write-methods-section

Skill: Write Methods Section

Purpose

Draft a publication-quality Methods section for a computational thermodynamics paper. Includes mathematical notation, algorithm pseudocode, and benchmark design description.

When to Use

• Drafting the Methods section of a flash algorithm paper
• Documenting algorithmic changes for peer review
• Writing supplementary material with implementation details

Methods Section Structure

A good Methods section for a flash algorithm paper has these subsections:

1. Thermodynamic Framework

State the phase equilibrium conditions:

## 2. Methods

### 2.1 Phase Equilibrium Conditions

At thermodynamic equilibrium, the fugacity of each component $i$ must be equal
across all phases:

$$f_i^V(T, P, \mathbf{y}) = f_i^L(T, P, \mathbf{x}) \quad \forall i = 1, \ldots, N_c$$

where $f_i^V$ and $f_i^L$ are the fugacities of component $i$ in the vapor and
liquid phases, $\mathbf{y}$ and $\mathbf{x}$ are the vapor and liquid mole
fraction vectors, and $N_c$ is the number of components.

The equilibrium ratio (K-value) is defined as:

$$K_i = \frac{y_i}{x_i} = \frac{\hat{\phi}_i^L(T, P, \mathbf{x})}{\hat{\phi}_i^V(T, P, \mathbf{y})}$$

where $\hat{\phi}_i$ denotes the fugacity coefficient.

2. Equation of State

Describe the EOS used:

### 2.2 Equation of State

Calculations in this work use the Soave-Redlich-Kwong (SRK) equation of state
\cite{Soave1972}:

$$P = \frac{RT}{V - b} - \frac{a(T)}{V(V + b)}$$

with classical van der Waals mixing rules:

$$a = \sum_i \sum_j x_i x_j (a_i a_j)^{1/2} (1 - k_{ij})$$

$$b = \sum_i x_i b_i$$

where $k_{ij}$ are binary interaction parameters.

3. Baseline Algorithm

Describe the algorithm being compared against:

### 2.3 Baseline TP Flash Algorithm

The baseline implementation follows the classical two-stage approach
\cite{Michelsen1982a, Michelsen1982b}:

**Stage 1: Successive Substitution (SS)**

Starting from Wilson K-value estimates:

$$K_i^{(0)} = \frac{P_{c,i}}{P} \exp\left[5.373(1 + \omega_i)\left(1 - \frac{T_{c,i}}{T}\right)\right]$$

the phase fraction $\beta$ is obtained from the Rachford-Rice equation:

$$g(\beta) = \sum_{i=1}^{N_c} \frac{z_i(K_i - 1)}{1 + \beta(K_i - 1)} = 0$$

K-values are updated by:

$$K_i^{(n+1)} = K_i^{(n)} \frac{\hat{\phi}_i^L(\mathbf{x}^{(n)})}{\hat{\phi}_i^V(\mathbf{y}^{(n)})}$$

This is repeated until convergence or a maximum of $N_{SS}$ iterations.

**Stage 2: Newton-Raphson (NR)**

If SS has not converged, the algorithm switches to Newton-Raphson using
Michelsen's u-variable formulation \cite{Michelsen1982b}:

$$u_i = \beta \cdot y_i$$

The residual vector is:

$$r_i = \ln \hat{\phi}_i^V(\mathbf{y}) - \ln \hat{\phi}_i^L(\mathbf{x})$$

and the Jacobian:

$$J_{ij} = \frac{\partial r_i}{\partial u_j}$$

The update is:

$$\mathbf{u}^{(n+1)} = \mathbf{u}^{(n)} - \alpha \mathbf{J}^{-1} \mathbf{r}$$

where $\alpha$ is a line search parameter (Armijo backtracking).