eduKateSG · How Science Works
Choose your reading route
Sound can generate a steady or slowly varying fluid flow through nonlinear interactions, often involving viscous boundary layers or attenuation in the liquid. The rapid acoustic oscillation and the slower mean circulation are different motions.
Contents
Understand the mechanism · Chapters 1–3
1. Separate the vibration from the circulation
Work the calculation · Chapters 4–5
Test the explanation · Chapters 6–8
6. A moving bead is not automatically a perfect flow meter
Imagine watching tiny tracers circulate inside a liquid while an acoustic transducer operates. The sound oscillates far too quickly to follow by eye, but the tracers complete visible loops. Acoustic streaming names the slower mean flow, not the rapid back-and-forth motion at the sound frequency.
The elementary statement that sound transports energy without carrying all the fluid along is a useful first approximation. It does not prohibit a smaller, higher-order mean flow. The distinction is between the leading oscillation and effects that become visible when the fluid equations are examined more closely.
This guide concerns fluid motion. The existing acoustic levitation guide concerns forces and trapping of objects. Both effects can occur in one apparatus, which makes measuring the liquid itself essential.
For an ideal oscillating velocity u = U cos(ωt), the average over complete cycles is zero. Its square, however, has average U²/2. Fluid momentum transport contains products of velocities, so averaging the governing equations is not equivalent to deleting every oscillating term.
This arithmetic is an entry point, not a complete derivation of streaming. Spatial structure, stresses and boundaries determine whether an averaged forcing can produce circulation or is balanced by a pressure adjustment. A nonzero average of a square alone does not prove a pump exists.
Original scale comparison: an oscillation amplitude of 0.020 m/s gives an average squared velocity of 0.00020 m²/s². Doubling the amplitude gives 0.00080 m²/s². That fourfold change explains why quadratic responses are plausible in a weak-forcing regime, while leaving the geometry-dependent flow calculation unfinished.
Near a solid wall, viscosity changes the amplitude and phase of the oscillating fluid motion over a thin region. The interaction of that layer with the surrounding acoustic field can generate boundary-driven streaming. In another common regime, attenuation of a travelling sound beam produces a distributed driving effect in the liquid.
These descriptions are not interchangeable diagrams of the same geometry. A microchannel, an open tank and a vibrating sharp edge can have different circulation patterns. Zhang and colleagues’ sharp-edge study explicitly investigated the relative sizes of edge curvature and the viscous layer.
Research on frequency-dependent streaming jets examined attenuation-driven flow and finite source size. The apparatus supplies boundary conditions as well as sound; neither can be omitted from the explanation.
Original calculation. For a simple oscillatory viscous layer, use δ = √(2ν/ω), with kinematic viscosity ν and angular frequency ω = 2πf. Assign ν = 1.0 × 10−6 m²/s and f = 1000 Hz. The result is δ ≈ 17.8 micrometres.
At 4000 Hz with the same viscosity, δ ≈ 8.92 micrometres. Quadrupling frequency halves this length because δ scales as f−1/2. Quadrupling viscosity at fixed frequency instead doubles it.
This is a local length scale, not the width of every visible vortex. Compare it with the channel height and the radius of a nearby edge before using a thin-layer approximation. If those dimensions are comparable to δ, a simplified boundary condition may no longer represent the detailed motion adequately.
Experimental work on the acoustic viscous layer measured amplitude and phase, illustrating that a layer theory can be tested through more than a picture of the eventual circulation.
Suppose a particular apparatus has measured mean speed 20 micrometres per second at acoustic velocity amplitude U₀. A provisional weak-forcing model predicts V = 20(U/U₀)² in those units. At 1.5U₀ it predicts 45; at 2U₀ it predicts 80.
These invented values define a test, not a universal performance claim. The acoustic amplitude in the fluid must be measured or calibrated. Doubling an electrical drive setting need not double U if the transducer or resonance changes.
An original study of vibration-driven streaming near a sharp edge identified a quadratic regime and investigated behaviour over a larger forcing range. A scaling law can be useful precisely because its departures expose a change in assumptions.
Also inspect the full velocity field. Two experiments can have the same maximum speed but different vortex positions, recirculation zones and net transport through an outlet.
| Relative acoustic amplitude | Predicted streaming speed |
|---|---|
| 1.0 | 20 micrometres/s |
| 1.5 | 45 micrometres/s |
| 2.0 | 80 micrometres/s |
A tracer can move with the fluid and also experience acoustic radiation forces, sedimentation, wall interactions or Brownian motion. Its observed speed is therefore not always identical to the local liquid speed.
Microparticle-image-velocimetry research compared streaming-induced and radiation-induced particle motion across particle sizes. The distinction matters because changing tracer size can change what the measurement represents.
For a thought experiment, release two tracer sizes into the same region. If both follow similar loops but one also drifts strongly towards an acoustic node, the common circulation and the size-dependent drift deserve separate explanations. Agreement between tracers is useful evidence, but is not enough without considering their other forces.
Report averaging time as well. A mean field can hide a wandering jet or switching vortices. The word steady should describe the measured timescale rather than an impression from one image.
Compare sound-off and sound-on observations while monitoring temperature. Heating can generate buoyancy-driven convection that resembles an acoustically driven circulation. A transducer may also shake the container mechanically. These are alternative contributions to investigate, not reasons to distrust every streaming experiment.
Use particle trajectories, acoustic-field information and temperature measurements together. Change one relevant feature, such as frequency or geometry, and predict how competing explanations would respond before looking at the result.
Fluid composition can matter too. A theoretical and experimental study of inhomogeneous fluids demonstrated suppression of boundary-driven streaming in particular conditions. A circulation pattern observed in a uniform liquid should not be transferred unchanged to one containing density and compressibility gradients.
These are measurement-design examples, not instructions for constructing a high-intensity ultrasonic source. Published fields and controlled laboratory datasets are sufficient to practise the reasoning.
Does a zero mean acoustic velocity rule out streaming? No. Nonlinear products and their spatial variation can contribute to a slower mean flow.
Does doubling frequency double streaming speed? Not generally. It changes several aspects of the acoustic and viscous problem; the layer-thickness formula is not a universal speed formula.
Does circulation mean liquid leaves the device? No. A closed vessel can contain strong loops with zero net discharge through its boundary. Mixing and pumping are different performance claims.
Continue to Acoustics for frequency, intensity and resonance, or Fluid Dynamics for momentum and boundary conditions.