Direct integration of dielectric all-ceramic thick films on a polymer substrate using room temperature fabrication

Direct integration of all-ceramic thick films and a polymer substrate has been realized for the first time without high temperature processing using the Room Temperature Fabrication method. Printable Li2MoO4-BaTiO3 composite pastes with 0, 10, and 20 vol.% of BaTiO3 were fabricated from the respective ceramic powders and water without organic additives or vehicles. The pastes were stencil printed on a polyimide substrate and dried at 120 °C without pressing or lamination. Using scanning electron microscopy, the films were observed to be in seamless contact with the substrate and to have a uniform microstructure. Relative permittivities of the ceramic films increased from 4.2 to 7.2 (at 2.5 GHz) and 4.5 to 7.5 (at 9.9 GHz) according to the vol.% content of the added BaTiO3, with corresponding dielectric losses from 10 to 10. The results show that the room temperature fabrication method enables 2D printing of all-ceramic thick films on temperature-sensitive substrates.


Introduction
Printed electronics technology offers the means for the easy and cost-efficient manufacture of electronic devices with a high output. The various techniques are all additive by nature which minimizes the material waste. The most simple, adaptable and cheap technique is screen printing. Using viscous paste-like materials, 0.1−100 μm thick layers called thick films are produced on top of a substrate. [1][2][3] In general, thick films are polymer composites, where the functional ceramic particles are bound by the polymer matrix enabling easy processing at relatively low temperatures [4][5][6]. However, there is a clear requirement, from the component point of view, for all-ceramic thick films due to their versatile electrical properties. On the other hand, allceramic films require a high manufacturing temperature (hundreds or over 1000°C) which hinders their integration with temperature-sensitive substrates.
The Room Temperature Fabrication method (RTF) has previously been introduced for bulk structures [7]. It offers the means for lowtemperature ceramics processing without the use of polymeric matrixes or organic additives. In these studies, the RTF utilized the water solubility of a ceramic material to form an aqueous phase, which aided the packing of the solid ceramic particles during pressing. The ceramic material studied for the RTF was lithium molybdate (Li 2 MoO 4 ), a non-toxic dielectric material with low relative permittivity (ε r ) and loss tangent (tan δ). The results showed ε r of 5.1 and tan δ of 0.0004 measured at 9.6 GHz [7].
In this paper, the RTF was adapted to the 2D printing of all-ceramic thick films on a polymer substrate for the first time. This offers a simple route to the manufacture of a printing paste without organic vehicles or solvents and, consequently, without a binder burnout stage. In addition, the process does not require any additional steps such as pressing or lamination. Li 2 MoO 4 -based pastes with different amounts of insoluble barium titanate (BaTiO 3 ) were prepared to modify the dielectric properties of the films. The pastes were characterized by their viscosity and yielding properties. The surface roughness and microstructure of the resulting all-ceramic thick films were studied as well as the dielectric properties at 2.5 and 9.9 GHz.

Preparation and characterization of the pastes
Three different Li 2 MoO 4 -based pastes with BaTiO 3 contents of 0, 10, and 20 vol.% were fabricated. The starting powders for the pastes were prepared according to Ramachandran et al. [8], except that a smaller mesh size of 90 μm was chosen to be used in sieving due to the stencil https://doi.org/10.1016/j.jeurceramsoc.2020.04.009 Received 4 December 2019; Received in revised form 12 March 2020; Accepted 5 April 2020 height (100 μm, see Section 2.2). First, the initial particle size of the Li 2 MoO 4 (99+ %, Alfa Aesar, Karlsruhe, Germany) was reduced by milling in ethanol using ZrO 2 milling medium and a planetary ball mill (Pulverisette 6, Fritsch, Idar-Oberstein, Germany), followed by drying and sieving. Secondly, predetermined amounts of BaTiO 3 powder (99.7 %, < 2 μm, Alfa Aesar, Karlsruhe, Germany) corresponding to 10 and 20 vol.% were sonicated in ethanol to break down possible agglomerates. Finally, the resulting BaTiO 3 slurries were mixed with the Li 2 MoO 4 in a planetary ball mill, followed by drying and sieving. The powder containing 0 vol.% BaTiO 3 was milled and sieved correspondingly.
The powders were characterized with a laser diffraction particle size analyzer (Beckman Coulter LS13320, Brea, CA) in isopropanol. For the powders containing 0, 10, and 20 vol.% BaTiO 3 , mean particle sizes of 10, 8.8, and 7.2 μm were measured, while 99 % of the particles were smaller than 31, 27, and 23 μm, respectively. No ultrasonic mixing was applied prior to the analysis because previously it had been observed to break down larger particles [9].
The starting point for the optimization of the paste compositions was to keep the amount of water as low as possible without compromising printability of the paste and the surface smoothness of the printed films. Extra water increases the drying time, resulting in larger shrinkage, and can increase porosity [10]. When suitable compositions were found, the viscosities of the pastes were studied and matched. As a result, recipes for optimized pastes were established (Table 1). Careful mixing with a spatula produced uniform viscous pastes.
A study of the viscosities and yielding behaviour of the pastes was conducted using a rheometer (Discovery HR-1, TA Instruments, New Castle DE) with Ø 40 mm parallel steel plate geometry and a 1 mm gap between the plates. To prevent drying of the pastes during the measurements, a plastic cover was employed. First, the pastes were preconditioned to 22°C. Then pre-shear with a 5 s −1 shear rate for 1 min and a subsequent equilibration period of 2 min was applied. To study the viscosity, flow ramps with shear rates of 1 to 200 s −1 were measured. Yielding behaviour analyses were conducted at 1 Hz changing the oscillation stress from 0.1 to 3000 Pa.
The solid contents of the pastes after mixing were calculated using the Li 2 MoO 4 solubility in water (44.81 wt.% at 25°C [11], meaning 81.2 g/100 mL of water) and a density of the Li 2 MoO 4 -water solution (1.4897 g/cm 3 ) [12].
The paste densities were measured with a pycnometer in 1-octanol (≥ 99.4 %, Sigma-Aldrich, St. Louis, MO) at the prevailing temperature of 22.3°C. Because 1-octanol is immiscible in water it is thus suitable to use with pastes containing an aqueous phase. The density of 1-octanol at this temperature was measured to be 0.822 g/cm 3 .

Printing and post-processing
The off-contact printing (gap 1 mm) of the pastes was conducted manually using a laser-cut steel stencil (thickness 100 μm) with an aperture size of 55 × 55 mm, and a rigid steel squeegee. The minimum sample size for measuring the dielectric properties at 2.5 GHz frequency (see Section 2.3) dictated the aperture size.
Polyimide film (thickness 127 μm, Kapton 500 HN, Lohmann, Neuwied, Germany), which is a common material in printed electronics, was used as a substrate. A suction table was employed to keep the substrate straight and in place during printing. Prior to printing, the substrate was wiped clean from grease and dust with ethanol. Because of the optimized paste compositions, a single pass was enough to produce a good quality print.
Six samples were fabricated using each paste. After the printing, the samples were taped to a steel plate from the corners of the substrate to keep the substrate straight and to ease handling during the drying process. The samples were first left to dry at room temperature for 18 h, after which post-processing at 60°C and 120°C for 6 h and 3 h, respectively, was applied to speed up the water evaporation and ensure complete drying [7].

Characterization of the ceramic films
The average thickness and surface roughness (arithmetic average of deviations in the surface profile, Ra) of the ceramic films were determined using a stylus profilometer (Dektak 8, Veeco, Plainview, NY) and polynomial curve fitting software. The surface profiles were studied with a laser microscope (VK-X200, Keyence, Osaka, Japan).
The microstructures of sample surfaces and cross-sections were studied with a field emission scanning electron microscope (Ultra Plus, Zeiss, Karlsruhe, Germany). The cross-sectional samples were prepared with epoxy casting. To allow safe positioning of the samples in the casting mould, the samples were first fixed on a piece of fibreglass with epoxy followed by application of epoxy on top of the ceramic films to secure their adhesion to the substrate. After casting and hardening, the cross-sections were revealed by polishing with abrasive papers. The final polishing was conducted using P4000 abrasive paper (MicroCut, Buehler, IL). A thin layer of carbon was then sputtered on to the sample surfaces to eliminate electrostatic effects during imaging.
The dielectric properties of the ceramic films on the substrate were measured using a vector network analyzer (ZVB 20, Rohde and Schwarz, Munich, Germany) equipped with dielectric split post resonators (QWED, Warsaw, Poland) having nominal resonance frequencies at 2.475 GHz or at 9.975 GHz. The measurements were conducted at the prevailing temperature of 24°C and a relative humidity of 41 %.
Densities of the ceramic films were determined with liquid impregnation using a vacuum method [13]. 1-Butanol (≥ 99.4 %, Sigma-Aldrich, St. Louis, MO) was used as the impregnation liquid due to the water-sensitivity of the samples. At the prevailing temperature of 22°C the density of 1-butanol was measured as 0.808 g/cm 3 . Calculation of the relative densities was done based on the theoretical densities of Li 2 MoO 4 (3.04 g/cm 3 , Powder diffraction file 21-0763) and BaTiO 3 (6.01 g/cm 3 , Powder diffraction file 05-0626).

Pastes and printing
In general, a paste consists of solid particles dispersed in a continuous liquid phase. The liquid fills the inter-particle voids and separates the particles from each other thus enabling the paste to flow. [14] Here, an aqueous solution saturated with the dissolved Li 2 MoO 4 , serves as the liquid phase. It is formed during the mixing of the starting materials when water partly dissolves the Li 2 MoO 4 according to its solubility. This causes the solids content to decrease compared to the initial state before mixing ( Table 2). The undissolved Li 2 MoO 4 and the insoluble BaTiO 3 constitute the solid phase.
The flow properties of a paste are affected by the amount of the liquid phase. However, the required amount of liquid needed to achieve certain flow properties is affected by the size distribution of the solid particles. [14] In this case, the starting powders for the pastes differed in mean particle size because of the different contents of BaTiO 3 . Hence, to achieve similar paste viscosities (Fig. 1a), different amounts of liquid phase, and thus water, were required. The larger the BaTiO 3 content, the smaller the mean particle size of the starting powder, and thus, the larger the amount of the needed water. In turn, the amount of added water upon the preparation of the paste determined how much Li 2 MoO 4 was dissolved.
An important property of a printing paste is to flow during printing and then to set immediately afterwards, meaning that the viscosity of the paste must decrease upon shear and again increase again when the shear is released [1,15]. In the current case, the pastes expressed this shear-thinning behaviour (Fig. 1a) even without the use of organic solvents or other additives. The pastes exhibited similar viscosities at low shear rates (< 20 s −1 ). At higher rates, the paste with 20 vol.% BaTiO 3 had a lower viscosity due to the slightly lower solid content.
The amount BaTiO 3 particles had an effect to the yielding behaviour of the pastes (Fig. 1b). The disruption of the paste network occurred at different oscillation stresses because of the differences in the paste composition. The yield point for the pastes increased with increasing BaTiO 3 content. It was observed to be at 21, 145, and 440 Pa for the pastes containing 0, 10, and 20 vol.% BaTiO 3 , respectively (Fig. 1b). This meant that 10 or 20 vol.% addition of BaTiO 3 particles was able to make the paste network significan tly stronger.
Previously, a similar Li 2 MoO 4 -based paste was designed for 3D printing using extrusion [9]. Compared to the pastes presented here, the 3D printing paste exhibited similar shear-thinning behaviour but a higher viscosity because of the larger solid content (60.0 vol.%). Differences in the optimal paste composition resulted from the printing technique used: with extrusion, the pressure forcing the paste to flow is larger, which is why a more viscous paste suffices in this case.

Ceramic thick films
The average thickness of the ceramic films was 73 μm. The films were slightly thicker at the edges, especially at the edge where the squeegee exits the aperture during printing. This indicated an accumulation of the paste against the stencil edge and its subsequent lifting with the stencil releasing back to the gap position after printing. Compared to the stencil height of 100 μm, the films were somewhat thinner because of shrinkage. With increasing BaTiO 3 content, the surface roughness decreased from 3.5 μm to 2.0 μm. This was due to the different water content (Table 2) and the mean particle size in the pastes. Smooth surfaces are desirable when dimensionally accurate conductive lines on top of the film are needed. The surface microstructures showed no cracking upon drying, however, some pinholes were observed (Fig. 2b).
Microstructural analysis of the sample cross-sections showed that the ceramic films were in seamless contact with the substrate (Fig. 3a). Furthermore, the BaTiO 3 particles were evenly distributed between the Li 2 MoO 4 particles as agglomerate-like clusters while some porosity could also be observed (Fig. 3b-d).
The dielectric properties of the samples are presented in Tables 3 (at 2.5 GHz) and 4 (at 9.9 GHz). For the plain polyimide substrate, ε r of 3.4 was measured at both frequencies with tan δ values of 0.007 at 2.5 and 0.008 at 9.9 GHz. The substrate affected the measured overall dielectric properties of the samples (ceramic film and substrate). The values of ε r and tan δ increased with respect to the amount of added BaTiO 3 . This was expected since BaTiO 3 is commonly used to adjust the dielectric properties of composites [4,[6][7][8][16][17][18][19][20].
Previously, Li 2 MoO 4 ceramics have been manufactured with RTF using pressing in a mould and extrusion based 3D printing [7,9]. At 9.6 GHz these samples exhibited ε r of 5.1 and 4.4, and tan δ of 0.0004 and 0.0006, respectively. Compared to the ceramic film with 0 vol.% of BaTiO 3 measured at a similar frequency of 9.9 GHz (Table 4), the value of ε r matched that of the 3D printed sample. However, tan δ was twice as high.
RTF manufacture of Li 2 MoO 4 -based composites from 10 vol.% of BaTiO 3 using pressing in a mould resulted previously in ε r of 8.2 and tan δ of 0.026 at 9.2 GHz [7]. Here, the corresponding sample, a ceramic film with 10 vol.% of BaTiO 3 , exhibited somewhat lower ε r while tan δ was the same magnitude of 10 −2 at a similar frequency of 9.9 GHz (Table 4). Similar samples with 9 vol.% of nanosized BaTiO 3 have also been manufactured for a weight optimized antenna, which exhibited ε r of 10.5 and tan δ of 0.006 at 1 GHz [8].
Despite the fact that RTF was applied in all of the above-mentioned cases, different manufacturing techniques were used for the sample  preparation and thus different paste compositions were needed. As a result, the dielectric properties of the resulting samples differed from case to case. This was due to the different densities. Moreover, the dielectric properties of BaTiO 3 depend on the measurement frequency as well as on the particle and grain sizes [21] which makes comparison of dielectric properties between different BaTiO 3 containing composites difficult.

Conclusions
The Room Temperature Fabrication method, used earlier to fabricate bulk ceramics with pressing in a mould and with 3D printing, is now shown for the first time to be feasible also in the 2D printing of allceramic thick films.
The compositions of the Li 2 MoO 4 -based printable pastes were optimized without organic additives or vehicles, just by controlling the Fig. 2. a) Surface profile of a corner of a sample with 20 vol.% BaTiO 3 (yellow to red) showing the accumulated paste at the sample edge (white arrow). b) Secondary electron microscope image of the same sample demonstrating the typical crack-free surface microstructure with a pinhole (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).  Table 3 The average values of dielectric properties of the samples measured at 2.5 GHz. BaTiO Table 4 The average values of dielectric properties of the samples measured at 9.9 GHz. amount of water. Composite manufacture with 10 and 20 vol.% of BaTiO 3 allowed adjustment of the dielectric properties. These results indicate that this approach allows the fabrication of all-ceramic composite thick films with a very broad variety of electrical properties by the selection of different ceramic fillers. Further investigations should also include detailed adhesion testing although the ceramic film and polymer substrate appeared to be in seamless contact.

Declaration of Competing Interest
None.