WMD-group · aronwalsh · Aug 25, 2026 · Aug 25, 2026 · Aug 25, 2026 · Aug 25, 2026
diff --git a/AGENT.md b/AGENT.md
@@ -39,7 +39,7 @@ C(T) = (V · 2π/ℏ) · g · W² · Σᵢⱼ pᵢ |⟨χᵢ|Q-Q₀|χⱼ⟩|²
 Where:
 - `V`: supercell volume (cm³)
 - `g`: degeneracy factor (spin/orbital)
-- `W`: electron-phonon coupling matrix element (eV)
+- `W`: electron-phonon coupling matrix element (eV/(amu^0.5·Å))
 - `pᵢ`: Boltzmann occupation of initial vibrational state `i`
 - `χᵢ, χⱼ`: vibrational wavefunctions (eigenfunctions of 1D Schrödinger equation)
 - `δ(εᵢ - εⱼ)`: energy-conserving delta function (Gaussian-broadened)
@@ -88,7 +88,7 @@ CarrierCapture.py/
 │   └── cli/                     # Command-line interface
 │       ├── main.py              # Click CLI entry point
 │       └── commands/            # Subcommands (fit, solve, capture, scan, viz)
-├── tests/                       # 88 tests (pytest)
+├── tests/                       # 169 tests (pytest)
 ├── examples/
 │   ├── notebooks/               # Jupyter tutorials
 │   └── data/                    # Example DFT data
@@ -142,8 +142,8 @@ Manages two-state capture calculation.
 ```python
 pot_i: Potential           # Initial state (excited)
 pot_f: Potential           # Final state (ground)
-W: float                   # Electron-phonon coupling (eV)
-g: int                     # Degeneracy factor
+W: float                   # Electron-phonon coupling (eV/(amu^0.5·Å))
+degeneracy: int            # Degeneracy factor
 overlap_matrix: ndarray    # ⟨χᵢ|Q|χⱼ⟩
 capture_coefficient: ndarray  # C(T) in cm³/s
 ```
@@ -169,20 +169,21 @@ cc.capture_coefficient  # Array of C(T) values
 
 High-throughput parameter scanning.
 
-**Key Functions**:
+**Key Classes**:
 ```python
-from carriercapture.analysis.parameter_scan import scan_parameters
+from carriercapture.analysis.parameter_scan import ParameterScanner, ScanParameters
 
-results = scan_parameters(
+params = ScanParameters(
     dQ_range=(0, 25, 25),      # (min, max, n_points)
     dE_range=(0, 2.5, 10),
-    hw_i=0.008,
-    hw_f=0.008,
-    W=0.068,
+    hbar_omega_i=0.008,
+    hbar_omega_f=0.008,
+    W=0.068,                   # eV/(amu^0.5·Å), required
     volume=1e-21,
     temperature=300.0,
-    n_jobs=-1                   # Parallel execution
 )
+scanner = ParameterScanner(params)
+results = scanner.run_harmonic_scan(n_jobs=-1)  # Parallel execution
 
 # Access results
 results.capture_coefficients  # 2D array [dQ, dE]
@@ -333,13 +334,16 @@ def solve(
 **Structure**:
 ```
 tests/
-├── test_potential.py         # Potential class tests
-├── test_schrodinger.py       # Solver validation (analytical solutions)
-├── test_config_coord.py      # Capture coefficient workflows
-├── test_parameter_scan.py    # High-throughput scanning
-├── test_visualization.py     # Plotting functions
-├── test_io.py                # File I/O
-└── test_cli.py               # Command-line interface
+├── test_potential.py            # Potential class tests
+├── test_schrodinger.py          # Solver validation (analytical solutions)
+├── test_config_coord.py         # Capture coefficient workflows
+├── test_transfer_coord.py       # Marcus theory (TransferCoordinate)
+├── test_sommerfeld.py           # Sommerfeld factor
+├── test_parameter_scan.py       # High-throughput scanning
+├── test_advanced_fitting.py     # Potential fitting methods
+├── test_visualization.py        # Plotting functions
+├── test_interactive_dashboard.py# Dash dashboard
+└── test_doped_integration.py    # doped interface
 ```
 
 **Run tests**:
@@ -351,7 +355,7 @@ pytest tests/ -x --pdb                           # Stop on first failure, debug
 ```
 
 **Current Status** (2026-01-18):
-- 88 tests passing
+- 169 tests (doped-integration tests skip without the optional doped package)
 - Core modules: >90% coverage
 - Python 3.9-3.12 supported
 
@@ -436,7 +440,7 @@ OCC_CUTOFF = 1e-5            # Max occupation for partition function convergence
 hw = 0.008 eV           # 8 meV phonon
 dQ = 10.5               # amu^0.5·Å shift
 dE = 0.5 eV             # Energy difference
-W = 0.068 eV            # Electron-phonon coupling
+W = 0.068 eV/(amu^0.5·Å)   # Electron-phonon coupling
 volume = 1e-21 cm³      # Supercell volume
 temperature = 300 K     # Room temperature
 nev_initial = 180       # Initial state eigenvalues
@@ -508,12 +512,16 @@ E_n = ℏω * (n + 1/2)
 
 **File**: `benchmarks/benchmark_sn_zn.py`
 
-**Results** (vs CarrierCapture.jl):
-- Initial eigenvalues: 0.005% difference ✓
-- Final eigenvalues: 0.02% difference ✓
-- Capture coefficient (300K): 1.5% difference ✓
+**Results** (three-tier comparison vs CarrierCapture.jl):
+- Tier 1: with Julia's grid/integration conventions emulated, eigenvalues and
+  C(300K) match the Julia reference to ~1e-12 (algorithmic equivalence) ✓
+- Tier 2: native eigenvalues match analytic ℏω(n+½) to ≤1.5e-4 ✓
+- Tier 3: native C(300K) differs from Julia by ~1.5%, entirely due to
+  CarrierCapture.jl's finite-difference spacing (ΔQ/N vs ΔQ/(N−1)) and
+  rectangle-rule overlaps; Python's native numerics are the more accurate ✓
 
-**Conclusion**: Python matches Julia within ~1-2% (floating-point precision)
+**Conclusion**: identical physics; the small native offset is a Julia grid
+convention, not floating-point noise.
 
 ---
 
@@ -696,7 +704,7 @@ pot.eigenvectors       # Array of χₙ(Q), shape: (nev, len(Q))
 ### Capture Calculation
 
 ```python
-cc = ConfigCoordinate(pot_i, pot_f, W=0.068, g=1)
+cc = ConfigCoordinate(pot_i, pot_f, W=0.068, degeneracy=1)
 cc.calculate_overlap(Q0=5.0, sigma=0.025, cutoff=0.25)
 cc.calculate_capture_coefficient(
     volume=1e-21,

diff --git a/README.md b/README.md
@@ -44,7 +44,7 @@ CarrierCapture.py started life as an automated rewrite of [CarrierCapture.jl](ht
 
 ### 🔬 Scientific Validation
 - Validated against CarrierCapture.jl
-- Comprehensive test suite (88 tests)
+- Comprehensive test suite (169 tests)
 - Tutorial notebooks with real examples
 
 ---
@@ -65,14 +65,14 @@ pip install -e ".[dev]"
 
 ### Optional Dependencies
 ```bash
-# Interactive dashboard
-pip install carriercapture[viz]
-
 # doped integration (for defect calculations)
 pip install carriercapture[doped]
 
-# All extras (recommended for development)
-pip install -e ".[all]"
+# Jupyter notebook support
+pip install carriercapture[notebook]
+
+# Development tools (tests, linting)
+pip install -e ".[dev]"
 ```
 
 ---
@@ -118,7 +118,7 @@ carriercapture solve excited.json -n 180 -O excited_solved.json
 carriercapture capture config.yaml -V 1e-21 --temp-range 100 500 50
 
 # High-throughput parameter scan
-carriercapture scan --dQ-min 0 --dQ-max 25 --dQ-points 25 \
+carriercapture scan --dQ-min 0 --dQ-max 25 --dQ-points 25 -W 0.05 \
                     --dE-min 0 --dE-max 2.5 --dE-points 10 \
                     -j -1 -o scan_results.npz
 
@@ -159,7 +159,9 @@ pot_final.solve(nev=60)
 ### Tutorial Notebooks
 
 - **[01_harmonic_sn_zn.ipynb](examples/notebooks/01_harmonic_sn_zn.ipynb)** - Basic workflow with harmonic oscillators
+- **[02_anharmonic_dx_center.ipynb](examples/notebooks/02_anharmonic_dx_center.ipynb)** - Anharmonic potentials (DX center)
 - **[03_parameter_scan.ipynb](examples/notebooks/03_parameter_scan.ipynb)** - High-throughput screening
+- **[04_interactive_viz.ipynb](examples/notebooks/04_interactive_viz.ipynb)** - Interactive visualization
 
 Full examples in [`examples/`](examples/) directory with detailed [README](examples/README.md).
 
@@ -195,7 +197,7 @@ $$C(T) = \frac{V \cdot 2\pi}{\hbar} \cdot g \cdot W^2 \cdot \sum_{i,j} p_i |\lan
 Where:
 - `V`: supercell volume
 - `g`: degeneracy factor  
-- `W`: electron-phonon coupling matrix element
+- `W`: electron-phonon coupling matrix element (eV/(amu^0.5·Å))
 - `pᵢ`: thermal occupation of initial state `i`
 - `χᵢ, χⱼ`: vibrational wavefunctions
 - `δ`: energy-conserving delta function (Gaussian broadened)
@@ -222,9 +224,9 @@ pytest tests/ --cov=src/carriercapture --cov-report=html
 ```
 
 **Test Statistics:**
-- 88 tests passing (53 Phase 3 tests skipped)
+- 169 tests (doped-integration tests skip without the optional doped package)
 - Core modules: >90% coverage
-- All tests pass on Python 3.11-3.12
+- Supported Python versions: 3.9-3.12
 - CI/CD with GitHub Actions
 
 ---
@@ -255,16 +257,26 @@ CarrierCapture.py has been validated against the original [CarrierCapture.jl](ht
 - Phonon energy: ℏω = 8 meV
 - Configuration coordinate shift: ΔQ = 10.5 amu^0.5·Å
 - Energy offset: ΔE = 0.5 eV
-- Electron-phonon coupling: W = 0.068 eV
-
-**Results**:
-| Observable | Python Value | Julia Value | Relative Diff | Status |
-|------------|--------------|-------------|---------------|--------|
-| Initial eigenvalues (E₀) | 0.504000 eV | 0.504001 eV | 0.005% | ✓ PASS |
-| Final eigenvalues (E₀) | 0.004000 eV | 0.004001 eV | 0.02% | ✓ PASS |
-| Capture coefficient (300K) | 1.339×10⁻¹⁰ cm³/s | 1.359×10⁻¹⁰ cm³/s | 1.5% | ✓ PASS |
-
-**Conclusion**: Python implementation matches Julia results within ~1-2% across all observables. Small differences (~0.01-1.5%) are due to floating-point arithmetic differences between language implementations and are well within acceptable tolerances for physical calculations.
+- Electron-phonon coupling: W = 0.068 eV/(amu^0.5·Å)
+
+**Results** (three-tier comparison):
+
+| Tier | Comparison | Max Relative Diff | Status |
+|------|------------|-------------------|--------|
+| 1 | Julia-convention emulation vs Julia reference (eigenvalues) | 3×10⁻¹² | ✓ PASS |
+| 1 | Julia-convention emulation vs Julia reference (C at 300 K) | 5×10⁻¹² | ✓ PASS |
+| 2 | Native eigenvalues vs analytic ℏω(n+½) | 1.5×10⁻⁴ | ✓ PASS |
+| 3 | Native C(300 K) = 1.339×10⁻¹⁰ vs Julia 1.359×10⁻¹⁰ cm³/s | 1.5×10⁻² | ✓ (informational) |
+
+**Conclusion**: When Python is run with CarrierCapture.jl's numerical
+conventions, the two codes agree to near machine precision (Tier 1) — they
+implement identical physics. The ~1.5% native difference (Tier 3) is entirely
+due to two CarrierCapture.jl conventions: its finite-difference kinetic term
+uses grid spacing ΔQ/N while its grid actually has spacing ΔQ/(N−1), and it
+integrates overlaps with the rectangle rule. Python uses the true grid spacing
+and the trapezoid rule, and its native eigenvalues are closer to the analytic
+harmonic result (Tier 2). Both codes converge to the same answer with
+increasing grid density.
 
 ### Running the Benchmark
 
@@ -340,7 +352,7 @@ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file
 | Parameter Scanning | ✅ Complete |
 | doped Integration | ✅ Complete |
 | Documentation | ✅ Complete |
-| Test Coverage | ✅ 88 tests |
+| Test Coverage | ✅ 169 tests |
 | PyPI Release | 🔄 Planned |
 
 ---