Stability¶
ChE 621 — Thermodynamics.
Classification of equilibrium states¶
Equilibrium states are classified based on the local curvature around the equilibrium state.
Figure 1 The four classes of equilibrium state. Four small plots of \(U\) against a generalized displacement, each with a marble resting on the curve. Top left, stable: a single well, with the marble at the bottom and a short arrow pointing from the marble up the inside of the right-hand wall of the well. Top right, metastable: a shallow well on the left, then a barrier, then a deeper well on the right; the marble sits in the shallow well, and a vertical double-headed arrow between two dashed guide lines measures the barrier height \(E_{\mathrm{a}}\) from the bottom of the shallow well up to the top of the barrier. Bottom left, unstable: the marble balances on the summit of a hill, with a well on either side of it. Bottom right, neutrally stable: the curve is a horizontal line and the marble rests anywhere on it.¶
Stability criteria¶
Consider the stability of phase \(α\) with respect to the formation of another phase, \(β\), as shown in Figure 2.
Figure 2 A nucleus of phase \(β\) inside phase \(α\). A rectangular enclosure with hatching along all four outer edges, marking rigid walls. The interior is phase \(α\), labelled at its upper right. A small dashed oval near the lower left is a nucleus of the new phase \(β\), labelled inside the oval. The dashed outline marks it as an incipient region rather than a physical wall.¶
Expansion around equilibrium of the total derivative of
Then
with
and the displacements measured from the equilibrium value:
\(U\) is at a minimum at \(\vec{z}^{\mathrm{e}}\) if and only if
\(δU = 0\);
\(δ^{2}U \ge 0\), and if \(δ^{2}U = 0\), then
\(δ^{3}U \ge 0\), and if \(δ^{3}U = 0\), then \(δ^{4}U\), and so on, until the lowest non-vanishing term \(δ^{n}U > 0\).
The second variation for two phases¶
where
Because phases \(α\) and \(β\) are initially identical, two consequences follow together:
The second follows because
Therefore
Drop the superscript \(α\), so that
The Hessian form¶
For \(n\)-component systems,
The second variation is the quadratic form
The matrix in the middle, \(\bar{\bar{H}}\), is the \(m \times m\) Hessian matrix of second derivatives; it is also Hermitian. In the compact form,
and \(\bar{\bar{H}}\) is positive definite if and only if \(|\bar{\bar{H}}| > 0\).
Example: two variables¶
For \(\tilde{U} = \tilde{U}\left(\tilde{S}, \tilde{V}\right)\), the determinant of the two-by-two Hessian must be positive:
Reducing Equation (1) to a diagonal form¶
Back to Equation (1), one can eliminate the cross terms and get
where
Stability therefore requires \(U_{SS} > 0\) and \(A_{VV} > 0\), which are
Derivation¶
Squaring,
Rearranging gives
Substituting into \(\mathrm{d}^{2}U\) gives
We have eliminated \(\mathrm{d}S\) terms but still have cross terms involving \(V\) and \(N\). Note that the quantities in parentheses can be expressed as second order derivatives of \(A\):
The second group is a triple product, so that
Now
Dividing by \(\mathrm{d}V\) at constant \(T\) and \(N\) gives
so that
Similarly, we can show
so that
Still one more cross term to remove. Consider, at constant \(T\),
so that
and
Now
The third term is zero:
So finally
With \(y^{\left(0\right)} = U\left(S, V, N\right)\), Equation (2) can be written as
Outline of a general stability criterion¶
Take \(y^{\left(0\right)} = U\), so that \(δ^{2}U > 0\) becomes
where \(y_{ij}^{\left(0\right)}\) is the second derivative \(\frac{\partial ^{2}y^{\left(0\right)}}{\partial z_{i}\cdot \partial z_{j}}\). Define
One can show
which holds if and only if
Question: which of the \(m - 1\) criteria is violated first? There is a relationship,
As \(y_{\left(k-1\right)\left(k-1\right)}^{\left(k-2\right)}\) approaches 0, \(y_{kk}^{\left(k-1\right)}\) becomes negative first. Therefore
Binary mixture¶
For \(U = f\left(S, V, N_{1}, N_{2}\right)\),
and that last derivative can be rewritten as
The compact determinant form¶
The most compact and convenient form for expressing the stability criteria uses a sequence of determinants. Let
Here \(m = n + 2\), where \(n\) is the number of species, and \(0 \le i \le m - 2\).
Then \(\mathcal{L}_{i} > 0\) for stability, and \(\mathcal{L}_{i} = 0\) at the limit of stability.
Example: the internal energy of a one-component system¶
Take \(y^{\left(0\right)} = U\left(S, V, N\right)\), so \(m = 3\). For stability, \(0 \le i \le 1\), so there are two criteria of interest.
For \(i = 0\),
which gives
For \(i = 1\), \(\mathcal{L}_{1}\) is the one-by-one determinant \(|y_{22}^{\left(1\right)}| = A_{VV} > 0\), that is
Since \(κ_{T} = - \frac{1}{V}\left(\frac{\partial V}{\partial P}\right)_{T}\) is the isothermal compressibility,
Equivalent stability criteria¶
Equivalent stability criteria with \(y^{\left(0\right)} = U\), one for each ordering of the independent variables:
Ordering |
Criterion |
|---|---|
\(S, V, N\) |
\(A_{VV} = - \left(\frac{\partial P}{\partial V}\right)_{T,\, N} > 0\) |
\(S, N, V\) |
\(A_{NN} = \left(\frac{\partial μ}{\partial N}\right)_{T,\, V} > 0\) |
\(V, S, N\) |
\(H_{SS} = \left(\frac{\partial T}{\partial S}\right)_{P,\, N} > 0\) |
\(V, N, S\) |
\(H_{NN} = \left(\frac{\partial μ}{\partial N}\right)_{P,\, S} > 0\) |
\(N, S, V\) |
\(U'_{SS} = \left(\frac{\partial T}{\partial S}\right)_{μ,\, V} > 0\) |
\(N, V, S\) |
\(U'_{VV} = - \left(\frac{\partial P}{\partial V}\right)_{μ,\, S} > 0\) |
All will be violated at the same time.
\(U' = U - μ\cdot N = f\left(μ, S, V\right)\).