Two-orbital SU(N) magnetism with ultracold alkaline-earth atoms
A. V. Gorshkov, M. Hermele, V. Gurarie, C. Xu, P. S. Julienne, J. Ye, P. Zoller, E. Demler, M. D. Lukin, A. M. Rey
aa r X i v : . [ c ond - m a t . qu a n t - g a s ] S e p Two-orbital SU(N) magnetism with ultracold alkaline-earth atoms
A. V. Gorshkov ∗ , M. Hermele , V. Gurarie , C. Xu , P. S. Julienne ,J. Ye , P. Zoller , E. Demler , , M. D. Lukin , , and A. M. Rey Physics Department, Harvard University, Cambridge, MA 02138 Department of Physics, University of Colorado, Boulder, CO 80309 Joint Quantum Institute, NIST and University of Maryland, Gaithersburg, MD 20899-8423 JILA, NIST, and Department of Physics, University of Colorado, Boulder, CO 80309 Institute for Theoretical Physics, University of Innsbruck, A-6020 Innsbruck,Austria and Institute for Quantum Optics and Quantum Information of the Austrian Academy of Sciences, A-6020 Innsbruck, Austria Institute for Theoretical Atomic, Molecular and Optical Physics,Harvard-Smithsonian Center of Astrophysics, Cambridge, MA 02138 and ∗ e-mail: [email protected] (Dated: October 29, 2018) Fermionic alkaline-earth atoms have unique properties that make them attractive candidates for therealization of novel atomic clocks and degenerate quantum gases. At the same time, they are attractingconsiderable theoretical attention in the context of quantum information processing. Here we demon-strate that when such atoms are loaded in optical lattices, they can be used as quantum simulators ofunique many-body phenomena. In particular, we show that the decoupling of the nuclear spin from theelectronic angular momentum can be used to implement many-body systems with an unprecedented de-gree of symmetry, characterized by the SU(N) group with N as large as 10. Moreover, the interplay ofthe nuclear spin with the electronic degree of freedom provided by a stable optically excited state allowsfor the study of spin-orbital physics. Such systems may provide valuable insights into strongly correlatedphysics of transition metal oxides, heavy fermion materials, and spin liquid phases.
The interest in fermionic alkaline-earth atoms [1, 2, 3, 4, 5,6, 7, 8] stems from their two key features: (1) the presence ofa metastable excited state P coupled to the ground S statevia an ultranarrow doubly-forbidden transition [1] and (2) thealmost perfect decoupling [1] of the nuclear spin I from theelectronic angular momentum J in these two states, since theyboth have J = 0 . This decoupling implies that s-wave scat-tering lengths involving states S and P are independentof the nuclear spin, aside from the restrictions imposed byfermionic antisymmetry. We show that the resulting SU(N)spin symmetry (where N = 2 I + 1 can be as large as 10) to-gether with the possibility of combining (nuclear) spin physicswith (electronic) orbital physics open up a wide field of ex-tremely rich many-body systems with alkaline-earth atoms.In what follows, we derive the two-orbital SU(N)-symmetric Hubbard model describing alkaline-earth atomsin S and P states trapped in an optical lattice. We fo-cus on specific parameter regimes characterized by full orpartial atom localization due to strong atomic interactions,where simpler effective spin Hamiltonians can be derived.The interplay between orbital and spin degrees of freedomin such effective models is a central topic in quantum mag-netism and has attracted tremendous interest in the condensedmatter community. Alkaline earth atoms thus provide, onthe one hand, a unique opportunity for the implementationof some of these models for the first time in a defect-freeand fully controllable environment. On the other hand, theyopen a new arena to study a wide range of models, many ofwhich have not been discussed previously, even theoretically.We demonstrate, in particular, how to implement the Kugel-Khomskii model studied in the context of transition metal ox-ides [9, 10, 11, 12, 13], the Kondo lattice model [14, 15, 16,17, 18, 19, 20, 21, 22, 23, 24, 25, 26] studied in context of manganese oxide perovskites [20] and heavy fermion materi-als [25], as well as various SU(N)-symmetric spin Hamilto-nians that are believed to have spin liquid and valence-bond-solid ground states [27, 28, 29, 30, 31, 32, 33, 34]. For ex-ample, we discuss how, by appropriately choosing the initialstate, a single alkaline-earth atom species with I = 9 / (suchas Sr) can be used to study experimentally such a distinc-tively theoretical object as the phase diagram as a function of N for all N ≤ .Before proceeding, we note that, while an orthogonal sym-metry group SO(5) can be realized in alkali atoms [35],proposals to obtain SU(N > >
2) symmetry.
FIG. 1:
Interaction parameters between g (green) and e (yel-low) atoms loaded in the lowest vibrational state of the corre-sponding optical lattice. Here we assumed I = 1 / , and the ar-rows indicate the m I = ± / spin states. | s, t i denote the singletand triplet nuclear spin states of the two atoms (only one of threetriplet states - | ↑↑i - is shown). The dashed circle represents anti-symmetrization of the nuclear spin state (i.e. | s i ). The interaction en-ergy U X ( X = gg, ee, eg + , eg − ) is proportional to the correspond-ing scattering length a X . Many-body dynamics of alkaline-earth atoms in an opticallattice
We begin with the Hamiltonian describing cold fermionicalkaline-earth atoms in an external trapping potential: H = X αm Z d r Ψ † αm ( r )( − ~ M ∇ + V α ( r ))Ψ αm ( r ) (1) + ~ ω Z d r ( ρ e ( r ) − ρ g ( r )) + g + eg + g − eg Z d r ρ e ( r ) ρ g ( r )+ X α,m To understand the properties of the Hamiltonian in Eq. (2),we consider its symmetries. We define SU(2) pseudo-spin al-gebra via T µ = X j T µj = 12 X jmαβ c † jαm σ µαβ c jβm , (3)where σ µ ( µ = x, y, z ) are Pauli matrices in the { e, g } basis.We further define nuclear-spin permutation operators S mn = X j S mn ( j ) = X j,α S mn ( j, α ) = X j,α c † jαn c jαm , (4)which satisfy the SU(N) algebra [ S mn , S pq ] = δ mq S pn − δ pn S mq ,and thus generate SU(N) rotations of nuclear spins ( N = 2 I +1 ).In addition to the obvious conservation of the total numberof atoms n = P j ( n je + n jg ) , H exhibits U (1) × SU ( N ) sym-metry (see Methods for the discussion of enhanced symme-tries), where U (1) is associated with the elasticity of collisionsas far as the electronic state is concerned ( [ T z , H ] = 0 ) andSU(N) is associated with the independence of scattering andof the trapping potential from the nuclear spin ( [ S mn , H ] = 0 for all n , m ). The two-orbital SU(N)-symmetric HubbardHamiltonian in Eq. (2) is a generalization to N > of itsSU(2)-symmetric counterpart [9] and to two orbitals of itssingle-orbital counterpart [28]. The SU(N) symmetry and thelargely independent spin and orbital degrees of freedom aretwo unique features present in alkaline-earths but absent inalkalis due to strong hyperfine interactions.One important consequence of SU(N) symmetry is the con-servation, for any m , of S mm , the total number of atoms withnuclear spin m . This means that an atom with large I , e.g. Sr ( I = 9 / ), can reproduce the dynamics of atoms withlower I if one takes an initial state with S mm = 0 for some m values. To verify SU(N) symmetry of the interaction ex-perimentally, one could, thus, put two atoms in one well inspins m and m ′ and confirm that collisions do not populateother spin levels. This feature of SU(N) symmetry is in starkcontrast to the case of weaker SU(2) symmetry, where the de-pendence of scattering lengths on the total spin of the two col-liding particles allows for scattering into spin states other than m and m ′ . We note that although collisions are governed byelectronic interactions and obey the nuclear-spin SU(N) sym-metry, the nuclear spins still indirectly control the collisionsvia fermionic statistics and give rise to effective spin-orbitaland spin-spin interactions.One can alternatively implement the two-orbital Hubbardmodel with two ground-state species of alkaline-earth atoms(e.g. Yb and Yb, or Yb and Sr). If we still re-fer to them as | g i and | e i , the nuclear distinguishability andthe fact that both atoms are in the ground state will resultin a + eg = a − eg , corresponding to an enhanced symmetry (seeMethods). While experimentally more challenging, the useof two different ground state species will solve the prob-lem of losses associated with collisions of two excited state FIG. 2: Young diagrams describing the irreducible representa-tions of SU(N) on individual sites. a, A general diagram consistsof n j boxes arranged into at most two columns (to satisfy fermionicantisymmetry with only two orbital states) whose heights we willdenote by p and q , such that N ≥ p ≥ q and p + q = n j . SeeSupplementary Information for a brief review of Young diagrams. b, The Young diagrams for the two special cases discussed in the maintext: (1) ( p, q ) = (1 , and (2) ( p, q ) = ( p, on a bipartite lattice. atoms and will reduce the (already very weak) nuclear-spin-dependence of a ee and a eg . Spin Hamiltonians One of the simplest interesting limits of Eq. (2) is thestrongly interacting regime ( J/U ≪ ) where the Hilbertspace is restricted to a given energy manifold of the J g = J e = 0 Hamiltonian (with a fixed number of atoms on eachsite), and tunneling is allowed only virtually, giving rise toan effective spin (and pseudo-spin) Hamiltonian. Single-siteenergy manifolds can be classified according to the numberof atoms n j = n jg + n je , the pseudo-spin component T zj ,and the spin symmetry (SU(N) representation) described bya Young diagram. As shown in Fig. 2a, each diagram con-sists of n j boxes and at most two columns of heights p and q ,representing two sets of antisymmetrized indices.The U (1) × SU ( N ) symmetry of Eq. (2) restricts the order J spin Hamiltonian to the form H ( p,q ) = X h i,j i ,α h κ ijα n iα n jα + λ ijα S nm ( i, α ) S mn ( j, α ) i + X h i,j i h κ ijge n ig n je + λ ijge S nm ( i, g ) S mn ( j, e )+˜ κ ijge S emgm ( i ) S gnen ( j ) + ˜ λ ijge S engm ( i ) S gmen ( j ) + { i ↔ j } i , (5)where the sum over n and m is implied in all but the κ termsand S αmβn ( j ) = c † jβn c jαm . { i ↔ j } means that all 4 precedingterms are repeated with i and j exchanged. The coefficients κ , λ , ˜ κ , and ˜ λ are of order J /U with the exact form determinedby what single-site energy manifolds we are considering. κ terms describe nearest neighbor repulsion or attraction, while λ , ˜ κ , and ˜ λ terms describe nearest neighbor exchange of spins,pseudo-spins, and complete atomic states, respectively. With-out loss of generality, κ ijα = κ jiα and λ ijα = λ jiα . In manycases (e.g. case (2) below), the Hilbert space, which H ( p,q ) a (0,0)0 b FIG. 3: The ground-state phase diagram for the SU(N=2) Kugel-Khomskii model restricted to two wells, left (L) and right (R). a, The phase diagram for T z = − (two g atoms). | gg i = | gg i LR . | s i and | t i are spin singlet and triplet states, respectively. b, Thephase diagram for T z = 0 (one g atom and one e atom). | Σ i = √ ( | eg i LR − | ge i LR ) and | τ i = √ ( | eg i LR + | ge i LR ) are anti-symmetric and symmetric orbital states, respectively. See Supple-mentary Information for a detailed discussion of both of these dia-grams. acts on, has n ie and n ig constant for all i , which not onlyforces ˜ κ ijge = ˜ λ ijge = 0 but also allows one to ignore the con-stant κ ijα and κ ijge terms. We now discuss two special cases of H ( p,q ) shown in Fig. 2b. A third case, ( p, q ) = (1 , , whichreduces for N = 2 to the spin-1 Heisenberg antiferromagnetis discussed in the Supplementary Information. (1) In the case of one atom per site, ( p, q ) = (1 , . H ( p,q ) isthen a generalization to arbitrary N of the SU( N = 2 ) Kugel-Khomskii model [9, 13], and we rewrite it as (see Supplemen-tary Information) H (1 , = X h i,j i h κ ge + ˜ λ ge S ij )( T xi T xj + T yi T yj ) + λ ge S ij − [ A + BS ij ]( T zi T zj + 14 ) + h (1 − S ij )( T zi + T zj ) i , (6)where S ij = P mn S nm ( i ) S mn ( j ) is +1 ( − ) for a symmetric(antisymmetric) spin state, A = 2 κ ge − κ e − κ g , B = 2 λ ge + κ e + κ g , and h = ( κ e − κ g ) / . The N = 2 Kugel-KhomskiiHamiltonian is used to model the spin-orbital interactions (notto be confused with relativistic spin-orbit coupling) in transi-tion metal oxides with perovskite structure [13]. Our imple-mentation allows to realize clean spin-orbital interactions un-altered by lattice and Jahn-Teller distortions present in solids[13].To get a sense of the competing spin and orbital orders[10, 11, 12] characterizing H (1 , , we consider the simplestcase of only two sites ( L and R ) and N = 2 (with spin statesdenoted by ↑ and ↓ ). To avoid losses in e - e collisions, we set U ee = ∞ (see Supplementary Information). The double-wellground-state phase diagram for T z = 1 (two e atoms) is thentrivial, while the T z = − (two g atoms) and T z = 0 (one g atom and one e atom) diagrams are shown in Fig. 3. One cansee that, depending on the signs and relative magnitudes of theinteractions, various combinations of ferromagnetic (triplet)and antiferromagnetic (singlet) spin and orbital orders are fa- ba FIG. 4: Probing the phases of the SU(N) antiferromagnet on a 2Dsquare lattice. a shows the phase diagram for the case n A + n B = N . Some points on this diagram have been explored in earlier nu-merical studies [29, 30, 31] and are marked according to the groundstate obtained: Neel (circles), columnar-valence-bond solid (VBS)[shown schematically in b ] (squares), and possibly critical spin liq-uid (triangle) [30, 31]. Since for sufficiently large N quantum fluc-tuations tend to destabilize long-range magnetic ordering, it is likelythat VBS ordering characterizes the ground state for all N > (i.e.above the wavy line). vored. In the Methods, we propose a double-well experimentalong the lines of Ref. [39] to probe the spin-orbital interac-tions giving rise to the T z = 0 diagram in Fig. 3b. Multi-wellextensions of this experiment may shed light on the model’smany-body phase diagram, which has been studied for N = 2 and mostly at mean-field level or in special cases, such as inone dimension or in the presence of enhanced symmetries (seee.g. [10, 11, 12]). (2) In order to study SU(N) spin physics alone, we con-sider the case of g atoms only. On a bipartite lattice withsublattices A and B, we choose A sites to have n A < N atoms [ ( p, q ) = ( n A , ] and B sites to have n B < N atoms[ ( p, q ) = ( n B , ]. This setup can be engineered in cold atomsby using a superlattice to adjust the depths of the two sublat-tices favoring a higher filling factor in deeper wells. H ( p,q ) then reduces to H ( p, = 2 J g U gg U gg − ( U gg ( n A − n B ) + ∆) X h i,j i S ij , (7)where ∆ is the energy offset between adjacent lattice sites.The coupling constant can be made either positive (antifer-romagnetic) or negative (ferromagnetic) depending on thechoice of parameters [39]. Three body recombination pro-cesses will likely limit the lifetime of the atoms when n j ≥ (see Supplementary Information).We focus on the 2D square lattice in the antiferromag-netic regime. The case n A + n B = N shares with theSU(2) Heisenberg model the crucial property that two ad-jacent spins can form an SU(N) singlet, and has thus beenstudied extensively as a large-N generalization of SU(2) mag-netism [27, 28]. Fig. 4a shows the expected phase diagramfor the case n A + n B = N , which features Neel (circles),valence-bond-solid (VBS) (squares) [Fig. 4b], and possiblecritical spin liquid (triangle) [30, 31] ground states. To ac-cess various ground states of the system, the initial state mustbe carefully prepared so that the conserved quantities S mm takevalues appropriate for these ground states. Another interestingand experimentally relevant case, n A = n B = N/ , whichcan also exhibit spin liquid and VBS-type ground states, isdiscussed in the Supplementary Information and in Ref. [34].Since one can vary N just by choosing the number of ini-tially populated Zeeman levels ( e.g. via a combination of opti-cal pumping and coherent manipulation), alkaline-earth atomsoffer a unique arena to probe the phase diagram of H ( p, , in-cluding exotic phases such as VBS [Fig. 4b], as well as com-peting magnetically ordered states. We propose to load a bandinsulator of N g atoms per site, then slowly split each well intotwo to form an array of independent SU(N) singlets in a pat-tern shown in Fig. 4b. The intersinglet tunneling rate shouldthen be adiabatically increased up to the intrasinglet tunnelingrate. As N increases, the magnetic or singlet nature of thestate can be probed by measuring the Neel order parameter(see the description of the Kugel-Khomskii double-well ex-periment in the Methods) and spin-spin correlations via noisespectroscopy in the time of flight [40] (which directly mea-sures P i,j h S mn ( i, g ) S nm ( j, g ) i e IQ ( i − j ) ). The Kondo lattice model (KLM) The SU(N) Kondo lattice model (KLM) [15, 17] is anotherexample of the rich physics, beyond the Mott regime, whichcould be simulated with alkaline-earth atoms. The KLM isone of the canonical models used to study strongly corre-lated electron systems, such as manganese oxide perovskites[20] and rare earth and actinide compounds classed as heavyfermion materials [25].For its implementation with cold atoms (for N = 2 , seealso Refs. [23, 24]), we propose to put one e atom (localizedspin) per site in a deep lattice such that J e ≪ U ee , so thatwe can set J e = 0 and n je = 1 for all j in Eq. (2). We alsosuppose that we can set U gg = 0 , e.g. by taking a very shallow g -lattice (see Fig. 5a). The resulting Hamiltonian is the SU(N)KLM [15, 17] H KLM = − X h j,i i m J g ( c † igm c jgm + h.c. )+ V ex X j,m,m ′ c † jgm c † jem ′ c jgm ′ c jem . (8)The magnitude of V ex can be adjusted by shifting the e and g lattices relative to each other [7].The properties of the SU(N) KLM depend crucially on thesign of the exchange interaction. For concreteness, we focuson the antiferromagnetic (AF) case ( V ex < ), which favorsformation of spin-antisymmetric states (singlets, for N = 2 )between mobile fermions and localized spins. This regimedescribes the physics of heavy fermion materials [25], and,in the case of a single localized spin, gives rise to the Kondoeffect.In the limit | V ex | ≪ J g , g atoms mediate long-range RKKY ab FIG. 5: Kondo lattice model for the case N = 2 . a, The schematicof the setup. g atoms are green; e atoms are yellow; the spin ba-sis is {↑ , ↓} . b, Schematic representation of the competition be-tween RKKY magnetism vs Kondo singlet formation in the SU(2)AF KLM (see [16, 25, 26] and references therein). In this model,the localized spin- / e atoms couple antiferromagnetically to thedelocalized g atoms, via an on-site exchange interaction propor-tional to V ex . This coupling favors the formation of localized Kondosinglets between e and g atoms, with characteristic energy scale k B T K ∼ J g exp( − cJ g / | V ex | ) , with c a dimensionless constant oforder one [25]. On the other hand, the g atoms can mediate long-range RKKY interactions between the e atoms, giving rise to mag-netic order (which can be antiferromagnetic (AF) or ferromagneticdepending on the density of g atoms), where the characteristic en-ergy is k B T RKKY ∼ V ex /J g . The competition between Kondo ef-fect and RKKY magnetism leads to very rich physics. For small val-ues of | V ex | /J g , the RKKY interaction is dominant and the systemorders magnetically. At intermediate values of | V ex | /J g , the energyscales T K and T RKKY are of comparable strength, and a variety ofnovel quantum phenomena are expected to arise, including quantumcriticality and non-Fermi liquid (NFL) physics [25, 26]. With furtherincrease of the | V ex | /J g coupling, magnetic order is suppressed, thelocalized e atoms become screened into singlet states and melt intothe g -atom Fermi sea, forming the so called heavy Fermi liquid state(HFL). The large Fermi volume [21], which is set by the total num-ber of g atoms plus e atoms, can be directly probed by measuring themomentum distribution via time of flight imaging. interactions [14] between localized spins and tend to inducemagnetic ordering (antiferromagnetic or ferromagnetic de-pending on the density of g atoms) of the latter, at least for N = 2 . The engineering of RKKY interactions can be testedin an array of isolated double wells (see Methods). At in-termediate and large | V ex | , the formation of Kondo singletsdominates the RKKY interaction and favors a magneticallydisordered heavy Fermi liquid (HFL) ground state with a sig-nificantly enhanced effective quasiparticle mass (see Fig. 5b).The competition between RKKY interactions and the Kondoeffect in the regime where both are comparable is subtle, andthe resulting phases and phase transitions [25, 26] are notwell-understood. Ultracold alkaline-earth atoms provide apromising platform to study these phases and phase transi-tions.In the large- N limit [15, 17], the SU(N) HFL can be con-trollably studied, and /N expansions have successfully re-produced the experimentally observed properties of the HFL.However, very little is known about the SU(N) model outsidethe HFL regime. Several very interesting parameter regimesin this domain can be directly probed with our system, as dis-cussed in the Methods. Experimental Accessibility The phenomena described in this manuscript can be probedwith experimental systems under development. Indeed, weshow in the Methods that SU(N)-breaking terms are suffi-ciently weak, and here we discuss the temperature require-ments.The key energy scale in the spin Hamiltonians [Eq. (5)]is the superexchange energy J /U , while the RKKY en-ergy scale is k B T RKKY ∼ V ex /J g . In their region of va-lidity ( J < U and | V ex | < J g , respectively), these en-ergy scales are limited from above by the interaction energy( U and | V ex | , respectively), which typically corresponds totemperatures T . nK [39]. Thanks to the additionalcooling associated with certain adiabatic changes [41, 42], T ∼ nK and the Mott insulating regime have already beenachieved with fermionic alkali atoms [43], and are thereforeexpected to be achievable with fermionic alkaline-earths, aswell (a bosonic alkaline-earth Mott insulator has already beenachieved [44]). Furthermore, the requirement to reach k B T smaller than J /U or V ex /J g can often be relaxed. First,the double-well experiments, such as the ones discussed inthe Methods in the contexts of the Kugel-Khomskii and theKondo lattice models, are performed out of thermal equilib-rium, and can, thus, access energy scales far below the tem-perature of the original cloud [39]. Second, for SU(N) anti-ferromagnets, the energy range between J /U and N J /U may also exhibit intriguing physics: in this regime, SU(N)singlets, which require N J /U energy to break, stay intactbut can diffuse around. Finally, in the V ex < Kondo latticemodel, exotic heavy Fermi liquid behavior is expected when J g . | V ex | and the temperature is below the Kondo tempera-ture, i.e. k B T . J g exp( − cJ g / | V ex | ) with c is a dimension-less constant of order one [25]. Thus, with J g chosen to be onthe order of | V ex | , k B T as high as ∼ | V ex | may be sufficient. Outlook The proposed experiments should be regarded as bridgesaiming to connect well-understood physics to the complex andpoorly understood behavior of strongly correlated systems. It is important to emphasize that, except for the one dimensionalcase, the phase diagram of most of the models considered isonly known at mean field level or numerically in reduced sys-tem sizes. Therefore, their experimental realization in cleanand controllable ultracold atomic systems can provide majoradvances.Our proposal motivates other new lines of research. Ultra-cold bosonic or fermionic diatomic molecules [45] may giverise to similar SU(N) models with large N and with the possi-bility of long-range interactions. Ions with alkaline-earth-likestructure, such as Al + could also be considered in this context.It would also be interesting to explore the possibility of real-izing topological phases with SU(N) models for applicationsin topological quantum computation [34]. Beyond quantummagnetism, the fact that the formation of SU(N) singlets re-quires N partners might give rise to novel exotic types of su-perfluidity and novel types of BCS-BEC crossover [37]. Prac-tical applications of our Hubbard model, such as the calcula-tion of the collisional frequency shift in atomic clocks [46],can also be foreseen. Note added in proof. After the submission of this paper, atheoretical study of the SU(6)-symmetric Yb system wasreported [50]. MethodsExperimental tools available for alkaline-earth atoms Many experimental tools, such as tuning the interactionstrength by adjusting laser intensities [39], are common toboth alkali and alkaline-earth atoms. There are, however,some experimental tools specific to alkaline earths; we reviewthem in this Section.First, a combination of optical pumping [2] and direct co-herent manipulation of the | g i − | e i transition in the presenceof a magnetic field [1, 2] can be used [8] to prepare any desiredsingle-atom state within the 2 (2 I + 1)-dimensional manifoldwith basis | αm i , where α = g or e and m = − I, . . . , I . Thiscoherent manipulation can also be used to exchange quantuminformation between nuclear spin states and electronic states.Second, by using far-detuned probe light or a large magneticfield to decouple the electronic angular momentum J and thenuclear spin I , the electronic | g i − | e i degree of freedom canbe measured by collecting fluorescence without destroying thenuclear spin state [8]. Fluorescence measurement of the nu-clear spins can be achieved by mapping nuclear spin statesonto electronic states [7, 8]: for example, for a spin- / nu-cleus, a π pulse between | g, m = 1 / i and | e, m = − / i allows one to accomplish a swap gate between the nuclear { / , − / } qubit and the electronic { e, g } qubit. Single-sitespatial resolution during the coherent manipulation and fluo-rescence measurement can be achieved using magnetic fieldgradients [7] or dark-state-based techniques [8, 47] that relyon an auxiliary laser field whose intensity vanishes at certainlocations. Third, an appropriate choice of laser frequenciesallows one to obtain independent lattices for states g and e [7]. Finally, optical Feshbach resonances [48] may be used tocontrol scattering lengths site-specifically and nearly instanta-neously. Enhanced Symmetries While in the general case, our Hubbard model [Eq. (2)]satisfies U (1) × SU ( N ) symmetry, for particular choices ofparameters, higher symmetry is possible. In particular, if J g = J e and the interaction energies for all states within thepseudo-spin triplet are equal ( U gg = U ee = U + eg ), the fullSU(2) symmetry (not just U(1)) in the pseudo-spin space issatisfied. Alternatively, if V ex = 0 , then both S mn ( i, g ) and S mn ( i, e ) generate SU(N) symmetries resulting in the overall U (1) × SU ( N ) × SU ( N ) symmetry. Finally, if both condi-tions are satisfied, i.e. all four U X are equal and J g = J e , then H satisfies the full SU(2N) symmetry ( N can be as high as20) generated by S αmβn = X j S αmβn ( j ) = X j c † jβn c jαm , (9)in which case the interaction reduces to U P j n j ( n j − ,where n j = n jg + n je .In the case when | e i and | g i correspond to two ground statesof two different atoms (with nuclear spin I e and I g , respec-tively), we will have a + eg = a − eg (i.e V ex = 0 ), which is equiva-lent to imposing U (1) × SU ( N g = 2 I g +1) × SU ( N e = 2 I e +1) symmetry, where SU (2 I α + 1) is generated by S mn ( i, α ) .While for I g = I e , the m index in c jαm will run over a dif-ferent set of values depending on α , the Hubbard Hamiltonianwill still have the form of Eq. (2) (except with V ex = 0 ). Ifone further assumes that J g = J e and U gg = U ee = U eg ,the interaction satisfies the full SU ( N g + N e ) symmetry. Itis worth noting that for the case of two different ground stateatoms, this higher symmetry is easier to achieve than for thecase of two internal states of the same atom, since a + eg = a − eg automatically. Thus, in particular, it might be possible to ob-tain SU (18) with Sr ( I = 9 / ) and Ca ( I = 7 / ) simplyby adjusting the intensities of the two lattices (to set J g = J e and U gg = U ee ) and then shifting the two lattices relative toeach other (to set U eg = U gg ).Enhanced symmetries of the Hubbard model [Eq. (2)] areinherited by the spin Hamiltonian [Eq. (5)]. In particular, im-posing SU (2) × SU ( N ) instead of U (1) × SU ( N ) forces κ ijge = κ jige , ˜ κ ijge = ˜ κ jige , κ ijg = κ ije = κ ijge + ˜ κ ijge ≡ κ ij , λ ijge = λ jige , ˜ λ ijge = ˜ λ jige , λ ijg = λ ije = λ ijge + ˜ λ ijge ≡ λ ij .Alternatively, imposing U (1) × SU ( N ) × SU ( N ) forces ˜ κ ijge = λ ijge = 0 . Finally, imposing the full SU(2N) forcesthe satisfaction of both sets of conditions, yielding H = X h i,j i h κ ij n i n j + λ ij S βnαm ( i ) S αmβn ( j ) i , (10)which is, of course, equivalent to restricting Eq. (5) to g -atoms only and extending labels m and n to run over N states in-stead of N . Double-well Kugel-Khomskii and RKKY experiments In the main text and in the following Methods Section, wediscuss the open questions and previously unexplored regimesassociated with the SU(N) Kugel-Khomskii and Kondo latticemodels (KLM) that can be studied with ultracold alkaline-earth atoms. As a stepping stone toward these many-body ex-periments, we propose in this Section two proof-of-principleexperiments in an array of isolated double wells with N = 2 (with the spin basis {↑ , ↓} ): one to probe the spin-orbital in-teractions of the Kugel-Khomskii model and one to probe theRKKY interactions associated with KLM.We first propose an experiment along the lines of Ref. [39]to probe the spin-orbital interactions giving rise to the T z = 0 diagram in Fig. 3b. In the Supplementary Information, we de-scribe how to prepare an array of independent double wells inthe state | e, ↑i L | g, ↓i R , which is a superposition of the foureigenstates featured in Fig. 3b. The energies of these foureigenstates [Eqs. (S4-S7)] can be extracted from the Fourieranalysis of the population imbalance as a function of time: ∆ N ( t ) = n eR + n gL − n gR − n eL = − cos h tJ e J g ~ U − eg i − cos h tJ e J g ~ U + eg i . ∆ N can be measured by combining the dump-ing technique, band mapping, and Stern-Gerlach filtering ofRef. [39] with the use of two probe laser frequencies to distin-guish between | g i and | e i .We now turn to the double-well experiment aimed at prob-ing RKKY interactions. After preparing the state √ ( | g, ↓i L + | g, ↓i R ) | e, ↓i L | e, ↑i R (see Supplementary Informationfor how to prepare this state), we propose to monitor theNeel order parameter for the e atoms, N ez = [ n e ↑ L − n e ↓ L − ( n e ↑ R − n e ↓ R )] . In the limit | V ex | ≪ J g , N ez ( t ) = − cos (cid:0) V ex t ~ (cid:1) − cos (cid:16) V ex t ~ − V ex t J g ~ (cid:17) [in the Supplemen-tary Information, we present the plot of N ez ( t ) for V ex = − J g / ]. It exhibits fast oscillations with frequency ∼ V ex ,modulated by an envelope of frequency ∼ V ex /J g inducedby RKKY interactions. In order to probe RKKY interactionsonly, it is important to suppress super-exchange ∼ J e /U ee andthus to choose J e /U ee small. To study the full spatial depen-dence of RKKY interactions, one must of course go beyondthe double-well setup. We also note that recent experimentsusing alkali atoms populating the lowest two vibrational levelsof a deep optical lattice have measured the local singlet-tripletsplitting induced by V ex [49]. Physics accessible with the alkaline-earth Kondo lattice model The alkaline-earth atom realization of the AF KLM is well-suited to access a number of parameter regimes that are outof reach in solid state materials. One example is the onedimensional (1D) limit, since, to our knowledge, real solidstate materials exhibiting KLM physics are restricted to 2D or3D. Another example is the regime of large Kondo exchange( | V ex | ≫ J g ), which is interesting even for N = 2 . In thislimit the system is well described by the U → ∞ Hubbardmodel [18] by identifying the Kondo singlets with empty sites(holes) and the unpaired localized spins with hard core elec-trons. From this mapping, possible ferromagnetic ordering isexpected at small hole concentration (small n g ), however thestability of this phase for increasing hole concentration andfinite | V ex | values remains unknown. For general N , in theextreme limit J g = 0 , the ground state is highly degenerate:for any distribution of the g atom density n jg < N , there isa ground state (with further spin degeneracy), where on eachsite the spins combine antisymmetrically to minimize the ex-change interaction. Lifting of such extensive degeneracies of-ten leads to novel ground states; this will be addressed in fu-ture studies using degenerate perturbation theory in J g /V ex .For N > , AF SU(N) spin models have a different kind ofextensive degeneracy, which was argued to destroy antiferro-magnetism and to lead to non-magnetic spin liquid and VBS-like ground states [34]. Similar expectations are likely to ap-ply to the KLM at small | V ex | /J g , where the N = 2 antifer-romagnetism may give way to situations where the localizedspins form a non-magnetic state that is effectively decoupledfrom the mobile fermions [22].Even though we have set U gg to zero in Eq. (8), it can betuned, for example, by adjusting the g -lattice depth and cangive rise to interesting physics. For example, the n g = 1 case, which is known to be for N = 2 either an antifer-romagnetic insulator or a Kondo insulator depending on theratio | V ex | /J g [19], will become for large enough U gg and N > a Mott insulator, because the two atoms on each sitecannot combine to form an SU(N) singlet. If n g is reducedfrom unity, the doping of this Mott insulator can be studied,and it will be interesting to understand how this physics, usu-ally associated with cuprate superconductors, is related to theother ground states of the KLM, usually associated with heavyfermion compounds. Experimental Accessibility Immediate experimental accessibility makes our proposalparticularly appealing. Having shown in the main text that the temperature requirements of our proposal are within reach ofcurrent experimental systems, here we show that the nuclear-spin dependence of interaction energies is sufficiently weak tokeep the SU(N) physics intact.In the Supplementary Information, nuclear-spin-dependentvariation in the interaction energies is estimated to be ∆ U gg /U gg ∼ − and ∆ U ee /U ee ∼ ∆ U ± eg /U ± eg ∼ − .Since the scale of SU(N) breaking is at most ∆ U , a very con-servative condition for the physics to be unaffected by SU(N)breaking is that all important energy scales are greater than ∆ U . In particular, in the spin models with more than one atomper site, the condition is ∆ U ≪ J /U , which can be satisfiedsimultaneously with J ≪ U even for ∆ U/U ∼ − . Withone atom per site, the SU(N) breaking scale is not ∆ U butrather ( J/U ) ∆ U , which relaxes the condition to the imme-diately satisfied ∆ U/U ≪ . Similarly, in the Kondo latticemodel, the conditions ∆ V ex ≪ J, | V ex | can be satisfied for ∆ V ex / | V ex | ∼ − . Acknowledgments We gratefully acknowledge conversations withM. M. Boyd, A. J. Daley, S. F¨olling, W. S. Bakr, J. I. Gillen,L. Jiang, G. K. Campbell, and Y. Qi. This work was supportedby NSF, CUA, DARPA, AFOSR MURI, NIST. Author contributions All authors contributed extensively to the work presented inthis paper. Additional information [1] Boyd, M. M. et al. Nuclear spin effects in optical lattice clocks. Phys. Rev. A , 022510 (2007).[2] Campbell, G. K. et al. Probing interactions between ultracoldfermions. Science , 360-363 (2009).[3] Lemke, N. D. et al. Spin-1/2 optical lattice clock. Phys. Rev. Lett. , 063001 (2009).[4] Fukuhara, T., Takasu, Y., Kumakura, M., & Takahashi, Y. De-generate fermi gases of ytterbium. Phys. Rev. Lett. , 030401(2007).[5] Reichenbach, I. & Deutsch, I. H. Sideband cooling while pre-serving coherences in the nuclear spin state in group-II-likeatoms. Phys. Rev. Lett. , 123001 (2007). [6] Hayes, D., Julienne, P. S., & Deutsch, I. H. Quantum logic viathe exchange blockade in ultracold collisions. Phys. Rev. Lett. , 070501 (2007).[7] Daley, A. J., Boyd, M. M., Ye, J., & Zoller, P. Quantum com-puting with alkaline-earth-metal atoms. Phys. Rev. Lett. ,170504 (2008).[8] Gorshkov, A. V. et al. Alkaline-earth-metal atoms as few-qubitquantum registers. Phys. Rev. Lett. , 110503 (2009).[9] Kugel, K. I. & Khomskii, D. I. Crystal structure and magneticproperties of substances with orbital degeneracy. Sov. Phys.-JETP , 725-730 (1973).[10] Arovas, D. P. & Auerbach, A. Tetrahis(dimethylamino)ethylene-C : Multicomponent su-perexchange and Mott ferromagnetism. Phys. Rev. B ,10114-10121 (1995).[11] Li, Y. Q., Ma, M., Shi, D. N., & Zhang, F. C. SU(4) theoryfor spin systems with orbital degeneracy. Phys. Rev. Lett. ,3527-3530 (1998).[12] Pati, S. K., Singh, R. R. P., & Khomskii, D. I. Alternating spinand orbital dimerization and spin-gap formation in coupled spin-orbital systems. Phys. Rev. Lett. , 5406-5409 (1998).[13] Tokura, Y. & Nagaosa, N. Orbital physics in transition-metaloxides. Science , 462-468 (2000).[14] Ruderman, M. A. & Kittel, C. Indirect exchange coupling ofnuclear magnetic moments by conduction electrons. Phys. Rev. , 99-102 (1954).[15] Coqblin, B. & Schrieffer, J. R. Exchange interaction in alloyswith cerium impurities. Phys. Rev. , 847-853 (1969).[16] Doniach, S. The Kondo lattice and weak antiferromagnetism. Physica B+C , 231-234 (1977).[17] Coleman, P. /N expansion for the Kondo lattice. Phys. Rev. B , 5255-5262 (1983).[18] Tsunetsugu, H., Sigrist, M. & Ueda, K. The ground-state phasediagram of the one-dimensional Kondo lattice model. Rev. Mod.Phys. , 809-863 (1997).[19] Assaad, F. F. Quantum Monte Carlo simulations of the half-filled two-dimensional Kondo lattice model. Phys. Rev. Lett. ,796-799 (1999).[20] Tokura, Y., editor. Colossal Magnetoresistive Oxides (Gordonand Breach, New York, 2000).[21] Oshikawa, M. Topological approach to Luttinger’s theorem andthe Fermi surface of a Kondo lattice. Phys. Rev. Lett. , 3370-3373 (2000).[22] Senthil, T., Sachdev, S. & Vojta, M. Fractionalized Fermi liq-uids. Phys. Rev. Lett. , 216403 (2003).[23] Duan, L.-M. Controlling ultracold atoms in multi-band opticallattices for simulation of Kondo physics. Europhys. Lett. , 721-727 (2004).[24] Paredes, B., Tejedor, C. & Cirac, J. I. Fermionic atoms in opti-cal superlattices. Phys. Rev. A. , 063608 (2005).[25] Coleman, P. Heavy fermions: Electrons at the edge of mag-netism. In Handbook of Magnetism and Advanced Magnetic Ma-terials , Kronm¨uller, H. and Parkin, S., editors, volume 1, 95-148(John Wiley & Sons, New York, 2007).[26] Gegenwart, P., Si, Q. & Steglich, F. Quantum criticality inheavy-fermion metals. Nature Phys. , 186-197 (2008).[27] Read, N. & Sachdev, S. Valence-bond and spin-Peierls groundstates of low-dimensional quantum antiferromagnets. Phys. Rev.Lett. , 1694-1967 (1989).[28] Marston, J. B. & Affleck, I. Large-n limit of the Hubbard-Heisenberg model. Phys. Rev. B , 11538-11558 (1989).[29] Harada, K., Kawashima, N., & Troyer, M. Neel and spin-Peierlsground states of two-dimensional SU(N) quantum antiferromag-nets. Phys. Rev. Lett. , 117203 (2003).[30] Assaad, F. F. Phase diagram of the half-filled two-dimensionalSU(N) Hubbard-Heisenberg model: A quantum Monte Carlostudy. Phys. Rev. B , 075103 (2005).[31] Paramekanti, A., & Marston, J. B. SU(N) quantum spin models:a variational wavefunction study. J. Phys.: Condens. Matter ,125215 (2007).[32] Greiter, M. & Rachel, S. Valence bond solids for SU(n) spinchains: Exact models, spinon confinement, and the Haldane gap. Phys. Rev. B , 184441 (2007).[33] Xu, C. & Wu, C. Resonating plaquette phases in SU(4) Heisen-berg antiferromagnet. Phys. Rev. B , 134449 (2008).[34] Hermele, M., Gurarie, V., & Rey, A. M., Mott insulators of ultracold fermionic alkaline earth atoms: underconstrainedmagnetism and chiral spin liquid. Phys. Rev. Lett. , 135301(2009).[35] Wu, C., Hu, J. P., & Zhang, S. C. Exact SO(5) symmetry in thespin-3/2 fermionic system. Phys. Rev. Lett. , 186402 (2003).[36] Honerkamp, C. & Hofstetter, W. Ultracold fermions and theSU(N) Hubbard model. Phys. Rev. Lett. , 170403 (2004).[37] Rapp, A., Hofstetter, W., & Zarand, G. Trionic phase of ultra-cold fermions in an optical lattice: A variational study. Phys.Rev. B , 144520 (2008).[38] Affleck, I., Arovas, D. P., Marston, J. B., & Rabson, D. A.SU(2n) quantum antiferromagnets with exact C-breaking groundstates. Nucl. Phys. B , 467-506 (1991).[39] Trotzky, S. et al. Time-resolved observation and control of su-perexchange interactions with ultracold atoms in optical lattices. Science , 295-299 (2008).[40] Altman, E., Demler, E., & Lukin, M. D. Probing many-bodystates of ultracold atoms via noise correlations. Phys. Rev. A ,013603 (2004).[41] Hofstetter, W., Cirac, J. I., Zoller, P., Demler, E., & Lukin,M. D. High-temperature superfluidity of fermionic atoms in op-tical lattices. Phys. Rev. Lett. , 220407 (2002).[42] Werner, F., Parcollet, O., Georges, A., & Hassan, S. R.Interaction-induced adiabatic cooling and antiferromagnetism ofcold fermions in optical lattices. Phy. Rev. Lett. , 056401(2005).[43] Schneider, U. et al. Metallic and insulating phases of repul-sively interacting fermions in a 3D optical lattice. Science ,1520-1525 (2008).[44] Fukuhara, T., Sugawa, S., Sugimoto, M., Taie, S., & Takahashi,Y. Mott insulator of ultracold alkaline-earth-metal-like atoms. Phys. Rev. A , 041604(R) (2009).[45] Ni, K. K. et al. A high phase-space-density gas of polarmolecules. Science , 231-235 (2008).[46] Rey, A. M., Gorshkov, A. V., & Rubbo, C. Many-body treat-ment of the collisional frequency shift in fermionic atoms.Preprint at < http://arxiv.org/abs/0907.2245 > (2009).[47] Gorshkov, A. V., Jiang, L., Greiner, M., Zoller, P., & Lukin,M. D. Coherent quantum optical control with subwavelengthresolution. Phys. Rev. Lett. , 093005 (2008).[48] Ciuryło, R., Tiesinga, E., & Julienne, P. S. Optical tuning of thescattering length of cold alkaline-earth-metal atoms. Phys. Rev.A , 030701(R) (2005).[49] Anderlini, M. et al. Controlled exchange interaction betweenpairs of neutral atoms in an optical lattice. Nature , 452-456(2007).[50] Cazalilla, M. A., Ho, A. F., & Ueda, M. Ultracold gases ofytterbium: ferromagnetism and Mott states in an SU(6) Fermisystem. Preprint at < http://arxiv.org/abs/0905.4948 > (2009). SUPPLEMENTARY ONLINE MATERIALSKnown Scattering Lengths Very few scattering lengths a X ( X = gg, ee, eg + , eg − ) be-tween g ( S ) and e ( P ) states of alkaline-earth-like atomsare known at the moment. a gg is known for all isotopic com-binations of Yb [51] and Sr [52]. Estimates of a ee for Sr[53] and of a − eg for Sr [54] also exist. Finally, there is aproposal describing how to measure a + eg via clock shifts [55]. Nuclear-Spin Independence of the Scattering Lengths Independence of scattering lengths from the nuclear spin isa key assumption of the paper. This feature allows us to ob-tain SU(N)-symmetric models with N as large as 10 and dis-tinguishes alkaline-earth atoms from alkali atoms, which canexhibit at most an SO(5) symmetry [56, 57, 58, 59], a symme-try that is weaker than SU(4). The assumption of nuclear-spinindependence of scattering lengths is consistent with recentexperiments, where - within experimental precision - the clockshift does not depend on how the Zeeman levels are populated[60, 61]. In this Section, we present the theoretical justifica-tion of this assumption.Direct magnetic dipole-dipole coupling between the nu-clear spins of two atoms sitting on the same site of an op-tical lattice is negligible: even for two magnetic dipole mo-ments as large as 10 nuclear magnetons at a distance of 10 nm(which is significantly smaller than the confinement typicallyachieved in optical lattices [62]), the interaction energy stillcorresponds to a frequency smaller than one Hertz. Therefore,nuclei can affect the collisions only via the electrons. All fourscattering lengths ( a gg , a ± eg , and a ee ) are, thus, expected tobe independent of the nuclear spin because both g and e havetotal electronic angular momentum J equal to zero, which re-sults in the decoupling between nuclear and electronic degreesof freedom during the course of a collision. The decouplingduring a collision is a consequence of the fact that each of thefour molecular electronic states that correlate with the J = 0 separated atom pair has zero projection Ω of total electronicangular momentum on the molecular axis. The nuclear spinsin this case can only couple very weakly to other molecularstates, even if there is a molecular curve crossing.While the short-range potential energy structure for amolecule like Sr is very complex for the excited states[63, 64], we will now show that scattering length differencesamong different combinations of nuclear spin projections forthe same Ω = 0 potential are expected to be very small. Thescattering length a can be computed as a = ¯ a [1 − tan(Φ − π/ , where ¯ a is the average scattering length governed bythe asymptotic behavior of the potential and Φ is the semiclas-sical phase computed at zero energy from the classical turn-ing point R to infinity: Φ = R ∞ R dR p M [ − V ( R )] / ~ , where − V ( R ) is the (positive) depth of the interaction potential atseparation R and M/ is the reduced mass [65]. Defining R ( t ) as the classical trajectory from time t = −∞ to time t = ∞ of a particle of mass M/ at zero energy in the potential V ( R ) , we can rewrite the phase as Φ = − R ∞−∞ dtV ( R ( t )) / ~ .The order of magnitude of the change δ Φ in the phase asso-ciated with different nuclear spin projections can, thus, be es-timated as δ Φ ∼ ∆ tδV / ~ , where ∆ t is the total time in theshort-range part of the collision and δV is the typical energydifference associated with different nuclear spin projectionsduring this time. Since δV vanishes at R → ∞ , only the shortrange molecular region contributes to the phase difference.Therefore, assuming δ Φ ≪ , a ∼ ¯ a , and | cos(Φ − π/ | ∼ , the nuclear-spin-dependent variation δa in the scatteringlength can be estimated as δa/a ∼ δ Φ ∼ ∆ tδV / ~ .Turning to the actual numbers, ∆ t can be estimated fromthe depth ( ∼ cm − hc ) and the range ( ∼ Bohr radii)of the appropriate interatomic potential (see e.g. [63, 64]) tobe ∆ t ≈ ps. For g - g collisions, δV /h can be estimatedby the second-order formula E hf / ( hE opt ) ∼ Hz, where E hf /h ∼ MHz is the approximate value for the hyper-fine splittings in P in Sr and E opt /h ∼ THz is theoptical energy difference between S and P in Sr. Thisyields the following estimate for the dependence of a gg on thenuclear spin: δa gg /a gg ∼ δ Φ ∼ − . For e - e and e - g colli-sions, an analogous second-order formula would use the finestructure splitting between P and P in Sr ( E f /h ∼ THz) instead of E opt to yield δ Φ ∼ − . However, the latterestimate ( δ Φ ∼ − ) is too optimistic since molecular statesthat are split by E f at large interatomic separations may comeorders of magnitude closer at short range [66]. Therefore, amore realistic conservative estimate would use the first-orderformula δV ∼ E hf to yield δa ee /a ee ∼ δa ± eg /a ± eg ∼ δ Φ ∼ − . It is important to note, however, that these are all onlyvery rough estimates. For example, hyperfine coupling in amolecule will differ from the hyperfine coupling in separatedatoms. In fact, since it is very difficult to predict δa/a accu-rately, these values would need to be measured. To concludethis Section, we would like to emphasize that, as mentionedin the main text, if the small nuclear-spin dependence of a ee and a ± eg is not negligible for some applications, one can usetwo different ground state atomic species instead of a groundand an excited state of one species. Likelihood of Lossy e - e Collisions and Possible Solutions Collisions of two e atoms are likely to be accompanied bylarge loss [53]. This can occur if the molecular + g potentialthat correlates with the e - e atoms undergoes an avoided cross-ing with a potential curve that correlates with a lower energypair of separated atoms (see, for example, Ref. [64]). Simi-lar crossings that result in inelastic energy transfer collisionswere examined for P + S collisions of alkaline earth atomsin Ref. [67]. The likelihood of a relatively high probability ofan inelastic event during such a crossing with species suchas Sr or Yb means that the imaginary part b ee of the scatter-ing length is expected to be large. However, just like a ee , b ee can not be calculated accurately from the potentials but wouldneed to be measured.The possible effects of b ee on the four examples we discuss1FIG. S1: A general Young diagram. [Eqs. (6-8) and Eq. (S1)] are as follows. H ( p, [Eq. (7)] is,of course, not affected because it involves only g atoms. In H (1 , [Eq. (S1)] and H KLM [Eq. (8)], the e lattice is assumedto be so deep that J e is negligible compared to U ee + V ex and U ee , respectively, or to the experimental timescale, thus, fullysuppressing tunneling of e atoms and occupation of one siteby more than one e atom. The presence of an imaginary part b ee of the e - e scattering length will give an effective nonzerowidth to the state with more than one e atom per site and can,therefore, only further suppress this tunneling by a Zeno-likeeffect [68, 69, 70].Therefore, H (1 , [Eq. (6)] is the only example that can beaffected by large b ee . In order for H (1 , to contain a non-negligible term proportional to J e /U ee , the ratio | b ee /a ee | would need to be very small [71]. Several approaches toavoid the losses associated with b ee in H (1 , are possible.First, the large variety of stable atoms with two valence elec-trons (which includes not only alkaline-earths, but also Zn,Cd, Hg, and Yb) may have coincidentally an isotope withsmall | b ee /a ee | , which is more likely for lighter atoms [67].Second, while obtaining a good optical Feshbach resonance[51, 72, 73, 74, 75] to reduce | b ee /a ee | might not be possible,it should be possible to use optical Feshbach resonances toenhance b ee and, thus, suppress [68, 69, 70] the virtual occu-pation of one site by two e atoms; H (1 , would then have thesame form as in Eq. (6), except with U ee effectively set to in-finity. Notice that here we suggest to use optical Feshbachresonances to affect e - e scattering, which is different fromthe typical application to g - g scattering [51, 72, 73, 74, 75].Third, one can consider using a different ground state atom torepresent state | e i , which would set V ex = 0 in H (1 , . Fi-nally, one could simply use an e -lattice that is deep enough tomake J e negligible, which would, however, lead to the loss ofterms in H (1 , that exchange the pseudospin between neigh-boring sites. Brief Review of Young Diagrams Irreducible representations of SU(2) are classified accord-ing to the total half-integer angular momentum J and have di-mension J + 1 . On the other hand, a (semistandard) Youngdiagram, instead of a single value J , is used to describe anirreducible representation of SU(N) for a general N [76, 77].As shown in the example in Fig. S1, a Young diagram has allits rows left-aligned, has the length of rows weakly decreasingfrom top to bottom, and has at most N rows. The dimensionof the representation corresponding to a given diagram is the FIG. S2: (p,q) = (1,1) Young diagram. number of ways to fill the diagram with integers from to N such that the numbers weakly increase across each row andstrictly increase down each column. For our purposes, thenumber of boxes in the diagram is the number of atoms on thesite, and the diagram describes the (nuclear) spin symmetry ofthe particular chosen single-site energy manifold. In particu-lar, columns represent antisymmetrized indices, while rowsare related to (but do not directly represent) symmetrized in-dices. It is the relation between antisymmetrized indices andthe columns that limits the number of rows to N . On the otherhand, since the full wavefunction (spin and orbital) on eachsite must satisfy complete fermionic antisymmetry, the rela-tion between rows and symmetrized indices and the fact thatwe have only two orbital states ( g and e ) force all our diagramsto have at most two columns. The ( p, q ) = (1 , spin Hamiltonian and the spin-1 Heisenbergantiferromagnet In the main text, we discussed two special cases of thespin Hamiltonian H ( p,q ) , both of which had a single-columnSU(N) representation on each site (i.e. q = 0 ). In this Sec-tion, we discuss the simplest SU(N) representation with twocolumns, ( p, q ) = (1 , [see Fig. S2]. It can be obtainedwhen there is one g and one e atom per site in the electronicsinglet | ge i − | eg i configuration. Setting J e = 0 to avoid e - e collisions, H ( p,q ) reduces to H (1 , = J g U gg + V ex ) X h i,j i S ij . (S1)The case of N = 2 is the spin-1 antiferromagnetic Heisen-berg model. This model has a 1D ground state with hiddentopological structure [78]. Recently, applications of relatedmodels in one-way quantum computation have been proposed[79, 80]. Models with more complicated two-column repre-sentations may have exotic chiral spin liquid ground states thatsupport non-Abelian anyons and that might thus be used fortopological quantum computation [81]. The Kugel-Khomskii model and the double-well phase diagram In the main text, we omitted the values of the parame-ters in H ( p,q ) that characterize the Kugel-Khomskii model H (1 , [Eq. (6)]. In this Section, we present these parame-ters. We also present a detailed discussion of the double-wellcase phase diagram.2The parameters in H ( p,q ) that characterize the Kugel-Khomskii model H (1 , [Eq. (6)] are λ ijg = − κ ijg = J g U gg ≡− κ g , λ ije = − κ ije = J e U ee ≡ − κ e , κ ijge = − J e + J g U + eg − J e + J g U − eg ≡ κ ge , λ ijge = J e + J g U + eg − J e + J g U − eg ≡ λ ge , ˜ κ ijge = J e J g U − eg − J e J g U + eg ≡ ˜ κ ge , ˜ λ ijge = J e J g U − eg + J e J g U + eg ≡ ˜ λ ge . To avoid loss in e - e collisions,we assume for the rest of this Section that U ee = ∞ (seeSupplementary Information for a discussion of losses in e - e collisions).The nontrivial orbital-orbital, spin-spin, and spin-orbital in-teractions in H (1 , [Eq. (6)] result in competing orders, withthe actual ground-state order dependent on the parameters ofthe Hamiltonian H (1 , . To get a sense of the possible orders,we consider the case N = 2 (with the spin states denoted by ↑ and ↓ ) and discuss the double-well problem, with the wells de-noted by L (left) and R (right). Due to the large optical energyseparating e and g , which we have ignored after Eq. (1), thethree manifolds of constant T z = T zL + T zR ( T z = − , , )should each be considered separately.The four states in the T z = 1 manifold, the subspace oftwo e atoms, are | ee i| s i and | ee i| t i . Here | ee i = | ee i LR isthe orbital (or pseudo-spin) state, while | t i = | ↑↑i LR , | ↓↓i LR , √ ( | ↑↓i LR + | ↓↑i LR ) and | s i = √ ( | ↑↓i LR − | ↓↑i LR ) are the triplet and singlet spin states. Since U ee = ∞ ,all four of these states have zero energy and the ground-statephase diagram is trivial.The four states in the T z = − manifold (two g atoms) aresplit by H (1 , into two energy manifolds: | gg i| t i , E = 0 , (S2) | gg i| s i , E = − J g U gg . (S3)Only | gg i| s i can take advantage of the virtual tunneling sincetwo g atoms in the triplet spin states cannot sit on the samesite. Which of the two manifolds is the ground manifold de-pends on the sign of U gg , as shown in the ground-state phasediagram in Fig. 3a. It is important to emphasize that for U gg < , the subspace of one g atom per site may be sub-ject to extra loss down to the lower energy states that haveboth g atoms in the same well. It is also worth noting that thediagram is only valid for J g ≪ | U gg | .Finally, the eight states in the T z = 0 manifold (one g atomand one e atom) are split by H (1 , into four energy manifolds: | Σ i| t i , E = − ( J g + J e ) U − eg , (S4) | τ i| s i , E = − ( J g + J e ) U + eg , (S5) | τ i| t i , E = − ( J g − J e ) U − eg , (S6) | Σ i| s i , E = − ( J g − J e ) U + eg , (S7) where | Σ i = √ ( | eg i LR − | ge i LR ) and | τ i = √ ( | eg i LR + | ge i LR ) are anti-symmetric and symmetric orbital states, re-spectively. The denominators U − eg and U + eg in the energies ofthe | t i and | s i states, respectively, reflect the fact that tunnel-ing preserves the nuclear spin. At the same time, the ± signsin the numerators can be understood by considering the case J g = J e , when all states with overall symmetry under par-ticle exchange must have zero energy since for these statestunneling is forbidden due to the Pauli exclusion principle.The corresponding ground-state phase diagram as a functionof the signs and relative magnitude of U + eg and U − eg is shown inFig. 3b. As in the case of the T z = 1 phase diagram, negativeinteraction energies may lead to increased losses. Effects of Three-Body Recombination Three-body recombination [70, 82, 83, 84, 85] is a pro-cess during which three atoms come together to form a di-atomic bound state and a single atom, and both final productshave enough kinetic energy to leave the trap. While in certaincases, three-body recombination can be an asset [70], usuallyit results in the loss of atoms and, thus, limits the duration ofthe experiment. For our purposes, we can describe three-bodyrecombination by a decay rate γ [70] resulting in a loss ofthree particles from one site. This rate will likely depend onwhat atomic states are involved and, to the best of our knowl-edge, has not yet been measured or calculated for fermionicalkaline-earth atoms.Out of the four examples [Eqs. (6-8) and Eq. (S1)] that wediscuss, only H (1 , [Eq. (S1)] and H ( p, [Eq. (7)] may beaffected by three-body recombination ( H KLM [Eq. (8)] as-sumes negligible g - g interactions, such as in a very shallow g lattice or with a low density of g atoms). In the case of H (1 , ,two g atoms and one e atom occupy the same site virtuallyin the intermediate state that gives rise to the second orderspin Hamiltonian with interaction strength ∝ J g / ( U gg + V ex ) .Thinking of γ as an effective linewidth for the intermediatestate, H (1 , will be valid and losses small provided that γ issmaller than the effective ”detuning” U gg + V ex . Since scat-tering lengths for alkaline-earth atoms [51, 52, 54] are compa-rable to those for alkali atoms, U gg + V ex can be on the orderof several kHz [62]. At the same time, /γ for bosonic al-kali atoms in deep traps can be on the order of 1 s [86]. If γ were the same in our case, γ ≪ U gg + V ex would be satis-fied. Ways of controlling the interactions via optical Feshbachresonances [51, 72, 73, 74, 75] may also be envisioned.In the case of H ( p, [Eq. (7)], ( n A , n B ) = (1 , does notsuffer from three-body recombination. ( n A , n B ) = (1 , and (2 , may have three atoms per site virtually. As in the dis-cussion of H (1 , , provided γ associated with three g atomsper site is smaller than U gg , these configurations should beaccessible. For the case ( n A , n B ) = (1 , , γ ≫ U gg isalso acceptable, since it will effectively prohibit the tunnel-ing of the atoms to the state with 3 atoms on a site [70], butthe interaction can still take place through the intermediatestate, in which an atom from a B site tunnels to an A siteand back. One can also envision ways to use optical Feshbach3FIG. S3: Square lattice valence plaquette solid for N = 4 . When N = 4 and n A = n B = 1 , four sites are required toform an SU(4) singlet; these singlets can in turn form theschematically shown plaquette-ordered state or a disorderedphase made of resonant plaquette states [87].resonance techniques [51, 72, 73] to induce large γ . To beable to resolve the superexchange coupling ∼ J g /U gg in caseswhere n A or n B is equal to 3, one must have γ < J g /U gg .Given that superexchange coupling can be as high as 1 kHz[62], this condition should also be achievable. Although n A or n B greater than 3 will result in even shorter lifetimes [84],there is a good chance that relatively large n A and n B can beachieved: at least, for bosonic alkali atoms in an n = 5 Mottinsulator state, the lifetime can still be as long as . s [86]. The ( p, spin Hamiltonian with n A = n B = N/ In the main text, we focused on one special case of the anti-ferromagnetic ( p, spin Hamiltonian on a square lattice, thatwith n A + n B = N (where n A and n B denote the numberof atoms per site on the two sublattices). In this Section, wedescribe another interesting and experimentally relevant case, n A = n B = N/ [81, 87, 88, 89, 90, 91, 92, 93]. Poten-tial ground states include states built from valence plaquettes (Fig. S3) [88, 89], resonant plaquette states [87], and topo-logical spin liquids [81, 90]. Valence plaquette states and res-onant plaquette states are the natural generalization of VBSstates and resonant valence bond states (RVB) [94], respec-tively; for example, when n A = n B = 1 , N lattice sites areneeded to form a SU(N) singlet. Fig. S3 depicts a square lat-tice valence plaquette solid for n A = n B = 1 and N = 4 .Techniques for detecting some of these phases are discussedin Ref. [81]. The experiment described in the main text forthe case n A + n B = N can also be generalized to probe the n A = n B = N/ phase diagram including exotic phases suchas valence plaquette solids [Fig. S3], as well as competingmagnetically ordered states. The main difference is that afterpreparing a band insulator of N g atoms per site, each siteshould be split not necessarily into two sites but into the num-ber of sites that is appropriate for the case being considered(e.g. 4 for the case shown in Fig. S3). L R RL FIG. S4: A schematic diagram describing the preparationof the double-well state | e, ↑i L | g, ↓i R . Double-well Kugel-Khomskii and RKKY experiments In the Methods, we have omitted the description of howto prepare the initial states for the proof-of-principle double-well Kugel-Khomskii and RKKY experiments. We presentthis description in this Section.We first describe how to prepare an array of independentdouble wells in the state | e, ↑i L | g, ↓i R , which we use for theproof-of-principle experiment to probe the spin-orbital inter-actions in the Kugel-Khomskii model [95, 96, 97, 98, 99, 100,101]. After loading a band insulator of | g, ↓i atoms in a deepoptical lattice, an additional lattice for both g (green) and e (yellow) atoms with twice the spacing of the first lattice isturned on in one direction to create an array of independentdouble wells [62]. Then, as shown in Fig. S4, in the presenceof an e -lattice bias, σ + polarized light on resonance with the | g, ↓i L → | e, ↑i L transition can be used to prepare the state | e, ↑i L | g, ↓i R . For examples of earlier orbital physics stud-ies with ultracold atoms, where the orbitals are distinguishedonly by the different motional states of the atoms, we refer thereader to Refs. [102, 103, 104, 105, 106, 107] and referencestherein.We now describe how to prepare the initial state √ ( | g, ↓i L + | g, ↓i R ) | e, ↓i L | e, ↑i R (see Fig. S5a) for the double-well proof-of-principle RKKY experiment, whose expectedNeel order parameter N ez ( t ) for V ex = − J g / is show inFig. S5b. The first step to prepare the initial state √ ( | g, ↓i L + | g, ↓i R ) | e, ↓i L | e, ↑i R is to load a band insulator withthree | g, ↓i atoms per site on the long lattice and then slowlyramp up the short lattice with a bias so that it is energeticallyfavorable to have two atoms in the left well and one in theright well. Next one can change the state of the right atomfrom | g, ↓i R to | e, ↑i R by applying a π pulse of σ + polar-ized light resonant with this single-atom transition. The leftwell will be unaffected because the spectrum is modified bythe interactions (if interactions alone do not provide the de-sired selectivity, one could, for example, change the bias ofthe e -lattice). The next step is to change the state of theleft well from two | g, ↓i L atoms populating the lowest twovibrational states to | e, ↓i L | g, ↓i L both populating the low-est vibrational state. This can be accomplished by using π -polarized traveling wave laser light to apply a π pulse reso-nant with the transition between these two many-body states[106]. This results in | e, ↓i L | g, ↓i L | e, ↑i R . One can then4 ba RL FIG. S5: Proof-of-principle experiment to probe RKKYinteractions in an array of isolated double wells. a, Schematic representation of the initial state √ ( | g, ↓i L + | g, ↓i R ) | e, ↓i L | e, ↑i R . b, In the limit | V ex | ≪ J g , the Neel order parameter for the e atoms[ N ez ( t ) = [ n e ↑ L − n e ↓ L − ( n e ↑ R − n e ↓ R )] ] is N ez ( t ) ≈ − cos (cid:0) V ex t ~ (cid:1) − cos (cid:16) V ex t ~ − V ex t J g ~ (cid:17) , which isshown in red for V ex = − J g / . It exhibits fast oscillationswith frequency ∼ V ex , modulated by an envelope offrequency ∼ V ex /J g induced by RKKY interactions( − − cos (cid:16) V ex t J g ~ (cid:17) shown in blue). temporarily shift the g and e lattices relative to each otherto set U ± eg interactions to zero, then make J g nonzero, andwait until the g atom evolves into the desired superposition √ ( | g, ↓i L + | g, ↓i R ) via tunneling. This yields the desiredstate √ ( | g, ↓i L + | g, ↓i R ) | e, ↓i L | e, ↑i R . [1-50] References for the main text are listed at the end of themain text above. [51] Enomoto, K., Kasa, K., Kitagawa, M., & Takahashi, Y. OpticalFeshbach resonance using the intercombination transition. Phys.Rev. Lett. , 203201 (2008).[52] Martinez de Escobar, Y. N. et al. Two-photon photoassociativespectroscopy of ultracold Sr. Phys. Rev. A , 062708 (2008).[53] Traverso, A. et al. Inelastic and elastic collision rates for tripletstates of ultracold strontium. Phys. Rev. A , 060702 (2009).[54] Campbell, G. K. et al. Probing interactions between ultracoldfermions. Science , 360-363 (2009).[55] Rey, A. M., Gorshkov, A. V., & Rubbo, C. Many-body treat-ment of the collisional frequency shift in fermionic atoms.Preprint at < http://arxiv.org/abs/0907.2245 > (2009).[56] Wu, C., Hu, J. P., & Zhang, S. C. Exact SO(5) symmetry in thespin-3/2 fermionic system. Phys. Rev. Lett. , 186402 (2003).[57] Wu, C. Competing orders in one-dimensional spin-3/2Fermionic systems. Phys. Rev. Lett. , 266404 (2005).[58] Chen, S., Wu, C., Zhang, S.-C., & Wang, Y. Exact spontaneousplaquette ground states for high-spin ladder models. Phys. Rev.B , 214428 (2005).[59] Wu, C. Hidden symmetry and quantum phases in spin-3/2 coldatomic systems. Mod. Phys. Lett. B , 1707–1738 (2006).[60] Boyd, M. M. et al. Sr lattice clock with inaccuracy below − . Phys. Rev. Lett. , 083002 (2007).[61] Ludlow, A. D. et al. Sr lattice clock at × − fractionaluncertainty by remote optical evaluation with a Ca clock. Science , 1805-1808 (2008).[62] Trotzky, S. et al. Time-resolved observation and control of su-perexchange interactions with ultracold atoms in optical lattices. Science , 295-299 (2008).[63] Boutassetta, N., Allouche, A. R., & Aubert-Fr´econ, M. Theo-retical study of the electronic structure of the Sr molecule. Phys.Rev. A , 3845-3852 (1996).[64] Czuchaj, E., Kro´snicki, M., & Stoll, H. Valence ab initio cal-culation of the potential energy curves for the Sr dimer. Chem.Phys. Lett. , 401-409 (2003).[65] Gribakin, G. F. & Flambaum, V. V. Calculation of the scatter-ing length in atomic collisions using the semiclassical approxi-mation. Phys. Rev. A , 546-553 (1993).[66] Wang, Y. & Dolg, M. Pseudopotential study of the ground andexcited states of Yb . Theor. Chem. Acc. , 124-133 (1998).[67] Machholm, M., Julienne, P. S., & Suominen, K.-A. Calcula-tions of collisions between cold alkaline-earth-metal atoms in aweak laser field. Phys. Rev. A , 033425 (2001).[68] Syassen, N. et al. Strong dissipation inhibits losses and inducescorrelations in cold molecular gases. Science , 1329-1331(2008). [69] Daley, A. J., Boyd, M. M., Ye, J., & Zoller, P. Quantum com-puting with alkaline-earth-metal atoms. Phys. Rev. Lett. ,170504 (2008).[70] Daley, A. J., Taylor, J. M., Diehl, S., Baranov, M., & Zoller,P. Atomic three-body loss as a dynamical three-body interaction. Phys. Rev. Lett. , 040402 (2009).[71] Tiesinga, E., Williams, C. J., Mies, F. H., & Julienne, P. S. In-teracting atoms under strong quantum confinement. Phys. Rev. A , 063416 (2000).[72] Ciuryło, R., Tiesinga, E., & Julienne, P. S. Optical tuning of thescattering length of cold alkaline-earth-metal atoms. Phys. Rev.A , 030701(R) (2005).[73] Naidon, P. & Julienne, P. S. Optical Feshbach resonances ofalkaline-earth-metal atoms in a one- or two-dimensional opticallattice. Phys. Rev. A , 062713 (2006).[74] Zelevinsky, T. et al. Narrow line photoassociation in an opticallattice. Phys. Rev. Lett. , 203201 (2006).[75] Martinez de Escobar, Y. N., Mickelson, P. G., Yan. M.,& Killian, T. C. Modication of atom scattering using anintercombination-line optical Feshbach resonance at large detun-ing. Preprint at < http://arxiv.org/abs/0906.1837 > (2009).[76] Jones, H. F. Groups, Representations and Physics (Institute ofPhysics Publishing, London, 1998).[77] Fulton, W. Young tableaux. With applications to representa-tion theory and geometry (Cambridge University Press, London,1997).[78] Girvin, S. M. & Arovas, D. P. Hidden topological order in inte-ger quantum spin chains. Phys. Scr. T27 , 156-159 (1989).[79] Brennen, G. K. & Miyake, A. Measurement-based quantumcomputer in the gapped ground state of a two-body Hamiltonian. Phys. Rev. Lett. , 010502 (2008).[80] Verstraete, F. & Cirac, J. I. Valence-bond states for quantumcomputation. Phys. Rev. A , 060302 (2004).[81] Hermele, M., Gurarie, V., & Rey, A. M., Mott insulatorsof ultracold fermionic alkaline earth atoms: underconstrainedmagnetism and chiral spin liquid. Phys. Rev. Lett. , 135301(2009).[82] Esry, B. D., Greene, C. H., & Burke, J. P. Recombination ofthree atoms in the ultracold limit. Phys. Rev. Lett. , 1751-1754(1999).[83] Bedaque, P. F., Braaten, E., & Hammer, H. W. Three-bodyrecombination in Bose gases with large scattering length. Phys.Rev. Lett. , 908-911 (2000).[84] Jack, M. W. & Yamashita, M. Signatures of the quantum fluc-tuations of cold atoms in an optical lattice in the three-body lossrate. Phys. Rev. A , 033605 (2003).[85] Kraemer, T. et al. Evidence for Efimov quantum states in anultracold gas of caesium atoms. Nature , 315-318 (2006).[86] Campbell, G. K. et al. Imaging the Mott insulator shells byusing atomic clock shifts. Science , 649-652 (2006).[87] van den Bossche, M., Zhang, F. C., & Mila, F. Plaquetteground state in the two-dimensional SU(4) spin-orbital model. Eur. Phys. J. B , 367-370 (2000).[88] Pankov, S., Moessner, R., & Sondhi, S. L. Resonating singletvalence plaquettes. Phys. Rev. B , 104436 (2007). [89] Xu, C. & Wu, C. Resonating plaquette phases in SU(4) Heisen-berg antiferromagnet. Phys. Rev. B , 134449 (2008).[90] Wang, F. & Vishwanath, A. A Z spin-orbital liquid statein the square lattice Kugel-Khomskii model. Preprint at < http://arxiv.org/abs/0806.1743 > (2008).[91] Greiter, M. & Rachel, S. Valence bond solids for SU(n) spinchains: Exact models, spinon confinement, and the Haldane gap. Phys. Rev. B , 184441 (2007).[92] Rachel, S., Thomale, R. F¨uhringer, M., Schmitteckert, P.,& Greiter, M. Spinon confinement and the Haldane gap inSU(n) spin chains. Preprint at < http://arxiv.org/abs/0904.3882 > (2009).[93] Sutherland, B. Model for a multicomponent quantum system. Phys. Rev. B , 3795-3805 (1975).[94] Anderson, P. W. The resonating valence bond state in La CuO and superconductivity. Science , 1196-1198 (1987).[95] Arovas, D. P. & Auerbach, A.Tetrahis(dimethylamino)ethylene-C : Multicomponent su-perexchange and Mott ferromagnetism. Phys. Rev. B ,10114-10121 (1995).[96] Li, Y. Q., Ma, M., Shi, D. N., & Zhang, F. C. SU(4) theoryfor spin systems with orbital degeneracy. Phys. Rev. Lett. ,3527-3530 (1998).[97] Pati, S. K., Singh, R. R. P., & Khomskii, D. I. Alternating spinand orbital dimerization and spin-gap formation in coupled spin-orbital systems. Phys. Rev. Lett. , 5406-5409 (1998).[98] Zasinas, E., Sushkov, O. P., & Oitmaa, J. Phase diagram of thespin-orbital model on the square lattice. Phys. Rev. B , 184431(2001).[99] Li, P. & Shen, S.-Q. Spin and orbital valence bond solids ina one-dimensional spin-orbital system: Schwinger boson mean-field theory. Phys. Rev. B , 214439 (2005).[100] Itoi, C., Qin, S., & Affleck, I. Phase diagram of a one-dimensional spin-orbital model. Phys. Rev. B , 6747-6756(2000).[101] Zhang, G.-M. & Shen, S.-Q. Ordered valence-bond statesin symmetric two-dimensional spin-orbital systems. Phys. Rev.Lett. , 157201 (2001).[102] Wu, C. Orbital ordering and frustration of p-band Mott insula-tors. Phys. Rev. Lett. , 200406 (2008).[103] Zhao, E. & Liu, W. V. Orbital order in Mott insulators of spin-less p-band fermions. Phys. Rev. Lett. , 160403 (2008).[104] Xu, C. & Fisher, M. P. A. Bond algebraic liquid phase instrongly correlated multiflavor cold atom systems. Phys. Rev. B , 104428 (2007).[105] Wang, L., Dai, X., Chen, S., & Xie, X. C. Magnetism of coldfermionic atoms on the p band of an optical lattice. Phys. Rev. A , 023603 (2008).[106] M¨uller, T., F¨olling, S., Widera, A., & Bloch, I. State prepa-ration and dynamics of ultracold atoms in higher lattice orbitals. Phys. Rev. Lett. , 200405 (2007).[107] Anderlini, M. et al. Controlled exchange interaction betweenpairs of neutral atoms in an optical lattice. Nature448