3.1.

Fabrication of the Optical Fibers

To investigate PTEP induced via POFs, three different optical fibers were prepared—bare fiber (BF), gold-coated fiber (GCF), and plasmonic fiber (PF), based on silica multi-mode optical fibers ($200/220\mu \mathrm{m}$ core/cladding diameter, 0.22 NA, FG200LEA, Thorlabs, Inc., Germany). All fibers were cut to 96 mm long using an automatic optical fiber cleaver (CT-101, Fujikura Ltd., Japan). Gold thin film (50 nm) deposition for the GCF was carried out by sputtering (HEX thin film deposition system, Korvus Technology, United Kingdom). The PF [Fig. 1(a)] was fabricated by creating polymeric microstructure arrays using two-photon polymerization, followed by 50 nm gold thin film sputtering. Cross spike arrays [Fig. 1(b)], which were successfully validated to support effective plasmon excitation in previous works,^19,26 were fabricated using the Photonic Professional GT system (Nanoscribe GmbH., Germany) and photoresist (IP-Dip, Nanoscribe GmbH., Germany).

Fig. 1

Experimental setup configurations and image analysis procedures. SEM images of (a) the plasmonic fiber (PF) and (b) the plasmonic microstructure array. (c) Vertical and (d) horizontal setup schematic. (e) PSs tracking in the vertical experimental setup: (e-1) conversion into a grayscale image; (e-2) binarization and particle detection; (e-3) particle trajectory tracing; and (e-4) out-of-frame particle rejection. (f) Procedures of quantifying the particle concentration change near the fiber tip: (f-1) Conversion into a grayscale image; (f-2) brightness/contrast adjustment; (f-3) $4\times 4$ filter windows generation; and (f-4) pixel intensity measurement.

3.2.

Vertical Configuration Setup

One experimental setup was designed to mount the optical fibers vertically in a suspension of polystyrene microspheres (PSs) and to provide a view of PS behavior in suspension in the horizontal observing direction [Fig. 1(c)]. A fiber holder was designed and 3D printed to clamp the fiber while dipping it into the test suspension vertically. A three-axis translational stage connecting with the metal post by a 3D-printed right-angled optical post-holder was used to provide six degrees of freedom of the optical fiber. Additionally, a cuvette holder fixed to the sample stage of the microscope (VH-Z100R, Keyence Co., Japan) was 3D printed to secure stability and clear access from the side view. The proximal end of the fiber was connected to a laser diode (785 nm continuous wave, L785P090, Thorlabs, Inc., Germany) mounted within a laser diode mount with integrated temperature controller (LDM9T/M, Thorlabs, Inc., Germany), operated by a laser diode current controller (LDC202C, Thorlabs, Inc., Germany).

3.3.

Horizontal Configuration

A horizontal experimental configuration allowed horizontal immersion of the optical fiber into suspension ($10\mu \mathrm{L}$) contained within an imaging spacer (thickness $360\mu \mathrm{m}$). The imaging spacer accommodated the diameter of the optical fiber, and the proximal end was connected to a laser source. The imaging was taken by an upright optical microscope (LCD Digital Microscope II, Celestron, LLC, United States) with a 10× magnification lens (resulting in a pixel size of $0.89\times 0.89\mu {\mathrm{m}}^2$). The schematic of the horizontal configuration is shown in Fig. 1(d).

3.4.

Bacteria Sample Preparation

Five bacteria species were investigated in this work. Escherichia coli (EC) (ATCC BAA-769), Staphylococcus epidermidis (SEp) (ATCC 12228), Staphylococcus aureus (SAu) (ATCC 29313), and Klebsiella pneumoniae (KP) (ATCC 13883) were purchased from ATCC (United States) and Streptococcus agalactiae (SAg) (NCTC 8181) was purchased from NCTC (United Kingdom). Bacteria were aliquoted in glycerol stocks. They were cultured overnight in tryptic soy broth (Sigma-Aldrich, Germany) at 37°C with constant shaking (200 rpm). Afterward, the bacterial cultures were washed three times using phosphate-buffered saline (PBS). Finally, the precipitates (bacterial cells) were resuspended in fresh PBS using a vortex mixer.

3.5.

Experiment

3.6.

Image Segmentation (PSs)

To characterize the behavior of PSs in suspension using three different fibers, images from the recorded videos (28 frames per second, Figs. 2Fig. 3–4) were processed using MATLAB. To obtain particle velocity and other relevant parameters, the positions of the PSs must first be detected in each frame. The center points of the PSs in each image frame were found using the circular Hough transform algorithm (which detects circle patterns and returns the center points and radii). This approach worked well for the PS experiments due to the spherical shape of the PSs.

Fig. 2

$1.1\mu \mathrm{m}$ PS particles motion under laser irradiation through different optical fibers. BF, bare fiber; GCF, gold-coated fiber; PF, plasmonic fiber (Video 1, mp4, 10 MB [URL: https://doi.org/10.1117/1.JBO.28.7.075003.s1]).

Fig. 3

$5\mu \mathrm{m}$ PS particles motion under laser irradiation through different optical fibers. BF, bare fiber; GCF, gold-coated fiber; and PF, plasmonic fiber (Video 2, mp4, 9.95 MB [URL: https://doi.org/10.1117/1.JBO.28.7.075003.s2]).

Fig. 4

$15\mu \mathrm{m}$ PS particles motion under laser irradiation through different optical fibers. BF, bare fiber; GCF, gold-coated fiber; PF, plasmonic fiber (Video 3, mp4, 9.88 MB [URL: https://doi.org/10.1117/1.JBO.28.7.075003.s3]).

Having detected the PS positions in each image frame, the first frame was set as the reference frame and tracked PS trajectories were then defined relative to their initial position in the reference frame. The trajectories for individual PSs were determined by finding the closest position in the next frame to the detected position in the previous frame. After the trajectories for all PSs were recorded, the lengths of the trajectories were filtered to remove PSs that were not in the focal plane for the duration of the experiment [Fig. 1(e)].

After obtaining the position data of the PS trajectories, particle velocities were calculated according to the change in position and the time between frames. The Reynolds number was calculated based on the velocity to confirm whether the PS behavior satisfied Stokes’ law. The Stokes number was then used to verify the PS motion mode theoretically. Average velocities (across entire trajectories) were also calculated to compare velocities for different PS sizes and laser excitation powers.

3.7.

Quantification of the Concentration Change

To assess the accumulation of bacteria and PSs near the fiber tip, particle concentrations were determined by measuring the changes in image pixel intensity in selected areas. The greater the number of bacteria and PSs concentrate in the vicinity of the fiber tip (particle concentration increase), the darker the image gets (pixel intensity decreases) because the particles block or scatter the transmission illumination. Prior to analysis, all image frames were converted to grayscale (i.e., 0 to 255) and then normalized to provide a constant brightness and contrast across all frames. To assess the concentration of particles/cells at the fiber tip, we defined 16 “filter windows” ($10\times 10\text{pixels}$ per window) in a $4\times 4$ grid in the region in which particle accumulation was expected (i.e., at the fiber tip).

The concentration of particles or bacterial cells ($C$) was calculated by the highest pixel intensity (255) minus the average pixel intensity ($I$) in each filter window (i.e., $C=255-I$). The concentration versus time ($C-t$) profiles from all filter windows were then plotted to facilitate analysis and interpretation of accumulation behavior. To reduce the noise of the $C-t$ curves, a moving average window (size of $300=10.7\mathrm{s}$) was applied. Figure 1(f) illustrates the procedure of quantifying concentration changes. Finally, the accumulation ability for the bacteria and the $1.1\mu \mathrm{m}$ PSs was quantified based on both their accumulation speed (i.e., time to reach saturation) and final accumulation area.

The accumulation area was measured from the last frame of the laser-on period ($t=192\mathrm{s}$), where the swarm size was saturated [Fig. 5(a)]. The images were binarized to facilitate the segmentation of the swarm areas from the background [Fig. 5(b)]. Subsequently, rectangular region of interest (ROI) windows were set to filter the binarized images. The positions of the ROI windows were adjusted with respect to the positions of the optical fibers in each image to cover the swarms observed in each experiment [Fig. 5(b)]. Finally, the dark pixels within the ROI windows were counted to obtain the size of the accumulation area in each experiment.

Fig. 5

Image processing for characterization of particle accumulation. (a) Example images of accumulating particles following grayscale conversion and normalization. (b) Rectangular ROIs used for calculation of accumulation areas.

The accumulation speed was extracted from the $C-t$ profiles. As the particles were expected to start to accumulate once the laser was turned on, the initial slope of the concentration curve was defined as the accumulation speed. The accumulation speed values of the first 10 s were used to calculate the slope of the curve.

3.8.

Characterization of Bacteria Accumulation

We examined particle shape and zeta potential to determine the cause of the different accumulation behaviors. The influence of the shape of bacterial cells was investigated via microscopic imaging [scanning electronic microscope (SEM), LYRA3 FIB-SEM, TESCAN, Czechia; Fig. S1 in the Supplementary Material]. The zeta potential of all particles was measured using a zeta potential analyzer (Zetasizer Nano ZSP, Malvern Panalytical, United Kingdom).

4.1.

PS Behavior Under Plasmonic Heating

The motion of $15\mu \mathrm{m}$ PSs under illumination through the three types of optical fibers (i.e., BF, GCF, and PF) was analyzed to characterize the plasmonic heating of the optical fibers and resulting particle behavior (Fig. 6). Since the measurements were conducted using the vertical measurement setup [Fig. 1(c)], there was a gravitational force acting ($-y$ direction in the images). Therefore, PS motion in the $-y$ direction was observed for all fibers when there was no laser illumination [Figs. 6(a)–6(c)]. However, when the laser was on, different PS behaviors were observed [Figs. 6(d)–6(f)]. While no significant change in particle motion was observed with laser illumination through the BF, an apparent directional change was detected in experiments with the GCF and PF. This indicates that additional external force(s) were generated against the gravitational force (more accurately, the differential net force between gravity and buoyancy). The 50 nm gold thin film coated onto the GCF was sputtered, and it is well known that sputtered gold contains relatively smaller and larger grains resulting in nanoscale roughness. This can support plasmon excitation on the gold surface (and, hence, plasmonic heating), creating a local temperature gradient and thus generating convection currents. Notably, the $y$-displacement change of the particles under the PF was more significant than that under the GCF [Fig. 6(g)], indicating a more substantial plasmonic heating effect by the PF. This is attributed to the higher plasmon excitation efficiency with the extra microscopic features of the plasmonic structures at the fiber tips.¹⁹

Fig. 6

(a)–(f) PSs particle tracking under laser irradiation through various optical fibers. BF, bare fiber; GCF, gold-coated fiber; and PF, plasmonic fiber. $d_p=15\mu \mathrm{m}$, Scale bars: $100\mu \mathrm{m}$. (g) Average $y$-displacements before and after laser irradiation over 5 s of time intervals of tracked particles in the images (a)–(f) and their differences.

4.2.

Particle Motion Analysis

Next, PSs behavior under laser irradiation through the PF with various laser powers was characterized to analyze the relationship between the plasmonic heating and the resulting force on the PSs due to the generated convection current (Figs. 7 and 8). The laser diode current was varied from 30 to 50 mA in 5 mA increments, resulting in 2.42, 86.0, 206, 327, and $449\mu \mathrm{W}$ in measured irradiation power at the distal end of the PF. Additionally, 31.8 mA (yielding an irradiation power of $7.85\mu \mathrm{W}$) was included because it was observed that $1.1\mu \mathrm{m}$ PSs started to move upward ($+y$ direction) at that value. The larger particles exhibited upward motion under the laser irradiation power of $86.0\mu \mathrm{W}$. Furthermore, the particle travel distance increased over the same elapsed time with increased laser power. It is important to note that the laser power drastically jumped under the electric current between 31.8 and 35 mA, as this range contained the lasing threshold. Thus the exact threshold laser power in which the upward force was in equilibrium with the gravitational force was not directly measurable for the larger PSs (5 and $15\mu \mathrm{m}$). However, the dataset was informative enough to derive the mechanical properties of the particles to understand the PSs behavior under plasmonic heating (Fig. 9).

Fig. 7

Extracted frames of the video data of $1.1/5/15\mu \mathrm{m}$ PSs under varying laser power at $t=2\mathrm{s}$. The red lines indicate the tracked PSs trajectories. Scale bars: $100\mu \mathrm{m}$.

Fig. 8

Representative visualizations of PS particle (1.1, 5, and $15\mu \mathrm{m}$) tracking under $206\mu \mathrm{W}$ laser irradiation via PFs. (Video 4, mp4, 3.19 MB [URL: https://doi.org/10.1117/1.JBO.28.7.075003.s4]).

Fig. 9

Relevant mechanical properties of PSs under plasmonic heating. (a) PSs velocity, (b) Reynolds number, and (c) Stokes drag force. (d) The force and motion diagrams of PSs experiencing a gravitational force ($F_g$), buoyancy ($F_b$), and the Stokes drag force ($F_D$) without and with the convection current caused by plasmonic heating.

First of all, the velocity of particles was calculated from the particle travel distance ($\mathrm{\Delta }s$) divided by the time interval ($\mathrm{\Delta }t=1/\text{frame rate}=1/(30\mathrm{fps})=0.033\mathrm{s}$) over consecutive frames. The velocities of particles were then averaged along the whole trajectory and the number of tracked particles [Fig. 9(a)]. The particle velocity was approximately linearly proportional to the laser power for all PSs. Furthermore, the larger the particle size, the higher the velocity that was achieved. This can be explained by the fact that the upward drag force on the particle is proportional to the cross-sectional area perpendicular to the fluid flow. To further investigate the particle motion, the Reynolds number and Stokes number were also calculated.

The Reynolds number (Re, the ratio of inertial forces to viscous forces within a fluid) was calculated to investigate the flow regime around the optical fibers. It is defined as $\mathrm{Re}={\rho }_{f}uL/\mu$, where ${\rho }_f$ is the density of the fluid (${\rho }_f=997\mathrm{kg}/{\mathrm{m}}^3$ for DIW), $u$ is the fluid speed, $L$ is a characteristic length (optical fiber diameter in this study, $L=220\mu \mathrm{m}$), and $\mu$ is the fluid dynamic viscosity ($\mu ={10}^{-3}\mathrm{Pa}·\mathrm{s}$ for DIW). Here the tracked particle velocity was used as the fluid speed ($u$) because the particle motion was instantaneous with the laser irradiation and constant over time, representing the fluid flow adequately (Fig. S2–S4 in the Supplementary Material). Figure 9(b) displays the Re for each size of PS with respect to the varying laser power. $\mathrm{Re}\ll 1$ in all cases indicates that the fluid in the system behaved as a laminar flow. Together with this laminar flow condition since the PSs are spherical, the particle behavior was assumed to meet Stokes’ law.

According to Stokes’ law, the Stokes drag force is defined as $F_D=-3\pi \mu d_{p}v$, where $d_p$ and $v$ are the diameter and the velocity of PSs, respectively. The calculated drag forces applied on the PSs with different particle diameters under varying laser irradiation power are displayed in Fig. 9(c). As mentioned above, PSs experience a gravitational force ($F_g$), buoyancy force ($F_b$), and drag force [Fig. 9(d)]. In the case of the absence of the convection current due to plasmonic heating, the net force between $F_g$ and $F_b$ ($F_g-F_b$) is in equilibrium with $F_D$, resulting in particles traveling downward (due to gravity) at their terminal velocity, $v_T$ [Fig. 9(d) left]. However, when plasmonic heating starts creating the convection current, $F_D$ becomes larger and exceeds the net force between $F_g$ and $F_b$, leading to upward motion [Fig. 9(d) right]. As $F_g$ and $F_b$ exerting on the PSs ranging from a few fN ($d_p=1.1\mu \mathrm{m}$) to a few tens of pN ($d_p=15\mu \mathrm{m}$), it is rationally derived that $F_D$ values (which were observed to push the particles upward) range from a few hundred fN to a few tens of pN.

4.3.

PTEP Accumulation of Bacteria

Although the plasmonic heating and the resulting local temperature gradient were successfully demonstrated to be effective in generating convection currents (resulting in far-field particle motion toward the PFs), particle accumulation at the fiber tip was not apparent in suspensions in DIW. This is where the PTEP effect ought to take place in order to facilitate local particle accumulation with the help of the interactions between the particle surface charge and the surrounding ions in the medium. Indeed, it was evident that KP suspension in PBS exhibited accumulation while those in DIW did not [Fig. 10(a)]. Therefore, PBS was chosen as the default medium for bacteria accumulation experiments.

Fig. 10

(a) Comparison between KP in DIW and PBS. (b) Overall concentration changing results of seven samples. (c) Smoothed profiles [“robust quadratic regression (rlowess)” smoothing window, filter size = 200 (7.1 s)] during the early time period for extraction of the accumulation speed values. (d) Summary results of accumulation area, speed, and zeta potential for the analyte suspensions. Inset: binary images of particle accumulations at $t=192\mathrm{s}$.

Figure 10(b) displays the concentration–elapsed time ($C-t$) profiles for five different bacterial species alongside the $C-t$ profile for the $1.1\mu \mathrm{m}$ PSs (PS1.1) (Fig. 11). KP demonstrated the most substantial accumulation ability (i.e., concentration increase) followed by EC, and SAu and SEp showed relatively low accumulation ability. Interestingly, almost negligible accumulation was observed from SAg and PS1.1. Recalling the observation that PS1.1 experienced a drag force of $F_D>1.2\mathrm{pN}$ at the laser power over $449\mu \mathrm{W}$ [Fig. 9(c)], it can be deduced that the force attracting the PS1.1 particles toward the fiber tip was weaker than 1.2 pN in this case. To further quantify the accumulation ability of the particles, we analyzed the size of the bacterial swarms [accumulation area = the number of dark pixels × pixel size ($0.89\times 0.89\mu {\mathrm{m}}^2$)] and the accumulation speed (concentration increase divided by the time until the saturation in ${\mathrm{s}}^{-1}$) extracted from the $C-t$ profiles [Figs. 10(c) and 10(d)]. The accumulation area and speed trends follow each other—increasing with the particles demonstrating stronger accumulation. It should be noted that SAg tended to form scattered clusters near the PF tips, but no recognizable single swarm as the indication of an effective accumulation [Fig. 10(d) inset]. Therefore, even though the accumulation area values of SAg and SEp were comparable (because the total number of dark pixels are similar), it needs to be carefully interpreted that SAg did not produce apparent accumulation, whereas SEp did.

Fig. 11

Accumulation behavior of bacterial/synthetic particle suspensions (KP, EC, SAg, SAu, SEp, and PS1.1) at the tip of PF (Video 5, mp4, 9.16 MB [URL: https://doi.org/10.1117/1.JBO.28.7.075003.s5]).

To explain the particle behavior in the PTEP environment, the electrophoretic mobility also needs to be investigated (even though the temperature gradient effect is also important). In an electric field, the electrophoretic mobility for charged particles relies on the ion composition of the media, zeta potential (i.e., surface charge), and the shape of the particles.^27–30 Accordingly, the difference in the electric double layer (EDL) can be regarded as an entry point to analyze the accumulation ability. Hence, it is essential to measure the zeta potential of the particles to understand their accumulation behaviors better.

As such, the zeta potentials of the bacteria cells and PS1.1 were measured and co-plotted in Fig. 10(d). All the particles were verified to have negatively charged surfaces ranging between $-4$ to $-24\mathrm{mV}$, and greater (negative) surface charge values were correlated with stronger accumulation ability.

Bacteria typically have a net negative electrostatic charge due to the chemical composition of the cell wall.³¹ In general, lipopolysaccharide and teichoic acid on the outermost layer of Gram-negative (e.g., EC and KP) and Gram-positive (e.g., SAg, SEp, and SAu) bacteria, respectively, contribute to the net negative surface charge. In addition, it was reported that Gram-negative bacteria possess more negatively charged surfaces than Gram-positive bacteria,^32,33 which supports our observations.

Moreover, the ion concentration in the medium is another crucial variable in PTEP because it determines the direction of the thermoelectric field.²³ Different ions have different Soret coefficients ($S_T$), and thus there can be a local electric field formed by the separation of the ions at the vicinity of the PF tip (Fig. 12). The $S_T$ of the ions comprising PBS can be found in the literature.^34,35 Since 95% of the ion composition of PBS is ${\mathrm{Na}}^{+}$ and ${\mathrm{Cl}}^{-}$, the system can be approximated as a medium containing only these two ions. $S_T$ of ${\mathrm{Na}}^{+}$ and ${\mathrm{Cl}}^{-}$ are $4.69\times {10}^{-3}$ and $7.18\times {10}^{-4}{\mathrm{K}}^{-1}$, respectively, meaning that the cation (${\mathrm{Na}}^{+}$) is more responsive to the temperature gradient. This disparity in S_T of these ions [“Seebeck effect,” Fig. 12(b)] eventually leads to an ionic concentration gradient, resulting in a local electric field generation [Fig. 12(c)]. Finally, this local electric field generates an induced dipole on a particle at the near field and promotes electrostatic attraction to create accumulation [Figs. 12(d) and 12(e)].

Fig. 12

Schematic illustration of particle motion under PTEP effect near the tip of PF. (a) Particles suspended in electrolyte solution (Brownian motion); (b) differential ionic thermophoresis (Seebeck effect); (c) local electric field generation; (d) induced polarization of the double layer; and (e) induced dipole attraction toward the plasmonic tip.

Another factor that could play a role in PTEP accumulation is the shape of the particles. Drag force is proportional to the drag coefficient $C_D$ associated with the particle shape.³⁶ The drag coefficient of particles with a shape with more indentations is larger than particles with smooth surfaces. The rod-shaped bacteria (KP and EC) are smoother than connected spherical shapes (SAu and SEp) and chain shapes (SAg). The chain shape contains the most indentations over the surface, and thus the drag force exerted on SAg can potentially exceed the PTEP attraction force, towing the bacteria along the convection current streamline, eventually resulting in poor accumulation at the PF tips.

The overall results, together with the properties of the particles investigated in this work, are summarized in Table 1. Notably, while the rod-shaped Gram-negative bacteria (EC and KP) showed effective accumulation ability, the round-shape particles, including Gram-positive bacteria (SAg, SEp, and SAu) and PSs, either did not show noticeable accumulation (PS1.1 and SAg) or exhibited very weak accumulation performance (SEp and SAu). This is attributable to the interplay between the particle surface charge (zeta potential) and the particle shape. The motility of particles did not seem to contribute to the effect. Interestingly, as we observed strong accumulation for Gram-negative bacteria and weak (or no) accumulation for Gram-positive bacteria, these results may also suggest that the PTEP effect can be used to aid bacterial detection and classification (alongside the original target application for bacterial manipulation).

Table 1

Summary of particle properties and the PTEP accumulation performances.