary excitation Q = [0.5 0.5 1],E = 8.287meV, in RbMnF3. − 28 2 NRSE resolution theory
Example 2: Transverse acoustic phonon in Pb
Using equation (2.95) and assuming an excitation with zero linewidth and perfectly satisfied spin echo conditions the instrumental resolution function for the transverse acoustic phonon [2 0.1 0] in Pb can be calculated numerically (see Fig. 2.9). For the calculations an energy ± of ~ω0 = 0.88meV, a slope of the dispersion of ~ qω0 (q0) = 6.9meVÅ at T = 290K and 1 | ∇ | an incident wavevector of ki = 1.7Å− were assumed. The resolution function changes only slightly with different TAS configurations. This is due to the small dependence of the beam divergence on the scattering senses. For this specific phonon the spin echo time for the NRSE option of the cold TAS V2/FLEX at BER II is limited to τ = 230ps. Therefore, the depolarizing effects arising from the scattering senses of the background TAS do not play a role in the accessible τ-range of the NRSE spectrometer.
1 SM=−1 SS=−1 SA=−1 SM=−1 SS=−1 SA=+1 0.8 SM=−1 SS=+1 SA=−1 SM=−1 SS=+1 SA=+1
0.6
0.4 echo amplitude
0.2
0 0 1000 2000 3000 4000 τ [ps]
Fig. 2.9: Calculated instrumental resolution for the [2 0.1 0] TA phonon in Pb assuming zero linewidth and perfectly satisfied spin± echo conditions. The depolarization is calculated for different scattering senses. The small difference for different instrument settings arise from small changes of the beam divergence due to the scattering senses. Note that for this specific phonon the spin echo time for the NRSE option of the TAS V2/FLEX at BER II is limited to τ = 230ps.
2.3.3 Quantitative description of depolarization due to sample imperfec- tions
In this section the additional term +τ ∆G ω (q ) in the total Larmor phase given by 2′′ c ·∇ c0 equation (2.81) arising from sample imperfections is considered in more detail. As pointed out by Pynn [19] the mosaicity of the sample introduces a further limit to the resolution. Fig. 2.10 shows a schematic drawing of this effect for a transverse phonon. For each lattice vector G within the mosaic spread the dispersion will have a different orientation in q-space. 2.3 Extended NRSE resolution function 29
This smearing of the dispersion leads to a broadening of the linewidth, which is different from the intrinsic linewidth of the excitation.
ω
Δω
Δq
q
lattice vector G
Fig. 2.10: Effect of the mosaic spread of the sample on the width of the dispersion. An angular variation in the lattice vector G (mosaic) leads to a variation of the linewidth of the dispersion and thus to an artificial broadening of the linewidth.
Correction for sign errors and using the correct second order term rather than that stated in [6] yield:
∆G = G G (2.96) c c − c0 (Gc0 +∆Gc)cos∆ν cos∆η = (Gc0 +∆Gc)cos∆ν sin∆η . (2.97) (Gc0 +∆Gc)sin∆ν Expanding small variations in the lattice vector Gc and neglecting terms higher than second order leads to
∆G 1 G ∆ν2 +∆η2 c − 2 c0 ∆Gc = Gc0∆η +∆ Gc∆η . (2.98) Gc0∆ν +∆Gc∆ν Since the scattering function S (Q,ω) is proportional to (Q ξ)2 where ξ is the phonon · polarization vector, the optimum choice for q0 for transverse phonons is an orientation perpendicular to G0, whereas for a longitudinal phonon the optimum choice for q0 is parallel 30 2 NRSE resolution theory
to G0. Equation (2.98) shows that for longitudinal phonons the mosaic spread contributes only in second order. Therefore, the sample can have a rather large mosaic spread before these effects limit the resolution. For a more detailed discussion see [6]. Since this vector has linear and quadratic terms in the variables ∆η, ∆ν and ∆Gc the linear and quadratic parts are defined separately:
∆Gc ∆Gc′ = G0c∆η (2.99) G0c∆ν 1 G ∆ν2 +∆η2 − 2 0c ∆Gc′′ = ∆ Gc∆η . (2.100) ∆Gc∆ν The complex resolution matrix can now be written as
1 LM = I− LI I + N + W, (2.101) with the non-zero elements of the (6 9) matrix I: ×
I11 = I22 = I33 = I44 = I55 = I66 = 1, (2.102)
I17 = Cx, I18 = CyG0c, I19 = CzG0c (2.103) with the definition C = qω (q ). The non-zero elements of the (9 9) matrix N are: ∇ c0 × 1 1 1 N77 = 2 , N88 = 2 , N99 = 2 . (2.104) ΥS ηS νS
Here ΥS is the 1σ standard deviation for the Gaussian distribution of lattice vectors. The quadratic terms in ∆η, ∆ν and ∆G are taken into account by the symmetric (9 9) matrix c × W. Its non-zero elements are:
W = W = iτ ′′C , W = W = iτ ′′C (2.105) 78 87 − 2 y 79 97 − 2 z W88 =+iτ2′′CxG0c, W99 =+iτ2′′CxG0c. (2.106)
Note that equation (2.106) is the corrected version of equation (83) in [6]. The linear terms T in ∆η, ∆ν and ∆Gc arising from ∆Gc′ can be written as +iτ2′′Tg JM , with the column 2.3 Extended NRSE resolution function 31 vectors
Tg = (0, 0, 0, 0, 0, 0,Cx,CyG0c,CzG0c) (2.107)
JM = (∆ω, ∆kin,y1,y2,z1,z2, ∆Gc, ∆η, ∆ν) . (2.108)
All linear terms can be taken into account by introducing
τ1 NI ei2 NF ef2 1,NI Ci Cf , , ,NI ei3, NF ef3, T = − − τ2 cos θ1 − cos θ2 − . (2.109) M Cx, CyG0c, CzG0c − − Analog to the treatment in the previous subsection, terms in ∆ω higher than linear are neglected, leading to
1 A = S (Q,ω)exp iτ ′′∆ω d∆ω (2.110) E N − 2 Z T 1 T 8 exp iτ ′′T J exp J L J d J , (2.111) × 2 M M −2 M M M n Z e e e e e where JM = (∆kin,y1,y2,z1,z2, ∆Gc, ∆η, ∆ν) and TM are the 8-dimensional sub vectors of
JM and TM , respectively. LM is the corresponding symmetric (8 8) sub matrix of LM . e e × Applying the general matrix theorem (2.94), leads to the resolution function in the form e
det LM (τ2′′ = 0) 1 2 T 1 FM τ2′′ = exp τ2′′ TM LM− τ2′′ TM , (2.112) s det LM (τ ) −2 2′′ e e e e which includes general lattice imperfections.e Figure 2.11 shows numerical calculations for the depolarizing effects of different mosaic spreads using (2.112). The calculations were done for the [2 0.1 0] TA phonon in Pb at the energy E = 0.879meV, assuming a zero linewidth. Note that the mosaic spread ηs is the FWHM. As discussed above, the higher the mosaic spread of the sample, the faster the polarization decreases.
2.3.4 Quantitative description of depolarization due to curvature of the dispersion surface
1 T In this subsection the additional term τ ′′∆q′ Hc ∆q′ in the total Larmor phase given − 2 2 c |qc0 c by equation (2.81), arising from the curvature matrix of the dispersion surface, is discussed. A curved dispersion surface within the TAS resolution ellipsoid leads to additional depo- larization effects, since neutrons corresponding to different points on the dispersion surface 32 2 NRSE resolution theory
1 η =1’ 0.9 S η =2.5’ S 0.8 η =5’ S 0.7
0.6
0.5
0.4 echo amplitude 0.3
0.2
0.1
0 0 50 100 150 200 250 300 τ [ps]
Fig. 2.11: Depolarization effects due to mosaic spread calculated for the [2 0.1 0] TA phonon in Pb. The mosaic spread ηs is the FWHM.
will have a different spin echo phase Φ. The formalism discussed here is also applicable for crystallographic systems with a symmetry lower than cubic. The Hessian matrix Hc is ex- pressed in the Cartesian coordinate system related to the reciprocal lattice by the B-matrix [16]. The elements of the Hessian H are defined by
∂2 Hijc = ω (qc) i,j = 1, 2, 3. (2.113) ∂qic∂qjc
If the dispersion ω (q) is differentiated in the frame of the reciprocal lattice with basis vectors b1, b2 and b3, i.e. the elements of H are defined as
∂2 Hij = ω (q) , (2.114) ∂qi∂qj and the Hessian H needs to be transformed into Cartesian coordinates using the B matrix:
1 Hc = BHB− . (2.115)
Here only second order curvature terms and no sample imperfections are considered. Thus
∆Gc = 0, ∆qc′ reduces to ∆qc and LI and J are used. Terms arising from sample imper- fections will be included in the next subsection. Following [6], the modified matrix reads:
1 1 LC = LI + iτ ′′I− Θ− Hc ΘC IC (2.116) 2 C C |qc0 2.3 Extended NRSE resolution function 33
where the matrix ΘC IC transforms the Hessian Hc into the coordinate space of LI . The matrix ΘC describes the transformation
∆qθ = ΘC Y (2.117) with the column vector Y = (x1,y1,z1,x2,y2,z2). The variation of the wavevector ∆qθ = ∆k ∆k is expressed in the basis of the Cartesian system (i , j and l ). In this coordinate i − f 0 0 0 system the total wavevector transfer Qθ is parallel to i0 (see Fig. 2.5):
QM 2π Qθ = 0 . (2.118) 0 Note that (i0, j0 and l0) are identical to the coordinate system as defined as θ-coordinate system in [17]. Here the same notation is adopted. In order to express the wavevector variation ∆qθ in the θ-coordinate system, ∆ki and ∆kf have to be rotated into the Qθ- frame. Using equations (2.43), (2.44) and the definitions made in Fig. 2.5 yields
∆k =(x cos φ y sin φ) i +(x sin φ + y cos φ) j + z l (2.119) i 1 − 1 0 1 1 0 1 0 ∆k =(x cosΞ y sinΞ) i +(x sinΞ + y cosΞ) j + z l (2.120) f 2 − 2 0 2 1 0 2 0 where φ is defined as the angle between Qθ and ki and Ξ is defined as the angle between
Qθ and kf . For equation (2.117) to hold
cos φ sin φ 0 cosΞ sinΞ 0 − − ΘC = sin φ cos φ 0 sinΞ cos Ξ 0 . (2.121) − − 0 010 0 1 − Note that here the definitions of φ and Ξ are different to [6] and the sign errors in the T definition of ΘC are corrected. In order to evaluate the expression ∆q Hc ∆qc, the c |qc0 Hessian Hc needs to be rotated into Hθ. According to [17] the transform of the vector Q is defined as
Qθ = ΩMNUBQ (2.122) 34 2 NRSE resolution theory and that of the matrix H is
1 1 1 1 1 Hθ = ΩMNUBHB− U− N− M− Ω− , (2.123)
T T since Qθ HθQθ = Q HQ must be invariant. Equation (2.123) is the generalization of the matrix transform given by equation (91) in [6]. In order to evaluate the resolution matrix, the Hessian needs a further transformation to the variable space of the six dimensional vector J. The matrix IC relates Y and
J = (∆ω, ∆kin,y1,y2,z1,z2) in the linear transformation [6]
Y = IC J. (2.124)
Since the aim is an expression for the linear relation between the variable vectors J and Y, the linear term in the expansion of the dispersion relation in combination with energy conservation relation (see equation 2.55) is used:
~ ~ ∆ω = k ω (q ) ∆k k ω (q ) ∆k (2.125) m I −∇ c0 · i − m F −∇ c0 · f = N ǫ ∆k N ǫ ∆k . (2.126) I i · i − F f · f
The substitution of ki,f and ǫi,f by using equations (2.69), (2.71) and (2.61) yields
∆ω = x (N e cos θ N e sin θ ) 1 I i1 1 − I i2 1 +y1 (NI ei1 sin θ1 + NI ei2 cos θ1)+ ei3NI z1 x (N e cos θ N e sin θ ) − 2 F f1 2 − F f2 2 y (N e sin θ N e cos θ ) e N z (2.127) − 2 F f1 2 − F f2 2 − f3 F 2 and
∆kin = x1 cos θ1 + y1 sin θ1. (2.128)
According to equation (2.124) the transformation reads
1 J = Ic− Y. (2.129)
1 The elements of the matrices IC− and IC are defined in Appendix B. For perfectly satisfied spin echo conditions equations (2.84) and (2.85) hold and IC reduces to the matrix given in [6].
With the equations (2.124) and (2.117) the additional term in the Larmor phase due to a 2.3 Extended NRSE resolution function 35 curved dispersion surface is given by
T T 1 T 1 1 ∆q Hc ∆qc = Y Θ− Hc ΘC Y = J I− Θ− Hc ΘC IC J. (2.130) c |qc0 C |qc0 C C |qc0 The resolution matrix as given in equation (2.116) can now be evaluated. Following the procedure described in section 2.3.2 the polarization can be written as
1 A = S (Q,ω)exp iτ ′′∆ω d∆ω E N − 2 Z T 1 T 5 exp iτ ′′T J exp J L J d J . (2.131) × 2 −2 C n Z e e e e e Using equation (2.94) the resolution function taking into account curvature effects and neglecting contributions from sample imperfections is then given by
det LC (τ2′′ = 0) 1 2 T 1 FC τ2′′ = exp τ2′′ T LC− τ2′′ T . (2.132) s det LC (τ ) −2 2′′ e e e e e
2.3.5 Quantitative description of depolarization due to sample imperfec- tions and curvature of the dispersion surface
The results of the previous two subsections can now be combined to give the resolution matrix T LMC = I LC I + N + W (2.133)
The resolution function can be expressed as
det LMC (τ2′′ = 0) 1 2 T 1 FMC τ2′′ = exp τ2′′ TM LMC− τ2′′ TM (2.134) s det LMC (τ ) −2 2′′ e e e e This resolution function FMC eincludes general lattice imperfections and the curvature of the dispersion surface. Here, the advantages of the matrix formalism are obvious, since cross terms arising from sample imperfections and the curvature are automatically taken into account by using (2.133)
T T ∆q′ Hc ∆q′ = ∆qc ∆G′ Hc ∆qc ∆G′ (2.135) c |qc0 c − c |qc0 − c T T = ∆q Hc ∆qc ∆G ′ Hc ∆G′ c |qc0 − c |qc0 c T 2∆q Hc ∆G′ (2.136) − c |qc0 c 36 2 NRSE resolution theory
1
0.9 0.8
0.7
0.6
0.5
0.4 echo amplitude 0.3 only instr 0.2 instr + sample 0.1 instr + curv instr + sample + curv 0 0 20 40 60 80 100 τ [ps]
Fig. 2.12: Numerical results for different depolarizing effects for the [2 0.1 0] TA phonon in Pb. Considering only instrumental resolution (black), instrumental resolution and sample imperfections (red), instrumental res- olution and curvature of the dispersion surface (blue). The depolarization due to all effects is shown by the green curve.
A comparison of the depolarization arising from the different effects discussed in the previous subsections is shown in Fig. 2.12. Since experimental tests on the numerical model discussed here were performed with Pb (see section 2.4), the numerical calculations were done for the [2 0.1 0] TA phonon in Pb, making the same assumptions as in section 2.3.2. Within the considered τ-range the depolarizing effects arising from the instrumental configuration of the background TAS (black) are negligible. The depolarizing effects due to sample mosaicity (red) are stronger compared to effects arising from the curvature of the dispersion surface (blue). However, the green curve shows, that it is important to consider all effects.
2.3.6 Dispersion surface not coinciding with the center of the TAS reso- lution ellipsoid
In this subsection the simplifying assumption that the center of the TAS resolution ellipsoid coincides with the dispersion surface of the excitation is dropped. The TAS is set such that the wavevector Q0 satisfies the scattering condition for a given set of wavevectors kI0 and kF 0 and an energy of ~ω0T AS. The wavevector transfer is
Q = k k (2.137) 0 I0 − F 0 2.3 Extended NRSE resolution function 37
and the TAS resolution ellipsoid is centered at the energy of (Q0,ω0T AS), where ~ ω (k , k )= k2 k2 . (2.138) 0T AS I0 F 0 2m I0 − F 0 This center does not need necessarily to coincide with the energy of the excitation ~ω0S (Q0), which is given by the dispersion relation
ω0S (Q0)= ω0 (qc0) . (2.139)
As a consequence all “mean” quantities, used in the derivation of the spin echo phase so far, have to be taken with respect to the sample dispersion. Since the TAS resolution ellipsoid is offset, the TAS resolution function needs to take into account the finite energy shift ∆ΩT AS:
∆ω ∆ω ∆Ω (2.140) → − T AS where ∆Ω = ω ω (Q ) . (2.141) T AS 0T AS − 0S 0 The resolution matrix reads
T LMC = I LC I + N + W (2.142) T 1 1 = I LT AS + iτ ′′Ψ + iτ ′′I− Θ− Hc ΘC IC I + N + W. (2.143) 2 2 C C |qc0 Since only the TAS transmission function changes, the following substitution has to be made
JT IT L IJ JT IT L IJ (2.144) T AS → T AS T AS T AS with the substitution
J = J J′=(∆ω ∆Ω , ∆k ,y ,y ,z ,z ), (2.145) T AS − − T AS in 1 2 1 2 where J′ = (∆Ω, 0, 0, 0, 0, 0). Therefore, the new TAS resolution matrix is defined as
T LS = I LT ASI. (2.146)
T T The Matrix LS is symmetric and therefore J′ LSJ = J LSJ′. Using this symmetry yields
T T T T J L J = J L J 2J L J′ + J′ L J′ (2.147) T AS S T AS S − S S = JT L J 2∆Ω (L ) J +(L ) ∆Ω2 . (2.148) S − T AS S 1n S 11 T AS 38 2 NRSE resolution theory
where (LS)1n defines the 6 dimensional row vector of the matrix LS and (LS)11 is the 2 (1,1) element of the matrix LS. The term proportional to (LS)11 ∆ΩT AS only produces a constant, which can be absorbed in the normalization factor N. With
1 1 JT L J = JT L J + ∆Ω (L ) J −2 T AS MC,T AS T AS −2 MC T AS S 1n 1 (L ) ∆Ω2 (2.149) −2 S 11 T AS the modified expression for the polarization reads
1 A = S (Q,ω)exp iτ ′′∆ω + ∆Ω (L ) ∆ω d∆ω E N − 2 T AS S 11 Z 1 exp TT J exp JT L J d5J . (2.150) × T AS −2 MC n Z Here the definition e e e e e
TT AS = iτ2′′TM + ∆ΩT AS LS . (2.151) 1n was used. Therefore, the finale result fore the resolution function,e including its normalization to 1 at τ = 0, is
′′ det LMC (τ2 =0) e ′′ det LMC (τ2 ) r e 1 T 1 FMC,T AS τ2′′ = exp T L− (τ ) T . (2.152) × 2 T AS MC 2′′ T AS