arXiv:1410.6383v1 [quant-ph] 23 Oct 2014 ntelmto uedmig h esni h missing the is reason The fails damping. huge equation a LL of the limit the while in physics correct the describes not is a motion damping. contain this the not by that origi- affected does means the which term In constant, precession damped. precessional damping the are the equation term LL terms, relaxation nal the both fact as equation the well is LLG as The equation the LLG reduced. in and be that LL will the energy between the field. difference effective relaxation the of re- this direction a During the term into second the the precessional of and a laxation field describes effective term an first within The motion spin. dynam- single the a equation. of describing ics terms (LLG) two equa- contain is Landau-Lifshitz-Gilbert this equations equation Both therefore called equation, Gilbert is LL the tion original of the form to explicit differential similar implicit The an is purely equation. equation, original a Gilbert Gilbert’s contains the the 1932 that equation, [19]. in fact term Lifshitz damping the M. phenomenological E. of L. and reaction by Landau proposed a D. was which as equation [18] (LL) Landau-Lifshitz 1955 in Gilbert equation. most LLG the the However, is [17]. equation equation important more Ishimori the are equation or Landau-Lifshitz-Bloch There [14–16] the the like [13], nearly life. equation of in Bloch daily dynamics dynamics our the describing the of equations devices Micromagnetism describes magnetic of which all equation field main the most the [11] [12], the is equations In Maxwell equation the this motion. with Together (LLG) of Landau-Lifshitz-Gilbert [10]. equation the equation an is equation have this to cases structures need magnetic is dy- of the dynamics it in the ferromagnetic focused describe To the mostly like is namics. interest areas the established fields [6–9] the new in an- these as field all well the In as or in [1] [2–5]. fields magnonics spintronics of new tiferromagnetic field more the and like more magnetism provides technology puter n15 iuh 2]hssonta h L equation LLG the that shown has [20] Kikuchi 1956 In L. T. by introduced originally was equation LLG The com- powerful more and faster of need increasing The h orsodn unu ehncltm eedn Schr¨ ta dependent to time operator mechanical Hamilton quantum corresponding non-Hermitian The a of assumption the is ASnmes 57.n 51.m 75.10.Hk 75.10.Jm, 75.78.-n, propose numbers: been PACS have dynamics spin mechanical quantum the a e into difference Furthermore, Landau-Lifshitz-Gilbert classical and discussed. the to similarities been corresponding have the dynamics and spin described chanical been have tion h lsia aduLfht qainhsbe eie fr derived been has equation Landau-Lifshitz classical The .INTRODUCTION I. ntttfu neadePyi n etu f¨ur Mikrostruk Zentrum und Physik f¨ur Angewandte Institut nvri¨tHmug ugusrse1,D235Hamburg D-20355 11, Jungiusstrasse Universit¨at Hamburg, unu pndynamics spin Quantum Dtd coe 4 2014) 24, October (Dated: .Wieser R. Γ oee,tenr a ecnevdb replacing by conserved be can norm the However, ltos h ulcto nswt umr (Sec. summary a with ends sim- temper- publication VI). computer include The the sec- to into how previous the ulations. fluctuations ways the quantum and two of and calculations describes results ature V numerical the by Sec. IV proved last be Sec. will system In multi-spin tions a III). single to a extended for and (Sec. motion derived of be equation will the spin II, Sec. in troduction, [27]. made mistakes publication the former correct the shall in and description dynamics spin mechanical the quantum of and classical the tween 26]. phe- a [10, later known using added are derivations description been complete classical has However, term In [23–25] damping term. nomenological. publications the precessional previous the cases, of as all derivation well the only as describe it Text 22] makes transition. [21, and quantum-classical physics books the underlying study the to possible in which mechanics insite quantum new from gives derived be can equation equation. LL the in precession the for damping nian h te adsc aitna osntcnev the function conserve wave not the does of Hamiltonian norm a such hand other the λ ˆ h − ∈ h auciti raie sflos fe hsin- this After follows. as organized is manuscript The be- difference the discusses paper the Furthermore, LLG the that show to is publication this of goal The ti elko atta o-emta Hamilto- non-Hermitian a that fact know well a is It uto n w ast nld temperature include to ways two and quation R n H Γ ˆ ˆ 0 + = i nti aew h olwn aefunction: wave following the we case this In . = mqatmmcais trigpoint Starting mechanics. quantum om dne,LovleadHiebr equa- Heisenberg and Liouville odinger, osat edt nrydsiain[83] On [28–30]. dissipation energy to lead constant, a ewe lsia n unu me- quantum and classical between s h eteeeg ispto noaccount. into dissipation energy the ke iedpnetSchr¨odinger equation dependent time ψ H ˆ d. ( I QAINO MOTION OF EQUATION II. − t ) | ψ iλ | ψ ( ,with Γ, ( ˆ t t ) ) i i = = turforschung, Germany , e h − ψ and H ˆ i 0 H ˆ | t e e | − ψ − 2 ( λ λ emta prtr and operators Hermitian Γ ˆ t Γ ˆ Γ ) ˆ t i t e | exp( = ψ λ h 0 Γ ˆ i i t = | ψ e 0 − − i i 2 , H λ ˆ h t Γ ˆ ) i | ψ t . 0 i : by Γ ˆ (2) (1) 2 which is conserved: n = 1. The corresponding Then, the Heisenberg equation appears after replacing Schr¨odinger equation is given by: the expectation values hSˆi by Sˆ: d d i |ψ(t)i = (Hˆ − iλ[Γˆ − hΓˆi])|ψ(t)i . (3) Sˆ = −i[Sˆ, H]ˆ − λ [Sˆ, H]ˆ + − 2hHˆiSˆ . (11) dt dt   The energy dissipation itself depends on λ and Γ.ˆ In the Now, the Heisenberg equation (11) as well as the Liouville following we assume that Γˆ is equal Hˆ and therefore, Hˆ = equation (5) can be used to derive the Landau-Lifshitz Hˆ − iλ(Hˆ − hHˆ i). With this assumption the Schr¨odinger equation. Therefore, we have to insert these equations equation becomes [31]: in (7) and to determine the traces. However, we need to know the exact form of the density operatorρ ˆ. d In the case of a Heisenberg spin system with spin quan- i |ψ(t)i = (Hˆ − iλ[Hˆ − hHˆi])|ψ(t)i . (4) dt tum number S the density operator is given by the fol- lowing multivector expansion [32]: We will see that this Schr¨odinger equation can be seen as the quantum mechanical analog of the classical Landau- 1 1 1 ρˆ = 1ˆ + hSˆmiSˆm + hSˆmliSˆml + ... 2S +1 nS n S Lifshitz equation, where λ is the damping constant. m 2 ml Now, it is quite easy to show that this equation corre- X X (12) sponds to the following Liouville (von Neumann) equa- The first term is the Identity matrix which behaves under tion [27]: rotation like a scalar. The next term is the sum over the three spin matrices Sˆ , Sˆ , and Sˆ . These matrices dˆρ x y z = i[ˆρ, H]ˆ − λ[ˆρ, [ˆρ, H]]ˆ . (5) behave under rotation like a vector. The next term is the dt sum over bivectors We assume that the system is in a pure state therefore 1 1 Sˆ = [Sˆ , Sˆ ] − S(S + 1)δ , (13) the density operator is given by: ml 2 m l + 3 ml ρˆ = |ψ(t)ihψ(t)| . (6) and the next term (not shown) is the sum over trivec- tors Sˆmlk and so on. All higher order terms have the Equation (5) has already the form of the Landau-Lifshitz same scheme as the previous ones but with higher order equation because the commutator [ , ] in the case of spin tensors. The prefactors nS and n2S are given by traces systems lead to a vector product. nS = Tr(SˆaSˆa) and n2S = Tr(SˆabSˆab) (these values only So far, we have derived the quantum mechanical Liou- depend on the spin quantum number S and are indepen- ville equation. The next step is the Heisenberg equation. dent of the direction a,b ∈ {x,y,z}). The prefactors of Therefore, we concentrate ourself first on the time de- the higher expansion terms are similar. Sˆ pendence of the expectation value h i: This expansion is highly related to the magnetic mul- tipol expansion [33], and the number of terms depend on d d hSˆi = Tr ρˆSˆ . (7) the spin quantum number S. In the case S = 1/2 the dt dt   expansion ends with the vector term Including the Liouville equation (5) in Eq. (7), written 1 in the alternative form: ρˆ = 1ˆ + hσˆmiσˆm . (14) 2 m ! dˆρ X = i[ˆρ, H]ˆ − λ [ˆρ, H]ˆ − 2ˆρHˆˆρ , (8) dt + Here,σ ˆm are the Pauli matrices. In the case S = 1 the   expansion ends with the bivector term and get immediately the following differential equation: 1 1 1 ρˆ = 1ˆ + hSˆ iSˆ + hSˆ iSˆ , (15) d 3 2 m m 2 ml ml hSˆi = −ih[Sˆ, H]ˆ i− λ h[Sˆ, H]ˆ i− 2hHˆihSˆi . (9) m ml dt + X X   and in the case S = 3/2 the expansion ends with the Here, the [ , ] is the anticommutator and we have used + trivector term, and so on. the fact that the trace is invariant under cyclic permuta- After this small digression let us come back to the tions. Furthermore: derivation of the Landau-Lifshitz equation. We have al- ready noticed that it is necessary to insert the Liouville Tr ρˆHˆˆρSˆ = hm|ψihψ|Hˆ |ψihψ|Sˆ|ψihψ|mi equation (5) or the Heisenberg equation (11) in Eq. (7) m   X to derive a complete solvable equation for the expecta- = hψ|Hˆ |ψihψ|Sˆ|ψihψ|mihm|ψi tion value hSˆi. The result is the following equation for m X the components of Sˆ (n ∈{x,y,z}): = hψ|Hˆ |ψihψ|Sˆ|ψi = hHˆihSˆi d ˆ Sˆ Tr ρˆSˆn = iTr [ˆρ, H]ˆ Sˆn − λTr [ˆρ, [ˆρ, H]]ˆ Sˆn (16). = Tr ρˆH Tr ρˆ . (10) dt           3

Insertingρ ˆ in (16) and calculating the traces lead to an A complete evidence of the previous statement and de- equation for hSˆi which is similar to the Landau-Lifshitz tails of the calculation can be found in the supplementary equation if the Heisenberg Hamilton operator can be of this publication [34]. written as external field H=ˆ −Beff · S.ˆ Now, it is known that the LL equation will lead to un- During the calculation we can ignore the first term ofρ ˆ physical results in the limit of a large damping λ ≫ 1. because the identity matrix 1ˆ commutates with any oper- The reason is that the precessional motion is not influ- ator and therefore all the commutators with the identity enced by the damping. The equation which holds in the matrix become zero. Furthermore, we can see that only limit of a large damping is the Landau-Lifshitz-Gilbert ˆ ˆ the vector term mhSmiSm participate to the differen- (LLG) equation. In the case of the LLG equation both tial equation of hSˆi. All the higher order expansion terms terms the precessional term as well as the relaxation term P (bivector, trivector, ...) are not contributing. The rea- are influenced by the damping. The easiest way to obtain son for that is the conservation of the rank k of a tensor the LLG equation from the LL equation is to make the 2 under rotation. Now, each term of the multivector ex- following transformation of the time t: t → t/(1 + λ ). pansion has its own rank k. The expansion starts with The same transformation in the case of Schr¨odinger equa- the Identity matrix with rank k = 0 (scalar). The vector tion (4) leads to: term has rank k = 1, the bivector rank k = 2, the trivec- d tor rank k = 3, and so on. The rank of each additional i(1 + λ2) |ψ(t)i = (Hˆ − iλ[Hˆ − hHˆ i])|ψ(t)i . (22) term increases by a factor 1. dt Sˆ Now, we assume that the trajectories of h i shall be- This equation can be seen as the quantum mechanical have classical. This means that the length of the spin is analog of the LLG equation: conserved, and the spin only fulfills precession and relax- ation. However, any precession as well as relaxation can d 1 λ hSˆi = hSˆi× B − hSˆi× hSˆi× B . be described as a rotation of the coordinate system and dt 1+ λ2 eff 1+ λ2 eff we have already mentioned that the rank of a tensor is  (23) conserved under such a rotation. This also means that The classical LLG equation itself appears by replacing ˆ ˆ the motion of hSni is only described by the tensors Sn the spin expectation value hSˆi by the classical spin S. which are responsible for hSˆni:

hSˆni = Tr(ˆρSˆn) III. SINGLE SPIN VS. MULTI-SPIN SYSTEM 1 1 = Tr(Sˆ ) + hSˆ i Tr(Sˆ Sˆ ) 2S +1 n n m m n S m This could be the end of the story. However, there is =0 X =nS δnm one point which has not been discussed so far. In the clas- 1 sical description we assume a constant length of the spin: + |hSˆ{zmli}Tr(SˆmlSˆn) +0+| ...{z+0}.(17) n S S 2 ml | | = 1. The LLG equation at zero temperature just de- X =0 Eq. (18) scribes a precessional motion resp. relaxation due to an B | {z } | {z } effective field eff . In the quantum mechanical descrip- The higher order expansion terms disappear due to tion we deal with the expectation values hSˆi = hψ|Sˆ|ψi

′ which strongly depend on the wave functions |ψi. The ˆk ˆk ′ Tr SαSβ = nSk δαβδkk . (18) absolute value |hSˆi| of these expectation values have not   necessarily to be constant: |hSˆi| ≤ ~S. Furthermore, due k and k′ are the rank of the tensor and α and β its to the time dependence of the wave function |ψi = |ψ(t)i, ˆ1 ˆ ˆ2 ˆ components: e.g. S3 = Sz, or S4 = Szx. |hSˆi| is also time dependent. This means that in general ˆ ˆ Inserting the vector term mhSmiSm ofρ ˆ in Eq. (16) we cannot expect an agreement of the classical with the we find after some algebra: quantum mechanical spin dynamics. On the other hand, P we find in textbooks the statement that we can expect a d classical behavior, at least for a single spin in an external hSˆni = hSˆi× Beff − λ hSˆi× hSˆi× Beff . dt n n field and without damping [22]. This textbook statement      (19) of a semiclassical description of the spin dynamics is not or written as complete vector: totally wrong. In the case of single spins and in the case of linear excitations of spin system in the ferromagnetic d hSˆi = hSˆi× Beff − λhSˆi× hSˆi× Beff . (20) ground state ( excitation) the classical and the dt quantum mechanical description can lead to the same   Sˆ This equation is similar to the classical Landau-Lifshitz trajectories if the Hamiltonian is linear in n (n stands (LL) equation [19]: for the lattice site and not for a spin component. For details see [25]). d To specify the description let us first assume a single S = S × B − λS × (S × B ) . (21) dt eff eff spin which is in a pure state |ψ(t)i. In this case the 4 density operator is given by Eq. (6) and we find Tr(ˆρ2)= where |ψi is given by (26) or (27), or by: Tr(ˆρ) = 1. For comparison, in the case of a mixed state: hSˆni = Tr ρˆSˆn . (29) ρˆ = pi|ψiihψi| (24)   i In the following we assume a pure state. Therefore, the X density operator of our multi-spin systemρ ˆ is defined by we findρ ˆ2 6=ρ ˆ, therefore Tr(ˆρ2) < Tr(ˆρ) = 1. Then, we ρˆ = |ψihψ|. 2 can write Tr(ˆρ ) in the case of a single spin with S =1/2: Alternativ we can calculate the expectation value hSˆni

with aid of the reduced density operatorρ ˆmn [35]: 1+ |hσˆi|2 Tr(ˆρ2)= ≤ 1 . (25) 2 Sˆ Sˆ Sˆ h ni = Trn Trj6=n ρˆ n = Trn ρˆmn n , (30) A similar formula can be given for higher spin quantum      numbers (S > 1/2) due to Eq. (18). However, in these where Trn is the partial trace over the sub Hilbert space cases it is impossible to give a geometric plot as for S = Hn and Trj6=n the partial trace over all the other sub 1/2 (Bloch sphere). Hilbert spaces: H = H1 ⊗ H2 ⊗ ...⊗ HN . The advantage of using the reduced density operatorρ ˆ is that we Tr(ˆρ2) = 1 appears in the case of a pure state which mn have to deal with smaller matricies, which reduces the corresponds to |hσˆi| = 1 resp. |hSˆi| = ~/2. In the case 2 numerical effort. of a spin ensemble in a mixed state we have Tr(ˆρ ) < 1 It is easy to verify that the reduced density operator corresponding to |hSˆi| < ~/2. This means, in the case ρˆmn is given by: of a single spin in a pure state the norm of the expecta- tion value |hSˆi| is constant and the dynamics can become ρˆmn = pmn |SnmnihSnmn| . (31) classical [see Eq. (23)]. mn In the case of a multi-spin system the wave function X is described by a product of the eigenstates of the single Eq. (31) clearly shows that the reduced density operator spins: corresponds to a mixed state, the pmn are the probabil- ities to find the system in the state |Snmni, even if the |ψi = |S1m1i ⊗ |S2m2i⊗ ... ⊗ |SN mN i . (26) complete system is in a pure stateρ ˆ = |ψihψ|. In the case of the product state (26) the reduced den-

Such a state is called product state. For example the fer- sity operator is given byρ ˆmn = |SnmnihSnmn|. Here, romagnet state is described by a product state: |FMi = the reduced density operatorρ ˆmn as well as the density | ↑↑ ... ↑i = |↑i⊗|↑i⊗ ... ⊗| ↑i. operator of the complete systemρ ˆ = |ψihψ| are pure. However, product states are normally not the general The consequences are as long as the system is described states of a magnetic system. The normal state in a quan- by a product state we find: |hSˆni| = ~S. However, tum mechanical system with more then one spin is a su- as mentioned before the normal states are superposition perpositions of the product states (26): states which are entangled and not product states. In these cases we find |hSˆni| < ~S (see discussion before about the density operators of pure and mixed states). |ψi = cm1m2...mN (|S1m1i ⊗ |S2m2i⊗ ... ⊗ |SN mN i) . ˆ m1...mN The explicit value of |hSni| corresponds to the strength of X (27) the entanglement: no entanglement means |hSˆni| = ~S The product states in this case cannot be separated which which is the maximal value and with entanglement |hSˆni| that we cannot write cm1m2...mN = cm1 cm2 ...cmN . This decreases and can become zero. Due to the fact that the means that we have a correlation between the spins which entanglement depends on |ψi and |ψi can change with do not exist in classical physics. This phenomenon is the time t, |hSˆni| can also change with the time. called entanglement [35, 36] and leads to the fact that Another way to quantify the entanglement is the von in the case of two entangled spins a measurement on the Neumann entropy: first spin also affects the second spin [37]. The same is true for more then two spins. In this case the measure- S(ˆρmn )= −Tr(ˆρmn log2ρˆmn ) , (32) ment on one spin affects all spins. Only in the case of a product state the subsystems are not correlated and can whereρ ˆmn is the reduced density matrix. be separated, which means that the measurement of a The von Neumann entropy is the extension of the clas- single spin in this case only affect this spin and do not sical Gibbs entropy concept to the quantum mechanics change the wave functions of the other spins as in the and can be seen as the quantum mechanical analog to the case of entanglement. classical Shannon entropy [38] of the information technol- Now, the expectation value hSˆni, where n is the lattice ogy. site, can be calculated similar to a single spin by: In principle the von Neumann entropy is defined for any density matrixρ ˜. In general, the von Neumann en- hSˆni = hψ|Sˆn|ψi , (28) tropy proves if the system is in a pure or in a mixed state. 5

Therefore, it makes sense to use the reduced density ma- 1 1 trixρ ˆmn because in these cases a pure state corresponds 0.5 ˆ to a product state, and therefore to no entanglement. hSyi 0 0.5 -0.5 In these cases we find S(ˆρmn ) = 0. A mixed state cor- -1 0 1 -0.5 ˆ 0.5 responds to an entangled superposition state. In these ˆ hSxi hSαi 0 cases we find 0 < S(ˆρmn ) ≤ 1. The maximum value ˆ hSxi Sx Sˆ ˆ S(ˆρmn ) = 1 corresponds to |h ni| = 0. Furthermore, as hSyi Sy -0.5 ˆ Sˆ hSzi Sz before |h ni|, S(ˆρmn ) is time dependent. -1 0 5 10 15 IV. NUMERICAL PROOF time t

FIG. 1: reversal of a single spin. For com- In the previous section we have seen that only in the parison the quantum mechanical expectation values hSˆni cases of a single spin or a spin ensemble which behaves n ∈ {x,y,z} and the classical spin components have been like a single spin and in the cases of a linear excita- plotted as function of time. The inset shows the precessional tion around the ferromagnetic ground state we can find motion during the reversal. (S = 1, J = 0, µS Bz = −5.1, x a quantum mechanical spin dynamics similar to the clas- µS B0 = 3.27, t0 = 2, TW = 0.02, λ = 0.2) sical spin dynamics. In these cases we have no entangle- ment and the absolute value |hSˆi| stays constant. In the cases of entanglement this is not the case. To proof this the first and the last spin. The Hamiltonian (33) is linear ˆ statement computer simulations have been performed. In in Sn and therefore we can expect a classical behavior if the previous publications only linear excitations [25] or we are close to the ferromagnetic ground state. single spins [27] have been investigated. In the following, Fig. 2 shows the trajectories of three ferromagnetic we will discuss the magnetization reversal of a single spin coupled spins, excited by a tiny field pulse (linear ex- as well as a linear excitation and a magnetization reversal citation), with S = 1. The field pulse acts on the first of a trimer. spin only. In the initial configuration is: all spins are ori- The Heisenberg Hamiltonian of our system is given by: ented in +z-direction, the direction of the external field Bz. ˆ Sˆ Sˆ ˆz ˆx H= −J n n+1 − µS BzSn + −µSBx(t)S1 .(33) To show the agreement of the classical spins Sn with n n the expectation values hSˆ i we have compared the curves X X n of the first spin: it can be seen that both curves lie on top The first term describes the nearest neighbor exchange of each other which means that the expectation values interaction with ferromagnetic (J > 0) or antiferromag- behave classical. The comparison between classical and netic (J < 0) coupling. quantum mechanical trajectory of the other two spins The second sum and third term describe couplings to show the same behavior (not shown). external magnetic fields. In this case a static field in z- direction and a field pulse in x-direction acting on the 0.2 ˆ first spin only: hS1i S1 ˆ hS2i t−t 2 0.15 hSˆ i − 1 0 3 x 2  TW  Bx(t)= B0 e . (34) 0.1 ˆy This field pulse will be used to bring the system out of hSni equilibrium and to start the reversal process or just for 0.05 linear excitations. For the moment we restrict ourself to a single spin [no 0 exchange term in (33)]. The spin is initially oriented in -0.05 +z direction: |ψ(t = 0)i = | ↑i. The static external field -0.1 -0.05 0 0.05 0.1 hSˆxi in −z-direction. The field pulse is needed to break the n symmetry and to initialize the magnetization reversal. Sˆ FIG. 2: Linear excitation and additional relaxation of three Fig. 1 shows the trajectories of h i calculated with the ferromagnetic coupled spins: only the first spin has been ex- Schr¨odinger equation (22) and for comparison the trajec- cited by the field pulse. A perfect agreement between quan- tory of the classical spin S calculated with the classical tum mechanical and classical trajectory can be seen for the Landau-Lifshitz-Gilbert equation. As expected, Fig. 1 first spin. The same is true also for the other spins. (S = 1, x shows a perfect agreement of both, the classical as well J = 1, µS Bz = 0.1, µS B0 = 3.27, t0 = 10, TW = 0.02, as quantum mechanical trajectories, which also means λ = 0.1) that |hSˆi| is conserved. For a detailed discussion please see Sec. II and III. As mentioned before such an agreement can be ex- Let us come to the trimer. All spins are aligned in a pected only for linear excitations of the ferromagnetic chain and we assume that there is no coupling between ground state, in this case |FMi = | ↑↑↑i. This is the 6 case within the simulation. This can be seen by the corresponds to the occupation of one of the basis states ˆx ˆy small amplitude of the hSni and hSni component. Fur- |m1m2m3i corresponding to |+3/2i = | ↑↑↑i (basis state thermore, the simulation shows that only the basis state with three spins up), | +1/2i (all basis states with two |Sm1i ⊗ |Sm2i ⊗ |Sm3i = |m1m2m3i = | ↑↑↑i gives spins up), |− 1/2i (all basis states with two spins down), a relevant impact: |h↑↑↑ |ψ(t)i|2 ≈ 1. All other basis and |− 3/2i = | ↓↓↓i (basis state with three spins down) ˆz ˆz states |m1m2m3i are not occupied or just marginal with [see Fig. 3b)]. The oscillation of hS1 i and hS3 i around 2 |hm1m2m3|ψ(t)i| < 0.01. Furthermore, additional sim- ˆz hS2 i is a direct result of the field pulse acting on the first ulations show that with increasing S the deviation from spin only and the exposed position of the second spin as |ψ(t)i = |FMi, up to which we can see a classical behav- the middle of this three spin cluster. ior, increases . This becomes clear because with S → ∞ ˆx ˆy During the reversal process hSni and hSni [Fig. 3c)] we find the classical limit. show a precessional motion of the spins similar to the

a) b) magnetization reversal of the single spin before [see 1 ✻ 1 ✻ Fig. 1]. Bx-pulse h↑↑↑ |ψ(t)i h↓↓↓ |ψ(t)i Bx-pulse i

0.5 ) 0.8 t

( The most important result is shown in Fig. 3d): the not ψ | 0.6 hSˆzi 3 Sˆ n 0 m conserved absolut value |h i|. This means, we cannot

2 n 0.4 1 1 m h+ |ψ(t)i h− |ψ(t)i 1 2 2 expect an agreement with the classical trajectories where -0.5 m h 0.2 1 2 3 we assume a constant spin length |Sn| = 1. As mentioned -1 0 0 20 40 60 80 100 0 20 40 60 80 100 Sˆ ~ time t time t before |h ni| < S means that the system is entangled, c) 1 d) which can be seen also by Fig. 3b). The three spins 0.75 1 are in a product state (FM) only at the beginning and 0.5 0.8 0.25 at the end of the reversal process. In between, we see ˆα 0.6 hS2 i 0 ˆ |hSni| superpositions of the basis states |m1m2m3i = |Sm1i⊗ -0.25 0.4 -0.5 Sˆx |Sm2i ⊗ |Sm3i which means entanglement. h 2 i 0.2 -0.75 ˆy hS2 i To strengthen this message the von Neumann entropy -1 0 0 20 40 60 80 100 0 20 40 60 80 100 time t time t S(ˆρm1 )= −Tr(ˆρm1 log2ρˆm1 ) , (35) FIG. 3: Magnetization reversal for three ferromagnetic cou- ple spins. a) z-component of the spin as function of time t, for the reduced density operatorρ ˆm1 , corresponding with b) projection of the wave function |ψ(t)i to the basis states the Hilbert space of the first spin of our three spin system, |m1m2m3i, c) precession motion of the second spin during has been calculated. For the details of the calculation the reversal process, d) spin length as function of time t. x please see the supplementary material [34]. (S = 1/2, J = 4, µS Bz = −2, µS B0 = 3.27, t0 = 10, Fig. 4 shows the von Neumann entropy S(ˆρm1 ) as TW = 0.02, λ = 0.1) function of time t. It can be seen that ferromagnetic states at the beginning and end: | ↑↑↑i and | ↓↓↓i show no entanglement. Furthermore, we see that the highest Sˆ 1 entanglement [max. S(ˆρm1 )] appears when the |h ni|, n ∈{1, 2, 3}, have their smallest values. 0.8 ) 1 m ρ

(ˆ 0.6 S V. TEMPERATURE 0.4

entropy 0.2 The previous sections describe the quantum spin dy- namics with energy dissipation at zero temperature. 0 Within this section we discuss two possibilities to include 0 20 40 60 80 100 temperature effects. The easiest way is to add to the time t Hamilton operator Hˆ [Eq. (33)] an additional stochastic field term: FIG. 4: Von Neumann entropy S(ˆρm1 ) vs. time t, correspond- ing to the magnetization reversal of the three spin system ˆ Sˆ presented in Fig. 3. Hξ = − ξn n . (36) n X x y z The situation changes if we leave the ferromagnetic ξn = (ξn, ξn, ξn) is a stochastic field characterized by α α β ′ ′ ground state. Fig. 3 shows the magnetization reversal of hξn (t)i = 0, and hξn (t)ξm(t )i = Dδnmδαβδ(t − t ), with three coupled spins with S =1/2. The initial configura- α, β ∈{x,y,z}, and lattice sites n,m. Depending on the tion is the same as before all three spins in +z-direction, prefactor D the stochastic field can be used to describe however the external field is now −z-direction. Again, temperature or quantum fluctuations [39]. the field pulse excites the first spin only. The result is a The advantage of such an stochastic field is that this stepwise magnetization reversal [see Fig. 3a)]. Each step term is already a field term. This means that this 7 term behaves classical. In other words: the Heisen- It is known that the Landau-Lifshitz equation fails berg Hamiltonian Hˆ = −(Beff + ξ)Sˆ inserted in the in the huge damping limit [20]. However, the Landau- Schr¨odinger equation (22) immediately leads to the fol- Lifshitz-Gilbert (LLG) equation can be simply obtained lowing Langevin equation which can be seen as the quan- from the Landau-Lifshitz equation by rescaling the time. tum mechanical analog of the stochastic Landau-Lifshitz- The LLG equation itself shows the correct physics. The Gilbert equation [40, 41]: same rescaling has been used to derive the time depen- dent Schr¨odinger Eq. (22) which corresponds to the clas- d 1 λ sical Landau-Lifshitz-Gilbert Eq. (23). hSˆi = hSˆi× Beff − hSˆi× hSˆi× Beff dt 1+ λ2 1+ λ2 During the complete derivation and later in the 1 λ   manuscript the similarities and differences of the clas- + hSˆi× ξ − hSˆi× hSˆi× ξ (37) 1+ λ2 1+ λ2 sical and quantum spin dynamics have been discussed.   It has been shown that only for single spins with a Zee- Similar, the quantum mechanical analog of the stochas- man Hamiltonian H=ˆ −B · Sˆ and in the case of a ferro- tic Landau-Lifshitz can be derived using the same Heisen- magnetic multi-spin system we can expect a classical be- berg Hamiltonian together with the Schr¨odinger equation havior, which means similar trajectories of the quantum (4) or by making the same transformation we have used mechanical expectation values hSˆni and classical spins to derive the Schr¨odinger equation (22). Sn. (The index n stands for the nth spin.) In all other An alternative way to include temperature, where we cases we expect a deviation of the trajectories of the ex- assume that the system is already in equilibrium, is to pectation values hSˆni from the classical behavior. The use statistical operator [35, 42]: reasons are the appearance of additional terms of the or- der of ~ in the Heisenberg equation which disappear in ˆ e−βH the classical limit: S → ∞ and ~ → 0 and in the case ρˆStat. = , (38) Tr e−βHˆ of a multi-spin system the appearance of entanglement. The entanglement itself is a pure quantum effect and also   disappears in the classical limit. with β the well known inverse temperature β = 1 . kB T All the results have been discussed theoretical with Then, the time dependence ofρ ˆ . is given by: Stat analytical calculations and proved with computer simu- + lations. In all cases the theoretical estimated effects have ρˆStat.(t)= Uˆ(t)ˆρStat.(0)Uˆ (t) . (39) been seen in the simulations. Uˆ(t) is the unitary operator In the last section of the manuscript two ways have been shown how to include temperature which has not ˆ ˆ ˆ Uˆ(t)= e−iHte−λHteλhHit , (40) been taken into account in the previous sections (Sec. II - IV). The first way is to add a random noise. This way which we have already seen in Eq. (2). At this point, we allows us to investigate the dynamics under the influence have to notice that Eq. (39) is a self-consistent equation, of temperature as well as quantum fluctuations. The sec- because the unitary operator Uˆ(t) contains hHˆ i which has ond way deals with the statistical operator. In this case fluctuations play no role. This way can be used if we to be calculated withρ ˆStat.: hHˆ i = Tr(ˆρStat.H).ˆ However, are interested in system which are always in the ther- ρˆStat. already needs hHˆ i to be calculated. Alternatively, we can use (38) in the Liouville equation (5) to describe modynamical equilibrium. This means, non-equilibrium the dynamics. effects cannot be described with this way.

VI. SUMMARY Acknowledgments In summary it has been show that the following non- Hermitian Hamilton operator Hˆ = Hˆ − iλHˆ inserted The author thanks D. Altwein, M. Krizanac, E. in the time dependent Schr¨odinger equation leads to Vedmedenko, Y. Saadi, M. Maamache, and P.-A. an equation of motion similar to the classical Landau- Hervieux for helpful discussions and comments. This Lifshitz Eq. (21). To show this the corresponding Liou- work has been supported by the Deutsche Forschungsge- ville Eq. (5) as well as Heisenberg Eq. (11) have been meinschaft in the framework of subproject B3 of the SFB derived and discussed. 668 and by the Cluster of Excellence “Nanospintronics”.

[1] V. V. Kruglyak, S. O. Demokritov, and D. Grundler, J. Phys. Rev. Lett. 101, 177202 (2008). Phys. D 43, 264001 (2010). [3] R. Wieser, E. Y. Vedmedenko, and R. Wiesendanger, [2] R. Wieser, E. Y. Vedmedenko, and R. Wiesendanger, Phys. Rev. Lett. 106, 067204 (2011). 8

[4] A. H. MacDonald and M. Tsoi, Phil. Trans. R. Soc. A Scientific, Singapore, 2008). 369, 3098 (2011). [23] M. Lakshmanan, Phil. Trans. R. Soc. A 369, 1280 (2011). [5] V. M. T. S. Barthem, C. V. Colin, H. Mayaffre, M. Julien, [24] A. Sakuma, arXiv, 0602075v2 (2006). and D. Givord, Nat. Commun. 4, 2892 (2013). [25] R. Wieser, Phys. Rev. B 84, 054411 (2011). [6] S. S. P. Parkin, M. Hayashi, and L. Thomas, Science 320, [26] M. F¨ahnle, D. Steiauf, and J. Seib, J. Phys. D 41, 164014 190 (2008). (2008). [7] C. Schieback, M. Kl¨aui, U. Nowak, U. R¨udiger, and [27] R. Wieser, Phys. Rev. Lett. 110, 147201 (2013). P. Nielaba, Eur. Phys. J. B 59, 429 (2007). [28] V. Weisskopf and E. Wigner, Z. Physik 65, 18 (1930). [8] D. C. Ralph and M. D. Stiles, J. Magn. Magn. Mat. 320, [29] R. L. Liboff and M. A. Porter, Physica D 195, 398 (2004). 1190 (2008). [30] R. Kosik, Ph.D. thesis, TU Wien (2004). [9] R. Wieser, E. Y. Vedmedenko, P. Weinberger, and [31] N. Gisin, Helv. Phys. Acta 54, 457 (1981). R. Wiesendanger, Phys. Rev. B 82, 144430 (2010). [32] H. F. Hofmann and S. Takeuchi, Phys. Rev. A 69, 042108 [10] T. L. Gilbert, IEEE Trans. Mag. 40, 3443 (2004). (2004). [11] J. D. Jackson, Classical Electrodynamics (John Wiley & [33] E. Y. Vedmedenko and N. Mikuszeit, Chem. Phys. Chem. Son, New York, 1999). 9, 1222 (2008). [12] W. F. Brown, (Wiley, New York, 1963). [34] See supplementary material. [13] F. Bloch, Phys. Rev. 70, 460 (1946). [35] D. A. Garanin, Adv. Chem. Phys. 147, 213 (2011). [14] D. A. Garanin, Physica A 172, 470 (1991). [36] P. Krammer, Ph.D. thesis, Universit¨at Wien (2009). [15] D. A. Garanin, Phys. Rev. B 55, 3050 (1997). [37] A. Einstein, B. Podolsky, and N. Rosen, Phys. Rev. 47, [16] R. F. L. Evans, D. Hinzke, U. Atxitia, U. Nowak, R. W. 777 (1935). Chantrell, and O. Chubykalo-Fesenko, Phys. Rev. B 85, [38] C. E. Shannon, The Bell System Technical Journal 27, 014433 (2012). 379 & 623 (1948). [17] Y. Ishimori, Prog. Theor. Phys. 72, 33 (1984). [39] V. L. Pokrovsky and N. A. Sinitsyn, Phys. Rev. B 67, [18] T. L. Gilbert and J. M. Kelly, in Proceedings of the Pitts- 144303 (2003). burgh Conference on Magnetism and Magnetic Materials [40] J. L. Garc´ıa-Palacios and F. J. L´azaro, Phys. Rev. B 58, (Am. Inst. Electr. Engrs., October, 1955), p. 253. 14937 (1998). [19] D. L. Landau, Phys. Z. Sowjetunion 2, 46 (1932). [41] U. Nowak, in Annual Reviews of Computational Physics [20] R. Kikuchi, J. Appl. Phys. 27, 1352 (1956). IX, edited by D. Stauffer (World Scientific, Singapore, [21] H. C. Fogedby, Theoretical Aspects of Mainly Low Di- 2001), p. 105. mensional Magnetic Systems (Springer-Verlag, Berlin [42] K. Blum, Density Matrix Theory and Applications Heidelberg New York, 1980). (Plenum Press, New York and London, 1996). [22] B. Guo and S. Ding, Landau-Lifshitz equations (World