Diffraction‐Based Strategy for Monitoring Topographical Features Fabricated by Direct Laser Interference Patterning

Process monitoring in laser‐based manufacturing has become a forward‐looking strategy for industrial‐scale laser machines to increase process reliability, efficiency, and economic profit. Moreover, monitoring techniques are successfully used in laser surface texturing workstations to improve and guarantee the quality of the produced workpieces by analyzing the resulting surface topography. Herein, dot‐like periodic surface structures are fabricated on stainless steel samples by direct laser interference patterning (DLIP) using a 70 ps‐pulsed laser system at an operating wavelength of 532 nm. A scatterometry‐based measurement device is utilized to indirectly determine the mean depth and spatial period of the produced topography by analyzing the recorded diffraction patterns. As a result, the average depth and the spatial period of the dot‐like structures can be estimated with a relative error below 15% and 2%, respectively. This new process monitoring approach enables a significant improvement in quality assurance in DLIP processing.


Introduction
Nature provides a wide variety of functional surfaces based on the unique combination of hierarchical surface topography and chemistry that can be observed in plants, animals, and insects. [1] One of the most well-known natural surface functions is the water-repellent and self-cleaning effect of the lotus leaf. [2] Inspired by the enormous biodiversity of flora and fauna, scientists strive to develop surface structures that replicate the functional properties found in nature. [3,4] In particular, laser surface texturing has been demonstrated to be a convenient manufacturing approach to obtain advanced functionalities on a wide range of materials, mainly when sources operate with ultrashort pulses. [5] For instance, specially tailored applications address the modification of optical properties, [6,7] wetting behavior, [8,9] and biomedical performances. [10,11] Among the laserbased manufacturing methods, direct laser interference patterning (DLIP) is an advanced method for periodic surface structuring due to its high flexibility, resolution, and high throughputs. [12][13][14] This laser technology is based on the phenomenon of interference. Here, a laser beam is split into several partial beams using, for example, a diffractive optical element. The partial beams are subsequently focused by a lens onto the sample surface at an overlapping angle, which ultimately determines the achievable texture spatial period and shape. By overlapping multiple partial beams, DLIP enables a wide range of surface textures with lateral feature sizes significantly smaller than the size of the spot diameter. [15,16] Monitoring technologies for laser processes and manufactured parts are essential to improve process efficiency by optimizing reproducibility and increasing the quality of the produced items. [17] Furthermore, integrating a monitoring system is necessary to assure high reproducibility of the produced micro-and submicrometer features in industrial environments and thus enhance the surface functionalities. To guarantee the quality of the machined surfaces, scanning electron (SEM), atomic force (AFM), or confocal microscopy (CM) are usually used to characterize the micro-and nanostructured surfaces ex situ. [18] However, integrating these characterization methods as real-time monitoring systems is not possible. Besides, such measurement methods are not fast enough for high-speed characterization and, therefore, are unsuitable for industrial applications.
Due to its potential for contactless and high-speed measuring methods, scatterometry is a competitive technique for characterizing laser-structured periodic patterns. This promising technology is fast, nondestructive, and can resolve features down to the submicroscale. [19,20] Particularly, a scatterometry-based method showed the potential to characterize the spatial periods and structure heights of low-spatial frequency laser-induced periodic surface structures (LIPSS). [19] The concept of the method is based on measuring and analyzing the light diffracted from the lasertreated surfaces to characterize the topography from the intensity DOI: 10.1002/adem.202201889 Process monitoring in laser-based manufacturing has become a forward-looking strategy for industrial-scale laser machines to increase process reliability, efficiency, and economic profit. Moreover, monitoring techniques are successfully used in laser surface texturing workstations to improve and guarantee the quality of the produced workpieces by analyzing the resulting surface topography. Herein, dot-like periodic surface structures are fabricated on stainless steel samples by direct laser interference patterning (DLIP) using a 70 ps-pulsed laser system at an operating wavelength of 532 nm. A scatterometry-based measurement device is utilized to indirectly determine the mean depth and spatial period of the produced topography by analyzing the recorded diffraction patterns. As a result, the average depth and the spatial period of the dot-like structures can be estimated with a relative error below 15% and 2%, respectively. This new process monitoring approach enables a significant improvement in quality assurance in DLIP processing.
distribution of the resulting diffraction orders (DOs). This phenomenon has been known for decades and several numerical approaches, such as the finite element method (FEM), finite difference time domain (FDTD), and rigorous coupled-wave analysis (RCWA), have been developed along with dedicated commercial software (e.g., Lumerical FDTD, PCGrate, GSolver) to simulate the resulting diffraction patterns. [21,22] When a periodic surface structure is illuminated by a coherent light source, the reflected rays form a diffraction pattern with discrete orders, whose positions or diffraction angle θ m , can be determined by Equation (1).
with m the DO, λ the wavelength of the incident light beam, and Λ the spatial period of the produced texture. [23] As a result, structures with small spatial periods yield large diffraction angles and vice versa. In this study, as a first step, stainless steel samples are structured by four-beam DLIP, producing periodical dot-like structures on the material's surface by local ablation. Afterward, an optical measurement system is used to evaluate the produced topography of the laser-structured areas. This advanced monitoring setup enables the ability to quantify the spatial period of the pattern and the average structure depth. The resulting topographies are analyzed by CM and SEM techniques.

Materials
Stainless steel sheets (AISI 304, 1.4301) with a mean surface roughness of 52 nm and a thickness of 0.7 mm were used for the structuring experiments. The samples were electrochemically mirror polished to minimize the influence of initial surface roughness on the patterning process. No further process steps for cleaning the surface were conducted.

DLIP Setup and Structuring Strategy
The stainless steel samples were structured by a four-beam DLIP setup using a pulsed Nd:YAG laser source (NeoLase GmbH, Hannover, Germany) with an operative wavelength of 532 nm. This four-beam technique allows the creation of a periodic dot-like topography. Figure 1 shows the used experimental setup, including the structuring strategy. During the experiments, the pulse duration was set to 70 ps, and the repetition rate f was kept at 1 kHz. The laser system provided pulse energy of up to %26.5 μJ. Depending on the angle of the overlapping laser beams, spatial periods Λ of 1.7, 2.6, and 5.3 μm were realized. The applied number of laser pulses in the same position N P varied between 1 and 5, and the diameter of the laser spot d S was set from 48 up to 118 μm, resulting in a fluence F between 0.66 and 11.36 J cm À2 . These process parameters were chosen according to previous research in which the same material and same setup were used to maximize the diffraction efficiency for obtaining high-intensity rainbow colors. [24] The movement of the samples in horizontal (x) and vertical (y) directions was realized by a positioning stage system (Aerotech Pro 115, Pittsburgh, USA) at a velocity of 100 mm s À1 . Figure 1b illustrates the applied structuring strategy. During the patterning process, the pulses were applied on the sample surface in the direction of motion (x) with a defined spacing o x . In addition to the overlap in x-direction, the vertical overlap o y of the individual laser spots was used to describe the motion in the secondary direction (y). The overlap distances were kept equal to the set spot diameter so that the pulse-to-pulse overlap was fixed at 0% for all samples during the structuring process. All the experiments were performed at room temperature and under ambient conditions.

Diffraction-Based Monitoring System
The processed surfaces were characterized ex situ using a selfdeveloped measurement system (TU Dresden, Germany) based on scatterometry. This device is described in detail elsewhere. [25] A schematic illustration of the measurement system, including www.advancedsciencenews.com www.aem-journal.com the required hardware components, can be seen in Figure 2a.
First, a low-power laser source (L) with an operating wavelength of 532 nm and an output power of 0.9 mW was used to illuminate the patterned surface (S). Then, the reflected light was focused onto the charge-coupled device (CCD) camera (UI-524°CP-M-GL, IDS, Obersulm, Germany) sensor by an optical system consisting of two lenses (LS). During the measurements, the camera constantly captured the diffraction pattern obtained due to the produced periodic surface topographies on the steel samples as a grayscale image. To avoid saturation of the recorded DO under analysis, leading to a loss of information, the intensity of the laser beam was adjusted by the two polarizers (P) before the measurements. To ensure same experimental conditions, the polarizers were kept in the set position for all samples. Five different locations were measured in each laser-structured field to provide statistical confidence. The image editing software "ImageJ" (National Institute of Health, Bethesda, USA) was used to analyze the captured diffraction patterns. For the evaluation process, a background image was subtracted from the recorded image to remove internal reflections of the measuring system from the image. This background picture was recorded with no sample below the measuring system. Afterward, a range of grayscale values (greater than 15 gray levels) was considered for the evaluation to eliminate the background noise of the camera sensor and environmental effects. Subsequently, the total area of the DOs was calculated at five different positions on each sample and the results were averaged. The linearity of the presented approach was validated by taking images at different light intensities and measuring the current from a photodiode (BPW21, Osram GmbH, Munich, Germany) attached to the lateral exit of the beam splitter cube, that is, without altering the optical path. As a sample, a mirror was placed. The photocurrent was measured with a Ketihley 2450 sourcemeter (Tektronix, Cleveland OH, USA) at a voltage of -1 V. The calibration curve along with a linear fit characterized by a coefficient of determination R 2 = 0.977 are shown in Figure S1 in the Supporting Information. Due to the linearity of the calibration curve, it can be stated that the determined intensities from the DOs using the CCD images are proportional to the real light intensity. Figure 2b shows an exemplary image of a diffraction pattern, including the 0th and one of the 1st DOs (DO). From the arrangement of the DOs, conclusions can be drawn about the type and orientation of the periodic surface structures. For instance, dot-like structures have a 2D distribution of the DOs and the spatial period of the texture can be determined by the distance between the 0th and 1st DO. For calibrating the spatial periods that can be determined with the system, a stainless steel sample was processed by two-beam DLIP yielding a linelike topography with a spatial period of 4.9 AE 0.14 μm, according to optical microscopy analysis. Considering that the camera had a sensor size of 6.784 mm Â 5.427 mm (1280 Â 1024 pixel), the optical setup was designed to collect light reflected at a maximum angle of AE25°from the normal of the surface under study. Considering the diode laser wavelength of 532 nm, the grating equation yielded a minimum detectable period of 1.3 μm. To select the optical components and their parameters, the system was simulated with a ray tracing software (OpticStudio, Zemax LLC, Kirkland, USA). [25]

Surface Characterization
The topography of the produced structures was evaluated by confocal microscopy (Sensofar S Neox 3D Surface Profiler, Barcelona, Spain). For this purpose, a stitched image containing six single images per structured area was captured with a 150x objective at a lateral and vertical resolution of 140 and 1 nm, respectively. Afterward, the measured topographies were analyzed by the software SensoMap 7 (Sensofar, Barcelona, Spain). Furthermore, high-resolution images of the surfaces were captured by a scanning electron microscope (Zeiss Sigma 300, Oberkochen, Germany) at an operating voltage of 6.0 kV and working distances between 7.2 and 12.6 mm.

Results and Discussion
Dot-like structures were fabricated on stainless steel with an average structure depth ranging from 0.04 to 0.42 μm, depending on the set spatial period. The fluence and the number of pulses (N P = 1, 3, and 5) were varied to create topographies with different structure depths. The resulting mean structure depth h as a function of the used pulse number N p and fluence F for the three spatial periods is shown in Figure 3. A polynomial fitting function was used for each spatial period to correlate both pulse number and fluences with the reached average depths. In general, by increasing the number of pulses and laser fluence, www.advancedsciencenews.com www.aem-journal.com deeper structures were obtained, as expected. The maximum reached depths were %0.42, 0.43, and 0.39 μm for the spatial periods of 1.7, 2.6, and 5.3 μm, respectively. Figure 4 shows exemplarily SEM images of different surface topographies (see caption for laser parameters) and the corresponding image of the 0th (dashed yellow circle) and 1st (solid red circle) DOs under analysis. The dot-like topography in Figure 4a(i) exhibits shallow craters and melt ejections. The more laser pulses were applied to the surface, the deeper the craters and the higher the rims around the crater due to melt ejection (Figure 4b,c(i)). This tendency is also confirmed by the corresponding profiles of the topographies, which are extracted from the middle of the DLIP pixel and are presented in Figure 5 (F = 1.39 J cm À12 , and a) N p = 1, b) N p = 3, and c) N p = 5). Figure 4(ii) and (iii) shows the corresponding diffraction patterns recorded by the CCD camera and used to estimate the intensities (areas) of the 0th and 1st DOs, respectively. It can be seen that the area, which is proportional to the intensity, and shape of the 0th and 1st DO change depending on the produced topography. It is worth mentioning that the fluctuations between the intensities of the first DOs in a given CCD image might be due to instrumental error (e.g., slight misalignments of one or several optical components). Although such fluctuations could also be attributed to asymmetries in the textured profiles, it can be seen from Figure 5 that the textures produced here are rather symmetric and this effect can be ruled out. Furthermore, the recorded diffraction patterns show a close similarity with the Fast Fourier transform (FFT), displayed in the insets of Figure 4(i), of the corresponding topography.
To determine the relationship between the intensity of the DOs and the fabricated structure depth, the mean area of the 0th and 1st DOs was calculated and considered as the corresponding intensities. Figure 6 exhibits the mean area of the 0th (a-c) and 1st (d-f ) DOs as a function of the mean structure depth for the three evaluated spatial periods (see caption). Various trends for both DOs can be interpreted from the graphs. As seen for the 0th DO (Figure 6a-c), the intensity drops following a near-exponential decay for all obtained spatial periods. In addition, the intensity of this DO increases with increasing spatial period. Furthermore, for the spatial period of 1.7 μm, the area of the 0th DO almost vanishes at a structure depth of %0.30 μm. In the case of the 2.6 and 5.3 μm period, the minimum value of 5 and 67 pixels occurs at a structure depth of 0.43 and 0.39 μm, respectively.
For the 1st DO, the relation between the mean area or intensity and the average depth of the structure shows a different tendency. At low structure depths, the mean area of the 1st DO (Figure 6e-f ) increases until it reaches a maximum value at a structure depth of 0.11, 0.16, and 0.20 μm for the periods of 1.7, 2.6, and 5.3 μm, respectively. For deeper structures, the diffraction intensities drop for all periods and vanish at depths between 0.40 and 0.45 μm. Contrary to the 0th order, the average www.advancedsciencenews.com www.aem-journal.com intensity (area) of the 1st order decreases with an increasing spatial period. This trend can be explained by the fact that as the spatial period decreases, and fewer DOs are available according to Equation (1); the 1st order becomes more intense at the expense of the vanishing orders. Overall, the behaviors of the 0th and 1st DOs described earlier are also commonly observed for periodic phase gratings. [26] The trends of the investigated DOs can be compared with the classical diffraction efficiency η m equations for sinusoidal phase gratings for the 0th and 1st order according to Equation (2). [26] where m is the corresponding DO, J is the Bessel function of the first kind, and a = 4π h/λ PD is the maximum phase shift between the incident and reflected waves, which depends on the structure height h and the wavelength of the light emitted by the laser diode λ PD . As the calculated intensity is given in an arbitrary unit (pixel), a fitting procedure was performed to normalize the  www.advancedsciencenews.com www.aem-journal.com analytical model in Equation (2) to the data. The resulting curves are shown in Figure 6 with blue lines. The data and the analytical model based on the Bessel functions show very similar trends for both orders and all spatial periods up to a structure depth of %0.20 μm. However, significant discrepancies between the data and model are observed for depths between 0.20 and 0.40 μm.
For instance, the model for the 0th order predicts a local maximum at 0.32 μm, which is not observed in the data (cf. Figure 6a-c). In the case of the 1st order, the analytical curve drops steeper than the data for structure depths in the range of 0.15 and 0.30 μm (cf. Figure 6d-f ). These discrepancies come to be probably explained by the difference between the real texture profile and the assumed sinusoidal profile in the model. As a consequence, this model cannot be used to accurately predict the structure depth upon fitting the diffraction intensities, especially for structure depths larger than 0.15 μm. Alternatively, two fitting curves are proposed for the 0th and 1st DOs for each evaluated spatial period, namely, an exponential decay function for the 0th DO and a 5th-order polynomial function for the 1st DO. The corresponding fitting curves are illustrated in red (Figure 6), resulting in a coefficient of determination R 2 between 0.85 and 0.96 (see Table S1-S3, Supporting Information).
Then, for an arbitrary sample textured by the present method, both fitting equations can be used to extract the mean structure depth of the topography from the measured intensity of the DOs. For low structure depths, that is, 0-0.20 μm, both fitting curves can be used for predicting the depth from the measured intensity. However, for depths larger than %0.20 μm, only the first-order intensity has to be used for the three studied spatial periods to allow the determination of the structure depth.
Finally, an evaluation routine was developed for extracting the topography spatial period and structure height from the collected diffraction patterns by the CCD camera. The spatial period was calculated from the distance between the 0th and 1st DO, whereas the depth was determined from the fitting curves already introduced. For example, the CCD image from the produced topography shown in Figure 4c(i) presented a mean area for the 0th and 1st DO of 50 and 631 pixels, respectively. The algorithm determined a structure depth of %0.14 μm that only differs by 6.5% from the depth obtained by the confocal microscope (0.15 μm). The corresponding spatial period of this structure was calculated to be 1.71 μm with a relative error of 1%. Table S4-S6 in the Supporting Information shows the calculated mean depth and spatial period of the dot-like topography, including the resulting errors for all laser-structured areas depending on the spatial period.
In summary, with the presented approach, it is possible to quantify the generated surface topography in terms of the average structure depth and the spatial period with a mean error smaller than %15% and 2%, respectively. This advanced monitoring approach could be implemented for quality control of the produced topography during the manufacturing process. Due to its compact and robust design, the presented technology is costeffective compared to other measurement methods, making it www.advancedsciencenews.com www.aem-journal.com easier to integrate into laser-texturing machines and highly interesting for industrial applications. [27][28][29]

Conclusion
In this study, periodic dot-like patterns were produced on stainless steel using DLIP technology. The applied laser fluence and the number of pulses were varied to achieve topographies with different structure depths. It was shown that an innovative approach based on scatterometry was capable to provide indirect topographical data from the analyzed laser-treated surfaces. This was performed by quantifying the intensity of the 0th and 1st DOs using a CCD camera. Precise information on the mean structure depth, as well as the spatial period for the dot-like topography, was obtained, with mean errors below 15% and 2%, respectively. In consequence, this simple method can be used to monitor the DLIP processing of metals, which is a requirement on an industrial-scale level. Future studies will focus on an implementation of the device in several laser systems using a computer-based algorithm for the automatic estimation of the mean structure depth depending on the area of the 0th and 1st DO. In addition, the goal of inline process monitoring using this approach will also be addressed.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.