Laser‐Guided Self‐Rolled Magnetic Microrobots for Targeted Biofilm Eradication in Severely Infected Medical Stents

ABSTRACT

Pathogenic bacterial biofilms on biological interfaces and implanted medical devices are highly resistant to conventional antimicrobial therapies, leading to persistent infections and device failure. Magnetically driven micro/nanomotors (MNMs) offer a promising platform for localized drug delivery and in situ biofilm eradication in complex anatomical environments. However, current MNMs face critical challenges, including the serious risks of retention in vivo and insufficient propulsion within viscoelastic biofilms. Here, a laser‐guided self‐assembly strategy is developed to assemble tubular magnetic micromotors from high‐entropy alloy/polyimide (HEA/PI) bilayers for hydrogel‐based drug delivery and biofilm eradication inside implantable medical tubes. Programmable direct laser writing converts PI into laser‐induced graphene (LIG) while simultaneously inducing controlled self‐rolling of the HEA/LIG bilayers into mechanically robust micro‐rolls. Under rotating gradient magnetic fields, these micro‐rolls display controllable oscillatory–spiral propulsion in confined microchannels, enabling fast transport and site‐specific drug release. When filled with an antibiotic‐loaded hydrogel, the HEA/LIG micro‐rolls achieve synergistic mechanical biofilm disruption and localized antibiotic release within E. coli–infected pancreatic duct stents, resulting in a 97% sterilization efficiency, 44% higher than that achieved by standard chemical sterilization. This work establishes an unprecedented laser manufacturing paradigm for medical micromotors, providing a minimally invasive approach for targeted biofilm removal from hard‐to‐reach anatomical sites.

A laser‐guided self‐assembly strategy manufactures high‐entropy alloy/laser‐induced graphene micro‐rolls. Guided by rotating magnetic fields, these hydrogel‐loaded microrobots navigate confined channels to execute synergistic mechanical peeling and targeted antibiotic release. This dynamic micro‐platform efficiently eradicates recalcitrant biofilms, offering a minimally invasive, combined physical‐chemical approach to resolve severe medical stent infections.

1. Introduction

Implantable tubes are widely used in clinical practice as stents, catheters, and vascular access devices (e.g., pancreatic duct stent, ureteral stent, central venous catheter, dialysis catheter). However, device‐associated bacterial infections remain common, often necessitating premature device removal and, in severe cases, progressing to life‐threatening bloodstream infections [1]. Implant tube‐associated infections are notably difficult to treat because systemic antibiotics poorly penetrate the tubular lumen and established bacterial biofilms strongly impede sterilization [2, 3, 4]. Bacterial biofilms are cohesive microbial aggregates comprising bacteria encased within a self‐secreted extracellular polymeric substance (EPS) matrix [5, 6, 7, 8]. The EPS presents substantial resistance against conventional antibiotics, thereby rendering biofilms extremely recalcitrant to eradication [9, 10, 11]. Biofilms frequently colonize inaccessible anatomical sites and adhere to confined implant surfaces (e.g., the luminal walls) [12, 13, 14], presenting significant challenges for conventional removal methods, such as high‐dose antibiotic therapies or radiation sterilization. Given the EPS barrier and the spatial constraints of biofilm location, eradication strategies require three main attributes: targeted site‐specific delivery; enhanced EPS penetrability; and retrievability or degradability of the deployed antimicrobial agent.

Wireless micro/nanomotors (MNMs) provide an innovative platform for the in situ targeted eradication of bacterial biofilms [15, 16, 17, 18]. MNMs exhibit autonomous propulsion in diverse fluidic environments through either chemical fuel conversion [19, 20] or externally applied fields (e.g., magnetic, electric, acoustic, optical) [21, 22, 23, 24, 25, 26, 27], enabling precise navigation to target sites for spatiotemporally controlled drug release or direct biofilm disruption. Development in this field fundamentally relies on the progress in materials science and micro/nanofabrication technologies. Considering critical requirements for directional controllability and biocompatibility of actuation sources, magnetically driven MNMs leveraging physiologically benign and high‐penetration‐depth magnetic fields represent one of the most promising candidates. To achieve high‐throughput fabrication of magnetic MNMs, researchers have developed multiple techniques, including bottom‐up chemical synthesis [28], 3D printing (two‐photon lithography, 2PP) [29], electrochemical deposition [30], strain‐engineered self‐rolling [31], conventional direct laser writing (DLW) [32], etc. Chemically synthesized magnetic MNMs typically exhibit sub‐micron to low‐micron scale dimensions (with diameters ranging from tens of nm to a few $\mu \mathrm{m}$) and mainly operate through collective swarm behaviors [12, 33]. While such nanomotor swarms demonstrate adequate biofilm clearance, their minute dimensions impede real‐time tracking and pose significant retention hazards following human body deployment. Conversely, 2PP, electrochemical deposition, self‐rolling, and conventional DLW enable fabrication of large‐sized magnetic micromotors (with diameters ranging from tens to hundreds of $\mu \mathrm{m}$), among which helical and scroll‐type magnetic architectures demonstrate exceptional movement stability in physiological solutions [32, 34, 35, 36]. However, despite its potential for biofilm removal, 2PP remains expensive and time‐consuming; the other methods enable 3D helical and tubular architectures with high surface area and good integration potential, but have not yet been exploited as active biofilm‐removal microtools due to the limited propulsion forces for sustainable motion within realistic viscoelastic biofilm environments. Consequently, cost‐effective and high‐yield manufacturing of mechanically robust magnetic micromotors for biofilm removal remains a formidable engineering challenge.

Herein, by combining conventional DLW with strain engineering, we develop a high‐throughput laser‐guided self‐rolling strategy for fabricating tubular magnetic micromotors, which can be used for gel‐based drug delivery and bacterial biofilm eradication in medical silicone tubes. Through programmable laser scanning of 2D high‐entropy alloy/polyimide (HEA/PI) bilayer precursors, the irradiated regions controllably delaminate from the substrate under photothermal effects, while the PI is converted into laser‐induced graphene (LIG), subsequently self‐assembling into 3D magneto‐responsive HEA/LIG micro‐rolls. The HEA/LIG micro‐rolls exhibit excellent resistance to plastic deformation. Crucially, key structural parameters, including wall thickness, diameter, length, and magnetic moment, are precisely tunable through precursor composition and laser setting. Therefore, micro‐rolls can be engineered with outer diameters below the target microchannel dimensions. The lumina serve as a high‐capacity container for drugs. Sodium alginate (SA) hydrogels containing either Rhodamine B or antibiotics have been loaded into these microcavities. The HEA layer (MnFeCoNiCu) enables magnetic‐field‐controlled precise navigation in physiological fluids. Under a gradient rotating magnetic field (RMF), HEA/LIG micro‐rolls translate, rotate, and undergo lateral oscillation within microchannels of comparable size. Appropriate crosslinking of the SA hydrogel reduces premature drug release, and the self‐rotation and lateral oscillation of the HEA/LIG micro‐rolls under the RMF promote rapid and localized drug release at the target sites. Drug‐loaded magnetic micro‐rolls enable concurrent mechanical removal and chemical eradication of biofilms. Taking medical‐grade silicone tubes for pancreatic stents as an example, gel‐loaded micro‐rolls oscillate laterally to contact, grip, and detach Escherichia coli (E. coli) biofilms from the inner wall of the tubes. The combined rotational and oscillatory motions fragment large bacterial aggregates and biofilm clumps, with the SA hydrogel simultaneously releasing antibiotics during this process. Our methodology establishes an unprecedented platform for customizable magnetic micromotor fabrication with transformative potential in spatiotemporally controlled drug delivery, precise biofilm eradication in inaccessible sites, and minimally invasive biomedical interventions.

2. Results and Discussion

2.1. Laser Manufacturing of 3D HEA/LIG Micro‐Rolls

Figure 1a illustrates the key steps for the fabrication of HEA/LIG micro‐rolls. First, a HEA thin film (MnFeCoNiCu, 20 mol% per element) was sputter‐deposited onto a commercial PI substrate (25 $\mu \mathrm{m}$ in thickness), thereby forming an HEA/PI bilayer structure (Figure 1a, left). The bilayer film was then affixed to a pre‐cleaned soda‐lime glass substrate using a polyvinyl alcohol (PVA) adhesive interlayer, as Figure 1a, right shows (as reported in our prior work [37, 38]). In this configuration, the PI side was in contact with the PVA adhesive. The assembled film underwent thermal curing at 65, inducing solvent evaporation and subsurface bubble nucleation within the PVA matrix. To ensure uniform interfacial adhesion, a precision‐engineered stainless‐steel squeegee was applied at 5‐min intervals to eliminate the generated bubbles. This cyclic compression process continued until the full polymerization of PVA (until no obvious bubbles were observed).

FIGURE 1.

(a) The illustration of key steps for manufacturing the HEA/LIG micro‐rolls, including sputter deposition of the HEA layer, attaching the HEA/PI bilayer thin film on a glass substrate with a PVA interlayer, and laser fabrication of HEA/LIG micro‐rolls. (b) The tilted view (55°) of fabricated HEA/LIG micro‐rolls. (c) The uniaxial compression tests of a HEA/LIG micro‐roll, and the 3D printed microtube, microspring, and reinforced tube. (d) The laser power‐diameter relationship of HEA/LIG micro‐rolls. The insets show 4 typical HEA/LIG micro‐rolls fabricated with laser powers of 1.18, 1.24, 1.41, and 1.50 W, respectively (n = 15 micro‐rolls per laser power). (e) Representative moment–magnetic field curves for HEA/LIG micro‐rolls fabricated under 5 different laser powers. Each curve was measured from a composite sample consisting of 5 micro‐rolls (effectively representing the average behavior of 5 micro‐rolls per condition). The inserted figure shows the details of coercivity. The scale bar (b, c) is 1 mm. The scale bar (d) is 100 $\mu \mathrm{m}$.

Then, a commercial Nd:YAG fiber laser system (wavelength: 1064 nm, operating in continuous wave mode) was used to perform laser processing on the HEA/PI bilayer (Video S1). The laser process consisted of two substeps. First, the HEA/PI bilayer was patterned by laser ablation under high power, isolating the target regions from other areas. Second, the target regions were scanned by a continuous laser under a much lower power, inducing high heat in these regions through photothermal transformation. During the scanning, localized pyrolysis converted the PI layer into LIG while thermal decomposition of the sacrificial PVA interlayer released the adhered HEA/LIG film. Differential thermal stresses between irradiated (LIG) and non‐irradiated (HEA) layers generated out‐of‐plane deformation in the irradiated HEA/LIG film. With the laser scanning line by line (Figure 1a bottom, red dashed lines), the irradiated HEA/LIG film was released line by line, enabling the gradual scrolling governed by the laser scanning trajectory. This single‐step process yielded 3D HEA/LIG micro‐rolls with HEA inner surface (facing the micro‐roll lumina), exhibiting soft magnetic properties (high permeability, low coercivity) for magnetic actuation [39, 40, 41]. The lumina can be used to load hydrogel‐based drugs.

As illustrated in Figure 1b and Figure S1a, multiple HEA/LIG micro‐rolls were fabricated through serial laser scribing processes. The optimized laser parameters consist of a scanning speed of 400 mm s⁻¹, a calibrated output power of 1.24 W, and a line spacing of 3 $\mu \mathrm{m}$. Figure 1b shows distinct black and gray regions on most HEA/LIG micro‐rolls, while corresponding yellow and gray residues are present on the glass substrate. We suppose that this phenomenon results from the incomplete decomposition of the PVA sacrificial layer. The residues on the glass substrate are tentatively identified as oxidized PVA (yellow) and carbonized PVA (gray). Notably, carbonized PVA residues with varying degrees are also observed on the HEA/LIG micro‐rolls, resulting in the distinct black and gray regions. Given the negligible mass ratio of PVA to HEA/LIG micro‐rolls, the localized nonuniform distribution of carbonized PVA residues exhibits no measurable impact on the mechanical or magnetic properties of HEA/LIG micro‐rolls. This phenomenon is systematically discussed in the subsequent section. After the laser‐induced rolling process, one extremity of the HEA/LIG micro‐roll remains anchored to the HEA/PI film (Figure S1b). These micro‐rolls can be effectively isolated from the HEA/PI film through either laser ablation or mechanical cutting. Figure S1c presents a heap of over 100 uniformly dimensioned HEA/LIG micro‐rolls, fabricated using the same laser parameters as those in Figure 1b, demonstrating high morphological consistency in both length and diameter. These HEA/LIG micro‐rolls were subsequently employed for drug loading, transport, and controlled release applications.

During the conversion of PI to LIG, a fraction of PI typically remained unconverted [42, 43]. Given that the HEA layer was very thin and the LIG layer was highly porous, this residual PI was expected to serve as the primary mechanical backbone of the micro‐rolls, such that their mechanical properties (apart from surface‐related behavior) should theoretically be close to those of the original PI film. We therefore designed uniaxial compression tests to evaluate the resistance of HEA/LIG micro‐rolls to plastic deformation. Video S2 and Figure 1c compare HEA/LIG micro‐rolls with 3D‐printed microtubes, microsprings, and reinforced tubes, which share similar geometrical parameters (length, outer diameter, and wall thickness). The 3D‐printed structures were fabricated using a common photocurable resin. The top row shows the initial, undeformed state. The middle row shows the samples under compression, where the maximum deformation along the gravity direction was set to approximately 90% of the initial height to assess their ability to recover after unloading. The bottom row displays their configurations 2 min after unloading, when the shapes had stabilized. The HEA/LIG micro‐rolls almost completely recovered their original shape along the gravity direction, demonstrating excellent resistance to plastic deformation. In contrast, the 3D‐printed microtube recovered to only 91% of its initial thickness and developed a prominent central crack (highlighted by a red dashed line). The 3D‐printed microspring fractured into two segments and exhibited severe permanent deformation. The 3D‐printed reinforced tube recovered to 87% of its initial thickness and also showed a clear central crack (indicated by a red dashed line).

These results indicate that HEA/LIG micro‐rolls possess superior resistance to plastic deformation under uniaxial compression. Such robustness is critical for practical handling, including gripping the micro‐rolls for drug loading, inserting them into microchannels, and retrieving them afterward. From an application standpoint, the high resistance to plastic deformation and fracture is also advantageous for future in vivo use, where mechanical reliability is essential.

Our previous studies indicated the existence of a critical processing window for laser‐guided thin film self‐assembly [38]. Subliminal laser energy densities fail to generate sufficient photothermal energy for film delamination, while excessive energy input leads to film ablation. Through systematic parameter optimization, we identified a viable processing window with laser power ranging from 1.18 to 1.50 W at constant scanning speed (400 mm s⁻¹) and line spacing (3 $\mu \mathrm{m}$). In this research, laser power served as the primary variable, with other variables maintained as constants throughout subsequent experimental procedures.

Figure 1d presents a systematic diameter analysis of HEA/LIG micro‐rolls fabricated under varying laser power (1.18 to 1.50 W), with all precursor films (patterned regions) maintaining identical dimensions of 1.8 mm $\times$ 1 mm (Figure S2). Quantitative measurements reveal an inverse correlation between laser power and micro‐roll diameter, decreasing from 553 $\pm$ 30 $\mu \mathrm{m}$ at 1.18 W to 441 $\pm$ 41 $\mu \mathrm{m}$ at 1.50 W, consistent with our prior studies. Below the threshold power of 1.18 W, incomplete film delamination from the glass substrate occurs, while exceeding 1.50 W induces significant surface ablation, manifested as multiple perforations. Inserts of Figure 1d show representative micro‐roll morphologies under different laser powers: (1) 1.18 W generates single‐turn micro‐rolls; (2) 1.24 W generates incomplete 1.5‐turn micro‐rolls with unsealed sidewalls, unsuitable for encapsulation applications; (3) 1.41 W generates optimal 2.5‐turn micro‐rolls with fully sealed sidewalls; (4) 1.50 W generates 3‐turn micro‐rolls with a slightly decreased diameter but significant structural degradation and irregular internal cavities caused by the excessive thermal stress.

Raman spectroscopic analysis (Figure S3a) confirms LIG formation across all power settings, evidenced by characteristic D (1350 ${\mathrm{cm}}^{-1}$), G (1580 ${\mathrm{cm}}^{-1}$), and 2D (2700 ${\mathrm{cm}}^{-1}$) peaks. Statistical evaluation of over 10 samples per condition demonstrates that the I_G/I_D ratios of samples increase from 1.13 $\pm$ 0.12 (1.18 W) to 1.84 $\pm$ 0.39 (1.41 W) due to enhanced graphitization, followed by a decline to 1.44 $\pm$ 0.29 at 1.50 W due to oxidative degradation (Figure S3b) [44].

As demonstrated in Figure 1e, the vibrating sample magnetometry (VSM) results of HEA/LIG micro‐rolls fabricated under varying laser powers (1.18 to 1.50 W) reveal a distinct trend in saturation magnetization (M_s). The M_s values, averaged across five micro‐rolls per parameter set, initially decline from 2.77 to 2.08 emu ${\mathrm{g}}^{-1}$ as the laser power increases from 1.18 W to 1.35 W, followed by a stabilization period when increases to 1.50 W. For magnetic measurements, each data point was obtained from a composite sample consisting of 5 HEA/LIG micro‐rolls, due to the small mass of individual micro‐rolls, so that the measured M–H curves effectively represent the average magnetic behavior of 5 micro‐rolls per parameter set. Notably, this behavior correlates with the diameter variation trend of micro‐rolls (Figure 1d), indicating potential magnetic shielding effects induced by the LIG layer [45, 46]. The coercivity (H_c) values remain consistently low (tens of Oe) across all samples, as evidenced by the minor hysteresis loop variations in the inset. These findings suggest that while the LIG layer may modulate the magnetic properties of HEA, the overall magnetic performance remains stable within all the samples.

Given the minor differences in magnetic performance, geometric parameters became the primary reference factor during application. The micro‐rolls fabricated under a laser power of 1.24 to 1.41 W were selected for subsequent drug loading and release experiments due to their optimal balance between structural integrity (closed‐side morphology) and internal cavity volume. All materials, including commercial polyimide films, glass substrates, HEA sputtering targets, and PVA solutions, were sourced from standard suppliers and underwent only cleaning treatment. The calculated manufacturing cost per micro‐roll is less than $0.01 USD, demonstrating significant scalability potential.

2.2. Mechanism and Characterization for the HEA/LIG Micro‐Rolls

There are numerous published papers on the fabrication of LIG using an Nd:YAG fiber laser [47, 48, 49], showing various practical applications due to its brilliant physical and chemical characteristics, including biosensors [50], solar evaporation [51, 52], and supercapacitors [38]. In our previous research, the mechanism for LIG micro‐roll fabrication has been investigated, and three types of microstructures, flake areas, line areas, and smooth areas, have been observed, characterized, and discussed [38]. This section focuses on the HEA/LIG micro‐rolls, especially the peeling mechanism.

Figure 2a–I shows the tilted view (55^°) of a typical HEA/LIG micro‐roll. Looking closely, there is a triangular area on the left side of the micro‐roll (marked with a green frame and amplified in Figure 2a–II), and a triangular crack area in the corresponding glass substrate (Figure S4). It is worth noting that the triangular area corresponds to the black area in Figure 2b. It is necessary to compare the microstructures on the micro‐roll surface with those on the corresponding substrate surface for analysis. Figure 2a–II is further amplified in Figure 2a–III, IV. The triangular area is a carbonized PVA layer attached to the micro‐roll surface. The surrounding area exhibits typical LIG microstructures (Figure 2a–IV), whereas the carbonized PVA displays numerous small and neat pores (Figure 2a–III). The PI under the carbonized PVA layer can release a significant amount of carbon oxides during the conversion to LIG, and these pores in the carbonized PVA are likely to serve as channels for the release of carbon oxides. Figure S4 shows the condition on the glass substrate. The surrounding area (Figure S4b) has many film pieces, whereas the triangular area (Figure S4c) shows a very smooth surface. Therefore, Figure S4 further supports our judgment.

FIGURE 2.

(a) A SEM image of a typical HEA/LIG micro‐roll (tilted view, 55°) (I), the triangular area (green frame) on the HEA/LIG micro‐roll (II), the amplified triangular area (III) and surrounding area (IV). (b) The PI‐PVA interfacial delamination condition (I), and PVA‐Glass interfacial peeling condition (II). (c) The EDS mapping results show the element C on the outer surface, while Fe, Ni, Mn, Co, and Cu are present on the inner surface. The scale bar (a‐I, II) is 50 $\mu \mathrm{m}$. The scale bar (a‐III, IV) is 1 $\mu \mathrm{m}$.

Figure 2b illustrates two possible conditions for the peeling of the thin film. As mentioned above, Figure 2b shows two different colored regions on the micro‐roll, and Figure 1a shows that these different regions have different microstructures. The microstructure of the light gray area is typical of highly porous LIG, while the black area is carbonized PVA with dense pores. Considering that peeling easily occurs at the interface, we think that the two microstructures correspond to the peeling at the PI‐PVA and PVA‐Glass interfaces, respectively. Figure 2b–I illustrates the condition of PI‐PVA peeling. The PVA layer is carbonized, and the HEA/PI is peeled off. Part of PI is converted into porous LIG due to the photothermal effect, while releasing gas (carbon oxides). Figure 2b–II illustrates the condition of PVA‐Glass peeling. Part of PI is converted into porous LIG due to the photothermal effect, while releasing gas (carbon oxides). There are many dense holes on the carbonized PVA layer, which are the conduits for the release of carbon oxides. Considering that these two types of microstructures are observed on almost every micro‐roll, these two types of interfacial peeling should be competitive.

Figure 2c is a group of energy‐dispersive X‐ray spectroscopy (EDS) images for the same micro‐roll. The outer surface of the micro‐roll is primarily composed of carbon, specifically LIG and carbonized PVA. The inner surface of the micro‐roll contains Mn, Fe, Co, and Ni, all of which are elements present in the HEA target. The only exception is Cu. Our EDS image only shows copper elements on the inner surface of the micro‐roll near the end. Perhaps the X‐ray signal from Cu has a low penetration ability, so the detector did not capture it. Additionally, Si, O, Na, Ca, Mg, Al, and K elements are also detected. However, these elements are in the area where the PI is peeled off and the glass is exposed. Therefore, it can be judged that these elements come from soda‐lime glass. Furthermore, Figure S5 shows a large micro‐roll with the same element distribution.

Figure S6 displays the X‐ray Photoelectron Spectroscopy (XPS) curve of the micro‐roll surface, which matches with the EDS result in Figure 2c. Only C, Na, Si, and O elements are detected. C is the LIG or carbonized PVA. Na and Si are from soda‐lime glass, and O is from glass or air. Figure S7 shows the Ultraviolet–visible (UV–vis) Spectroscopy result. Figure S7a,b shows the reflection and transmittance of HEA/PI/PVA/Glass and PI/PVA/Glass, respectively. Figure S7c shows the calculated absorption rates. The orange line is the HEA/PI/PVA/Glass sample, and the blue line is the PI/PVA/Glass (control sample). At 1064 nm, the light absorption rate of the control sample is less than 10%, while that of the experimental sample is about 80%. The HEA layer makes a massive contribution to the light absorption rate. Nevertheless, the workable laser setting range for fabricating HEA/LIG micro‐rolls is very similar to that for fabricating LIG micro‐rolls [38]. We suspect that the HEA layer is not in direct contact with the PVA or glass, and the thermal conductivity of the alloy is significantly greater than that of the PI. Hence, the energy absorbed by the HEA layer has a minimal contribution to the film peeling process. Therefore, the peeling process of the film is mainly dominated by the photothermal effect of the PI itself.

2.3. Control the Movement of HEA/LIG Micro‐Rolls in Microchannels With Magnetic Fields

Magnetic fields are often used to control the motion of magnetic microrobots. For example, RMFs are commonly employed in bionic spiral microswimmers, while gradient magnetic fields are often utilized in tasks that require precise navigation or positioning [21]. Here, we design a gradient RMF to achieve continuous motion of HEA/LIG micro‐rolls in narrow microchannels.

Figure 3a displays our magnetic field setup and the relative position of the micro‐roll and microchannel. In previous research, an RMF was typically produced with a triaxial Helmholtz coil system. However, considering that many common devices that use RMFs, such as magnetic stirrers, normally employ permanent magnets to simplify the device and reduce costs, we also use rotating permanent magnets here. Two long N52 magnets are used to generate a magnetic field. The N pole of one N52 magnet faces the S pole of the other N52 magnet. In previous studies, two Helmholtz coils were often placed completely parallel to make the magnetic field lines completely perpendicular to the axis of the tubular microswimmer. However, such a magnetic field setting could only make the tubular microswimmer rotate but not move forward. Therefore, a spiral tail was usually added to one end of the tubular microswimmer to break the symmetry and achieve tail propulsion (corkscrew motion). However, such a design significantly increases the difficulty of the process. In this work, we set up two magnetic fields. One type involves setting two magnets completely parallel to make the micro‐roll rotate purely. The other is shown in Figure 3a–I. Two magnets are placed at a certain angle $\theta$ (here, $\theta$ = $5^o$). Figure 3a–II illustrates the magnetic field lines at the micro‐roll and microchannel. The component of the magnetic field lines parallel to the micro‐roll axis ($B_{//}$) is not zero. The component of the magnetic field lines perpendicular to the micro‐roll axis ($B_{\perp }$) rotates with an angular velocity ($\mathrm{\Omega }$) via a stepper motor. In our experiment, the micro‐roll and the silicone microchannel have similar diameters, so the two are often partially in contact. When the magnet is not rotating, that is, in a static magnetic field, no forward movement of the micro‐roll is observed. This is due to the viscoelasticity of the silicone that prevents translation on the contact surface. When the magnet is rotating, that is, in a gradient RMF, we observe the forward and rotating movements of the micro‐roll. Because it is difficult for the micro‐roll to be on the ideal central axis, the micro‐roll tends to move laterally toward a certain magnet (perpendicular to the contact surface). Therefore, when the magnet rotates, the micro‐roll will be driven to do lateral oscillation. During lateral movements, it will briefly leave the inner wall of the microchannel, and the resistance to forward movement and rotation will be significantly reduced. Therefore, the micro‐roll can move forward while rotating.

FIGURE 3.

Control the movement of HEA/LIG micro‐rolls in microchannels. (a) The illustration of magnet setting around the microchannel (the angle between the two is $\theta$) and magnetic flux lines (orange lines) (I), the magnetic induction $B$ around the microchannel and its components parallel ($B_{//}$) and perpendicular ($B_{\perp }$) to the microchannel axis, and the rotating angular velocity ($\mathrm{\Omega }$) (II), which enables moving forward ($V_{//}$) and rotation ($\omega$) of HEA/LIG micro‐rolls in microchannels (III). (b) The real device for producing a predesigned RMF. (c) A typical Ecoflex microchannel corresponding to the red dotted area in (b), including the top and tilted views showing the circular shape. (d) Three key frames showing the rotation ($\omega$) of a HEA/LIG micro‐roll (I‐III), and 5 HEA/LIG micro‐rolls' rotation speed adjusted by magnet rotation (IV). The inserted image in (I) shows the HEA/LIG micro‐roll with a HEA/PI part (yellow color), which is pointed with red arrows in these key frames. (e) A HEA/LIG micro‐roll moves deeper into the microchannel under the gradient RMF (I), and returns from the depth of the microchannel to the opening under the reversed gradient RMF (II). The scale bar (c–e) is 1 mm.

Figure 3b shows the real device manufactured by 3D printing. The holder in the picture is used to place two N52 magnets. According to the above description, we have printed two holders to make the two magnets completely parallel and at a 5° angle, respectively. A stepper motor is connected to the holder to rotate it. The stepper motor is connected to a power supply, and the angular velocity ($\mathrm{\Omega }$) is controlled by adjusting the current and voltage. The right side of the device can hold a glass slide, on which a silicone microchannel containing a HEA/LIG micro‐roll is located. The silicone microchannel is marked with a red frame. Figure 3c–I is the microchannel in Figure 3b. Figure 3c–II shows that the microchannel is circular in cross section and about 0.7 mm in diameter. The microchannel is molded from a long sewing needle.

To ensure that the micro‐rolls move in the microchannels filled with deionized water, the micro‐rolls and the liquid must be wettable. Here, we first soaked the HEA/LIG micro‐rolls in deionized water for 1 min, and then placed them into a PDMS microchannel filled with deionized water. Video S3 shows a typical HEA/LIG micro‐roll placed in the microchannel. One end of the micro‐roll was inserted into one opening of a microchannel. Then the micro‐roll could automatically enter the microchannel in a very short time. This was attributed to the capillary attraction of the liquid, which automatically pulled the micro‐roll into the liquid and microchannel, manifesting as the spontaneous entry of the HEA/LIG micro‐roll into the channel.

The two magnets were set to be completely parallel to achieve pure rotation of the micro‐roll first. In Figure 3d and Video S4, the micro‐roll was placed at the air‐liquid interface. The image inserted in Figure 3d–I shows more details of the micro‐roll. When cutting the micro‐roll, we deliberately left a section of the HEA/PI film that was not scanned by the laser (the yellow part), so that we could determine whether the micro‐roll has rotational motion by the difference in light reflection. The red arrows in Figure 3d–I–III point to the location of HEA/PI. From Figure 3d–I–III, the micro‐roll rotated half a circle. We adjusted the rotation speed of the magnets by changing the current and voltage of the power supply, and determined the rotation speed of the micro‐roll by counting the position of LIG/PI. Figure 3d–IV shows five HEA/LIG micro‐rolls of different diameters rotating in the RMF. The micro‐roll rotation speed always remains consistent with the magnetic rotation speed, meaning the magnetic moment can stay synchronized with the rotating field [32, 53]. The effect of the rotating magnetic torque should be much stronger than the viscous drag of the liquid and the resistance of the microchannel wall. Additionally, the micro‐roll is very light, so it has minimal inertia.

Figure 3e and Videos S5 and S6 show the motion of a HEA/LIG micro‐roll in the microchannel in a gradient RMF. In Figure 3e–I and Video S5, the micro‐roll moved deeper into the microchannel. Within 30 s, the micro‐roll moved forward 10 mm, which is about 7 times its length. The micro‐roll as a whole always moved to the right, as indicated by the green arrow. However, both ends of the micro‐roll were observed to reciprocate in a plane perpendicular to it as it moves. This is due to the difficulty for the microchannel to maintain an entirely consistent distance from the two magnets, so the micro‐roll tends to move toward the closer magnet. The micro‐roll is restricted by the narrow micro‐channel, so it appears as a lateral oscillating motion. It is worth noting that when the magnets do not rotate (i.e., static magnetic field), the micro‐roll does not move at all. It supports our assumption that an interaction force between the micro‐roll and the microchannel wall hinders its movement. When the magnetic field rotates, the front end of the micro‐roll can separate from the microchannel wall during the lateral oscillating motion, significantly reducing the resistance to the micro‐roll and allowing it to move forward. Due to the high symmetry of the micro‐roll itself, reciprocal motion is possible. In Figure 3e–II and Video S6, the micro‐roll returns from the deep area of the microchannel to the opening under a reversed gradient RMF, thus enabling the recollection of the micro‐roll.

2.4. Drug Loading, Transport, and Release With HEA/LIG Micro‐Rolls

For tubular microrobots, the lumen is an ideal place to load drugs. Although LIG is a highly porous structure, PI is not completely converted to LIG [42, 43]. The unreacted PI can prevent the drugs inside the cavity from leaking out. Drugs can be dissolved in a gel‐forming solution, which is then applied to the openings of the HEA/LIG micro‐rolls. Driven by capillary force, the solution will be drawn into the lumina of the micro‐rolls. After air‐drying, a hydrogel‐based drug is formed inside, as illustrated in Figure 4a, left. Then, the drug‐loaded micro‐roll can be inserted into a microchannel filled with a liquid, such as deionized water or PBS solution. As the gel absorbs water and swells, the drug within it can be gradually released, as illustrated in Figure 4a, right. If the drug‐loaded micro‐roll is directed to the biofilm area within a microchannel using a magnetic field, it can then transport the drug to the biofilm, as Figure 4a (bottom) shows. A viscous SA solution was used as the gel‐forming solution, as SA is a commonly used hydrogel‐forming material for in vivo drug delivery [54, 55]. Additionally, alginate‐based systems can be engineered for targeted or site‐specific drug delivery [56]. Rhodamine B was used as the model drug because its vivid color makes it easy to visually track its release in microchannels.

FIGURE 4.

The loading and release of hydrogel‐based drugs. (a) The illustration of gel containing dissolved drug in the lumen of a HEA/LIG micro‐roll (left), gel swelling to release the drug (right), and localized drug release near the biofilm in a microchannel (bottom). (b) The drug (Rhodamine B) loading and release steps, including drug loading, air drying, crosslinking in a ${\mathrm{CaCl}}_2$ solution, air drying again, and finally releasing the drug in a microchannel. (c) Two drug‐loaded HEA/LIG micro‐rolls placed on a researcher's finger. (d) The drug loading on HEA/LIG micro‐rolls of different diameters (I) and their openings after air drying (II). (e) HEA/LIG micro‐rolls after drug release in microchannels. (f) Three HEA/LIG micro‐rolls with similar geometrical dimensions (one without crosslinking, one crosslinked in 0.22 wt% ${\mathrm{CaCl}}_2$ solution for 5 min, and another crosslinked in 1.11 wt% ${\mathrm{CaCl}}_2$ solution for 5 min; the other drug‐loading steps are the same) were placed at one end of a long microchannel, and drug diffusion at 2 min was shown (I), and the representative concentration‐time curve (out of 3 independent repeats) (II). (g) Three HEA/LIG micro‐rolls with different diameters (the drug loading steps are the same) were placed at one end of a long microchannel, and the drug diffusion at 2 min was shown (I), and the representative concentration‐time curve (out of 3 independent repeats) (II). (h) Drug release at different time points (4, 19, and 43 min) for a micro‐roll placed at one end of a microchannel under an RMF (I), and drug release at the same time point (4 min) for three micro‐rolls (of different diameters) placed at one end of a microchannel under an RMF (II). (i) A micro‐roll moved toward the deep region of the microchannel under an RMF (I), the magnetic field was adjusted to induce near‐pure rotational motion for the micro‐roll (II), drug release (III), and return of the micro‐roll (IV). The scale bar (b, d, e) is 1 mm.

Figure 4b shows the complete steps of drug loading and release for a micro‐roll. First, the middle part of a micro‐roll was adhered to one corner of a PDMS block, leaving both ends exposed and suspended. Subsequently, a small amount of SA solution containing dissolved Rhodamine B was picked up with tweezers and brought close to the openings of the micro‐roll. The openings could capture an appropriate amount of the solution (depending on the tube diameter at the openings), which then entered the cavity through capillary action, as Video S7 shows. Therefore, this method is applicable for drug loading into micro‐rolls with various diameters. The sample was then left to air‐dry, allowing the SA gel to form naturally. The sample was then immersed in a ${\mathrm{CaCl}}_2$ solution, where SA was crosslinked under the action of calcium ions [57]. As clearly shown in Videos S8 and S9, the SA rehydrates and swells during this step, accompanied by the release of a small amount of Rhodamine B. This step was crucial for sustained drug release. Otherwise, the drug‐loaded micro‐rolls would rapidly release the drug upon entering the microchannel, and the SA would dissolve and flow into the channel, potentially causing blockage. The sample was then left to air‐dry again. Finally, the micro‐roll could release the drug at designated locations within the microchannel under magnetic control. The rightmost image in Figure 4b shows that the SA gel became lighter in color after releasing Rhodamine B. The SA gel shows clear boundaries, indicating that it only absorbed water and swelled without dissolving.

Figure 4c shows two drug‐loaded HEA/LIG micro‐rolls placed on a researcher's fingertip. Figure 4d top row shows some micro‐rolls being coated with the drug solution, while the bottom row shows the openings of the micro‐rolls after the SA gel has dried. The SA gel can fill lumina of different diameters.

Figure 4e shows three micro‐rolls after complete drug release. They were kept in a long microchannel for 1 h to allow release, then magnetically guided away from the visibly discolored region for observation of the SA gel after release. The SA gel absorbed water and expanded from the micro‐rolls' openings, with clear boundaries indicating no dissolution of SA. The whitish appearance with some dark spots suggests that most of the Rhodamine B had been released.

In the ideal case of site‐specific drug release, microrobots should minimize drug leakage before reaching the target location. The crosslinking of SA is commonly used to achieve slow drug release, so it is necessary to identify optimized crosslinking conditions. Three drug‐loaded micro‐rolls with different crosslinking conditions (uncross‐linked, cross‐linked in 0.22 wt% ${\mathrm{CaCl}}_2$ solution for 5 min, and cross‐linked in 1.11 wt% ${\mathrm{CaCl}}_2$ solution for 5 min) were compared. Their diameters were all around 420 $\mu \mathrm{m}$. They were placed at one opening of a long Polydimethylsiloxane (PDMS) microchannel (filled with deionized water) for observation. The microchannel diameter was 1.2 mm, a constant value for all experiments in Figure 4. Figure 4f–I, which is captured from Video S10, shows the dissolution and diffusion of Rhodamine B 2 min after the micro‐rolls were placed in the microchannel. There was no magnetic field. It can be observed that the regions near the openings of all micro‐rolls exhibited a noticeably deeper color than other areas, indicating not only drug release at the openings but also swelling and even partial dissolution of the SA. Compared with cross‐linked samples, uncross‐linked samples exhibited markedly larger deeper color areas, indicating greater drug release and more pronounced dissolution of SA. The sample cross‐linked in a 1.11 wt% ${\mathrm{CaCl}}_2$ solution for 5 min exhibited minimal deeper color area with relatively well‐defined boundaries, meaning the SA underwent slight swelling due to water absorption without dissolving. Meanwhile, the surrounding areas appeared faint in color, suggesting minimal dissolution of Rhodamine B. Such samples meet the requirements for sustained drug release. Figure 4f–II shows the temporal evolution of the concentration of rhodamine B for the three samples presented in Figure 4f–I. These measurements were repeated 3 times, and Figure 4f–II,g‐II shows representative concentration–time curves. Because Rhodamine B diffused after being released from the micro‐roll opening, its concentration varied from region to region. To allow a consistent comparison, we quantified the average concentration of Rhodamine B within a 1 mm $\times$ 1 mm square region located 200 $\mu \mathrm{m}$ to the left of the micro‐roll. For all three samples, the drug concentration increases rapidly at the beginning, indicating that the drug release rate initially exceeds the diffusion rate. For practical applications, it is desirable to minimize drug release before the carriers reach the target site, so the concentration change in the first few minutes is particularly critical. For the first 2 min, the drug concentration of the sample crosslinked in 1.11 wt% ${\mathrm{CaCl}}_2$ solution is approximately 31% of that in the non‐crosslinked sample. Crosslinking with ${\mathrm{CaCl}}_2$ solution significantly benefits slow drug release. As time progresses, the increase in concentration gradually slows down and approaches saturation after 20 min, suggesting that the drug release rate and diffusion rate have become balanced, resulting in a quasi‐steady release profile.

In addition to the crosslinking extent of SA, the diameter of the micro‐roll also affects the drug release rate, as a larger diameter means a larger contact area between the hydrogel and the liquid. We selected three micro‐rolls with different diameters and subjected them to identical drug‐loading and crosslinking procedures. These samples were then placed at the openings of a long PDMS microchannel to observe the release rate of the loaded drug (Rhodamine B). Figure 4g–I, which is captured from Video S11, shows their drug release profiles after 2 min. As indicated by the color of the liquid, a smaller diameter indeed corresponds to a slower drug release rate. Figure 4g–II shows the temporal evolution of the Rhodamine B concentration for the three samples presented in Figure 4g–I. We also measured the average concentration of Rhodamine B within a 1 mm $\times$ 1 mm square region located 200 $\mu \mathrm{m}$ to the left of the micro‐roll. For all three samples, the drug concentration increased rapidly at the initial stage, indicating that the drug release rate initially exceeded the diffusion rate. Moreover, a larger diameter corresponded to a faster rise in concentration. For the first 2 min, the drug concentration of the smallest sample is approximately 45% of that in the largest sample. Over time, the increase in concentration gradually slowed and approached saturation after 20 min, suggesting that the release and diffusion rates had reached equilibrium, thereby forming a quasi‐steady‐state release profile.

As shown in Figure 3d, the micro‐rolls can undergo self‐rotation and lateral oscillation under a RMF. These motions accelerate the drug release from the micro‐rolls, similar to how magnetic stirring speeds up the dissolution of solid substances. A drug‐loaded micro‐roll (Rhodamine B) was placed at an opening of a long microchannel. The crosslinking was immersing the micro‐roll in a 1.11 wt% ${\mathrm{CaCl}}_2$ solution for 5 min. A purely rotational magnetic field was then applied at a rotation speed of approximately 200 r/min. Figure 4h–I, which is captured from Video S12, shows the drug release behavior of a micro‐roll at 4, 19, and 43 min. The drug distribution at 4 min is particularly noteworthy. Unlike in Figure 4f, g, the released drug did not accumulate near the openings of the micro‐roll. It was instead evenly distributed along the region swept by the micro‐roll's self‐rotation and lateral oscillation. This demonstrates that the micro‐roll acted as a miniature stirring rod, enhancing drug release and diffusion within that area. Moreover, a clear boundary was observed between the colored and uncolored regions of the microchannel, corresponding closely to the region traced by the left end of the micro‐roll—again differing from the behavior observed in Figure 4f,g. Because the microchannel is extremely narrow, the liquid inside primarily exhibits laminar flow. As a result, the liquid slightly away from the micro‐roll remained nearly stationary, leading to much slower diffusion of the drug compared to the agitated region. At 19 min, the liquid near the micro‐roll remained darker and more uniform in color, indicating that stirring had accelerated dissolution. In contrast, the liquid farther from the micro‐roll appeared noticeably lighter with a visible diffusion gradient, suggesting natural diffusion. A distinct boundary could be observed between these two regions. At 43 min, the liquid near the micro‐roll remained darker and more uniformly colored, while the liquid farther away appeared significantly lighter. However, no clear boundary was observed between the two regions. At this stage, the drug within the micro‐roll was likely depleted, resulting in no further replenishment of drug into the surrounding liquid, whereas diffusion into the deeper part of the microchannel continued. Figure 4h–II, which is captured from Video S13, shows the drug release behavior at 4 min for three drug‐loaded micro‐rolls with different diameters. The crosslinking conditions and magnetic field settings were the same as previously described. In all three samples, the released drug was uniformly distributed within the region swept by the micro‐roll's self‐rotation and lateral oscillation. A clear boundary was also observed between the colored and uncolored regions of the microchannel, indicating that this phenomenon is generally consistent. The Rhodamine B concentration in these regions was one to two orders of magnitude higher than that in more distant regions. This phenomenon facilitates localized drug release: by controlling the drug‐loaded micro‐roll to continuously rotate and oscillate at a designated position within the microchannel, targeted drug release can be achieved.

Video S14 and Figure 4i illustrate the process of localized drug release from a drug‐loaded micro‐roll inside a microchannel. The micro‐roll was first driven rapidly to the left in a gradient RMF (Figure 4i–I). After it reached the designated position, the field was switched to a purely RMF, causing the micro‐roll to undergo nearly pure rotation (Figure 4i–II). After several minutes of rotation, the liquid in the vicinity of the micro‐roll became visibly red, while the rest of the channel showed no obvious change (Figure 4i–III). Finally, the magnetic field was adjusted to guide the micro‐roll back along its original path (Figure 4i–IV). In the next section, Rhodamine B is replaced by antibiotics, while keeping the remaining steps unchanged, to achieve targeted bacterial killing within the microchannel.

2.5. Drug‐Loaded Micro‐Rolls for Targeted Biofilm Eradication in Medical Silicone Tubes

Biofilm formation in medical silicone tubes can cause lumen occlusion and recurrent infection, resulting in unnecessary implant removal and, in some cases, severe life‐threatening bloodstream infection. In the previous section, we demonstrated the fixed‐point drug release capability of the drug‐loaded micro‐rolls. In addition, the micro‐rolls can perform lateral oscillatory motion under an RMF, continuously impacting the walls of the tubes. The micro‐roll samples shown in Videos S12–S14 can collide with the inner wall thousands of times under the RMF without structural damage. Figure S12 shows SEM comparison of tube regions with and without such prolonged micro‐roll impacts, revealing markedly reduced biofilm coverage in the treated areas but no obvious additional damage to the inner surface, indicating that repeated collisions do not irreversibly damage the silicone substrate under the tested conditions. Therefore, the drug‐loaded HEA/LIG micro‐rolls could actively navigate within silicone tubes, mechanically disrupt the extracellular matrix, and enhance local delivery of antimicrobial agents, offering a promising minimally invasive strategy for in situ biofilm eradication in such devices.

Pancreatic duct stents are commonly used to treat main pancreatic duct strictures, but long‐term stenting predisposes to bacterial biofilm formation, resulting in recurrent infections and reduced stent patency [58]. Figure 5a depicts the anatomical relationship between the pancreas and duodenum, including the confluence and extension of the main pancreatic duct and main bile duct into the duodenal lumen. The location of the pancreatic duct stent within the main pancreatic duct is highlighted. A magnified inset focuses on the stent near the narrowest area of the main pancreatic duct, showing biofilm deposits attached to the inner wall of the stent, thus emphasizing stent‐associated microbial colonization within the pancreatic duct system.

FIGURE 5.

Biofilm eradication in medical silicone tubes (used as pancreatic duct stents). (a) An illustration of a pancreatic duct stent in the human body, with the inset showing potential biofilm formation on the inner surface after long‐term implantation. (b) Schematic of PBS rinsing through a curved silicone tube (with an attached biofilm), a micro‐roll rotates and moves toward the biofilm in an RMF, and the micro‐roll passes through and disrupts the biofilm (some biofilm attached to the micro‐roll ends and sidewall). (c) A tilted view (30°) of a medical silicone tube (I); Two HEA/LIG micro‐rolls loaded with SA gel (containing triple‐antibiotic drug) and subsequently air‐dried (II). (d) E. coli bacterial aggregate and biofilm in LB broth (I) and air (II), PBS rinsing of the tube (III), before (IV), during (V), and after (VI) disruption of bacterial aggregate and biofilm with a drug‐loaded HEA/LIG micro‐roll in an RMF, then before (VII) and after (VIII) PBS rinses again, and the rinse fluid (left: fluid after the first rinse; middle: fluid after the tenth rinse; right: fluid after biofilm disruption and the first rinse again) (IX). (e) A drug‐loaded HEA/LIG micro‐roll contacts E. coli biofilm (I), pulls up biofilm (II), moves and continuously pulls up biofilm (III), peels off the last bit of the attached biofilm (IV), carries biofilm and moves (V), and stops moving and biofilm becomes a mass (VI). (f) The silicone tubes bearing sparse biofilms underwent four treatments; The final rinsed solution was diluted 100‐fold and then coated onto LB agar plates; The viable E. coli percentage for each treatment was calculated (mean $\pm$ SD, n = 4 independent samples per treatment). The scale bar (c–e) is 1 mm.

Figure 5b illustrates the process of disrupting a pre‐existing biofilm within a curved microchannel using a drug‐loaded microrobot. In the upper figure, a continuous biofilm is arranged along the flow direction of phosphate‐buffered saline (PBS) on the inner wall of the silicone microchannel. Typically, bacterial cells, biofilms, secretions, etc., clump together in microchannels. When flushed with liquid, most of the bacterial cell clumps and a small amount of other substances are washed away, but the biofilm, a few clumps, and other secretions can adhere to the inner wall of the microchannel. In the middle Figure, the silicone microchannel is straightened. A biofilm is left on the inner wall in the middle of the microchannel, while the drug‐loaded micro‐roll translates along the channel axis at a velocity ${\mathrm{v}}_{//}$ under a gradient RMF, while simultaneously rotating at an angular velocity $\omega$, eventually contacting the biofilm. In the lower Figure, after the biofilm is disrupted, it no longer exists in a compact form but diffuses along the microchannel as dispersed fragments, while some biofilm adheres to the openings and sidewalls of the micro‐roll. In reality, the biofilm initially adheres to the openings and sidewalls of the micro‐roll. During micro‐roll's rotation and lateral oscillation, the adhered biofilm causes more biofilm to be torn off. As the micro‐roll moves forward, the entire biofilm is gradually mechanically broken up. Here, we use medical‐grade silicone tubes (pancreatic duct stents) as a demonstration. As shown in Figure 5c–I, its inner diameter is approximately 1 mm. Figure 5c–II, III presents two drug‐loaded HEA/LIG micro‐rolls, where triple‐antibiotic (Penicillin‐Streptomycin‐Amphotericin B) solution was used in place of Rhodamine B, while all other drug‐loading steps remained unchanged. Figure S8a shows dried SA films containing antibiotics and Rhodamine B. Figure S8b shows SA (containing antibiotics) loading onto micro‐rolls of different diameters. Considering that the SA hydrogel may swell upon water absorption, potentially increasing the micro‐roll diameter, micro‐rolls with diameters below 400 $\mu \mathrm{m}$ were selected for the experiments.

A medical‐grade silicone tube (a pancreatic duct stent) with a length of approximately 8 cm was used. LB medium containing E. coli was injected into the tube and incubated for 96 h to form abundant bacterial cell aggregates and biofilms. Figure 5d and Video S15 illustrate how the drug‐loaded micro‐rolls were used to clean medical tubes heavily covered with bacterial aggregates and biofilms. Figure 5d–I shows a short segment of the LB medium, where the dark region indicated by the black arrow corresponds to bacterial aggregates, biofilms, and other secretions. Subsequently, PBS solution was injected into the tube to replace the LB broth. Due to the presence of air bubbles in parts of the tube, Figure 5d–II shows the moment when the LB broth and some bacterial cells had been flushed out, while the incoming PBS solution had not yet reached that region. The areas indicated by the two black arrows correspond to sites where biofilms adhered, showing a noticeable increase in thickness and a slightly yellowish film. It can be observed that the biofilm was firmly attached to the inner wall of the tube and was not carried away by the flow of LB broth or its surface tension. The silicone tube could hold approximately 50 $\mu \mathrm{L}$ of liquid. It was rinsed with 30 times its volume of PBS solution for about 3 min. Figure 5d–III shows a representative scene during the rinsing process, where the black arrow indicates the flow direction of the PBS solution. The dark region inside the tube can be roughly divided into two areas, marked by blue and red arrows, respectively. The area indicated by the blue arrow shows a tendency to be flushed away by the PBS flow, while the area indicated by the red arrow remains firmly attached to the inner wall and appears to anchor the blue‐arrow region. Considering that biofilms tend to adhere to surfaces whereas bacterial cell aggregates are more easily removed, the blue‐arrow region likely contains more bacterial aggregates, whereas the red‐arrow region is richer in biofilm material. Overall, the residual dark regions not washed away by PBS represent a mixture of bacterial aggregates and biofilms.

Figure 5d and Video S15 illustrate the disruption of dense bacterial aggregates and biofilms inside a silicone tube using a drug‐loaded HEA/LIG micro‐roll. The bent silicone tube was locally straightened to facilitate precise control of the micro‐roll's motion, which was first guided to the dark, highly contaminated region under magnetic attraction and then driven through this region under a gradient rotating magnetic field. As the micro‐roll translated and rotated through the dark region, it mechanically disturbed and fragmented the adherent aggregates and biofilms, which detached from the inner wall and were transported into the lumen and into the effluent, whereas simple PBS rinsing alone left most of the firmly attached biofilm intact. Together with the colony‐count data in Figure 5f, these observations indicate that micro‐roll‐mediated mechanical debulking markedly enhances removal of biofilm biomass compared with PBS rinsing or antibiotics alone.

In a more clinically relevant situation, bulk bacterial aggregates and loosely attached clusters can be flushed out, leaving only a thin, tightly adherent biofilm on the inner surface of the silicone tube. Calcium‐crosslinked SA at the opening of the HEA/LIG micro‐roll absorbs water, forms a hydrated and slightly adhesive gel, and thus provides a soft interface for capturing such thin biofilms. As shown in Figure 5e and Video S16, when a drug‐loaded micro‐roll was guided to the biofilm region and actuated under a gradient rotating magnetic field, its combined forward translation, self‐rotation, and lateral oscillation caused the SA gel at the front end to repeatedly contact the inner wall, adhere to the biofilm, lift it from the substrate, and progressively peel and wrap the biofilm around the micro‐roll opening. This behavior is consistent with the viscoelastic nature of biofilms: the EPS‐based network can withstand transient shear, but repeated lateral oscillation and rotation of an adhered micro‐roll generate sufficient tensile and peeling forces to progressively detach the biofilm from the tube surface. As a result, most of the initially invisible, tightly attached E. coli biofilm was removed from the inner wall, confirming that the micro‐rolls can mechanically debulk both thick aggregates and thin adherent biofilms while simultaneously carrying the detached biomass and releasing antibiotics.

From Videos S15 and S16, the physical disruption of biofilms and bacterial aggregates by the drug‐loaded micro‐rolls can be clearly observed. In addition, the use of drugs (such as antibiotics) to inhibit or even kill bacteria should also be considered. Liquids flushed from medical tubes after different treatments were spread onto LB agar plates to evaluate the bacterial killing efficiency. The control treatment was rinsing the tube lumen with 60 times its volume of PBS solution and collecting the final 100 $\mu \mathrm{L}$ of the effluent. For the “PBS + Antibiotics” treatment, the tube lumen was rinsed with 60 times its volume of PBS solution containing antibiotics, and the final 100 $\mu \mathrm{L}$ of the effluent was collected. The concentration of the antibiotics in an equal volume of PBS solution after the drug‐loaded micro‐roll fully released the drug was estimated and used to prepare the antibiotic concentration in PBS solution. For the “PBS + Micro‐roll + Antibiotic‐loaded SA gel” treatment, the tube lumen was rinsed with 30 times its volume of PBS solution, followed by a drug‐loaded micro‐roll placed near the biofilm area and static release of drug for 10 min, then the tube lumen was rinsed again with 30 times its volume of PBS solution, and the final 100 $\mu \mathrm{L}$ of the effluent was collected. For the “PBS + Micro‐roll + RMF + Antibiotic‐loaded SA gel” treatment, the tube lumen was rinsed with 30 times its volume of PBS solution, followed by a drug‐loaded micro‐roll placed near the biofilm area, then the micro‐roll was led to mechanically disrupt the biofilm under the gradient RMF, during which the drug should also be released to do chemical disinfection. Then the tube lumen was rinsed again with 30 times its volume of PBS solution, and the final 100 $\mu \mathrm{L}$ of the effluent was collected. The above collected effluent was then diluted 100 times with PBS solution and evenly spread on LB agar plates. Each treatment corresponded to 4 samples. After incubation at 37 for 24 h, the bacterial colonies on these LB agar plates were counted. Using the average colony count of the control group as a baseline, the bacterial survival rates under other treatments were calculated. Figure 5f shows representative LB agar plates after 24 h of incubation and the corresponding viable E. coli statistics. Compared with rinsing the medical tube with PBS alone, rinsing with antibiotic‐containing PBS achieved a sterilization rate of about 53%, static antibiotic release from drug‐loaded micro‐rolls near the biofilm achieved about 82%, and RMF‐driven mechanical disruption combined with antibiotic release achieved about 97%. It can be seen that mechanical disruption of biofilms and bacterial aggregates, combined with the use of antibiotics, results in substantially fewer colonies, indicating the effectiveness of physical‐chemical hybrid sterilization.

From a translational perspective, the chemical composition of pancreatic fluid, including digestive enzymes and ionic conditions, may modulate the swelling and long‐term stability of ${\mathrm{Ca}}^{2+}$‐crosslinked SA hydrogels compared with the deionized water and PBS conditions used in the present in vitro release tests. Nevertheless, previous studies have shown that ${\mathrm{Ca}}^{2+}$‐alginate gels generally retain their gel‐like integrity over days to weeks in physiological environments in the absence of specific alginate lyases [59, 60], even though some gradual softening or partial degradation may occur, suggesting that rapid catastrophic degradation is unlikely within the short treatment window considered here. Consistent with this expectation, our additional experiments using artificial pancreatic fluid instead of PBS for flushing the biofilm‐contaminated tubes confirm that the drug‐loaded micro‐rolls still effectively disrupt biofilms and reduce bacterial viability under enzyme‐ and salt‐containing conditions (Video S17).

To further assess the sterilization performance under polymicrobial conditions, we also evaluated mixed‐species biofilms composed of E. coli, P. aeruginosa and S. aureus in artificial pancreatic fluid. After formation of the mixed biofilm inside the silicone tubes, the lumen was flushed with 20‐fold tube volume of artificial pancreatic fluid to remove planktonic cells and loosely attached aggregates while preserving adherent biofilm on the inner wall, followed by disinfection with drug‐loaded HEA/LIG micro‐rolls under the same magnetic actuation protocol (Video S17). Before and after treatment, the tubes were flushed again with artificial pancreatic fluid and the final effluent was collected, diluted and plated on species‐selective agar to independently quantify the survival of E. coli, P. aeruginosa, and S. aureus. The post‐treatment colony counts decreased to approximately 11.1%, 12.0% and 0.8% of the corresponding pre‐treatment values, respectively (Figure S13), which corresponds to approximately a one‐order‐of‐magnitude reduction for E. coli and P. aeruginosa and more than a two‐order‐of‐magnitude reduction for S. aureus. These results indicate that the combined mechanical disruption and localized triple‐antibiotic release achieved by the micro‐rolls can effectively reduce mature polymicrobial biofilms in an artificial pancreatic fluid environment within the tested time window.

In the present in vitro model, the artificial pancreatic fluid did not contain endogenous antimicrobial peptides or additional antibacterial agents, and the tubes were maintained under essentially static conditions after treatment. Under these conservative post‐treatment conditions, a 24 h follow‐up culture of the lumen fluid revealed renewed growth of E. coli, P. aeruginosa and S. aureus compared with the immediate post‐treatment state, with pronounced lysis observed in the E. coli colonies. These observations indicate that the current microrobot‐assisted treatment primarily demonstrates short‐term debulking and bactericidal efficacy, whereas long‐term suppression of biofilm recurrence in vivo will depend on the combined effects of microrobot intervention, antibiotic regimens and host factors such as continuous ductal flow and innate antibacterial components.

In addition to the antibacterial performance, a preliminary in vitro cytocompatibility assessment of the HEA/LIG micro‐rolls' eluates was carried out using a CCK‐8 assay with L929 mouse fibroblasts over a series of extract concentrations from 0.0001 to 0.01 g/mL. At low and intermediate extract concentrations (0.0001–0.0025 g/mL), the relative viability of L929 cells remained above 95% of the untreated control (about 99.8%, 98.4%, and 95.9%, respectively), indicating negligible impact on cell proliferation under these conditions, whereas at higher extract concentrations a clear concentration‐dependent decrease was observed, with the viability decreasing to approximately 86.1% and 79.6% at 0.005 and 0.01 g/mL, respectively (Figure S14). These results suggest that the eluates of the high‐entropy alloy layer induce only moderate cytotoxic effects even under conservative, worst‐case exposure conditions. We speculate that relatively mild cytotoxicity is associated with the immobilization of metallic components within a solid micro‐roll framework, which limits their dissolution. Nevertheless, future designs will aim to further optimize the magnetic material composition toward more biocompatible formulations.

From an application standpoint, although the present results demonstrate effective mechanical disruption of biofilms and localized antibacterial treatment in confined silicone tubes, the downstream fate of the detached biofilm fragments remains a factor that should be considered in future translational studies. In clinical practice, obstructed pancreatic stents are generally treated by endoscopic intervention to restore luminal patency and to remove retained debris or infected materials [61]. Therefore, the current microrobot system mainly serves as a proof‐of‐concept platform for local biofilm disruption and treatment. Further studies are needed to examine how such microrobot‐assisted disruption may be combined with clinically relevant fluid‐clearance procedures under more realistic physiological conditions [62]. In this proof‐of‐concept study, curved silicone tubes were locally straightened to provide a simple and reproducible platform for quantifying micro‐roll propulsion, guidance, and drug release under a gradient rotating magnetic field. The same magnetic actuation principle could, in principle, be applied in curved ducts by reorienting the external permanent magnets so that the resultant magnetic field is locally aligned with the local duct axis. A detailed implementation of adaptive field steering in patient‐specific duct geometries is beyond the scope of this work and will be investigated in future translational studies.

3. Conclusion

This study proposes a laser‐guided self‐assembly strategy for fabricating 3D HEA/LIG micro‐rolls from 2D HEA/PI bilayer films, achieved through simple DLW setting optimization and interface adjustment. The micro‐rolls could recover their original shape after undergoing 90% uniaxial compressive deformation. These magnetic micro‐rolls were used to load gel‐based drugs and achieved localized bactericidal effects through a combination of physical removal and chemical disinfection. The magnetically driven microrobots, fabricated via laser‐guided self‐assembly, showed an external diameter ranging from approximately 441 to 553 $\mu \mathrm{m}$ and a thickness of 25 to 50 $\mu \mathrm{m}$. Through optimized design, we engineered HEA/LIG micro‐rolls with diameters smaller than the inner diameter of the target microchannels. They exhibited a saturation magnetization of approximately 2.08 to 2.77 emu ${\mathrm{g}}^{-1}$, imparted by the HEA layer. These magnetic micro‐rolls were controllably loaded with SA gel containing Rhodamine B or antibiotics. The HEA/LIG micro‐rolls demonstrated precision navigation through a 700 $\mu \mathrm{m}$‐diameter microchannel with an average swimming speed of 1 mm ${\mathrm{s}}^{-1}$. Immersing SA gel in a 1.11 wt% ${\mathrm{CaCl}}_2$ solution for 5 min can reduce the average drug release rate by nearly 70% in the first 2 min; at the same time, the rotation of the micro‐roll (e.g., 200 r ${\min }^{-1}$) and lateral oscillation can ensure rapid local release of the drug at the target site, with a drug concentration one to two orders of magnitude higher than that in the surrounding area. The drug‐loaded micro‐rolls were capable of mechanically clearing medical silicone tubes that were heavily contaminated with biofilms and bacterial aggregates, as well as medical tubes bearing only sparse, barely detectable biofilms tightly adhered to the tube wall. By coupling mechanical clearance with simultaneous antibiotic release, a hybrid bactericidal mode of 97% E. coli killing rate was achieved, whereas that of the antibiotic‐containing PBS was only 53%. This scalable laser fabrication approach achieves good yields exceeding 180 units ${\mathrm{h}}^{-1}$ at material costs less than $0.01 per unit, representing a paradigm shift in large‐scale magnetically driven micromotor production.

4. Experimental Section

4.1. Sample Pretreatment

First, a cleaned 25 $\mu \mathrm{m}$‐thick PI film was coated with a thin layer of HEA ((MnFeCoNiCu, 20 mol% of each element, 160 nm thick) by sputter deposition (ARC12M, Plasma Sciences, USA). Then, a 0.5 wt% PVA solution was applied to a cleaned glass surface, followed by the attachment of the HEA/PI bilayer thin film to it (HEA side exposed to air). Lastly, the sample was heated on a hot plate (65, 25 min), and a precise scraper was used to scrape the film every 5 min to ensure that the surface of the film was flat and in close contact with the glass.

4.2. Laser‐Guided Self‐Rolling

This step consists of two substeps. The first is to define the pattern to be rolled. A Nd:YAG commercial laser marker was used. The laser settings include continuous mode, 8.8 W laser power, a scanning speed of 400 mm ${\mathrm{s}}^{-1}$, and writing 50 times. The second is that a much lower laser power (to prevent ablation) was applied to scan the patterned thin film line by line, during which PI was converted to LIG, and HEA/LIG was peeled, after which the film rolled up. After optimization, the laser power range was 1.18 to 1.50 W. Other laser settings include a scanning speed of 400 mm ${\mathrm{s}}^{-1}$, a line spacing of 3 $\mu \mathrm{m}$, and a single writing pass.

4.3. Uniaxial Compression Characterization

The HEA/LIG micro‐roll, 3D printed microtube, microspring, and reinforced tube were positioned horizontally and subjected to a uniaxial compressive load applied perpendicularly to their side surface, resulting in deformation exceeding 90% along that axis (ZP‐20, Ailigu, Hong Kong). The load was maintained for several seconds before being withdrawn. The 3D‐printed samples were fabricated using a precise printer (S230A, Boston MicroFabrication Corp., USA) and a matching photosensitive resin.

4.4. SEM and EDS Characterization

SEM images and EDS mapping images were obtained using a field‐emission scanning electron microscope (JSM‐7800F, JEOL, Japan).

4.5. VSM Characterization

VSM test was performed using a magnetic measurement system (Physical Property Measurement System, Quantum Design, USA).

4.6. Raman Characterization

Raman spectra were obtained with a Renishaw inVia Qontor Spectrometer System (RENISHAW, UK). The wavelength of the light source was 532 nm. The power was 2.5 mW.

4.7. XPS Characterization

XPS test was performed using a photoelectron spectrometer (Axis Ultra DLD, Kratos Analytical, UK). The sample used for XPS characterization was a square glass sheet with a side length of 5 mm, on which HEA/LIG micro‐rolls (prepared at a laser power of 1.18 W) were densely stacked.

4.8. UV–Vis Characterization

A UV–Vis test was performed using a UV/Vis/NIR spectrophotometer (Lambda 1050+, PerkinElmer, USA).

4.9. Rotating Magnetic Field Establishment

Two N52 magnets (60 mm $\times$ 20 mm $\times$ 10 mm) were inserted into a holder. The holder was controlled by a stepper motor. The stepper motor was connected to an adjustable power supply. The rotation speed of the magnet was controlled by adjusting the current and voltage. The rotation speed range was about 50 to 250 rpm. The holders for the magnets and the stepper motor were modeled with SolidWorks and then printed with a 3D printer (P1S, Bambu Lab, China). The printing material was commercial white PLA. Two types of magnet holders were used. In one design, the two magnets were aligned in parallel to generate a gradient‐free rotating magnetic field (RMF), whereas in the other, the magnets were fixed with a 10° angle between them to generate a gradient RMF.

4.10. Microchannel Fabrication

The microchannels were molded from long sewing needles. The mold material was PDMS (Sylgard 184, Dow, USA) or Ecoflex (00‐30, Smooth‐On, USA). The microchannels, with a diameter of 0.7 or 1.2 mm, were used to place magnetic microrolls and observe their movement and drug release within the microchannels.

4.11. Drug Solution Preparation

The drugs used were Rhodamine B (Macklin, China) or triple‐antibiotic solution (10 000 units/mL penicillin, 10 000 $\mu \mathrm{g}$/mL streptomycin, and 25 $\mu \mathrm{g}$/mL amphotericin B) (MiNE, China). Rhodamine B was dissolved in a SA solution under magnetic stirring to facilitate complete dissolution. The final solution contained 1 wt% Rhodamine B and 5 wt% SA. The triple‐antibiotic solution was dissolved in SA solution. The final solution contained 2000 units/mL penicillin, 2000 $\mu \mathrm{g}$/mL streptomycin, 5 $\mu \mathrm{g}$/mL amphotericin B and 2 wt% sodium alginate.

4.12. Loading and Release of Rhodamine B

HEA/LIG micro‐rolls were attached to one corner of a rectangular Ecoflex block, leaving both ends suspended. A pair of tweezers was used to pick up the drug‐containing SA solution and apply it to each end, and additional solution was applied after the liquid had retracted into the micro‐roll cavity. During this process, the solution gradually dried in the air and transformed into a gel. This procedure was repeated until the gel drug completely filled the micro‐roll cavity, followed by air‐drying for 2 h. The effect of different crosslinking conditions on drug release was evaluated. 0.2 wt% and 1.11 wt% ${\mathrm{CaCl}}_2$ solutions were prepared, and samples with comparable dimensions were immersed in these solutions for 5 min to crosslink the SA. The samples were then removed from the ${\mathrm{CaCl}}_2$ solutions and air‐dried again for 2 h. Subsequently, the samples were placed at one opening of a PDMS microchannel filled with deionized water, and their drug release behavior was recorded. The effect of micro‐roll diameter on drug release was investigated. Samples with different diameters were immersed in 1.11 wt% ${\mathrm{CaCl}}_2$ solution for 5 min for crosslinking, removed, and air‐dried for 2 h. A subset of these samples was then placed at one opening of a PDMS microchannel filled with deionized water, and the corresponding drug release processes were recorded. The influence of micro‐roll self‐rotation on drug release was examined. Samples with different diameters were immersed in 1.11 wt% ${\mathrm{CaCl}}_2$ solution for 5 min to achieve crosslinking, removed, and air‐dried for 2 h. A subset of the samples was subsequently placed at one opening of a PDMS microchannel filled with deionized water, and an RMF was applied to induce self‐rotation and lateral oscillation of the samples. The drug release behavior of the samples under these conditions was recorded.

4.13. Targeted Release of Rhodamine B

Gel‐based drug was loaded and crosslinked as described above, and the drug‐loaded micro‐roll was placed at one opening of the PDMS microchannel. A gradient RMF was applied to propel the micro‐roll into the channel to a prescribed position, after which the field was switched to a gradient‐free RMF to induce nearly pure self‐rotation with lateral oscillation. After several minutes, a pronounced local color change of the surrounding fluid was observed. The magnetic field was then switched to a reversed gradient RMF to drive the micro‐roll back along its original path. The magnetic field rotation speed was always 200 rpm.

4.14. Preparation of LB Broth

We took 20 g of lysogeny broth (LB) (without sugar) (HuanKai Microbial, China), added 1 L of deionized water, stirred for 10 s, boiled at high temperature until completely dissolved, and sterilized meanwhile (120, 30 min, in an autoclave), cooled, and set aside.

4.15. Preparation of Agar Plate

We took 36 g of LB Agar (HuanKai Microbial, China), added 1L of deionized water, shook for 10 s, then boiled at high temperature until completely dissolved and sterilized meanwhile (120, 30 min, in an autoclave). Then, the LB Agar solution was poured onto a sterilized petri dish and spread evenly with an inoculation rod. After being cooled, an LB agar plate was formed. In our research, we poured about 6.5 mL of LB Agar solution into each petri dish (6 cm diameter).

4.16. Preparation of PBS Solution

The PBS solution contained 0.28 wt% potassium dihydrogen phosphate and 0.14 wt% dipotassium hydrogen phosphate. The solution was sterilized in an autoclave at 120 for 30 min, then cooled to room temperature and stored until use.

4.17. E. Coli Culture in the Microchannel

We took 0.3 mL of the standard E. coli culture solution and added it to 7.5 mL of LB broth, vortexed to mix well, and then placed it in a shaker (150 rpm, 37) for 18 h to regenerate. We then repeated the above steps to obtain a new E. coli broth. Then we cut a medical‐grade silicone tube (provided by Shanghai Renji Hospital, commonly used for pancreatic duct stents) into 6.5 cm segments with an inner diameter of approximately 1 mm. Each tube was then filled with approximately 50 $\mu \mathrm{L}$ of E. coli broth that had been cultured for 24 h. Then, the tubes were bent into a U‐shape to facilitate the gathering of bacterial aggregates at the bottom of the U‐shape. We placed the tubes in 3D printed cylindrical containers (inner diameter: 4 cm). A humidifying solution was prepared by mixing 34 g of potassium chloride, 15 mL of glycerol, and 100 mL of ultrapure water, followed by vortexing at 800 rpm for 30 min (Vortex mixer, Joan Lab, China), and the supernatant was collected. Filter paper was cut to match the inner dimensions of the cylinder and briefly immersed in the humidifying solution, then removed to ensure no visible droplets remained on the surface. The filter paper was laid flat on the bottom of the cylinder. The samples were then incubated at 37. 50 $\mu \mathrm{L}$ of the humidifying solution was added onto the paper every 6 h. After 48 h, the inner surface of the silicone tubes exhibited loosely attached bacterial cells and biofilms, whereas after 96 h, a large amount of firmly attached bacterial cells and mature biofilms was observed.

4.18. Eradication of Bacterial Aggregates and Biofilm With a Drug‐Loaded Micro‐Roll

The medical tube containing bacterial broth was incubated for 96 h. Then the tube was rinsed with 1.5 mL PBS for 3 min, after which substantial bacterial aggregates and biofilms remained visible inside the silicone tube. The initial 0.15 mL and the final 0.15 mL of the rinsing PBS solution were collected for subsequent analysis. The tube was then straightened, and a drug‐loaded micro‐roll was placed at the tube opening and magnetically guided to a position approximately 1 cm from the target bacterial aggregate. A gradient RMF of approximately 200 rpm was applied to propel the micro‐roll forward, inducing simultaneous self‐rotation and lateral oscillation. The micro‐roll progressively contacted and disrupted the bacterial aggregates and biofilms, during which its translational speed noticeably decreased. Maintaining the same gradient field, the micro‐roll eventually passed through the fouled region after several minutes. The direction of the magnetic field was then reversed twice so that the micro‐roll repeatedly traversed this region to further disrupt the aggregates and biofilms. Finally, the tube was rinsed again with 1.5 mL PBS, and the initial 0.15 mL and final 0.15 mL of the effluent were collected. The micro‐roll was also rinsed out of the tube, and flocculent residues were clearly observed on its end face.

4.19. Eradication of Weakly Visible Biofilm From the Inner Surface of Silicone Tube

The medical‐grade silicone tube containing bacterial broth was incubated for 48 h. Then it was rinsed with 1.5 mL PBS for 3 min, after which only biofilms adhered to the inner wall remained visible. The tube was straightened, and a drug‐loaded micro‐roll was placed at the opening and magnetically guided to a position approximately 1 cm from the target biofilm. A gradient RMF (200 rpm) was applied to drive the micro‐roll forward while inducing simultaneous self‐rotation and lateral oscillation. The leading edge of the micro‐roll gradually contacted and peeled off the biofilm, and continued propulsion eventually resulted in complete detachment, with the micro‐roll carrying the peeled biofilm downstream. Finally, the tube was rinsed again with 1.5 mL PBS; the micro‐roll was rinsed out of the tube, and flocculent residues were clearly observed on its end face.

4.20. Comparison of E. coli Killing Efficiency Under Different Treatments

Four treatment protocols were evaluated for their bactericidal efficacy in silicone tubes bearing sparse biofilms. (1) The tube was rinsed with 3 mL PBS, and the final 100 $\mu \mathrm{L}$ of the effluent was collected. (2) A mixed solution was prepared by adding a small amount of triple‐antibiotic solution to PBS to obtain final concentrations of 1 unit/mL penicillin, 1 $\mu \mathrm{g}$/mL streptomycin, and 0.0025 $\mu \mathrm{g}$/mL amphotericin B. The tube was then rinsed with 3 mL of this solution, and the final 100 $\mu \mathrm{L}$ of the effluent was collected. (3) The tube was rinsed with 1.5 mL PBS first. Then a drug‐loaded micro‐roll was magnetically guided to the vicinity of the biofilm and left static for 10 min without applying an RMF, after which the tube was rinsed with 1.5 mL PBS, and the final 100 $\mu \mathrm{L}$ was collected. (4) The tube was rinsed with 1.5 mL PBS first. Then, a drug‐loaded micro‐roll was magnetically guided to the biofilm region, and a gradient RMF was applied to mechanically disrupt the biofilm. The tube was then rinsed with 1.5 mL of PBS, and the final 100 $\mu \mathrm{L}$ was collected. The collected effluents were diluted 100 times with PBS solution, and 50 $\mu \mathrm{L}$ of each dilution was spread uniformly onto agar plates using a spreader. The plates were incubated at 37 for 24 h and then imaged. The colony‐forming units (CFUs) on each plate were quantified using OpenCFU. The viable E. coli percentage for each treatment was calculated relative to the average CFU count obtained from treatment (1).

4.21. Polymicrobial Biofilm Formation and Disinfection in Artificial Pancreatic Fluid

A mixed bacterial suspension containing E. coli, P. aeruginosa and S. aureus was prepared by inoculating each strain into LB broth and incubating at 37 with shaking until the logarithmic growth phase, followed by mixing at equal volume ratios. Medical‐grade silicone tubes (inner diameter $\approx$ 1 mm) were filled with approximately 50 $\mu \mathrm{L}$ of the mixed LB culture and incubated at 37 for 96 h to allow formation of polymicrobial biofilms and bacterial aggregates on the inner wall. The tubes were then bent into a U‐shape and placed in sealed humidified containers as described above to prevent evaporation. After biofilm formation, the lumen was flushed with artificial pancreatic fluid at a volume equal to 20 times the tube volume to replace the LB medium and to remove planktonic bacteria and loosely attached aggregates, while retaining adherent biofilm on the tube wall. For disinfection, a drug‐loaded HEA/LIG micro‐roll containing the triple‐antibiotic‐loaded SA gel was introduced into the tube and magnetically guided to the fouled region. A gradient rotating magnetic field (about 200 rpm) was applied to drive the micro‐roll forward with simultaneous self‐rotation and lateral oscillation, enabling mechanical disruption of the polymicrobial biofilm and local antibiotic release in artificial pancreatic fluid (Video S17). Before and after microrobot treatment, the tube lumen was flushed with 20‐fold tube volume of artificial pancreatic fluid, and the final 100 $\mu \mathrm{L}$ of each effluent was collected and diluted 100‐fold with PBS. Aliquots of 50 $\mu \mathrm{L}$ were then spread onto cetrimide agar, MacConkey agar and mannitol–salt agar plates to selectively culture P. aeruginosa, E. coli, and S. aureus, respectively. After incubation at 37 for 24 h, the colony‐forming units on each plate were counted, and the viable cell fraction for each species was calculated relative to the corresponding pre‐treatment effluent.

4.22. In Vitro Cytotoxicity Test Based on CCK‐8 Assay

L929 mouse fibroblast cells were used to evaluate the in vitro cytotoxicity of the HEA–LIG micro‐rolls by a CCK‐8 assay. Cells in the logarithmic growth phase were harvested, counted, and seeded into 96‐well plates at a density of $6\times {10}^3$ cells per well, followed by incubation overnight in a humidified incubator (37°C, 5% ${\mathrm{CO}}_2$) to allow cell attachment. The samples were extracted in complete culture medium to obtain a series of working solutions at nominal concentrations of 0.00001, 0.0001, 0.0025, 0.005, and 0.01 g ${\mathrm{mL}}^{-1}$, and the cells were divided into a control group and several experimental groups. For the control group, 100 $\mu \mathrm{L}$ of fresh complete medium was added to each well, while for the experimental groups, 100 $\mu \mathrm{L}$ of the corresponding sample working solution was added; three replicate wells were used for each condition. After 24 h of incubation (37°C, 5% ${\mathrm{CO}}_2$), the medium in each well was removed and the cells were gently washed three times with PBS. Subsequently, 100 $\mu \mathrm{L}$ of complete medium containing 10% (v/v) CCK‐8 reagent was added to each well, and the plates were incubated for an additional 2 h under the same conditions. The absorbance at 450 nm was then measured using a microplate reader. Then the relative cell viability (%) was calculated.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File: smll74469‐sup‐0001‐SuppMat.pdf.

Supporting Video 1: smll74469‐sup‐0002‐VideoS1.mp4.

Supporting Video 2: smll74469‐sup‐0003‐VideoS2.mp4.

Supporting Video 3: smll74469‐sup‐0004‐VideoS3.mp4.

Supporting Video 4: smll74469‐sup‐0005‐VideoS4.mp4.

Supporting Video 5: smll74469‐sup‐0006‐VideoS5.mp4.

Supporting Video 6: smll74469‐sup‐0007‐VideoS6.mp4.

Supporting Video 7: smll74469‐sup‐0008‐VideoS7.mp4.

Supporting Video 8: smll74469‐sup‐0009‐VideoS8.mp4.

Supporting Video 9: smll74469‐sup‐00010‐VideoS9.mp4.

Supporting Video 10: smll74469‐sup‐00011‐VideoS10.mp4.

Supporting Video 11: smll74469‐sup‐00012‐VideoS11.mp4.

Supporting Video 12: smll74469‐sup‐00013‐VideoS12.mp4.

Supporting Video 13: smll74469‐sup‐00014‐VideoS13.mp4.

Supporting Video 14: smll74469‐sup‐00015‐VideoS14.mp4.

Supporting Video 15: smll74469‐sup‐00016‐VideoS15.mp4.

Supporting Video 16: smll74469‐sup‐00017‐VideoS16.mp4.

Supporting Video 17: smll74469‐sup‐00018‐VideoS17.mp4.

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.