Morning Session: Application Challenges

Talks

• Chair: Birgitta Whaley
• Speakers
  • Synge Todo
    • QMC challenges
      • Criticality with large correction to scaling
      • Supersolid: coexistence of diagonal or off-diagonal order (done on lattice)
      • Deconfined criticality
      • Long-range interacting systems
      • Strongly anisotropic systems
      • Bond-alternating spin chains (need good order parameter)
    • Go beyond memory barrier
    • Can one detect berry phases with QMC?
  • Tomotoshi Nishino
    • Classical lattice models
      • Low temperatures
      • Non-Hermitian problems (TMRG): violation of variational principle
        • Examples of such problems?
          • TM of vertex models are non-symmetric
      • Constraint models: constraints increase number of states kept
  • Didier Poilblanc
    • Generalized QDM for frustrated magnets
    • Only for Abelian particles
    • How does this expansion compare to CORE?
      • CORE is not exact
      • CORE breaks translational symmetry
    • How does one parallelize ED on distributed memory?
      •  ??? -> Andreas Laeuchli
  • Anatoly Kuklov
    • Free-space supersolidity
    • Ideal supersolid or relation to defects?
    • Shear-module
    • How to make people take ab-initio simulations be taken seriously?
    • Quantum mechanics of superfluid living on dynamical extended defects network: quantum metallurgy
    • Role of He3 in superfluid and mechanical properties of solid He4
    • Flow problems?
  • Valentin Murg
  • Guifre Vidal
    • Understand entanglement scaling
  • Bela Bauer
  • Adrien Feiguin
    • I/O constraints
    • Use of symmetries
    • DMRG in momentum space
  • Matthias Troyer
    • Band gap problem of silicon
      • Gap is completely wrong in LDA
      • GW approximation
    • What takes a cluster for the Hubbard model may take petaflop for material
  • Ian McCulloch
    • 1d challenges: parallelization
    • Current limitation: 10000 states (with good symmetries)
    • One million: maybe easier without quantum numbers
    • Most pure 1d problems are understood, except time evolution
    • Master equation evolution? Dissipation?
    • t-J on wide ladder - revisit with petascale?
      • Non-striped states out there?
      • Enforce stripelessness with translational invariance
      • Stripe in correlations
    • Update tensor sequentially? Independently?
    • DMRG to QFT?
      • QED in 2+1?
      • String is 1d object
    • Quantum Chemistry
      • Symmetry aspects: total spin excitations
    • Nanotubes
    • Nuclear shell models
      • Big local dimension
      • Complicated Hamiltonian

Discussion

• Which methods can scale to petaflop machines?
  • DMFT can (with disorder)
  • Loop algorithm can
  • DiagMC could
  • Worm for helium: much harder
  • Lanczos can
    • For QDM etc.: keep matrix in memory
  • DMRG: similar to Lanczos, more sceptical
    • Latency problems
    • I/O problems
    • Forget storing matrices on disk (2 GB/core)
    • Shared memory is easier: now about 64 cores, will not get beyond that
    • Heisenberg spin chain on a terascale machine?