[Advanced] — How to model continuous-inkjet droplet breakup in COMSOL Multiphysics: the physics (Rayleigh–Plateau instability), the step-by-step model build (two-phase laminar flow with a level set), calibration, post-processing, and the droplet-charging extension. Written as the model answer to SSG 511 exam Q3.
Why this is in a graphics course. It isn’t really graphics — it is computational fluid dynamics, and the exam took it from a published COMSOL paper rather than the lecture notes. It is here because it is on the paper. If you are short of time, skip Q3: the exam asks for Q1 and any three others, and the rest are covered by notes 02–06. The physics underneath is in continuum mechanics: computational methods and FEM.
Sources. The exam’s first sentence is the opening line of Modelling of Ink Jet Breakup using COMSOL Multiphysics® (Viry, Sturma, Namy and Barbet; Simtec and Markem-Imaje; COMSOL Conference 2024). This note follows that paper, the companion 2023 paper on droplet charging (Sturma et al.), and COMSOL’s application-library models Inkjet Nozzle: Level Set Method and Jet Instability: Level Set. Standard theory not taken from those sources is marked as such. [reference]
The kid version
Turn a tap on gently and the thin stream of water breaks into drops a little way down. Surface tension — the “skin” on the water — prefers drops to a long cylinder, so any small wobble grows until the stream pinches off. An inkjet printer does this on purpose, very fast, and shakes the nozzle at a steady rhythm so every drop comes off the same size at the same place. The computer model predicts where and how the stream breaks, so engineers can design the nozzle without building dozens of them.
Where the analogy stops working. A tap stream breaks at random spots. A printer’s stream is forced at a fixed frequency, so its breakup is regular — and the model has to be calibrated against experiments, because the shaking membrane itself isn’t simulated.
Terms used in the model
- CIJ: The letters stand for continuous ink jet. This is the printing technique where a stream of ink is broken into droplets continuously; chosen droplets are charged and deflected onto the print, and the rest go to a gutter.
- Rayleigh–Plateau instability: This is the tendency of a liquid cylinder, driven by surface tension, to break up when disturbed by wavelengths longer than its circumference.
- Breakup length: This is the distance from the nozzle to the point where a droplet has just separated.
- Satellite droplet: This is a small extra droplet that forms from the thin neck between two main droplets. Satellites hurt print quality.
- Level set function: This is a smooth field that marks which fluid is where — for example 1 in ink and 0 in air — with the interface at the contour.
- 2D axisymmetric: This is a model of a body of revolution drawn as one half-plane and spun about its axis, avoiding a full 3D model.
- Navier-slip condition: This is a wall boundary condition with no flow through the wall, but with the shear stress proportional to the fluid’s velocity along the wall, through a slip coefficient. Zero is no-slip; infinity is full slip.
1. Background to state first
- The process. In CIJ, ink leaves a nozzle at about 20 m/s and is broken into droplets at of the order of 100 kHz, typically about 100 µm in diameter. Chosen droplets are charged (about 1 pC), deflected by an electric field (about 1 kV/mm) onto the medium; uncharged droplets go to a gutter.
- Why it breaks up: the Rayleigh–Plateau instability. Surface tension makes a liquid cylinder unstable to disturbances longer than its circumference. A periodic perturbation applied at the nozzle by a piezoelectric membrane is amplified along the jet until it pinches off at a regular spacing.
- Standard theory, not from the papers: for an inviscid jet, the fastest-growing wavelength is about 4.5 × the jet diameter, and the droplet diameter is then about 1.9 × the jet diameter.
- What the model is for: the breakup length, the jet shape at breakup, and whether satellites form. These drive printhead size and print quality.
2. The step-by-step model build (the answer to write)
- Component: 2D axisymmetric. The nozzle and jet are bodies of revolution, so one half-plane suffices and the 3D cost is avoided.
- Geometry. Three connected parts: the stimulation (droplet generator) chamber, the nozzle with its chamfer, and a long free-surface region downstream, long enough to contain several formed droplets. The nozzle shape and operating point are the design variables.
- Physics. Add Laminar Flow and the Level Set interface through the multiphysics coupling Two-Phase Flow, Level Set (Fluid Flow > Multiphase Flow > Two-Phase Flow, Level Set > Laminar Flow). Flow is laminar and incompressible; gravity is neglected.
- Governing equations.
- Navier–Stokes — momentum change equals pressure, viscous and surface-tension forces, with incompressibility:
where includes the surface-tension force at the interface, . - The level set marks the phases, with the interface at . (COMSOL’s tutorials use in ink and 0 in air; the 2023 paper uses the opposite — state which you use.) It is carried by the flow and kept sharp by the reinitialised conservative level-set equation:
- Properties are blended across the interface: , .
- Level-set parameters: sets the amount of reinitialisation (COMSOL suggests the maximum velocity magnitude); sets the interface thickness and should match the mesh size. The 2023 paper uses m/s (the jet velocity) and µm.
- Navier–Stokes — momentum change equals pressure, viscous and surface-tension forces, with incompressibility:
- Materials. Ink density and viscosity, air density and viscosity, and the ink–air surface-tension coefficient.
- Boundary conditions.
- Upstream (top of the chamber): a periodic pressure . The constant (mean pressure drop) drives the flow; the oscillating at the stimulation frequency perturbs the jet and sets the breakup length.
- Downstream outer boundaries: open, at fixed pressure .
- Axis: axial symmetry.
- Walls: no-slip is the classical choice. Viry et al. use Navier slip — the slip coefficient lumps together uncertainties such as machining tolerance, roughness and non-Newtonian ink behaviour, and is fitted once.
- Initial conditions. Ink fills the chamber and nozzle; air fills the outside. Compute a smooth initial level-set field first (COMSOL’s inkjet tutorial does this as a separate step), then start the transient flow from it.
- Mesh. Triangular finite elements, refined where the interface moves most — in the 2023 paper from 10 µm down to 2 µm near the axis. Choose to match.
- Study. Time-dependent. In the 2023 paper the maximum time step is the pressure period ÷ 300, with a direct solver, over about 100 jet periods (1 ms) — solved until the solution becomes periodic.
- Calibrate. Fit the mean pressure drop and the slip coefficient together so the simulated mean flow rate and jet velocity match experiment. Then sweep the perturbation amplitude , and compare simulated and measured breakup lengths to find the relation between piezo signal amplitude and (linear for small amplitudes).
- Post-process. Plot the contour and the velocity field over time. Extract the breakup length and the jet shape at pinch-off. Check for satellites. Compare with experimental snapshots.
- Validate and discuss. The calibrated model matched breakup-length curves and jet-shape types for a low- and a high-viscosity ink; satellites were mostly predicted correctly, with some difference in their dynamics. Limitations: the piezo membrane is not modelled (so calibration is needed); rheology and roughness are lumped into the slip coefficient; and it is expensive — the 2023 model had about 620,000 degrees of freedom and took about 100 hours for 1 ms of physical time.
3. Extension: charging the droplets
Add the Electric Currents interface (AC/DC module) on a snapshot of the breaking droplet from the flow run. Assign conductivity by phase — about 0.1 S/m in ink, and a tiny non-zero S/m in air so the stationary problem is well posed. Set the droplet at 0 V and the charge electrode at a fixed potential (150 V in the paper). Solve a stationary study. The droplet’s charge is the integral of the normal displacement field over its surface (about −1.25 pC in the paper). The result depends on timing, so the time step must be small near pinch-off.
4. A short model answer for 10 marks
Build a 2D axisymmetric model of the nozzle and the ink–air region downstream. Couple Laminar Flow with the Level Set method (Two-Phase Flow, Level Set) to follow the interface () through pinch-off. Drive the jet with a periodic inlet pressure , with open fixed-pressure outer boundaries and Navier-slip (or no-slip) walls. Mesh finely near the axis, matching to the mesh and to the jet velocity. Solve a transient study with a small fixed maximum time step for about 100 periods until the jet breaks up periodically. Calibrate and the slip coefficient to the measured flow and jet velocity, sweep , and read off breakup length, jet shape and satellites to compare with experiment. Optionally couple an Electric Currents study to find each droplet’s charge.
Check your understanding
- Why is a 2D axisymmetric model justified here, and when would it stop being valid?
- What two quantities are calibrated, and against which measurements?
- Why must be matched to the mesh size?
Answers (after your attempt)
- The nozzle and jet are bodies of revolution and the flow is assumed symmetric about the axis. It stops being valid if the jet deflects or develops non-axisymmetric modes (for example, the electric deflection stage).
- The mean pressure drop and the slip coefficient, against the measured mean flow rate and jet velocity; then against measured breakup lengths.
- is the interface thickness; thinner than the elements and the interface cannot be resolved, much thicker and the interface is smeared and pinch-off timing is wrong.
Related
- Computational methods and FEM — what COMSOL is solving underneath
- Animation and Simulation — fluids in graphics, where visual plausibility replaces accuracy
- Question bank
- Computer graphics map
Software engineering
Infrastructure & security
Computer science
Hardware & engineering