Laser assisted direct ink writing of multi‐material bismuth molybdate—silver artificial dielectrics for radio frequency applications

Additive manufacturing of co‐fired low temperature ceramics offers a unique route for the fabrication of novel 3D radiofrequency (RF) and microwave components, embedded electronics and sensors. This study demonstrates the fabrication, materials analysis, and RF characterization of a multi‐material bismuth molybdate—silver (Bi2Mo2O9—Ag) artificial dielectric fabricated by laser assisted direct ink writing. The proposed fabrication technique enables 3D printing of dissimilar materials while minimizing inter‐material diffusion through the liquid phases. The permittivity of the artificial dielectrics increased up to 99% over the investigated frequency range (8–12 GHz) compared to the fabricated pure ceramic sample.

temperature below that of the electrode materials' melting point, and chemical compatibility with electrode material and so on.Furthermore, artificial dielectrics can be used to tailor specific dielectric properties such as the dielectric permittivity.Artificial dielectrics are composite structures of arrays of conductive material in a nonconductive matrix.][11][12][13] Reduction of the sintering temperature of the ceramic material to a level such that it can be co-fired with silver is essential.The potential use of the Bi 2 Mo 2 O 9 (BMO) for the fabrication of low temperature co-fired ceramics (LTCC) modules and novel RF substrates are worth considering due to its low sintering temperature and outstanding dielectric properties.Ag has high electrical conductivity which ensures a low conductor loss for RF applications.Although an interface reaction has been reported in literature 10,14 between BMO and Ag paste/ink, our previous work 15 showed a 3D printable material system based on co-fired Ag and BMO slurries based on the selective laser burnout (SLB) technique to create dense metal-ceramic parts.The work specifically emphasized the impact of employing the SLB technique prior to co-firing to reduce or eliminate both physical changes and chemical interactions between the BMO and Ag.
In the current study, BMO-Ag artificial dielectric structures with varying geometries have been fabricated and characterized.Material composition, properties, and interfaces have been investigated as well as the effect of metal island spacing on the dielectric properties of the final artificial dielectric structures.Even the relatively simple structures produced in this work could not be produced by conventional processes such as slip casting or screen printing as depositing the second material would distort the shape of the first, thereby proving the need for direct ink writing.The use of low viscosity inks to flow between a more solid first phase is not suitable either as solid loading would have to be low, leading to shrinkage and cracking.This process is the first step towards highly dense ceramics with great geometric design freedom, capable of complex metal and ceramic structures with geometric potentials beyond traditional LTCC, such as 3D metamaterials and metamaterial components.

2.1
Feed-stock materials

BMO slurry
Bi 2 Mo 2 O 9 (BMO) (American elements) was ball-milled in a PTFE jar with alumina balls to break up the agglomerates and reduce the average particle size to 1-5 μm.This was done to avoid nozzle clogging due to the presence of large particles, Figure 1.The ceramic slurry was prepared by mixing 81 wt% BMO powder in a ball-mill (Fritsch Pulverisette 7 Micro Mill) with 14 wt% ethylene glycol dimethacrylate, 3 wt% propylene carbonate and 2 wt% diisononyl phthalate to achieve appropriate viscoelastic properties for 3D printing. 15he slurry viscosity was adjusted to allow a slight flow after deposition to fill in gaps between the individual tracks or discontinuities in the print due to pressure drops from air bubbles.

(A) (B)
F I G U R E 1 SEM image of the BMO powder (A) before and (B) after ball-milling

Ag paste
The conductive Ag paste (PMC3 silver, Mitsubishi Materials Trading Corporation) was used to print the electrodes.According to the manufacturer, it consists of 90 wt% Ag particles (<5 μm), an organic binding agent and water which can be fired at temperatures between 600 • C and 900 • C with a low shrinkage (approximately 10%).The silver paste used in this study has been chosen due to its ideal rheological properties with a higher storage modulus than other inks which enables shape retention after dispensing.Through the SLB process and furnace sintering all the additives are burned off and what remains is pure silver unlike the conventional pastes containing glass frit.Conventional Ag pastes with glass frit also has a low storage modulus making only thin structures possible and is therefore not suitable to print in the third dimension.Additionally, the conductivity of the used paste is higher (5.7 × 10 6 S/m) than the commonly used pastes (1.75-4.63 × 10 6 S/m 16,17 ) and the shrinkage rate of that is very close to the shrinkage rate of the BMO. 15Thus, the shrinkage rate mismatch during sintering is reduced which should reduce the risk of fractures due to mechanical stress.

Sample preparation
Printing and laser processing experiments were performed using an OurPlant XTec system (Häcker Automation GmbH, Schwarzhausen, Germany) equipped with a DX-30 syringe dispensing module and using a Coherent-DILAS InGaAs fiber-coupled multimode diode laser (DILAS Diodenlaser GmbH, Mainz, Germany), operating at a wavelength of 980 ± 10 nm, producing a beam of 330 μm diameter.
The fabrication process uses layered slurry dispensing followed by selective laser burnout (SLB) to fabricate mono-material (ceramic only) and multi-material (ceramic/silver) green parts previously described in Reference 15.The slurry/paste was extruded onto an Al 2 O 3 substrate with a controlled flowrate (BMO 0.3 μL/s, Ag 0.645 μL/s) through a circular nozzle (d = 610 μm).Once the deposition of a layer was complete, it was scanned with a low laser energy density so that generated heat was conducted into the wet layer to uniformly burn out the organic components of the slurry with only ceramic/silver powders remaining without sintering/melting it.
Laser scanning was performed in X and Y directions for each layer.The vector spacing (VS) in X and Y direction was set to 200 μm.When laser scanning of a layer was complete, the dispensing head was offset by 100 μm in the Z direction which defined the thickness of the next layer to be deposited.Finally, multi-material BMO-Ag green bodies were co-fired at 665 • C for 4 h for further microstructural and EM characterization.

Physical end electromagnetic characterization
Micrographs were collected using a JEOL 7800F scanning electron microscopy (SEM).Powders samples were characterized using the same SEM device and a Malvern Mastersizer 2000 laser diffraction based particle size analyzer.The phase assemblage of sintered bodies identified using Bruker D2 PHASER x-ray diffractometer.Electromagnetic properties of the BMO and BMO-Ag samples 22.86 mm in length, 10.16 mm in width, and 1.5 mm in height were obtained using Anritsu ShockLine Vector Network Analyzers (MS46522B) between 8 and 12 GHz.

3D printing
Artificial dielectrics were 3D printed with arrays of metallic inclusions (Ag) embedded within the dielectric host material (BMO), Figure 2A,B.The distance between the inclusions was changed, thereby increasing the metal volume fraction, with the aim to control the permittivity of the artificial dielectric.
The printing was implemented in a layered manner which means base layers (BMO) were deposited first and then a BMO "ice cube tray" type host structure with Ag cuboid inclusions were printed on top of that layer-by-layer, Figure 2A,B.In terms of printing multi-material parts, the layer height alignment is a key consideration.An unaligned layer height for the BMO and Ag could cause a collision between the previously deposited layer and the nozzle while it is depositing a fresh layer which results in damaging the print.Figure 2C,D shows examples of failed prints as the BMO and Ag layers were not aligned causing a couple of the Ag layers to be peeled off or even the whole part was displaced by the nozzle traveling over it completely destroying the build.
Without using the SLB the multi-material parts failed during co-firing due to issues such as different shrinkage rates leading to cracking and warping as well as chemical interaction.The impact of employing the SLB technique prior to co-firing has already been reported. 15The interfacial reaction between the BMO and Ag was the result of the organic binders in both materials mixed into each other which caused the dark contrast in the ceramic body Figure 3A.It is hypothesized that Ag particles from the PMC leach into the BMO matrix during drying or co-firing.The warpage of the structures was due to the shrinkage rate mismatch of the BMO and Ag (Figure 3B), which was later reduced using the SLB technique.Figure 3C,D shows SEM images of the Ag paste before and after the SLB, respectively.It clearly indicates that the binding agents covering the particles (Figure 3C) were removed after performing the SLB and only the Ag particles left (Figure 3D).
An optimized binder burnout process led to the fabrication of multi-material structures free of cracks with significantly reduced interfacial reactions and shrinkage mismatch.The SLB process eliminated the outgassing issue and minor bursts within the body during the furnace sintering and resulted in crack-free parts.Although the interfacial reaction between the BMO and Ag was significantly reduced, the SEM analysis together with EDX elemental analysis showed a small amount of chemical reaction at the interface reaching <200 μm into the BMO. Figure 4A shows the interface between ceramic and silver, it was observed that the BMO matrix underneath the Ag layer has a darker color compared to the rest of the ceramic body which has a light color.Figure 4B-F depicts SEM images and the corresponding elemental analysis of four distinct areas labeled as P1, P2, P3, and P4, respectively.The labeled areas correspond to the Ag layer (P1), the interface between BMO and Ag (P2), towards the edge of the dark area in the BMO body (P3), and the light color area (P4).Figure 4G shows the EDS spectrum for P4.P1 was confirmed to be pure Ag through the EDS analysis while P2 and P3 revealed various amounts of Ag, Mo, and Bi in the interface.The area with the light color was defined to be pure BMO (P4) possibly with a trace of aluminum impurity from ball milling but all carbon has been burned off.Despite the interface reaction the effectiveness of the SLB technique in limiting the extent of this reaction is notably compared to Figure 4A printed without SLB where the extent of the interfacial reaction was significant.An EDS analysis indicated that the dark area at the interface (P2) belongs to the AgBi(MoO 4 ) 2 phase.
To further investigate the formation of new phases an XRD analysis was also performed.The XRD of both the raw BMO powder and printed BMO showed that the samples contained α-, β-, and γ-phases of BMO, which are known to have relative permittivities,  r = 19-40, 18 but the co-fired multi-material BMO-Ag samples exhibited new peaks in the XRD pattern, see Figure 5.As explained earlier, Ag reacted with MoO 3 and Bi 2 O 3 , and then formed new phases of AgBi(MoO 4 ) 2 which supports the EDS results shown above.Although the SLB process limited the interfacial reaction to a great degree, a minor reaction is inevitable during co-firing the multi-material structures, an observation also reported by Zhou et al. 14

EM Characterization of the artificial dielectrics
The EM property of the fabricated multi-materials structures composed of dielectric substrates (BMO) and metallic cuboid inclusions (Ag) was investigated.These multi-material structures allow the relative permittivity ( r ) to be controlled (e.g., by changing the size of the inclusions and spacing between them) and could be used for RF applications.Although there are different inclusion shapes can be used in the creation of such heterogeneous substrates such as spheres, disks, cubes, ellipsoids and so on, [19][20][21][22] it has been reported that cubic inclusions produce a significantly larger  r than other shapes. 23,24t is worth mentioning that adding the metallic inclusions generally gives rise to increased dielectric loss (tan) as well.Three samples were fabricated; one sample without Ag inclusions, one with 2 mm × 2 mm × 0.4 mm Ag inclusions with a 2 mm spacing (Ag1), and one with 3 mm × 2.5 mm × 0.4 mm Ag inclusions with a distance of 0.75 mm (Ag2).The Ag volume fraction for sample Ag1 and Ag2 was ≈ 4.5% and ≈ 8.5%, respectively.The fabricated samples were measured over the X-band frequency range (8-12 GHz) by using the rectangular waveguide-based transmission-reflection technique.The samples were fabricated with exterior dimensions of 22.86 mm × 10.16 mm so they could be fitted inside the WR90 rectangular waveguide cavity.The dielectric properties of a BMO-only substrate were also measured to be used as the reference value for the BMO-Ag artificial dielectric.
The X-band measurements of  r and tan for the BMO-only sample, Ag1 and Ag2 are shown in Figure 6.The addition of Ag inclusions gives rise to a more complex resonance patters as can be seen in Figure 6A,B.The measured average values of  r and tan were 17.47 and 0.012 for BMO-only, 22.29 and 0.019 for Ag1, and 34.84 and 0.020 for Ag2 as illustrated in Figure 7.As predicted by theory the relative permittivity increases with the addition of Ag inclusions and showed a 27% and 99% increase in relative permittivity over the pure BMO sample for Ag1 and Ag2, respectively.The loss tangent is also observed to increase with the addition of Ag inclusions as predicted.It is worth noting that since the ceramic samples are brittle and prone to breaking, the dimensions of the fabricated samples were slightly smaller than the waveguide cavity so the samples can be fitted into the waveguide without breaks.However, this could lead to an air gap between the samples and the cavity walls, which has previously been shown to result in lower permittivity compared to measurements performed in a TE01δ resonator, 15 but the dimensions of the artificial dielectrics did not permit the use of a TE01δ resonator.Therefore, the measured  r for the pure BMO sample is much lower than the bulk value.

CONCLUSIONS
The combination of direct ink writing and laser processing to make complete 3D products for RF/Microwave applications, for example, antennas or embedded circuit systems is a subject that both academia and industry would have interest in as it facilitates the fabrication process.A 3D printable material system based on co-fired Ag and BMO slurries has been developed and processed using SLB to create BMO substrates with Ag inclusions.The fabricated structures and the results showed that the SLB process can increase the quality of the printed multi-material structures, thereby limiting warping and interface leaching.The study showed that: • Artificial BMO-Ag dielectrics can be fabricated using direct ink writing in combination with the bespoke selective laser burn out process.The printed structures would be very difficult to make with conventional casting methods due the requirement to have a high solid loading to reduce shrinkage and cracking during co-sintering.
• The relative permittivity was increased significantly, up to 99% over the investigated frequency range (8-12 GHz) compared to the fabricated pure ceramic sample.Further increases in the metal to ceramic ratio would likely increase the relative permittivity further.

F I G U R E 3
U R E 2 (A) Illustration of the 3D printing process of the multi-material BMO and Ag structure, (B) printed green body BMO with Ag inclusions.Examples of failed prints (C) peeled off Ag layers, (D) displaced print Failed co-firing of multi-material BMO-Ag structures due to (A) interfacial chemical reaction, (B) warpage, Ag paste (C) before, and (D) after the SLB F I G U R E 4 (A) Optical microscope image, and (B) SEM images of the fracture surface of a multi-material BMO-Ag structure, (C-F) EDS analysis at P1, P2, P3, and P4 labeled in part (B), respectively, (G) EDS spectrum for P4 F I G U R E 5 X-ray diffraction patterns of raw BMO powder, sintered BMO, and BMO-Ag samples

F I G U R E 6
Relative permittivity and dielectric loss of (A) BMO sample without Ag inclusions, (B) BMO sample with 4.5% Ag inclusions (Ag1), and (C) BMO sample with 8.5% Ag inclusions (Ag2) F I G U R E 7 The relative permittivity and loss tangent for the pure sample and artificial dielectrics Ag1 and Ag2 over the 8-12 GHz frequency range