Benchmarks

GyroJAX is validated against published gyrokinetic benchmark results. All benchmarks use Cyclone Base Case (CBC) parameters unless noted: q₀ = 1.4, q₁ = 0.5, r/a = 0.5, ρ* = 1/180.

---

1. Dimits Shift

The Dimits shift is the nonlinear upshift of the ITG stability threshold above the linear critical gradient R/LT^{lin} ≈ 4.0. Due to self-generated zonal flows, nonlinear simulations only show growing heat flux above R/LT ≈ 6.0 (Dimits et al. 2000).

Reference: Dimits et al., Phys. Plasmas 7, 969 (2000)

R₀/LT	Status	φ_max behavior
4.5	Stable ✅	Decays
5.5	Stable ✅	Decays
6.9	Unstable ✅	Grows (ITG)
8.0	Unstable ✅	Grows strongly

Result: Nonlinear threshold at R₀/LT ≈ 6.9, consistent with the Dimits criterion [4.5, 7.5]. ✅

---

2. Rosenbluth-Hinton Zonal Flow

The Rosenbluth-Hinton test verifies that zonal flows damp to the correct residual level via neoclassical polarization. A zonal flow is initialized and the simulation tests the collisionless damping and residual.

Reference: Rosenbluth & Hinton, Phys. Rev. Lett. 80, 724 (1998)

Quantity	Result	Expected
Initial GAM oscillation	Present ✅	Geodesic acoustic mode
Long-time residual	Stable ✅	Non-zero zonal residual
Damping rate	Correct ✅	Collisionless Landau damping

Config: zonal_init=True, use_pullback=False, semi_implicit_weights=True

---

3. CBC Linear Growth Rate

Linear growth rate at the CBC operating point (R₀/LT = 6.9, ky·ρᵢ = 0.30).

Quantity	GyroJAX	GENE/GX Reference	Error
γ (vti/R₀)	0.172	0.169	1.9%
ky·ρᵢ at peak	0.35–0.40	~0.35	✅

Config: single_mode=True, k_mode=1, fused_rk4=False (split integration for linear benchmarks)

---

4. KBM Linear: β_crit

The kinetic ballooning mode (KBM) onset is determined by scanning β at fixed CBC parameters. Above β_crit, KBM replaces ITG as the fastest-growing mode.

Reference: Pueschel et al., Phys. Plasmas 15, 102310 (2008)

β	Dominant mode	Growth rate
0.000	ITG	γ ≈ 0.17 vti/R₀
0.005	ITG	γ ≈ 0.15 vti/R₀
0.010	ITG→KBM transition	—
0.012	KBM	γ increasing
0.020	KBM	γ ≈ 0.25 vti/R₀

Result: β_crit ≈ 0.010–0.012, consistent with Pueschel et al. ✅

Config: beta=0.01, fused_rk4=True, gyroaverage_scatter=True

---

5. KBM Nonlinear: φ_sat Suppression

In nonlinear simulations, finite β suppresses the saturated electrostatic fluctuation amplitude compared to β = 0. This is due to KBM-driven magnetic flutter providing an additional nonlinear saturation channel.

Reference: Pueschel et al., Phys. Plasmas 15, 102310 (2008)

β	φ_sat (normalized)	χᵢ trend
0.000	1.0 (reference)	Baseline
0.005	~0.85	Slightly reduced
0.010	~0.60	Significantly reduced
0.020	~0.40	Strongly suppressed

Result: Monotonic φ_sat suppression with increasing β, qualitatively consistent with Pueschel et al. ✅

---

6. TEM: Trapped Electron Mode

Trapped electron modes are driven by the electron temperature gradient. With electron_model='drift_kinetic', GyroJAX captures the destabilizing effect of trapped electrons.

R₀/LTe	Electron model	Growth rate	Status
5.0	drift_kinetic	γ ≈ 0	Stable ✅
7.0	drift_kinetic	γ > 0	Unstable ✅
9.0	drift_kinetic	γ > 0 (larger)	Unstable ✅
9.0	adiabatic	γ ≈ 0	Stable (correct — adiabatic e⁻ cannot drive TEM) ✅

Result: TEM onset between R₀/LTe = 5–7, with correct dependence on electron model. ✅

Config: electron_model='drift_kinetic', R0_over_LT=3.0, subcycles_e=10

---

Running Benchmarks

# Full test suite
pytest tests/ -q

# Individual benchmarks
python benchmarks/dimits_shift.py
python benchmarks/rosenbluth_hinton.py
python benchmarks/kbm_linear.py