Additive Manufacturing of Sub-Micron to Sub-mm Metal Structures with Hollow AFM Cantilevers

We describe our force-controlled 3D printing method for layer-by-layer additive micromanufacturing (µAM) of metal microstructures. Hollow atomic force microscopy cantilevers are utilized to locally dispense metal ions in a standard 3-electrode electrochemical cell, enabling a confined electroplating reaction. The deflection feedback signal enables the live monitoring of the voxel growth and the consequent automation of the printing protocol in a layer-by-layer fashion for the fabrication of arbitrary-shaped geometries. In a second step, we investigated the effect of the free parameters (aperture diameter, applied pressure, and applied plating potential) on the voxel size, which enabled us to tune the voxel dimensions on-the-fly, as well as to produce objects spanning at least two orders of magnitude in each direction. As a concrete example, we printed two different replicas of Michelangelo’s David. Copper was used as metal, but the process can in principle be extended to all metals that are macroscopically electroplated in a standard way.


Metal Additive Manufacturing at the Micro Scale
At the macro-and mesoscale (down to 100 µm), additive manufacturing (AM) of metal objects is recurrently performed with robust methods like selective laser and electron beam melting, both relying on the local fusion of metal particles to obtain a solid metal with the desired shape [1][2][3][4][5]. Typical voxel sizes are reported in the range of 100 to 400 µm [6] As alternatives, one can take into consideration the well-established localized electrochemical deposition (LED) [7,8] and laser chemical vapor deposition (LCVD) [9] methods, whose minimum feature sizes are continuously improved down to 10 µm [10,11] and correspondingly interpreted [12,13]. Restricting the depiction only to the fabrication of metal objects at the micron scale (i.e., with details smaller than 10 µm, µAM), three main strategies are being established [14]: fabrication of templates by micro-stereolithography to be successively metallized either by coating (positive templates) or by electroplating (negative templates) [15][16][17][18], transfer of metallic materials to be eventually sintered in a second step, or in situ metal synthesis. As an example of the "transfer" strategy, we can refer to direct ink writing [19][20][21], electrohydrodynamic printing [22][23][24], laser-assisted electrophoretic deposition [25], laser-induced forward transfer [26], melt droplets [27]. On the other hand, focused ion/electron beam methods (metal precursor dissociation) [28][29][30], as well as laser-induced photoreduction [31][32][33], together with electrochemical reduction (meniscus-confined) [34,35], local dispensing [36,37], electrohydrodynamic  [36] the simple case of two pillars side-by-side: we emphasize the layer-by-layer fabrication (layers I-III are labeled), i.e., the pillars are printed in parallel and not in series. a) The ion tip filled with CuSO4 solution is positioned over the first pillar at a set separation (e.g., 500 nm) where the metal voxel is to be deposited. Local electroplating is switched on at a given overpressure, leading to local pillar growth (voxel III). b) When the growing voxel touches the pyramidal apex, a cantilever deflection is detected on the photodiode via the moving laser beam. The inset graph shows the temporal evolution of the deflection signal (defl.) for a voxel touching event (yellow segment). c) As soon as this touching event is recognized by the software, the probe is moved to the next position, i.e., on top of the second pillar again with a typical separation of 500 nm, and the voxel ec deposition is started. Reprinted with permission from Ref. 36 We conceived this contribution to describe the FCEP in detail revisiting our two main publications [36,51] but also presenting complementary unpublished results.

The Force-Controlled Microprinting Tool
Microchanneled AFM probes. AFM cantilevers were produced with an embedded microchannel connecting the hollow pyramidal tip on one side and a macro reservoir on the other side (ion tips, Exaddon AG). Such microchannels were obtained either by etching a sacrificial polysilicon layer between two layers of Si3N4 according to the batch processes reported in [52,53]. The standard section of the microchannel is of 20 µm × 1 µm, whereas the pyramid is 10 µm × 10 µm × 7 µm with a circular aperture at the apex either of 300 nm or of 500 nm diameter ( Figure 2). Alternatively, for apex apertures of different sizes, we opened so called "closed" probes (i.e., having no aperture at the apex) by means of a focused ion beam (FIB) milling. In this case, the probes were mounted in a custom probe holder and coated with an 18-nm-thick carbon layer using a CCU-010 Carbon Coater (Safematic GmbH, Bad Ragaz, Switzerland) before milling by a FIB-scanning electron microscope (SEM) Nvision 40 device (Zeiss, Oberkochen, Germany) with the SmartSEM software (Zeiss). The milled face of the pyramidal probe was aligned in parallel with the FIB-beam equipped with a gallium ion source. Subsequently, the probes were milled with an acceleration voltage of 30 kV at milling currents varying from 10 pA up to 80 pA, depending on the desired opening. The active milling process was followed in live SEM mode and was only terminated when the desired opening size was reached. After the milling process, the probes were glued onto a dedicated printing holder (Exaddon AG). AFM approaches were always performed in contact mode in the standard optical beam deflection method.  [36] the simple case of two pillars side-by-side: we emphasize the layer-by-layer fabrication (layers I-III are labeled), i.e., the pillars are printed in parallel and not in series. (a) The ion tip filled with CuSO 4 solution is positioned over the first pillar at a set separation (e.g., 500 nm) where the metal voxel is to be deposited. Local electroplating is switched on at a given overpressure, leading to local pillar growth (voxel III). (b) When the growing voxel touches the pyramidal apex, a cantilever deflection is detected on the photodiode via the moving laser beam. The inset graph shows the temporal evolution of the deflection signal (defl.) for a voxel touching event (yellow segment). (c) As soon as this touching event is recognized by the software, the probe is moved to the next position, i.e., on top of the second pillar again with a typical separation of 500 nm, and the voxel ec deposition is started. Reprinted with permission from Ref. [36]. Copyright 2017 John Wiley & Sons Inc.
We conceived this contribution to describe the FCEP in detail revisiting our two main publications [36,51] but also presenting complementary unpublished results.

The Force-Controlled Microprinting Tool
Microchanneled AFM probes. AFM cantilevers were produced with an embedded microchannel connecting the hollow pyramidal tip on one side and a macro reservoir on the other side (ion tips, Exaddon AG). Such microchannels were obtained either by etching a sacrificial polysilicon layer between two layers of Si 3 N 4 according to the batch processes reported in [52,53]. The standard section of the microchannel is of 20 µm × 1 µm, whereas the pyramid is 10 µm × 10 µm × 7 µm with a circular aperture at the apex either of 300 nm or of 500 nm diameter ( Figure 2). Alternatively, for apex apertures of different sizes, we opened so called "closed" probes (i.e., having no aperture at the apex) by means of a focused ion beam (FIB) milling. In this case, the probes were mounted in a custom probe holder and coated with an 18-nm-thick carbon layer using a CCU-010 Carbon Coater (Safematic GmbH, Bad Ragaz, Switzerland) before milling by a FIB-scanning electron microscope (SEM) Nvision 40 device (Zeiss, Oberkochen, Germany) with the SmartSEM software (Zeiss). The milled face of the pyramidal probe was aligned in parallel with the FIB-beam equipped with a gallium ion source. Subsequently, the probes were milled with an acceleration voltage of 30 kV at milling currents varying from 10 pA up to 80 pA, depending on the desired opening. The active milling process was followed in live SEM mode and was only terminated when the desired opening size was reached. After the milling process, the probes were glued onto a dedicated printing holder (Exaddon AG). AFM approaches were always performed in contact mode in the standard optical beam deflection method. ec cell. The first version of the electrochemical cell consisted of gold 15-nm-thin films on glass with 3 nm Ti adhesion layer as WE, an Ag wire as quasi-reference electrode (RE), and a Pt wire as CE in a simple droplet of H2SO4 at pH 3 [36]. The second version of the printing chamber (i.e., the threeelectrode cell) was redesigned in a modular fashion with the parts made out of solid Teflon [51]. The working electrode is connected to the printing substrate of 12 mm × 12 mm; furthermore, a graphite counter electrode and a Ag/AgCl wire are used as reference. As WE, 100-nm-thick films of sputtered Cu on a 1.3 cm × 1.3 cm silicon substrate with 13 nm of Ti adhesion layer were used with a graphite CE and a Ag/AgCl wire as RE.
The supporting electrolyte was a 0.5 M H2SO4 with the addition of HCl to a concentration of 0.5 mM. The plating solution was a 0.8 M CuSO4 in 0.5 M H2SO4 solution.
Setup. The prototype printer was obtained assembling an AFM head (Nanowizard I by JPK, Berlin, Germany), a pressure controller (MFCS-4C by Fluigent, Villejuif, France), and a potentiostat (PalmSens, Utrecht, The Netherlands) to be synchronized together [36]. Automation was achieved by a custom-made LabVIEW program which controlled the probe position and read the deflection signal from the AFM via two data acquisition cards (NI-USB 6009 and NI-USB 6343, National Instruments, Ennetbaden, Switzerland). However, for sizes in the z direction exceeding the piezo range of 15 µm, further z increments could only be accomplished with the AFM stepper motors, whereby their operation introduced x-y offsets which had to be accounted for by printing an additional pillar close to each structure. The tip end of this continually increasing reference pillar had to be found after each stepper movement by automatically performing a rough AFM scan in the pillar area at a z value corresponding to the height of the current pillar. The x-y error introduced was then determined and successfully compensated, but at the cost of important time delays. To get rid of this deficiency, the printing tool was completely redesigned [51]. Here, we introduce a completely redesigned system as far as both the hardware and the software are concerned. A controller (Exaddon AG) was conceived ec cell. The first version of the electrochemical cell consisted of gold 15-nm-thin films on glass with 3 nm Ti adhesion layer as WE, an Ag wire as quasi-reference electrode (RE), and a Pt wire as CE in a simple droplet of H 2 SO 4 at pH 3 [36]. The second version of the printing chamber (i.e., the three-electrode cell) was redesigned in a modular fashion with the parts made out of solid Teflon [51]. The working electrode is connected to the printing substrate of 12 mm × 12 mm; furthermore, a graphite counter electrode and a Ag/AgCl wire are used as reference. As WE, 100-nm-thick films of sputtered Cu on a 1.3 cm × 1.3 cm silicon substrate with 13 nm of Ti adhesion layer were used with a graphite CE and a Ag/AgCl wire as RE.
The supporting electrolyte was a 0.5 M H 2 SO 4 with the addition of HCl to a concentration of 0.5 mM. The plating solution was a 0.8 M CuSO 4 in 0.5 M H 2 SO 4 solution.
Setup. The prototype printer was obtained assembling an AFM head (Nanowizard I by JPK, Berlin, Germany), a pressure controller (MFCS-4C by Fluigent, Villejuif, France), and a potentiostat (PalmSens, Utrecht, The Netherlands) to be synchronized together [36]. Automation was achieved by a custom-made LabVIEW program which controlled the probe position and read the deflection signal from the AFM via two data acquisition cards (NI-USB 6009 and NI-USB 6343, National Instruments, Ennetbaden, Switzerland). However, for sizes in the z direction exceeding the piezo range of 15 µm, further z increments could only be accomplished with the AFM stepper motors, whereby their operation introduced x-y offsets which had to be accounted for by printing an additional pillar close to each structure. The tip end of this continually increasing reference pillar had to be found after each stepper movement by automatically performing a rough AFM scan in the pillar area at a z value corresponding to the height of the current pillar. The x-y error introduced was then determined and successfully compensated, but at the cost of important time delays. To get rid of this deficiency, the printing tool was completely redesigned [51]. Here, we introduce a completely redesigned system as far as both the hardware and the software are concerned. A controller (Exaddon AG) was conceived with a low latency and a fast feedback loop in order to react to a touching event ( Figure 3) within 1 ms. In this way, possible clogging of the opening is minimized, thus increasing the probe lifetime while higher printing rates are also permitted. On the other side, the controller buffer is able to contain the next 16 voxel coordinates to avoid any delay by a lack of voxel coordinates in the controller for stable and long-term operation up to tens of hours. Further components of the hardware of the new system are two high-resolution optical systems (top and bottom view) for loading of the nozzle, for printer adjustment and calibration, as well as for imaging of the printed structures. The maximum volume of the printing space is 200 mm × 70 mm × 60 mm, assured by a three-axis positioning stage (Exaddon AG) consisting of three linear motors with an x-y positioning accuracy (full-range, repeatability) of 250 nm, as well as with 5 nm positioning accuracy in z at a 0.1 nm sensor resolution. The corresponding repositioning time from a voxel to the following one takes place in less than 50 ms (for a ∆z of 500 nm). with a low latency and a fast feedback loop in order to react to a touching event ( Figure 3) within 1 ms. In this way, possible clogging of the opening is minimized, thus increasing the probe lifetime while higher printing rates are also permitted. On the other side, the controller buffer is able to contain the next 16 voxel coordinates to avoid any delay by a lack of voxel coordinates in the controller for stable and long-term operation up to tens of hours. Further components of the hardware of the new system are two high-resolution optical systems (top and bottom view) for loading of the nozzle, for printer adjustment and calibration, as well as for imaging of the printed structures. The maximum volume of the printing space is 200 mm × 70 mm × 60 mm, assured by a three-axis positioning stage (Exaddon AG) consisting of three linear motors with an x-y positioning accuracy (full-range, repeatability) of 250 nm, as well as with 5 nm positioning accuracy in z at a 0.1 nm sensor resolution. The corresponding repositioning time from a voxel to the following one takes place in less than 50 ms (for a Δz of 500 nm).

Tuning the Voxel Size at Different Scales: from 0.5 to 20 µm
The ec µAM system is a multiparameter instrument, whereby the value of the probe aperture Deq, of the applied pressure p, of the electrodeposition potential E as well as the voxel height Δz can be precisely adjusted.
In a previous work [51], we investigated the effect of the ion tip aperture Deq and p on the diameter d of a pillar printed at fixed potential ( = −0.5 V vs Ag-AgCl) and geometry (i.e., distance between two consecutive voxels Δz, here referred to as voxel height) as in Figure 4a. The framework consists in measuring the pillar diameter from SEM images (Figure 4a, all the d values are compiled in Figure 4b) and the vertical speed from the printer logged z-axis position. These two experimental values can then be combined to calculate the volumetric deposition speed :  . Schematic of the force-controlled ec µAM system. On the left, a system computer sends commands to the system control unit on the right. The system control unit governs the printing process using an embedded controller. The ion tip is mounted on the z-stage in the printing head and is moved inside the ec deposition cell by the z-and x-y stages. A microfluidics control system with 1 mbar precision regulates the electrolyte flow through the cantilever aperture. Adapted with permission from Ref. [51]. Copyright 2019 John Wiley & Sons Inc.

Tuning the Voxel Size at Different Scales: from 0.5 to 20 µm
The ec µAM system is a multiparameter instrument, whereby the value of the probe aperture D eq , of the applied pressure p, of the electrodeposition potential E as well as the voxel height ∆z can be precisely adjusted.
In a previous work [51], we investigated the effect of the ion tip aperture D eq and p on the diameter d of a pillar printed at fixed potential (E = −0.5 V vs. Ag-AgCl) and geometry (i.e., distance between two consecutive voxels ∆z, here referred to as voxel height) as in Figure 4a. The framework consists in measuring the pillar diameter d from SEM images (Figure 4a, all the d values are compiled in Figure 4b) and the vertical speed . z from the printer logged z-axis position. These two experimental values can then be combined to calculate the volumetric deposition speed . V: as function of opening size d and p (Figure 4c). In this work, the same protocol is used to investigate the effect of the deposition potential E. Arrays of pillars were printed at pressures varying from 10 mbar to 210 mbar with a pressure step of 5 mbar. Each array was printed at a different deposition potential, the potential values were changed from −0.5 V to −0.42 V with a 20 mV difference between each array; the deposition potential was set against an Ag/AgCl pellet as RE and all the given potential values are referred to in this reference. Figure 5 shows a set of selected SEM images showing pillars printed at selected pressure values. In agreement with [51], increasing p while keeping E constant results in an increase of the diameter d. A larger flow of ions requires a larger area for the ions contained in it to be completely consumed by the reduction reaction so that the pillars grow with a wider cross-section. By contrast, an increase in E (i.e., more negative deposition potential) has an opposite effect: pillars printed at the same pressure but at increasing E have a smaller cross-section. In this work, the same protocol is used to investigate the effect of the deposition potential E. Arrays of pillars were printed at pressures varying from 10 mbar to 210 mbar with a pressure step of 5 mbar. Each array was printed at a different deposition potential, the potential values were changed from −0.5 V to −0.42 V with a 20 mV difference between each array; the deposition potential was set against an Ag/AgCl pellet as RE and all the given potential values are referred to in this reference. Figure 5 shows a set of selected SEM images showing pillars printed at selected pressure values. In agreement with [51], increasing p while keeping E constant results in an increase of the diameter d. A larger flow of ions requires a larger area for the ions contained in it to be completely consumed by the reduction reaction so that the pillars grow with a wider cross-section. By contrast, an increase in E (i.e., more negative deposition potential) has an opposite effect: pillars printed at the same pressure but at increasing E have a smaller cross-section. This result highlights that electrochemical printing is fully controlled by the electrodeposition potential and pressure, while it is monitored by measuring the retraction speed of the nozzle, which is automated with the force feedback. In principle, the arrangement of the ec cell resembles the socalled wall-jet electrode system, a tool in hydrodynamic voltammetry that utilizes a jet of solution of electroactive species, which is pushed from a circular nozzle to hit the working (collector) electrode perpendicularly [55]. The diffusion limited current Id is well-described in the wall-jet arrangement and is approximated by analytical expressions that relate it to the geometrical parameters of the system (nozzle, a, and working electrode radius, R) and the parameters of the fluid flow: where k, n, F, c0, D, ν and V specify geometrical proportionality factor, number of electrons in ec reaction, Faraday constant, concentration of electroactive species, their diffusion coefficient, dynamic viscosity of the medium and volumetric flow rate, respectively. This expression can be simplified by arranging all the constants into a proportionality factor α (different for each voltage value): The deposition current could not be measured directly because the active Cu 2+ reducing area is in the order of tens of µm 2 , while the total working electrode area is cm 2 , resulting in the deposition current being completely masked by the large background current on the working electrode. However, using the vertical growth speed (Figure 6a) along with experimentally measured pillar diameters (Figure 6b), it becomes possible to derive further insights of the printing process. For example, the vertical copper growth rate (using the given Cu molar mass and density, MCu and ρCu), can be estimated from the eq. 3 by a simple transformation: This result highlights that electrochemical printing is fully controlled by the electrodeposition potential and pressure, while it is monitored by measuring the retraction speed of the nozzle, which is automated with the force feedback. In principle, the arrangement of the ec cell resembles the so-called wall-jet electrode system, a tool in hydrodynamic voltammetry that utilizes a jet of solution of electroactive species, which is pushed from a circular nozzle to hit the working (collector) electrode perpendicularly [55]. The diffusion limited current I d is well-described in the wall-jet arrangement and is approximated by analytical expressions that relate it to the geometrical parameters of the system (nozzle, a, and working electrode radius, R) and the parameters of the fluid flow: where k, n, F, c 0 , D, ν and V specify geometrical proportionality factor, number of electrons in ec reaction, Faraday constant, concentration of electroactive species, their diffusion coefficient, dynamic viscosity of the medium and volumetric flow rate, respectively. This expression can be simplified by arranging all the constants into a proportionality factor α (different for each voltage value): The deposition current could not be measured directly because the active Cu 2+ reducing area is in the order of tens of µm 2 , while the total working electrode area is cm 2 , resulting in the deposition current being completely masked by the large background current on the working electrode. However, using the vertical growth speed (Figure 6a) along with experimentally measured pillar diameters Micromachines 2020, 11, 6 8 of 14 (Figure 6b), it becomes possible to derive further insights of the printing process. For example, the vertical copper growth rate (using the given Cu molar mass and density, M Cu and ρ Cu ), can be estimated from the Equation (3) by a simple transformation: metal ions upon electroplating, is in fact a function of the applied pressure R = R(p) (Figure 6b). For better fitting of the vertical growth data using the experimental values of pillar diameters Rexperiment(p), we introduced an empirical scaling coefficient that allows estimation of the radius of the area where ions are collected Rfit(p): The result of this estimation suggests that the area where the current is collected upon plating at higher pressures (p > 60 mbar) is somehow larger than the printed metal pillar. Volumetric deposition rate also deviates from the theoretical prediction, highlighting that at higher pressures some portion of the metal ions is electroplated around and not directly on the pillar. These results are in good agreement with electron microscopy images ( Figure 5) that depict circular «dunes» around the pillars, with sizes increasing at larger pressure magnitudes.  This relationship is complicated by the fact that the radius of the growing pillar R, which collects metal ions upon electroplating, is in fact a function of the applied pressure R = R(p) (Figure 6b). For better fitting of the vertical growth data using the experimental values of pillar diameters R experiment (p), we introduced an empirical scaling coefficient that allows estimation of the radius of the area where ions are collected R fit (p): R fit (p) = e p/1000 R experiment (p) The result of this estimation suggests that the area where the current is collected upon plating at higher pressures (p > 60 mbar) is somehow larger than the printed metal pillar. Volumetric deposition rate . V also deviates from the theoretical prediction, highlighting that at higher pressures some portion Micromachines 2020, 11, 6 9 of 14 of the metal ions is electroplated around and not directly on the pillar. These results are in good agreement with electron microscopy images ( Figure 5) that depict circular «dunes» around the pillars, with sizes increasing at larger pressure magnitudes.

Arbitrary-Shaped 3D Structures in a Layer-by-Layer Fashion: Michelangelo's David
The confined electrodeposition is not limited at strand printing of pillars and coils. The same system can also be using to produce complex 3D object by taking advantage of the merging of voxels placed side by side. To demonstrate the capabilities of this process, a 3D model of Michelangelo's David (author of the original model 'gabrielmda', distributed under CC-0 license on blendswap.com) was modified and sliced using the open-source 3D modeling software Blender 2.8 (Blender Foundation, Amsterdam, the Netherlands). The voxels defining the structure were routed in a layer-by-layer fashion using a custom clustering algorithm. The same model was then reproduced at a 10,000:1 and a 70,000:1 scale.
The two replicas of Michelangelo's David [56], at the two completely different scales shown in Figure 7, represent the current state-of-the-art in confined electrodeposition 3D printing. The 700 µm-tall 1:10,000 replica is shown in Figure 7a-c; the structure is defined by 130,000 voxels and the total print time required is 16 h using a pressure of 50 mbar and a deposition potential of −0.46 V. The 100-µm-tall, 1:70,000 version of the same model is shown in Figure 7a,b inset (to allow a direct comparison between the two objects) and 7d; this structure is defined by 25,000 voxels and the printing time required is 2 h using a pressure of 15 mbar and a potential of −0.5 V.

Arbitrary-Shaped 3D Structures in a Layer-by-Layer Fashion: Michelangelo's David
The confined electrodeposition is not limited at strand printing of pillars and coils. The same system can also be using to produce complex 3D object by taking advantage of the merging of voxels placed side by side. To demonstrate the capabilities of this process, a 3D model of Michelangelo's David (author of the original model 'gabrielmda', distributed under CC-0 license on blendswap.com) was modified and sliced using the open-source 3D modeling software Blender 2.8 (Blender Foundation, Amsterdam, the Netherlands). The voxels defining the structure were routed in a layerby-layer fashion using a custom clustering algorithm. The same model was then reproduced at a 10,000:1 and a 70,000:1 scale.
The two replicas of Michelangelo's David [56], at the two completely different scales shown in Figure 7, represent the current state-of-the-art in confined electrodeposition 3D printing. The 700 µmtall 1:10000 replica is shown in Figure 7a-c; the structure is defined by 130000 voxels and the total print time required is 16 h using a pressure of 50 mbar and a deposition potential of -0.46 V. The 100µm-tall, 1:70000 version of the same model is shown in Figure 7a,b inset (to allow a direct comparison between the two objects) and 7d; this structure is defined by 25000 voxels and the printing time required is 2 h using a pressure of 15 mbar and a potential of −0.5 V. The two different pressure and potential values were utilized to optimize the print time and resolution required: the larger pressure and lower deposition potential used on the 700 µm structure allowed for the definition of a geometry with larger voxels (~4 µm diameter, as from Figure 6b), thereby reducing the total voxel number and print time required, while keeping the resolution to an acceptable value for the scale of the object. By contrast, for the smaller 100 µm replica a higher resolution was required to target smaller voxels (~1.6 µm diameter, as from Figure 6b), therefore lower pressure and higher overpotential were employed. The two different pressure and potential values were utilized to optimize the print time and resolution required: the larger pressure and lower deposition potential used on the 700 µm structure allowed for the definition of a geometry with larger voxels (~4 µm diameter, as from Figure 6b), thereby reducing the total voxel number and print time required, while keeping the resolution to an acceptable value for the scale of the object. By contrast, for the smaller 100 µm replica a higher resolution was required to target smaller voxels (~1.6 µm diameter, as from Figure 6b), therefore lower pressure and higher overpotential were employed.

Intertwined Coils with Different Section Sizes
Herein we also demonstrate how it is possible to tune the voxel area by adjusting p on-the-fly to fabricate a structure composed of parts with different sections. For this purpose, we designed a structure of four intertwined coils, each wire having its own diameter. If fabricated with a single nozzle, such an entangled object can be produced only in a layer-by-layer fashion, on the condition that each of the four voxels of each layer is plated at a different value of pressure. A probe with a 500 nm aperture was operated at 20, 40, 60, and 80 mbar in succession (Figure 4b) obtaining the structure presented in Figure 8, whereby wires are presented, artificially colored for better visualization.

Intertwined Coils with Different Section Sizes
Herein we also demonstrate how it is possible to tune the voxel area by adjusting p on-the-fly to fabricate a structure composed of parts with different sections. For this purpose, we designed a structure of four intertwined coils, each wire having its own diameter. If fabricated with a single nozzle, such an entangled object can be produced only in a layer-by-layer fashion, on the condition that each of the four voxels of each layer is plated at a different value of pressure. A probe with a 500 nm aperture was operated at 20, 40, 60, and 80 mbar in succession (Figure 4b) obtaining the structure presented in Figure 8, whereby wires are presented, artificially colored for better visualization. The main difficulty of growing multiple threads in parallel each at different pressure values is the time response required to establish a stable pressure; specifically, whether or not the pressure and thus the cloud of Cu 2+ ions can change quickly enough when the nozzle is moved from one strand to the next (which takes just a few milliseconds). As evincible from Figure 8, each printed wire has indeed a different diameter, which is homogeneous along all its length (5, 6, 7 and 8 µm, respectively). From Figure 4b it would be expected that the diameters at those particular pressure values are 2, 3, 3.7 and, 4.4 µm. Experimentally, the diameters of the strands of the intertwined helices are consistently larger than those measured for the individual pillar. This bigger voxel size is probably due to the "additional" plating taking place during the lateral motions from a strand to the next. Such lateral motions are absent in the printing of pillars which occurs only in the z direction.

Conclusions and Outlook
In conclusion, we have introduced a new one-step method to do additive micromanufacturing of metallic structures using microchanneled AFM probes. The key novelty of this approach is that the voxel growth can be detected in situ by monitoring the real-time deflection of the hollow cantilever. This is essential to avoid clogging of the probe aperture and to automate the process. Furthermore, it removes the need to calibrate parameters such as growth speed. Finally, the fact that the manufacturing takes place in a supporting electrolyte allows for the reliable voxel-wise printing on arbitrary positions without the restrictions of methods relying on meniscus formation. Combining these advantages, an ec 3D metal printing is achieved in a true voxel-by-voxel layer-by-layer fashion, The main difficulty of growing multiple threads in parallel each at different pressure values is the time response required to establish a stable pressure; specifically, whether or not the pressure and thus the cloud of Cu 2+ ions can change quickly enough when the nozzle is moved from one strand to the next (which takes just a few milliseconds). As evincible from Figure 8, each printed wire has indeed a different diameter, which is homogeneous along all its length (5, 6, 7 and 8 µm, respectively). From Figure 4b it would be expected that the diameters at those particular pressure values are 2, 3, 3.7 and, 4.4 µm. Experimentally, the diameters of the strands of the intertwined helices are consistently larger than those measured for the individual pillar. This bigger voxel size is probably due to the "additional" plating taking place during the lateral motions from a strand to the next. Such lateral motions are absent in the printing of pillars which occurs only in the z direction.

Conclusions and Outlook
In conclusion, we have introduced a new one-step method to do additive micromanufacturing of metallic structures using microchanneled AFM probes. The key novelty of this approach is that the voxel growth can be detected in situ by monitoring the real-time deflection of the hollow cantilever. This is essential to avoid clogging of the probe aperture and to automate the process. Furthermore, it removes the need to calibrate parameters such as growth speed. Finally, the fact that the manufacturing takes place in a supporting electrolyte allows for the reliable voxel-wise printing on arbitrary positions without the restrictions of methods relying on meniscus formation. Combining these advantages, an ec 3D metal printing is achieved in a true voxel-by-voxel layer-by-layer fashion, enabling the fabrication of arbitrarily shaped geometries. This is a breakthrough in metal 3D printing on the micrometer scale, enabling a new range of printable shapes and structures.
In a second step, the force-controlled ec µAM was completely redesigned to optimize the ec printing of copper by reducing the response and positioning times and increasing the motor range. We gained further understanding of the deposition process by investigating the pillar growth systematically varying the ec deposition potential, the applied pressure and the probe aperture at a fixed the voxel height (∆z = 500 nm). The influence of the applied pressure and the probe aperture on the lateral diameter of the electroplated copper voxels was also investigated and rationalized. In this way, the voxel area could be spanned over two orders of magnitude using the same probe aperture. This multi-scale capability was exploited to fabricate copper microstructures of two different replicas of Michelangelo's David. Lastly, a 50 µm × 50 × µm 170 µm object consisting of four helical wires was obtained by changing the pressure on a voxel-by-voxel basis so that each printed wire had a different diameter, yet homogeneous along its entire length. A significant throughput increase is now imaginable with the possibility of tuning the voxel size depending on the feature dimensions within the same design. Copper is the most compliant metal for electrodeposition, nonetheless this protocol can be applied in a one-to-one way to other metals.