Development of Low-Dielectric Modified Polyimide with Low-Temperature Radical Curing for High-Frequency Flexible Printed Circuit Boards

Abstract

This study presents the development of a modified polyimide (MPI) with low dielectric properties and low-temperature curing capability for high-frequency flexible printed circuit boards (FPCBs). MPI was cured using dicumyl peroxide (DCP) at 80–140 °C through a radical process optimized via DSC analysis, while Fourier-transform infrared (FT-IR) confirmed the elimination of C=C bonds and the formation of imide structures. The MPI film exhibited low dielectric constants (D_k) of 1.759 at 20 GHz and 1.734 at 28 GHz, with ultra-low dissipation factors (D_f) of 0.00165 and 0.00157. High-frequency S-parameter evaluations showed an excellent performance, with S11 of −32.92 dB and S21 of approximately −1 dB. Mechanical reliability tests demonstrated a strong peel strength of 0.8–1.2 kgf/mm (IPC TM-650 2.4.8 standard) and stable electrical resistance during bending to ~6 mm radius, with full recovery after severe deformation. These results highlight MPI’s potential as a high-performance dielectric material for next-generation FPCBs, combining superior electrical performance, mechanical flexibility, and compatibility with low-temperature processing.

1. Introduction

FPCBs are essential components in next-generation high-frequency communication systems, including 5G, 6G, Internet of Things (IoT) devices, and high-speed data transmission equipment [1,2,3]. These advanced systems necessitate materials with superior electrical, thermal, and mechanical properties to enable high-density and high-speed signal transmission [4,5,6,7]. Polyimide (PI) is widely utilized as an FPCB substrate as it has several key advantages [8,9]. Firstly, PI exhibits exceptional thermal stability, maintaining structural integrity at temperatures exceeding 400 °C, which is crucial for ensuring the reliability of electronic devices. Secondly, PI possesses high mechanical strength, chemical resistance, and a low coefficient of thermal expansion (CTE), contributing to dimensional stability in high-density circuit designs. Thirdly, PI offers excellent electrical insulation properties, effectively minimizing crosstalk between signal lines [10,11,12].

However, conventional PI exhibits a relatively high D_k ≈ 3.0–4.0 and (D_f), resulting in signal attenuation and transmission losses at high frequencies [13,14,15]. To overcome these limitations, PI-based low-D_k materials have been actively developed [16,17,18,19,20,21]. Strategies to achieve a low dielectric constant include incorporating nanofillers, fluorination, and designing porous structures [22,23,24]. For instance, fluorinated PI (FPI) reduces the dielectric constant by introducing C–F bonds, which decrease molecular polarity [15,25]. Additionally, incorporating low-D_k fillers such as polytetrafluoroethylene (PTFE) or silica (SiO₂) further enhances dielectric performance [26,27,28].

To improve manufacturing efficiency, low-temperature curable polyimide (PI) has garnered significant interest. Conventional PI requires curing at 300–400 °C, which increases production costs, induces thermal stress, and limits compatibility with temperature-sensitive substrates [29,30]. To address these challenges, a modified polyimide (MPI) utilizing a radical curing mechanism was developed, enabling curing below 140 °C. This low-temperature curable MPI reduces manufacturing costs, enhances energy efficiency, and broadens compatibility with various substrates, including plastics, glass, and metal foils [31,32,33,34].

In this study, MPI was synthesized using DCP as a curing agent and processed through a three-step thermal curing procedure: 80 °C for 10 min, 110 °C for 20 min, and 140 °C for 30 min. DSC analysis confirmed the optimal curing conditions, ensuring effective polymerization and structural stability. FT-IR spectroscopy further verified successful chemical transformation by confirming the elimination of C=C bonds and the formation of imide structures, validating the curing mechanism at the molecular level.

Dielectric measurements at 20 GHz and 28 GHz revealed low D_k values of 1.759 and 1.734, respectively, with ultra-low D_f of 0.00165 and 0.00157, indicating minimal signal loss. High-frequency S-parameter evaluations demonstrated an S11 of −32.92 dB, confirming excellent impedance matching, while an S21 of approximately −1 dB validated the material’s low transmission loss characteristics.

Mechanical reliability was assessed through a 90° peel test, which exhibited an average peel strength of 0.8–1.2 kgf/mm, satisfying the IPC TM-650 2.4.8 standard and confirming strong adhesion between the MPI and the copper layer. Additionally, bending deformation tests confirmed the mechanical flexibility of MPI, maintaining stable electrical resistance down to a curvature radius of approximately 6 mm and demonstrating full recovery even after severe bending to ~3 mm. The MPI-based transmission line structure was further evaluated using vector network analyzer (VNA) measurements, which showed strong agreement with simulation results, further validating its suitability for high-frequency signal transmission.

These results confirm that MPI is a highly promising material for next-generation high-frequency FPCB applications, providing an optimal balance of low dielectric loss, robust adhesion, mechanical flexibility, and compatibility with low-temperature processing. Its superior performance meets the stringent demands of advanced communication systems, including 5G, 6G, and high-speed data transmission applications.

2. Materials and Methods

The PI resin (95 wt%) was cured using DCP (1 wt%) as a curing agent, with anisole (4 wt%) added as a solvent for ink preparation. Figure 1 illustrates the low-temperature radical curing mechanism of the low-D_k PI mixed with DCP. During the curing process, DCP decomposes, breaking the O–O bond and generating free radicals. These radicals migrate to the C=C bonds within the imide groups, initiating polymerization. The radicals propagate through the imide structure, continuously reacting with adjacent imide groups until all C=C bonds are consumed. This reaction results in an extended PI network, which exhibits not only low dielectric properties but also enhanced flexibility. This curing approach enables the fabrication of MPI without the need for additional polymers, simultaneously achieving low-temperature curing and low-D_k characteristics.

To clarify the reaction pathway in greater detail, the crosslinking mechanism underlying the radical curing of MPI is described as follows. The curing of the modified polyimide (MPI) proceeds through a thermally initiated free-radical mechanism, primarily triggered by the thermal decomposition of dicumyl peroxide (DCP). Upon heating, DCP undergoes homolytic cleavage of the O–O bond, generating cumyloxy radicals. These radicals initiate polymerization by abstracting hydrogen atoms from labile sites or adding directly to residual C=C bonds in the precursor. The macroradicals generated propagate through the polyimide network, resulting in the formation of a three-dimensional crosslinked structure. This radical-mediated propagation is illustrated schematically in Figure 1. In addition, UV exposure applied during the pre-curing stage contributes to surface-level densification and partial crosslinking. Although the UV-induced radical formation is not the dominant curing pathway, it enhances mechanical integrity at the surface and helps stabilize the film prior to full thermal curing. The combination of UV-initiated and thermally activated crosslinking ensures that MPI achieves robust mechanical and electrical properties while being processable below 150 °C.

The optimal curing conditions for MPI were determined based on DSC analysis conducted over a temperature range from room temperature to 330 °C.

FT-IR spectroscopy was employed to monitor the chemical transformation of MPI during the curing process. Spectra were recorded in the range of 600–4000 cm⁻¹ using an FT-IR spectrometer equipped with an attenuated total reflectance (ATR) accessory. Samples were prepared by drop-casting MPI ink and by using cured MPI films, enabling direct comparison between pre-cured and post-cured states. Key peaks corresponding to C=C stretching, imide C=O stretching, and C–N–C vibrations were analyzed to confirm successful crosslinking and imidization.

Based on the DSC results, the volatile solvent anisole in the MPI was evaporated. To minimize surface void formation during evaporation, the drying process was conducted in two stages: 80 °C for 10 min and 110 °C for 20 min. To enhance the surface durability of the dried film, a UV exposure process was performed. The film was then pre-cured at 140 °C for 30 min to achieve a B-stage state, facilitating subsequent lamination with low-profile copper foils for CCL fabrication. The final curing profile for the MPI dielectric layer during vacuum lamination was applied as shown in Figure 2f.

Figure 2 presents a schematic of the MPI film fabrication process for dielectric property measurements. The process begins with coating a 100 µm thick layer of MPI using a bar coater, followed by leveling at room temperature for 10 min (Figure 2a). The coated film then undergoes a drying process at 80 °C for 10 min and 110 °C for 20 min, during which most of the solvent evaporates (Figure 2b). To enhance film durability, UV exposure is performed at an energy of 14 mJ for 70 s (Figure 2c). This step facilitates surface densification and induces partial crosslinking through UV-initiated radical reactions, improving mechanical robustness and enabling stable film handling prior to full thermal curing. The film is subsequently pre-cured at 140 °C for 30 min (Figure 2d). Finally, a vacuum lamination process is conducted to evaluate its applicability for copper-clad laminate (CCL) fabrication (Figure 2e).

Figure 3 displays images of the 100 µm thick MPI film connected to a VNA and the coaxial resonators used for dielectric property measurements at 20 GHz and 28 GHz. The D_k and D_f of the MPI-based CCL transmission lines were measured within the 20–30 GHz (Ka-band) frequency range to validate the simulation results and assess material performance.

3. Fabrication and Design of Transmission Line

To effectively utilize the measured dielectric properties, the structure was designed based on the thickness of the MPI, as shown in Figure 4. For accurate simulation that matches the actual sample conditions, copper (Cu) pads for ground–signal–ground (GSG) probes were fabricated on both sides of the signal line. Vias were incorporated to interconnect the ground plane and the Cu pads, with each via having a diameter of 100 µm, achieved through UV laser drilling. The vias were arranged in a 2 × 2 configuration per pad to ensure reliable electrical connectivity and signal integrity.

The fabrication process flow of the designed transmission line is illustrated in Figure 5. A 100 µm thick MPI layer was laminated with 11 µm low-profile copper foils on both sides to form a CCL (Figure 5a). To minimize signal loss and enhance transmission efficiency at high frequencies, low-profile copper was used to reduce surface roughness, as the current tends to flow closer to the surface due to a reduced skin depth [35,36].

UV laser drilling was employed to create vias in the CCL (Figure 5b). Any smear generated during the drilling process was removed using an O₂ plasma ashing process (Figure 5c). The vias were then metallized through electroless plating (0.4 µm), followed by electro-plating (10 µm) to ensure proper electrical connection between the ground line and vias (Figure 5d).

For patterning, a 40 µm dry film resist (DFR) was laminated at 110 °C using a roll lamination process (Figure 5e). The pattern was defined by using an i-line UV lamp for exposure (Figure 5f). The exposed resist was developed (Figure 5g), and unwanted copper was removed through an etching process (Figure 5h). Finally, the remaining resist was stripped, completing the fabrication process (Figure 5i).

The mechanical properties of the MPI-based CCL were evaluated by conducting peel strength and bending deformation tests.

The peel strength was measured according to the IPC TM-650 2.4.8 standard by applying a constant force to detach the copper foil from the MPI layer. A tensile testing machine was used with a peel angle of 90°, and the peel resistance was recorded as a function of extension length to assess the interfacial adhesion between the copper electrode and the MPI dielectric layer.

To evaluate the mechanical flexibility of the MPI-based CCL, a bending deformation test was conducted. Samples measuring 12 cm × 12 cm were prepared and subjected to bending with various radii of curvature, ranging from approximately 20 mm to 3 mm. The electrical resistance of the samples was measured in real time using a four-probe method while progressively increasing the bending strain. The change in resistance (R/R₀) was recorded to assess the electrical stability under mechanical deformation. Photographic documentation of the bending states was captured to visually illustrate the deformation levels.

4. Results and Discussion

Figure 6a presents the DSC analysis conducted to determine the curing conditions for MPI. The analysis was performed by heating the sample from room temperature to 330 °C at a rate of 10 °C per minute. A broad exothermic reaction was observed around 150 °C, corresponding to the curing reaction between the PI and DCP, resulting in a distinct exothermic peak.

A secondary exothermic reaction appeared near 250 °C, indicating the curing of incompletely reacted regions within the material. This reaction is associated with crystallization, which enhances the strength and hardness of the PI while improving its thermal stability. The presence of this additional exothermic peak confirms the completion of the curing process and the formation of a robust PI network structure.

Further chemical insights are provided by the FT-IR spectra in Figure 6c,d. When comparing the MPI ink (black curve) and the cured film (red curve), it can be seen that the characteristic C=C stretching vibration at around 1640 cm⁻¹, which represents unsaturated bonds in the precursor, significantly diminished after curing. Concurrently, the absorption peaks of imide C=O stretching (~1720 cm⁻¹) and the C–N–C stretching (~1375 cm⁻¹) intensified, confirming the formation of the imide structure and the progression of crosslinking.

Additionally, the high wavenumber region (Figure 6d) shows noticeable changes in C–H stretching vibrations (~2900 cm⁻¹), indicating the densification of the polymer network and completion of the curing process. These complementary results from DSC and FT-IR analyses confirm the successful low-temperature radical curing of MPI, which is essential for achieving targeted dielectric and mechanical properties.

The dielectric properties of the fully cured MPI film were measured at 20 GHz and 28 GHz, as shown in Figure 6b. The D_k values were 1.759 and 1.734, respectively, while the D_f values were 0.00165 and 0.00157.

The structure shown in Figure 4 was used for simulation in HFSS, incorporating the measured dielectric constant and dissipation factor at 20 GHz. The simulated S-parameters were compared with the measured S-parameters of the fabricated transmission line, as shown in Figure 5. Resonant frequencies corresponding to different signal line lengths were identified, and the experimental results closely aligned with the theoretical and simulated values. Measurements were performed over the 20–30 GHz frequency range, and the resonant frequencies for each mode order n are presented in

Table 1. Comparison of resonant frequency theory and simulation.

Resonant Mode Order n	Theory	Simulation
9	21.22 GHz	21.2212 GHz
10	23.34 GHz	23.3433 GHz
11	25.47 GHz	25.4855 GHz
12	27.60 GHz	27.5976 GHz
13	29.72 GHz	29.7297 GHz

.

The theoretical and simulation results exhibited excellent agreement, with discrepancies remaining below 0.005% for all resonant mode orders. This high accuracy confirms the precise design of the signal line length. Resonance was verified to occur when the signal line length (L) corresponded to an integer multiple of the wavelength (λ). The signal velocity, c = 3 × 10⁸ m/s, and the wavelength-based resonant frequency calculations fell within the expected range for the 20–30 GHz frequency band. The resonant mode order n was confirmed to range from 9 to 13.

The signal velocity equation is given by [37]

$$\lambda =\frac{C}{f\cdot \sqrt{{\epsilon }_r}}$$

where c = 3 × 10⁸ m/s, $f$ is the frequency (20–30 GHz), and ${\epsilon }_r$= 1.759.

The resonance condition is expressed as [38,39]

$$L={\displaystyle \frac{n\cdot \lambda }{2}}$$

Resonance occurs when the signal line length $L$ is an integer multiple of the signal wavelength $\lambda$ (typically $\lambda /2$ or $\lambda /4)$. At these frequencies, the voltage or current reaches its maximum, resulting in resonance.

Figure 7a shows the VNA used to measure S-parameters, while Figure 7b displays the fabricated transmission line structure corresponding to the design in Figure 4. Figure 7c,d compare the measured and simulated S-parameters, demonstrating a high degree of correlation. The S11 results exhibit matching dips at resonant frequencies, with the simulated data (black curve) and measured data (red curve) closely aligning. The measured return loss reaches a minimum of −32.92 dB, indicating excellent impedance matching the corresponding frequencies [40]. The periodic nature of the dips reflects the frequency-selective characteristics of the structure, which are consistent with the theoretical design for resonant modes. Although minor deviations can be observed at certain frequencies, they can be attributed to fabrication imperfections or measurement errors. The strong correlation between the simulation and experimental data validates the proposed design [41,42,43,44], confirming the effective minimization of reflection losses within the evaluated frequency range.

Figure 7d presents the S21 characteristics, comparing simulated and measured results. The data show a consistent performance across the entire 20–30 GHz frequency range, with the measured insertion loss remaining around −1.0 dB, confirming the low-loss nature of the designed transmission line. Minor fluctuations and deviations may result from fabrication variability or measurement inaccuracies but do not significantly impact overall performance. These results demonstrate that the fabricated structure reliably supports signal transmission with low loss, validating the accuracy of the simulation predictions. The transmission line’s performance confirms its suitability for high-frequency applications in the 20–30 GHz range.

The mechanical flexibility of the MPI-based CCL was further evaluated through a bending deformation test, as presented in Figure 8. The electrical resistance remained stable under bending conditions down to a radius of curvature of ~6 mm, and exhibited recoverable behavior even after severe bending to ~3 mm. This demonstrates that the MPI film possesses excellent mechanical robustness, which is critical for flexible printed circuit applications.

To evaluate the durability of MPI, whose excellent dielectric properties were confirmed through measurements of the dielectric constant, dissipation factor, and S-parameters, a 90° peel test was conducted to measure the peel strength between the copper electrode layer and the MPI dielectric layer. The peel strength was assessed following the IPC TM-650 2.4.8 standard [45], which specifies a quantitative method for determining the bond strength between metallic foils and substrates in printed circuit boards, including FPCBs.

Figure 9 shows the peel resistance results of the MPI-based CCL, obtained from the 90° peel test. The peel strength graph illustrates the variation in peel resistance with mechanical extension. An initial sharp increase in peel resistance was observed, followed by stabilization after approximately 5 mm of extension. The average peel strength ranged from 0.8 kgf/mm to 1.2 kgf/mm, meeting the adhesion strength requirements specified by IPC TM-650 2.4.8. These results indicate a consistent bonding performance between the conductor and dielectric material, demonstrating the material’s durability and reliability.

Minor fluctuations in peel resistance were attributed to slight inconsistencies in bonding or localized defects. However, the overall performance remained stable, confirming that the MPI material provides reliable adhesion and mechanical strength. Therefore, the developed MPI is suitable for high-reliability FPCB applications, meeting the demands of next-generation electronic packaging and high-frequency communication systems.

5. Conclusions

In this study, we successfully synthesized and comprehensively characterized a modified polyimide with a low dielectric constant and low-temperature curing capability, specifically designed for high-frequency flexible printed circuit boards. Using dicumyl peroxide as the curing agent, MPI was processed through a radical curing mechanism, with optimal curing conditions determined via DSC analysis. The fabrication process included bar coating, UV exposure, and vacuum lamination, resulting in a structure suitable for copper-clad laminate applications.

Dielectric property measurements at 20 GHz and 28 GHz confirmed the superior electrical performance of MPI, exhibiting dielectric constants of 1.759 and 1.734, respectively, and ultra-low dissipation factors of 0.00165 and 0.00157. These values ensure minimal signal attenuation, positioning MPI as a strong candidate for advanced high-frequency communication applications. Furthermore, high-frequency S-parameter simulations and experimental validations demonstrated excellent agreement, with a return loss of −32.92 dB and an insertion loss of approximately −1 dB, confirming reliable impedance matching and low transmission loss.

Mechanical robustness was also thoroughly validated. The bending deformation test demonstrated that the electrical resistance of the MPI-based CCL remained stable down to a curvature radius of approximately 6 mm, with full recovery even after severe bending to around 3 mm. This confirms the excellent flexibility and mechanical durability of MPI, critical for reliable performance in dynamic FPCB applications. Additionally, the 90° peel test, conducted in accordance with the IPC TM-650 2.4.8 standard, confirmed strong adhesion between the MPI dielectric layer and the copper electrode, achieving peel strengths ranging from 0.8 to 1.2 kgf/mm.

Collectively, these findings confirm that MPI offers an optimal balance of electrical performance, mechanical flexibility, and adhesion strength, while being compatible with low-temperature, energy-efficient manufacturing processes. Thus, MPI is a highly promising dielectric material for next-generation FPCBs, well-suited for 5G, 6G, and high-speed data transmission systems, as well as emerging advanced packaging technologies.