Fabrication of polyetheretherketone (PEEK)-based 3D electronics with fine resolution by a hydrophobic treatment assisted hybrid additive manufacturing method

Additive manufacturing (AM) is a free-form technology that shows great potential in the integrated creation of three-dimensional (3D) electronics. However, the fabrication of 3D conformal circuits that fulfill the requirements of high service temperature, high conductivity and high resolution remains a challenge. In this paper, a hybrid AM method combining the fused deposition modeling (FDM) and hydrophobic treatment assisted laser activation metallization (LAM) was proposed for manufacturing the polyetheretherketone (PEEK)-based 3D electronics, by which the conformal copper patterns were deposited on the 3D-printed PEEK parts, and the adhesion between them reached the 5B high level. Moreover, the 3D components could support the thermal cycling test from −55 °C to 125 °C for more than 100 cycles. Particularly, the application of a hydrophobic coating on the FDM-printed PEEK before LAM can promote an ideal catalytic selectivity on its surface, not affected by the inevitable printing borders and pores in the FDM-printed parts, then making the resolution of the electroless plated copper lines improved significantly. In consequence, Cu lines with width and spacing of only 60 µm and 100 µm were obtained on both as-printed and after-polished PEEK substrates. Finally, the potential of this technique to fabricate 3D conformal electronics was demonstrated.


Introduction
Three-dimensional (3D) electronics, which enable the integration of functional circuits into 3D devices, possess the advantages of lightweight, high integration and small size, meeting the development demands of the electronic products, especially in the aerospace field [1,2]. Thus, how to fabricate high-quality 3D conformal electronics efficiently has attracted widespread attentions of both academia and industries recently.
In the past years, 3D components with embedded electronics such as sensors [3,4], batteries [5] and antennas [6,7], have been fabricated successfully by using hybrid additive manufacturing (AM) technologies [8,9], which mainly consist of the 3D printing of insulators and direct writing (DW) of conformal circuits thereon. The former is usually conducted by the fused deposition modeling (FDM), stereolithography (SLA), selective laser sintering (SLS) and slurry dispensingsintering methods [10][11][12], etc, and the typical techniques of the latter are inkjet printing, dispensing, aerosol jet printing [13][14][15], etc. Tehrani et al [16] and Li et al [17] prepared the circuits on SLA-printed 3D objects by inkjet printing and micro-dispensing, respectively, by which the mm-wave wireless packaging device and embedded electronics were obtained. Aerosol jet printing was used to fabricate conformal silver circuits on the SLS-printed safety cage for drone [11], which replaced the conventional connecting wires and saved the weight of system. Zhou et al [18] proved the feasibility of fabricating embedded electronics and smart structures by using FDM and inkjet printing. However, the electronic ink or pastes used for the above DW processes are mainly composed of conductive particles, organic stabilizers and glass binder, which usually provide lower electrical conductivity than pure metal [8]. Additionally, in order to achieve the desired properties of the conductive layer, a bulk-sintering treatment for the component is essential, which shows low versatility on different substrates.
Laser activation metallization (LAM) is a promising method of depositing high-conductivity metallic patterns on insulators [19], and it is usually conducted by the procedures including laser selective treatment and electroless copper plating (ECP). The laser selective treatment can modify the substrates and activate the metallic precursors preset thereon, then the ECP process can create conductive patterns on the laseractivated zone. The major advantages of LAM technique have been verified as follows: (1) the laser activation can be realized on different substrates with the assistance of precursor film. So far, LAM has been applied to fabricate metallic patterns on polymer [20][21][22][23][24][25], ceramic [26,27] and glass [28,29] substrates, etc; (2) The laser treatment is able to fabricate micro-groove structures on the surface of substrate, which can lead to a firm mechanical bonding strength between the coating and substrate [30-32]; (3) The conductivity of the metal layer is basically equal to the bulk value, much better than that obtained by DW technologies.
Recently, Li et al [33] have prepared the 3D directcurrent electronics and microwave/millimeter-wave antennas by combining the SLA and LAM technologies. It enabled the free-form creation of high-conductivity patterns on the 3D parts with strong adhesion and high accuracy, showing great potential in the application of functional 3D electronic devices. Yet, the supporting materials or dielectric substrate used in the above method can only be limited to photo-curable polymers, whose heat distortion temperatures and dielectric properties are relatively low, limiting its wide applications in high-temperature and high-frequency service conditions [34]. Aiming this, Wang et al [12] proposed a hybrid AM technology by combining dispensing 3D printing process and LAM for fabricating functional ceramic electronics. Though the ceramic matrix was able to endure high temperature and high voltage, it was more brittle and showed lower toughness compared to polymer.
Poly-ether-ether-ketone (PEEK), with a melting point of 343 • C, a dielectric constant of 2.95 (24 GHz) and a dielectric loss tangent of 0.0038 (24 GHz), is an ideal circuit packaging material for highly integrated electronics. And, lots of studies on the printing of PEEK parts by FDM have been reported recently [35,36], however, its related applications in 3Dprinted electronics are few up to now.
In this work, we develop a hybrid AM technology combining the FDM and LAM to create conformal metallic patterns on the 3D-printed PEEK parts. The forming characteristics of PEEK in FDM process, the laser-PEEK interaction mechanism in LAM process, and their effects on the quality of the plated Cu patterns were investigated systematically. Particularly, aiming to overcome the deficiencies in the above method, a hydrophobic treatment on the FDM-printed PEEK was performed before LAM process to modify its surface properties, by which the resolution of the metallic patterns deposited thereon was improved significantly.

Preparation of PEEK substrates
A FDM 3D printer produced by Jugao Corporation (China) was used to fabricate the PEEK substrates. It mainly consisted of an extrusion head, a filament feeder, a temperature control system, a modeling platform and a 3-axis motion mechanism, as shown in figure 1. PEEK filaments with diameter of 1.75 mm were used as the raw materials of FDM, and they were extruded from the PEEK pellets (551G, ZYPEEK, China). All the FDM processes were conducted using a nozzle with diameter of 0.4 mm at room temperature.
According to the previous studies [35], the infill spacing is an important factor that affects the forming characteristics of FDM-printed components. Therefore, different infill spacings (D = 0.36-0.50 µm) were conducted, while other parameters including the nozzle temperature, printing speed and layer thickness were pre-optimized to the constant value of 460 • C, 40 mm s −1 and 0.2 mm, respectively. All the samples were tempered at 300 • C for 3 h in a furnace and then cooled down to room temperature, after which better mechanical properties could be obtained [37]. In order to improve the surface accuracy, some of the as-printed PEEK substrates were polished by silicon carbide abrasive papers up to 2500 CCR/R. In addition, commercial injection-molded PEEK substrates with the same raw material were adopted for a comparative study.

Fabrication of copper patterns
After the fabrication of PEEK substrates, the copper patterns were deposited as the following procedures, and the schematic diagram is shown in figure 2.
(1) Firstly, the PEEK substrates after ultrasonically cleaning were immersed in a commercial silicone solution (SJ-2600, Jining Tangyi Chemical Co., China) for 5 min, and then dried in the air for 60 min to form a hydrophobic coating on the surface. (2) Secondly, a femtosecond fiber laser (FemtoYL ® -IR-100W, China) with the wavelength of 1064 nm, pulse duration of ∼300 fs and focused spot diameter of ∼25 µm was conducted to selectively ablate the hydrophobic coating in the areas where the copper layer was intended to be deposited. A dynamic focus system which is mainly composed of the dynamic focusing galvanometer (Z-axis working distance: ±∼10 mm, RAYLASE, Germany), a control board and software was conducted for achieving the conformal scanning of laser on the 3D surface.
(3) Next, the sample was immersed in a 5.0 g l −1 PdCl 2 aqueous solution with pH = 1.42 for 5 min. Afterwards, PdCl 2 precursor films were formed in the laser-treated zones after they were dried in the air. (4) The laser activation was conducted in the areas where the PdCl 2 precursor films had been formed. (5) The injection of laser energy led to the modification of substrate and PdCl 2 simultaneously, and embedded active seeds were generated in the laser-irradiated zone. (6) Then, the laser-activated samples were ultrasonically cleaned in the deionized water for 5 min. (7) The samples obtained in step (6) were immersed in a commercial ECP solution at 57 • C for 1-2 h to realize the selective deposition of copper on the substrates. The detailed constitute of the ECP solution was: 6 40 mg l −1 , NiCl 2 10 g l −1 , HCHO 25 ml l −1 , NaOH 2-3 g l −1 . (8) Finally, the sample was immersed in a 500 g l −1 sodium hydroxide solution at 80 • C for 20 min, breaking the Si-O bond and consequently removing the hydrophobic coating on the unmetallized areas, which showed no effect on the quality of the deposited copper.

Measurements and characterizations
After FDM printing, the actual densities of different PEEK samples were measured according to the Archimedes' drainage method by using an electronic balance (FA2204, Lichen, China). Then, the relative density (ρ r ) of FDM-printed PEEK samples, which is a dimensionless constant, was calculated from the ratio of their actual densities to the theoretical density of pure PEEK (about 1.35 kg m −3 in this work). The static contact angles of the deionized water droplet (4 µl) on the surface of different samples were measured with a contact angle equipment (Kruss-k100, Germany) at room temperature. Each sample was tested on five different zones and the average value was taken. After laser treatments, the microstructure of surface textures was examined by an environmental scanning electron microscope (Quanta 200 ESEM, FEI, Netherlands), and the corresponding 3D topography was observed by a 3D profilometer (VK-X200K, Keyence, Japan). Surface chemical analysis was conducted via an energy dispersive spectroscopy (EDS, Oxford x-act, UK), an x-ray photoelectron spectroscopy (XPS, AXIS-ULTRA DLD-600W, Japan) and a fourier transform infrared spectroscopy (FT-IR, Nicolet Is50R, Thermo Scientific, USA).
After ECP, the macroscopic morphologies of Cu patterns were observed by a stereomicroscope (Stemi 508, Carl Zeiss, Germany). The adhesion of Cu patterns was evaluated by the tape test (Scotch610, 3M) according to the ASTM D3359 standard. The resistivity of Cu patterns was tested by the fourprobe conductivity measurement (ST2258C, China). Then, the thermal cycling performance of the integrated components was measured in a temperature change test chamber (TS-120SW, China) according to the IPC-TM-650 standard, in which the testing temperature range was −55 • C to 125 • C, and the testing cycle was 100.

Forming characteristics of PEEK substrates
Figures 3(a) and (b) demonstrate the relative densities (ρ r ) and surface roughness of PEEK substrates formed by FDM with different infill spacings (D). When D ⩽ 0.36 mm, the sample was hardly to form because the overlapping ratio of adjacent filaments was too high to cause the clogging of nozzle. When D distributed in the range of 0.38 mm-0.40 mm, ρ r reached the highest of about 94%-95%, which was close to the optimal value in the previous work [36]. In the meantime, the lowest surface roughness (Sa) was obtained, however, it was still as high as about 3.38-3.50 µm. As shown in figure 3(c), printing borders could be observed at the overlapping zone of adjacent printing lines in the as-printed PEEK substrate with D = 0.38 mm, and they disappeared completely after polishing treatment, along with the decrease of Sa to about 0.47 µm. However, some minor pores could be still detected therein, as shown in figure 3(d). With the further increase of D, ρ r decreased rapidly, accompanied by the generation of serious depressions and larger pores at the overlapping zone, as well as the re-enlargement of Sa, as shown in figures 3(b), (e) and (f). As a comparison, the injection-molded PEEK showed the highest ρ r of ⩾99% and the lowest surface roughness of Sa = 0.11 µm. Combined with the above results, it was obtained that the as-printed PEEK substrates usually had rough surface due to the existence of printing borders, which may be harmful to the uniformity and continuity of the circuits [18]. The polishing treatment was capable of removing the borders. When D = 0.38-0.40 mm, the polished PEEK substrates possessed a relatively smoother and denser surface among all the printed samples, however, it was still inferior to that of the injectionmolded PEEK.

Laser-PEEK interaction mechanisms.
Since the raw materials of all PEEK samples in this work were the same, the PEEK substrate in figure 3(c) (D = 0.38 mm, after polishing treatment) was selected as a representative to study the effects of laser fluence (E) on the physicochemical properties of PEEK and PdCl 2 .
Figures 4(a 1 )-(c 1 ) present the surface morphologies of the PEEK substrates treated with different laser fluence, and the corresponding 3D topographies are shown in figures 4(a 2 )-(c 2 ), in which the laser scanning speed (V), repetition rate (F) and scanning spacing (S) were set to 1000 mm s −1 , 500 kHz and 30 µm, respectively. When the laser fluence was relatively low (E = 0.78 J cm −2 ), only slight melting occurred on the surface of PEEK, resulting in an increase of Sa to 1.09 µm (figures 4(a 1 ) and (a 2 )). When treated with a higher laser fluence (E = 2.49 J cm −2 ), a porous groove array structure with depth of ∼8.5 µm was generated on the surface of PEEK, which improved the Sa to about 2.94 µm (figures 4(b 1 ) and (b 2 )). The roughened zone could increase the in-contact area between the subsequent plated coating and substrate, contributing to a stronger mechanical interlocking, thus improving the adhesion strength of the coating [38]. As the laser fluence (E) increased to 12.35 J cm −2 further, both bulges and depressions were detected on the surface, in which the maximum height difference was as high as about 100 µm and the Sa increased to about 10.28 µm synchronously (figures 4(c 1 ) and (c 2 )). This may be due to the extreme high injection of laser energy, which caused the local burning of PEEK material.
The surface chemical identification of different samples was performed with the XPS analysis and the results are given in figure 5, in which the calibration was done with the C 1 s peak at a value of 284.6 eV. Figures 5(a 1 )-(d 1 ) gives the C 1 s spectra that fitted by XPSPEAK 41. Three components were detected on the surface of unirradiated PEEK, which were assigned to C-H/C-C/C=C (284.6 eV), C-O (286.0 eV) and C=O (287.5 eV), respectively, as shown in figure 5(a 1 ). After the laser treatment with laser fluence (E) at 0.78-2.49 J cm −2 , a new peak corresponding to O=C-O (288.6 eV) was observed, accompanied by the increase of the polar functional group's amount (C-O) (figures 5(b 1 ) and (c 1 )). The polarization of the surface was attributed to the chemical modification of the injected laser energy to the PEEK, and the enhanced surface polarity could usually promote the adhesion strength of copper on the substrate [39]. When the laser fluence (E) further increased to 12.35 J cm −2 , there was a significant  increase in the carbon bonds and a decrease in the polar groups ( figure 5(d 1 )). This may be caused by the carbonization of the PEEK surface, which was consistent with the result in the previous work [40]. Figures 5(a 2 )-(d 2 ) show the fitted results of Pd spectra. The Pd 3d spectra of the unirradiated PdCl 2 film on PEEK ( figure 5(a 2 )) mainly displayed as a doublet, in which the Pd 3d 5/2 and Pd 3d 3/2 peaks were located at 337.8 eV and 343.0 eV, respectively, indicating that it was mainly attributed to the Pd 2+ in PdCl 2 [41]. After treatment with the laser fluence of E = 0.78 J cm −2 , the spectra were divided into two doublets with the first peak located at Pd 3d 5/2 (337.8 eV) and the second peak located at Pd 3d 5/2 (335.6 eV), as shown in figure 5(b 2 ). The former was known as PdCl 2 and the latter was corresponding to Pd 0 [42]. This demonstrated that the PdCl 2 film was not completely decomposed after the irradiation at laser fluence of E = 0.78 J cm −2 . When the laser fluence increased to E = 2.49 J cm −2 , only a doublet located at Pd 3d 5/2 (335.6 eV) was detected (figure 5(c 2 )), indicating that almost all of the PdCl 2 in the laser-irradiated zone had been converted to Pd 0 . With the further increase of laser fluence to E = 12.35 J cm −2 , the Pd 3d peak disappeared (figure 5(d 2 )), indicating that the PdCl 2 had vaporized directly at the high laser energy input, and no Pd-based active seeds existed in the laser-treated zone.

Electroless plating of Cu patterns.
After ultrasonically cleaning, all the laser-activated samples were put into the ECP solution for copper deposition. Firstly, an initial redox reaction as equation (1) was motivated by the catalysis of Pdbased active seeds, making the copper (Cu 0 ) generated on the laser-activated zone, Then, the reaction was proceeded at the autocatalytic effect of the copper. The reaction equation is as follows: (2) Figure 6 shows the images of the plated-copper on different PEEK substrates, in which the plating time of all samples was about 90 min. For the sample treated by the laser fluence of E = 0.78 J cm −2 , copper particles only deposited in part of the laser-treated area ( figure 6(a)). This was due to that plenty of undecomposed PdCl 2 particles have remained on the surface, which was soluble in water and easy to be removed during the ultrasonic cleaning, leading to the incomplete catalytic activity therein. When the laser fluence increased to E = 2.49 J cm −2 , a uniform copper layer with thickness of about 12-20 µm was formed on the PEEK surface, as shown in figure 6(b). Its resistivity was tested to be about 2.9 × 10 −6 Ω.cm, close to that of the Cu bulk. Moreover, none of the lattices has detached from the substrate after the tape test, as demonstrated in figures 6(d) and (e), indicating that the adhesion of the copper layer to the PEEK substrate has reached the 5B high level. With the further increase of laser fluence to E = 12.35 J cm −2 , no copper was deposited in the laser-treated areas due to the active particles therein have vaporized (figure 6(c)), consistent with the results in figure 5(d 2 ).
By summing up above results, the optimized laser fluence (E) for laser activation was determined to 2.49 J cm −2 , while other laser processing parameters were: V = 1000 mm s −1 , F = 500 kHz, S = 30 µm.   patterns were directly fabricated on the PEEK substrates, and the surface morphologies are presented in figure 7. It is seen that the uniformity and deposition accuracy of the copper pattern on the surface of as-printed PEEK was extremely poor (figure 7(a)), in which plenty of copper has deposited in the un-activated area. The ESEM image inset shows that they mainly distributed in the printing borders or pores. After the removal of printing borders by polishing, the surface uniformity and deposition accuracy were enhanced significantly, however, a few of copper particles could still be detected in the un-activated area ( figure 7(b)). As indicated by the inset ESEM image, it was mainly caused by the inevitable pores in the FDM-printed PEEK substrates. Comparatively, the copper patterns on the injection-molded PEEK substrate (ρ r = ∼99%) showed much higher deposition accuracy ( figure 7(c)).

Fabrication of fine-resolution Cu patterns on
The above results suggest that the pores in the FDM-printed PEEK substrates are the main cause that leads to the inaccurate deposition of Cu patterns. Next, the influence mechanism is analyzed in detail. Figure 8 shows the wetting behavior of water on FDMprinted PEEK substrates with different density, in which the water contact angle on the PEEK substrate with ρ r = 95% was about 75 • , and it decreased to about 64 • as ρ r decreased to 88%, indicating that the PEEK substrate with more pores on the surface was more hydrophilic. The adhesive work (W a ) between the surface and water can be calculated by the followed Ownes-Wendt equations: where γ H2O and θ H2O is the water surface energy and water contact angle, respectively. Obviously, a lower θ H2O can lead to a larger W a . It indicates that the PEEK substrate with more pores on the surface has a larger adhesion force with water or aqueous solutions. As indicated in figures 9(a) and (b), the PdCl 2 aqueous solution was easy to seep into the pores of FDM-printed PEEK and form a film, which may be the main reason that led to a smaller θ H2O on the surface of PEEK substrate with higher porosity. The Pd wt% therein was detected to be about 21.8%. After ultrasonically cleaned for 5 min, a large number of tiny PdCl 2 particles with content of 5.2 wt% still attached to the pores, as shown in figure 9(c). With the extending of ultrasonic cleaning time, the content of Pd decreased, however, it could be still detected even the cleaning time had increased to 30 min (figure 10). The residual PdCl 2 therein could be reduced by the HCHO in the ECP solution and became the active centers   Pd 0 , which then catalyzed the ECP reaction. Therefore, the deposition of Cu particles on the un-activated area of FDMprinted PEEK is inevitable since the pores therein cannot be completely avoided.

Improving the resolution of Cu patterns with the assist-
ance of a hydrophobic coating. As indicated by the above analysis, the inaccuracy of selective metallization on the FDM-printed PEEK substrates was mainly caused by the defects therein. Therefore, a silicone hydrophobic coating was introduced to modify the surface properties of printed PEEK before LAM, and the implementation details can refer to the description in section 2.2 and in figure 2. Figure 11 illustrates the reaction mechanism during this process. The silicone solution exposed in the air underwent a hydrolysis reaction (step 1) and condensation reaction successively (step 2). Then, a uniform hydrophobic coating with the thickness of about 20 µm was uniformly covered on the surface of PEEK substrate, and the water contact angle thereon  increased to 137 • , as indicated in figure 12(a). The corresponding EDS spectrum shows the Si wt% in the coating was about 16.61% ( figure 12(b)). Figure 12(c) shows the FT-IR spectra of the as-printed and coating-covered PEEK substrate. The group vibration modes corresponding to the infrared characteristic peaks are shown in tables 1 and 2, respectively. For the coating-covered PEEK, the peaks associated with Si-CH 3 (1258 cm −1 , 796 cm −1 ) and Si-O-Si (1073 cm −1 , 1015 cm −1 ) were detected. Based on the orientation of methyl groups, the coating was identified to be hydrophobic.  . When E = 0.96 J cm −2 or 1.70 J cm −2 , the injected laser energy was not enough to remove the hydrophobic coating completely. As E increased to 2.75 J cm −2 , PEEK and the hydrophobic coating thereon were both ablated, and a slightly roughened surface of PEEK was exposed (as indicated in figure 13(d)). When E = 0.96 J cm −2 or 1.70 J cm −2 , the injected laser energy was not enough to remove the hydrophobic coating completely. As E increased to 2.75 J cm −2 , PEEK and the hydrophobic coating thereon were both ablated. A slightly roughened surface of PEEK was exposed, as shown in figure 13(d). As a result, the PdCl 2 solution could only spread in the laser-ablated zone. Figures 14(a) and (b) show the obtained metallic patterns thereon, which possess much clearer edges and higher resolution than those in figures 7(a) and (b). And, Cu lines with a minimum width and spacing of about 60 µm and 100 µm were fabricated on  the surface of PEEK substrates successfully, including the asprinted specimens with obvious printing borders and that after polishing treatment, as shown in figures 14(c) and (d). Here, it should be noticed that the minimum width of Cu lines was proportional to the spot size, and a finer Cu line could be obtained by adjusting the focusing diameter of laser beam. show the PEEK-based component with embedded Cu circuits and its application in illumination. Above all, the hybrid AM technique composed of the FDM and hydrophobic treatment assisted LAM shows great potential in the fabrication of highaccuracy 3D structural electronics, and it is little affected by the surface quality of substrates. The research has established sufficient basics for its application in the environment needing high temperature, high frequency and fine resolution.

Conclusion
In this work, a hybrid AM technology that integrated FDM and hydrophobic treatment assisted LAM was proposed to fabricate 3D structural electronics. The forming characteristics of the FDM-printed PEEK substrates, the laser-PEEK interaction mechanism, and the adhesion strength and deposition accuracy of copper patterns were investigated systematically.
PEEK parts with highest relative density of about 94%-95% could be achieved by using FDM technology with the optimized infill spacing of 0.38 mm-0.40 mm. Uniform copper patterns exhibited low resistivity of 2.9 × 10 −6 Ω.cm and high adhesion of 5B level could be conformally fabricated on the FDM-printed PEEK substrates. And, the integrated components could support the thermal cycling test from −55 • C to 125 • C for more than 100 cycles. However, compared to the injection-molded PEEK, the FDM-printed PEEK possessed lower surface quality due to the inevitable defects therein, such as printing borders and pores, making the deposition accuracy of Cu patterns reduced significantly. Aiming this, a removable hydrophobic coating was introduced to modify the surface properties of FDM-printed PEEK before LAM process. It could promote the highly selective wetting of PdCl 2 aqueous solution to the surface of FDM-printed PEEK, then making the resolution of the plated copper lines improved significantly.
The smallest width and spacing of Cu lines reached about 60 µm and 100 µm, whether on the surface of as-printed PEEK substrate or that after polished. On this basis, series of fine conformal copper patterns were successfully fabricated on the FDM-printed 3D PEEK parts.
This study demonstrates the possibility of fabricating 3D structural electronics with high adhesion and fine resolution by the hybrid AM technology including FDM and LAM, and we will explore its applications in the manufacturing of multimaterial and multilayer 3D devices later.