Bug in property calculations for saturated liquid/vapor

[ischoegl]

Problem description

The implementation of property calculations for edge cases (saturated liquid / vapor) is incorrect.

Steps to reproduce / Behavior

In [1]: import cantera as ct
   ...: w = ct.Water()
   ...: w.TP
Out[1]: (300.0, 101324.99996785645)

In [2]: w.cp_mass, w.thermal_expansion_coeff, w.isothermal_compressibility
Out[2]: (4180.788860505055, 0.0002739534956286111, 4.5370965768089256e-10)

In [3]: w.TP = 300, w.P_sat + 1 # compressed liquid
   ...: w.cp_mass, w.thermal_expansion_coeff, w.isothermal_compressibility
Out[19]: (4181.302071450938, 0.0002736968713794185, 4.5385398006526666e-10)

In [4]: w.TQ = 300, 0 # saturated liquid
   ...: w.cp_mass, w.thermal_expansion_coeff, w.isothermal_compressibility
Out[4]: (81268775.17346017, 66.66325231711855, 110675.13197217113)

In [5]: w.TQ = 300, 1 # saturated vapor
   ...: w.cp_mass, w.thermal_expansion_coeff, w.isothermal_compressibility
Out[5]: (81266465.29549846, 66.66325263277206, 2.834506650014112)

In [6]: w.TP = 300, w.P_sat - 1 # superheated vapor
   ...: w.cp_mass, w.thermal_expansion_coeff, w.isothermal_compressibility
Out[6]: (1871.4780865974458, 0.0033559522559820127, 0.00028393341634016384)

I.e. values jump to unreasonable values for edge cases (where properties are calculated based on one-sided differentials).

System information

• Cantera version: 2.5.0b1
• OS: Ubuntu 18.04
• Python/MATLAB/other software versions: 3.6.9

Additional context

Documenting another bug found while investigating #907. There is a question whether these properties should be calculated at all (see #919)..

PS: The cv calculation is likewise affected (calculations for saturated mixtures yield unreasonable results).