---
title: 'r2SCAN Meta-GGA: Enhanced SCAN Functional'
url: https://www.emergentmind.com/topics/r2scan-meta-generalized-gradient-approximation-meta-gga
type: topic
---

# r2SCAN Meta-GGA: Enhanced SCAN Functional

The r²SCAN Meta-Generalized Gradient Approximation (meta-GGA) is a non-empirical, constraint-restoring enhancement of the SCAN (Strongly Constrained and Appropriately Normed) functional in density functional theory. It preserves the formal rigor and transferability of SCAN, improves numerical robustness, and, in its deorbitalized (Laplacian-level) form r²SCAN-L, further enables high-throughput solid-state calculations at computational cost comparable to standard GGAs. r²SCAN has been extensively validated for molecular energetics, transition-metal solid-state properties, spin-crossover energetics, band structure predictions, defect physics, and magnetic systems.

## 1. Mathematical Formulation and Key Ingredients

The r²SCAN meta-GGA functional defines the exchange–correlation (xc) energy as
$$
E_{xc}[n] = \int d^3r\, n(r)\, \epsilon_{xc}^{r^2\mathrm{SCAN}}( n(r), \nabla n(r), \tau(r) )
$$
where:
- $n(r)$ is the electron density,
- $\nabla n(r)$ is its gradient,
- $\tau(r) = \frac{1}{2} \sum_i |\nabla\psi_i(r)|^2$ is the positive-definite kinetic energy density.

Central to its construction are the **reduced density gradient**
$$
s(r) = \frac{|\nabla n|}{2\,(3\pi^2)^{1/3}\, n^{4/3}}
$$
and the **iso-orbital indicator**
$$
\alpha(r) = \frac{\tau(r) - \tau_W(r)}{\tau_U(r)} \,,
$$
with $\tau_W = |\nabla n|^2/(8 n)$ (von Weizsäcker limit) and $\tau_U = (3/10) (3\pi^2)^{2/3} n^{5/3}$ (uniform-gas limit).

In r²SCAN, $\alpha$ is regularized to $\bar\alpha$ to avoid singularities, typically via
$$
\bar\alpha = \frac{\tau - \tau_W}{\tau_U + \eta\,\tau_W}
$$
with a small parameter $\eta \sim 10^{-3}$, ensuring smooth limiting behavior and uniform scaling.

The exchange–correlation enhancement factor takes the form
$$
\epsilon_{xc}^{r^2\mathrm{SCAN}} = \epsilon_x^{\rm LDA}(n) F_x^{r^2\mathrm{SCAN}}(s,\bar\alpha) + \epsilon_c^{\rm LDA}(n) F_c^{r^2\mathrm{SCAN}}(s, \bar\alpha)
$$
where $F_x$ and $F_c$ are constraint-satisfying, regularized polynomials, designed to ensure smoothness and recovery of physical limits. No additional empirical parameters are introduced relative to SCAN [2008.03374, 2110.00647].

## 2. Restoration of Exact Constraints and Numerical Robustness

The principal limitation of SCAN in large-scale or plane-wave DFT calculations is the presence of nonanalytic features (“kinks”) in its switching functions around $\alpha \approx 1$. These features lead to pronounced grid sensitivity and instabilities in self-consistent field (SCF) convergence. rSCAN introduced regularization but sacrificed some exact constraints.

r²SCAN was designed to restore—except for the fourth-order gradient expansion term (GE4X)—all of SCAN’s exact constraints while implementing:
- A smooth, polynomial switching function in the iso-orbital indicator $\bar\alpha$.
- Uniform scaling and recovery of the correct second-order gradient expansion (GE2) in both exchange and correlation.
- Consistent treatment of “norms” including the uniform electron gas, hydrogen atom, and jellium surface energies.

As a result, r²SCAN delivers smooth and grid-stable xc potentials, facilitating convergence in both Gaussian and plane-wave codes, and eliminating pathological SCF behavior [2008.03374, 2009.07808, 2110.00647, 2208.02841].

## 3. Deorbitalization: The r²SCAN-L Functional

Meta-GGAs are, by construction, partially orbital-dependent through τ, yielding a generalized Kohn–Sham (gKS) potential. For practical reasons (chiefly speed and implementation), r²SCAN has been deorbitalized to produce r²SCAN-L, a Laplacian-level meta-GGA:
- τ is replaced by a semilocal kinetic energy density τ_L[n, ∇n, ∇²n], constructed to reproduce τ_U in the uniform limit and τ_W in the iso-orbital limit.
- The iso-orbital indicator becomes $\alpha_L = (\tau_L - \tau_W)/\tau_U$.
- The exchange–correlation energy is then expressed as a pure density functional, delivering a *multiplicative* xc potential.

Recent work further improved the smoothness of the potential in r²SCAN-L via the use of carefully constructed orbital-free kinetic energy deorbitalizers, notably the SRPP and SRPP2 forms [2509.00953].

r²SCAN-L retains nearly all of the transferability of r²SCAN but at GGA-like cost in periodic systems, enabling high-throughput studies.

## 4. Benchmark Performance: Molecular and Solid-State Properties

Systematic benchmarks have established r²SCAN and r²SCAN-L as accurate, stable, and efficient.

### Molecular Validation

- On the G3/99X thermochemistry set: r²SCAN achieves mean absolute errors (MAE) of ≈4.5 kcal/mol, matching or improving on SCAN but with superior grid-insensitivity [2008.03374].
- Heats of formation, bond lengths, and vibrational frequencies: r²SCAN and r²SCAN-L show similar or improved accuracy versus SCAN-L, with r²SCAN-L slightly less accurate but much faster [2008.12420, 2509.00953].

### Solid-State Properties

- Lattice constants: r²SCAN reaches MAE ≈0.037 Å over 55 solids, closely matching SCAN and outperforming PBE and PBEsol [2008.12420, 2208.02841, 2309.12554].
- Cohesive energies: r²SCAN MAE ≈0.20–0.24 eV/atom, with nearly unbiased mean errors (MPE ≈0) [2208.02841, 2309.12554].
- Bulk and shear moduli: r²SCAN provides the most accurate values among semilocal DFAs (MAE ≈6–8 GPa) [2208.02841].
- Thermal properties: Debye temperatures, thermal expansion coefficients, and other thermophysical quantities are predicted with errors comparable to or better than PBEsol [2309.12554].
- Magnetic moments: r²SCAN tends to overestimate moments in itinerant magnets (e.g. Fe), but r²SCAN-L restores GGA-level accuracy [2008.12420, 2512.04995].

### Large-Scale Materials Datasets

For >1000 solids, r²SCAN achieves MAE in formation enthalpies ≈92 meV/atom and is systematically more reliable than SCAN or GGA for structural properties and energies, except in a few late transition metal intermetallics, where GGAs can perform comparably [2208.02841].

## 5. Applications: Transition Metals, Defects, Spin Crossover, and Magnetism

### Transition Metals and High-Throughput Workflows

- r²SCAN delivers unbiased predictions of equilibrium volumes, cohesive energies, and elastic moduli across 3d, 4d, and 5d transition metals (e.g. MAPE on volumes ≈2%; cohesive energies ≈8%) [2309.12554].
- Its stability enables robust QHA phonon calculations and efficient high-throughput screening for alloys, refractory metals, and intermetallics [2309.12554].

### Spin Crossover and Energetics

- r²SCAN markedly improves spin-crossover (SCO) energetics over SCAN, reducing mean errors in high-spin/low-spin gaps by more than a factor of two (ΔE_HL MAD ≈2.8 kcal/mol) and brings all errors into the entropy-compensated ±10 kcal/mol window [2009.07808].
- r²SCAN-L is nearly as accurate (MAD ≈7.9 kcal/mol for SCO) but is much faster, especially in periodic boundary conditions [2009.07808].
- Optimal workflow: geometry optimization with r²SCAN-L, final energetics with r²SCAN (“optimize in r²SCAN-L, evaluate ΔE in r²SCAN”) [2009.07808].

### Defect Physics and Quantum Materials

- r²SCAN enables efficient, accurate screening of defect formation energies, charge transition levels, and spin/optical properties in wide-gap materials such as hexagonal boron nitride (hBN), with deviations from hybrid functionals generally within 0–0.5 eV [2509.09002].
- Absolute band gaps and charge transition levels are systematically underestimated due to the semilocal nature, but r²SCAN accurately reproduces key defect properties (e.g. ZPL, ZFS) at low cost. Hierarchical workflows recommend r²SCAN for screening and hybrids/GW for final characterization [2509.09002].

### Magnetism and Altermagnetism

- In itinerant systems, r²SCAN-L avoids spurious magnetic ground states seen in SCAN and r²SCAN, offering physically realistic predictions for magnets like RuO₂. The onset of altermagnetism under strain, doping, or Hubbard U is systematically mapped within a constraint-adhering framework [2512.04995].

## 6. Numerical Efficiency and Implementation

- r²SCAN converges 5–10× faster than SCAN on integration grids in molecular codes, is stable with medium/fine grids in plane-wave codes, and eliminates SCF failures seen in SCAN [2008.03374, 2208.02841].
- r²SCAN-L delivers 3–4× speedup in periodic codes and 20% acceleration in molecular codes relative to r²SCAN. In benchmark tests, wall times for r²SCAN-L approach those of PBE [2008.12420, 2009.07808].
- Deorbitalized forms (e.g. SRPP/SRPP2) further harmonize accuracy, smoothness, and speed for high-precision solid-state calculations [2509.00953].

| Functional    | SCF Stability | Computational Cost | Magnetization Accuracy (Fe) |
|---------------|--------------|-------------------|-----------------------------|
| SCAN          | Poor (grid-sensitive)      | Moderate-high      | Overestimates (2.63 μB)    |
| r²SCAN        | Excellent     | Moderate-high      | Overestimates (2.63 μB)    |
| r²SCAN-L      | Excellent     | Low (GGA-like)     | Accurate    (2.28 μB)       |
| PBE           | Excellent     | Very low           | Accurate    (2.22 μB)       |

## 7. Applicability, Limitations, and Recommendations

r²SCAN and r²SCAN-L are recommended as general-purpose, numerically robust meta-GGA functionals for:
- High-fidelity structural, energetic, mechanical, and thermophysical predictions in both molecules and solids.
- Systems where computational efficiency and stability are imperative (large solids, high-throughput, defect screening).
- Spin-crossover and noncollinear magnetic systems (r²SCAN for energetics, r²SCAN-L for structure).

Identified limitations include:
- Systematic underestimation of band gaps, especially for wide-gap insulators (requiring hybrid or GW corrections for spectroscopy).
- Remaining minor grid sensitivity in certain molecular properties for r²SCAN-L.
- Overestimation of magnetic moments in metallic systems for non-deorbitalized meta-GGAs (addressed by r²SCAN-L).
- In warm-start molecular dynamics, Laplacian-level meta-GGAs may require more SCF cycles per step, partly negating per-cycle efficiency gains [2509.00953].

For most applications, r²SCAN (and its dispersion-corrected forms, e.g., r²SCAN+rVV10) represents a balanced “workhorse” functional, superseding most GGA and early meta-GGA approaches [2208.02841, 2309.12554]. For properties requiring explicit KS potential accuracy or hybrid-level treatment, a hierarchical protocol is advised: structural and preliminary screening with r²SCAN/r²SCAN-L, and final spectroscopic or fine-accuracy properties with hybrids or many-body techniques [2509.09002].

Source: https://www.emergentmind.com/topics/r2scan-meta-generalized-gradient-approximation-meta-gga