Running the Build¶

From inside the working directory containing 2-bisgma-styrene-thermoset.yaml:

$ htpolynet run -diag diagnostics.log 2-bisgma-styrene-thermoset.yaml &> console.log &

The stage layout under proj-N/systems/ is the same as example 1 — init/, densification/, precure/, iter-K/ (one per CURE iteration), capping/, postcure/, final-results/ — plus plots/ and profile.json at the project root.

Setup¶

The setup stage is where the param-stage reactions actually run. After parameterizing the three primitive species, htpolynet assembles GM1 and then GMA, then enumerates the 16 GMA diastereomers implied by HIE’s stereocenters declaration, and finally builds the 32 chain-expanded cure templates. An abbreviated view of the diagnostic log during this stage:

INFO> Templates in proj-0/molecules/parameterized
INFO> 32 molecules detected in 2-bisgma-styrene-thermoset.yaml
INFO>                       explicit: 11
INFO>     implied by stereochemistry: 21
INFO>            implied by symmetry: 0
INFO> AmberTools> generating GAFF parameters from BPA.mol2
INFO> BPA: 228.28 g/mol
INFO> AmberTools> generating GAFF parameters from HIE.mol2
INFO> HIE: 146.18 g/mol
INFO> AmberTools> generating GAFF parameters from STY.mol2
INFO> STY: 106.16 g/mol
INFO> AmberTools> generating GAFF parameters from GM1.mol2
INFO> GM1: 372.44 g/mol
...
INFO> AmberTools> generating GAFF parameters from GMA.mol2
INFO> GMA: 516.61 g/mol
...
INFO> Built GMAS-1 using topology of GMA; copying GMA.top to GMAS-1.top
INFO> GMAS-1: 516.61 g/mol
...
INFO> 32 molecules implied by chaining
...
INFO> Generated 64 molecule templates
INFO> Initial composition is STY 150, GMA 75
INFO> 100% conversion is 300 bonds

GMA itself is fully GAFF-parameterized; the diastereomers GMAS-1 … GMAS-15 are built (atom positions reflected) but reuse GMA’s topology directly — they’re geometry-only variants.

Densification¶

Eight consecutive 100 ps NPT segments compress the initial 100 kg/m³ liquid up to roughly the polymer’s bulk density. Sample trace from the diagnostic log:

INFO> Initial density: 100.0 kg/m^3
INFO> Initial box side lengths: 9.683 nm x 9.683 nm x 9.683 nm
...
INFO> Repeat 6 out of 8
INFO> Current box side lengths: 4.462 nm x 4.462 nm x 4.462 nm
INFO> Density                      1021.60
...
INFO> Repeat 8 out of 8
INFO> Current box side lengths: 4.465 nm x 4.465 nm x 4.465 nm
INFO> Density                      1020.05

By the end of densification the liquid is at roughly 1.02 g/cm³.

CURE iterations¶

With desired_conversion: 0.95 and a max of 300 bonds, the target is 285 bonds. Early iterations grab dozens of bonds at a time at the default 0.5 nm search radius; as conversion approaches the target, the remaining reactive pairs get sparser and htpolynet has to grow the radius (and sometimes drag pairs in) to find anything. A typical run needs ~15–20 iterations. Excerpted from a representative log:

INFO> ~~~~~~~~~~~~~~ Iteration 1 begins ~~~~~~~~~~~~~~
INFO> Bond search using radius 0.5 nm initiated
INFO> Iteration 1 will generate 81 new bonds
...
INFO> Iteration 1 current conversion 0.270 or 81 bonds
INFO> ~~~~~~~~~~~~~~ Iteration 2 begins ~~~~~~~~~~~~~~
INFO> Bond search using radius 0.5 nm initiated
INFO> Iteration 2 will generate 78 new bonds
...
INFO> Iteration 2 current conversion 0.530 or 159 bonds
...
INFO> ~~~~~~~~~~~~~~ Iteration 15 begins ~~~~~~~~~~~~~~
INFO> Bond search using radius 0.5 nm initiated
INFO> Radius increased to 0.75 nm
INFO> Iteration 15 will generate 1 new bond
...
INFO> Iteration 15 current conversion 0.943 or 283 bonds
INFO> ~~~~~~~~~~~~~~ Iteration 16 begins ~~~~~~~~~~~~~~
INFO> Bond search using radius 0.5 nm initiated
INFO> Radius increased to 0.75 nm
INFO> Iteration 16 will generate 2 new bonds
INFO> Step "cure_drag" initiated on 2 distances (max 0.691 nm)
...
INFO> Iteration 16 current conversion 0.950 or 285 bonds

Tip

This is exactly the workflow CURE.controls.min_bonds_per_iteration was designed to compress. Setting it to e.g. 10 would make htpolynet grow the radius further on the late iterations so each one picks up at least 10 bonds before relaxation. Whether that’s a win depends on whether your relax cascade can absorb the larger strained-bond batches — try it on this example and watch how many iterations are needed.

Capping and postcure¶

After the last cure iteration, htpolynet looks for any unreacted C1–C2 sites:

INFO> Capping begins
INFO> Capping will generate 0 new bonds
INFO> Connect-Update-Relax-Equilibrate (CURE) ends

At 95% conversion in this run, all radicals had paired up — but on a build that hits max_iterations before reaching desired_conversion, expect the cap reactions styCC and hieCC to do non-trivial work.

Postcure runs two more 300/600 K annealing cycles plus a 100 ps NPT at 300 K, then htpolynet writes final.gro, final.top, final.tpx, final.grx plus the VMD viz pair into proj-N/systems/final-results/ and emits the run profile.

Now to look at some results.