Properties at the solidus
In this tutorial you evaluate the heat capacity of the mantle just below and just above the solidus, with and without the blend of the material properties. You see why Aragog uses a blend width matprop_smooth_width above 0, and what the configuration does with a width of 0.
What you will learn
- How to evaluate material properties at a given pressure and entropy with
EntropyPhaseEvaluator. - What the blend width does at the solidus.
- Which widths the configuration accepts.
- Where the expansivity of a single phase comes from.
What you need
- Aragog installed (see Installation).
- The SPIDER P-S test tables, with
ARAGOG_TEST_EOS_DIRset to their folder (see Build tests).
1. Compare the heat capacity on both sides of the solidus
The evaluator takes a pressure \(P\) [Pa] and a specific entropy \(S\) [J/kg/K] for each node and returns the blended properties. We evaluate it at 65.7 GPa, 10⁻⁶ J/kg/K below and above the solidus entropy \(S_\mathrm{sol}\), for the widths 0 and 0.01.
import os
import numpy as np
from aragog.eos.entropy import EntropyEOS
from aragog.eos.entropy_phase import EntropyPhaseEvaluator
eos = EntropyEOS(os.environ['ARAGOG_TEST_EOS_DIR'])
P = np.array([65.7e9]) # Pa, in the lower mantle
S_sol = float(eos.solidus_entropy(P)[0])
S_liq = float(eos.liquidus_entropy(P)[0])
print(f'solidus {S_sol:.1f} J/kg/K, liquidus {S_liq:.1f} J/kg/K')
def heat_capacity(width, S):
"""Heat capacity [J/kg/K] at pressure P and entropy S for a blend width."""
ev = EntropyPhaseEvaluator(eos, gravitational_acceleration=10.0, matprop_smooth_width=width)
ev.set_pressure(np.full_like(S, P[0]))
ev.set_entropy(S)
ev.update()
return np.ravel(ev.heat_capacity())
S = S_sol + np.array([-1e-6, 1e-6])
for width in (0.0, 0.01):
below, above = heat_capacity(width, S)
print(f'width {width}: c_p below {below:.1f}, above {above:.1f}, ratio {above / below:.3f}')
The output is
solidus 1350.2 J/kg/K, liquidus 9497.9 J/kg/K
width 0.0: c_p below 1373.5, above 29704.8, ratio 21.627
width 0.01: c_p below 15539.2, above 15539.2, ratio 1.000
At width 0 the heat capacity jumps by a factor of 21.6 between two entropies 2 × 10⁻⁶ J/kg/K apart: below the solidus it is the value of the solid table, above it the two-phase value, which holds the latent heat. At width 0.01 both sides have the mean of the two values. The evaluator called directly keeps a default width of 0; a run built from a configuration uses the width of the configuration.
2. Plot the blend
import matplotlib.pyplot as plt
dS = np.linspace(-300.0, 300.0, 601)
for width, style in ((0.0, ':'), (0.01, '-')):
plt.plot(dS, heat_capacity(width, S_sol + dS), style, label=f'width {width}')
plt.xlabel('S - S_sol [J/kg/K]')
plt.ylabel('c_p [J/kg/K]')
plt.legend()
plt.savefig('phase_blend.png')
The width is a fraction of the melting interval \(S_\mathrm{liq} - S_\mathrm{sol}\) = 8148 J/kg/K at this pressure, so a width of 0.01 spreads the change over about 81 J/kg/K. The dotted line (width 0) steps at \(S = S_\mathrm{sol}\); the solid line (width 0.01) passes smoothly through the mean value.
3. Check which widths the configuration accepts
The configuration classes refuse a width that is not strictly between 0 and 1, and a width below 10⁻³⁰⁰, where the JAX gradient of the blended heat capacity overflows.
from aragog.config.phases import MixedPhaseConfig
required = dict(
latent_heat_of_fusion=4e5,
rheological_transition_melt_fraction=0.4,
rheological_transition_width=0.15,
solidus='solidus.dat',
liquidus='liquidus.dat',
phase='mixed',
phase_transition_width=0.01,
grain_size=1e-3,
)
print(MixedPhaseConfig(**required).matprop_smooth_width)
try:
MixedPhaseConfig(**required, matprop_smooth_width=0.0)
except ValueError as err:
print(err)
The output is
0.01
matprop_smooth_width must lie in (0, 1), got 0.0
A TOML file read by Parameters.from_file refuses the same values.
4. See where the expansivity of a single phase comes from
Outside the mushy zone the evaluator derives the thermal expansivity \(\alpha\) [1/K] from the density, the heat capacity, the temperature and the adiabatic gradient of the same phase, \(\alpha = \rho\, c_p\, (\partial T/\partial P)_S / T\). The test tables also carry two thermal_exp tables, one per phase, which the solver does not read. We evaluate a melt state 200 J/kg/K above the liquidus with a width of 10⁻⁶, so that the blend does not enter.
S = np.array([S_liq + 200.0]) # in the melt, outside the blend
ev = EntropyPhaseEvaluator(eos, gravitational_acceleration=10.0, matprop_smooth_width=1e-6)
ev.set_pressure(P)
ev.set_entropy(S)
ev.update()
alpha = float(ev.thermal_expansivity()[0])
rho, c_p, T = (float(np.ravel(f())[0]) for f in (ev.density, ev.heat_capacity, ev.temperature))
identity = rho * c_p * float(np.ravel(ev.dTdPs())[0]) / T
table = float(eos._tables['thermal_exp_melt']['interp']([[P[0], S[0]]])[0])
print(f'alpha {alpha:.6e} 1/K, identity {identity:.6e} 1/K')
print(f'thermal_exp table {table:.6e} 1/K, relative gap {abs(table - identity) / identity:.1e}')
The output is
alpha 8.433257e-05 1/K, identity 8.433257e-05 1/K
thermal_exp table 8.433308e-05 1/K, relative gap 6.1e-06
The evaluator gives the identity value. The table value is close to it but not equal; SPIDER reads the table, so a comparison of Aragog with SPIDER shows this gap outside the mushy zone. Thermal expansivity of a single phase explains the choice. Phase transitions explains the blend, and Validation: Phase transitions shows the checks.