Nanofountain Pen for Writing Hybrid Plasmonic Architectures

Self‐assembled colloidal nanostructures enable various interesting fabrication alternatives to current top‐down lithographic approaches for supporting localized surface plasmon resonance (LSPR). Recently, there have been numerous efforts to manipulate the LSPR in structural or compositional engineering using clustered nanoparticles (NPs). However, creating heterogeneous colloidal clusters comprising two or more types of NPs requires a complicated chemical synthesis of the linker molecules on the NP surfaces. Moreover, the traditional assembly methods pose several challenges to 3D manufacturing. Herein, a fountain pen‐inspired open‐microfluidic approach demonstrating an exquisite balance between evaporation and capillary action is reported. This approach enables the direct writing of binary NP clusters. Microcapillaries are designed to precisely control the evaporation and capillary action of the binary NP suspension ink at the tip. This approach effectively guides the growth of NP clusters along the out‐of‐plane direction to fabricate a freeform microarchitecture. The growth characteristics are theoretically explained using a simplified balancing model. In addition, the optical properties can be tuned precisely using multicolloidal NP mixing. A Janus pillar comprising plasmonic NPs and biomolecules is showcased as a microactuator operating under chemical stimuli. It is expected that the proposed method paves the way for manufacturing numerous interesting structures in nanophotonics.


Introduction
[15] One prerequisite for high-performance, functional photonic devices concerns the design and manipulation of the LSPR.[18][19] When the dipolar plasmons of all the NPs oscillate in phase (super-radiant mode), the increased radiative damping results in spectral broadening.In contrast, out-of-plane oscillations decrease radiative damping, resulting in a spectral dip.The ensembles of the plasmonic modes determine the LSPR of an NP cluster.These plasmonic modes can be manipulated according to factors, such as the shape, size, dielectric surroundings, and interparticle distances of the NPs. [16,17]Furthermore, recent studies have Self-assembled colloidal nanostructures enable various interesting fabrication alternatives to current top-down lithographic approaches for supporting localized surface plasmon resonance (LSPR).Recently, there have been numerous efforts to manipulate the LSPR in structural or compositional engineering using clustered nanoparticles (NPs).However, creating heterogeneous colloidal clusters comprising two or more types of NPs requires a complicated chemical synthesis of the linker molecules on the NP surfaces.Moreover, the traditional assembly methods pose several challenges to 3D manufacturing.Herein, a fountain pen-inspired open-microfluidic approach demonstrating an exquisite balance between evaporation and capillary action is reported.This approach enables the direct writing of binary NP clusters.Microcapillaries are designed to precisely control the evaporation and capillary action of the binary NP suspension ink at the tip.This approach effectively guides the growth of NP clusters along the outof-plane direction to fabricate a freeform microarchitecture.The growth characteristics are theoretically explained using a simplified balancing model.In addition, the optical properties can be tuned precisely using multicolloidal NP mixing.A Janus pillar comprising plasmonic NPs and biomolecules is showcased as a microactuator operating under chemical stimuli.It is expected that the proposed method paves the way for manufacturing numerous interesting structures in nanophotonics.[18][19] Using electron-beam or ion-beam lithography enables the fabrication of nanometer-sized metallic objects with designed shapes and interparticle distances but suffers from cost inefficiencies and low throughput.Moreover, most of the resulting structures are constrained within a plane; thus, it is challenging to construct a vertically stacked configuration.[26] In particular, solution-processed self-assembly provides a simple, scalable, and high-throughput route for fabricating 2D or 3D plasmonic clusters.A nonuniform evaporation-induced surface tension gradient from the solvent drives a Marangoni flow of colloidal NPs toward a two-or three-phase interface, producing densely packed clusters at the interface.Based on this solution-mediated mass transfer process, innovative methods have been devised for fabricating NP clusters, such as dip coating, [24] the Langmuir-Blodgett process, [25] and template-assisted self-assembly. [26]For improved efficiency and variety of assembly, a chemical approach has been employed to tailor the molecular linkers on the NP surface. [27]lthough molecular linkers deliver outstanding optical properties, their use inevitably requires complicated chemical syntheses, reducing the simplicity of the approach.Despite enormous efforts, the technological challenges associated with the 3D structuring of plasmonic clusters remain unresolved.
This study proposes a nanofountain pen (NFP) for fabricating functional architectures (such as 3D plasmonic architectures) with tunable optical properties and dynamic actuation.Conventional self-assembly-based NP cluster fabrication methods are performed over a solid-liquid-air interface (1D) (dip-coating process) or liquid-air interface (2D) (Langmuir-Blodgett process).Recently, Tan et al. proposed that a meniscus-guiding 3D printing method (3D) enables fabrication of micro-to macroscale freestanding pillars composed of colloidal clusters. [28]However, this method does not guarantee uniform binary colloid assembly because the self-assembly occurs over a large area.Until now, binary colloidal clusters have been fabricated by means of a template-assisted self-assembly and chemical synthesis of linker molecules on the surfaces of the NPs. [29]In contrast, the proposed self-assembly method, inspired by a fountain pen, enables the direct writing of NP clusters at the tip of the pen (0D); this tip localizes the meniscus.By taking advantage of the femtoliter-scale meniscus localization, binary NP clusters and freeform 3D architectures can be fabricated without desired template or chemical synthesis for guiding NP assembly.In the results and discussion, we report on fountain pen-inspired architectures consisting of NPs and the related fabrication mechanisms, as well as their optical and mechanical properties.

Results and Discussions
2.1.Fountain Pen-Inspired Self-Assembled Architectures Consisting of Nanoparticles (NPs) By scaling down the writing of a fountain pen to the micrometer scale, it is possible to create a 3D architecture consisting of NPs.When writing on paper using a fountain pen, the evaporation and absorption of ink occur simultaneously at the point where the tip of the fountain pen and the paper come into contact (Figure 1a).On a porous surface, such as paper, water is absorbed within seconds by capillary action, [28] but several minutes are required for all the water to evaporate. [30]In other words, in general, when writing on paper with a fountain pen, the ink is first absorbed.Then, as the absorbed ink evaporates, the fine particles dispersed in the ink leave traces on the paper to complete the writing.For a spherical point ink source with an initial radius R 0 , the time required to fully evaporate is proportional to R 0 2 of the source; this is known as the radius square law. [31]In this case, the volume loss of the source per unit time is proportional to R 0 , and the evaporation-induced accumulation of ink particles per unit of time is also proportional to R 0 .The writing area of the fountain pen is proportional to R 0 2 ; thus, the accumulated height of the particles per unit of time is proportional to R 0 À1 .Through this qualitative prediction, we can expect that if the size of the tip of a fountain pen is reduced to the micrometer level (Figure 1b), the deposition of the particles dispersed in the ink on the nonabsorbing substrate will be enhanced 10 3 times more than a tip sized at the millimeter level.In this study, we refer to the scaled-down fountain pen as customized with a capillary tube as a NFP.Thus, the NFP can help vertically grow NP deposition on a substrate.Indeed, the evaporation of an NFP and self-assembly of NPs can create unique 2D and 3D structures at the micrometer scale.At the meeting point of the tip and glass substrate, a capillary bridge (CB) is formed, and only evaporation occurs at the meniscus.The resultant continuous ink inflow causes NPs to accumulate on the glass substrate (Figure 1c).Due to the rapid deposition of NPs from this scaled-down approach, half donut-type patterns can be formed, or dotted text-type letters can be written (Figure 1d).Moreover, by slowly lifting the tip in a vertical direction, the NPs can be deposited continuously in a direction perpendicular to the substrate (Figure 1e).Using this strategy, a straight pillar can be constructed (first image of Figure 1f ).Besides, applying both horizontal and vertical motions of the substrate enables to form helical structures, as shown in the second and third images of Figure 1f.These half-donut patterns or 3D architectures consist of the regular stacking of NPs (fourth image of Figure 1f ).

Half-Donut Structures Fabricated by the Stamp-Like Process
A stamp-like process is applied to construct the self-assembled structures (Figure 2a and Movie S1, Supporting Information).The process consists of three steps: 1) accumulation, 2) contact, and 3) detachment.The top and bottom of Figure 2a show a side-view image and illustration of each process, respectively.The black dashed line represents an interface of the substrate mounted on a motorized three-axis stage.The dark shadow under the black dashed line is a mirror image of an NFP, as reflected from the substrate.The yellow spheres represent 80 nm gold NPs (AuNPs).In step (1), the NFP tip maintains a certain distance far from the substrate; the waiting time in this step is defined as the accumulation time (t A ).In step (II), the substrate approaches close to the tip (to %Δz); then, the ink droplet hanging on the NFP tip touches the substrate.Notably, the NFP tip and substrate do not make mechanical contact.After that, the substrate proceeds rapidly back to the initial position in step (3).Repeating this stamp-like process with a horizontal transportation of the substrate enables the 2D patterning of self-assembled structures, as shown in Figure 1d.
The structures change according to the accumulation time (t A ) and AuNP number density (n 1 ) of the ink. Figure 2b shows two phases of structures resulting from the stamp-like process: a half donut (marked green) and short pillar (marked light green).At a low number density and short accumulation time (marked sky blue), the stamp-like process cannot form structures.At a high number density and long accumulation time (marked orange), the process breaks due to the clogging of the ink and thus becomes unavailable under these conditions (marked light orange).
In particular, the shape and volume of the structures can be adjusted according to t A and n 1 .Figure 2c shows scanning electron microscope (SEM) images of the structures with the Roman numbers (I-VI) corresponding to Figure 2b.Images (I)-(III) of Figure 2c show that the volume of the structures (V HD ) is increasing at t A ¼ 10s.The structures change to a coffee ring-like pattern (I), half donut (II), and short pillar (III).The images (I), (IV), and (V) of Figure 2c indicate that the volume of half-donut structure changes following an increment of accumulation time at n 1 ¼ 0.2particles=fL (a femtoliter (fL) corresponds to μm 3 ).The short pillar is cut off because excessive accumulation leads to clogging of AuNPs at the NFP tip end ((VI) of Figure 2c).These results demonstrate qualitative characteristics; for example, the numbers of stacking NPs in structures increase with an increment of the number density and/or accumulation time.This is because of the linearity between the structure volume and the number of stacking NPs.The stacking structure of AuNPs consists of a hexatic (red line) and pentagon packing (blue line, Figure 2d).Even if the pentagon packing generates more void space than the hexatic packing in the structure, the ratio of the void space can be maintained with minimal variation.Therefore, we carefully assume the volume fraction (α) of the self-assembled structure as α ¼ 0.74, that is, the value of the hexagonal-close-packed lattice.Notably, the tip radius (R Tip ) of the NFP used in the experiment is the same for each number density, with the R Tip varying slightly for the different number density conditions.
In the stamp-like process, the half-donut structures form under narrow conditions.Here, we focus on the forming characteristics of the half-donut shape at n 1 ¼ 0.2particles=fL for an 80 nm AuNP.The shape and volume of the half-donut structure are characterized by the height profile (h xy ) as measured by atomic force microscopy (AFM) (shown in Figure 2e), where the indexes represent the x and y positions, respectively.The radius of the half donut is defined as a distance from the center of mass, as calculated using h xy .Figure 2f shows the height profile hðrÞ of the half donut, where hðrÞ denotes the average value for Δr ¼ 10 mm.The half-donut structures are formed at t A > 30s.In their normalized height profiles (hðrÞ=h max , h max is a maximum of hðrÞ and exhibits a similar shape, as shown in the inset of Figure 2f.With an excessively short accumulation time (t A ≤ 10s), the NPs get spread out on the substrate, as shown in (I) of Figure 2c.The volumes of the structures generated using the stamp-like process (V HD ) are measured based on h xy for the number density (n 1 ) of 0.2 and 1particles=fL (Figure 2g).The increment of V HD is nearly proportional to the accumulation time (t A ) and the slopes are s 1 ¼ 0.106 AE 0.002 μm s À1 for n 1 ¼ 0.2particles=fL and s 2 ¼ 0.635 AE 0.053 μm s À1 for n 1 ¼ 1particles=fL.The number density exhibits an almost linear relationship with the volume of the structure, indicating that the ink uniformly evaporates.

Vertical Growth of Self-Assembly of NPs
The continuous pulling of the substrate enables the vertical growth of the self-assembled structures consisting of NPs. Figure 3a shows four steps to constructing the 3D architecture cluster using the 80 nm AuNP-dispersed ink with a number density of n 1 ¼ 2particles=fL (see Movie S2, Supporting Information).The steps of (I) approach, (II) contact, and (IV) detachment proceed in the same manner, as shown in Figure 2a.Step (III) (growth) denotes an additional continuous process of pulling the substrate; then, the NPs are stacked continuously to form a freestanding 3D architecture.Notably, the resulting structures are nanoporous due to the voids between the AuNPs.This porosity significantly influences the assembly dynamics, as discussed according to Figure 4.
The growth speed of the pillar depends on the AuNP composition of the ink. Figure 3b shows a plot of the growth speed when using 80 nm AuNP dispersed ink as a function of the number density (n 1 ).A pillar can be stably grown at a pulling speed of %200 nm s À1 .The growth becomes unstable as the pulling speed exceeds 200 nm s À1 .Above 300 nm s À1 , the growth becomes impractical at the given n 1 .Notably, adding 20 nm AuNPs makes an apparent change in the self-assembly dynamics.Figure 3c shows that the pillar growth depends on the number density of the 20 nm AuNP (n 2 ) for n 1 ¼ 2particles=fL.There are two distinguishable trends.First, the maximum allowable growth speed increases linearly as n 2 increases from 0 to 75particles=fL.Secondly, the maximum speed converges and becomes constant at %500nm=s for n 2 > 75particles=fL.Figure 3d shows exemplary binary AuNP pillars with different exterior shapes for 50-150 s of fabrication time under conditions I-IV (shown in Figure 3c).The fabrication time depends on structure height, such as the fabrication of microstructure using plasmonic bubbles. [32,33]At a relatively low pulling speed (I, II, and III), the pillars exhibit uniform exterior diameters along the height.However, at a high pulling speed (IV), the growth becomes unstable, with the pillars assuming various wavy shapes.
The packing of the pillars depends on the mixing ratio of two different-sized AuNPs.The packing configurations for I, II, and III are shown in the magnified field-emission (FE) SEM images in Figure 3e (left, middle, and right in order), respectively.The 20 nm AuNPs are sandwiched between 80 nm AuNPs (I) and sufficiently surround the larger NPs when sufficiently numerous (II and III).The 80 nm AuNPs exhibit hexagonal packing (red lines), with the lattice distance increasing when the number of mixed 20 nm AuNPs increases.The 20 nm AuNPs exhibit hexagonal packing, as shown in the inset of III (Figure 3e).To investigate the internal packing configuration, the pillars were cut with a focused ion beam (FIB), as shown in Figures 3f,g  The cross-sectional FESEM image of the single-composition 80 nm AuNP pillar (left image of Figure 3g) shows a close-packed hexagonal structure well matched to that shown in Figure 2d.In the binary pillar, the distance between the 80 nm AuNPs is increased by embedding the 20 nm AuNPs (right image of Figure 3g).

Evaporation-Induced Accumulation of NPs for Stamp-Like Process
The evaporation of the ink at the tip of the NFP is crucial to the creation of half-donut structures.Figure 4a shows the evaporation-induced AuNP accumulation at the NFP tip before the contact process.Here, t A ¼ 0s denotes the start of the water evaporation right after detachment, as shown in step (III) of Figure 2a.Thus, the accumulated 80 nm AuNPs before stamping are ejected fully by this process.Correspondingly, the number of accumulated 80 nm AuNPs at t A ¼ 0s is approximately zero.The accumulated number of 80 nm AuNPs increases according to t A , as shown in the dark region H A (second and third images of Figure 4a).Therefore, this observation qualitatively proves the monotonic incremental process when accumulating 80 nm AuNPs depending on t A (Figure 4b).
To estimate the number of NPs accumulated at the NFP tip, we consider the water evaporation based on a diffusion equation expressed as ∂c= ∂t ¼ D∇ 2 c. [34] Here, c is the vapor concentration and D is the diffusion coefficient of the water in air.For simplicity, we assume that the water at the tip of NFP is a micrometer-sized spherical water source with a constant radius, as the evaporation occurs at the tip and the water (liquid phase) supplied continuously from the NFP to the tip (left image of Figure 4c).Near the source surface, c is saturated due to the evaporation (defined as c S Þ.In contrast, in the air very far from the source, c converges to the value of the vapor concentration (defined as c 0 ).As the reservoir is sufficiently large, the evaporation from the source quickly reaches a quasisteady condition.Then, the left term of the diffusion equation becomes 0. In this case, the diffusion equation becomes the Laplace equation (D∇ 2 c ¼ 0), with the boundary conditions of cðr ¼ RÞ ¼ c S and cðr ≫ RÞ ¼ c 0 , where r is the radial distance and R is the radius of the water source.The solution of the Laplace equation is cðrÞ ¼ c 0 þ ðc S À c 0 ÞR=r for r ≥ R in spherical coordinates.
The evaporation leads to the volume loss of the water in the ink at the NFP tip.In Figure 4c, the flux ( J) of the water vapor represented by the yellow arrow flowing from the outward source is calculated based on the gradient of the concentration, which is expressed as J ¼ ÀD∇c.Then, the mass loss per unit of time ( Ṁ ¼ dM=dt) caused by the evaporation is estimated by integrating the flux passing through a certain enclosing surface surrounding the source (expressed as Ṁ ¼ H J⋅dã). Next, the volume loss rate ( V) of water in the NFP is calculated as where ρ w is the mass density of the liquid water.Then, the evaporating volume rate of the water at the NFP tip is VTip % V=2 due to the half-exposed surface of the water source.
The evaporation volume of water helps to control the number of NPs comprising a structure.The number of accumulated NPs per unit of time is expressed as follows where n i is the number density of the NPs.The index i represents the type of NP dispersed in the ink, as distinguished by its size The fabrication of structures requires the proper accumulation time and number density, as determined by the injected volume in the contacting step.When the accumulated NPs are injected in the contacting step according to t A and n i , the NPs with VNP ¼ VTip P i n i v i are stacked on the substrate after detachment.Here, the volume of a particle is v i ¼ πD i 3 =6, with D i denoting the diameter of each NP.Then, the volume of an assembled structure depending on the accumulation time by the stamp-like process is expressed as follows where α is applied due to the porosity of the structure.The assembly conditions for the stamp-like process can be understood through the volume of the structure using Equations ( 1) and (2). Figure 2b is based on the 80 nm AuNPs dispersed ink (D 1 ¼ 80 nm).The radius of NFP tip is R % 1.5 μm, as estimated from Figure 2f for n 1 ¼ 0.2particles=fL.For the temperature (25°C) and relative humidity (RH) (40%) of the reservoir, the parameters are c S ¼ 19.96 g m À3 , c 0 ¼ 7.99 g m À3 , D ¼ 2.6 Â 10 À5 m 2 s À1 , and ρ w ¼ 10 3 kg m À3 . [35,36]Using Equations ( 1) and (2) with these parameters, we can estimate VSt % 0.205 μm 3 s À1 .However, this is larger than the s 1 ¼ 0.106 AE 0.002 μm 3 s À1 in Figure 2g.Indeed, c S decreases more significantly, as the picoliter level of water evaporation cools down the NFP tip.Then, VSt approaches s 1 when the NFP tip cools to %5 °C.Similarly, VSt is proportional to the accumulation time and can be evaluated for n 1 ¼ 1particles=fL.Naturally, the value of s 2 is approximately five times larger than the value of s 1 , as shown in Figure 2g.Based on the analysis for evaporation-induced accumulation, the accumulated NPs during t A contribute to the fabricated volume of the half-donut structure (slopes of Figure 2g).Additionally, VSt is useful for understanding the fabrication limit.Here, we use the average height of the structure expressed as H St ¼ ðt A Â VSt Þ=S, with a cross-sectional area of the structure S ¼ πR 2 , as shown in the left image of Figure 4d.For the breaking condition as shown in (VI) of Figure 2b, the parameters are n 1 ¼ 4particles=fL and t A ¼ 25s.This leads to H St % 8 μm.In this overaccumulation case, the accumulated volume cannot escape at once from the NFP tip in the detachment step.As such, the remaining volume clogs the NFP tip, as shown in the right image of Figure 4d.We note that these moderate accumulation conditions are dependent on diverse parameters regarding water evaporation and NP stacking, and thus the conditions should be considered always when each parameter is changed.
The shape of the half-donut structure is influenced by the accumulation before the contacting step.In a contacting step with a moderate accumulation volume (marked to box in Figure 4d), the evaporation (yellow arrows) occurring at the meniscus of the CB creates a fluid flow (red arrows) of ink that escapes from the center of the CB, causing the NPs to relocate near the walls, as shown in Figure 4e.After contact and upon detachment, the CB changes into a sessile droplet while maintaining the relocation of the NPs (Figure 4f ).At this time, the edge of the sessile droplet meeting the substrate is fixed by the pinning effect of the stacked NPs at the edge.The micrometer-scale sessile droplet dries in an instant; in this process, the number density of the NPs in the sessile droplet determines the shape of the self-assembled structure.If the number density is excessively low, the surface of the sessile droplet will gradually lower while drying, leaving the edges fixed (top of Figure 4f, where the solid, dashed, and dotted lines represent the lowering sequences of the surface).Within the sessile droplet, the NPs circulate according to Marangoni (black arrow) and capillary flows (red arrow), preventing the stacking of NPs at the edges.In contrast, for a moderate number density of the sessile droplet, the numerous NPs rearranged at the edge do not escape easily during the drying process of the sessile droplet (bottom of Figure 4f ).Accordingly, numerous NPs accumulate at the vicinity of the edges due to the capillary flow.Therefore, the surface of the sessile droplet changes to that as presented by the solid, dashed, and dotted lines, forming the half-donut structure.

Evaporation-Capillary Action-Balanced Fabrication Mechanism for 3D Architectures
The vertical growth of the 3D architectures is achieved by balancing between evaporation and capillary action.Unlike the stamping-like process, vertical growth requires a continuous pulling of the substrate after the contacting step, as shown in Figure 3a.Immediately after the NFP contacts the substrate, the NPs accumulated on the NFP tip migrate to the substrate to form the stump of the pillar, as shown in the first image of Figure 4g.The continuous evaporation on the meniscus of the CB leads to the additional stacking of NPs upon the stump; this process comprises the vertical growth.
To characterize the vertical growth speed, it is necessary to quantify the volume loss because of water evaporation from the CB.To avoid complex entanglements among the tube shapes, the CB interface curvature, substrate nonslip boundary condition (s), contact angle, temperature, humidity, atmospheric pressure, and quasisteady water evaporation from a CB surface with a cylindrical geometry are assumed.This results in a Laplace equation in the cylindrical coordinate system with rotational symmetry.The general solution is cðrÞ ¼ A ln r þ B for r ≥ R (r: the radial distance, A and B: coefficients).The water vapor is saturated at the water interface and converges when the distance is sufficiently far, for example, for the system size L.Then, the boundary condition for the vapor concentration becomes cðr ¼ RÞ ¼ c S and cðr ¼ LÞ ¼ c 0 , and thus the coefficients are A ¼ ðc S À c 0 Þ=ðln R À ln LÞ and B ¼ ðc 0 ln R À c S ln LÞ= ðln R À ln LÞ.In particular, the vapor concentration decreases as a function of ln r as the distance increases.Then, we can determine the volume of water evaporating from a cylindrical water column of height H CB in the same way as in the previous section, and the result is expressed as follows This evaporation-induced water inflow contributes to the vertical growth of the pillar.In the pulling process, the height of the CB is H CB % 2R (the CB is disconnected above this height).
The growth speed can be estimated based on the evaporation volume.When the ink comprises two types of NPs (i ¼ 1, 2), the number of stacked NPs per unit time is Ṅi ¼ n i VCB and the corresponding volume per unit time is v i Ṅi .The total stacking volume of NPs per unit of time becomes VNP ¼ P 2 i¼1 v i Ṅi .As the pillar structure is porous, the growing structure volume per unit of time is expressed as α VPillar ¼ VNP .Then, the growth speed of the pillar is expressed as Here, S Pillar ¼ πR 2 is the cross-sectional area of the pillar.The capillary rise (CR) caused by the porous structure of the pillar improves the growth speed.As shown in the second image of Figure 4g, the CR expands the wetting surface.The wetting height is expressed as . [37] Here, χ is the sorptivity with a unit of length per squared time, and ϕ ¼ 1 À α is the porosity.Due to the CR, additional water evaporation occurs on the expanded wetting surface.The wetting height H CR does not continuously increase but converges to a specific value due to the evaporation from the expanded wetting surface.Using Equation (3), the evaporating volume per unit time according to the CR is VCR ¼ À2πADH CR =ρ w , and the water volume flowing into the pillar is VFlow ¼ ϕS Pillar ḢCR .These two are in equilibrium ( VCR ¼ VFlow ).Thus, it is possible to determine the balancing time τ ¼ ϕρ w S Pillar =4πAD and balancing height H CR ðτÞ.Then, the total evaporation volume of water per unit of time is expressed as Equations ( 3) and ( 5) indicate that the porous pillar as well as RH change improves evaporation, and thus, the volume of water supplied per unit of time from the NFP increases.In particular, the stacked volume of NPs per unit of time is enhanced and expressed as VNP ¼ VTotal P 2 i¼1 n i v i , which also satisfies α VPillar ¼ VNP .The tilde symbol represents the values enhanced by the CR.Finally, the growth speed as improved by the CR is expressed as Based on the above information, the growth of pillars can be quantitatively analyzed in greater detail.The growth speed of the pillar fabricated using 80 nm AuNPs is shown in Figure 3b.The growth speed without the CR effect (Equation ( 4)) is calculated as ḢPillar % 107 nm s À1 when the parameters for the temperature (40C) and RH (15%) are used: and α ¼ 0.74. [35,36]Due to the cooling down at the CB, as mentioned earlier, the value of c S ¼ 37.24 g m À3 is applied for 35 °C.This growth speed is enhanced more by the CR effect.As compared to Equation (6) and Figure 3b, one can find an additional enhancement in the growth speed due to the CR effect on H CR =H CB % 0.86.As expected, the porosity of the pillar also contributed to the growth.It is also possible to determine how the mixing ratio enhanced the growth of the pillar.In Figure 3, 80 and 20 nm AuNPs are used as the colloidal particles dispersed in the ink.The former and latter are defined as i ¼ 1 and i ¼ 2, respectively.Then, the parameters n 2 and D 2 ¼ 20nm are added to Equations ( 4)-( 6).Adding 20 nm AuNPs increases the growth speed.The calculated growth speed using Equation ( 6) is ḢPillar % 523 nm s À1 for n 1 ¼ 2particles=fL and n 2 ¼ 200particles=fL.Here, α ¼ 0.74 and H CR =H CB ¼ 0.86 are applied.This calculated growth speed matches well when the smaller NPs enclose a sufficient number of larger ones (the third image of Figure 3e).
Although the growth speed should be increased by n 2 , it grows faster for n 2 ≤ 100particles=fL.This faster growth is caused by the void volume, as shown in the first image of Figure 3e.The small number of 20 nm AuNPs leads a large void volume ratio when they are embedded between 80 nm AuNPs, correspondingly leading to a faster growth of the height.However, the void volume ratio gradually decreases when the number of 20 nm AuNPs is sufficiently larger than that of 80 nm AuNPs because the 20 nm AuNPs sufficiently enclose the 80 nm AuNPs (second and third images of Figure 3e).Using a simple calculation, one can determine the enclosing condition as 0.5 Here, 0.5 is multiplied because the neighboring 80 nm AuNPs share the enclosing surface.The number of required 20 nm AuNPs is 50 times more than the number of required 80 nm AuNPs, with the faster growth decreasing and converging while reaching the perfect enclosing conditions (100n 2 for 2n 1 , Figure 3e).
Ultimately, we can observe that four factors affect the growth speed of NP clusters: 1) the diffusion coefficient D (as a function of the temperature or atmospheric pressure), 2) the difference in the vapor concentration (c S À c 0 ) (as a function of the humidity), 3) the concentration of AuNPs, and 4) the tip size of the capillary tube.Equation (6) shows the state of the pillar growth with respect to the pulling speed of the substrate ḢS : 5) thickened for ḢPillar > ḢS , 6) stably built for ḢPillar ¼ ḢS , and 7) unstable and broken ḢPillar < ḢS .

Tuning the Resonant Wavelength Using Heterogeneous Structures
Fabricating self-assembled structures using a method based on evaporation can be useful for forming heterogeneous structures.
Figure 5a shows a binary heterogeneous half-donut structure where the ink consists of 80 nm AuNP and 80 nm silver NPs (AgNPs) with a number density of %5ðAuNPÞ∶1ðAgNPÞ.In Figure 5a, the SEM image (I) shows the self-assembled structure, with the magnified image (II) marked in a red box, representing the distribution of AgNP (dark spots) and AuNP (bright spots).It appears that, locally, the AgNP and AuNP are evenly distributed.The overall distribution of AgNP (III) and AuNP (IV) in the structure, obtained by energy-dispersive X-ray spectroscopy (EDS), shows that both are evenly distributed in the self-assembled structure, as shown in Figure 5a.Similar to the half-donut structure, a heterogeneous pillar composed of Au and Ag NPs can also be made into a structure in which the two types of NPs are evenly distributed, as shown in (V) and (VI) of Figure 5a.
The even distribution of a self-assembled structure is determined by the distribution states of the NPs dispersed in the ink.A homogeneous ink comprising AuNPs and AgNPs can be prepared by vortexing.As the NP have no charge, they exhibit Brownian motion without phase separation.That is, the ink retains homogeneity regardless of time.This leads to the accumulation distribution of NPs at the tip and represents a similar trend distributed in the ink.The accumulated ink transfers to the substrate in the contacting step, and then the distribution of NPs comprising the self-assembled structure also follows the accumulation process.
The optical properties of 3D binary colloidal clusters can be programmed based on their composition.Figure 5b shows the LSPR tuning of a binary colloidal cluster.A 3D binary colloidal cluster with a sufficiently high proportion of 20 nm AuNPs is expected to have the largest surface area per unit volume and the number of contact points (nanogaps).In higher-order plasmonic clusters, a subradiant mode appears at the same wavelength as the superradiant mode, but the full width at half maximum is narrower as the content of 20 nm AuNPs increases.The subradiant mode suppresses the scattering (and emission) of surface plasmon polaritons, and thus, a strong spectral dip appears in the center of the LSPR (the left of Figure 5b).The depths of the spectral dip is proportional to the content of the 20 nm AuNPs.We simulated models ranging from bare NP clusters to satellite NP clusters to understand the optical properties at the subwavelength scale.The densities of smaller NPs vary from low to high in satellite NP clusters and help analyze the plasmonic property differences.The geometrical (diameter) parameters are 80 nm for larger NPs and 20 nm for smaller NPs.The simulated scattered results for the nanostructures are displayed on the right of Figure 5b.In the absence of smaller NPs, the bare NP clusters scatter at a longer wavelength.The difference occurs upon the introduction of the smaller NPs.To explain the plasmonic characteristics, we can divide them into two categories.With an increasing density of smaller NPs, a dip at the vicinity of The magnified image of (II) from the (I) marked as red box exhibits the distribution of AuNPs (bright particles) and AgNPs (dark particles).The EDS images in (III) and (IV) directly show the distributions of the AuNPs and AgNPs, respectively.In (I), (III), and (IV), the scale bar is 2 μm.The scale bar of (II) is 0.5 μm.Images (V) and (VI) show EDS maps for a heterogeneous pillar fabricated by the ink mixed with 1.2 particles=fL of AuNP and 0.8particles=fL of AgNP, and the scale bar is 10 μm.b) Measured (left) and simulated LSPR spectra (right) of NP cluster micropillars fabricated with 0particles=fL (0), 50particles=fL (50), 100particles=fL (100), and 150particles=fL (150) of 20 nm AuNP solution mixed with 2 particles=fL of 80 nm AuNP solution.c) Quantitative EDS analysis results showing the composition of the heterogeneous cluster (left) according to the particle number density ratio of the heterogeneous colloidal solution, and LSPR spectra of the heterogeneous clusters (right).mode; in addition, the NP mode characteristics become visible near the 580 nm wavelength due to the increasing density of smaller NPs.Enhanced plasmonic properties (e.g., resonance scattering wavelength, optical modes, near-field enhancement properties) are observed in the nanogap, particularly with regard to smaller contact areas (e.g., hot spots or decreased interparticle distances), which result in increased near-field enhancement intensity.However, this intensity diminishes in 3D binary colloidal clusters, particularly when the interparticle distance increases to greater than 20 nm.Once optimal assembly quality is achieved, consistent plasmonic properties, such as scattering resonance wavelength and plasmonic modes, are expected.Our observations from 3D-printed architectures indicate that plasmonic scattering properties are inversely proportional to the concentration of 20 nm AuNPs.This can be understood based on the appearance of subradiant modes in higher-order plasmonic clusters.The nonradiative behavior exhibited by these subradiant modes results in a reduction of near-field intensity compared to that produced by superradiant modes, which feature radiative behavior.
The subradiant and superradiant modes of the NP cluster can be tuned according to the material composition.For example, a heterogeneous colloidal solution prepared by mixing two different materials (80 nm AuNPs and 80 nm AgNPs) was used to fabricate a heterogeneous cluster.An EDS mapping of a heterogeneous cluster fabricated with a heterogeneous colloidal solution composed of 1.2 particles/fL of AuNP solution and 0.8 particles/fL of AgNP solution showed well-distributed AuNPs and AgNPs across the pillar (V and VI of Figure 5a).In this case, the particle number density ratio between AuNP and AgNP corresponds to "3:2."Subsequently, a series of heterogeneous colloidal solutions was prepared with various particle number density ratios (7:3, 3:2, 2:3, and 3:7 heterogeneous colloidal solution).The total particle number density of all heterogeneous colloidal solution was fixed at 2 particles/fL.The detailed experimental conditions used to prepare the heterogeneous colloidal solutions are displayed in Experimental Section.The quantitative EDS results showed that the composition of the heterogeneous cluster was well controlled by the particle number density ratio of the heterogeneous colloidal solution (in the left of Figure 5c).The LSPR spectra of the heterogeneous clusters are also displayed (the right of Figure 5c).The colloidal cluster composed of AuNPs indicated a broad super-radiant mode across red in the visible-to-near-infrared regime.As the AgNP composition increased, the super-radiant mode of the heterogeneous cluster was blueshifted by plasmon hybridization. [38,39]

Dynamic Response of Janus Pillar
The micrometer fountain pen (i.e., the NFP) makes it possible to fabricate Janus pillars combined with biomaterial-based structures.Figure 6a and Movie S3, Supporting Information, show the fabrication of a Janus pillar using Ag NPs and the biomaterial of the M13 bacteriophage (M13).The top and bottom show a side-view image and illustration of each process, respectively.The Janus pillar is produced through the same manufacturing process as the previous 3D pillar.Two types of ink, where 80 nm AgNPs and M13 are dispersed, flow out through the NFP tip with two exit ports (inset of Figure 6a).Because the two inks exiting through the NFP tip dry very quickly, they do not mix in the pillar-making process.Thus, a Janus pillar with a clear boundary between the two materials is created, as shown in Figure 6b.The black area exhibits M13, and the bright area represents the AgNPs.The 3D-printing technique using this evaporation is highly suitable for making a structure with combined biomaterials and NPs.This indicates that it can also aid in developing numerous structures composed of polymeric materials or metallic NPs.Notably, M13 is NP with a diameter of 6.6 nm and a length of 880 nm.It is a functional biomaterial that can help detect various target substances by changing 2700 of the p8 proteins (that make up the body) using genetic engineering technology. [40,41]he Janus pillar can also be employed as a dynamic actuator capable of mechanical motion based on the different expansion rates of materials.M13 shows humidity-dependent swelling, whereas metal NPs do not.Therefore, the Janus pillar composed of M13 can expand due to changes in humidity; the corresponding Janus pillars shown in the pictures are bent (the first and second images of Figure 6c).Conversely, if the humidity is excessively low, M13 shrinks.Thus, the shape of the Janus pillar taken in a vacuum is bent toward M13 (the third image of Figure 6c).Figure 6d shows the bending of the Janus pillar with humidity.The bending increases rapidly when the humidity increases above 60%.This bending can be repeatedly conducted according to increases or decreases in humidity (Movie S4, Supporting Information).In the Janus pillar, the length of the part composed of the AuNPs does not change, but the length increases by up to 12% in the part composed of M13 (Figure 6e).This proves that M13 expands by %10% when exposed to high humidity, that is, the cause of the bending.In other words, this dynamic change from exposure to external gas means that it is possible to develop a Janus pillar composed of M13 and AgNPs as a dynamic actuator.2][43][44]

Conclusion
In this work, we presented 3D binary colloidal clusters fabricated by capillary action in a customized capillary tube.The customized capillary tube confined the NP solution at the femtoliter scale with a high evaporation rate originating from the high surface-to-volume ratio.The customized capillary tube makes possible the uniform self-assembly of colloidal clusters exhibiting a single or binary composition.In particular, we presented a step-by-step theoretical analysis of the self-assembly features.The balanced capillary action and evaporation of water from the CB and CR of the porous nanostructure were used to analyze each process, and the results agreed well with the experimental results.We also performed an LSPR characterization of the binary colloidal clusters in which the dark and bright modes were tuned for precise control of the nanostructure and composition, respectively.An NFP was used to fabricate a Janus pillar-like dynamic actuator consisting of M13 bacteriophages and AuNPs.Due to the capillary action balancing-based fabrication of colloidal clusters composed of well-distributed functional NPs, we expect binary colloidal clusters to pave the way for various applications, including those in sensing, photocatalysis, metamaterials, and 3D-integrated photonics.

Experimental Section
Preparation of NPs: We utilized polyvinylpyrrolidone (PVP)-coated 80 nm AuNPs, 20 nm AuNPs, and 80 nm AgNPs dispersed in water (Nanocomposix, San Diego, USA) as building blocks for the self-assembly processes.The concentration of the NP solution was increased by means of a centrifuge-based solvent transfer protocol.To maintain a quality of commercial NP solution, we followed the solvent transfer protocol provided by distributor.(https://cdn.shopify.com/s/files/1/0257/8237/files/nanoComposix_Polystyrene-Coated_Gold_Solvent_Transfer_Protocol.pdf ) (Nanocomposix, San Diego, USA).We applied 14 500 RCF for all of the solvent transfer processes while different centrifuge time was applied depending on the size of NPs (15 min for 20 nm NPs and 4 min for 80 nm NPs).After centrifugation, a clear and colorless supernatant was removed using a pipette with a fresh tip.As much supernatant as possible was removed while leaving the pellet intact.Finally, the pellet was resuspended with a desired volume of the solvent.We determined the NP concentration using UV-vis spectrometer (Evolution 300, Thermo Fisher Scientific, Waltham, USA).We measured the optical density (OD) of the new suspension by UV-vis followed by dividing the new suspension's OD by the material's OD*mL/mg value to calculate an approximate gold mass concentration in mg/ml.The OD*mL/mg value was determined by certificate of analysis provided by the distributor.In the case of heterogeneous colloidal solution, the particle number density of used AuNP solution and AgNP solution corresponded to the following: 3D Self-Assembly of the NPs: The 3D self-assembly setup was prepared by aligning the instrumental parts including a three-axis (xyz configuration) motorized stage equipped with the backlash-free ballscrew and the DC motors (M-VP-25XA-XYZR, Newport, Irvine, CA 92 606, USA), three-axis (xyz configuration) manual stage (PT3A/M, Thorlabs, Newton, NJ 07860, USA), home-made pipette holder attached to the manual stage, light source equipped with a halogen lamp (OSL2, Thorlabs, Newton, NJ 07860, USA), 20Â objective lens (MY20X-804, Mitutoyo, Kanagawa, 213-8533, Japan), and charge-coupled device camera.A commercial glass pipette puller (P-97, Sutter Instruments, Novato, CA 94949, USA) was used to prepare the tapered glass capillary tube with the micrometer-sized tip.The pulling-up process was carefully controlled with a motorized stage controller (XPS-D4, Newport, Irvine, CA 92606, USA).The micropipette fabricated with a glass pipette puller was installed to the home-made pipette holder.To observe and control the printing process, the tip of the micropipette was positioned to the focal point of the objective lens using the manual stage.The NP solution was injected to the micropipette using the syringe.Because the micropipette was fabricated with the glass capillary equipped with the filament at an inside wall (1B100F-6, World Precision Instruments, Sarasota, USA), the NP solution immediately moved to the tip of the micropipette by means of capillary force.Using the motorized stage, the micropipette approached the substrate (silicon wafer) until it directly touched the substrate.The fluctuation of the NP solution at the tip of the micropipette was observed when the micropipette directly touched the substrate.Fabrication of the NP cluster was carried out with a desired pulling-up speed which was applied and controlled by the motorized stage and motorized stage controller, respectively.After fabrication of the NP cluster, the micropipette was detached from the NP cluster by applying fast pulling-up speed (25 mm s À1 ).
Scanning Electron Microscope and Atomic Force Microscope: The SEM images were captured using FESEM (Hitachi, S-4700, PNU Center for Research Facilities, Busan, Korea).Before every measurement, a titanium coating was used to capture high-quality images.Additionally, FIB milling (Tescan, LYRA 1 XMH, Busan Technopark, Busan, Korea) was conducted to observe the cross sections.
Images of the self-assembled structures were obtained using an SEM (JSM-7900F, JEOL Ltd.), which was also used to perform the EDS.An NX10 AFM system was used to measure the structural shapes of the self-assembled structures in a noncontact mode using the data acquisition program XEP 3.0.4.The AFM images were analyzed using the XEI 1.8.2 image processing program (Park Systems).

Figure 1 .
Figure 1.Fountain pen-inspired writing on micrometer scales.a) Scheme of writing using a fountain pen.b) Ultrafine fountain pen for NP dispersed ink.The scale bar represents 5 μm.c) Scheme of point-like colloidal assembly.d) Micrometer-sized text written by the colloidal assembly (left) and SEM image of a half-donut structure (right).The scale bars of the left and right images represent 50 and 1 μm, respectively.e) Scheme of 3D colloidal assembly.f ) Diverse 3D colloidal assemblies (left) and packing of NP (right).The scale bar represents 10 μm (black) and 1 μm (white).

Figure 2 .
Figure 2. Characteristics of point-like colloidal assembly.a) Fabrication process of the half-donut structure.The bottom pictures depict four steps of the fabrication process observed in the top pictures.The black dashed line represents the interface of the substrate.The black shadow under the black dashed line is a mirror image of the NFP.The scale bar represents 20 μm.b) Diagram of colloidal assembly in the context of the accumulation time and AuNP concentration.c) SEM images of assembled structures.The images are obtained under 45°tilted conditions.The scale bar represents 2 μm.d) AuNP packing of assembled structures.The assembled structures exhibit hexagonal (red line) and pentagonal packing (blue line).The scale bar represents 100nm.e) Schematic diagram for analyzing the half-donut structures.f ) Height profile functional to radial distance from the center of mass.The height profiles are measured for 0.2particles=fL of 80 nm-AuNPs to display varying half-donut structures.The inset exhibits the normalized height profile where the maximum height is fit to 1. e) Volumes of half-donut structures according to accumulation time.The right-handed axis represents the total number of 80 nm AuNPs in the self-assembled structures. .

Figure 3 .
Figure 3. Vertical growth of a binary colloidal cluster.a) Series of optical micrographs showing the vertical growth of the binary colloidal cluster.The scale bar represents 50 μm.b) Available growth speeds based on the 80 nm AuNP solution.c) Available growth speeds based on the 20 nm AuNP solution mixed with 2particles=fL of 80 nm AuNP solution.d) SEM images of the micropillars marked as I, II, III, and IV in (c).The scale bar represents 10 μm.e) SEM images of the nanostructures of micropillars marked as I, II, and III in (d).The scale bar represents 200nm.f ) FESEM image of a micropillar milled with FIB.The scale bar represents 5 μm.g) FESEM image of the cross section of a micropillar composed of a single (left) and binary composition (right).The scale bar represents 200nm.

Figure 4 .
Figure 4. Femtoliter-scaled fabrication mechanism of half-donut structure.a) Side-view images of 80 nm AuNPs accumulation at the tip.H A represents the accumulation height of 80 nm AuNPs at the NFP tip inside.The scale bar represents 20 μm.b) Accumulation height according to accumulation time t A .c) Evaporation-driven AuNP accumulation depending on t A .The direction and magnitude of the green arrow represent the fluid flow of the ink.The orange arrows show the flux of the water vapor ( J) at the interface (marked with a red dotted line).d) Clogging of the NFP due to overaccumulation.e) Evaporation-induced relocation of AuNPs in the CB.f ) Self-assembly of half-donut structure depending on the number density of the AuNPs.The black and red arrows represent Marangoni and capillary flows, respectively.The dashed and dotted lines sequentially show the receding interface of the sessile ink droplet due to drying.g) Enhanced growth speed of 3D pillar due to CR.

Figure 5 .
Figure 5. Optical properties of a binary colloidal cluster.a) Self-assembled heterogeneous structure consisting of AuNPs and silver NPs (AgNPs).Image (I) shows the SEM image of the half-donut structure consisting of 80 nm AuNPs and 80 nm AgNPs.The number density ratio of AuNP to AgNP is %5:1.The magnified image of (II) from the (I) marked as red box exhibits the distribution of AuNPs (bright particles) and AgNPs (dark particles).The EDS images in (III) and (IV) directly show the distributions of the AuNPs and AgNPs, respectively.In (I), (III), and (IV), the scale bar is 2 μm.The scale bar of (II) is 0.5 μm.Images (V) and (VI) show EDS maps for a heterogeneous pillar fabricated by the ink mixed with 1.2 particles=fL of AuNP and 0.8particles=fL of AgNP, and the scale bar is 10 μm.b) Measured (left) and simulated LSPR spectra (right) of NP cluster micropillars fabricated with 0particles=fL (0), 50particles=fL (50), 100particles=fL (100), and 150particles=fL (150) of 20 nm AuNP solution mixed with 2 particles=fL of 80 nm AuNP solution.c) Quantitative EDS analysis results showing the composition of the heterogeneous cluster (left) according to the particle number density ratio of the heterogeneous colloidal solution, and LSPR spectra of the heterogeneous clusters (right).

Figure 6 .
Figure 6.Dynamic Janus pillar comprising M13 bacteriophages (M13) and AgNPs.a) Fabrication process of Janus pillar (top) and schematics (bottom).The scale bar represents 25 μm.The inset shows a dual injecting tip with 10 μm for the scale bar.The sphere and short line represent the AgNPs and M13, respectively.b) SEM image showing a cross section of the Janus pillar.The scale bar of the top and bottom image represents 4 and 1 μm, respectively.In the bottom image, the dark and bright areas represent the M13 and AgNPs, respectively.c) Dynamic motion of the Janus pillar depending on humidity.The left and middle image show a bending motion of the Janus pillar according to the RH, where the scale bar represents 20 μm.The SEM image (right) shows the bending of the Janus pillar in a vacuum with a scale bar of 10 μm.d) RH-dependent dynamic motion of the Janus pillar.e) Length change ratio of an M13/AgNP scaffold.