Automated pick-and-place of single nanoparticle using electrically controlled low-surface energy nanotweezer

Techniques capable of building functional nanodevices with synthesized nanomaterials have been dreamed of for decades.¹ Today, various high-quality functional nanoparticles have been successfully synthesized, but there is still a lack of techniques to assemble these nanometer building blocks into 3D nanostructures. One way to address this issue is to use the tip-based particle manipulation technique. For 2D structures, one can simply build them by pushing and pulling particles using an atomic force microscope (AFM) tip.^2–4 To build 3D structures, the pick-and-place technique was developed using dual-probe systems, which function like nano-chopsticks.^5,6 However, it is extremely challenging to handle particles smaller than the tip using the dual-probe system, and the smallest particle manipulated by this technique is 80 nm.⁷ Interestingly, it was shown that the pick-and-place of a nanoparticle may actually be carried out with a single tip by mechanically pressing the nanoparticle,⁷ but this method suffers a low success rate (only 1/3 of particles can be released after being picked). It is because the pick step and the place step require contradictive relative interaction strengths between the tip–particle interaction and particle–substrate interaction.

To address the above-mentioned issue, techniques capable of tuning the tip–particle interaction from being stronger to being weaker than the substrate–particle interaction are required. One of such techniques is using electrical forces. It has been demonstrated that nanoparticles can be captured and released by a conductive AFM tip using dielectrophoretic forces,^8–10 but this method only works in solution, limiting them from being applied to many important applications. For ambient conditions, one needs to find a way to make the surface energy of the tip considerably lower than that of the substrate in order to be able to place a nanoparticle to a substrate after the pick step. Fortunately, the authors have found a way to do so using the fluoride deposition technique, and by applying the surface energy engineering technique together with electrostatic interactions, we have recently demonstrated a highly efficient nanoxerography technique with an extremely low nonspecific absorption ratio.¹¹ In this work, we extend the method to AFM tips and develop reliable electrically controlled low surface energy nanotweezers for the pick-and-place of single nanoparticles. Moreover, in order to improve the efficiency of the pick-and-place technique, a computer-vision-based automation program is built, which allows us to automatically pick, move, and place single luminescent nanoparticles.

Figure 1(a) shows the schematic diagram of the experimental setup and the major steps of the pick-and-place process. It consists of an inverted fluorescence microscope (Olympus IX73) and a commercial atomic force microscope (AFM, NT-MDT Netegra), which are used for guiding and performing the pick-and-place process, respectively. To increase the accuracy and signal strength, an oil immersion objective lens (Olympus Apo N 60XO, NA = 1.49) was used to collect the optical signal from the tip apex and nanoparticles, and an EM-CMOS camera (Tuscen) was used to record all the signals.

In this work, upconversion luminescent nanoparticles (NaYF₄:Yb, Er nanoparticles) were used. They were synthesized using the heating-up technique, and the resulting nanoparticles were rod-shaped (aspect ratio of about 3:2, 30 nm wide, 45 nm long) with great uniformity (size variation <7%), dispersed in a nonpolar solvent. Before using them, the particles were spin-casted on an indium tin oxide (ITO) coated cover glass and then imaged using an inverted microscope with 980 nm excitation (Thorlabs CLD1015). The particles were sparsely distributed on the substrate without aggregation, as shown in Fig. 1(b).

Commercial conductive probes (NT-MDT&TipsNano HA_NC) were used for the pick-and-place experiments [Fig. 1(c)]. Before use, a thin layer (10 nm) of fluorocarbon film was conformally deposited on the tip using plasma deposition to reduce the van der Waals interactions between the tip and the nanoparticles. The key advantage of the low surface energy conductive AFM probe is that the tip-particle interaction strength can be tuned over a broad range, from being lower to being higher than the substrate–particle interaction strength, when a voltage is applied. This will allow us to pick up and release a nanoparticle from and to a substrate by simply tuning the voltage applied to the tip.

The pick-and-place procedure was guided by the optical imaging system. With white light illumination from the side, the tip apex can be clearly visualized, allowing us to locate the tip apex. To locate the NaYF₄:Yb, Er nanoparticles, fluorescence images were recorded by using the 980 nm excitation.

Using the method described above, we can pick and place single nanoparticles with a high success rate (>90%, 103 in 112 tries) and manipulate single nanoparticles into an arbitrary pattern. Figure 1(d) shows an example, in which a mini-“NJU” was built by picking and placing 21 NaYF₄:Yb, Er nanoparticles one-by-one.

Figure 2 shows the detailed pick-and-place process, which can be divided into four steps, namely, positioning, picking, transferring, and placing.

1. Rough positioning: Move the tip into the field of view of the optical microscope, land the tip, and find the position of the tip apex using the optical image with the help of white light illumination from the side [Fig. 2(a)]. Then, switch to the fluorescence mode, and find a target nanoparticle. After that, move the tip apex to the vicinity (within a range of 3 µm) of the target particle.

   At this time, no voltage is applied on the tip, and the fluoride polymer coating on the tip guarantees a low surface energy and consequently a small possibility for particle absorption. Comprehensively, a free nanoparticle in air does not experience any interactions, and therefore, its potential U_air = 0. After the nanoparticle is absorbed into the surface of a tip or the substrate, it will release a small amount of energy, $Etip0$ or E_sub, respectively, because of the van der Waals forces, and its potential will become $Utip=−Etip0$ or U_sub = −E_sub, respectively [Fig. 2(b)]. Here, due to the fluoride coating, U_tip > U_sub, and the nanoparticle will stay on the substrate.

2. Particle pick-up: Apply voltage (typically −10 V) on the tip, and scan the tip over the location of the target nanoparticle (∼5 µm²) using tapping mode. When the tip scans over the target nanoparticle, the particle will be absorbed into the tip apex due to the dielectrophoretic interactions [Fig. 2(c)]. After that, retract the tip from the substrate, and the particle will be lifted from the substrate due to the larger absorption energy of the tip.

   From the energy point of view, the dielectrophoretic interactions lead to additional absorption energy, E_die, and then, the total energy released during absorption becomes $Etip=Etip0+Edie$⁠. Since it is larger than E_sub, the particle pick-up step becomes possible, as shown in Fig. 2(d).

3. Particle manipulation: After the pick-up step, we are able to manipulate the nanoparticle in 3D freely using the probe, which is controlled by a nanometer-precision piezo tube. Here, for simplicity, we lift the nanoparticle ∼2 µm from the surface of the substrate and move it to the predefined position with the help of the guidance of the optical microscope [Fig. 2(e)].

   When the nanoparticle is moved, the bright fluorescent spot will become a bright line since the particle will move together with the tip for several micrometers during the long exposure time (∼1 s), as depicted in Fig. 2(e). It is worth mentioning that the particle can be located easily by simply using the optical image precision,¹² making this method simple and precise.

4. Particle placing: After moving the nanoparticle to the predefined position, we turn off the voltage applied to the probe, and land the tip on the substrate. The nanoparticle will jump back and absorb into the substrate [Fig. 2(g)] because the substrate surface offers a larger absorption energy (E_sub) than the tip (⁠$Etip0$⁠), as shown in Fig. 2(h). Finally, we can lift the tip up from the substrate, and the nanoparticle will stay on the substrate.

The above-mentioned steps can be technically divided into a series of simple operations of the tip, which are suitable for automation. Here, the automation was realized using Python script (website: https://github.com/weihualab/auto-nanomanipulation). The program can control the commercial hardware and software, pick a pre-assigned nanoparticle, and further move and place it to a given location automatically without any human intervention (see the attached video in the supplementary material).

Comprehensively, the program can control the illumination mode and directly read out the locations of the tip apex and luminescent nanoparticles in the field of view using computer vision techniques. By using the PyAutoGUI library (website: https://github.com/asweigart/pyautogui), the program can also read the interface of the AFM commercial software and further automatically operate it (e.g., tip approach, retraction, scan, and move) following the steps described above. Here, we would like to emphasize that this technique can be applied to any commercial AFM instruments (or any instruments running on Windows OS) since the method used here can be used for automation of any software running on Windows OS.

From the above-mentioned discussion, we can see that, fundamentally, whether the particle is on the tip or the substrate is dependent on the relative absorption energy of the particle on the tip and the substrate. In the particle pick-up and placing steps, the tip is tapping the substrate at 235 kHz, and the nanoparticle has enough chance to overcome the energy barrier and hop from one surface to another. Assuming that jumping occurs randomly and the number of particles in contact with the tip and the substrate is large (>1 × 10⁴ times considering that operation time is typically longer than 0.1 s), the possibility of the nanoparticle staying on the tip or the substrate will follow Boltzmann distribution P ∼ exp(−U/kbT), where U is U_tip = −E_tip or U_sub = −E_sub, respectively.

In this work, U_tip and U_sub can be quantitatively determined. We first calculate the van der Waals interactions in the system. For simplicity, we use spheres to model the nanoparticle and the tip apex [Fig. 3(a)]. Then, the absorption energy, E_abs, caused by tip–particle and tip–substrate interactions can be calculated using a unified sphere–sphere interaction model,¹³ 

where W is the adhesion work, and $a0=(6πR2W/K)1/3$ is the contact radius, which is decided by the geometrical parameter R = R₁R₂/(R₁ + R₂), combined elastic modulus K, and adhesion work W. In this experiment, R₁ = 15 nm is the radius of the nanoparticle, R₂ = 40 nm and ∞ are the radius of the tip and the substrate, respectively,^13,K = 4584 Mpa and 4593 Mpa are the combined elastic modulus at the tip and the substrate, respectively,^14–16 and W = γ_sl + γ_pl − γ_sp. Here, γ_sl, γ_pl, and γ_sp are the surface energy of the solid–liquid, particle–liquid, and solid–particle system (solid corresponds to the substrate/tip material), respectively.

The surface energy can be obtained from the contact angle experiment. In our case, the contact angle of water on the ITO covered glass (i.e., the substrate) and fluoride polymer coated Pt/Au (the tip surface) is 42° and 72° [Fig. 3(b)], respectively. The surface energy of water is 72.8 mJ/m² at a temperature of 20 °C, and the surface energy of the substrate and the tip can then be calculated using the Owens–Wendt–Kaelble method,¹⁷ γ_sub = 184.7 mJ/m² and γ_tip = 104.1 mJ/m². As to the surface energy of the NaYF₄:Yb, Er nanoparticle, we assume that it is the same as that of the bulk material¹⁸ (γ_np = 97.6 mJ/m²). After incorporating all these values into Eq. (1), we can finally obtain the absorption energy E_sub = 3.02 × 10⁻¹⁹ J and $Etip0=2.21×10−19J$⁠.

The dielectrophoretic interaction induced additional tip-particle absorption energy can be calculated with the help of static electric modeling. The dielectrophoretic forces, F_die, are caused by the polarization of a particle in external inhomogeneous electric fields, E,

where

Here, r is the radius of the nanoparticle, and ɛ_r and ɛ₀ are the dielectric constant of NaYF₄:Yb, Er and vacuum, respectively.¹⁹ The absorption energy can then be written as E_die = 1/2 α |E|².

At the tip apex, strongly localized electric fields will be formed when a static voltage is applied to the tip due the “lightning rod” effect,²⁰ and it will induce strong dielectrophoretic interactions. This process can be numerically modeled using the finite-element method (FEM). Comprehensively, we calculated the field distribution at the tip apex using the realistic 3D geometrical parameter with the help of a commercial FEM solver (Comsol Multiphysics). Fine (minimum 0.2 nm) meshes were used at the tip apex to guarantee the accuracy of the simulation. Figure 3(c) shows the result. As expected, the electric fields are highly localized at the tip apex with the field intensity reaching 3.89 × 10⁸ V/m. This leads to significant modulation of absorption energy of the nanoparticle at the tip apex, E_die = 1.55 × 10⁻¹⁹ [Fig. 3(d)]. Here, E_die is considerably larger than the difference of the van der Waals interaction induced absorption energy between the substrate and the tip, $Esub−Etip0$⁠, and this large energy modulation allows us to effectively control the pick-and-place of a NaYF₄:Yb, Er nanoparticle.

It worth noting that the dielectrophoretic interaction and van der Waals force follow different scaling laws. The dielectrophoretic interaction is proportional to polarizability and scales up cubically with the particle size, while the van der Waals interaction is approximately proportional to the 4/3 power of the particle diameter. As a result, when the particle size decreases, the modulation term, E_die, will become too small to make the pick-and-place process efficient, as shown in Fig. 4. Indeed, we observed this effect in our experiment. When smaller particles [e.g., 10 nm quantum dots (QDs)] were used, the pick-and-place experiment becomes challenging.

Finally, in the process of placing, we found that a considerable proportion (30%–40%) of the nanoparticles could not be desorbed by simply turning off the voltage on the tip. One possible reason is that in an ambient environment, there is always a thin layer of water film at the tip, which further complicates the interactions.²¹ To solve this issue, we tried applying triangular pulses (∓10 V, 100 Hz) instead of simply turning off the DC voltage on the tip and found that it can significantly improve the particle-placing effectiveness (from 60% to >90%).

In summary, we developed automated AFM-based electric nanotweezers for the pick-and-place of single luminescent nanoparticles. The nanotweezers are made of fluoride-coated conductive AFM probes, and the strength of tip–particle interaction can be tuned smaller or larger than that of substrate–particle interaction by simply switching the voltage applied on the tip off or on. This electrical tuning technique makes the pick-and-place of nanoparticles controllable and efficient. Furthermore, we automized the full pick-and-place process using a Python program. It is capable of automatically finding and monitoring the locations of the tip apex and the nanoparticle and further completing the pick-and-place operation by controlling both software and hardware of the AFM, as well as other accessories including the camera and illumination. We believe that, with such simplicity, high reliability, and versatility, the automatic nanotweezers developed in this work can be applied to many important fields.

See the supplementary material for the automatic pick-and-place procedure of a single upconversion luminescent nanoparticle.

This work was supported by the National Key Research and Development Program of China (Grant No. 2016YFA0201104).