Critical points

Critical points

A critical point is the state where two coexisting phases become indistinguishable: their compositions, molar volumes, and densities converge, and the boundary between them disappears. For a pure substance it is a single point in pressure-temperature space. For a mixture it also depends on composition, so the critical states form a critical line in pressure-temperature-composition space.

Near a critical point the usual “which phase is this?” questions stop having sharp answers. That is why critical calculations are useful: they mark the end of a phase envelope, and they explain why a flash or a saturation solver becomes ill-conditioned in that region.

yaeos provides:

Theory

For a pure fluid, the critical point is the point where the pressure-volume curve develops an inflection:

For a mixture, composition fluctuations matter as well as changes in density; the pure-fluid pressure-inflection test is not a general mixture criticality criterion. Write the total Helmholtz energy as an ideal-gas contribution plus a residual (real-fluid) contribution,

The mixture loses stability when the matrix of second derivatives of the total Helmholtz energy with respect to the component mole numbers becomes singular:

The ideal contribution is not optional. In the Hessian divided by , it supplies the positive diagonal elements . An ideal gas with positive component amounts is stable against these fluctuations: this term does not make it unstable at low temperature. In a real fluid, the residual contribution can offset that stabilizing curvature. Using the residual Hessian alone would therefore test a different, generally incorrect condition.

This is how the implementation proceeds. The fugacity derivative returned by lnfug_vt satisfies

that is, the ideal-mixing term is already included. The stability matrix built from it is

and the implementation selects the eigenvalue closest to zero in absolute value, denoted here. This selection is an implementation choice; it is not by itself a test that all other eigenvalues are nonnegative. The implementation evaluates at the state and along a small composition perturbation , where is the eigenvector associated with the selected eigenvalue. The solver imposes

plus the pressure identity and the requested specification. The second condition is evaluated by a central difference with , and the whole system is solved by Newton iteration.

Critical-point specification

Critical-point calculations solve a system of equations involving the stability limit and the criticality conditions. One variable must be specified. The spec_CP object provides the allowed specifications:

Field Meaning
spec_CP%a Composition parameter
spec_CP%V Volume
spec_CP%T Temperature
spec_CP%P Pressure

The composition of the mixture is defined as:

where and are the two limiting compositions. With and , is simply the mole fraction of the second component. With more general limiting compositions, is a mixing parameter and the actual mole fractions are the linear combination above.

critical_point

type(EquilibriumState) function critical_point( &
   model, z0, zi, spec, S, max_iters, V0, T0, a0, P0 &
   )
   class(ArModel), intent(in) :: model
   real(pr), intent(in) :: z0(:)   !! First limiting composition
   real(pr), intent(in) :: zi(:)   !! Second limiting composition
   integer, intent(in) :: spec      !! spec_CP field
   real(pr), intent(in) :: S        !! Specification value
   integer, intent(in) :: max_iters
   real(pr), optional, intent(in) :: V0
   real(pr), optional, intent(in) :: T0
   real(pr), optional, intent(in) :: a0
   real(pr), optional, intent(in) :: P0
end function critical_point

The specification value S follows the natural variables of the solver, :

  • For spec_CP%a, S is the composition parameter itself, in .
  • For spec_CP%T, S is , with in kelvin.
  • For spec_CP%P, S is , with in bar.
  • For spec_CP%V, S is , with in L/mol.

The V0, T0, P0, and a0 arguments are initial guesses, given in the ordinary units (K, bar, L/mol, and respectively), not as logarithms. Only the specified variable S uses the logarithmic convention. Passing a temperature in kelvin as S for spec_CP%T therefore produces a nonsense state rather than an error.

Example

program critical_example
   use yaeos, only: pr, CubicEoS, PengRobinson76, EquilibriumState
   use yaeos, only: critical_point, spec_CP
   implicit none

   type(CubicEoS) :: model
   type(EquilibriumState) :: cp
   real(pr) :: tc(2)
   real(pr) :: pc(2)
   real(pr) :: w(2)
   real(pr) :: z0(2)
   real(pr) :: zi(2)
   real(pr) :: a

   ! Carbon dioxide + n-hexane
   tc = [304.21_pr, 507.6_pr]
   pc = [73.83_pr, 30.25_pr]
   w = [0.223621_pr, 0.301261_pr]

   model = PengRobinson76(tc, pc, w)

   z0 = [1.0_pr, 0.0_pr]  ! Pure carbon dioxide
   zi = [0.0_pr, 1.0_pr]  ! Pure n-hexane
   a = 0.5_pr             ! Equimolar mixture

   cp = critical_point( &
      model, z0, zi, spec=spec_CP%a, S=a, max_iters=1000, a0=a &
      )

   print *, "Critical temperature: ", cp%T, " K"
   print *, "Critical pressure: ", cp%P, " bar"
   print *, "Critical volume: ", cp%Vx, " L/mol"
   print *, "Critical composition: ", cp%x
   print *, "Iterations: ", cp%iters
end program critical_example

The result is an EquilibriumState with kind="critical", x == y == z, and Vx == Vy. beta is zero. With the parameters above the calculation converges in a few tens of Newton iterations to about 461.2 K and 76.4 bar.

critical_point has no separate failure flag: kind is set to "critical" when the iteration ends, whether or not the residuals are small. Check the returned state instead: the pressure must be positive and finite, the composition must be normalized and nonnegative, and the result must be consistent with an independent calculation.

critical_line

critical_line continues a critical point along the composition parameter . It uses the same continuation method as the phase envelopes.

type(CriticalLine) function critical_line( &
   model, a0, z0, zi, ns0, S0, dS0, &
   v0, t0, p0, max_points, maxP, first_point, stability_analysis &
   )
   class(ArModel), intent(in) :: model
   real(pr), intent(in) :: a0       !! Initial alpha
   real(pr), intent(in) :: z0(:)
   real(pr), intent(in) :: zi(:)
   integer, intent(in) :: ns0       !! Specification variable
   real(pr), intent(in) :: S0       !! Initial specification value
   real(pr), intent(in) :: dS0      !! Initial step
   real(pr), optional, intent(in) :: v0
   real(pr), optional, intent(in) :: t0
   real(pr), optional, intent(in) :: p0
   integer, optional, intent(in) :: max_points
   real(pr), optional, intent(in) :: maxP
   type(EquilibriumState), optional, intent(in) :: first_point
   logical, optional :: stability_analysis
end function critical_line

ns0 is usually spec_CP%a. In that case S0 is the initial and dS0 is the step in . A negative dS0 scans in the opposite direction. The optional first_point allows the continuation to start from an already converged critical point; without it the routine starts from an initial guess built from the component critical constants and a volume estimate.

When ns0 is not spec_CP%a, S0 follows the same logarithmic convention as critical_point: log(T), log(P), or log(V).

The optional first_point is used for two things: the volume, temperature and pressure guesses, and the composition parameter, which is taken from first_point%x(2). That second element is the mole fraction of the second component only for a binary mixture whose limiting compositions are the pure components. For other mixtures, set the starting point with a0 and S0 instead.

use yaeos, only: CriticalLine, critical_line, critical_point, spec_CP

type(CriticalLine) :: cl
type(EquilibriumState) :: cp
real(pr) :: a

a = 0.5_pr

cp = critical_point(model, z0, zi, spec=spec_CP%a, S=a, max_iters=1000, a0=a)

cl = critical_line( &
   model, a0=a, z0=z0, zi=zi, &
   ns0=spec_CP%a, S0=a, dS0=0.01_pr, &
   first_point=cp, max_points=1000 &
   )

print *, "Number of points: ", size(cl%a)
print *, "Alpha range: ", cl%a(1), cl%a(size(cl%a))

The CriticalLine type contains:

Field Meaning
a Composition parameter at each point
z0, zi Limiting compositions
P Pressure [bar]
V Molar volume [L/mol]
T Temperature [K]
ns Specification variable used at each point
iters Iterations used at each point
CEP Critical end point, when one is found

The number of points is decided by the continuation itself: it stops when the line turns, when a point fails to converge, when the pressure exceeds maxP (default 2500 bar), or when max_points is reached. An empty line is a legitimate outcome and means the starting point was not usable, not that the mixture has no critical behaviour. Starting from a converged critical point matters: with the example above, a continuation started from a poor point returned an empty line.

CEP is only filled when stability_analysis is requested and a critical end point is actually detected. The detection routine is limited to binary mixtures and scans a fixed grid of trial compositions, so a null CEP is not evidence that no critical end point exists.

get_critical_constants

get_critical_constants calculates the pure-component critical constants of an ArModel and updates the model’s components object in place:

use yaeos__equilibria_critical, only: get_critical_constants

call get_critical_constants(model)

! The critical constants are now available in the model
print *, model%components%Tc
print *, model%components%Pc
print *, model%components%w

This routine is not re-exported by the top-level yaeos module; it must be imported from the implementation module yaeos__equilibria_critical as shown above. Use it only with a model that has already allocated Tc, Pc, and w for its components.

The routine solves a critical point for each pure component, then estimates the acentric factor from the saturation pressure at :

It is used internally by models such as PC-SAFT, where the critical constants are not known analytically and must be obtained from the equation of state itself.

Practical notes

  • Critical-point calculations are sensitive to the initial guesses. Use a0, T0, P0, and V0 when they are available.
  • Specify temperature or pressure in preference to volume. A run with spec_CP%V for the example mixture returned a state with a pressure of order bar, which is a converged iteration of the wrong problem.
  • The specification value is logarithmic for spec_CP%T, spec_CP%P, and spec_CP%V; the initial guesses are not.
  • critical_point does not report convergence failure. Validate the returned temperature, pressure, and composition, and compare with a second specification when possible.
  • A critical line can cross a critical end point or a liquid-liquid region. Check the CEP field and the stability of the calculated points.
  • The critical line and the phase envelope should meet at the critical point. Use this as a consistency check.
  • Near a critical point the two phases are nearly identical, so the saturation and flash routines lose their conditioning: phase fractions, volumes, and compositions all become uncertain at the same time. Treat results close to a critical point as qualitative.
  • Critical calculations are expensive because they require second and third derivatives of the Helmholtz energy through numerical differentiation of the criticality conditions.