Open Access
Add to BibSonomy
Abstract
Laser micromachining is the chosen method for vertical interconnect access point (VIA) formation in flex PCB layers. Even so, this method suffers from several inherent physical issues as a result of the intense localized heating causing strong Marangoni convection and the buildup of recast along the VIA upper crater walls while also scattered particle debris and oxidation of copper across the surface. The mitigation of the height and radius of this recast layer is critical for the following build-up process and device functionality and reliability. This is currently a major technology inhibitor to the adoption of flex PCBs for high-power electronics. In this study, we present experimental results showing the use of engineered sacrificial layers that coat the surface of the flex PCB substrate during the laser micromachining process. Optimization of this engineered sacrificial layer resulted in a major improvement in recast quality and debris control as well as reducing the oxide formation while increasing the laser drilling efficiency, attributable to increased surface pressure on the substrate. In this paper, we describe the methods and materials used in the development of sacrificial layers and show the positive impact it achieves on improving and modifying the plasma characteristics throughout the overall laser drilling process.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
The value of Flex PCBs (FPCB) lies in the capability to construct a complex network of VIAs over multiple layers and by so adding important functionality in compact form factors. The increased penetration of Flex PCBs in consumer electronics is a major area of growth as in-depth market analysis show for mobile devices [1], as well as for the PCB market as a whole [2], however more vigorous applications such as power electronics and automotive are seeing a major increase as well [3]. These applications come with much higher requirements regarding thermal cycling and the main risk of failure increases due to delamination effects in the layered substrates. Since for these applications, device failure may result in real-life personal injury, improving the robustness of the PCBs to follow higher standards is of utmost importance.
VIA formation in Flex PCB substrates is done with UV nanosecond lasers. The main reasons for this choice are the good quality of copper penetration due to high absorption efficiency, resulting in low heat-affected zones (HAZ) demonstrated in the whitepaper [4,5], and long depth of focus relative to the spot size as compared to longer. Additional optimization of the temporal pulsewidth within this regime was previously shown to further benefit the ablation process as described in [6]. The process consists of locally irradiating the substrate by a high-intensity pulsed beam, causing a phase explosion and removing the target material in the forms of melt, and vaporization referred to as laser ablation. Throughout this process, the absorption of extremely high laser intensities up to ${\sim} 1\frac{{GW}}{{c{m^2}}}$ results in the melting and vaporization of the copper and polyimide substrate reaching localized temperatures greatly exceeding the vaporization temperature of copper ${\sim} 3,000{\; ^o}C\; $, where [7] produced measurements of the plasma temperature in a similar laser ablation process. While it is possible to ablate the copper and polyimide layers at lower intensities as investigated by [8] who also showed the advantages of this with less deformation with a more photochemical process, this decreases the achievable application throughput which is undesirable for cost reasons. This high thermal load and mixture of vaporized and molten material cause a strong surface tension driven Marangoni flow as described in [9] resulting in recast formation along the outer crater rim for various cases as seen in [10] who also related this effect to the characteristics of the resulting plasma. This recast formation can directly result in device failure due to either fatigue performance or delamination.
An added direct result of the laser ablation process on the copper surface due to the high temperatures is in the ring formation of the cupric oxide layer. While carefully grown $CuO$ layers of $< 50\; nm$ have been applied in the past to improve adhesion between PCB layers, oxidation of copper metal at the elevated temperatures reached in the copper plasma/melt led to a decrease in the adhesion strength resulting in delamination as described by [11].
In this study, we have introduced a sacrificial layer to mitigate the recast formation and minimize the HAZ around the site all while not hindering application throughput in terms of time and energy efficiency. The use of a sacrificial layer has been done before in the field of laser micromachining, notable with the use of $\textrm{C}{\textrm{O}_2}$ laser drilling for the compound effect of increasing the process resolution while mitigating undesired heating effects as shown by [12]. The use of coating layers has also been shown and patented for processes such as micromachining thin glass substrates statistically analyzed in the patent [13], shown to minimize fracturing as well as improve the overall VIA geometry. Predating this patented study, [14] describes the general advantage of introducing a sacrificial layer above any substrate, while the patent by [15] relates the advantages of laser drilling for PCB interconnects. In the field of VIA formation for Flex PCB, process throughput is a key factor as PCB designs become increasingly complex with increased VIA densities and an overall increase in demand. In this study, we have developed a sacrificial layer that is easily and rapidly applicable and removable, while investigating the effects on the VIA formation process from a physical perspective from the plasma emission signals and comparing this to the 3D measurements of the VIA profile. This helps to provide insight into the desired material’s physical and chemical characteristics for optimizing the sacrificial layer for the process while also optimizing throughput.
2. Materials and methods
2.1 Sacrificial layer material
Polyvinyl Butyral (PVB) is an amorphous thermoplastic polymer obtained by a condensation reaction of polyvinyl alcohol and butyraldehyde. The resin is renowned for its high flexibility, film-forming, and good adhesion properties. PVB is mostly known for its use in laminated glass, retaining high levels of adhesion even under high stress described in [16]. The material also showed good optical transmission (>98%) in the UV-VIS-NIR ($330 - 900nm)$ spectrum range. PVB has a low thermal conductivity and therefore can also be used as a good insulator, localizing the heat-affected zone.
Optimization of the sacrificial layer was performed to achieve a uniform thin layer. Mowital B 60 HH resin (by Kuraray) was dissolved in Isopropanol (IPA) by sonication. This material allows for easy removal by simply washing off with IPA. The thickness of the resulting polymer was controlled by altering the concentration of the polymer powder and the applicator roll-spacer. An optimized concentration of 11% and a 30um gap applicator resulting in a total thickness of ∼1.95 µm produced the best results in regards to full process ablation as elaborated further. This thin layer design allows for alternative application and removal procedures that can prove more attractive for production procedures such as spray-on/spray-off. Further development of this process is planned in future studies.
2.2 Flex PCB substrate
The target substrate used in this study consisted of three layers, typical for FPCB rolls, a top layer of 12 µm electroplated copper (see Fig. 1), a middle layer of 25 µm of polyimide, and a bottom of 12 µm electroplated copper. The general process of VIA formation is done by penetration of the top copper layer and ablation of the underlying polyimide while stopping on the bottom copper layer. Requirements of the process include minimal penetration of bottom copper layer <2 µm, steep taper $\frac{{{\emptyset _{bottom}}}}{{\emptyset top}} > 0.7$, minimal recast formation and HAZ to not interfere with next processes of layer-up including any material formation within the VIA.
2.3 Experimental setup
Laser drilling was carried out with an experimental setup for laser micromachining. From the laser source (Pulseo by Spectra-Physics), the beam propagates through a beam expander followed by galvo-scanning mirrors (excelliScan by Scanlab) and a 175 mm F-theta scanning lens (by Sill Optics) resulting in a measured beam diameter of 15 µm at focus, see Fig. 1. Sets of lines were scanned to compare quality and provide more accurate measurements for data analysis. VIA formation was done using the trepanning method common for producing large VIAs with small laser spot sizes. Each set was drilled while parameterizing the pulse energy, number of repetitions, and scanning velocity.
Fig. 1. Left: Laser drilling experimental setup with beam expander, galvo scanning mirrors, and f-theta lens. Right: Sample of FLEX PCB buildup 12:25:12(µm)-Cu:PI:Cu.
3. Results
Multiple tests were carried out to optimize the process, and the optimal working points demonstrate a clear advantage when using the PVB sacrificial layers. A significant response to the process that is immediately evident is the radius of the heat-affected zone causing discoloration on the surface as well as recast and debris along the edges of the scanned line. Figure 2 presents the comparison of both cases spanning over several parameter settings such as scanning velocity (100-500 mm/s), pulse energy (24-194µJ), and repetitions (1-3). In Fig. 2(a) the discoloration due to local oxidation on the copper surface forms up to several tens of microns around the scanned area for the reference case while being confined to the immediate area of the scan for the PVB coated sample. While this optical discoloration may seem irrelevant it is precisely this effect that may cause delamination in the buildup process as the adhesion properties are affected by the variation in oxidation levels. Figure 2(b-d) shows SEM images, measured 3D profile (by a Bruker Contour GT-X), and analyzed data of the solidified recast created during the ablation process. Again, the PVB layer manages to localize the recast effect as well as capture the ‘splattering’ of copper particles on the surface. This reduces the chance of void formation during later steps in the build-up process. The distribution of all data points along with the various parameter settings, in Fig. 2(d) shows the broad distribution and large recast widths of the reference sample over all cases. The PVB coating skews the distribution to reach lower recast widths, again resembling the confinement of the process with less damage around the area overall. Important to note is that while the copper particles remain and are only deposited on the PVB surface which is removed, the measurement shows that also the recast formation is minimized in the case of ablation with the PVB layer. All of these advantages for producing cleaner laser processes are compounded when the efficiency also shows a preference for the PVB coating layer. The slight statistical significance in favor of using the PVB layer to achieve higher efficiency, notably for parameter combinations that produce deeper trenches (higher energy and lower velocities) is substantial in demonstrating a cleaning process that can also increase throughput. The difference in the distributions across a span of laser scanning parameters proves the significance of the improvement with the PVB layer overall.
Fig. 2. (a) Microscope images of scan lines, Top: Reference without a sacrificial layer, Bottom: Drilling with 1.95 µm PVB layer, imaged after removal of the layer in isopropanol. Settings were: E = 194µJ, 3 Reps, 100 mm/s velocity. (b) SEM images of scan lines from (a), Top: Reference, Bottom: with PVB layer. (c) Sample from 3D optical profiler measurement. (d) Distribution of recast radius and trench depth measured with a 3D optical profiler. Box length shows the interquartile range with upper and lower fences showing the total range, the centerline indicates the median while the 95% confidence interval centers around the average. Kernel density is also shown relaying the distribution of the data points. Data includes various parameter settings of pulse energy and scanning velocity and shows the comparison of all points related. Data available in Dataset 1 Ref. [17].
For VIA formation, methods such as trepanning the beam are used for producing high-quality VIAs where large diameters are required. This process allows for much higher taper ratios (steeper sidewalls) to be achieved. This process utilizes a spot size of 12 µm while scanning the beam within the circumference of the 100 µm diameter, see Fig. 3(a). The result of trepanning motion, imaged in Fig. 3(b) and (c), shows a debris reduction, and recast demonstrates an even greater advantage for using the PVB layer since for the same factor settings a full ‘pop-out’ is achieved, requiring no additional post-processing. This unique phenomenon of ‘popping out’ the inner Cu and PI pillar occurs when a significant thermal load accumulates due to high pressures within the center of the VIA being formed. This makes a very efficient process since no post-processing is required to clean the inner area. However, specific conditions that produce this effect show strong correlations such as requiring a fast enough scanning speed to complete the rotation before the thermal load dissipates in the substrate while not too high as to limit the effective fluence $\left( {\frac{J}{{c{m^2}}}} \right)$ due to the large pitch between consecutive pulses. This directly benefits from an increase in the pulse repetition rate (PRR), minimizing the pitch while allowing for higher velocity scanning speeds. Issues arise as high PRRs for these types of lasers result in a much lower THG (Third Harmonic Generation) conversion efficiency, thus lowering the pulse energy with a nonlinear relation, and decreasing the output power drastically. Figure 4 helps us understand this relation by defining the more efficient process in terms of throughput.
Fig. 3. Sample comparison of trepanned VIAs at identical parameter settings, 300 mm/s velocity, 1 rep, 100µJ pulse energy. (a) Diagram showing trepan laser drilling method using a small spot size for drilling a 100 µm diameter VIA. (b) VIA drilled with PVB layer, (c) VIA drilled without sacrificial layer. The scale bar in (b) is consistent for all images in the figure.
Fig. 4. Analysis of specific ablation rate and drilled trench depth to scanning velocity for two pulse repetition rates. Error bar is standard error over multiple samples and measurements. The pulse energy was held at $25{\mathrm{\mu}} J$ by adjusting for total output power at both rates. Dotted lines are depth measurements of trenches, solid lines are the specific ablation rate. Measurements were done using a 3D optical profiler.
We can see how the depth of the trenches decreases for both PRRs as a result of the pulse pitch increasing and the effective fluence per unit area decreasing near linearly. Even so, the higher PRR manages to maintain a preference as the pitch is the ratio of $\frac{{Velocity}}{{PRR}}$. The specific ablation rate has been observed to help determine the optimized factor settings for efficiency and high throughput ablation processes, see [20]. Observing the results in Fig. 4 shows the significance of achieved throughput when increasing the velocity and PRR. Combining this with the use of thin PVB sacrificial layers can prove its value in terms of increased throughput while also improving VIA quality.
4. Analytical analysis
We attribute the effects observed by referencing from several fields that incorporate what is referred to here as ‘sacrificial layers.’ The process of laser-shock peening incorporates thermal coatings as well as confined mediums in contact with the processing surface, such as water or glass. This has been shown to improve the peening process attributed to an increase in the peak pressure within the shock wave by at least one order of magnitude shown in water in [21] and under various laser pulse conditions in [22]. The explanation for this result is shown in accounting for the dynamic pressure of the confined plasma. We can consider two cases (a) solid copper interface with solid PVB layer, and (b) solid copper with ambient conditions. Citing the similarities with the cases in [23] we can recalculate the Eqs. (8) and (9) from the study for peak pressures with our values:
Case (a):
Case (b):
The results of these calculations for our study are shown in Fig. 5. A clear increase in peak pressure is displayed, showing the disadvantage in the direct ablation model where the substrate properties do not affect the model, Eq. (4). Once the substrate is enclosed with geometry, the significant factor affecting the peak pressure of the shock wave is the reduced acoustic impedance, described by Eq. (1). Also significant is the plasma length as calculated for a simplified model in Eq. (2), and Eq. (4). The reasoning for this effect is due to the confinement limiting the dispersion of the plasma, thus assuring the plume is a more localized event. This can explain the increased visual oxidation within the surrounding heat-affected zone for the case of no sacrificial layers as compared to the case using the PVB layer, see the microscope images in Fig. 2(a). While these effects may not result in deformation can cause delamination in the following processes for build-up layers. An additional issue with long plasma lengths, while consisting of a decreased density, can still result in light absorption far above the target area, causing lower pulse energies of consecutive pulses to reach the target surface, inducing plasma shielding effects.
Fig. 5. (a)Peak pressure and (b)Plasma length as calculated in [21] for the cases studied with PVB and Copper interface layers. The laser intensity matched the values of the experiment done. Blue lines indicate the results with the PVB layer, red lines indicate the direct ablation case, and green lines represent the ratio of both PVB layers over direct ablation (right vertical axis).
The plasma signals were captured using a similar setup as described in [6], by adding a photomultiplier tube (PMT) with a 400 nm low-pass filter to capture the scattered plasma emission lines from the copper. With this method, we can only capture a small cross-section area of the emission signal, not enough to quantify the total intensity but effective in analyzing the temporal behavior. Figure 6 shows the data from measuring the plasma emission signal from several cases with and without a sacrificial layer of several thicknesses as experimented with in this study. The laser parameters were also like the study, for this case a single 100µJ pulse from a UV – 355 nm laser of ∼32 ns pulsewidth. Consistency over several measurements confirmed the signal shown below is an accurate presentation of the study case. Disregarding the peak power measured, for the reasoning described above, the length of the signal translating to the plasma lifetime shows a significant and consistent difference with and without the sacrificial layer. The direct ablation case has a substantially longer lifetime, nearing a factor of x2 as with sacrificial layers while drilling with a sacrificial layer is consistent with having a shorter lifetime, and as shown previously, this is correlated to having a shorter plasma length. While the plasma lifetime is extended in the direct ablation case, the pressure is shown to be applied only during the laser pulse time described in [24] and explicitly modeled by [25]. In the case of confined plasma, the induced pressure is applied throughout the plasma lifetime, several times the laser pulse time in our case.
Fig. 6. Temporal behavior of plasma emission during direct ablation and with PVB layers of various thicknesses. (a) Time-dependent signal as measured on Photomultiplier Tube (PMT). (b) Determination of plasma lifetime from $1/{e^2}$from the power signal. Data available in Dataset 2 Ref. [17], Dataset 3 Ref. [17].
From these studies, we attribute the difference in the ablation results with the use of sacrificial layers as compared to the direct ablation to the differences in the plasma characteristics. In the case of substantial material ablation, this is may be referred to as laser plume, consisting of a more complex mixture of plasma and large particles of the various materials in the target and the immediate environment described in-depth in chapter 6 of [26] referencing the Knudsen layer. We reason that the most significant response to the plume interactions results in substantially modifying the surface pressure resulting in the phenomena that we have observed of increased ablation efficiency while producing more confined heat-affected zones. Since in the application case of laser trepanning each subsequent pulse impacts a new site still covered with the sacrificial layer this accumulation of surface pressure to the substrate results in reaching the ‘pop-out’ effect as shown previously for lower pulse energies. Future studies are intended to focus on analyzing the plume dynamics when using PVB layers.
5. Conclusions
In the study, we investigated laser drilling processes for VIA formation in Flex PCB substrates using sacrificial layers made of thin-film PVB. Optimization of the PVB material concentration and thickness improved the VIA quality by reducing recast and debris while eliminating the oxidation of the copper layer and not reducing ablation rates. The use of these passive layers has also shown promising results for improving the throughput of the laser drilling process. While this study builds upon experiments and processes that are common in laser peening, the introduction of this process for laser ablation of Flex PCBs and the development of the engi6eered sacrificial layer is also effective for the process and friendly for the materials involved institutes a newfound processing solution that alleviates critical issues which have inhibited this technology adoption for applications. Recommended use of these layers can again prove the advantages of laser micromachining for PCB substrate as well as other applications that required increased quality for micromachining of metallic surfaces.
Funding
Israel Innovation Authority (65500).
Acknowledgments
We thank our research groups for their owing collaboration between institutes in supporting and developing this study. Ultimately producing intriguing results as shown here.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are available in Dataset 1 [17]. Dataset 2 [18], and Dataset 3 [19].
References
1. Prismark Discovery Series: Mobile Phones in Prismark Industry Report (2019).
2. The Printed Circuit Report in Prismark Industry Report (2020).
3. “Automotive Electronics Council Documents,” http://www.aecouncil.com/AECDocuments.html.
4. “High-power UV lasers for precision micromachining | Laser Focus World,” https://www.laserfocusworld.com/industrial-laser-solutions/article/14221734/highpower-uv-lasers-for-precision-micromachining.
5. “Printed circuit board processing with UV lasers | Laser Focus World,” https://www.laserfocusworld.com/industrial-laser-solutions/article/14215774/printed-circuit-board-processing-with-uv-lasers.
6. J. Linden, S. Cohen, Y. Berg, Z. Kotler, and Z. Zalevsky, “Influence of Nanosecond Pulse Bursts at High Repetition Rates on Ablation Process,” J. Laser Micro/Nanoeng. 16(1), 1–5 (2021). [CrossRef]
7. A. Chen, Y. Jiang, T. Wang, J. Shao, and M. Jin, “Comparison of plasma temperature and electron density on nanosecond laser ablation of Cu and nano-Cu,” Phys. Plasmas 22(3), 03301 (2015). [CrossRef]
8. B. S. Shin, J. Y. Oh, and H. Sohn, “Theoretical and experimental investigations into laser ablation of polyimide and copper films with 355-nm Nd:YVO 4 laser,” J. Mater. Process. Technol. 187-188, 260–263 (2007). [CrossRef]
9. M. Hendijanifard and D. A. Willis, “Phase explosion and Marangoni flow effects during laser micromachining of thin metal films,” in Photon Processing in Microelectronics and Photonics VII, A. S. Holmes, M. Meunier, C. B. Arnold, H. Niino, D. B. Geohegan, F. Träger, and J. J. Dubowski, eds. (SPIE, 2008), 6879, p. 687907.
10. L. Tunna, A. Kearns, W. O’neill, and C. J. Sutcliie, Micromachining of Copper Using Nd : YAG Laser Radiation at 1064, 532, and 355 Nm Wavelengths (2001), 33.
11. P. Nothdurft, G. Riess, and W. Kern, “Copper/Epoxy Joints in Printed Circuit Boards: Manufacturing and Interfacial Failure Mechanisms,” Materials 12(3), 550 (2019). [CrossRef]
12. L. Gu, G. Yu, and C. W. Li, “A fast and low-cost microfabrication approach for six types of thermoplastic substrates with reduced feature size and minimized bulges using sacrificial layer assisted laser engraving,” Anal. Chim. Acta 997, 24–34 (2018). [CrossRef]
13. Aric Bruce Shorey, Garrett Andrew Piech, Xinghua Li, John Christopher Thomas, John Tyler Keech, Jeffrey John Domey, and Paul John Shustack, “Sacrificial cover layers for laser drilling substrates and methods thereof,” U.S. patent US9758876B2 (November 27, 2013).
14. Leonard F. Altman and Thomas N. Johnson, “Method of laser drilling a substrate,” U.S. patent US4948941A (February 27, 1989).
15. W. H. Koh, “Thin substrate micro-via interconnect,” U.S. patent US5493096A (May 10, 1994).
16. A. D. Godwin, “Plasticizers,” in Applied Plastics Engineering Handbook (Elsevier, 2011), pp. 487–501.
17. J. Linden, A. Hoch, A. Levy, I. Sakaev, G. Bernstein Toker, O. Fogel, M. Hod, and Z. Zalevsky, “Enhanced throughput and clean laser drilling with sacrificial polymer layer: Dataset 1,” figshare (2022), https://doi.org/10.6084/m9.figshare.20069279.
18. J. Linden, A. Hoch, A. Levy, I. Sakaev, G. Bernstein Toker, O. Fogel, M. Hod, and Z. Zalevsky, “Enhanced throughput and clean laser drilling with sacrificial polymer layer: Dataset 2,” figshare (2022). https://doi.org/10.6084/m9.figshare.20069282.
19. J. Linden, A. Hoch, A. Levy, I. Sakaev, G. Bernstein Toker, O. Fogel, M. Hod, and Z. Zalevsky, “Enhanced throughput and clean laser drilling with sacrificial polymer layer: Dataset 3,” figshare (2022). https://doi.org/10.6084/m9.figshare.20069303.
20. B. Neuenschwander, B. Jaeggi, D. J. Foerster, T. Kramer, S. Remund, and Dr. A. Oehler, “Influence of the burst mode onto the specific removal rate for metals and semiconductors,” J. Laser Appl. 31(2), 022203 (2019). [CrossRef]
21. L. Berthe, R. Fabbro, P. Peyre, L. Tollier, and E. Bartnicki, “Shock waves from a water-confined laser-generated plasma,” J. Appl. Phys. 82(6), 2826–2832 (1997). [CrossRef]
22. D. Devaux, R. Fabbro, L. Tollier, and E. Bartnicki, “Generation of shock waves by laser-induced plasma in confined geometry,” J. Appl. Phys. 74(4), 2268–2273 (1993). [CrossRef]
23. R. Fabbro, J. Fournier, P. Ballard, D. Devaux, and J. Virmont, “Physical study of laser-produced plasma in confined geometry,” J. Appl. Phys. 68(2), 775–784 (1990). [CrossRef]
24. R. D. Griffin, B. L. Justus, A. J. Campillo, and L. S. Goldberg, “Interferometric studies of the pressure of a confined laser-heated plasma,” J. Appl. Phys. 59(6), 1968–1971 (1986). [CrossRef]
25. A. Rondepierre, S. Ünaldi, Y. Rouchausse, L. Videau, R. Fabbro, O. Casagrande, C. Simon-Boisson, H. Besaucéle, O. Castelnau, and L. Berthe, “Beam size dependencyof a laser-induced plasma in confined regime: Shortening of the plasma release. Influence on pressure and thermal loading,” Opt. Laser Technol. 135, 106689 (2021). [CrossRef]
26. J. Dowden and W. Schulz, The Theory of Laser Materials Processing, Second Edi, Springer Series in Materials Science (Springer International Publishing, 2017), 119.
