Methacrylate

The experimental system is similar to (28 (link)) but with the addition of a 3D printed cylindrical obstacle. The microrollers are TPM [3-(trimethoxysilyl)propyl methacrylate] spheres with a diameter of 2.1 μm with an embedded magnetic hematite cube (Fig. 1, C and D) (28 (link)) suspended in water. The TPM spheres had a total diameter of 2.1 ± 0.1 μm, and the hematite cubes had a side length of 0.77 ± 0.1 μm, both measured using scanning electron microscopy. The TPM spheres where fluorescently labeled for fluorescence microscopy using 4-methylaminoethylmethacrylate- 7-nitrobenzo-2-oxa-1,3-diazol (NBDMAEM) (28 (link)). To reduce the Debye length b, lithium chloride (LiCl) was dissolved in the water. The obstacles were printed using a photopolymer resist (IP-Dip) on microscope coverslips using a Nanoscribe Professional GT two-photon printer (47 (link), 48 ). The autofluorescent obstacles were printed as open cylinders with height H = 20 μm, where the wall thickness was 2.5 to 2.8 μm, in a periodic array with a square lattice, with a lattice constant of 100 μm. A sample chamber (~120 μm by 2 cm by 2 cm) was constructed from the coverslip with the printed pillars, two spacers, and a microscope slide (28 (link)), which was filled with the microroller suspension and sealed using an ultraviolet (UV) glue (Norland Adhesives, no. 68). The glue was cured using a UV light, after which the sample was placed on the microscope with the coverslip down. Before imaging, the colloids were allowed to sediment toward the coverslip on which the pillars were printed.
We imaged the microrollers and obstacles using a bright-field or fluorescence microscope (fig. S1) while applying a rotating magnetic field (40 G, 9 Hz; see Fig. 1D). The microscope was an Olympus IX83 inverted widefield microscope equipped with a 20×/0.7 numerical aperture air objective. A 555-nm light-emitting diode lamp was used for excitation during fluorescence imaging. During the acquisition of long image sequences, the particles were kept in focus using the Olympus IX3-ZDC2 drift compensation module.
The magnetic field was generated using a home-built triaxial Helmholtz coil set mounted on top of the microscope stage [see (28 (link)) for details and images]. Thorlabs C-Mount extension tubes were used to raise the objective close to the sample in the center of the coil set. To create the rotating magnetic field, two out-of-phase sinusoidal signals were created by a Python script, a data acquisition system (Measurement Computing), and two (audio) AC amplifiers (EMB Professional) and fed into the coil set. The phase difference between the two signals was π/2, and the signals were sent to one coil parallel and one perpendicular to gravity, resulting in a rotating magnetic field with its rotation axis parallel to the bottom wall.
Escape time measurements were done using bright-field microscopy and analyzed manually using ImageJ. Microrollers that remained adjacent to an obstacle after crossing its center coordinate were considered trapped; when these microrollers moved more than five particle diameters downstream from the post, they were considered released; the time difference between trapping and release was recorded for multiple interactions (sample size varied between 4 and 16 escapes per obstacle size). The times reported represent the mean and SE of the distribution of trapping times per obstacle size.
The 2D positional histograms of rollers passing obstacles in experiments were made from two fluorescence microscopy image sequences (sequence length of 1800 frames, captured at 1 frame/s with a 500-ms exposure time) in the following way. We tracked the positions of all the rollers using TrackPy (42 (link)) and filtered out rollers with very short trajectories (<100 frames), particles that did not move substantially, and clusters of rollers by their intensity. Next, we transformed the particle coordinates so that each particle’s position was shifted to the reference frame of the nearest obstacle; obstacle centers were identified using scikit-image’s contour finding algorithm (49 (link)). To remove instances where two rollers came closer than 15 μm (or 15r), we removed these instances from the trajectory as these microrollers likely were hydrodynamically interacting. Last, all the positions of the remaining rollers around ∼30 individual pillars from both image sequences were then combined and plotted in a 2D histogram.

All the reagents used in this study were of analytical grade. Acrylamide (AM), sodium p-styrene sulfonate (SSS), 3-(trimethoxysilyl)propyl methacrylate (MPS), sodium hydroxide (NaOH), hydrochloric acid (HCl, 37wt%) and ammonium persulfate were purchased from Aladdin Reagent Co., Ltd. (Shanghai, China). Anhydrous ethanol (99.7% purity) and tetracycline hydrochloride (TCH) were supplied by Macklin Biochemical Technology Co., Ltd. (Shanghai, China). Deionized Milli-Q water (18 μS cm⁻¹) was used to prepare solutions.