ALIGNN Predictor¶

Predict 15+ materials properties simultaneously using ALIGNN (Atomistic Line Graph Neural Network) pretrained models. Input by JARVIS ID or POSCAR structure. Properties span energy/stability, electronic band gaps, mechanical moduli, dielectric/piezoelectric constants, topological spillage, SLME, magnetic moment, superconducting Tc, and interface band offsets.

Open App

Overview¶

The ALIGNN Predictor runs 15 pretrained ALIGNN models in one call to predict a comprehensive set of materials properties from crystal structure alone. Max 50 atoms per structure. Three endpoints are available: GET (API key auth), POST (session auth for web UI), and web_predict (structured JSON response with all 15 properties organized by category).

Data Source

ALIGNN pretrained models — 15+ models from alignn.pretrained, trained on JARVIS-DFT data.

Predicted Properties¶

Category	Property	Unit	Model Key
Energy	Formation energy	eV/atom	`jv_formation_energy_peratom_alignn`
Energy	Total energy (OptB88vdW)	eV/atom	`jv_optb88vdw_total_energy_alignn`
Energy	Energy above hull	eV/atom	`jv_ehull_alignn`
Electronic	Band gap (OptB88vdW)	eV	`jv_optb88vdw_bandgap_alignn`
Electronic	Band gap (mBJ)	eV	`jv_mbj_bandgap_alignn`
Mechanical	Bulk modulus (Kv)	GPa	`jv_bulk_modulus_kv_alignn`
Mechanical	Shear modulus (Gv)	GPa	`jv_shear_modulus_gv_alignn`
Dielectric	Dielectric constant (εx)	—	`jv_epsx_alignn`
Piezoelectric	Piezoelectric constant	—	`jv_dfpt_piezo_max_dielectric_alignn`
Topological	SOC spillage	—	`jv_spillage_alignn`
Optical	SLME	%	`jv_slme_alignn`
Magnetic	Magnetic moment	μB	`jv_magmom_oszicar_alignn`
Superconducting	Tc	K	`jv_supercon_tc_alignn`
Interface	CBM (InterMat)	eV	`intermat_cbm`
Interface	VBM (InterMat)	eV	`intermat_vbm`

Endpoints¶

`GET /alignn/query` — Predict by JARVIS ID or POSCAR (API key)¶

curl "https://atomgpt.org/alignn/query?jid=JVASP-1002&APIKEY=sk-XYZ" \
  -H "accept: application/json"

With POSCAR string (URL-encoded):

curl "https://atomgpt.org/alignn/query?poscar=Si%0A1.0%0A0%202.734%202.734%0A2.734%200%202.734%0A2.734%202.734%200%0ASi%0A2%0Adirect%0A0%200%200%0A0.25%200.25%200.25&APIKEY=sk-XYZ" \
  -H "accept: application/json"

Param	Description
`jid`	JARVIS ID (e.g. `JVASP-1002`) — provide `jid` OR `poscar`
`poscar`	URL-encoded POSCAR string
`APIKEY`	API key (query parameter auth)

Returns raw prediction dict with all model keys.

`POST /alignn/query` — Predict from file or string (session auth)¶

Upload a POSCAR file:

curl -X POST "https://atomgpt.org/alignn/query" \
  -H "Authorization: Bearer sk-XYZ" \
  -F "file=@POSCAR"

Or submit as form string:

curl -X POST "https://atomgpt.org/alignn/query" \
  -H "Authorization: Bearer sk-XYZ" \
  -F "poscar_string=Si
1.0
0 2.734 2.734
2.734 0 2.734
2.734 2.734 0
Si
2
direct
0 0 0
0.25 0.25 0.25"

`POST /alignn/web_predict` — Structured predictions (JSON body)¶

Returns all 15 properties organized in a clean structure with formula and atom count.

curl -X POST "https://atomgpt.org/alignn/web_predict" \
  -H "Authorization: Bearer sk-XYZ" \
  -H "Content-Type: application/json" \
  -H "accept: application/json" \
  -d '{
    "poscar": "Si\n1.0\n0 2.734 2.734\n2.734 0 2.734\n2.734 2.734 0\nSi\n2\ndirect\n0 0 0\n0.25 0.25 0.25"
  }'

Response:

{
  "formula": "Si",
  "num_atoms": 2,
  "properties": {
    "formation_energy": -0.005,
    "total_energy": -5.412,
    "ehull": 0.0,
    "bandgap_optb88": 0.611,
    "bandgap_mbj": 1.156,
    "bulk_modulus": 88.9,
    "shear_modulus": 51.5,
    "epsx": 12.3,
    "piezoelectric": 0.0,
    "spillage": 0.32,
    "slme": 5.2,
    "magmom": 0.0,
    "supercon_tc": 0.0,
    "intermat_cbm": -4.1,
    "intermat_vbm": -5.3
  }
}

Python Examples¶

Query by JIDWeb predict (structured)Upload POSCAR file

import requests

response = requests.get(
    "https://atomgpt.org/alignn/query",
    headers={"accept": "application/json"},
    params={"jid": "JVASP-1002", "APIKEY": "sk-XYZ"},
)
data = response.json()
print(f"Band gap (OptB88): {data['jv_optb88vdw_bandgap_alignn'][0]:.3f} eV")
print(f"Formation energy: {data['jv_formation_energy_peratom_alignn'][0]:.3f} eV/atom")

import requests

response = requests.post(
    "https://atomgpt.org/alignn/web_predict",
    headers={
        "Authorization": "Bearer sk-XYZ",
        "accept": "application/json",
        "Content-Type": "application/json",
    },
    json={"poscar": open("POSCAR").read()},
)
data = response.json()
print(f"{data['formula']} ({data['num_atoms']} atoms)")
p = data["properties"]
print(f"  Formation energy: {p['formation_energy']:.3f} eV/atom")
print(f"  Band gap (OPT): {p['bandgap_optb88']:.3f} eV")
print(f"  Band gap (MBJ): {p['bandgap_mbj']:.3f} eV")
print(f"  Bulk modulus: {p['bulk_modulus']:.1f} GPa")
print(f"  Supercon Tc: {p['supercon_tc']:.2f} K")

import requests

with open("POSCAR", "rb") as f:
    response = requests.post(
        "https://atomgpt.org/alignn/query",
        headers={"Authorization": "Bearer sk-XYZ"},
        files={"file": ("POSCAR", f)},
    )
data = response.json()
for key, val in sorted(data.items()):
    if key.endswith("_alignn") or key.startswith("intermat"):
        print(f"  {key}: {val}")

AGAPI Agent¶

from agapi.agents import AGAPIAgent
import os

agent = AGAPIAgent(api_key=os.environ.get("AGAPI_KEY"))

# Predict properties
response = agent.query_sync("Predict all ALIGNN properties for JVASP-1002")
print(response)

# Compare materials
response = agent.query_sync("Compare the band gap of Si and GaAs using ALIGNN")
print(response)

References¶

K. Choudhary, B. DeCost, npj Comp. Mat. 7, 185 (2021) — ALIGNN DOI
K. Choudhary, Comp. Mat. Sci. 259, 114063 (2025) DOI
atomgptlab/alignn