A function generator (Wavetek Model no. 145; Willtek, Ismaning, Germany) produced a sinusoidal signal with an amplitude of 1.00 ± 0.02 V. The speaker volume control was set to maximum, and a button microphone (Genexxa 33–3003; InterTAN, Barrie, Canada) was mounted via a vibration isolator on the acrylic plate to provide a reproducibility check on the volume of sound emitted. The experimental setup, including microphone details, is shown schematically in Figure 1. Numerical experimental values we report should not be taken as being universal; in the present proof-of-concept study they are likely to be features of the resonances of the particular plate, mounting, and cavity geometry used.
At an applied frequency of 140 ± 0.5 Hz, the microphone output showed several frequencies as well as the main 140 Hz response. These were clearly due to a combination of plate and speaker resonances: they could be changed by putting a finger on the plate. The microphone output was 55 mV peak-to-peak (p-p). At 195 Hz (±0.5 Hz), the microphone output was equally polyphonic and generated a 73-mV p-p spike at the main 195 Hz response, although most of the waveform power was ~55 mV p-p, as in the 140 Hz case.
The pre-loaded “reagent” liquid was modeled using a solution of 50 mL glycerol [molecular weight (MW) = 92.0, viscosity 1.5 Pa s at 20°C] and 10 mL deionized (DI) water, into which had been dissolved 4 g of KCl. A drop of 15 ± 1 µL glycerol solution was placed in the well with a precision micropipet (10 µL; Research 3111 000.122; Eppendorf South Pacific, North Ryde, NSW, Australia). Acoustic streaming was noticeable for any microliter quantity placed in the well, but a convex meniscus created a dark halo obscuring some of the flow. Thus, a procedure of slight overfilling (so the meniscus reached the well rim), followed by liquid removal, formed a concave meniscus that facilitated imaging. The patterns generated (discussed under the “Results and discussion” section) were not sensitive to these small variations in volume. Images of the chamber were captured in this “background” state to ensure that any variations in illumination, camera gain, etc. could be corrected for if necessary.
Blue dye was made by dissolving 2.523 g Brilliant Blue (CI 42090; Asia Pacific Specialty Chemicals Limited, Seven Hills, NSW, Australia; MW = 792.86 Da) in 300 mL DI water. To begin the experiment, the dye was formed into a 0.100 ± 0.005–µL drop “sample,” which was placed in the well. In the images shown in Figure 2 (discussed later, in the “Results and discussion” section), the initial drop sizes appear to vary somewhat, but this was due to small variations in the time taken to switch on the sound, plus the fact that a drop initially spreads rapidly on the surface only. The dye has a much higher diffusivity than most biological macromolecules (MW of proteins are usually in the range of 10–100 kDa), and it is expected that any improvements noticeable in the dye experiments would be greater with macromolecules.
Experiments were run with diffusion only, at two constant frequencies (discussed in the “Results and discussion” section) that both generated vigorous but different streaming patterns, and in a regime that alternated between these patterns. The time scale for the alternating regime was 15 s of vortex and 50 s of dipole.
Bright field microscopy was done at 2× magnification. The light source unit (Microlight 150 W; Shanxi Changcheng Microlight Equipment Co. Ltd, Taiyuan City, China) was a tungsten lamp fed via a fiber-optic cable. A digital camera capable of 60 frames per second (fps) (Model no. A602FC; Basler, Highland, IL, USA) was set to 7.5 fps. The white balance was set on background images and the gain and shutter speed were fixed during the experiments. Most micromixing efficiency assessments rely on estimates of mixing time, which, in general, is the time taken for introduced dye to appear uniformly dispersed through a mixing vessel (25,26).Results and discussion
We found that the presence of a curved liquid-drop surface open to the atmosphere was sufficient to produce excellent streaming. If a coverslip was placed over an overfilled well—so that there was no meniscus, no trapped air, and thus no locally large gradient in the acoustic field—streaming was eliminated entirely. Small, surface-tension vibrations of the drop surface were probably occurring; the frequencies at which flows were observed were of the lowest linear shape-mode frequency of the entire drop, assuming it was spherical (27). If amplitudes higher than those reported here were used, the ‘dimpled’ pattern characteristic of Faraday waves (11,20,28,29) appeared. Such waves may in turn generate useful mixing flows, but raised the risk of droplets atomizing from the open well, which is undesirable in a biological laboratory application. In tests using a range of well sizes, it was found that if a well were filled with liquid so that the radius of curvature of the meniscus was similar to the well radius, which requires that the well be small, acoustic microstreaming was readily and reproducibly driven.