3D Printing by Two‐Photon Polymerization of Polyimide Objects and Demonstration of a 3D‐printed Micro‐Hotplate

Polyimides are polymeric materials with outstanding thermal, chemical, and mechanical properties. For this reason, they find applications in several engineering sectors, including aerospace, microsystems, and biomedical applications. For realizing 3D structures made of polyimides, 3D printing is an attractive technique because it overcomes the limitations of polyimide processing using conventional manufacturing techniques such as molding and subtractive manufacturing. However, current polyimide 3D printing approaches are limited to realizing objects with the smallest dimensions of the order of a few hundred micrometers. 3D printing of polyimide objects featuring sub‐micrometer resolution using two‐photon polymerization by direct laser writing is demonstrated here. A negative photosensitive polyimide is applied that is widely used in microsystems applications. To demonstrate the utility of this polyimide 3D printing approach and the compatibility of the 3D objects with operation at elevated temperatures, a micro‐hotplate is 3D printed and characterized at operating temperatures of above 300 °C.


Introduction
Polyimides are high-performance polymeric materials used in several engineering fields such as the automotive, aerospace, and oil and gas sectors. [1]Within the field of microsystems, polyimides have been widely used in various applications, including microelectronics packaging, [2] flexible electronics, [3] and biomedical applications. [4]The widespread use of polyimides is owed to their outstanding thermal, mechanical, and chemical properties [1] and their biocompatibility. [4]Kapton is the most broadly used polyimide, thanks to the very broad range of temperatures it can withstand, from −270 to 400 °C. [5]The fabrication DOI: 10.1002/admt.202300229 of 3D shaped objects made of polyimide using conventional manufacturing methods, such as injection molding, is impeded by the insolubility and high thermal stability of polyimides.To fabricate polyimide objects, soluble polyimide precursors are used, generally poly-amic acids, and cured at high temperatures to activate the imidization reaction of the precursors, usually above 300 °C for complete transformation into a polyimide.The imidization reaction generates water molecules inside the material that are highly volatile at the curing temperature and escape the polyimide, causing a considerable loss of both mass and volume of the synthetized polyimide part.For this reason, the fabrication of polyimide objects using conventional manufacturing methods is limited to simple shapes that can easily tolerate homogeneous shrinking, such as sheets, cylinders, and hollow tubes.Recently, additive manufacturing of 3D object made of polyimide and related high-performance materials have attracted increasing attention.Both researchers and industry have developed materials and technologies for polyimide printing by stereolithography, [6,7] inkjet printing, [8] direct ink writing, [9] and fused deposition modeling. [10]However, 3D printing of polyimide objects at the micro-scale has, up to now, received limited attention.
In microsystem applications, polyimides are being used in the form of sheets, as substrates, and as thin films.For example, thin polyimide films are employed as stress-distribution layers, [11] and dielectric interlayers [12] in electronics and microsystems.The thin polyimide films are typically realized by spin-coating a soluble polyimide precursor on a silicon or glass substrate.The polyimide precursor coating is then cured at temperatures between 200 and 400 °C, [13] typically above 300 °C, to induce the imidization of the polyimide precursor.Prior to the curing step, the thin polyimide coating can be patterned using a mask in combination with oxygen plasma etching, [14] or using photo-sensitive polyimide formulations enabling patterning in a single lithographic step. [15]Both positive and negative photosensitive polyimide precursors are available.However, the patterns created by conventional UV lithography techniques are inherently 2D and not suitable for realizing free-form 3D objects.
To 3D print polymer objects at the micro and nanoscale, two-photon polymerization of negative photoresist materials has been used extensively.In this approach, a focused femtosecond laser beam is used for direct writing inside the photoresist, thereby inducing localized cross-linking of the photoresist by two-photon polymerization.Micro 3D printing by two-photon polymerization can achieve sub-micrometer resolution [16] and has been widely used for realizing complex geometries [17] and a variety of devices, including photonic crystals, [18] biomedical scaffolds, [19] and MEMS devices. [20]However, the photosensitive polymers used in these approaches have limited temperature stability and micro 3D printing of polyimide structures by two-photon polymerization has to the best of our knowledge not been reported in the literature.Micro 3D printing by twophoton polymerization of temperature-stable inorganic materials has been achieved previously using sintering of 3D-printed composite materials [21] and by direct 3D printing of silica glass. [22]imilarly, temperature-stable conductive structures have been realized by pyrolysis of 3D-printed polymers. [23]However, the required sintering temperatures often exceed 900 °C, and the resulting materials are brittle and do not feature the characteristics of polymers.Here, we report micro 3D printing of micro-scale polyimide objects by two photon-polymerization using a commercially available photosensitive negative polyimide precursor, HD-4100 (HD Microsystems GmbH, Germany).We present 3Dprinted structures with micrometer and sub-micrometer resolution and demonstrate the utility of our approach by fabricating a micro-hotplate and characterizing it at high temperatures of above 300 °C.The wide usage of the HD-4100 polyimide in the microsystem industry and its compatibility with semiconductor processing make it an excellent micro 3D printing material for integration with conventional microsystem applications.This polyimide offers a high glass transition temperature of 340 °C, a high degradation temperature of 430 °C, good substrate adhesion, and elongation at break of up to 45%. [24]

Polyimide 3D Printing by Two-Photon Polymerization
To demonstrate the possibility of 3D printing of polyimide objects by two-photon polymerization, we applied the process sequence outlined in Figure 1.In our experiments, we used the commercially available polyimide precursor HD-4100 (Figure 1a) in combination with the 3D printer Nanoscribe Photonic Professional GT2 (Nanoscribe GmbH, Germany) in the "oil-immersion" configuration (Figure 1b). [25]The details of the 3D printing process sequence are provided in the Experimental Section.In our experiments, we first investigated and optimized the material pro-cessing conditions (i.e., the soft-baking, development, and curing of the polyimide precursor) and the laser processing window (i.e., the laser power and scanning speed) suitable for fabricating 3D-printed objects with micrometer-scale features in the polyimide precursor.Hence, we used a 63× objective in the Nanoscribe tool (Experimental Section).When using this objective, the basic voxels have the shape of an elongated ellipsoid, which is typical for two-photon polymerization processes, [26] with sizes >1 μm x 1 μm x 1 μm according to tool specifications.Therefore, we selected slicing and hatching distances of 0.1 μm during the processing of the CAD models in these experiments (Experimental Section).We performed 3D prints with different laser powers and laser scan speeds.This experimental work was required to find the optimal printing parameters because we cannot predict the correct exposure dose for two-photon polymerization theoretically since the exact polymer formulation of the commercial product used here is not publicly available.We found that objects printed with a laser scan speed of 12 mm s −1 and a laser power of 4 mW show good structural integrity and dimensional control of the printed features (Figure 2a).This can be compared to objects printed with laser powers of 2 mW, showing a lack of structural stability due to the low level of cross-linking of the polyimide precursor because of the low laser dosage (Figure 2b).In contrast, objects printed with laser powers of 6 mW result in features with dimensions that are wider than designed, which is especially prominent for features that are printed next to each other (Figure 2c), for instance between the top and the bottom horizontal beams.The structural integrity of the printed structures is directly related to the laser dosage that depends on both the laser power and the laser scanning speed.From these results we can compute the required laser power at different laser scan speeds, considering that the laser dosage deposited into the polyimide precursor varies inversely with the laser scan speed, and quadratically with the laser power. [27,28]o synthesize the polyimide from the 3D-printed polyimide precursor, the precursor needs to be cured at a temperature of 375 °C (Experimental Section).The curing causes thermal imidization of the polyimide precursor, while water is released as a by-product of the chemical reaction, resulting in material shrinkage during polyimide synthesis.We measured the linear shrinkage of suspended cantilevers made of the polyimide precursor that are 20 μm long and with a cross-section of 6 μm x 6 μm to be 19 ± 1% the after the curing process.An optical microscope was used to measure the cantilever length before and after the curing process.Such shrinkage is unavoidable in polyimide synthesis, although many of the commercially available polyimide precursors are optimized for low shrinkage.Hence, shrinkage must be considered and compensated for in the design of the 3D geometry to be printed.In addition to the expected size reduction of structural features, shrinkage can also cause the breaking of 3D printed objects that are anchored to the substrate at multiple points, since the substrate does not shrink at the same rate as the polyimide.Common strategies to avoid breaking of 3D printed objects due to material shrinkage are the introduction of stress relieving structures, such as a thick base plate that shrinks together with the 3D object and alleviates the stresses experienced by the 3D object, or meander structures that can stretch out when the 3D printed material is shrinking.When curing our 3D-printed objects to form the polyimide, we found that objects printed on top of a 10 μm-thick polyimide precursor base plate, which was 3D printed in the same step as the object itself (Experimental Section), preserved their structural integrity and did not significantly suffer from deformations due to the material shrinkage during curing (Figure 2d).This is because the surface of the base plate to that the object is attached shrinks by approximately the same amount as the object itself.3D-printed objects that were directly printed on the glass surface without a 3D-printed base, deformed more readily during the curing process due to the material shrinkage and related stresses created in the polyimide.The resulting distortions of the shape of the 3D-printed objects were particularly pronounced for objects featuring several distributed attachment points to the glass substrate as the attachment points restrain the freedom of movement of the object during curing.
Furthermore, we determined the size of the smallest polyimide structures that we can 3D print using our approach.We printed test structures composed of suspended nanowires that are anchored on both sides to 3D-printed supporting blocks (Figure 3).For each pair of supporting blocks, we printed, in sequence, five identical 10 μm-long nanowires using a single laser scan for each nanowire.We performed several experiments using different laser powers and laser scan speeds.Specifically, we swept the laser power between 3.5 and 22.5 mW, in steps of 5 mW.The selected laser scan speeds were 10, 30, and 50 mm s −1 , respectively.We printed two identical arrays composed of 3 × 5 test structures, one for each parameter combination.Both arrays of test structures were printed on the same substrate to ensure that any difference in nanowire width resulted from the curing process only, and not from batch-to-batch variations of the fabrication process.After splitting the substrate, only one of the two arrays of nanowire test structures was cured to transform the polyimide precursor into polyimide.We show the average measured widths of the five printed nanowires for each combination of laser power and scan speed before and after curing in Figure 3a.After printing (and before curing), we measured the smallest nanowire to be 330 nm wide when using a laser power of 7.5 mW and a laser scan speed of 10 mm s −1 (Figure 3b).Smaller wires were still resolved but were either broken or collapsed (identified in Figure 3a by the white markers).When using a laser power of 3.5 mW, the structures were not printed or resolved.The width of the smallest nanowires was measured to be smaller than the diffraction limit for the used wavelength of the light, as is typical for features printed by two-photon polymerization. [16]ll nanowires were printed using the same optical system and microscope objective and thus, the different resulting nanowire widths are not due to differences in the laser spot size.The resulting nanowire width depends on the width at the focal plane where the laser fluence is sufficiently high to trigger two-photon polymerization, i.e., increasing the laser power typically results in increased width of the nanowires.After curing, the widths of the nanowires were consistently smaller than before curing, with a reduction in width of up to 42%.The smallest cured polyimide nanowires were 190 nm wide when using a laser power of 7.5 mW and a laser scan speed of 10 mm s −1 (Figure 3d,e).From Figure 3e-g it can be seen that the printed nanowires have an elliptical cross-section with the long side extended in the vertical direction.This is due to the elongated shape of the laser voxel that is typical for two-photon polymerization processes. [26,29]Similar to the nanowires, we also printed double-sided clamped beams that are 1 and 3 μm-wide, respectively.After curing, the polyimide beams were 0.87 and 2.4 μm wide, respectively (Figure 3h,i).The width reduction of the nanowires and beams is affected by several interacting factors such as the water removal rate, which is proportional to their surface-to-volume ratio, and the strain exerted on the wires by the shrinking supporting blocks.Our optimized polyimide 3D printing process entails process parameters and features that diverge from the parameters employed in conventional processing of the HD-4100 polyimide precursor, which , after laser exposure and before curing at 375 °C) printed with a laser scan speed of 10 mm s −1 and a laser power of 7.5 mW.c) Nanowires of polyimide precursor (i.e., after laser exposure and before curing at 375 °C) printed with a laser scan speed of 10 mm s −1 and a laser power of 12.5 mW.d,e) Polyimide nanowires (after curing at 375 °C) printed with a laser scan speed of 10 mm s −1 and a laser power of 7.5 mW.f,g) Polyimide nanowires (after curing at 375 °C) printed with a laser scan speed of 10 mm s −1 and a laser power of 12.5 mW.h,i).Double-sided clamped polyimide beams (after curing at 375 °C) that are 1 and 3 μm wide, respectively.
we discuss in the following.The soft bake time of 22 min used in our proposed process is somewhat longer than what is typically used and recommended in the HD-4100 material process guide. [24]The recommended soft bake time is optimized to work for thin polyimide precursor coatings with thicknesses of below 10 μm, where volatile compounds can quickly diffuse through the coating and escape from its outer surface.In our case, however, the solvent must diffuse through a polyimide precursor droplet with a typical thickness of a few hundred micrometers at its thickest point.Thus, a longer soft bake time is needed to ensure sufficient removal of the solvent from the central volume of the deposited droplet, where the printing typically is performed.In situations where the solvent was not sufficiently removed, we observed two different failure modes in the 3D printing process, depending on the remaining solvent concentration.At high solvent concentrations and high laser power, the laser heats up the solvent and causes the formation and bursting of gas bubbles inside the polyimide precursor droplet.At lower solvent concentrations and lower laser power, when 3D printing long and loosely supported structures, these structures can shift position because of the semi-liquid properties of the polyimide precursor inside the droplet.Similar to the soft bake time, we use a development time of 12 h for the polyimide precursor after the laser exposure, which is significantly longer than the development time that is recommended for the patterning of HD-4100 polyimide precursor coatings. [24]The longer development time is owed to the larger volume of resist that must be dissolved by the developer in our process.Large structures made of fully cross-linked polyimide precursor can tolerate even longer development times than the 12 h we used, without substantial damage or size modifications of the pattern.However, if the polyimide precursor is only partially cross-linked, for example when realizing high-resolution features such as the nanowires shown in Figure 3, these structures are more sensitive to the action of the developer.Thus, if the initial droplet thickness of the polyimide precursor is different in different process batches, the 3D-printed objects may be exposed to the developer for different time intervals, which can then yield inconsistent results from the development of the 3Dprinted objects.For this reason, in our experiments on 3D printing of high-resolution structures, we printed all nanowire arrays (Figure 3) in the same polyimide precursor droplet, and at the same time.
Moreover, in our experiments exploring optimized 3D printing process parameters, we found that for printing of objects with small features, the process parameter window is relatively narrow (Figure 2).At too low laser exposure dosage, the printed polyimide precursor is too weakly cross-linked, resulting in mechanically unstable structures.At too high exposure doses, crosslinking of the polyimide precursor also occurs outside the volume scanned by the voxel defined by the laser focus, particularly in spaces in between adjacent structures.This is known as the proximity effect, [30] which is due to inefficient quenching of the polymerization reaction from oxygen or other polymerization inhibitors.This effect limits the minimum distance at which adjacent structures can be printed.It also means that the resulting size of a printed structure is influenced by the presence of other structures within the vicinity of a few microns or even a few tens of microns.However, the proximity effect is less significant when printing large, isolated structures that are not located nearby other structures.For such structures, the laser dosage can be increased to values above the one used for printing the object shown in Figure 2c.In summary, the proximity effect depends on the laser dosage used to 3D print the structures, and thus, laser power, laser scan speed, and the slicing and hatching distances must be tailored to the specific structures, to ensure correct printing.
Finally, we found that the curing step to synthesize the polyimide from the 3D-printed polyimide precursor induces volume changes and stresses in the 3D-printed structures.For 3D-printed objects featuring multiple connection points to the substrate, a 3D-printed base plate on top of which the object is printed is suitable to avoid unwanted deformation of the object as the uniform shrinking of the base during curing minimizes the stresses generated in the object (Figure 2d).In contrast, the supporting blocks on which the nanowires are suspended (Figure 3) are not printed on a base plate, but directly on the glass substrate.In this case, during curing, the distance between the blocks increases and causes stresses in the nanowires as the nanowires elongate.As a result of the nanowire elongation, their widths decrease.On the other hand, the thicker double-sided clamped beams exert a large pulling force on the supporting blocks, which can deform the blocks (Figure 3i).We also observed that the curing step affected the color of the polyimide in the 3D-printed objects, which ranges from dark orange as typically expected for polyimides, to almost black for structures that are larger than 50 μm.A dark polyimide color is often attributed to residues from the pyrolysis of photoinitiator cross-links that occur during curing. [9]Because some of our 3D-printed structures are thicker than the conventional 2-10 μm thick polyimide coatings, we hypothesize that the volatile pyrolyzed particles are impeded from escaping the thicker 3D-printed structures.

3D-Printed Micro-Hotplate Demonstrator
To demonstrate the utility of our polyimide 3D printing approach and the suitability for operating the 3D-printed polyimide objects at elevated temperatures, we fabricated and characterized 3D-printed proof-of-concept micro-hotplates.Micro-hotplates are used in several applications, including gas sensing, [31] biochemistry, [32] and wearable devices. [33]To realize a functional 3D-printed micro-hotplate, we used an approach that combines micro 3D printing of shadow-masking structures with undercut cross-sections and subsequent directional metal deposition to form electrically addressable transducers, [20] i.e., the electrical heaters of the 3D-printed micro-hotplate (Figure 4a-c).We designed the micro-hotplate to have two contact electrodes connected to a wire featuring the typical meander shape used for metal lines on flexible substrates. [34]On flexible and stretchable substrates, the meander shapes are used as a stress-relief mechanism that increases the maximum stress tolerated by the metal lines.Here, we used the meander shape to mitigate the impact of stresses induced by the curing of the polyimide precursor, and limit deformation of the 3D-printed meander shape.The meander-shaped heating wire of the micro-hotplate is suspended on eight supporting pillars that are 60 μm high, to prevent collapse of the soft 3D-printed polyimide precursor structure onto the substrate surface.The position and size of the pillars, as well as the amplitude of the bending of the meanders, were optimized to minimize the bending and the deflection of the suspended wire caused by shrinkage during the curing process.Suspending of the micro-hotplate heater significantly reduces the thermal losses by heat conduction to the substrate, which is desirable for many applications such as thermal anemometers, [35] because this reduces energy consumption and increases performance.We 3D printed the micro-hotplate on top of a 30 μm thick base plate that was printed in the same step.For the printing, we used a printing recipe that allowed faster printing as compared to the recipes used in the previous experiments, since the micro-hotplate structure is significantly larger than the previously printed parts.We used a 25× objective and the Shell-&-Scaffold printing mode, with a slicing distance of 0.5 μm, a hatching distance of 0.4 μm, a laser power of 10 mW, and a laser scan speed of 100 mm s −1 .After the development and curing of the micro-hotplate structure, we deposited a 20 nm thick Titanium adhesion layer and a 200 nm thick platinum layer, using a directional (perpendicular to the substrate surface) e-beam evaporation process.The self-shadow masking mechanism of the hotplate design ensured the electrical isolation of the contact electrodes and the heating wire from the substrate (Figure 4c,f). [20]We used platinum as a heater material because of its advantages for high-temperature applications, such as high degradation temperature in air [30] and the linear change of its resistance over temperature. [36]e used a Source Measure Unit 6240B (ADCMT, Japan) to drive the micro-hotplate by forcing a current through the thin film platinum wire (resistive heater).Prior to operation, we measured the resistance of the platinum wires to be 150 and 130 Ω, respectively for the two printed micro-hotplates, using a fourpoint probe configuration.In all our experiments, the temperature of the micro-hotplates was evaluated by taking thermal infrared images of the micro-hotplates using an infrared camera (Experimental Section) and computing the average temperature across the measurement area defined in Figure 4g.We characterized the response time of one hotplate by applying 250 ms long current pulses while taking infrared images of the hotplate every 20 ms.The average temperature of the hotplate across the measurement area was computed and plotted for the different applied currents in Figure 4d.From this plot, we extracted the ) Optical microscope images of the 3D-printed hotplate before and after the curing step, respectively.c,f) SEM image of the hotplate after operation at a temperature of 220 °C.d) Plot of the measured operating temperature of the hotplate as a function of time, using different applied current pulses.e) Average operating temperature of the hotplate measured in the device area of interest (marked in (g)) as a function of applied power.g-i) Thermal infrared images of the micro-hotplate at different applied heating currents.j) SEM image of a damaged hotplate after operation at an excessive heating current of 30 mA, which we estimate to result in a heater temperature of ≈430 °C.
response time of the micro-hotplate to be ≈80 ms.Thus, in the subsequent experiments we used 100 ms long current pulses.For the thermal characterization of the micro-hotplates, we applied a series of current pulses starting from 2 mA and increasing the current by 2 mA for each pulse, while taking infrared images of the hotplate every 20 ms.We plotted the average temperature across the measurement area of the resistor against the power consumed by the resistor (Figure 4e).The consumed power was computed by multiplying the applied current with the resistance at that temperature.The resistance variation was measured up to a driving current of 14 mA.Combining the resistance measurements with the thermal measurements, we could extract a linear thermal coefficient of resistance of 0.002 ± 0.0001 K −1 for the two micro-hotplates, which is somewhat lower than the typical thermal coefficient of resistance of bulk platinum (0.0039 K −1 ).The lower thermal coefficient of resistance is expected for thin platinum films, [37] and is in the same range as previously reported for thin platinum films deposited on polyimide. [38]For the first micro-hotplate (blue graph in Figure 4e), we applied a maximum current of 20 mA, which corresponded to a maximum measured temperature of 220 ± 20 °C.For the second hotplate (red graph in Figure 4e), we applied increasing current pulses until the resistor failed, which occurred at an applied current of 30 mA (Figure 4j).At heating currents of above 22 mA, the high-temperature areas of the infrared images saturated at a temperature of 350 °C, thus the infrared temperature measurements at heating currents of above 22 mA were not reliable.However, we estimated the maximum temperature reached by the micro-hotplate to be ≈430 °C, assuming a linear increase of the resistance with temperature and fitting the temperature to the applied power.

Conclusion
In this work, we explored an approach for 3D printing of micro-objects made of polyimide.This approach involves direct laser writing by two-photon polymerization to 3D print inside a negative photosensitive polyimide precursor.Then, the 3Dprinted objects are cured at a temperature of 375 °C to turn the cross-linked precursor into polyimide.We successfully demonstrated the feasibility of this approach and characterized the process windows for printing of 3D objects with resolved micrometer-scale features, and of nanowires that are below 200 nm wide.We presented printing strategies to minimize proximity effects in laser writing and to mitigate the deformation of the 3D-printed objects during curing as a result of material shrinkage.Finally, we 3D printed a polyimide micro-hotplate and coated the heater with a thin platinum layer.We successfully operated the micro-hotplate using heating currents, and measured heater temperatures of above 300 °C using a thermal infrared camera.Thus, our approach for 3D printing of polyimide objects with sub-micrometer resolution enables the realization of hightemperature compatible microsystems with novel and more efficient geometries, with promising applications in research and industry, including in robotics, chemistry, and medicine, all of which hold great promise for future research.

Experimental Section
Materials: The photosensitive polyimide precursor HD-4100 (HD Microsystems GmbH, Germany) was stored in an opaque glass bottle in a freezer at −20 °C.Before all printing operations, the bottle was removed from the freezer and left to rest at room temperature for a minimum of 8 h.This waiting time was necessary to avoid water condensation inside the bottle during the opening, which would irreversibly damage the material.Based on the information provided in the Material Safety Data Sheet (MSDS), [39] the polyimide precursor contained the two solvents N-Methyl-2-pyrrolidone (NMP) and methanol that were removed during the soft-bake, O-(Ethoxycarbonyl)-N-(1-methyl-2-oxo-2phenylethylidene)hydroxylamine that was part of the polyimide backbone, and 2-Hydroxyethyl methacrylate, which was a reactive monomer that participates in the cross-linking mechanism but was not photo-active per se.According to the HD-4100 material specifications, the polyimide precursor was not photosensitive to the infrared wavelength (780 nm) of the laser used in the Nanoscribe tool.The polyimide precursor was drawn from the bottle with a 5 mL plastic pipette and a single droplet was deposited on a 170 μm-thick borosilicate glass substrate (Thermo Fisher Scientific, US).Prior to use, the glass substrate was cleaned with acetone and isopropanol, and no adhesion promoter was applied.After the application of the polyimide precursor droplet on the glass substrate, the sample was soft-baked on a hotplate at 110 °C for 22 min (Figure 1a).The temperature of the hotplate was raised gradually from 70 °C, at a ramp rate of 10 °C min −1 , to avoid excessive temperature gradients across the droplet that would result in additional thermal stresses and inhomogeneous shrinking.The initial temperature was chosen to be below the soft-bake temperatures recommended by the material supplier to ensure a minimum temperature gradient across the polyimide precursor droplet, which is about one order of magnitude thicker than the layer thickness used in typical application use cases of this polyimide precursor.After soft baking, the sample was cooled down to room temperature and mounted into the 3D printer.
3D Printing Experiments: All printing experiments were performed using the commercially available 3D printer Nanoscribe Photonic Professional GT2 (Nanoscribe GmbH, Germany).The 3D printer was used to pattern photosensitive polymers by two-photon polymerization, using a femtosecond laser with a central wavelength of 780 nm and a repetition rate of 80 MHz.The printer was used in the "oil-immersion" configuration (Figure 1b), i.e., oil was placed between the objective of the laser system and the glass substrate to increase the numerical aperture of the objective.The sample was mounted in the 3D printer having the polyimide precursor on the opposite side with respect to the objective.This printing configuration was chosen over the alternative "Dip-In" configuration, where the objective was directly immersed in the photosensitive polymer because here the soft-baked polyimide precursor forms a dense matrix that does not allow the movement of the objective inside the precursor.Then, the laser was focused on the interface between the glass and the polyimide precursor using either the automatic interface finding function of the 3D printer, or manually.The printing started from that interface and formed the printed structure layer-by-layer.In this configuration, the laser light passed through already cross-linked polymer areas, which could distort the incoming laser beam and degrade the resolution of the printing process.However, no significant distortions were observed during our experiments.The distance between the adjacent layers (slicing distance) and the distance between the adjacent lines forming the layers (hatching distance) was set by the DeScribe software (Version 2.5.5, Nanoscribe GmbH, Germany), which was used to process the 3D CAD models and generate the instruction files for the 3D printer, thereby optimizing the laser trajectories and the irradiation intervals to ensure the correct printing of the desired objects.The selected slicing and hatching distances, as well as the other printing parameters (laser power and laser scan speed) and settings (Solid mode or Shell-&-Scaffold mode), depended on the size of the specific printed object and were detailed in the Results section.The structures in Figures 2d and 4 were printed on a base plate of 10 and 30 μm thickness, respectively, to reduce the tensile stress on the 3D-printed structure caused by the material shrinkage in the imidization (curing) process.In both cases, the base plates were printed with the same printing configuration as the printed object.Objects were printed with sizes spanning from a few hundred nanometers (Figure 3) to a few micrometers (Figure 2), up to millimeters (Figure 4).The microscope objective was chosen according to the size of the smallest desired features in the object to be printed.Either a 25x/NA 0.8 (Zeiss, Germany) or a Plan-Apochromat 63x/1.4Oil DIC M27 (Zeiss, Germany) was used.
After laser exposure, the sample was removed from the 3D printer, and the oil was cleaned off from the backside of the glass substrate using isopropanol.A post-exposure bake was carried out on a hotplate at 80 °C for 2 min.Then, the sample was immersed in a beaker containing the developer solution PA-401D (HD Microsystems GmbH, Germany) based on cyclopentanone (Figure 1c).After 12 h, the sample was removed from the beaker, flushed with clean developer solution using a pipette, rinsed with the rinse solution PA-400R (HD Microsystems GmbH, Germany) based on PGMEA, to remove developer residues, and dried using an air gun.When the polymer structure was printed with the Shell-&-Scaffold printing mode, a UV flood exposure of 5 min was carried out using an LED unit (12 mW cm −2 at 365 nm).
After laser exposure and development, the 3D-printed structures were composed of the cross-linked polyimide precursor.To turn the precursor into a polyimide, the samples were cured in an oxygen-free atmosphere (Figure 1d) in the Polyimide Bake Oven YES-450PB8-2/6-2 (Yield Engineering Systems, CA, US).Before this thermal treatment, the gas in the oven chamber was evacuated and the chamber was then purged with nitrogen gas.Next, the oven temperature was raised at a rate of 2 °C min −1 up to 200 °C and held at 200 °C for 30 min.Then, the temperature was further increased at the same rate up to 375 °C and held at 375 °C for 1 h.Finally, the temperature was reduced by natural cooling inside the oven until it reached 100 °C when the sample was removed from the oven.
Scanning Electron Microscopy (SEM): The samples were inspected by SEM (Gemini Ultra 55, Carl Zeiss AG, Germany) after gold deposition in a table-top sputtering system (JFC-110, JEOL, Japan).
Infrared Imaging: The thermal characterization of the micro-hotplate was carried out using the thermal camera FAST M350 (Telops, Canada) equipped with a 4x lens (NEOS Inc., US), offering a spatial resolution of 3.75 μm.The camera was calibrated between 0 and 350 °C, and measured infrared radiation with wavelengths between 3 and 5 μm.The collected images were analyzed with the software Reveal IR (version 1.6.0,Telops, Canada), which was used to compute the radiometric temperature from the measured data and convert it into real temperature by setting a material emissivity.The infrared radiation emitted by the micro-hotplate was dominated by radiation emitted from the polyimide, as polyimides had

Figure 1 .
Figure 1.Fabrication process of micro 3D-printed polyimide objects.a) A drop of polyimide precursor (HD-4100) is deposited on a glass slide and soft baked on a hotplate at 110 °C.b) The polyimide precursor is locally cross-linked by two-photon polymerization using a femtosecond laser beam traveling through the glass substrate and focused inside the polyimide precursor.c) The non-cross-linked polyimide precursor is dissolved by the developer solution.d) The cross-linked polyimide precursor is cured in an oxygen-free atmosphere and thus, imidized to become polyimide.

Figure 2 .
Figure 2. Process parameters for 3D printing of polyimide objects by two-photon polymerization.The laser scan speed was set to 12 mm s −1 for all 3D-printed structures.The laser power was set in the script fed to the 3D printer to the values of a) 8%, b) 4%, and c) 12%, respectively, of the maximum laser power of 50 mW.d) Fully cured polyimide structure 3D printed on top of a 20 μm-thick polyimide base plate with the optimized printing parameters used in (a).

Figure 3 .
Figure 3. 3D-printed nanowires and suspended double-sided clamped beams.a)The plot of the average measured width of five suspended nanowires produced with single laser scans at different laser speeds and laser powers before and after curing of the polyimide precursor.b) Nanowires made of polyimide precursor (i.e., after laser exposure and before curing at 375 °C) printed with a laser scan speed of 10 mm s −1 and a laser power of 7.5 mW.c) Nanowires of polyimide precursor (i.e., after laser exposure and before curing at 375 °C) printed with a laser scan speed of 10 mm s −1 and a laser power of 12.5 mW.d,e) Polyimide nanowires (after curing at 375 °C) printed with a laser scan speed of 10 mm s −1 and a laser power of 7.5 mW.f,g) Polyimide nanowires (after curing at 375 °C) printed with a laser scan speed of 10 mm s −1 and a laser power of 12.5 mW.h,i).Double-sided clamped polyimide beams (after curing at 375 °C) that are 1 and 3 μm wide, respectively.

Figure 4 .
Figure 4. Thermal characterization of the 3D-printed polyimide micro-hotplate.a,b) Optical microscope images of the 3D-printed hotplate before and after the curing step, respectively.c,f) SEM image of the hotplate after operation at a temperature of 220 °C.d) Plot of the measured operating temperature of the hotplate as a function of time, using different applied current pulses.e) Average operating temperature of the hotplate measured in the device area of interest (marked in (g)) as a function of applied power.g-i) Thermal infrared images of the micro-hotplate at different applied heating currents.j) SEM image of a damaged hotplate after operation at an excessive heating current of 30 mA, which we estimate to result in a heater temperature of ≈430 °C.