Inkjet printing is a versatile and cost-effective process for creating a wide range of electronic devices [1–5]. In particular, this process has facilitated the production of low-cost, large-area, and flexible electronic devices [6]. One of its greatest advantages is that it offers the ability to precisely position materials on both planar and non-planar substrates, which enables the fabrication of microstructures without the need for photolithography [7]. Moreover, the thickness of the structures can be continuously controlled using a multi-printing technique [8]. However, inkjet-printed structures often exhibit surface cavities and undulated surface morphology due to the coffee ring effect, which occurs during the drying and sintering processes after inkjet printing [9–13].

In our previous study [11], we reported on the development of layer-by-layer printing and low-temperature drying techniques to obtain silver (Ag) conductive lines without undulated morphology for use in electrodes for displays. Nonetheless, the printed Ag line, which had a thickness of less than 0.6 μm and a width of 18 μm, exhibited high electrical resistivity owing to the formation of surface cavities after two overprintings.

Transmission lines for radio frequency (RF) applications require thicker conductor lines than those used for display electrodes to ensure reduced conductor loss at high frequencies. Moreover, in coplanar waveguide (CPW) lines, the width of the signal lines and the gap between the signal lines and the ground are designed to achieve a characteristic impedance of 50 ohms, which is suitable for measurement setups such as a Sub- Miniature version A (SMA) connector or a ground-signal-ground (GSG) probe. Therefore, using previously developed printing technology [11], we investigated the effect of drying temperature on the formation of surface cavities and geometrical changes in inkjet-printed transmission lines (CPW) up to 3 GHz [12]. Our findings confirmed that as the drying temperature increases, the surface roughness and insertion loss of the transmission line tend to decrease.

In this paper, we conduct a more in-depth study by implementing atomic force microscopy (AFM) and cross-sectional profile analysis to examine the effects of drying temperature before the sintering process, using Ag inkjet printing to ascertain the relationships among surface roughness, surface cavities, morphological changes, and insertion loss. Additionally, we propose a new printing method that combines low-temperature and high-temperature drying to improve morphology and achieve fewer surface cavities.

We designed a CPW with a characteristic impedance of 50 Ω, operating up to 3 GHz, using a 0.5-mm-thick glass substrate (ɛ_r = 5.4, tanδ = 0.003). As shown in Fig. 1, the signal lines width and gap were determined to be 180 μm and 30 μm, respectively, through calculation and simulation. The length of the CPW was set to 2 mm. Meanwhile, the thickness of the signal lines was designed to be 3 μm, since it had to be greater than or equal to 3 μm to avoid conductor losses at 3 GHz due to the skin effect [14].

Fig. 2 illustrates the fabrication process of the CPW. First, a positive photoresist was patterned on the glass substrate (EAGLE XG; Corning Inc., Corning, NY, USA) based on the CPW’s geometric design by carrying out a conventional photolithographic process (Fig. 2(a)). Next, a non-wetting layer was spin-coated with a fluorocarbon (FC) solution, which was prepared by mixing FC-722 (Fluorad; 3M, Saint Paul, MN, USA) and FC-40 (Fluorinert; 3M) at a ratio of 1:4 (Fig. 2(b)), after which the FC pattern was formed by performing a lift-off process using acetone solution (Fig. 2(c)). Ag ink was then printed into the FC pattern while controlling the jetting pitch (10 μm) and the number of printings (Fig. 2(d)). Notably, the Ag ink used in this study is a commercial Ag ink (DGP 40LT-15C; ANP Inc., Seongnam, Korea) containing Ag nanoparticles in triethylene glycol monoethyl ether. The viscosity and surface tension of the ink were 12.8 mPa·s and 35.9 mN/m, respectively. We employed a DMP2800 inkjet printer (Fujifilm Dimatix Inc., Santa Clara, CA, USA) and a 2.4 pL inkjet head (Samba Cartridge; Fujifilm Dimatix Inc.) for the inkjet-printing process. After printing, the ink drying process was carried out at 45°C for 180 minutes and then at 100°C for 30 minutes in a convection oven. Finally, the dried Ag inkjet patterns were sintered at 150°C for 30 minutes (Fig. 2(e)). Since the FC pattern formed a non-wetting barrier, it inhibited the spreading of the Ag ink, allowing it to build up inside the FC pattern during the inkjet-printing process. This implies that the thickness of the Ag pattern can be increased by repeating the number of printings. Subsequently, the FC patterns were removed using O₂ plasma (Fig. 2(f)).

We prepared four samples characterized by different drying temperatures and number of printings. The printed surface and morphology were measured using an optical microscope, an atomic force microscope, and a surface profiler. Fig. 3 depicts the optical images of the four samples. L1 refers to the sample printed once and dried at a low temperature (45°C), H1 indicates the sample printed once and dried at a high temperature (100°C), LL denotes low-temperature drying (45°C) after first printing and low-temperature drying (45°C) again after second printing, while LH refers to low-temperature drying after first printing (45°C) and high-temperature drying (100°C) after second printing.

The optical images indicate that the samples dried at a low temperature have considerably more microcavities than those dried at a high temperature. AFM images of the four samples are depicted in Fig. 4, with R_a, R_q, R_p, and R_v denoting the average roughness, root mean square (RMS) of roughness, maximum peak height, and valley depth of surface roughness, respectively. R_a, R_q, and R_p for the four samples showed similar values, regardless of the number of printing passes and drying temperature. However, numerous microcavities were observed in the L1 and LL samples, exhibiting a three-fold increase in their R_v compared to those of H1 and LH.

A surface profiler was employed to measure the surface morphology (cross-section view) of each sample. The fabricated samples (H1, LH) and the measured cross-section view (A–A′) of the four samples are presented in Fig. 5. Fig. 5(b) shows that the samples dried at a high temperature (H1, LH) exhibit a more rabbit-ear geometry compared to the L1 and LL samples. Notably, the LL and LH samples were thicker than the L1 and H1 samples, especially because they were printed twice consecutively.

The rabbit-ear morphology was formed by the accumulation of Ag nanoparticles at the edges of the printed Ag lines, which occurred due to the coffee ring effect during the drying process. Generally, capillary flow and Marangoni flow are known to contribute to the coffee ring effect in deposited ink droplets. Capillary flow occurs due to the presence of an evaporation gradient across the droplet surface, which causes the liquid to flow from the center to the edge, compelling the particles to accumulate in the edge region [15]. Marangoni flow refers to a recirculating flow toward the center inside the droplet that continuously supplies particles to the edge region. Marangoni flow occurs under surface tension gradients generated by changes in chemical composition or temperature along the free liquid surface [16].

The insertion loss of the CPW was simulated over the frequency range of 300 kHz to 3 GHz using the High-Frequency Structure Simulator (HFSS; ANSYS Inc., Canonsburg, PA, USA), and then measured using a vector network analyzer (VNA). Fig. 6 presents the on-wafer measurement setup, where a pair of GSG probes contact the CPW surface and are connected to a VNA.

As shown in Fig. 7, the simulated insertion loss was 0.054 dB at 3 GHz, while the simulated return loss remained greater than 35 dB across the frequency range. A GSG probe connected to the port of an E5061B VNA (Agilent Technologies Co., Santa Clara, CA, USA) was calibrated via short-open-load-through (SOLT) calibration over the frequency range from 300 kHz to 3 GHz. The GSG probe maintained contact with the GSG of the printed CPW, while the VNA measured the insertion loss at room temperature. The measured insertion losses of the L1, H1, LL, and LH samples were 0.128 dB, 0.167 dB, 0.092 dB, and 0.082 dB at 3 GHz, respectively. In most cases, the measured return loss was greater than approximately 30 dB across the entire frequency range. Among the four samples, H1 and LH exhibited the highest and lowest insertion losses, respectively.

Fig. 8 depicts the transmission-line model used to analyze the differences in insertion loss among the samples (L1, LL, H1, and LH), considering conductor loss and parasitic capacitance. Here, R denotes the additional resistance from the thinner signal lines caused by high-temperature drying, which increases conductor loss at a high frequency. Meanwhile, C refers to the shunt capacitance from the edge wall located between the printed signal lines and the ground. The additional R and C in the sample that was printed once and underwent high-temperature drying (H1) resulted in a degradation of the insertion loss up to 3 GHz. Notably, the simulated and measured data up to 3 GHz were compared using ANSYS Electronics Desktop Passive (AEDT) RF Circuits to validate the proposed equivalent circuit model. The simulated results demonstrated that the differential resistance and capacitance values observed in the H1 and LH samples correlated with their measured insertion loss trends observed up to 3 GHz. Furthermore, circuit simulation analysis using AEDT revealed that the LH sample exhibited approximately 0.13 Ω lower series resistance and 50 fF reduced shunt capacitance relative to the H1 sample.

Our investigation into the ways in which microcavities and roughness degrade the insertion loss at a high frequency [12] revealed that these factors have only minor effects under 30 GHz, with surface morphology after printing being a more critical factor. Since the wavelength in free space at 3 GHz was 10 cm, the microscale cavities were too small compared to the wavelength. We also calculated the roughness factor based on impedance calculations [17–19]. In this context, the important factor that had to be measured was the ratio of skin depth to the RMS value of roughness. The skin depth of the printed Ag transmission line at 3 GHz was calculated to be around 3 μm (the conductivity of the printed Ag ink was 5 × 10⁶ S/m [11]) and the RMS value of the roughness was less than 20 nm (Fig. 4). It is evident that the roughness value of the printed transmission line is too small compared to the skin depth, meaning that attenuation due to roughness is nearly zero. This also implies that roughness would affect attenuation loss only from approximately 30 GHz [17–19].

The RF characteristics of CPW transmission lines printed on both rigid and flexible substrates have already been investigated in several studies [20–23]. These CPW lines fabricated using Ag nanoparticle ink or Ag paste have been electrically characterized up to 40 GHz. Table 1 summarizes the insertion loss, surface roughness, and thickness at 3 GHz achieved by various CPW printing methods, including the method proposed in this study. Insertion losses achieved by the screen-printed CPW lines were approximately 0.03 dB/mm at a thickness of 10 μm, whereas inkjet- or aerosol-jet-printed CPW lines exhibited an insertion loss of approximately 0.3 dB/mm at a thickness of approximately 1 μm. However, these prior studies did not examine the correlation between surface roughness, surface porosity, morphological change, and insertion loss based on the drying temperature, which is the primary focus of this study. Overall, the comparative results in Table 1 confirm that the proposed printing and drying method significantly lowers the insertion loss of inkjet-printed CPWs by optimizing printing thickness.

Theoretically, the insertion losses of LL and LH exceeded those of L1 and H1, because the thickness of the L1 and H1 samples did not match that of the skin depth. In H1, both the central region of the signal line and the ground edges adjacent to the signal line were too thin. Consequently, its insertion loss was higher than that of L1, mainly due to the increased conductor loss in the signal line and the additional parasitic capacitance between the signal and ground regions. Additionally, the sidewall morphology of the printed CPW plays a critical role in determining RF performance. The sidewalls of both the signal and ground conductors exhibited a rabbit-ear shape, whose thickness and width were dependent on the drying temperature. The combination of low- and high-temperature drying in the LH resulted in a broader and smoother rabbit-ear profile with fewer surface microcavities. In contrast, the LL, which was dried twice at a low temperature, exhibited a relatively flat and uniform morphology but contained larger microcavities due to slow solvent evaporation during the low-temperature drying process. Consequently, the insertion loss of LH was slightly better than that of LL, despite the morphology of the center signal lines being better in the case of LL (less of the rabbit-ear shape). We also found that a thicker and broader rabbit-ear shape formed at the edges of both the signal line and ground contributed to a slight improvement in insertion loss. This improvement is attributed to the concentration of surface current density on the facing surfaces of the signal line and ground conductors. The surface current distribution along the edge sidewalls of the signal line was analyzed using HFSS, as shown in Fig. 9. It was checked that the LH shows a higher surface current density than the LL.

Although it cannot be generalized that the side wall effect between the facing surfaces of the signal lines and ground in the CPW results in improved insertion loss, the simulation and measurement results obtained from the design and fabrication proposed in this study support this argument. Nonetheless, these results must be verified both theoretically and experimentally through further in-depth research.

The results of this study imply that controlling the drying temperature between the steps involved in inkjet printing can facilitate the use of microwave devices with low loss and a well-defined structure. We found that drying process at a low temperature created more surface cavities on the printed surface. It was also observed that the difference between the maximum and minimum heights of the printed structure was smaller than that of the samples dried at a high temperature. Furthermore, the printed structures dried at a high temperature had fewer microcavities but exhibited very serious rabbit-ear morphology resulting from the coffee ring effect. This implies that a trade-off must be made between microcavities and the surface morphology of the printed transmission line structure based on the drying temperature during the inkjet printing process. Notably, microcavities are more critical in thin electrodes and microwave devices operating over 30 GHz, while surface morphology issues caused by the coffee ring effect are more critical in inkjet-printed RF transmission lines. Furthermore, we found that a thicker and broader printed region formed along the sidewalls of the signal and ground lines in the inkjet-printed CPW, achieved by controlling the printing steps and drying temperature, reduced the insertion loss up to 3 GHz and resulted in fewer microcavities on the surface.

We believe that the proposed method offers a basic and practical tool for inkjet printing applications in microwave devices that require low loss levels and mechanical structures that require fewer microcavities and a relatively smooth morphology.