Spin response in HTSC cuprates: generalized RPA approach with projection operators method Текст научной статьи по специальности «Физика»

ISSN 2072-5981

Volume 16, Issue 2 Paper No 14206, 1-21 pages

2014

http://mrsej.kpfu.ru http://mrsej.ksu.ru

Established and published by Kazan University Sponsored by International Society of Magnetic

Resonance (ISMAR) Registered by Russian Federation Committee on Press,

August 2, 1996 First Issue was appeared at July 25, 1997

«I5

© Kazan Federal University (KFU)

"Magnetic Resonance in Solids. Electronic Journal" (MRSey) is a

peer-reviewed, all electronic journal, publishing articles which meet the highest standards of scientific quality in the field of basic research of a magnetic resonance in solids and related phenomena. MRSey is free for the authors (no page charges) as well as for the readers (no subscription fee). The language of MRSey is English. All exchanges of information will take place via Internet. Articles are submitted in electronic form and the refereeing process uses electronic mail. All accepted articles are immediately published by being made publicly available by Internet (http://MRSez.kpfu.ru).

Editors-in-Chief Jean Jeener (Universite Libre de Bruxelles, Brussels) Boris Kochelaev (KFU, Kazan) Raymond Orbach (University of California, Riverside)

Executive Editor Yurii Proshin (KFU, Kazan) mrsei@kpfu.ru

editor@ksu.ru

Editors

Vadim Atsarkin (Institute of Radio Engineering and Electronics, Moscow) Yurij Bunkov (CNRS, Grenoble) Mikhail Eremin (KFU, Kazan) David Fushman (University of Maryland,

College Park)

Hugo Keller (University of Zürich, Zürich) Yoshio Kitaoka (Osaka University, Osaka) Boris Malkin (KFU, Kazan) Alexander Shengelaya (Tbilisi State University, Tbilisi) Jörg Sichelschmidt (Max Planck Institute for Chemical Physics of Solids, Dresden) Haruhiko Suzuki (Kanazawa University,

Kanazava) Murat Tagirov (KFU, Kazan) Dmitrii Tayurskii (KFU, Kazan)

In Kazan University the Electron Paramagnetic Resonance (EPR) was discovered by Zavoisky E.K. in 1944.

*

Spin response in HTSC cuprates: generalized RPA approach with projection operators method^

M.V. Eremin1'*, I.M. Shigapov1, I.M. Eremin1'2 1Kazan Federal University, Kremlevskaya 18, 420008 Kazan, Russia 2Institut fur Theoretische Physik III, Ruhr-Universit at Bochum, D-44801 Bochum, Germany

*E-mail: meremin@kpfu.ru

(Received: March 26, 2014; accepted: April 19, 2014)

We derive the dynamical spin susceptibility in the t-J-G model combining the random phase approximation (RPA) and projection operator method, which allows describing the mutual interplay between the local and the itinerant components of susceptibility. Near the antifer-romagnetic wave vector the calculated dispersion of the spin excitations reproduces well the so-called hour-glass dispersion, characteristic for several layered cuprates. It is formed as a result of competition between the original spin-gap in magnon-like excitations spectrum and the superconducting gap, which affects the itinerant component of the susceptibility. Furthermore, the calculated collective spin excitations along (0, 0)-(0,n) are in agreement with the positions of the absorption peaks in the inelastic X-ray scattering spectra. They refer to the paramagnon-like modes, characteristic to the itinerant spin system, rather than magnon-like excitations that originate from short range order effect in the system of local spins at Cu sites.

PACS: 71.27.+a, 74.72.-h

Keywords: HTSC, cuprates, dynamical spin susceptibility, collective spin excitations 1. Introduction

The magnetic properties of high-temperature superconductors such as YBa2Cu3O6+y are quite unusual. These materials contain charge carriers, distributed mainly over the oxygen positions in CuO2 plane and localized spins at the copper sites. Correspondingly, two approaches are used for the description of the dynamic spin susceptibility. When one starts from overdoped regime, it is naturally to employ the conventional Fermi liquid type description with effective on-site Coulomb repulsion of the carriers at the same site. If one considers the lightly doped regime, both the tendency towards Mott physics and strong antiferromagnetic correlations have to be taken into account. Here we focus on intermediate doping level, when there are both local and itinerant spins, but the system is still uniform. Possible phase separations in underdoped part of the phase diagram and existence of the spin density waves or charge density waves are not considered.

Previously [1, 2], combining projection Mori method and Green's function technique we derived an analytical formula for the spin susceptibility in superconducting cuprates. It allowed to take self consistently into account both the itinerant and the localized components of magnetic susceptibility. However, the spectrum of collective spin excitations, which is in focus of many experimental investigations, was not investigated in details. A dispersion of the collective spin excitations is needed for the construction of the microscopic theory of high temperature superconductivity, at least, when the spin fluctuations mechanism is assumed. In the present paper we systematize our results. We also discuss the possible improvement of the general expression

^This paper is originally written by authors on the occasion of eightieth birthday of Professor Boris I. Kochelaev.

for the spin susceptibility and will present new numerical results for real and imaginary parts of the dynamic spin susceptibility along symmetry routes of Brillion zone in the normal and superconducting states.

2. Model and general expression for susceptibility

The Hamiltonian of the model is written as [1, 2]

H

Uiiij

] + ô E G%SiSi = Ht + Hj + HG. (1)

%3,<r

1- ,3

hJ

Here, ipf'a (ip°'pd) are the creation (annihilation) operators for composite quasiparticles. Symbol pd means that there is a strong coupling between copper and oxygen holes at each Cu-site in Cu-0 plane, which is resembled in the formation of the so-called Zhang Rice [3-6] (copper-oxygen) singlet band at strong enough doping level. The second term Hj = \ E Jij[SiSj — ^p-]

S3

describes the superexchange interaction between Cu-spins at site i and j [7], rii = ipj'^ + tpf'^ is operator of the number of spins at site i. The last term refers to screened Coulomb repulsion between doped oxygen holes, 6i = il)pd'pd is the operator of the number of copper-oxygen singlets per one unit cell.

For derivation of the spin susceptibility we employ the Green's function method and projection formalism. It is convenient to start from the equation

u>((S+ \szq)) = E (ttf-* - tk>){{4'+q€dA \s-*)) + E J^-^dSfS* - SfS+ \S-_q)). (2)

k'

3,1

Here it is assumed that (Sf) = 0 i.e. the long range spin order is absent, tk = Y1 hj exp(ikRjj) is

j

usual Fourier transform of the hopping integral. On the right-hand side we get two new Green's functions. The first can be naturally referred to itinerant spins, which accompany the motion of copper-oxygen singlet correlations over Cu-0 plane, whereas the second term in Eq. (2) is related to localize spins at Cu-sites.

Linearization of the anticommutator

E J3ie~^(S+S]-SfS+),Hj

3,1

r^j O2 C +

(3)

was discussed in many papers [8-18], where

= 2jfa\KX\ (2 - 7i)(A+ 2 + lq)]

(4)

is the typical expression for collective local spin excitations in layered aniferromagnets, Asp is dimensionless spin-gap parameter, jq = cos qxa + cos qya, a is decoupling parameter, which is usually calculated self-consistently via the sum rule = — S) which is discussed in

Appendix C, ô is a number of carriers per one unit cell, K\ = Kq\ = 4(SqSI) is the spin-spin correlation function of nearest neighbors.

Calculation of the anticommutator

E Jiie~iqRi(S+Sï - SfSf), Ht

is given in Appendix A.

It is approximated as follows

E J^~iqRl OT - Sfsf), E t3,m^d'armpd

Jiti [2 - cos qxa - cos 5"+ - - E (Jfc+g ~ Jfc) (¿fc+g ~ ¿fc) V'lf^

(5)

fc •

Here Jfc = ^ Ja exp(ikR^) and J^ = Jn(ipi V^ )• Angular brackets denote the thermody-i

namic average.

Using Eqs. (3), (5) we derive the equation for Fourier transform of Green's function as

UJ

E (Wsf - Ui+I5-,» =

- -JiKi (2 - 7,) + {Q2 + Jiii [2 - cos qxa - cos qya]} ((5+ ))

- \ E G7^-^) - ifc) «vi^f4|srff».

k

Appling projection method as it was described in Ref. [1] we write

^ = ek№* + Al^ +

dt 1

2Ar E + ^ E G'M'-ir/^ + )

q q

The expression for the energy of quasiparticales is written as [1, 2]

?G , 7 tpJ

ek = E № + (1 + 2i$)<S?S?)/P] "

(6)

(7)

(8)

Here, P = (1 + 5)/2, is projection parameter, which will be calculated later via a number of holes per one unit cell 5 and spin-spin correlation function (SjSf). It is interesting to compare the quantity in square brackets with Gutzwiller's projection factor 25/(1 + 5), which was introduced for the phenomenological description of the doping dependent bandwidth [19]. In contrast to Hubbard 1 approximation for 5 —> 0 the bandwidth shrinks to zero. The same result gives expression P + (1 + 2Fjl)(SjSf)/P. For this particular case one expects that the spin-spin correlation for nearest neighbors on the square lattice will be given by (S^Sf) = —1/4 and F\ —> 0. Note that the role of the last term in the bracket (8) at 5 = 0.2 is relatively small.

The superconducting gap equation is given by [1, 2]

k' ^

(9)

t ,pd„iA,pd\

where t'k = Y1 tji^n exp^kR^) is Fourier transform of the reduced hopping integral and Gq = i

G— Ji/4. The analysis of this equation for Jk-k' > 2Gk-k>, which we assume hereafter, reveals Ak = A(T)(coskxa — coskya)/2. Having expression (7) one can construct the equation

for functions and ((ip^t'^iplf^qlSlq)), which enter in (2) and (5). However,

one has to be careful because the rule of differentiation of product composite operators, as it was pointed in Ref. [10], is different from the usual case. Before applying expression (7) we write the following relations

= jf E C1 " S*i) + (10)

rkdA4fq = ^ E 4 (1 - tij) e~i^Ri+ikR'. (11)

ij

Then carry out the differentiation of operators with lattices index (site representation) and taking into account (10) and (11) one gets

^(4fq€d'11 szq)) = ^ ((4fg€d4) - (4pd€M))

+ (sk+q-sk){{4fqrkdA\Slq}

(i2)

~[Fj\jq - tk+q{i - FM4fq€il)Mst |SZq))

- jf E ^ - \szq)) + |srff »,

k'

"((€M4fq ISzq)) = ^ (OK> - (rkdM'pd) + (sk+q-sk){{rkdA4fq\Slq»

i

N \ 2 v /JNr/c r/c ' (13)

1 2*

N

~[Fj^Jq ~ tk+q( 1 - ))

1

y

These equations are rewritten as follows [2]

mdA4fq\szq)) = ¿XM + jjT/k,q((S+ \SZq)) + ^(k,qD>(uj,q), (14)

{{4+^ ISZq)) = -¿XM - jjT/'k,q((S+ \SZq)) ~ jjtk,qD"(u,,q). (15)

Here

_ Uk+q - Uk / I tk+qnk+q ~ tkUk

Xk,q — , f]k,q — 0JqXk,q ,

, _ 1 " _ ' . p(tk+q -tk) -w — ,. , _ I i ^IkM — VkM +

(16)

W + £fc — £k+q ' ' W + £k ~ £k+q nk = Pf (a/.) are occupation numbers, / (ek) is Fermi function. New Green's functions, which are appeared on right-hand side (14) and (15) are:

D>(u,q) = -J2(ek+q-ek){{rkdA4fq \SZq)), (17)

D"(UJ, q) = J2 (ek+q ~ ek)((4idg€d'1 \S~* »• k

To compute these functions the following exact relations are used [1, 2]

k

k i Combining Eqs. (14), (15) and (19), (20) one finds

D'(co, q) = - | l£-X(uj, q) + r/(W, q)((S+ \Slq)) j /((to, q)

and

Here,

D"(uj,q) = - [^XM + [1 + »/>,?)] ((5+ |5lff))} /((uj,q).

(18)

(19)

(20)

(21) (22)

(23)

N

k,q-

Taking into account that J2(£k+q — £k) = 0, it is easy to prove that D'(uj,q) = D"(uj,q) and

k

derive the relation [2]

F*. =

v

IK,

v I

(l + sy - 2\Kl3\ Substituting D"(uj,q) functions in (15) we get

(24)

((4fq€dAKq)) = -

2ir

+ l([l + r]"(uj,q)}- Cfc*

Xk'q CM

Then multiplying (2) by frequency and using Eq. (5) we find

~ri"k,?j ((S+\SZq)).

(25)

{w2 - Q2 - Jiii [2 - cos qxa - cos qya]} ((S+ \Slq)) = -¿2JXKX (2 - lq)

- \ E -- 1 s-_q))

k

+ u E (4'+, - |5Zff))

(26)

k> i

IT

hKx (2 - 7g) - - E (Jk+q ~Jk~ 2u) {tk+q - tk) {{4fqrkdA \SZg)).

Substituting here the expression for ((V^+gipvk'^ l^-^}} function as given by (25), we get the equation

(w2 - Q2 - tiJi [2 - cos qxa - cos qya]) ((5+ )) =

iiV

" 27T

2 JXKX (2 - 7g) + tJ " XtJ

«5+151.)).

Since NXtotaii00>Q) = ^idSq for the dynamical spin susceptibility per one unit cell we

get

= X(tJ + [2J1K1 (2-_lq)-Xtj}( ( 8)

Xtotal + ^ (u + [w2 _ _ Jlh (2 _ lq) _ ^ (* ^ )

Here the symbol (uj,q), which accompanies the functions (, rj, is dropped for short. The new functions with index tJ are:

Xt.j(oj, q) = ^ E (tk+i ~ ifc) 2u] Xkq, (29)

k

i/'tq) = E (tk+i ~ ifc) - Jfc - 2w] r/'kq, (30)

k

Ct.j(0j, = (tk+1 ~ tfc) 20J] (kg- (31)

k

The expression (28) can be rewritten also as follows

= X(u + [2Jj-fij (2 - lq) - XtJ} C

Xtotal + A] (tj + _ Q2 _ Jlh (2 _ lq) _ AtJ] C' ^ )

where A = rf' — uj( (this is a function like defined in [20]):

, , P{t'k+g ~ t'k)

Xk,q = V k,q H--1-,

to + £k — £k+q A = A (w, q) =

N k

Atj(w, q) = —— E (tk+q - tk) \Jk+q — Jfc — 2w] Afcg.

2N

k

It seems this form is more convenient for numerical calculations. Another form for susceptibility, which has more clear properties under electron-hole transformation, is discussed in Appendix B. Note that in numerical calculations the substitution uj —> uj + iT is assumed in all entering functions. T is dumping factor, which can be anisotropic along the Fermi contour. Factor [1 + A] in denominator (32) reminds the corresponding Stoner's factor in random phase approximation schema (RPA) for itinerant electron system. On the other hand, the quantity [Q2 — uj2 + J\t\ (2 — 7q) + Atj] is typical for localized spin-subsystem. Collective spin excitations are determined by equation

[1 + A] tJ + [to2 ~n2q~ Hh (2 - lq) - AtJ] = 0. (33)

Therefore, local spin excitations (magnon-like) and itinerant holes (paramagnon-like) are coupled to each other self-consistently. In other words the total susceptibility (32) can be considered as a possible variant of description magnetic susceptibility compounds with interplay between local and itinerant spin subsystems.

3. Superconducting state

The expression for spin response function (32) retains its form in superconducting state as well. The entering functions, of course, modify due to Bogolyubov transformations in the superconducting state. The method of calculation can be found in Ref. [1]. The x(w,<?) function at T <TC is written like in BSC theory, except additional factor by P = (1 + S)/2:

P \ ^ n _fk+q — fk__P_ \ ^ n _fk — fk+q_

N ^ xx to + iT + Ek — Ek+q N vvu + iT -Ek + Ek+q

. P y^ 1k + ^k+q ~ 1 \ ^ 1 ~ /fc ~ fk+q ^ ^

N1- yx UJ + lT _ Ek - Ek+g N xy uj + ir + Ek + Ek+q>

fk = { 1 + exp[Ek/k,BT]}~1 is the usual Fermi function. For the sake of simplicity, we are using the following abbreviation for the coherence factors:

Sxx — XkXk-\-q ZkZk+q, &yy — UkUk+q ^kZk+q, S^) = %kUk+q ~ ZkZk+g, SyX^ = Ukxk+q ~ ZkZk+q,

(35)

where

1

fc 2

1 +

Ek

1

Vk = ~

1 _ £k~ ^ Ek

2 Ek

The function A(uj,q) is written as [20]

zk = Ek = \J(ek - fx)2 + |Afc|2

(36)

Hu>Q) = O JqFJX(u,

_ (ST^ c t'k+q(fk+q ~ 1) ~ t'k{fk ~ 1) V^ c ^fc/fc ~ t'k+q fk+q

N \^ UJ + ¿r + Ek — Ek+q ' UJ + ¿T — Ek + Ek+q ^ '

I V^ q(-) t'kfk — t'k+q{ 1 — fk+q) ~ fk) ~ t'k+qfk+q\

+ V* UJ + iT-Ek-Ek+q xy UJ + iT + Ek + Ek+q )■

Physically last term in Eq. (37) corresponds to an effective molecular field of kinematic origin. It is strong correlation effect, because in our case the anticommutate relations are different from conventional Fermi liquid.

The function ((uj,q) is written as follows:

tit ? n\_ 1 \ "___, 1 Y^ Syy

C(W) Q) T7 , .-p , rp _ TP, , _ + n2

N ^ uj + iT + Ek- Ek+q N ^ uj + iT-Ek + Ek+q (-) c(+)

1 „_, q\ ) i __o1

± V Jyx -L X ^

(38)

, \ " _*fyx__\ ^

+ N Z^ m -I- -)T - R„. - R„. , . N 2-^t

xy

N ^ uj + iT-Ek- Ek+q N ^uj + iT + Ek + Ek+q

Let us now for the moment assume that we do not have any conduction electrons (holes). Substituting zero instead Fermi functions, from expression (26) one gets

XiocalQ) = -Q2 _• (39)

This expression is identical to those found by many authors for lightly doped cuprates [8-17]. It is remarkable that magnetism of localized spins at T <TC is strongly suppressed "or in other words frozen out" due to the superconducting gap, which naturally incorporated in function ((uj,q). In opposite limit, when spin-spin correlation functions are small, and conducting bandwidth is

large enough and correspondent^ function ((uj,q) is small the expression (32) corresponds to generalized random phase approximation (GRPA) schema [21-23]. Moreover, in limit q —> 0, uj —> 0 the expression (32) converts to static susceptibility expression, which corresponds to those one, which was derived in Ref. [24], beyond Green's function method.

4. Numerical results

Calculated imaginary and real parts of susceptibility along triangle contour in Brillouin zone are displayed in Figs. 1-10. The chosen parameters are: T = 4meV, 5 = 0.33. We have neglected J-2 and K-2 as they are small numerically. The energy dispersion was chosen according to photoemission data [25]

£k = 2ii(cos kxa + cos kya) + 41-2 cos kxa cos kya — ¿t,

(40)

where (in meV): ii = 139, ¿2 = —33, n = 88. The superconducting gap function was set in agreement with analyzes of the temperature dependencies of nuclear relaxation rate, Knight shift and superfluid density, which were discussed in Ref. [23] and [26];

Afc = ^(cos kxa - cos kya) tanh (l.76 \/Tc/T - l) . (41)

The decoupling parameters a, (5 and Fj below (except of specific cases in Fig. 9 ) were set by 1.

u

Figure 1. (Color online) Imaginary part of the dynamical susceptibility for the normal state. Input values are: T = 100K, Ji = 66meV, I\\ = —0.2, Asp = 0.1, Ft = 0.1. Upward dispersion can be interpreted as a damped magnon-like collective spin excitations. Vertical scale is in 1/eV.

Figure 2. (Color online) The real part of the dynamical spin susceptibility for the superconducting state. Input values are: T = 10K, A0 = 25meV, J1 = 66meV, K1 = -0.2, Asp = 0.14, Ft = 0.1. Vertical scale is in 1/eV.

(0,0) !-0) (q ,q ) MM (0,0)

Figure 3. (Color online) The imaginary part of the dynamical spin susceptibility for the superconducting state. Input values are: T = 10K, A0 = 25meV, J1 = 66meV, K1 = -0.2, Asp = 0.14, Ft = 0.1. Vertical scale is in 1/eV.

Figure 4. (Color online) Zoomed in region of point the imaginary part of the spin susceptibility for the superconducting state . This aria is usually probed by inelastic neutron scattering [27-33]. Input values are: T = 10K, A0 = 25meV, J1 = 66meV, K1 = -0.2, Asp = 0.14, Ft = 0.1. Vertical scale is in 1/eV.

Figure 5. (Color online) Real part of the spin susceptibility for the superconducting state. Input values are: T = 10K, A0 = 30meV, J1 = 66meV, K1 = -0.2, Asp = 0.14, Ft = 0.1. Vertical scale is in 1/eV.

(0,0) (71,0) , | inri (0,0)

x y

Figure 6. (Color online) Imaginary part of the spin susceptibility for the superconducting state. Input values are: T = 10K, A0 = 30meV, J = 66meV, K = -0.2, Asp = 0.14, Ft = 0.1. Vertical scale is in 1/eV.

0 I I I

(7t,7t/2j /a a I fcaf (71/2,Tt/2)

y

Figure 7. (Color online) The imaginary part of the spin susceptibility for the superconducting state in region of (n,n) point. This region is usually probed by inelastic neutron scattering [27-33]. Input values are: T = 10K, A0 = 30meV, J = 66meV, K = -0.2, Asp = 0.14, Ft = 0.1. Vertical scale is in 1/eV.

0 (0,0)

Figure 8. (Color online) Fragment of imaginary part susceptibility in superconducting state along the line in Brillouin zone. This region is studied by resonance inelastic X-ray scattering (RIXS) [34-38]. Input values are: T = 10K, A0 = 30meV, J1 = 66meV, K1 = -0.2, Asp = 0.14, Ft = 0.1. Vertical scale is in 1/eV.

fcogII fa m SUSI)

3jP-jr

Figure 9. (Color online) Fragment of the imaginary part of the spin susceptibility for the superconducting state near (n, n) point. Input values are: T = 10 K, A0 = 25meV, J1 = 70meV, K1 = -0.2, Asp = 0.147, Ft = 0.1, ft = 0.6. Vertical scale is in 1/eV.

(71,71/2) (71,71.) (7l/2,7tQ)

(qy,qv)

Figure 10. (Color online) Fragment of the imaginary part of the spin susceptibility for the superconducting state near (n,n) point. Input values are: T = 10 K, A0 = 25meV, J1 = 76meV, K1 = -0.152, Asp = 0.17, Ft = 0.1, Fj = 0.89. Vertical scale is in 1/eV.

0.16

(0,0) i-0! (ti,TI;) (0,0)

(qx!qy)

Figure 11. (Color online) Dispersion of the collective spin excitations calculated as a solution to the equation (32) along the route (0, 0) - (0, n) - (n,n) - (0, 0) in the Brillouin zone for the normal state (dashed line) and superconducting one (solid line). Parameters are the same as in Fig. 3.

5. Conclusions and final remarks

The calculated results, shown in the previous section for the spin response near (n,n) point of the BZ resemble very strongly the experimental results reported by inelastic neutron scattering for hole doped cuprates [27-33]. Both the upward and the downward dispersions are successfully reproduced by our calculations. We have found that the fine features of the dynamical spin susceptibility near (n, n) point around the frequency 40 meV are very sensitive to the possible variation of the spin-gap (Asp = 0.14 or Asp = 0.16) and superconducting gap (A0 = 25 meV or A0 = 30 meV) parameters. Relatively small changes of these values lead to quite essential modification in the calculated picture. Under small changes of the input parameters the picture with the intersection of upward and downward dispersions (or in other words X-shape ) in Fig. 4 and in Fig. 7 transforms to the so-called hour-glasses picture (Fig. 9).

As for the shape of the imaginary part of the spin susceptibility along the route (0,0) — (n, 0) is very similar to those which was observed by inelastic resonance X-ray scattering (RIXS) [34-38]. Generally, its dispersion is not so sensitive to the possible variation of the input parameters. Double shape form calculated by us here and in Ref. [20] was not observed experimentally, probably because of the error bars in the current RIXS technique, which is about 300 meV [34-38]. It is interesting to note that the splitting between two mountain chain-like features, found in Fig. 8, is sensitive to the superconducting gap values. This fact clearly demonstrates that these mountain chain-like features are more related to the itinerant part of spin subsystem rather than to the collective spin excitation of the localized spins. Another interesting observation is that the calculated frequency plot of the collective spin excitations (Fig. 11) does not exactly corresponds to the maximum positions of imaginary part, shown in Fig. 3. Especially this can be easily seen for dispersions along the route (0,0) — (n, 0). This observation strongly supports the idea that peaks in RIXS spectra correspond to the spin excitations, which originate from paramagnon-like features of the itinerant origin rather than magnon-like excitations, which would originate from the short range order of the local spins at Cu sites. The latter explains better the nature of the high-energy spin excitations near (n,n) point with frequencies which are above 2A0/h.

Acknowledgments

This work is supported by RFBR Grant 13-02-00492_a

Appendix A

Let us consider the commutator:

tqRi(SfSi — SfSf), E Tpfrf1

],m

Using relation [AB, CD] = A[B, C]D + AC[B, D] + [A, C]DB + C[A, D]B one has

E J«e~iqRi ^st - sist)> E hm4d'armpd

i,l j,m

= E - ipf>Hkpd)

+ \ E 'hie~^thlSn^dHlpd ~ ^fpd)

- E J*e~iqRitl,mfldA№dS; + E Jue-^n^'^S! + E Jue-^mSf^'^ - E Jue~iqRitj,iSfijjPjd^ipJ'pd

(la)

(2a)

Doing averaging we focus on the quantities Ju (ippd'a ip'['pd) and tu(ijjpd'aip'['pd) as suggested by Kuz'min [16]. However, assuming that tiim(ippd^iplfd) = tiim(ippd'^ipifd), we can see that the first two and the last two amounts in (la) can be discarded. Further following Kuz'min suggestion for third, fourth, fifth and sixth terms we get following expressions:

-E Jue-^t^f'Hlf'S!

(3a)

= \ E ^VWV) - \ E Jiie^h^ti**^,

,pd,i

(4a)

(5a)

Spin response in HTSC cuprates: generalized RPA approach with projection operators method

- E Jae-^t],tSfr3d'i4pd = E Jue-^t^Sf^r/'1

i i (6a)

= \ E " \ E ^e-^i^^i^4-.

Keeping terms containing Ju(ippd^ipj'pd) = Ju we have

= 1 -^ 4 (7a)

Next we turn to the Fourier components:

= (9a)

^ fc

After straightforward calculation the expression (7a) can be rewritten as

-\(Jk+q - Jk){tk+q ~ tk)4fq€d'1 + O^i) I2 - cos - cos qx] (10a)

Now, consider the remaining in Eqs. (3a)-(6a) four terms:

\ E ^xrw - \ E Jn^t^^i^

+ ^ E - \ E Jiie~iqRitj,i'iiji'^'iiji'pd'iijljd'^

(lla)

= ^ E ^^^^¿^(^Vl^)^^^4 - ^ E 4#pd>CUVi'P<i

+ \ E ^e-^ii.^'^i^^'Vif') " \ E

Let us start with the following:

(12a)

E " ^ E ^ue~iqRitj,i(^pd'^ipj'pd)ipJ'pdippd'^■

The largest terms are:

\ E 'he-^M^l'^f'1 + \ E Jue~iqR%Md'PVi'Pd^d'4

(13a)

E ^e-^ii,»(^rf^4 " 5 E Jne~iqRi{^pd'pd)^}'pd■

In the uniform lattice (ippd'pd) = (ipfd'pd). Therefore, the first and fourth (second and third) do cancel each other.

Now let us discuss the role of the three-site correlations. After Fourier transform these terms are written as

\ E + \ E

(14a)

к ■

Performing summation over the square lattice we find

i,m

- Jltl (2+ (Фр0ал4ра)) (cos qx + COS qy) ^

- Jiii + {rodM'Pd}) (cos qx + COS qy) = 0.

Here {r>pQd'^rip2Pd) refers to the next nearest neighbor correlation. The hopping parameters t2=t02 and t3=t03 are not included into consideration.

Finally the anticommutator (la) is written as follows:

E Jue~^(S+St - SfS+), E 4miPpd>aympd

1

(16a)

= Jiti [2 - cos qxa - cos qya] 5"+ - ^ E (Jk+g ~ Jk) (tk+q ~ tk) Ф^Ф^'1-

Here is, of course, an addition decoupling factor /3 = 1 is occurred, like factor a in Eq. (4). Therefore here it is logical to change the definition Ji as follows

Ti = Ji^fHYd). (17a)

It should be noted that our result (16a) is different from those which presented in the original paper [16] by factor 2. Furthermore, we do not approximate additionally the last term in Eq. (16a). The system of self-consistent equations for Green's functions can be obtained (see the main text), using expression for aticommutator, as it is given by (16a). Entering correlation functions in the normal state are calculated as follows

(vf' H\,pd) = ^ E {ФркллФ\:рЛ)е~гкЩ+гк'К1 к,к'

ж/а ж/а / у\ (18а)

For nearest neighbors it is written as:

0 0

Here tetragonal symmetry in Cu-0 plane is assumed.

Appendix B

The Green's function of itinerant part in Eq. (2) can be rewritten as follows [2]

E (tk+q-tk)((4fq€d'l\szq)) k

= \ E - **)«teX"4 - Iszq)).

(lb)

In equation (16a) the last term can be rewritten as

1

i E (Jk+i ~ Jfc) (tk+i ~ ifc) = ^ E {Jk+q~ Jk) (tk+q - tk) 4id,i4>,

k+q

2 N

E fc+9 ~~ Jk) ~~ Sq-

k

In case of square lattice

1 E (Jk+i ~ Jk) (tk+i ~ tk"> ~ ~2iiJi (2 ~ cos qxa - cos qya).

2 N

k

Therefore the equation (16a) becomes

E Jue-** (S+Sf - SfSf), E hm4d'armpd i,i

= — J\t\ [2 — cos qxa — cos qya] S+

- 7 E ^ " lk) «K-H ~ tk) tor*4 - VT^I^

(2b)

(3b)

(4b)

Using (lb) and (4b) we get the following expression for spin response function

X(u + [2Jitfi (2 - 7g) - xtj] C

where

Xtotal

[i + r?] Ctj + [w2 - Q2 - Jih (2 - lq) - ntj] C'

v = ^ E = q)

k

+ E

¿'fc+g(/fc+g - 1/2) - t'kjfk - V2)

uj + iT + Ek- Ek+q

t'k+q(I/2 ~ fk+q) ~ t'k( 1/2 ~ fk) uj + ¿r - +

, o(-) ¿'fc+g(/fc+g ~ l/2) ~ ¿;fc(l/2 - fk)

^ yx uj + iT-Ek-Ek+q

4- Y^ <?(+) t/fc+g(1/2 ~ fk+q) - t'kUk ~ 1/2) ^ u + iT + Ek + Ek+q

as it was introduced in [2] and

T]tJ(uj, q) = — E (tk+q ~ tk) \Jk+q ~ Jk ~ 2w] r]kq.

(5b)

(6b)

(7b)

Note the denominator of the dynamical spin susceptibility can be written also via rf(w.q) function. Corresponding expression for rf(w.q) in superconducting state can be found in Ref. [1]

Appendix C

Spin-spin correlation functions.

The spin-spin correlation functions are calculated self-consistently via expression for the dynamical spin susceptibility. It is useful to start discussion from the Fourier transform:

= Jp E (SZqS+yiR-. (lc)

q

Then to use Green's function technique

(S_qS^) = J g/3w _ -y [((S^ IS-q))oj+i£ —

= J i^TTI w + - (<?>w - i£)] (2c)

N f dUJ T +,-, N

= 77

Since

lmx+~(q,uj) = -lmx+~(q, -w), (3c)

integration can be carried out only by positive values uj. Thus, one finds

= / J^lmX+'-Me*««

q

=^ (¿) 7// ^Im x+,~ (<?) - Im x+,~ (<?) -u)]eikRii ^^ (4c) = 2 IIICth Im%+'" (<l>u)eikRiid"dk*dkv

Sum rule.

The case i = j is used for self-consistent control. Since

ipf'1 + + 2 ^d'pd = 1 + 5,

J I I oo

f f fcth №£) lmx+'" (9,0;)

(5c)

(^I) = 0) and (ipfd'pd) = 5 one gets

cth ( ^T ) ImX+'" (<?>e1^ d(jjdqxdqy. (6c)

ooo

References

1. Eremin M.V., Aleev A.A., Eremin I.M. JETP 106, 752 (2008) [Zh. Eksp. Teor. Fiz. 133, 862 (2008), in Russian]

2. Eremin M.V., Shigapov I.M., Eremin I.M. Eur. Phys. J. B 85, 131 (2012)

3. Zhang F.C. Rice T.M. Phys. Rev. B 37, R3759 (1988)

4. Eremin M.V., Solov'yanov S.G., Varlamov S.V., Brinkman D., Mali M., Markendorf R., Roos J. JETP Lett. 60, 125 (1994) [Pisma Zh. Eksp. Teor. Fiz. 60, 118 (1994), in Russian]

5. Eremin M.V., Solovjanov S.G., Varlamov S.V. J. Phys. Chem. Solids 56, 1713 (1995)

6. Plakida N.M., Hayn R., Richard J.-L. Phys. Rev. B 51, 16599 (1995)

7. Anderson P.W. Science 235, 1196 (1987)

8. Shimahara H., Takada S. J. Phys. Soc. Jpn. 61, 989 (1992)

9. Kondo J., Yamaji K. Prog. Theor. Phys. 47, 807 (1972)

10. Zavidonov A.Yu., Brinkmann D. Phys. Rev. B 58, 12486 (1998)

11. Jackeli G., Plakida N.M. Theor. Math. Phys. 114, 335 (1998) [TMF 114, 426 (1998), in Russian]

12. Winterfeldt S., Ihle D. Phys. Rev. B 58, 9402 (1998)

13. Winterfeldt S., Ihle D. Phys. Rev. B 59, 6010 (1999)

14. Sherman A., Schreiber M. Phys. Rev. B 68, 094519 (2003)

15. Sherman A., Schreiber M. Eur. Phys. J. B 32, 203 (2003)

16. Kuz'min E.V. Low Temp. Phys. 31, 143 (2005)

17. Vladimirov A., Ihle D., Plakida N. Theor. Mat. Phys. 152, 1331 (2007) [TMF 152, 538 (2007), in Russian]

18. Mikheenkov A.V., Barabanov A.F. JETP 105, 347 (2007) [Zh. Eksp. Teor. Fiz. 132, 392 (2007), in Russian]

19. Anderson P.W., Lee P.A., Randeria M., Rice T.M., Trivedi N., Zhang F.C. J. Phys.: Condens. Matter 16, R755 (2004)

20. Eremin M.V., Shigapov I.M., Thuy H.T.D. J. Phys.: Condens. Matter, 25, 345701 (2013)

21. Field J.J. J. Phys. C: Solid State Phys. 5, 664 (1972)

22. Auslender M.I., Irkhin V.Yu., Katsnelson M.I. J. Phys. C: Solid State Phys. 21, 5521 (1988)

23. Mayer T., Eremin M., Eremin I., Meier P.F. J. Phys.: Condens. Matter 19, 116209 (2007)

24. Eremin M.V., Solov'yanov S.G., Varlamov S.V. JETP 85, 963 (1997) [Zh. Eksp. Teor. Fiz. 112, 1763 (1997), in Russian]

25. Norman M.R. Phys. Rev. B 63, 092509 (2000)

26. Eremin M.V., Larionov I.A., Lyubin I.E. J. Phys.: Condens. Matter 22, 165704 (2007)

27. Bourges P., in The Gap Symmetry and Fluctuations in High Temperature Superconductors, ed. by J. Bok, G. Deutscher, D. Pavuna, and S. A. Wolf, Plenum, New York, pp. 349-371 (1998)

28. Reznik D., Bourges P., Pintschovius L., Endoh Y., Sidis Y., Masui T., Tajima S. Phys. Rev. Lett. 93, 207003 (2004)

29. Reznik D., Ismer J.-P., Eremin I., Pintschovius L., Wolf T., Arai M., Endoh Y., Masui T., Tajima S. Phys. Rev. Lett. 78, 132503 (2008)

30. Pailhes S., Sidis Y., Bourges P., Hinkov V., Ivanov A., Ulrich C., Regnault L.P., Keimer B. Phys. Rev. Lett. 93, 167001 (2004)

31. Arai M., Nishijima T., Endoh Y., Egami T., Tajima S., Tomimoto K., Shiohara Y., Taka-hashi M., Garrett A., Bennington S.M. Phys. Rev. Lett. 83, 608 (1999)

32. Dai P., Mook H.A., Hunt R.D., Dogan F. Phys. Rev. B 63, 054525 (2001)

33. Bourges P., Sidis Y., Fong H.F., Regnault L.P., Bossy J., Ivanov A., Keimer B. Science 288, 1234 (2000)

34. Le Tacon M., Ghiringhelli G., Chaloupka J., Moretti Sala M., Hinkov V., Haverkort M.W., Minola M., Bakr M., Zhou K.J., Blanco-Canosa S., Monney C., Song Y.T., Sun G.L., Lin C.T., De Luca G.M., Salluzzo M., Khaliullin G., Schmitt T., Braicovich L., Keimer B. Nat. Phys. 7, 725 (2011)

35. Le Tacon M., Minola M., Peets D.C., Moretti Sala M., Blanco-Canosa S., Hinkov V., Liang R., Bonn D.A., Hardy W.N., Lin C.T., Schmitt T., Braicovich L., Ghiringhelli G., Keimer B. Phys. Rev. B 88, 020501(R) (2013)

36. Braicovich L., van den Brink J., Bisogni V., Moretti Sala M., Ament L.J.P., Brookes N.B., De Luca G.M., Salluzzo M., Schmitt T., Strocov V.N., Ghiringhelli G. Phys. Rev. Lett. 104, 077002 (2010)

37. Dean M.P.M., James A.J.A., Springell R.S., Liu X., Monney C., Zhou K.J., Konik R.M., Wen J.S., Xu Z.J., Gu G.D., Strocov V.N., Schmitt T., Hill J.P. Phys. Rev. Lett. 110, 147001 (2013)

38. Dean M.P.M., Dellea G., Springell R.S., Yakhou-Harris F., Kummer K., Brookes N.B., Liu X., Sun Y-J., Strle J., Schmitt T., Braicovich L., Ghiringhelli G., Bozovic I., Hill J.P. Nat. Mater. 12, 1019 (2013)