Light-induced manipulation of passive and active microparticles

Abstract We consider sedimented at a solid wall particles that are immersed in water containing small additives of photosensitive ionic surfactants. It is shown that illumination with an appropriate wavelength, a beam intensity profile, shape and size could lead to a variety of dynamic, both unsteady and steady state, configurations of particles. These dynamic, well-controlled and switchable particle patterns at the wall are due to an emerging diffusio-osmotic flow that takes its origin in the adjacent to the wall electrostatic diffuse layer, where the concentration gradients of surfactant are induced by light. The conventional nonporous particles are passive and can move only with already generated flow. However, porous colloids actively participate themselves in the flow generation mechanism at the wall, which also sets their interactions that can be very long ranged. This light-induced diffusio-osmosis opens novel avenues to manipulate colloidal particles and assemble them to various patterns. We show in particular how to create and split optically the confined regions of particles of tunable size and shape, where well-controlled flow-induced forces on the colloids could result in their crystalline packing, formation of dilute lattices of well-separated particles, and other states. Graphic Abstract Supplementary Information The online version contains supplementary material available at 10.1140/epje/s10189-021-00032-x.

Extensive efforts have gone into investigating particles at the gas-liquid and liquid-liquid interface [14,[22][23][24]. The body of theoretical and experimental work studying particles at the solid surface is much less than that for gas-liquid interface, although there is a growing literature in this area. For instance, it has been shown that a self-generated solvent flow at the solid-liquid interface can be used to establish long-range attractions on the colloidal scale leading to the aggregation [25]. Electroosmotic flow near the wall has been reported to drive sedimented particles [26]. Different mechanisms of the colloidal motion can be induced by employing the socalled diffusioosmosis (DO), where the flow tangent to the wall is driven by the gradient of the solute concentration (or the osmotic pressure gradient) [27,28]. It is traditionally considered that the DO flow takes its origin in the adjacent to the wall thin surface layer, such as, for example, the diffuse layer of counter-ions. It is known that if the settled at the wall particles dissolve into ions, this produces local ion gradients driving microflows and particle movement along the charged solid surface [29]. The DO flow can also be induced by light allowing to manipulate inert (i.e. nondissolvable) particles [30], and it has recently been reported that some type of particles, such as porous, can even participate in its generation mechanism, acting as micropumps and thus representing a novel type of active colloids [31].
Previous studies of light-driven DO, usually referred to as LDDO, have addressed the motion and aggregation of either passive or active particles under illumination of focussed or uniform light. Here we show how to induce changes in dynamic pattern of colloidal ensemble of nonporous (passive) and porous (active) particles, as well as temporary steady states, by using the irradiation with light of combined wavelength and simple adjustment of its parameters, such as wavelengths, intensity and its distribution, and the size of the laser beam. Since porous particles can self-generate the LDDO flow near the wall leading to their repulsion or attraction, and this is easily controlled by irradiation parameters, the variety of surface patterns becomes extremely rich. Our work opens up many intriguing avenues, such as manipulating ensembles of different particles with fine-tuning of both their interaction forces and confining areas.
Our paper is arranged as follows. In Sec. II a general concept of light-induced diffusio-osmosis is introduced. Some general considerations concerning DO velocities of a liquid induced by the laser spot that drives passive and active particles is presented. The description of the experimental system, materials and methods used is given in Sec. III. Sec. IV describes the experimental results.
Here we explore and contrast the light-induced aggregation properties of active and passive particles, and show that by tuning the irradiation with a combined wavelength one can remotely control their assembly and, as a result, generate complex dynamic patterns of particles at the solid surface. We conclude in Sec. V.

II. GENERAL CONCEPT OF LIGHT-DRIVEN DIFFUSIO-OSMOSIS
We first present the theoretical basis and physical ideas underlying the approach, which we have developed before and extend here to manipulate microparticles, sedimented under gravity to the bottom wall, but not really stick, as known from numerous hydrodynamic and electrokinetic experiments [32][33][34]. Therefore, they can move along the wall when a lateral force is applied. In our experiment the particles are driven by the fluid DO flow generated by the photosensitive surfactant concentration gradient near a bottom wall. The velocity of this flow, v DO , is generally not equal to that of particles, v p . Below we estimate these theoretically, by focusing on two specific limiting configurations that can be used to manipulate particles.
A. Passive particles.
The first typical case we address is the radially directed DO flow. Such a situation happens when an electrolyte solution containing photosensitive (azobenzene) ionic surfactant is irradiated with the focused light as shown in Fig. 1. Since only a portion of the wall is irradiated, by choosing an appropriate wavelength one can generate trans-isomers inside the focussed beam and cis -outside of it, and vice versa [30]. For example, focused UV light (converting trans-to cis-) induces the flow out of the irradiated area as seen in Fig. 1(a, c). By contrast, when the solution of cis-isomers is irradiated with focused green light (converting fast cis-to trans-) (see Fig. 1(b, d)), the flow is directed towards the laser spot. Note that above a certain critical (bulk) micellar concentration, abbreviated as cmc, the majority of transisomers finds itself aggregated in micelles.
Near the bottom wall an electrostatic diffuse layer, i.e. a cloud of counter-ions balancing its surface charge, is formed. A measure of the thickness of the diffuse layer is the Debye screening length of the solution, λ D , and the excess of mobile (thermal) ions in the diffuse layer compared to the bulk can be denoted as Γ(x), where the x-axis is defined along the wall. The value of Γ(x) can be related to the surface potential of the wall, φ s , as [30] Γ where c 0 is the bulk concentration of surfactants, e is the elementary (positive) charge, k B is the Boltzmann constant, T is the temperature. Naturally, in the absence of irradiation Γ does not depend on x, and the system is at equilibrium. However, when it is illuminated, the lateral gradient of the excess (mobile) surfactant ions in the diffuse layer ∂Γ/∂x induces the liquid flow of the velocity [30] v with the index {t, c} standing for trans-and cis-isomers, and derivatives ∂Γ/∂c t,c characterizing the enrichment a diffuse layer by trans-and cis-molecules. Below cmc the surfactants exist in a form of isolated molecules, so that c c = c 0 − c t , and Eq.(2) can then be reduced to where we assumed that ∇c c ≃ c 0 /L with the length scale of inhomogeneous concentration field L. It is evident that L is roughly equal to the radius of the laser spot R for a configuration shown in Fig. 1, but it can be smaller for a more complex irradiation when finite ∇c c is generated only in some portion of the beam. We will return to this point in Sec. IV A. Overall, we conclude that the magnitude of v DO should grow linearly with c 0 , and that the direction of flow is generally defined by the enrichment of the diffuse layer by trans-or cis-isomers. Another conclusion from Eq. (3) is that v DO reduces with added salt.
All this was indeed observed in prior experiments. Finally, we note that P = c 0 k B T is the osmotic pressure of a surfactant solution, so that everything can be equally discussed in terms of variations in osmotic pressure. In Fig. 1 and some further figures we have included the arrows of different size that intend to indicate the value of P (but, importantly, not a direction of the pressure force) in different part of the diffuse layer. When c 0 reaches cmc, a fraction of trans-isomers does remain in a non-aggregated form. Their concentration is constant (∇c t ≃ 0) and equal the cmc so as to maintain the micelle/trans-isomer equilibrium. Eq.(2) can be then transformed to Then v DO can be estimated using Eq.
Note that since surfactants are ionic, at much higher than cmc concentrations they can simultaneously reduce λ D and increase η, and, consequently, decrease v DO , compared to predictions of Eq.(5). Some experimental measurements [30] lend some support of this. Therefore, to maximize v DO it would be reasonable to keep c 0 ≃ cmc and to avoid high concentration of surfactants. Another advice would be to use pure water to provide largest possible λ D . Finally, we stress that since the laser spot radius R is normally of the order of ten µm, the fluid velocity could be of the order of ten µm/s. This long-range flow appears as homogeneous on the scale of small passive (i.e. smooth impermeable) particles of size a. Such a flow captures them, and particles move with the velocity comparable to that of the DO flow, v p ≃ v DO .
The previous consideration is, of course, grossly simplified, but it provides us with some guidance on a control of DO velocities. We also recall that our model assumes ideal, impenetrable bottom wall, typical for most ordinary applications. In this case the (dynamic) ζ-potential, which, in fact, controls the induced velocity, is equal to φ s . Therefore, one can increase the velocity by using coatings that are permeable to water and ions, such as porous or rough ones, for which ζ-potential is normally augmented compared to the surface electrostatic potential [35], which should lead to an increase in v DO given by Eq.(5).
B. Active particles.
An unfocused uniform illumination cannot, of course, drive the passive impermeable particles described in Sec. II A since in this situation ∂Γ/∂x = 0. However, a light-induced DO flow can be generated when sedimented particles are porous [31]. At equilibrium such particles absorb the ionic species [36][37][38][39], but irradiation with blue light induces a rapid escape of cis-isomers out the pores. Consequently, each particle becomes a source of a laterally inhomogeneous excess of cis-isomers that leads to a concentration inhomogeneity along the wall. This in turn induces the (local) flow along the wall, which is directed away from the particle. In other words, the porous particle acts as a micropump, so that it can be seen as active. An illuminated isolated particle cannot, of course, translate due to a symmetry of emerging flow. However, a region of enhanced pressure near a bottom pole of the porous particle with excess of cis-isomers should inevitably induce an (Archimedes) buoyancy force ( Fig. 2(a)). Consequently, such particles become effectively lighter than passive ones, and, therefore, have smaller support reaction force. Besides, porous particles are slippery [40]. All these should reduce their friction on the bottom wall. One can, therefore, speculate that when the flow is induced by the focused beam (as in Sec.II A), the porous particles would translate faster than nonporous ones.
In the case of ensemble of particles at the wall they repel ( Fig. 2(a)). Such a DO repulsion of active porous particles can be used for a remote control of their twodimensional assemblies at the solid wall, and in particular, simply using different illumination wavelengths (eg. UV light) it is possible to reversibly switch the state of porous particle dispersion from a periodic lattice of particles separated by distances on the order of tens of micrometers to densely packed surface aggregates ( Fig. 2(b, c)) [41]. Note that a similar mechanism of a flow generation near active particles has been reported for calcium carbonate particles [29]. An important difference of our case is that the magnitude and direction of the flow is fully controlled by the intensity and wavelength of irradiation, and that the whole process is reversible.
The phenomenon is identical to described in Sec. II A, but now the length scale of the local LDDO flow is the particle size a, so that the velocity becomes larger than that for the flow induced by the focused light: The axisymmetric flow around a single particle can not, of course, induces its motion, but can drive the motion of neighbouring passive particles of a smaller size. We stress that in this case the velocities of passive particles could be much higher compared to discussed in Sec. II A, where the DO flow was induced by focused laser beam [31].
It is naturally to assume that the fluid velocity near the particle decreases inversely proportional to the distance from it, so that the motion of active porous particles is generated by their long-range DO interactions, i.e. it originates due to the superposition of flows induced by  3. The self-propelled Janus particle under uniform irradiation. The inset shows an SEM micrograph of the particle consisting of porous silica colloid with one hemisphere covered with a gold layer. each particle. The particle velocity then becomes where d is the characteristic interparticle distance. The mutual repulsions of many active particles result in a formation of a stationary periodic lattice of particles that are separated by distances on the order of tens of micrometers (see Fig. 2(c)). Forming such a lattice particles exhibit some hydrodynamic fluctuations resembling the Brownian motion.
If the symmetry of a single porous particle is broken, it can demonstrate a self-propelled motion even when irradiation is unfocussed (see Fig. 3). For example, when a half of an active (porous) particle is coated by metal, it becomes self-propelled, and this self-propulsion can be controlled by the wavelength of light [42]. The motion of such particles is similar to known for Janus particles, generated by diffusiophoresis, thermophoresis or catalytic reactions [43][44][45]. The scaling expression for a DO velocity at the wall generated due to presence of such a Janus particle is also given by Eq. (6), but now this short-range DO flow at the wall is asymmetric. This should lead to a rolling of the Janus particle on the bottom wall. Since the rotation inevitably induces changes in orientation of the Janus axis relative to a wall, this should result in fluctuations in the particle velocity [46].

A. Materials
Azobenzene containing cationic surfactant consists of charged head (trimethyl-ammonium bromide) and hydrophobic tail in which azobenzene unit is incorporated [47] (see Fig. 4(a)). The azobenzene undergoes photo-isomerization from the trans-to the cis-state under irradiation with UV light (λ = 365 nm). The photostationary state with 90% of cis-isomers is achieved within several minutes (intensity dependent) of irradiation. UV-Vis absorption spectra recorded in dark (trans-state) [48] and during UV illumination at the photo-stationary state (cis-isomers) are presented in Fig. 4(b). Under blue light, the trans/cis ratio at the photo-stationary state is 67%/33%. Under illumination with longer wavelengths (green light), the photoisomerization from the cis-to the trans-state takes place within seconds, while thermal back relaxation in the dark takes more than 48 hours [48]. The red light (λ = 625 nm) illumination does not affect photo-isomerization of the surfactant.
The aqueous dispersion of nonporous silica particles of a = 2.5 µm (purchased from microparticle GmbH, Germany) or mesoporous silica particles of the same radius (pore diameter: 6 nm and BET value = 850 m 2 /g , purchased from micromod, Germany) is mixed with surfactant aqueous solution of c 0 = 1 mM, which is twice above cmc. The particle concentration is adjusted to 0.1 mg/ml. The dispersion is kept for equilibration at least for 1 hour and later introduced to glass sealed chamber of height ca. 1 mm and sample volume of 40 µl. All samples are kept in the dark or in red light to prevent undesired photo-isomerization.

B. Methods
An inverted microscope (Olympus IX73) equipped with UV (M365L2-C1, Thorlab Gmbh), blue M455L3, Thorlab Gmbh) and red (M625L1-C1, Thorlabs Gmbh) light source is used for optical measurements. Additionally, three lasers (λ = 532 nm and 488 nm, Cobolt, Sweden; λ = 375 nm, Coherent.inc, USA) are incorporated in microscope where all the beams are focused through the objective to the sample. The illuminated intensity of light is measured using optical power meter PM100D with sensor S170C (Thorlabs Gmbh, Germany). Micrographs are acquired with a Hamamatsu Orca-Flash 4.0 LT (C11440) at a rate of 1 frame per sec. The setup is kept in the dark to prevent the uncontrolled isomerization. When required, red light (M625L1-C1, Thorlabs Gmbh) is used for imaging in dark as it does not affect the photo-isomerization, i.e. the azobenzene molecules do not change their isomerization state. The intensity of irradiation for all light source is kept constant during the experiments.
Image processing and analysis of image data is performed by Image J plugins (Mosaic Single Particle Tracking) using the tracking algorithm described by Sbalzarini and Koumoutsakos [49].

A. Choice of experimental parameters
We first ascertain the range of intensities leading to the LDDO flow. Note that since the LDDO is the consequence of absorption of the optical energy by dissolved surfactant molecules, at certain intensities the local temperature could increase resulting in convection. In Fig. 5 we show how the LDDO flow depends on the light intensity under exposure to focused UV light (R = 15 µm). These results refer to the intensity varied in a broad range, from 30 mW/cm 2 to 10 6 mW/cm 2 . The LDDO flow is visualized by the passive particles's displacement ( Fig. 5(a-c)). Figure 5(d) plots the velocity of particles, v p , in the ring near the edge of the cleaned area as a function of light intensity. It is seen that the velocity of colloids, moving outwards of the beam, increases with the intensity up to 10 4 mW/cm 2 . Starting from ca. 10 5 mW/cm 2 the particles move towards light spot being driven by the temperature gradient. In the intermittent range of the intensity the counter propagating DO and convection flow are balanced and there is no particle displacement. Consequently, for all measurements of LDDO phenomena we use here the intensities confined in a red rectangle in Fig. 5(d). In Fig. 5(e), we plot the illumination time dependence of the average distance of passive particles (located in two rings of width 30 µm) from the center of a laser spot. The dark symbols indicate the adjacent to the cleaned area ring, whilst bright symbols -the next ring, where the magnitude of v p is, naturally, smaller. In this plot the increase in this distance with time reflects cleaning (red and black symbols), while a decrease (blue symbols) corresponds to a collecting of the particles at the center of the laser spot. Corresponding videos (Fig. S1) is provided in supplemental materials.
At large laser spot radii (greater than 100 µm) another scenario of particle motion can be observed. At small intensities, c c is everywhere nonuniform, and all particles move slowly out of the irradiated area ( Fig. 6(a)), so that we have a growing cleaned area similar to Fig. 5(a). However, when I becomes larger than 14 mW/cm 2 , a large fraction of particles still move outward, but some particles are trapped inside the central circle ( Fig. 6(b)) (see corresponding video in Fig. S2 in supplemental materials). This can be explained by the fast photoisomerization of the surfactant resulting in an uniform distribution of c c (see inset Fig. 6(b)). Clearly, the DO can only be generated outside of this region of a constant c c , and the length scale of inhomogeneous concentration field L for such a beam becomes smaller than R. We measure the velocity of nonporous particles at the bor- der of the cleaned area, which is obviously the largest velocity in the system. Fig. 6(c,d) shows v p as a function of irradiation intensity for different spot radii and wavelengths. Our results show that velocities increase with I and are independent of R (Fig. 6(c)) which implies that L is also independent of R in this case. The maximum values are attained for UV light (Fig. 6(d)).
The velocity of LDDO flow for particle manipulation can be adjusted by either a choice of surfactant concentration c 0 (see Sec. II) or by pre-irradiation of the sample with homogeneous UV light for a certain time. Indeed, after exposure of the surfactant solution of c 0 = 1 mM to homogeneous UV light of 0.128 mW/cm 2 during 100 seconds, the amount of the trans-isomers in the bulk decreases to 80%(see Fig. 7(a)). The particle ve-  locity in such pre-irradiated sample is the same as in solution with c 0 = 0.8mM exposed to irradiation with focused UV light (see Fig. 7(b) and S3 in supplemental materials). Correspondingly, with increase of the preirradiation time, the amount of the trans-isomers decreases (blue line in Fig. 7(a) depicts the kinetic of photoisomerization), and the velocity drops. By exposure of the sample to longer wavelength, one converts the majority of the surfactant back to the trans-state and thus restore the system again. The restoring is completely done when the photo-stationary state is achieved, and this takes place within few minutes of irradiation depending on the light intensity and can be at maximum of 10 minutes. Thus, one of the advantages of using LDDO particle manipulation is the "re-charging" of the "battery" within only few minutes.
B. Passive nonporous versus active porous

Triggering LDDO flow by changing the wave length of light
We next examine the displacement of a mixture of passive nonporous and active porous particles illuminated by focused light. The results are shown in Fig. 8. The direction and the velocity of the LDDO flow depend sensitively on the local distribution of the trans-and cis-isomers. Namely, the flow points out of the area with increased cis-concentration and in to the trans-enriched area. The corresponding concentration gradients can be generated in different ways. For instance, under UV exposure more than 90% of the surfactant is isomerized to the cis-state. Also, smaller portions of surfactant molecules in the cisstate form under irradiation with blue and green lights (see Fig. 8 (c)). In all these cases the outwards flow is generated with the velocities and the radii of the cleaned (i.e. free of particles) area which depend on the wavelength (see Fig. 8 (a,b,d)). An important observation is that the velocities of the porous and nonporous particles shown in Fig. 8(d) are different. The porous particles, in general, are faster than similar impermeable colloids. This is exactly what we have suggested in Sec. II B by arguing that active particles should have a smaller friction on the bottom wall. This experimental result requires further theoretical studies, which are beyond the scope of the present paper.
We next address another important question of how long does the LDDO flow sustain under irradiation with focused light? Fig. 9(a) illustrates an extension of the cleaned of nonporous particles area under the irradiation with focused UV light at different moments of time. The area of the cleaned region grows with time strictly monotonically (see uppermost curve in Fig. 9(c) and also Fig. S4 in supplemental materials for the growth during 5 hours). However, when the same focused UV light is applied together with an homogeneous blue light, the growth of the cleaned area stops already after 10 minutes as seen in Fig. 9(b), but by switching off the blue light one can restore the growth of the cleaned area. As a result, the area of the cleaned region grows weakly monotonically (see lowermost curve in Fig. 9(c), which shows a distinct plateau at the time range from ca. 10 to 20 min). These results suggest that one can easily control the size of the cleaned area simply by switching on and off the homogeneous blue light while continuously irradiating with focused UV beam (see supplemental materials for a movie in Fig. S5).
This behavior is likely related to the restricted concentration gradient of the cis-isomers at the irradiated area. Indeed, under continuous exposure to focused UV beam with switch-off blue light, the concentration gradient is defined by the diffusion of the cis-isomers out of the irradiated area (see inset in Fig. 4(c)). In the case of blue light added, the gradient is cut by the continuous photoisomerization out of the UV irradiated area, so that the cis-isomer produced in the focused light diffuses out and isomerizes back to the trans-one. The size of the cleaned area is then defined by the ratio of the blue/UV light intensities since the amounts of the trans-and cis-isomers vary significantly.
Similar response is observed in the case of active porous particles (see Fig. 10). We remark that the size of the cleaned area is always smaller than for nonporous particles. At first sight this is somewhat surprising, since, as discussed above, the porous particles should generally move faster under irradiation with light of one wavelength, either UV or blue (see Fig. 8(c)). We recall, however, that in this particular experiment we have a combination of the two wavelengths, homogeneous blue and focused UV. In this case the cleaned area for active particles decreases faster when the blue light is switched on (see Fig. 10(e) and corresponding video in Figure S6 of  supplemental materials). This effect is due to particle mutual DO repulsion (described in Sec. II B and illustrated in Figs. 10(a,c)) outside the cleaned area.

C. Controlled particle manipulation by optically triggered restrictions
We are now on a position to propose the recipes of selecting and further confining a subset of particles in a confined region, without using a chamber with real walls. Fig. 11 illustrates how the expanding of the laser spot or changing its parameters, such as the shape and/or intensity profile, could potentially manipulate arrested nonporous (grey spheres) or active porous particles (white spheres). For example, a Gaussian beam should tightly gather particles [50] as seen in Fig. 11(a), but using flat-top beams of circular or rectangular shapes would almost likely lead to more dilute ensembles, with much larger separations between active particles ( Fig. 11(b,c)). Moreover, the confinement (and the ideality of the flattop beam) could be dynamically controlled by the frequency ω (Fig. 11(c)). Thus, since the size of a confined area and the uniformity of the illumination can be easily tuned, one may generate situations from rather dilute "gases" of colloids to almost crystalline arrangements. Below we present some specimen experimental results illustrating the power and variability of this approach. Fig. 12 shows that upon reducing the radius of the circular laser spot, R, from 500 µm down to 190 µm, one can generate a rapid change of the particle surface density, from a dilute state to a densely packed monolayer ( Fig. 12(a, b, c)). On reducing R further one can even force the particles to form a multi-layer crystal structure (see Fig. 12(d)). Note that in this specimen example we increase the density of the confined particles in 17 times. We stress that the process is completely reversible.
In the case of irradiation with homogeneous UV followed by focused blue light, i.e. when the blue light is applied to a cis-enriched solution, the outer particles are concentrated at the boarder of the irradiated spot forming a densely packed shell. The "incapsulated" inner particles, i.e. the ones in the confined area, randomly move if they are nonporous ( Fig. 13(a)), but when porous, a kind of lattice of well-separated (due to the DO repulsion) particles is formed (Fig. 13(b)). When these systems are irradiated by the combined light of two wavelengths, namely, when the UV light is homogeneously exposed over the sample, while the selected area is illuminated by blue light, steady configurations of particles become different from described above. In this case, all particles are collected to the focused spot area, and their package becomes denser then in Figs. 13(a, b). The passive particles form a crystalline structure (Fig. 13(c)). The active ones remain separated due to the DO repulsion ( Fig. 13(d)), but the interparticle distance becomes quite small. Fig. 14 demonstrates that one can easily "cut" an optically confined dispersion of sedimented particles, by acting as optical "scissors". When particles are passive nonporous (Fig. 14(a)), the irradiation of the cisenriched solution with modulated blue beam of elliptic shape (R 1 = 215 µm, R 2 = 36 µm) leads to a formation of the elongated, densely packed cluster ( Fig. 14(b)). By applying then the UV light this cluster of nonporous particles can be splitted by two parts (see corresponding video in Fig. S7 of supplemental materials). Porous particles can be manipulated similarly, and we illustrate this by using other illumination conditions. When the particles are irradiated by focused blue light together with homogeneous UV light, the elongated clusters arise as seen in Fig. 14(d). If we apply homogeneous UV preirradiation and then focused blue light, one can ealily cut the cluster of particles (see Fig. 14(e)). However, an homogeneous UV pre-irradiation with the consequent illumination with focused green light lead to a tightly arranged elongated ensemble of the porous particles resembling the situation with nonporous colloids (Fig. 14(f)). Note that the generated patterns can again be easily reversed by tuning the parameters of the laser beam. (see corresponding videos in Figure S8 in supplemental mate-  14. (a,b,c) Optical micrographs of nonporous silica particles during exposure to laser spot of an elongated shape of either blue or UV light. Insets in (a) and (c) show the beam shape for blue and UV light, respectively, with arrows indicating the direction of the DO flow. (d,e,f) Optical micrographs of porous silica colloids during irradiation with modulated laser spot: (d) blue light with additional homogeneous irradiation with UV light, (e) blue light applied to pre-irradiated sample with homogeneous UV light, (f) green light applied to pre-irradiated sample with UV light. rials).

V. CONCLUSION
In our work we have demonstrated that a combination of two distinct primitive scenarios of the light-induced diffusio-osmosis (LDDO) can be exploited to generate many non-trivial dynamic configurations of particle ensembles at a solid wall, at once, seamlessly and quickly. For instance, it can be used to create the restricted, but "fenceless", areas of tunable size, where the motion of passive and active particles can be studied and/or remotely controlled. Importantly, we can easily confine a subset of particles simply by using the laser spot, i.e. without introducing them into a chamber with physical walls. We also note that by simply varying the radiation intensity, focusing or wavelength is sufficient to alter the diffusio-osmotic flow and thus to induce various patterns of colloid particles at the bottom wall, from rather dilute "gases" of colloids to almost crystalline, densely packed structures.
One of the main experimental results of our work is the difference in behaviour of nonporous and porous particles. The trends that have been observed are consistent with our simple theoretical models that treat nonporous particles as passive, and porous -as actively participating in the mechanism of the flow generation.
Our strategy can be employed to manipulate more complex particles, such as asymmetric Janus ones that contain active and passive parts. Such particles are selfpropelling with velocity fluctuations. A systematic study of their manipulation in the spirit of the present work appears to be very timely and would constitute its significant extension.