Layer detection in ultrashort pulsed multilayer laser ablation by analyzing ultrasonic process emission

The average power of ultrashort pulsed (USP) lasers has increased steadily over the last decades as a result of the continuous development of higher pulse energies and repetition rates up to the megahertz regime. In consequence, the scanning speed has also increased in order to distribute the available energy over the workpiece in efficient micromachining processes. For any process monitoring approach, these very fast machining processes require very high sensitivity and temporal resolution from the used sensors. While numerous commercial sensor systems are available for laser macro material processing such as laser welding and laser cutting, there are hardly any systems for process monitoring of USP laser-based material processing. A direct transfer of known systems from macro material processing is often not suitable, as the requirements of very small processing geometries and very fast processes represent an impediment. For example, optical coherence tomography (OCT), which is commonly used to monitor welding and cutting processes⁠,^1–3 requires a large keyhole for monitoring compared to the geometry of the interaction zone in USP laser processes, which precludes the use of this system, for example, for drilling microholes in the production of modern printed circuit boards⁠.^4–7 One possibility for process monitoring of short pulse processes is the detection of the generated plasma by laser induced breakdown spectroscopy, which allows the detection of different layers of multilayer materials.^⁠8 Laboratory-scale methods exist for monitoring USP laser processes, such as the setup used by Kunze et al. which is based on a silicon-based line camera and allows the material transition to be detected by changing the secondary emitted intensity spectrum in the range of 200 and 600 nm.^⁠8 However, the low sampling rate compared to today’s repetition rates and the limited spectral range are significant limitations. In addition to the possibility of detecting optical signals using photodiodes or cameras, acoustic process monitoring via the detection of acoustic process emissions using a microphone is known, which has already been successfully demonstrated for processes, such as laser welding,^9,10 laser cutting,¹¹ and additive manufacturing,¹² and for crack detection.^13,14 However, based on various research results,^13,15,16 it can be concluded that conventional microphones have the disadvantage of a relatively narrow bandwidth with high sensitivity for frequencies up to 20 kHz, but have very low or no sensitivity for higher frequency ranges.

Material processing using USP laser systems is characterized by high acoustic emissions in the pulse frequency range, i.e., the repetition rate and its higher harmonics, which cannot be detected by conventional microphones due to the use of repetition rates beyond 50 kHz or even in the megahertz range. New commercially available optical microphones offer the advantage of achieving sampling rates of up to 4 MHz, enabling the detection and analysis of acoustic process emissions with much higher temporal resolution and at higher frequency ranges. These microphones are based on detecting the sound pressure induced changes of the refractive index in the air with a Fabry–Perot resonator, microassembled in front of the tip of an optical fiber, thus basically representing a refractive index sensor. These optical microphones have recently been used, for example, to detect acoustic emissions from high-voltage power lines, where the absence of moving parts has shown advantages over conventional membrane-based microphones.¹⁷ Another application of optical microphones is the monitoring of the connectors assembly in combustion engine production lines,¹⁸ where their high sensitivity and the use of neural networks are intended to achieve automated quality control. In addition to these examples and the application in nondestructive material testing,^19–21 optical microphones have also been used in laser material processing using continuous wave lasers in laser welding. Tomcic et al. used the acoustic emission in the welding process to obtain indications of the predicted and actual welding depth through experimental investigations and data analysis.²² In a further study by Authier et al., a comparison of OCT and an optical microphone for monitoring the welding depth and the formation of the required keyhole is successfully presented.²³ The optical microphone has also been successfully used to detect stress-induced cracks during laser welding of glass, which often occurs due to the low thermal conductivity of glass.¹⁴ In addition to the above applications of laser welding, other applications are known from additive manufacturing,^24,25 laser metal deposition,²⁶ jet-guided laser micro hole drilling,²⁷ and further processes.^28–30 The outstanding properties in terms of high bandwidth and very high temporal resolution are key indicators that also motivate the use for monitoring USP laser processes. Within this context, this work investigates the detection of different materials in the ablation process of multilayer materials using the fiber coupled Fabry–Pérot-interferometer based optical microphone.

The processed three-layer materials contain two 35–70 μm thick copper layers separated by a 50 μm thick layer of polyimide or a 300 μm thick layer of FR4. The investigated materials are used for typical printed circuit board (PCB) based products and are shown schematically in Fig. 1(b). Measurement of the materials in cross section using an optical microscope revealed a tolerance of about 0.2–0.4 μm for the actual thickness of the copper layers, which is negligible for the following investigations. For convenience, the layers of the three-layer materials will be referred to as layers and the passes of the scanning laser system will be referred to as passes in the following document. A schematic illustration of the experimental setup is depicted in Fig. 1(a). A Yb:YAG USP laser (Amplitude Tangor, Amplitude Systems, Bordeaux, France) with an infrared emission at 1030 nm, variable repetition rate between single shot and 40 MHz, pulse durations between 800 fs and 10 ps, and an average power of 100 W was used to ablate the PCB materials. The laser is integrated into a micromachining system (RDX-1000 FBS, Pulsar Photonics GmbH, Herzogenrath, Germany) equipped with a galvonometer scanner (IntelliSCANse 14, SCANLAB GmbH, Pucheim, Germany) and an F-theta lens (LINOS FTheta Ronar, QiOptiq, Photonics GmbH & Co. KG, Goettingen, Germany) with a focal length of 100 mm. The resulting focal diameter of 43 μm (1/e²), measured with a CCD camera (UI149xLE, IDS Imaging Development Systems GmbH, Obersulm, Germany), is moved across the workpiece with the galvonometer scanner at a speed of 1512 mm/s. The scanning speed, a hatch of 15.48 μm and a repetition rate of 100 kHz, results in a pulse overlap of 64% for the exemplary rectangle of 2 × 0.2 mm², resulting in a constant energy input. The fluence is calculated as $Φ= E P/( π ⋅ r 0 2)$⁠, where $E P$ is the pulse energy and $r 0$ is the radius of the focal beam. The ablation process was performed in the experiments with fluences between 0.3 and 3.3 J/cm² to investigate different acoustic emissions caused by different ablation rates. Based on the laser parameters and the optical setup, the Rayleigh length of the focused laser beam is calculated to be 1409 μm. The acoustic emission of the ablation processes is detected by an optical microphone (Eta450 Ultra, XARION Laser Acoustics GmbH, Vienna, Austria) and analyzed in real time using a field programmable gate array (FPGA) based evaluation unit. Although the FPGA-based fast Fourier transformation can be assumed to calculate in real time, the latency of the acoustic signal from the processing point to the microphone and the transfer of the calculation result to the display unit will result in short delays. In the performed experiments at an ambient temperature of 21 °C, a speed of sound of 344 m/s, and an assumed air pressure of 1013.25 hPa in dry air, the sound propagation time for a distance of 100 mm is 0.2906 ms. While the analysis tools can access the spectrum as soon as it has been calculated, there is a delay in transferring the results to the display area. The frame rate of the display area during measurement is 10 frames per second, resulting in a delay of 0.1 s. However, it should be noted that this only applies to the display area, whereas process events can still be detected and reacted to in quasireal time. The membrane-free optical microphone, which is constructed based on a Fabry–Perot interferometer, enables a bandwidth of 50 kHz–2 MHz with a dynamic range of 100 dB. The sensor head of the microphone, not moved during the ablation process, is positioned at a distance of 100 nm from the center of the ablated rectangular surface, which results in an angle of 53.5°. This position was selected to avoid positioning the microphone in the range of the laser beam to prevent damage and to keep the distance between the microphone and the processing zone as small as possible. Furthermore, initial results (Fig. 4) show that the acoustic detection is dependent on the alignment of the sensor head in the scanning field. With these findings, an alignment along the Y axis at X = 0 was selected in order to achieve a symmetric acoustic pattern for positive or negative X-values. The ablated structures are measured using a laser scanning microscope (VK-X3000 series, Keyence AG, Osaka, Japan).

First, five passes of the different materials are ablated with their acoustic emission being analyzed in a fluence range between 0.3 and 3.3 J/cm² and for a pulse duration of 800 fs. For this application, the optical microphone is used with a sampling rate of 1 MHz and a high-pass filter of 50 kHz, allowing for a spectrogram with a bandwidth of 50–500 kHz to be calculated in real time. A frame width of 400 values for a spectrum of 50–150 kHz is used for data cleaning, from which the average value is calculated in data postprocessing. The average acoustic energy of each ablation process shown in Fig. 2 represents the maximum of each data sequence. As can be seen in Fig. 2, the ablation of copper and the resulting acoustic emission show an almost linear characteristic except the boundaries of very low and very high fluences. The emitted acoustic energy at a fluence of 0.9 J/cm² is 0.73 a.u. and increases to a maximum of 2.36 a.u. at a fluence of 3.3 J/cm². Compared to the acoustic emission of copper, the two dielectrics polyimide and FR4 show a higher acoustic emission, but no linear relationship can be deduced. This characteristic is attributed to the comparatively high ablation rate for the two polymers, which leads to a higher depth of ablation during the ablation process, resulting in slightly defocused processing, in turn resulting in a lower ablation rate and, thus, lower acoustic emission. In addition, the acoustic emission of polyimide is higher as compared to FR4 and, therefore, shows a greater difference to the emission of the copper material.

Overall, the investigations show that the integrated energy of the ultrasound emissions of the ablation process is strongly dependent on the respective material. Therefore, the material transition between copper and polyimide layers can be clearly detected and monitored with the optical microphone in quasireal time by monitoring the acoustic energy. This property could be advantageous for subsequent experiments on material specific ablation or layer detection.

Furthermore, the acoustic energy is analyzed in relation to the ablation depth of the ablation process at a fluence of 1.5 J/cm². For this study, ablation processes with a number of 15–1000 passes are performed with a repetition rate of 100 kHz. A sampling rate of 1 MHz and a high-pass filter of 50 kHz are defined for the optical microphone, resulting in a bandwidth of 50–500 kHz. In order to calculate the average acoustic energy of the respective ablation process between 15 and 1000 passes, a moving average of 16 values for a frequency band of 50–150 kHz is used in this evaluation, from which the average value over the entire ablation sequence is calculated. The ablation process of 1000 passes, represented by a continuous graph in Fig. 3, is calculated by taking the moving average of 16 values, without subsequent generation of a mean value. As shown in Fig. 3, at the beginning of the ablation process, the maximum acoustic energy of about 23 a.u. is reached with a slight decrease of about 23% up to an ablation depth of about 150 μm. After reaching this level, the acoustic energy continues to decrease, although with a slightly lower gradient, and reaches a value of 16 a.u. at a depth of 300 μm, corresponding to 68% of the maximum at the beginning of the ablation. The decrease in acoustic energy from the beginning of ablation to the end of ablation at a depth of 300 μm can be explained by the defocusing of the focal spot and the resulting reduction in fluence and ablation rate. Starting from a spot size of 43 μm on the surface of the material defocusing by 300 μm results in a measured spot diameter of 45.3 μm. The resulting larger focal area reduces the fluence from the initial 1.5 to 1.35 J/cm², which leads to a lower ablation rate. As shown in Fig. 2, both the fluence and the ablation rate have a direct impact on the acoustic energy, which explains the described characteristics. Comparing the described path except for the higher deviation in the range below 50 μm, the mean values calculated for the ablation processes with different numbers of passes agree quite well with the process with 1000 passes. The shift toward a lower value for the sub 50 μm ablations can be explained by the calculation of the mean value. A process with only 15 passes, for example, has a comparatively short process time, and, therefore, a fewer number of spectra can be detected. The formation of a representative average also includes the rising and falling edges of the signal, which have a greater influence with a smaller number of measured amplitudes. As the number of available values increases, the influence of the edge areas decreases and the deviations become significantly smaller, as can be seen in Fig. 3.

The used microphone is positioned at a fixed distance of 100 mm from the center of the scan field at an angle of 53.5° to the material surface to protect the microphone from damage by the laser beam incident vertically, while at the same time being positioned as close as possible to the machining process. As a later application in printed circuit board manufacturing is not limited to a small ablation area in the center of the scan field, the acoustic emission is investigated depending on its position in a scan field of 50 × 50 mm². The ablation processes are performed within this scan field with a step size of 2.5 mm, where the respective position corresponds to the center of the 0.2 × 2 mm² rectangle. While the sampling rate and the high-pass filter remain unchanged from the previous experiments, only a maximum of 32 serial values are taken and the acoustic energy is averaged with a bandwidth of 50–150 kHz in order to reduce the data value. The maximum value of the resulting data series consisting of 21 ablations along the respective y-position is used. The respective maxima over the analyzed scan field are shown in Fig. 4.

Figure 4 shows that the acoustic emission across the scan field is not constant but depends on both the distance to the microphone and the direction. The microphone positioned at X = 0 mm in the direction of the Y axis detects the maximum acoustic emission for positions in the range X = ±2.5 mm and Y = 22.5 mm–25 mm. The measured values in this area represent the shortest distances to the microphone and have a small angular deflection compared to other areas in the examined scan field. Considering the values from X = 0 mm to Y = 25 mm toward higher or lower X-values, a decrease of the acoustic energy toward the limits of the scan field can be observed, which corresponds a decrease to 79.2% for X = 25 mm and 75.5% for X = −25 mm compared to the maximum value (X = 0 mm and Y = 25 mm). This reduced acoustic emission shows the effect of the directionality of the observed ablation process toward the microphone. In addition to the directional dependency, a distance dependency can also be seen, as the emissions detected at short distances are generally higher than those at longer distances. These characteristics are based on the attenuation of high frequencies by air, which is known as a general law in the field of acoustics.

For further evaluation of the acoustic emission in the scan field, Fig. 5 shows the sections along the X- and Y-axes through the scan field origin. The secondary X axis shows the distance from the microphone of the ablations, recorded along the Y axis for X = 0 mm for easier analysis and evaluation of the acoustic values. Analyzing the cross section of the X axis shows a decrease in acoustic emission toward higher and lower X-values, i.e., toward the edges of the scan field. In the range of X = ±10 mm, an almost flat level exists where the acoustic emission is only slightly reduced (±2.8%) as compared to the center of the scan field (X = 0 mm and Y = 0 mm). Based on the maximum acoustic emission for X = 0 mm, the reduction is 24.3% for X = −25 mm and 20.4% for X = 25 mm. In addition to the slightly higher distance of 3.1 mm to the microphone compared to the center of the scan field (X = 0 mm and Y = 0 mm), these positions are also characterized by an angular deviation of 22.6° in the XY-plane. This relatively large angular difference and the associated sound waves entering the sensor head at a nonperpendicular angle are the main influencing variables for these positions. Compared to the almost symmetric path along the X axis of the X-section discussed previously, the section along the Y axis at X = 0 mm shows an increasing path for higher Y-values and a decreasing path for smaller Y-values. The maximum acoustic energy is recorded for the ablation process positioned closest to the microphone at Y = 25 mm, which has a distance of 87.3 mm and an angle of 66.4°. Due to the fixed height of the microphone above the component surface, the angle changes toward the center of each ablation and is 43.3° for the ablation furthest from the microphone at Y = −25 mm with a distance of 116.7 mm. As a result of the change in position, the acoustic energy is reduced to 86.6% of the maximum value at Y = 25 mm.

Following these fundamental investigations of the acoustic emission characteristics in the laser process, the application of layer detection during multilayer printed circuit board material processing is investigated. A sampling rate of 6.25 MHz is used to analyze these processes in detail and a moving average is calculated to simplify the data. The acoustic energy shown in Fig. 6(b) is calculated over a bandwidth of 50–1050 kHz by averaging two consecutive values. First, the ablation process is performed with a fluence of 1.5 J/cm². At this fluence, the ablation depth in copper is approximately 180 nm per scanned pass, which is commonly denoted as a pass. The ablation process with 300 passes is shown in Fig. 6(a), which illustrates the different ablation properties of the copper and polyimide layers. Approximately, 185 passes are required to ablate the first 35 μm thick copper layer, whereas the subsequent 50 μm thick polyimide layer is ablated after only 30 more passes, corresponding to an ablation of 1.6 μm per pass. As can be seen in Fig. 6(a), due to the material structure of the layers, there are two material transitions in the ablation process, at depths of 35 and 85 μm, which are characterized by different ablation rates due to the material properties. The slightly lower removal rate in the second copper layer compared to the first copper layer, which is characterized by a slightly lower gradient, results from the defocusing of the laser beam as it is not guided during the process. As a result of the laser-material interaction and the different ablation characteristics, a variation in the acoustic process emission can also be detected by the high bandwidth microphone. As Fig. 6(b) clearly shows, there is a sudden increase in acoustic energy in the region of the first material transition (copper–polyimide). The sudden increase in acoustic energy is detected at 3.921 s after the start of the ablation process, corresponding to 185 passes, demonstrating that the 6.25 MHz sampling rate and data postprocessing are sufficient for this purpose. After reaching the maximum of the acoustic energy, it remains at a high level during the ablation of the polyimide layer, followed by a sudden drop caused by reaching the second material transition (polyimide–copper) after approximately 210 passes. With respect to fast process monitoring, the abrupt change of the emitted acoustic energy in the area of the material transitions is worth highlighting. The recognizable transition area is related to tolerances in the thickness of the base material as well as slightly different ablation depths in the ablated surface. As a result of these process conditions, the ablation surfaces in the transition area show a combination of copper and polyimide, which can lead to hybrid acoustic emission.

To illustrate the ablation process at 1.5 J/cm² in detail and relate it to the measured acoustic energy, Fig. 7 depicts cross sections of the ablated areas taken with a laser scanning microscope at 50× magnification for a number of passes between 50 and 220. The top layer visible in the microscope image represents the material used to embed the sample. Between the embedding material, there are two 35 μm thick copper layers, separated by the 50 μm thick polyimide layer. As can be seen in Figs. 7(a)–7(c), the copper layer is removed in the first 175 passes, followed by the first material transition between polyimide and copper after about 185 passes. Figure 7(e) shows the immediate transition or combination between the copper and the polyimide surface resulting from slight accumulation effects in the central area of the ablation geometry. The acoustic energy [Fig. 6(b)] reaches its maximum level in this area after 185 passes, which allows a unique determination of the emission and represents the detection of the polyimide layer. After the rapid ablation of the 50 μm polyimide layer in the following 30 passes, the second material transition is reached, which corresponds exactly to the sudden drop in acoustic energy in this process area. The microscope images shown in Fig. 7 and the derived conclusions are consistent with the time course of ablation rate and acoustic emission and confirm the described approach for layer detection.

For comparison, the ablation process was also performed with a fluence of 3 J/cm² and otherwise identical parameters. The experimental data shown in Fig. 8 were generated using the same evaluation methods. Due to the increased fluence, a pass thickness of approximately 380 nm is ablated per pass on copper, which corresponds to a 2.1-fold ablation rate compared to the previous experiment. Due to the higher ablation rate, the first copper layer is ablated in 92 passes when the first material transition (copper–polyimide) is achieved. The 50 μm polyimide layer is removed in 18 passes, resulting in a pass thickness of 2.7 μm which corresponds to a 1.7-fold ablation rate compared to the precious experiment with a fluence of 1.5 J/cm², leading to the second material transition (polyimide–copper). As already observed, the removal rate in the polyimide material is significantly higher than in the copper cladding, which leads to a much steeper progression of the removal rate in this area. Fig. 8(b) shows the path of the averaged acoustic energy during the ablation of the PCB material at a fluence of 3 J/cm² and shows similar characteristics to the previous experiments. Again, there is a significant increase in acoustic energy in the area of the first material transition, which decreases after reaching the second material transition. The sudden increase in acoustic energy occurs after approximately 1.87 s, which corresponds to the ablation of approximately 90 passes. It should be noted that the temporal resolution of the sensor recording and evaluation is the same in both experiments. However, as the recording time in the experiment is significantly shorter due to the higher ablation rate at 3 J/cm², this results in a less detailed mapping of the acoustic energy in the diagram. In a subsequent comparison of the two experiments [Figs. 6(b) and 8(b)] on the same material, a general increase in acoustic energy can be detected with the higher fluence of 3 J/cm², which can be attributed to the higher ablation rate. As a result of processing the material with a low ablation rate, the acoustic energy drops sharply. This behavior can be observed independently of the used material and, therefore, allows conclusions to be drawn about the correlation between the ablation rate and the measured level of acoustic energy.

In addition to polyimide, other intermediate layers such as FR4 are used in today’s printed circuit boards to isolate the copper layers. This epoxy resin composite material is characterized by the use of a glass fiber mesh. In order to demonstrate the broadband applicability of layer detection, experiments were performed on FR4 based PCB material with a copper cladding of 35 and 70 μm using a fluence of 3 J/cm². As Fig. 9 reveals, the ablation of FR4 base material shows a slightly lower acoustic emission compared to the ablation of polyimide, which is consistent with the results shown in Fig. 2. Figure 9(b) also shows that after removing the 70 μm copper layers after approximately 170 passes, i.e., slightly more than twice as in Fig. 9(a), the acoustic energy is again slightly lower during the removal of the FR4 material. Defocusing the laser beam through the two ablated layers of 335 and 370 μm, respectively, results in a measured increase in spot diameter to 45.6 and 45.8 μm (1/e²), which reduces the resulting fluence to 2.66 and 2.64 J/cm², respectively. A direct comparison of the detailed acoustic energy curves in Figs. 9(a) and 9(b) shows slight differences in level and deflection. It should be noted that both the laser process and the material are subject to slight process variations. Similarly, the propagation of sound waves or process emissions is generally not completely constant in a process that is permanently changing slightly, resulting in differences in detail. Nevertheless, in the here presented context, a direct comparison of processes with different parameters is not necessary or suitable as demonstrated by the fact that layer detection or differentiation is obviously possible due to the sudden increase in acoustic energy. However, these further experiments show that the presented layer detection can be implemented on different materials, depending on the thickness of the copper layer and the base material.

The characteristics of an optical microphone with high bandwidth and high sampling rate allow the detection of acoustic energy and serve as an indicator of the material transition in multilayer materials.

We have demonstrated the possibility of process monitoring in the ablation of multilayer materials using ultrashort pulsed lasers with a membrane-free optical microphone. The presented results show the possibility of layer detection by analyzing the acoustic process emissions in relation to the material transition in quasireal time, thus facilitating reliable process monitoring. In particular, the influence of the ablation depth on the detection of acoustic energy down to a depth of 300 μm was experimentally investigated, in addition to the investigation of the separate base materials for the printed circuit boards, copper, polyimide, and FR4, with respect to their acoustic emission as a function of fluence between 0.3 and 3.3 J/cm². Furthermore, the detection of the acoustic emission of the ablation geometry scanned with a laser repetition rate of 100 kHz and a scanning speed of 1512 mm/s over the 50  × 50 mm² scan field was characterized for the practical use of the sensor system, whereby both a distance dependence and a direction dependence between the process and the sensor head were characterized. In the context of these fundamental experiments, the layer detection was performed on different three-layer printed circuit board materials consisting of two 35–70 μm copper layers separated by a 50 μm polyimide or a 300 μm thick FR4 layer with fluences of 1.5 and 3 J/cm². By comparing the microphone response, a direct correlation was found between the increase in acoustic energy and a higher ablation rate in the different material layers. The material transition between the different material layers can be resolved by the pronounced increase in ablation rate and the significant increase in acoustic energy when reaching and ablating the polyimide or the FR4 layer. Cross-sectional images of the multilayer material showing the ablation progress of the copper and intermediate layers confirm the temporal locations of the material transitions and reinforce the conclusions from the ablation rate and acoustic emission. The developed layer detection method is based on the broadband acoustic monitoring of the frequency spectrum between 50 kHz and 2 MHz in air and the detection of a high gradient change in acoustic energy. In summary, it can be concluded that the investigated sensor technology with the presented monitoring method represents a promising basis for the automated control of layer- or material-selective ablation using an ultrashort pulsed laser.

We acknowledge funding by the German Federal Ministry of Education and Research (BMBF project MOSES, Grant No. FKZ 13N16330 and project KIsS, Grant No. FKZ 13FH013KI2).

The authors have no conflicts to disclose.

Christian Lutz: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Funding acquisition (supporting); Investigation (lead); Methodology (lead); Project administration (equal); Software (lead); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). Cemal Esen: Writing – original draft (equal); Writing – review & editing (equal). Ralf Hellmann: Funding acquisition (lead); Methodology (supporting); Project administration (lead); Resources (lead); Supervision (lead); Writing – original draft (supporting).