REPORTS

which the weak-valued photon trajectories are trajectory reconstruction is carried out further 10. D. Bohm, Phys. Rev. 85, 180 (1952).
reconstructed. and further from the initial plane at z = 2.75 m. 11. N. W. M. Ritchie, J. G. Story, R. G. Hulet, Phys. Rev. Lett.
66, 1107 (1991).
For the experimentally reconstructed trajecto- Single-particle trajectories measured in this fash- 12. R. Mir et al., N. J. Phys. 9, 287 (2007).
ries for our double slit (Fig. 3), it is worth stress- ion reproduce those predicted by the Bohm–de 13. J. S. Lundeen, A. M. Steinberg, Phys. Rev. Lett. 102,
ing that photons are not constrained to follow Broglie interpretation of quantum mechanics (8), 020404 (2009).
these precise trajectories; the exact trajectory of although the reconstruction is in no way depen- 14. K. Yokota, T. Yamamoto, M. Koashi, N. Imoto, N. J. Phys.
11, 033011 (2009).
an individual quantum particle is not a well- dent on a choice of interpretation. 15. G. J. Pryde, J. L. O’Brien, A. G. White, T. C. Ralph,
defined concept. Rather, these trajectories rep- Controversy surrounding the role of mea- H. M. Wiseman, Phys. Rev. Lett. 94, 220405 (2005).
resent the average behavior of the ensemble of surement in quantum mechanics is as old as the 16. O. Hosten, P. Kwiat, Science 319, 787 (2008);
photons when the weakly measured momentum quantum theory itself, and nowhere have the 10.1126/science.1152697.
17. P. B. Dixon, D. J. Starling, A. N. Jordan, J. C. Howell,
in each plane is recorded contingent upon the paradoxes been thrown into such stark relief as Phys. Rev. Lett. 102, 173601 (2009).
final position at which a photon is observed. The in the context of the double-slit experiment. Our 18. A. Peres, Phys. Rev. Lett. 62, 2326 (1989).
trajectories resemble a hydrodynamic flow with a experimentally observed trajectories provide an 19. A. J. Leggett, Phys. Rev. Lett. 62, 2325 (1989).
central line of symmetry clearly visible: Trajec- intuitive picture of the way in which a single par- 20. Y. Aharonov, L. Vaidman, Phys. Rev. Lett. 62, 2327
(1989).
tories originating from one slit do not cross the ticle interferes with itself. It is of course impos-
21. J. Dressel, S. Agarwal, A. N. Jordan, Phys. Rev. Lett. 104,
central line of symmetry into the opposite side of sible to rigorously discuss the trajectory of an 240401 (2010).
the interference pattern. Trajectories at the edges individual particle, but in a well-defined opera- 22. Materials and methods are available as supporting

Downloaded from https://www.science.org at Indian Institute of Technology Bombay on June 15, 2023
of bright fringes tend to cross over to join more tional sense we gain information about the aver- material on Science Online.
central bright fringes, thus generating the ob- age momentum of the particle at each position 23. R. P. Mirin, Appl. Phys. Lett. 84, 1260 (2004).
24. R. H. Hadfield et al., Opt. Express 13, 10846
served intensity distribution because of interfer- within the interferometer, leading to a set of “av- (2005).
ence. The trajectories cross over dark fringes at erage trajectories.” The exact interpretation of 25. R. Loudon, The Quantum Theory of Light (Oxford Univ.
relatively steep angles; there is a low probability these observed trajectories will require continued Press, Oxford, ed. 3, 2000).
of finding a photon in these regions that cor- investigation, but these weak-measurement results 26. M. Born, E. Wolf, Principles of Optics (Cambridge Univ.
Press, Cambridge, 1999).
respond to postselecting on a state nearly or- can be grounded in experimental measurements 27. R. Feynman, The Feynman Lectures on Physics
thogonal to the initial state of the system. The that promise to elucidate a broad range of quan- (Addison-Wesley, Boston, 1989).
separation of the imaging planes sets the scale tum phenomena (7, 11–13, 15–17). By using the Acknowledgments: A.M.S. conceived of the experiment and
over which features in the trajectories can be power of weak measurements, we are able to pro- supervised the work; M.J.S. and R.P.M. designed and
fabricated the single-photon sources; S.K. constructed the
observed. The evolution of the interference in vide a new perspective on the double-slit experi- experiment and acquired the data with assistance from
our double-slit apparatus takes place over a scale ment, which Feynman famously considered to B.B. and S.R. and much guidance from L.K.S.; B.B., S.K.,
that is much longer than the separation between have in it “the heart of quantum mechanics” (27). and S.R. carried out the data analysis and generated
imaging planes, and our trajectories can accu- the figures; S.K. and L.K.S. wrote the text with input
from all the other co-authors. This work was supported
rately track the evolution of this interference. The
References and Notes by the Natural Sciences and Engineering Research
one place where the accuracy can suffer is when 1. N. Bohr, Collected Works (North-Holland, Amsterdam, Council of Canada, the Canadian Institute for Advanced
a trajectory quickly passes through a dark fringe; 1972). Research, and QuantumWorks. S.K. thanks B. Higgins
here the fine scale behavior is smaller than the 2. W. K. Wootters, W. H. Zurek, Phys. Rev. D 19, 473 (1979). and L.K.S. thanks K. J. Resch for useful discussions.
spacing between imaging planes. By overlaying 3. N. Bohr, Naturwissenschaften 16, 245 (1928). The data from the experiment has been archived
4. M. O. Scully, B.-G. Englert, H. Walther, Nature 351, 111 and is available at www.physics.utoronto.ca/~aephraim/
the trajectories on top of the measured intensity (1991). data/PhotonTrajectories.
distribution (Fig. 4), we observe that the trajec- 5. C. Brukner, A. Zeilinger, Phys. Rev. Lett. 83, 3354
tories reproduce the global interference pattern (1999).
6. V. Jacques et al., Science 315, 966 (2007).
Supporting Online Material
well. The tendency of the reconstructed trajecto- www.sciencemag.org/cgi/content/full/332/6034/1170/DC1
7. Y. Aharonov, D. Z. Albert, L. Vaidman, Phys. Rev. Lett. 60,
ries to “bunch” together within each bright in- 1351 (1988).
Materials and Methods
terference fringe is an artifact of measurement 8. H. Wiseman, N. J. Phys. 9, 165 (2007). 27 December 2010; accepted 11 April 2011
noise with the position error accumulating as the 9. D. Bohm, Phys. Rev. 85, 166 (1952). 10.1126/science.1202218

Spin-Liquid Ground State of the S = 1/2 no spontaneously broken symmetries (4). A key
problem in searching for spin liquids in two-
dimensional (2D) models is that there are no ex-
Kagome Heisenberg Antiferromagnet act or nearly exact analytical or computational
methods to solve infinite 2D quantum lattice sys-
Simeng Yan,1 David A. Huse,2,3 Steven R. White1* tems. For 1D systems, the density matrix renor-
malization group (DMRG) (5, 6), the method we
We use the density matrix renormalization group to perform accurate calculations of the ground use here, serves in this capacity. In addition to
state of the nearest-neighbor quantum spin S = 1/2 Heisenberg antiferromagnet on the kagome its interest as an important topic in quantum mag-
lattice. We study this model on numerous long cylinders with circumferences up to 12 lattice netism, the search for spin liquids thus serves
spacings. Through a combination of very-low-energy and small finite-size effects, our results as a test-bed for the development of accurate and
provide strong evidence that, for the infinite two-dimensional system, the ground state of this widely applicable computational methods for
model is a fully gapped spin liquid. 2D many-body quantum systems.

e consider the quantum spin S = 1/2 lattice (Fig. 1A). This frustrated spin system has

W
isotropic
→
kagome Heisenberg antiferromagnet
(KHA) with only nearest-neighbor
exchange
→
interactions
→ →
(Hamiltonian
long been thought to be an ideal candidate for a
simple, physically realistic model that shows a
spin-liquid ground state (1–3). A spin liquid is a
1
Department of Physics and Astronomy, University of Cali-
fornia, Irvine, CA 92617, USA. 2Department of Physics,
Princeton University, Princeton, NJ 08544, USA. 3Institute
for Advanced Study, Princeton, NJ 08540, USA.
H ¼ S S i ⋅ S j , where S i and S j are the spin magnetic system that has “melted” in its ground *To whom correspondence should be addressed. E-mail:
operators for sites i and j, respectively) on a kagome state because of quantum fluctuations, so it has [email protected]

www.sciencemag.org SCIENCE VOL 332 3 JUNE 2011 1173

 REPORTS
The KHA has been studied with approxi- cylinders of different length to remove end have minimal finite-size effects and thus provide
mate approaches for decades (1), with proposals effects. reliable estimates of the 2D energy.
for spin-liquid and valence-bond–crystal (VBC) Our DMRG results for ground-state energies, The ground state that we find has only short-
ground states (2, 3). In the past few years, nu- along with results from other approaches, are range correlations, with a nonzero energy gap
merical evidence had suggested that the KHA shown in Fig. 1B. The DMRG energies are con- for any excitations, including spin-singlet excita-
ground state might be a VBC with a 36-site unit sistent with the Lanczos results (16, 17) and well tions. We have tested the response of this ground
cell of bonds with spatially nonuniform spin-spin below the energies of MERA (10) and the series state to many sorts of perturbations that would
correlations in the ground state that break the expansion for the HVBC (8, 9). The previous select out ordered states, if they exist, without
translational symmetry of the kagome lattice. DMRG result (12) is close to the true ground detecting any signs of any ordering. Thus, we con-
First proposed in (2), and explored in more detail state (17) for a torus with a circumference of six clude that the ground state is a gapped spin liquid.
in (7), this honeycomb VBC (HVBC) was also (circumference-6) lattice spacings. The entangle- Such a state can be represented as some sort of
studied with a perturbative series expansion (8, 9). ment across a cut that separates a circumference- short-range resonating–valence-bond (RVB) state
The series for the ground-state energy rapidly six torus into two parts is roughly the same as that (18–24). The shortest resonant loops of singlet
converges to a low energy (8, 9). Subsequently, across a cut that separates a circumference-12 dimers in a nearest-neighbor RVB state on a
Evenbly and Vidal (10) used the multiscale en- cylinder—our current limit. Thus, it is not sur- kagome lattice each surround only one hexagon of
tanglement renormalization ansatz (MERA), ob- prising that the DMRG energies for tori (12) are the lattice (7). If the RVB state had all dimer covers
taining the HVBC and an energy close to the series overestimates for circumferences larger than six equally weighted, all 32 of these elementary res-

Downloaded from https://www.science.org at Indian Institute of Technology Bombay on June 15, 2023
result. Another approximate numerical approach lattice spacings. We also obtain a new rigorous onant loops would be equally present. Instead, the
based on an effective quantum dimer model also upper bound on the ground-state energy of the ground state appears to substantially overemphasize
yielded the HVBC as the ground state, but noted infinite 2D system using the DMRG on strips certain eight-site loops of a diamond shape (Fig.
that a Z2 spin liquid is close by in a generalized with open boundary conditions (15); this is well 2B). To test the response of the ground state to
parameter space (11). A recent DMRG study (12), below the previous bound from MERA (10). enhancing each of these 32 elementary resonant
in contrast, found a spin-liquid ground state, but The 36-site unit cell of the HVBC does fit on loops, we slightly increased the exchange cou-
on the largest lattices the energy obtained was sub- the XC8 cylinder (1/c ≅ 0.14, where c is the cir- plings along the bonds of such a loop at the center
stantially above that of the HVBC, suggesting cumference), allowing a direct comparison be- of a YC8 cylinder and measured how much this
that the method had not found the true ground tween the HVBC series and our DMRG: the enhanced the spin-spin correlations along the
state. The simple KHA is clearly near a number of DMRG energy is lower by 0.004(1). Comparing loop and elsewhere. It is the eight-site diamond
different possible ground-state phases upon add- this cylinder to the corresponding torus, the en- loop that elicits the strongest response (Fig. 2B).
ing small perturbations to its Hamiltonian [such ergy shows strong finite-size effects in the HVBC The six-site “perfect hexagon” loop (Fig. 2A)
as a second-neighbor exchange interaction (13)], series (8, 9), but the same comparison between (2, 7, 8), shows a much smaller response, sug-
and it is this proximity of many competing states Lanczos (16) and our DMRG shows that the true gesting that this resonant loop is actually under-
that has made this such a challenging system. A finite-size effect is much smaller (15). These weighted in the ground state.
variational state that may serve as a multicrit- finite-size effects remain small for even smaller One cannot tile a kagome lattice with just
ical state between all these possible ground-state circumferences (17). This is consistent with the these favored resonant diamonds. However, a
phases is the critical gapless spin liquid of (14). small correlation length that we find: less than particular “diamond-pattern” VBC (Fig. 2C), ap-
We performed an extensive DMRG study 1.5 lattice spacings (15). We conclude that the pears to be closely related to the spin liquid, and it
of the KHA with important differences in tech- ground-state energies of our widest cylinders is useful to think of the spin liquid as a melted
nique from the previous study. Most important-
ly, we studied long cylinders with open ends,
thus avoiding fully periodic (toroidal) boundary
conditions, which are known to greatly magnify
the truncation errors in the DMRG (6). We also
show why the series expansions appear to con-
verge well, yet reach the incorrect HVBC state:
The path connecting the series expansion start-
ing point and the KHA is interrupted by a first-
order phase transition (15). For details on the
competition between the spin liquid and HVBC
states, see figs. S1 to S3.
DMRG efficiently finds the ground state for
long 1D systems, but for 2D systems one must
study cylinders or strips of limited width. We
studied mostly cylinders with a maximum circum-
ference of 12 lattice spacings, obtaining ground-
state energies with uncertainties of less than 0.1%,
with much higher accuracy for the narrower cyl-
inders. We labeled the cylinders by their orien-
tation, circumference, and any shift in wrapping
the cylinder periodically (Fig. 1A) (15). For ex- Fig. 1. (A) A section of kagome lattice. The frustration is indicated by the spins shown on one triangle. To
ample, “YC9-2” denotes a cylinder (C) with some make a cylinder the top (a) is connected to the bottom (b) by the dotted bonds indicated; this example
of the bonds oriented in the y direction (Y), with makes cylinder YC6 (15). To make a torus, similarly connect sides c and d. (B) Comparison of energies per
circumference of nine lattice spacings and a shift site for various lattices and methods. For cylinders, the horizontal axis is the inverse circumference in units
of two columns when connected periodically. We of inverse lattice spacings. For tori (12, 16, 17), the smallest circumference is used. The MERA (10) and our
estimate the energy per site for each type of in- DMRG upper bound (15) results apply directly to an infinite 2D system, as does the series HVBC result (8, 9)
finitely long cylinder by subtracting energies of on the axis. Error bars indicate estimated uncertainties.

1174 3 JUNE 2011 VOL 332 SCIENCE www.sciencemag.org

 REPORTS
state of this crystal. We have measured the cor- so we do not have gap estimates for our widest winding numbers by choosing how many spins
relation length for VBC energy correlations in cylinder (15). See table S1 for energies and gaps to leave at each end. For a finite circumference c,
this diamond pattern along cylinder YC8 and of all the cylinders studied. these two sectors have different ground-state
find that it is less than 1.5 lattice spacings (15). The singlet gap is 0.050, within the errors, for energies (E ) per site; for YC8, their difference
Unlike the HVBC, which shows a first-order the even XC8 and YC8 cylinders, and it remains is dE ≈ 0.00069(3). For odd cylinders, the two
phase transition to the spin liquid (15), this dia- near this value for the wider even XC12-2 cyl- sectors are related by translation along the length
mond VBC evolves smoothly into the spin liquid inder. The odd cylinders come in at least two of the cylinder and, thus, are degenerate. If an
without any phase transition as one changes the families: YC6 and YC10 are not spirals, whereas even cylinder is topologically ordered, one ex-
strengths of the exchanges that are altered to YC5-2, YC7-2, and YC9-2 are a series of spirals pects dE ∼ exp(−c/x) (where x is a correlation
favor it (we call this “pinning it,” Fig. 2C). For with increasing circumference. In each of these length ) at large c, but we do not yet have re-
the even cylinders on which this diamond VBC (small) families of odd cylinders, the singlet gap liable dE results for larger cylinders to test this
does fit, this process allows a careful production increases as the circumference increases, support- hypothesis. This is partly because the singlet
of the spin liquid by approaching it in a smooth ing our conclusion that the singlet gap is ~ 0.05 gap above the ground state in the higher-energy
fashion from the diamond VBC (15). in the 2D limit. This is quite different from the sector is substantially smaller than the singlet gap
The infinitely long cylinders may be viewed exact diagonalization results, where there are above the overall ground state, slowing DMRG’s
as one-dimensional systems with a unit cell con- many lower-lying singlets, and the lowest singlet convergence.
taining Nc spins. Even-Nc cylinders (e.g., Nc = 12 gap is only ~0.01 on the standard 36-site torus A domain wall along the cylinder where the

Downloaded from https://www.science.org at Indian Institute of Technology Bombay on June 15, 2023
for YC8) are compatible with the diamond VBC, (26). The singlet gap is a strong function of a Z2 winding number changes is a spinon. For the
and the ground state of the infinite cylinder ap- second-neighbor exchange coupling (J2) (13), odd cylinders, the degeneracy of the two sectors
pears to be nondegenerate. Odd-Nc cylinders are with an apparent phase transition at a very small means the spinons are unconfined. However, two
not compatible with the diamond VBC, and the ferromagnetic value of J2. The location of this
Lieb-Schultz-Mattis theorem implies that the nearby transition is quite sensitive to circumfer-
ground state must be degenerate (25). Ground ences, and this produces the large differences in
states on these odd cylinders weakly break transla- the singlet gap between the torus and our cylinders.
tional invariance, spontaneously doubling the unit The triplet (spin) gap on the 36-site torus is
cell, which produces a pair of degenerate ground 0.164 from exact diagonalization (26). Although
states, still with a gap to higher excited singlet XC8 and YC8 have gaps that are quite close to
states. The symmetry breaking is in a “striped” this, our results on other cylinders suggest that
pattern (Fig. 3). For YC6 and YC10, the stripes the 2D triplet gap is smaller (Fig. 4). The triplet
run around the circumference, whereas for the excitations are composed of two spinons, but we
other odd spiral cylinders the stripes are spirals. cannot resolve whether or not the two spinons
The ends of our cylinders may have low-lying bind, although in some cylinders any binding
states below the bulk singlet and triplet gaps. The must be very weak. This composite nature of the
following DMRG procedure avoids these edge excitation seems to make the finite-size effects
states: First target only one state, and sweep and variation between the cylinders more pro- -0.1 -0.05
enough to obtain a high-accuracy ground state. nounced. We do not yet understand the details of 0.0
Then restrict the range of bonds that are updated these effects. As for the singlet gap, the spiral odd 0.0
0.1
in the DMRG sweeps to the central half of the cylinders have the smallest triplet gaps, and the 0.05
sample and target the two lowest-energy states, even cylinders the largest. The triplet gap remains
again sweeping to high accuracy, but keeping the above the singlet gap in all the systems we have Fig. 3. Ground-state energy pattern for a YC9-2
end regions of the samples locally in the ground studied; thus, we believe it remains nonzero in cylinder. The colors of the triangles and their in-
state. This technique is particularly important for the 2D limit. tensities indicate the deviation of the sum of the
obtaining the singlet gap. For the triplet gap, one A nearest-neighbor RVB wave function is a spin-spin correlations on the three bonds forming
can also keep the excitation away from the ends linear combination of nearest-neighbor singlet the triangle from 3e0, with e0 = −0.219.
with local magnetic fields. With this approach, dimers covering the kagome lattice (24). For a
we can target both states together—one with total kagome lattice wrapped on a cylinder, such dimer
Sz = 0 and the other with Sz = 1—or run them covers are in two topologically distinct sectors that
separately. These different approaches allowed differ by a Z2 winding number, and dimer reso-
for fairly independent checks on the results. nances on finite loops do not change this winding
Figure 4 shows the measured bulk gaps. Gaps number (19, 20, 22–24). We can force our even
are more demanding than ground-state energies, cylinders to have one or the other of the two
Fig. 2. Response of the spin
liquid to small perturbations on
cylinder YC8. In each case, cer-
tain bonds have been strength-
ened: (A) a six-site hexagon by
1%, (B) an eight-site diamond
by 1%, and (C) the wide strong
and weak vertical bonds (12 of
each are shown) by T0.5%. Line → →
widths indicate subtracted bond spin-spin correlations, |〈 Si ⋅ Sj 〉 − ea|, with dashed lines indicating that
the quantity is positive. Here, a signifies bond direction, and the approximate unperturbed bond cor- Fig. 4. Spin triplet (solid symbols) and singlet (hol-
relations are ea = −0.223 for horizontal bonds and −0.217 for diagonal. Note the different scales used: low symbols) gaps for various cylinders with circum-
the response increases strongly from (A) to (B) to (C). Surrounding dimers arise in response to the diamond ferences c. The type of cylinder (15) is indicated in
pinning in (B), and diamonds arise in response to dimer pinning in (C). the key (inset).

www.sciencemag.org SCIENCE VOL 332 3 JUNE 2011 1175

 REPORTS
spinons might bind with a finite binding energy— References and Notes 20. N. Read, B. Chakraborty, Phys. Rev. B 40, 7133
it appears that this might be happening on cyl- 1. V. Elser, Phys. Rev. Lett. 62, 2405 (1989). (1989).
2. J. B. Marston, C. Zeng, J. Appl. Phys. 69, 5962 (1991). 21. N. Read, S. Sachdev, Phys. Rev. Lett. 66, 1773 (1991).
inder YC10 and, thus, it might also occur in the 3. S. Sachdev, Phys. Rev. B 45, 12377 (1992). 22. X. G. Wen, Phys. Rev. B 44, 2664 (1991).
2D limit. For the even cylinders, on the other hand, 4. L. Balents, Nature 464, 199 (2010). 23. R. Moessner, S. L. Sondhi, Phys. Rev. Lett. 86, 1881
the spinons are confined by an effective potential 5. S. R. White, Phys. Rev. Lett. 69, 2863 (1992). (2001).
that grows linearly with the distance, because 6. S. R. White, Phys. Rev. B 48, 10345 (1993). 24. G. Misguich, D. Serban, V. Pasquier, Phys. Rev. B 67,
7. P. Nikolic, T. Senthil, Phys. Rev. B 68, 214415 (2003). 214413 (2003).
the domain between them is in the higher-energy 25. E. Lieb, T. Schultz, D. Mattis, Ann. Phys. 16, 407 (1961).
8. R. R. P. Singh, D. A. Huse, Phys. Rev. B 76, 180407 (2007).
sector. As a result, the excitation across the spin 9. R. R. P. Singh, D. A. Huse, Phys. Rev. B 77, 144415 (2008). 26. C. Waldtmann et al., Eur. Phys. J. B 2, 501 (1998).
gap for the even cylinders is a bound spinon pair. 10. G. Evenbly, G. Vidal, Phys. Rev. Lett. 104, 187203 Acknowledgments: We thank A. Lauchli for Lanczos results;
It remains to be determined if they remain bound (2010). R. Singh for series results; and C. Lhuillier, M. P. A. Fisher,
11. D. Poilblanc, M. Mambrini, D. Schwandt, Phys. Rev. B 81, T. Senthil, E. Sorensen, S. Sondhi, G. Vidal, X.-G. Wen,
in the 2D limit. O. Tchernyshyov, and M. Stoudenmire for discussions.
180402(R) (2010).
Much remains to be understood concerning 12. H. C. Jiang, Z. Y. Weng, D. N. Sheng, Phys. Rev. Lett. This work was supported by NSF grants DMR-0907500
the low-energy behavior of the KHA, particularly 101, 117203 (2008). and DMR-0819860.
the detailed structure, exchange statistics, and dis- 13. P. Sindzingre, C. Lhuillier, Europhys. Lett. 88, 27009
Supporting Online Material
persion relations of the various excitations. It will (2009).
www.sciencemag.org/cgi/content/full/science.1201080/DC1
14. Y. Ran, M. Hermele, P. A. Lee, X.-G. Wen, Phys. Rev. Lett.
be instructive to also explore the phase diagram 98, 117205 (2007).
Materials and Methods
in the vicinity of this simple nearest-neighbor– SOM Text
15. Supporting material is available on Science Online.

Downloaded from https://www.science.org at Indian Institute of Technology Bombay on June 15, 2023
Figs. S1 to S3
only Heisenberg model by changing the Hamil- 16. P. W. Leung, V. Elser, Phys. Rev. B 47, 5459 (1993).
Table S1
tonian in various ways (13) to find what other 17. A. M.Läuchli, J. Sudan, E. S. Sørensen, http://arxiv.
References (5, 6, 8–10, 12, 16, 17, 27–29)
org/abs/1103.1159.
phases are nearby and perhaps to move “deep- 18. P. W. Anderson, Mater. Res. Bull. 8, 153 (1973). 30 November 2010; accepted 7 April 2011
er” into this spin liquid where it might be easier 19. S. A. Kivelson, D. S. Rokhsar, J. P. Sethna, Phys. Rev. B Published online 28 April 2011;
to study. 35, 8865 (1987). 10.1126/science.1201080

(9, 10). Here we report the creation of an artifi-

Two-Dimensional Mott-Hubbard cial lattice with honeycomb geometry for trap-
ping electrons, and we demonstrate the formation
Electrons in an Artificial of HBs through strong correlations. We nano-
fabricated the artificial lattice on the surface of

Honeycomb Lattice a gallium arsenide (GaAs) heterostructure that

hosts a high-quality two-dimensional electron gas
(2DEG) (11–13). Similar nanostructures have
A. Singha,1* M. Gibertini,1 B. Karmakar,1 S. Yuan,2 M. Polini,1,3† G. Vignale,4,3
been studied in the past in the context of the
M. I. Katsnelson,2 A. Pinczuk,5 L. N. Pfeiffer,6 K. W. West,6 V. Pellegrini1†
Hofstadter’s fractal energy spectrum (14, 15). We
Artificial crystal lattices can be used to tune repulsive Coulomb interactions between electrons. probed the electron excitation spectrum by in-
We trapped electrons, confined as a two-dimensional gas in a gallium arsenide quantum elastic light-scattering and observed signatures
well, in a nanofabricated lattice with honeycomb geometry. We probed the excitation spectrum stemming from strong Coulomb interactions,
in a magnetic field, identifying collective modes that emerged from the Coulomb interaction in which we could tune by applying an external mag-
the artificial lattice, as predicted by the Mott-Hubbard model. These observations allow us to netic field. Carriers in the patterned structures
determine the Hubbard gap and suggest the existence of a Coulomb-driven ground state. supported anpffiffiffiunusual collective mode; its energy
scales like B, where B is the component of the
he conduction electrons of a crystal ex- phenomena and metal-insulator phase transitions magnetic field perpendicular to the 2DEG. A

T perience the potential created by a period-

ic arrangement of coupled quantum units
(such as ions and atoms). Numerous properties of
(3, 4). Mott showed (5) that interaction-induced
insulators are better described in real space (rath-
er than in momentum space), in which the solid is
theoretical analysis based on a minimal Hubbard
model reveals that the mode energy is determined
by the onsite Coulomb interaction and represents
solids can be explained in terms of Bloch bands viewed as a collection of localized electrons bound direct evidence of the existence of HBs in the
(1), neglecting Coulomb interactions between elec- to atoms with partially filled shells. Electrons hop- 2DEG subjected to the artificial lattice. At low
trons. These interactions often lead to qualitative ping through the lattice are absorbed and emitted temperatures and large B fields, we found evi-
changes that are particularly pronounced in solids from the atoms, thus originating two bands, which dence for the opening of an unexpected gap in the
with narrow energy bands (2–4). These strongly are split by the energy cost of having two electrons spin excitation spectrum. We argue that the ob-
correlated materials display exotic ordering with antiparallel spin on the same atomic site. served gap reveals the occurrence of a new cor-
Hubbard subsequently introduced a model related phase of electrons in a honeycomb lattice
1
National Enterprise for nanoScience and nanoTechnology,
Hamiltonian with onsite interactions that displays akin to one of those discussed in the context of
Istituto Nanoscienze–CNR, and Scuola Normale Superiore, split bands (Hubbard bands, or HBs) in the strong- graphene in high magnetic fields (16–18). These
I-56126 Pisa, Italy. 2Radboud University Nijmegen, Institute ly correlated (or atomic) limit (6). HBs (and their findings pave the way for the possibility to ex-
for Molecules and Materials, NL-6525 AJ Nijmegen, Nether- coexistence with quasiparticle bands in the cor- plore graphene-like physics in the ultrahigh mag-
lands. 3Kavli Institute for Theoretical Physics China, Chinese
related metallic phase) are characteristic features netic field limit, in which the magnetic length is
Academy of Sciences, Beijing 100190, China. 4Department of
Physics and Astronomy, University of Missouri, Columbia, MO of such strongly correlated systems. Fermions on smaller than the lattice constant of the artificial
65211, USA. 5Department of Applied Physics and Applied a honeycomb lattice, in particular, have been pre- crystal—a regime not accessible in graphene.
Mathematics and Department of Physics, Columbia University, dicted to display unusual correlated phases of The sample used in this study was the host of
New York, NY, USA. 6Department of Electrical Engineering, matter such as topological Mott insulating (7) a 2DEG in a 25-nm-wide, one-side modulation-
Princeton University, Princeton, NJ, USA.
and quantum spin liquid phases (8). doped Al0.1Ga0.9As/GaAs quantum well. The pro-
*Present address: Department of Physics, Bose Institute, 93/1
Acharya Prafulla Chandra Road, Kolkata 700009, India.
The creation of artificial systems with a cedures for nanofabricating the artificial lattice
†To whom correspondence should be addressed. E-mail: high degree of tunability offers a way to explore are detailed in (11, 19, 20). The artificial honey-
[email protected] (V.P.); [email protected] (M.P.) Mott-Hubbard (M-H) physics systematically comb lattice extended over a 100-mm–by–100-mm

1176 3 JUNE 2011 VOL 332 SCIENCE www.sciencemag.org

 Spin-Liquid Ground State of the S = 1/2 Kagome Heisenberg Antiferromagnet
Simeng Yan, David A. Huse, and Steven R. White

Science, 332 (6034), .

DOI: 10.1126/science.1201080

Downloaded from https://www.science.org at Indian Institute of Technology Bombay on June 15, 2023
View the article online
https://www.science.org/doi/10.1126/science.1201080
Permissions
https://www.science.org/help/reprints-and-permissions

Use of this article is subject to the Terms of service

Science (ISSN 1095-9203) is published by the American Association for the Advancement of Science. 1200 New York Avenue NW,
Washington, DC 20005. The title Science is a registered trademark of AAAS.
Copyright © 2011, American Association for the Advancement of Science