toolbox_chaos
v0.1.0
← All guides

Toolbox Chaos practical guide

2D Phase Portraits

Inspect selected pairs of state variables, compare projections from the same run, and avoid confusing a flattened view with the full state space.

Level Beginner
GUI tab Retratos 2D
Research question How do pairs of state variables evolve together, and which features survive across different projections?

Objective

What you will accomplish

Use pairwise projections to reveal relationships, folds, loops, and return geometry while preserving the identity of the source trajectory.

Before you begin

  • Have either a compatible last trajectory or a complete local simulation setup.
  • Know which state labels correspond to each plotted axis.

Reference outputs

What these views can show

Current Retratos 2D tab with three selected Lorenz projections
The current tab lets the user select pairwise planes, reuse a compatible trajectory, and save the complete grid.
Grid of Lorenz xy, xz, and yz phase portraits
The same Lorenz trajectory looks different in each pairwise projection; read the grid as a coordinated set.
Pairwise projections of a Chua trajectory
Projection comparisons are also useful for piecewise-linear circuit models such as Chua.

Procedure

Step-by-step workflow

  1. Select the system and source

    Choose the same system as the shared trajectory and press Usar última trayectoria, or enter a fresh numerical contract and run locally.

    If the interface reports a system mismatch, do not force reuse; simulate the selected system locally or return to the correct model.

  2. Choose variable pairs

    Enable the available pairwise planes. For three displayed states these are normally x–y, x–z, and y–z; higher displayed dimensions can provide more combinations.

    Keep at least two projections when diagnosing whether an apparent crossing or loop is a flattening artifact.

  3. Set a consistent visual style

    Choose a color with enough contrast and retain equal aspect where provided so geometric proportions are not visually distorted.

    Use the same style for a controlled series unless color itself encodes a declared experimental condition.

  4. Read projection-specific structure

    Look for repeated loops, folded bands, symmetry, narrow channels, clusters, and regions visited only during the transient.

    Compare every observation with another projection or with the time series before assigning a dynamical meaning.

  5. Repeat under a controlled change

    Vary one parameter, initial-state component, step, or duration. Keep the remaining contract fixed and name the changed quantity in the export.

    For parameter intervals, use the bifurcation workflow rather than treating a few handpicked portraits as a sweep.

  6. Export the grid

    Use Guardar gráfica to save the complete selected projection grid. Confirm that axis labels remain readable at the intended publication size.

    If individual panels are needed, keep the complete grid as the provenance record and state how any crop was produced.

Result

Expected output

  • A grid of selected pairwise portraits generated from one known trajectory.
  • A list of features that are consistent across projections and features that appear only after flattening.

Interpretation

How to read it

A projected crossing can represent states separated in an omitted coordinate. A closed-looking loop may also conceal slow drift outside the plane.

Pairwise portraits are strong descriptive tools, but dimensional reduction can hide state dependence and cannot by itself classify the full invariant dynamics.

Export

Reproducibility checklist

  • Name every variable pair and state whether the trajectory was reused.
  • Record the full simulation contract even though the panel shows only geometry.
  • Keep aspect-ratio and crop choices consistent across comparisons.

Applications

Where this workflow helps

  • Visual comparison of coupled variables.
  • Preparation of figures for lectures, posters, and exploratory reports.
  • Selecting a useful observable pair before a basin, section, or return-map study.

Limits

What it does not establish

  • Projection overlap is not a state-space intersection.
  • A portrait does not measure sensitivity, spectral content, or local stability.
  • The panel visualizes known trajectories; it does not perform hidden-attractor localization.