JSON Configuration File

Overview

Most aspects of the analysis algorithm can be configured through a JSON file. This allows users to change the target elements, neighbor-selection rules, bond parameters, angle parameters, and FNC types without modifying the program.

The current configuration targets SiO₂, but the design allows adaptation to other materials by changing JSON settings.

Configuration Flow

POSCAR
  ↓
JSON
  ↓
QnNetworkBuilder
  ↓
QnNetworkAnalysis
  ↓
RmcConstraintWriter

The builder passes JSON configuration to the relevant classes. The details of the analysis are determined by the configuration.

network

The network configuration defines the Central → Bridge → Second network through central_types, bridge_types, and second_types.

For SiO₂:

Central = Si
Bridge = O
Second = Si

These three atomic categories are the targets of network analysis.

Neighbor Search

Neighbor searches are configured separately for first and second.

First

The first settings control the Central → Bridge search, including the search method, minimum distance, maximum distance, and number of nearest neighbors.

Second

The second settings control the Bridge → Second search, including minimum and maximum distances.

second_use_first_network

This is an important configuration option.

false: perform a new distance-based Bridge → Second search.
true: reverse the previously constructed Central → Bridge network to identify Second atoms.

For SiO₂, true provides better consistency between the two sets of connections.

Bond Parameters

Bond settings for Topology generally specify atom types, equilibrium distances, and force constants. The writer exports these settings to [ bonds ].

Angle Parameters

Angle settings generally specify angle types, equilibrium angles, and force constants. The writer exports these settings to [ angles ].

Atomic Information

The [ atoms ] section requires properties such as atom type, mass, and charge. These are also specified in JSON. The current implementation uses settings for Si and O.

FNC Types

FNC types map identifiers to distance ranges. For example:

Type 1 → 1.55–1.75 Å

The writer converts these settings into a type_id → (rmin, rmax) mapping and generates the FNC header.

Role of JSON

Both the analysis and the writer use JSON, but they consume different settings.

The analysis uses neighbor-search rules, Qn classification settings, and atomic species. The writer uses bond and angle parameters, masses, charges, and FNC distance ranges.

SiO₂-Specific Settings

The current JSON configuration assumes:

Central = Si
Bridge = O
Second = Si
nearest = 4

This constructs SiO₄ tetrahedra. O–Si–O angle settings likewise assume tetrahedral geometry. Thus, although the configuration can be changed, its current values are specific to SiO₂.

Applying the Library to Other Materials

The analysis algorithm uses the general Central → Bridge → Second model. It can therefore be adapted relatively easily to bridging-oxygen networks such as:

P-O-P
B-O-B
Ge-O-Ge
Al-O-Si

Typically, the settings that need adjustment are atomic species, neighbor distances, bond parameters, angle parameters, and FNC ranges.

Materials That Are Difficult to Handle

For metallic glasses containing Fe, Ni, or Cu, and for some liquids or simple crystals, the Central → Bridge → Second model may not be meaningful.

Changing JSON alone is insufficient for such systems; the analysis algorithm in QnNetworkAnalysis must be extended.

Design Philosophy

The design is based on algorithm + JSON, not an algorithm wholly defined by JSON. JSON controls elements, distances, and parameters, while the program implements the Central–Bridge–Second network model.

This gives the library substantial reusability for SiO₂ and related network materials, but materials with fundamentally different network concepts require algorithmic extensions.

The next chapter examines how far the current SiO₂ implementation can be extended to other materials and discusses possible future development.