Open-source optimization toolkit for carbon capture and artificial photosynthesis systems.
Carbon capture costs between $500-1,300/ton for direct air capture and $40-120/ton for point-source industrial capture. These costs are too high for most of the world to adopt at scale.
The core problem is engineering complexity: a carbon capture system has 5-7 efficiency stages chained together (light harvesting, charge separation, electron transport, water splitting, CO2 fixation...). Each stage multiplies against the others, so a small improvement in one step has outsized effects on total system output. But today, finding which step to improve requires either:
- $20,000-100,000/year simulation tools (Aspen Plus, COMSOL) that need PhD-level expertise
- Months of trial-and-error lab work iterating on material compositions
Researchers, startups, and engineering teams need a way to identify bottlenecks instantly and rank materials quantitatively without enterprise simulation software.
Phi-Engine is a Python SDK that converts multi-step efficiency cascades into a simple additive space (D-space) where:
- Bottlenecks become obvious — the step with the largest D-value is always the bottleneck
- Energy conservation is guaranteed — every computation is validated against a proven algebraic law
- Materials are scored and ranked — 8 MOF (Metal-Organic Framework) materials scored by capacity, selectivity, cost, and abundance
- Full systems are modelled — cascade efficiency + MOF capture + quantum coherence + environmental factors, all in one call
You feed it numbers, it tells you what to improve and by how much.
Primary users:
- Carbon capture researchers and R&D engineers
- Cleantech startups designing CO2 capture or solar fuel systems
- Materials scientists evaluating MOF and catalyst candidates
- University labs teaching photosynthesis or electrochemistry
- Process engineers optimizing multi-step chemical systems
Market sizing:
| Segment | Size (2025) | Size (2030) | Growth |
|---|---|---|---|
| TAM — Global CCUS market | $5.8B | $17.8B | 25% CAGR |
| SAM — Carbon capture software + simulation tools | $800M | $2.4B | 24% CAGR |
| SOM — Open-source optimization tools for SMB/research | $50M | $200M | 32% CAGR |
The CCUS market is projected to reach $17.8B by 2030 (MarketsandMarkets). Carbon dioxide removal alone is forecast to grow from $2B today to $50B by 2030 (GlobeNewsWire). Investment in carbon capture tripled since 2022 to $6.4B in 2024.
1. Find the bottleneck in your photosynthesis system You built an artificial leaf prototype and measured each stage's efficiency. Feed the numbers to Phi-Engine. In one call, it tells you that carbon fixation (RuBisCO) consumes 40% of your total losses and that improving it from 45% to 63% would hit your 20% overall target.
2. Pick the right MOF material for your CO2 filter You have 8 candidate MOF materials. You need abundant elements only (no rare earths), water stability, and CO2/N2 selectivity above 50x. Phi-Engine scores all 8, applies your constraints, and ranks them — with a key insight: a pore diameter of 0.363 nm (PHI-optimal) sits exactly between CO2 (0.330 nm) and N2 (0.364 nm), creating a natural molecular sieve.
3. Model a complete capture-and-convert system Combine your photosynthesis cascade with a MOF pre-filter and quantum coherence parameters. Phi-Engine returns a single number: kg of CO2 captured per square meter per day, with a breakdown of which subsystem (cascade, MOF, coherence, temperature, CO2 supply) is the weakest link.
4. Calibrate laboratory instruments Upload sensor readings against a known reference. D-space linearises drift patterns and detects systematic bias that raw statistics miss.
5. Teach D-space analysis Interactive Streamlit demo with 5 tabs and Jupyter notebooks. Students manipulate sliders and see how efficiency changes propagate through a cascade in real time.
Phi-Engine is built on one mathematical insight:
D(x) = -ln(x) / ln(PHI) where PHI = (1 + sqrt(5)) / 2
This maps any efficiency value (0-1) to a positive number on a logarithmic scale. The key property:
D(eta_1 * eta_2 * ... * eta_n) = D(eta_1) + D(eta_2) + ... + D(eta_n)
Multiplicative cascades become additive sums. The step with the largest D-value is always the bottleneck. This is mathematically guaranteed, not a heuristic.
Input data D-Space Analysis Output
----------- ---------------- ------
Step efficiencies --> D-transform each step --> Bottleneck identification
[0.95, 0.99, 0.85, [0.11, 0.02, 0.34, "carbon_fixation" (D=1.66)
0.80, 0.66, 0.45, 0.46, 0.86, 1.66, = 40% of total loss
0.72] 0.68] Target: raise to 0.63
MOF materials --> Score + rank by --> Ranked candidates
capacity, selectivity, adapted integrity + PHI pore analysis
cost, abundance formula + constraint filtering
Full system --> Cascade * MOF * coherence --> kg CO2/m2/day
all parameters * temp * CO2 corrections + system bottleneck
The SDK has zero core dependencies — only Python's stdlib math module. Optional extras add FastAPI, Streamlit, and Jupyter support.
git clone https://github.com/Cloudhabil/phi-engine.git
cd phi-engine
pip install ".[dev]"from phi_engine import PhiEngine
from phi_engine.adapters.photosynthesis import PhotosynthesisAdapter
from phi_engine.photosynthesis_constants import NATURAL_STEPS
engine = PhiEngine()
engine.register_adapter("photosynthesis", PhotosynthesisAdapter())
# Find the bottleneck in plant photosynthesis
result = engine.run("photosynthesis", {
"mode": "cascade",
"steps": NATURAL_STEPS,
"target_efficiency": 0.20,
})
print(result["bottleneck"]["step_name"]) # "carbon_fixation"
print(result["bottleneck"]["d_value"]) # 1.659 (40% of total D)
print(result["recommendations"][0]) # "Consider engineered RuBisCO..."pip install ".[demo]"
streamlit run demo/app.pycd demo && docker-compose up --build
# Streamlit: http://localhost:8501
# API: http://localhost:8200pip install ".[api]"
phi-engine # http://localhost:8200pip install ".[notebooks]"
jupyter notebook notebooks/Phi-Engine is free and open source (MIT license).
| Tier | Price | Includes |
|---|---|---|
| Open Source | $0 | Full SDK, all adapters, API server, Streamlit demo, notebooks |
| Hosted Demo (coming soon) | $0 free tier / $49/mo pro | Cloud-hosted Streamlit app, 100 analyses/day free |
| Enterprise Support (coming soon) | Custom | Priority support, custom adapters, on-premise deployment |
What you need:
- Python 3.10+ (any OS)
pip install .(zero dependencies for core)- 5 minutes to run first analysis
What you do NOT need:
- No GPU, no cloud, no Docker (optional)
- No license keys
- No training — the Streamlit demo is self-explanatory
- No data migration — feed it numbers, get results
Switching cost from current tools:
| From | Effort | Time |
|---|---|---|
| Manual spreadsheets | Copy efficiency values into a Python dict | 10 minutes |
| Aspen Plus / COMSOL | Export stage efficiencies, run Phi-Engine alongside | 1 hour |
| No existing tool | Start with NATURAL_STEPS defaults, adjust | 5 minutes |
You do not need to replace your existing simulation tools. Phi-Engine runs alongside them as a fast bottleneck finder and material ranker.
For a research lab spending $20K-100K/year on Aspen Plus or COMSOL:
- Phi-Engine replaces the bottleneck analysis step (typically 20-30% of simulation time)
- Saves $4K-30K/year in license cost for this specific workflow
- Saves weeks of iteration by identifying the right step to improve before running expensive simulations
For a carbon capture startup at pilot scale (1-30 tons CO2/day):
- A 5% efficiency improvement at the bottleneck step translates to $2-6/ton cost reduction
- At 10 tons/day, 365 days: 3,650 tons/year = $7,300-21,900/year savings from one optimization cycle
- Cost of Phi-Engine: $0
For a MOF materials group evaluating candidates:
- Screening 8 MOFs manually (synthesis + testing): ~6 months, $50K-200K
- Pre-screening with Phi-Engine to narrow to 2-3 candidates: 5 minutes, $0
- Saves 50-70% of screening cost by eliminating poor candidates before synthesis
| Tool | Price/yr | Open Source | Carbon Capture Focus | Expertise Required |
|---|---|---|---|---|
| Aspen Plus (AspenTech) | $20K-100K | No | General process simulation | PhD-level |
| COMSOL Multiphysics | $3,500-10K | No | General multiphysics | PhD-level |
| CCSI2 (DOE/NETL) | Free | Yes | Carbon capture | Expert |
| SimCCSPRO (Carbon Solutions) | Custom | No | CCS infrastructure routing | Specialist |
| KBC Petro-SIM | Enterprise | No | Oil & gas decarbonization | Specialist |
| OLI Systems | Enterprise | No | Process chemistry | Specialist |
| Phi-Engine | Free | Yes | Photosynthesis + MOF + system | Any engineer |
How Phi-Engine differentiates:
- Zero cost, zero dependencies — Aspen Plus costs $20K+/year and requires weeks of training. Phi-Engine is
pip installand 5 minutes. - Bottleneck-first approach — Other tools simulate the full system and leave you to interpret results. Phi-Engine directly tells you which step to improve and by how much.
- MOF material ranking — No other open-source tool scores MOF candidates against cost, abundance, self-healing, and PHI-optimal pore geometry in one call.
- Algebraic guarantees — D-space analysis is mathematically proven (sum-rule conservation), not a numerical approximation that can diverge.
- Interactive demo included — Streamlit UI + Jupyter notebooks ship with the SDK. Competitors require separate visualization tools.
Companies implementing similar solutions:
- Twelve (formerly Opus 12) — CO2-to-chemicals electrochemistry, raised $645M
- Svante — MOF-based carbon capture, pilot scale (1-30 tons/day)
- Climeworks — DAC at $1,000/ton, building megaton-scale plants
- Nuada / Captivate Technology — MOF modular capture systems
These are hardware companies. They need software tools to optimize their systems. Phi-Engine is the optimization layer that sits on top of their engineering workflow.
| Method | Path | Description |
|---|---|---|
| GET | /health |
Health check |
| POST | /transform |
D-space transform |
| POST | /analyze |
Run adapter analysis |
| GET | /constants |
Browse 53 physics constants |
| GET | /ladder |
PHI-power energy ladder |
| POST | /validate |
Sum-rule validation |
| POST | /decompose |
Fibonacci decomposition |
phi_engine/
core.py # D(x), Theta(x), Energy(x), PHI constants
engine.py # PhiEngine facade (transform, validate, run)
analyzer.py # Sum-rule, Fibonacci decomposition, consistency
ladder.py # PHI-power energy ladder
constants_db.py # 53 physics constants predicted from PHI
photosynthesis_constants.py # Photon yields, 8 MOF materials, helper functions
adapters/
base.py # BaseAdapter ABC
calibration.py # Instrument calibration adapter
sensor_fusion.py # Multi-sensor fusion adapter
photosynthesis.py # Cascade + MOF + coherence (3 modes)
api/
server.py # FastAPI (port 8200)
schemas.py # Pydantic request/response models
demo/
app.py # Streamlit (5 tabs)
database.py # SQLite persistence
payment.py # Stripe integration
webhook.py # Stripe webhooks
notebooks/
01_getting_started.ipynb # D-space basics
02_calibration_demo.ipynb # Calibration workflow
tests/
test_core.py # 9 core tests
test_photosynthesis.py # 16 adapter tests
PHI = (1 + sqrt(5)) / 2 = 1.6180339887498949
D(x) = -ln(x) / ln(PHI) Dimension from value
x(D) = 1 / PHI^D Value from dimension
E(x) = PHI^D(x) * 2*pi*x = 2*pi Energy conservation (proven)
MIT -- free for commercial and research use.