This repository contains a minimal, reproducible example for running METTS (Minimally Entangled Typical Thermal States) with ITensors on the 2D Fermi–Hubbard model wrapped as a cylinder. The example computes finite‑temperature observables and writes results to HDF5.

examples/hubbard_cylinder_metts.jl   <-- runnable example script
src/                                <-- package code
outfiles.metts/                     <-- created at runtime; outputs written here

• Julia 1.11 or newer
• Packages installed into this project: ITensors, ITensorMPS, HDF5

Note: Do not add METTS from the registry when working inside this repo. Activating this project makes using METTS work.

# 1) Enter the project
cd /path/to/METTS.jl

# 2) Instantiate and add required packages *into this project*
julia --project -e 'using Pkg; Pkg.instantiate(); Pkg.add(["ITensors","ITensorMPS","HDF5"])'

# 3) Run an example (L W U filling t T tau maxD seed nmetts Nwarm)
julia --project examples/hubbard_cylinder_metts.jl   24 4 10.0 0.9375 1.0 0.10 0.3 2000 1 100 5

# 4) Find produced HDF5 outputs
find outfiles.metts -maxdepth 8 -name outfile.h5

Notes

• Measurements are written only after warm‑up.
• Ensure nmetts > Nwarm.

Default output root: ./outfiles.metts (or the directory set by METTS_OUTDIR).

Each run is saved under:

outfiles.metts/L.<L>.W.<W>/U.<U>/filling.<f>/T.<T>/D.<maxD>/tau.<tau>.seed.<seed>/

Within a run directory you will find:

• outfile.h5 — parameters under meta/; groups step_xxxxx/ for each measured step
  (datasets include: energy, Ntotal, DO, SZSZ, SPSM, SMSP, NtotNtot, SS).
• samples.txt — product‑basis states per measured step. Codes 1..4 map to Emp, Up, Dn, UpDn.

Resuming a run is automatic if samples.txt already exists.

Sanity check:

julia --project examples/hubbard_cylinder_metts.jl   4 4 10.0 0.9375 1.0 0.10 0.3 100 1 10 5

Paper‑grade template:

julia --project examples/hubbard_cylinder_metts.jl   32 4 10.0 0.9375 1.0 0.10 0.2 2000 1 2000 100

• ERROR: Package ITensors not found
  Use the project environment: run with julia --project (or Pkg.activate(".")).

• No measured data appears
  Set nmetts > Nwarm so that measurements are taken after warm‑up.

• Slow / memory‑heavy
  Decrease maxD or increase tau slightly.

• A short DMRG warm‑up avoids long initial transients.
• Choose tau and maxD carefully; typical tau = 0.1–0.3.