Magnetic monopole and string excitations in two-dimensional spin ice

Download Magnetic monopole and string excitations in two-dimensional spin ice

Post on 15-Apr-2017




0 download


  • Magnetic monopole and string excitations in two-dimensional spin iceL. A. Ml, R. L. Silva, R. C. Silva, A. R. Pereira, W. A. Moura-Melo, and B. V. Costa Citation: Journal of Applied Physics 106, 063913 (2009); doi: 10.1063/1.3224870 View online: View Table of Contents: Published by the AIP Publishing Articles you may be interested in Spin-wave spectra and stability of the in-plane vortex state in two-dimensional magnetic nanorings J. Appl. Phys. 114, 233906 (2013); 10.1063/1.4851695 New physics in frustrated magnets: Spin ices, monopoles, etc. (Review Article) Low Temp. Phys. 39, 901 (2013); 10.1063/1.4826079 Formation of thermally induced ground states in two-dimensional square spin ices J. Appl. Phys. 112, 043909 (2012); 10.1063/1.4747910 Emergent magnetic monopoles, disorder, and avalanches in artificial kagome spin ice (invited) J. Appl. Phys. 111, 07E103 (2012); 10.1063/1.3670441 Gravitational magnetic monopoles and Majumdar-Papapetrou stars J. Math. Phys. 47, 042504 (2006); 10.1063/1.2184766

    [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: Downloaded to ] IP: On: Sun, 21 Dec 2014 01:11:26

  • Magnetic monopole and string excitations in two-dimensional spin iceL. A. Ml,1,2 R. L. Silva,1 R. C. Silva,1 A. R. Pereira,1 W. A. Moura-Melo,1,a andB. V. Costa21Departamento de Fsica, Universidade Federal de Viosa, Viosa, Minas Gerais 36570-000, Brazil2Departamento de Fsica, ICEX, Universidade Federal de Minas Gerais, Caixa Postal 702, Belo Horizonte,Minas Gerais 30123-970, Brazil

    Received 17 April 2009; accepted 12 August 2009; published online 23 September 2009

    We study the magnetic excitations of a square lattice spin ice recently produced in an artificial formas an array of nanoscale magnets. Our analysis, based on the dipolar interaction between thenanomagnetic islands, correctly reproduces the ground state observed experimentally. In addition,we find magnetic monopolelike excitations effectively interacting by means of the usual Coulombicplus a linear confining potential, the latter being related to a stringlike excitation binding themonopoles pairs, which indicates that the fractionalization of magnetic dipoles may not be so easyin two dimensions. These findings contrast this material with the three-dimensional analog, wheresuch monopoles experience only the Coulombic interaction. We discuss, however, two entropiceffects that affect the monopole interactions. First, the string configurational entropy may lose thestring tension and then free magnetic monopoles should also be found in lower dimensional spinices; second, in contrast to the string configurational entropy, an entropically driven Coulomb force,which increases with temperature, has the opposite effect of confining the magnetic defects. 2009American Institute of Physics. doi:10.1063/1.3224870


    Geometrical frustration among spins in magnetic mate-rials can lead to a variety of cooperative phases such as spinglass, spin liquid, and spin ice behaving like glass, liquid,and ice in nature. The description and understanding of suchstates are becoming increasingly important not only in con-densed matter but also in other branches such as field theo-ries. In a crystal at low temperature excitations above theground state often behave like elementary particles carryinga quantized amount of energy, momentum, electric charge,and spin. Several of these objects arise as a result of thecollective behavior of many particles in a material, which ismost effectively described in terms of the fractions of theoriginal particles. The emergence of these excitations is anexample of the phenomenon known as fractionalization.This occurrence is often tied to topological defects1 and iscommon in one-dimensional systems polyacetylene, nano-tubes, etc. Higher dimensional fractionalization is more dif-ficult to be found. In two spatial dimensions the only con-firmed case is the involvement of quasiparticles with onethird of an electrons charge in the fractional quantum Halleffect in strong magnetic fields. Among several suggestions,2

    there is also the proposal that the merons forming a skyrmionin two-dimensional 2D Heisenberg antiferromagnets arespinons and therefore, they are neutral spin-halfexcitations.3,4 More recently, examples of fractionalization inthree-dimensional 3D systems were provided in spin icematerials.2,5 Particularly, Castelnovo et al.5 have shown howthe famous magnetic monopole may emerge in these materi-als. Despite some exciting suggestions for its existence from

    the realms of quantum mechanics, a single magnetic poleremains elusive after decades of searching in particle accel-erators and cosmic rays. Now, Castelnovo et al.5 indicated anunexpected but, perhaps, better place to look. Under certainconditions, spin ice magnets behave like a gas of free mag-netic poles. There is even a phase transition at which a thinvapor of these poles condenses into a dense liquid. An ex-perimentally measurable signature of monopole dynamics ona diamond lattice in the grand canonical ensemble was pre-sented in Ref. 6. The existence of these excitations in a con-densed matter system is exciting in itself. Our aim in thispaper is to study spin ice materials, but in two spatial dimen-sions. Such structures have been artificially produced in ageometrically frustrated lattice of nanoscale ferromagneticislands.79 Here, we examine the excitations magneticmonopoles and how they interact in this 2D system.

    3D spin ice materials have the pyrochlore structure inwhich magnetic rare-earth ions form a lattice of corner-sharing tetrahedra. To minimize the spin-spin interaction en-ergy, the ice rules are manifested: two spins point inward andtwo spins point outward on each tetrahedron. A similar sys-tem was built in two dimensions with elongated permalloynanoparticles. This artificial material consists of elongatedmagnetic nanoislands distributed in a 2D square lattice. Thelongest axis of the islands alternate its orientation pointing inthe direction of the two principal axes of the lattice.7 Themagnetocrystalline anisotropy of permalloy is effectivelyzero, so that the shape anisotropy of each island forces itsmagnetic moment to align along the largest axis thus, makingthe islands effectively Ising-like. The intrinsic frustration onthis lattice is similar to that in the spin ice model and can bebest seen by considering a vertex at which four islands meet.A pair of moments on a vertex can be aligned either to maxi-mize or to minimize the dipole interaction energy of the pair.

    aAuthor to whom correspondence should be addressed. Electronic Tel.: 55-31-3899-3417. FAX: 55-31-3899-2483.

    JOURNAL OF APPLIED PHYSICS 106, 063913 2009

    0021-8979/2009/1066/063913/5/$25.00 2009 American Institute of Physics106, 063913-1

    [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: Downloaded to ] IP: On: Sun, 21 Dec 2014 01:11:26

  • As shown in Ref. 7, it is energetically favorable when themoments of a pair of islands are aligned so that one is point-ing into the center of the vertex and the other is pointing outthe two arrangements on the right, Fig. 1, while it is ener-getically unfavorable when both moments are pointing in-ward or both are pointing outward the pair of arrangementson the left, Fig. 1. This artificial system exhibits short-rangeorder and icelike correlations on the lattice, which is pre-cisely analogous to the behavior of the spin ice materials.However, it should be stressed that the fundamental interac-tion among the islands is the long-range dipole-dipole force,once the short-ranged exchange is negligible in this case,where the islands are spaced by around 320 nm, muchgreater than the permalloy exchange length, around 57 nm.Here, we consider an arrangement like that experimentallyinvestigated in Ref. 7. In our scheme the magnetic momentspin of the island is replaced by a point dipole at itscenter. To do this, in each site xi ,yi of a square lattice twospin variables are defined: Shi with components Sx= 1,

    Sy =0 located at rh= xi+1 /2,yi, and Svi with componentsSx=0, Sy = 1 at rv= xi ,yi+1 /2. Therefore, in a lattice ofvolume L2= l2a2 a is the lattice spacing one gets 2 l2

    spins see Fig. 2. Representing the spins of the islands by S i,which can assume either Shi or Svi, then the 2D spin ice isdescribed by the following equation:

    HSI = Da3

    ijS i S j


    3S i rijS j rijrij

    5 , 1where D=0

    2 /4a3 is the coupling constant of the dipolarinteraction. The sum is either over all l22l21 pairs of

    spins in the lattice for the case with open boundary condi-tions OBC or over all spins and their images for the casewith periodic boundary conditions PBC. We study thesetwo possibilities; OBC is more related to the artificial spinice fabricated in Ref. 7, while using PBC we minimize theborder effects. In the system with PBC the Ewaldsummation11,12 is used.


    To start, we consider the ground states obtained from Eq.1 describing the 2D spin ice. To do this we use a simulatedannealing process,13 which is a Monte Carlo calculationwhere the temperature is slightly reduced in each step of theprocess in order to drive the system to the global minimum.Our Monte Carlo scheme consist of a simple Metropolisalgorithm.13 In each Monte Carlo step MCS we attempt toflip all spins in the lattice sequentially or randomly whichgives the same results. Several tests for systems with differ-ent sizes L6aL80a were studied. In each simulation10 l2 MCSs were done at each temperature starting at T=3.0 and decreasing the temperature in steps T=0.2 untilT=0.2 the temperature is measured in units of D /kB. Weobserved that for T0.4 the system freezes, in the sense thatall trial moves are rejected. The final configuration groundstate was found to be twofold degenerate see Fig. 2a for alattice with L=6a. If we consider the vorticity in eachplaquette, assigning a variable =+1 and 1 to clockwiseand anticlockwise vorticities, respectively, the ground statelooks like a checkerboard, with an antiferromagnetic ar-rangement of the variable. Note that the ground stateclearly obeys the ice rule. We remark that it is impossible tominimize all dipole-dipole interactions. Actually, on eachvertex there are six pairs of dipoles and only four of themcan simultaneously minimize the energy. It is important tomention that although there are other possible configurationsthat also obey ice rules, these are not the ground state. In-deed, the state shown on the right side of Fig. 2 has energyabout four times larger than that of the ground state. Thedifference between these two states is related to the distincttopologies for the configurations of the four moments seeFig. 3. It was experimentally shown in Ref. 7 that while thetopologies of types a and b obey the ice rule, case a hassmaller energy than case b. Our theoretical calculationsconfirm this fact. The same ground state was also reported inRefs. 8 and 9. We would like to remark that although this isthe ground state, its thermal equilibration in experimentsseems to be very difficult.810

    FIG. 1. Color online The 2D square lattice studied in this work. Only afew islands are shown. The arrows inside the islands represent the localdipole moments Shi or Svi.

    FIG. 2. a Configuration of the ground state obtained for L=6a, in exactagreement with that experimentally observed. Note that the ice rules aremanifested at each vertex. This is the case in which the topology demandsthe minimum energy see Fig. 3. b Another configuration also respectingice rules, but displaying a topology which costs more energy.

    FIG. 3. The four distinct topologies and the 16 possible magnetic momentconfigurations on a vertex of four islands. Although configurations a andb obey the ice rule, the topology of a is more energetically favorable thanthat of b. Equation 1 correctly yields to the true ground state based ontopology a, without further assumptions. Topologies c and d do notobey the ice rule. Particularly, c implies in a monopole with charge QM.

    063913-2 Ml et al. J. Appl. Phys. 106, 063913 2009

    [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: Downloaded to ] IP: On: Sun, 21 Dec 2014 01:11:26

  • Once the system is naturally frustrated, in the two-in/two-out configuration, the effective magnetic charge QM


    number of spins pointing inward minus the number of spinspointing outward on each vertex i , j is zero everywhere.The most elementary excited state involves inverting a singlespin to generate localized dipole magnetic charges, whichimplies in a vortex-pair annihilation. Such an inversioncorresponds to two adjacent sites with net magnetic chargeQM

    i,j = 1, which is like a nearest-neighbor monopole-antimonopole pair. In principle, such monopoles can beseparated from one another without violations of local neu-trality by flipping a chain of adjacent spins. One can easilysee that in this process, a string of spins pointing from thepositive to the negative charge is created see Fig. 4. Thepresence of a stringlike excitation joining these poles is evi-denced by an extra energy cost behaving as bX, where X isthe length of the string and b0 is the effective string ten-sion, as below. In order to establish a link between themonopole-antimonopole distance R and the string length Xwe choose two basic string shapes to move the charges asshown in Fig. 4. Of course, the shortest strings will beformed around the straight line joining the monopoles and,therefore, we choose two different ways in which they maybe created as the charges are separated see Fig. 4. First,using the string shape 1 and starting in the ground state wechoose an arbitrary site and then the gray spins in Fig. 4 areflipped, thus creating a monopole-antimonopole separated byR=2a. In sequence, the spins marked in blue are flipped andthe separation distance becomes R=4a and so on. In thiscase X=4R /2. Note that the string surges in the system be-cause in the separation process, the topology is locally modi-fied, although still keeping the ice rule; in the region betweenthe two poles, the topology of type b, which has largerenergy than that of type a, prevails. Being essentially local-ized along the line joining the monopoles this additionalamount of energy increases as the distance between the mag-netic charges increases, justifying the bX term.

    The potential Vr the energy of the excited configura-tion minus the energy of the ground state as a function ofr=R /a can be obtained by simple evaluation of the energy ofeach configuration. It is shown in the inset of Fig. 5 for thestring shape presented in Fig. 4a. The behavior is appar-ently linear but the function fqR=q /R+bR+c, with q

    0.00122Da ,b=b /20.00305D /a ,c0.00734D, fits bet-ter the data than the purely linear possibility gR=R+,with 0.00611D /a ,0.00702D. This difference be-comes clearer when we analyze the 2 /Dof , which is equalto 1.04108 for the linear fitting and 4.51013 for fqR.Also, in Fig. 5 we draw a baseline of the potential using thelinear fit. One can clearly see that fqR describes the databetter and, therefore, Vr fqR. The same method wasrepeated using the string shape 2. In this case, the charges areseparated diagonally and X=2R /2. The results are qualita-tively the same and the values of the constants are: q0.00125Da ,b=b /20.00317D /a ,c0.00724D. Notethat the quantitative changes are small. The results are alsoqualitatively the same if PBC are used instead of OBC. Fur-thermore, quantitative differences between PBC and OBCcalculations are smaller than 1% for constants b and c, whileit is smaller than 9% for q. The larger difference for constantq can be understood if one remembers that the use of PBCwill imply that the charges interact also with their images.

    Our calculations yield the total energy cost of amonopole-antimonopole pair, separated by R, as the sum ofthe usual Coulombic-type term, q /R q0 is a constant,and an extra contribution behaving like bX, brought aboutfrom the stringlike excitations that bind the monopoles, sothat VR=q /R+bXR+c XR is the string length, while cis a constant associated to the monopole pair creation. Untilnow we have only considered the shortest strings connectingtwo poles. However, many dipole strings of arbitrary shapeand size can be identified that connect a given pair of mono-poles. The associated energy cost increases with X and di-verges with the length of the string. So, the monopolesshould be confined in the artificial material. As we will arguelater, it is possible that the string tension vanishes at a criticaltemperature proportional to b and hence, free magneticmonopoles may also be found in the 2D system.

    For concreteness, the magnetic charge may be easily es-timated if we take into account experimental values of someparameters. Considering the usual expression for the Cou-lombic interaction in mks units 0QM

    2 /4R, we get, q


    2 /4, or QM 4q /0 0.035 /a. Now, us-

    FIG. 4. Color online The two basic shortest strings used in the separationprocess of the magnetic charges: pictures 1 and 2 exhibit strings 1 and 2,respectively. The left circle is the positive charge north pole while the rightcircle is the negative south pole.

    0 5 10 15 20 25 30 35r








    0 10 20 30r





    FIG. 5. Color online Inset: the interaction potential between two magneticcharges with opposite signs as a function of r=R /a. The baseline of Vris also plotted: the curves are obtained by fitting the data to R+ andq /R+bX+c minus R+.

    063913-3 Ml et al. J. Appl. Phys. 106, 063913 2009

    [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: Downloaded to ] IP: On: Sun, 21 Dec 2014 01:11:26

  • ing data from Ref. 7 such as a320 nm and 2.791016 JT1, the fundamental magnetic charge of an exci-tation in the array of ferromagnetic nanoislands reads QM31011 cm /s, which is about 6103 times smaller thanthe fundamental charge of the Dirac monopole QD=2 /0e. Such a charge can even be tuned continuouslyby changing the lattice spacing.


    Before concluding, it is important to analyze the behav-ior of the string tension as some parameters are varied in thesystem. The string tension for the artificial system built inRef. 7 is approximately given by b2.261015 J /s4.5103 eV /a. Therefore, it is necessary a relatively largeamount of energy about 103 eV to separate the 2Dmonopoles by one lattice spacing, regardless of how farapart they are. Consequently, at low temperature, there isinsufficient thermal energy to create long strings, and so themonopoles would be bound together tightly in pairs. Thestring tension can be artificially reduced by increasing theparameter ab1 /a. However, it has also the effect of de-creasing the magnetic charge since QM is proportional to1 /a. A way to reduce b without affecting QM is increasingthe temperature. By using the random walk argument, onecan see that the many possible ways of connecting a pair ofmonopoles with a string give rise to a string configurationalentropy proportional to R. Then, as the temperature in-creases, the string tension should decrease like bkBT, with=Oa1. It means that the string may lose its tension byentropic effect and, therefore, it should vanish at some criti-cal temperature kBTc, of the order of ba4.5103 eV. An-other important point in this discussion is that as the tem-perature increases the monopoles density also increases.Indeed, if the pair creation energy is of the order of Ec=Va1.3102 eV, one expects that, for temperaturesabove this value, the description in terms of monopoles itselfcould break down in fact, the Boltzmann factor expEcwould increase considerably for kBTEc. Then, a possibledeconfined phase would live between the melting tempera-ture of the ordered and the dense monopole phases. A com-parison between ba and Ec suggests that a temperature win-dow between the confined and deconfined phases could beperfectly plausible in the range 4.5103 eVkBT1.3102 eV. Our expectation is that the window is stillgreater starting at a much lower temperature, since the ar-gument based on the balance of energy versus entropy mayoverestimate the critical temperature for instance, theBerezinskiiKosterlitzThouless critical temperature esti-mated by this argument for the planar rotator model is muchhigher than the correct value obtained by Monte Carlo simu-lations, which is TBKT0.89J,

    14 where J is the coupling con-stant of the model. Once the deconfinement realizes, thequestion of technological applications of this system is rel-evant. For instance, learning how to move the magneticmonopoles around would be of importance toward technolo-gies involving magnetic analogous of electric circuits.

    However, there is another entropic effect, discussed inprevious works of purely ice rule problem and related short-

    range problems1518 for strictly 2D systems, which maychange the scenario of free monopoles. That is, the entropicinteractions between monopoles due to the underlying spinconfiguration. Really, two monopoles should be attracted be-cause there are more ways to arrange the surrounding dipolesin the lattice when they are close together. These entropicinteractions, in a strictly 2D system, results in a 2D effectiveCoulomb attraction like ln R, between oppositely chargedmonopoles, whose strength vanishes proportionally to T, atlow temperatures of course, in 3D materials, such entropiceffect should result in a 1 /R attraction. In our case, thislogarithmic interaction could be present in addition to the 3DCoulombic, q /R, and linear bR interactions discussed in thiswork. Thus, at a temperature high enough to destroy thestring tension, this entropically driven 2D Coulomb forcewould become crucial for keeping the monopole-antimonopole pairs bounded, in such a way that no freemonopole phase would occur at any temperature. Neverthe-less, how these monopoles precisely experience such an ef-fect in their local dynamics should be investigated in moredetails. We should recall that our calculations to obtain theinteraction potential between monopoles have been per-formed at zero temperature and, consequently, this entropiccontribution could not be directly or even indirectly presentin VR. The precise effect of the temperature on VR isunder investigation and will be communicated elsewhere.Here, it should be remarked that the present monopoles arenot actually 2D objects: their physical interaction is given bythe usual 3D Coulomb force, which means that they shouldaffect magnetic test particles placed at relatively large dis-tances along the direction perpendicular to the plane of is-lands we remember that in a strictly 2D space, the magneticfield should be a pseudoscalar field. In addition, a genuine2D monopole, as a counterpart of the Dirac pole, appears tobe not magnetic charge; it rather looks like an exotic electriccharge, giving rise to a rotational electric field, instead ofradial-like, as usual charges do. For details, see Refs. 1921.The dipoles forming the 2D lattice are genuinely 3D objectsand their long-range dipolar interaction propagates in the 3Dspace see Eq. 1. It is an important difference of this sys-tem when compared to the strictly 2D models such as vertexmodels and others. Now, it is also important to say that, suchentropic interaction will not be accompanied by a magneticfield, it will not renormalize the monopole charge and it willnot be felt by a stationary magnetic test particle.5 Therefore,all calculations concerning the energy scales involved in thephysical interactions between the defects will not be altered.In addition, it seems that this force has not been measureddirectly, yet. The peculiarities between strictly 2D modelsand the system studied here have not been considered previ-ously. Thus, it is not completely clear if and how the entropi-cally driven 2D Coulomb force acts in the spin ice with aninherent 3D spatial behavior.

    Finally, we should remark that the above scenario in-volving the monopole physical interactions may be drasti-cally changed if one considers these excitations in Fig. 2b.As experimentally shown in Ref. 8, this metastable state is avery real possibility when magnetic fields are applied. In this

    063913-4 Ml et al. J. Appl. Phys. 106, 063913 2009

    [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: Downloaded to ] IP: On: Sun, 21 Dec 2014 01:11:26

  • case only the topology of Fig. 3b is present in the separa-tion process. Further investigation is demanded for sheddingextra light on this subject.


    The authors thank CNPq, FAPEMIG, and CAPES Bra-zilian agencies for financial support.

    1J. Goldstone and F. Wilczek, Phys. Rev. Lett. 47, 986 1981.2Z. Nussinov, C. D. Batista, B. Normand, and S. A. Trugman, Phys. Rev. B75, 094411 2007.

    3G. Baskaran, Phys. Rev. B 68, 212409 2003.4A. R. Moura, A. R. Pereira, and A. S. T. Pires, Phys. Rev. B 75, 0144312007.

    5C. Castelnovo, R. Moessner, and S. L. Sondhi, Nature London 451, 422008.

    6L. D. C. Jaubert and P. C. W. Holdsworth, Nat. Phys. 5, 258 2009.7R. F. Wang, C. Nisoli, R. S. Freitas, J. Li, W. McConville, B. J. Cooley, M.

    S. Lund, N. Samarth, C. Leighton, V. H. Crespi, and P. Schiffer, NatureLondon 439, 303 2006.

    8A. Remhof, A. Shumann, A. Westphalen, H. Zabel, N. Mikuszeit, E. Y.Vedmedenko, T. Last, and T. Kunze, Phys. Rev. B 77, 134409 2008.

    9X. Ke, J. Le, C. Nisoli, P. E. Lammert, W. McConville, R. F. Wang, V. H.Crespi, and P. Schiffer, Phys. Rev. Lett. 101, 037205 2008.

    10G. Mller and R. Moessner, Phys. Rev. Lett. 96, 237202 2006.11J.-J. Weis, J. Phys.: Condens. Matter 15, S1471 2003.12Z. Wang and C. Holm, J. Chem. Phys. 115, 6351 2001.13D. Landau and K. Binder, A Guide to Monte Carlo Simulations in Statis-

    tical Physics Cambridge University Press, New York, 2005.14J. Tobochnik and G. V. Chester, Phys. Rev. B 20, 3761 1979.15J. L. Jacobsen and J. Kondev, Nucl. Phys. B 532, 635 1998.16M. F. Fisher and J. Stepheson, Phys. Rev. 132, 1411 1963.17C. Moore, M. G. Nordahl, N. Minar, and R. Shalizi, Phys. Rev. E 60, 5344

    1999.18C. L. Henley, Phys. Rev. B 71, 014424 2005.19M. Henneaux and C. Teitelboim, Phys. Rev. Lett. 56, 689 1986.20R. Pisarski, Phys. Rev. D 34, 3851 1986.21W. A. Moura-Melo and J. A. Helayl-Neto, Phys. Rev. D 63, 065013


    063913-5 Ml et al. J. Appl. Phys. 106, 063913 2009

    [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: Downloaded to ] IP: On: Sun, 21 Dec 2014 01:11:26


View more >