Open Access
Add to BibSonomy
Abstract
We present an in-line metrology solution for dimensional characterization of roll-to-roll imprinted nanostructures. The solution is based on a scatterometric analysis of optical data from a hyperspectral camera deployed at a production facility, where nanostructures are produced at speeds of 10m/min. The system combines the ease of use of a real-space imaging system with the spectral information used in scatterometry. We present nanoscale dimensional measurements on one-dimensional line gratings with various periods and orientations. The depths of the produced structures are accurately characterized with uncertainties on the scale of a few nanometers. The hyperspectral imaging capabilities of the system can also be used to avoid vibrational effects.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Nanostructured surfaces have a range of applications in fields such as optics [1] and medicine [2], and can add specific functionalities to a surface, including hydrophobicity [3], structural coloring [4,5], improved heat transfer [6], and optical anti-reflectivity [7]. Several methods have been employed to mass-produce nanostructures [8–10]. However, few, if any, can compete with the speed of roll-to-roll embossing.
In roll-to-roll embossing, nanostructures are copied from the surface of a cylinder to the surface of a moving foil, referred to as the web, by applying heat and pressure. This requires that the processed material is thermoformable (at least on the surface). The nanoimprinting temperature, pressure, and contact time (the speed of the embossing machine) are critical parameters for a successful replication, but the optimal parameter values can be challenging to find. Nanoscale periodic structures give a visual appearance of the surface, and therefore hint towards the quality of the imprinting: the stronger the visual effect, the better the nanoimprinting. However, a more quantitative measure of the nanostructure dimensions is often needed.
The production speed enabled by roll-to-roll fabrication presents a challenge to current metrology for characterization of the fabricated nanostructured surfaces. Conventional nanoscale characterization techniques such as atomic force microscopy (AFM) and scanning electron microscopy (SEM) are simply too slow to keep up with the production rate, and one must therefore resort to offline sampled tests often performed at remote laboratories. Since the replication fidelity of structures is strongly dependent on the production parameters, real-time feedback of the produced structures is highly desirable.
In this paper, we present an instrument for real-time characterization of roll-to-roll produced nanostructured surfaces. We employ optical scatterometry, a fast technique capable of measuring the structural parameters of periodic nanostructures, with the precision of a few nanometers [11]. We demonstrate that the technique is robust enough, and fast enough to be used for in-line measurements in a roll-to-roll production environment with an accuracy comparable to conventional metrology techniques. Other scatterometry-based solutions suitable for roll-to-roll production of nanostructures have been demonstrated previously [12–14]. These methods are based on the scattered signal from a monochromatic light source collected at a series of angles, rather than using a broadband light source and collecting only the specular transmission/reflection. Using broadband light, one gets more data for the inverse modelling, making the reconstruction more robust, and the angle of the specular reflection/transmission is independent of the sample surface, reducing sensitivity to vibrations, and making universal detection hardware simpler. The presented instrument has the added versatility of an imaging system, making it simple to visualize the measurement area.
2. Methods
2.1 Shim fabrication
In this study, a series of nanoscopic gratings were investigated. The design of a single unit cell is illustrated in Fig. 1(C). The unit cell consists of 16 fields of ${4}\;\textrm {mm} \times {4}\;\textrm {mm}$ in size, and covers 8 different periods from 700 nm to 1400 nm in steps of 100 nm, with two different orientations (horizontal and vertical) for every period. The gratings are designed as 50/50 gratings, i.e. to have a width equal to half the period, and a height of 230 nm.
Fig. 1. (A) Si substrate after dry etching. (B) ${150}\;\mathrm{\mu} \textrm {m}$ thick Ni shim prior to welding onto the embossing cylinder. (C) Illustration of a single unit cell consisting of 16 fields that cover 8 different periods from 700 nm to 1400 nm. A 50/50 grating is defined as having linewidths of half the period as shown on the right. The depicted orientation corresponds to that of the periods $p=[800, 1000, 1200, 1400]$ nm (the others are flipped upside down). (D) AFM measurement of the Ni shim. Pattern heights of 230 nm for the 700 nm period and 233 nm for the 1400 nm period (shown) have been measured. (E) SEM image (${30}^{\circ }$ tilted) of the Ni shim; periods of 700 nm and 800 nm are shown. (F) SEM image (${30}^{\circ }$ tilted) of the Ni shim; periods of 1300 nm and 1400 nm are shown.
A deep ultraviolet stepper (Canon FPA-3000EX4) was utilized to pattern a photoresist on a 6 inch silicon wafer. The pattern was subsequently transferred to the silicon substrate by dry etching in an inductively coupled plasma etching tool with SF$_6$ and C$_4$F$_8$ gasses. The patterned silicon substrate after dry etching is shown in Fig. 1(A).
A Technotrans microform 200 Ni electroplating tool was used for the Ni shim fabrication. A 100 nm thick Ni-Va-layer was used as the seed layer prior to Ni electroplating onto the Si wafer. To maintain the necessary flexibility of the Ni shim for welding onto the embossing cylinder (see Section 2.2), a Ni shim of only ${150}\;\mathrm{\mu} \textrm {m}$ thickness was fabricated. A maximum current of 3.5 A and total charge amount of 18.4 Ah was necessary to electroplate a ${150}\;\mathrm{\mu} \textrm {m}$ thick Ni shim. After electroplating, the Si wafer was removed using KOH etching. The Ni shim prior to welding is shown in Fig. 1(B). AFM inspection of the Ni-shim shows a pattern height of $h_{\textrm {shim}} = {230}\;\textrm {nm}$ for the 700 nm period and $h_{\textrm {shim}} = {233}\;\textrm {nm}$ for the 1400 nm period, see Fig. 1(D). Fig. 1(E) and Fig. 1(F) show SEM images of the Ni shim (periods of 700 nm, 800 nm, 1300 nm and 1400 nm are shown). The SEM inspection confirmed a high pattern fidelity for all periods.
2.2 Roll-to-roll embossing
The roll-to-roll embossing machine used is a narrow-web machine (width of 200 mm) capable of running from 3 m/min up to 30 m/min. The machine has an unwinder unit that takes a roll of material up to 300 mm in diameter and controls the unwinding tension. The web goes from the unwinding unit to a nip that is formed by two cylinders: the embossing cylinder and the counter cylinder. The shim is attached around the embossing cylinder and heated by an electrical resistor inside the cylinder. The temperature can be adjusted from room temperature up to ${180}^{\circ }\textrm {C}$. The counter cylinder, which is rubber coated with a hardness of Shore A 99, is pushed against the embossing cylinder by a pneumatic system, resulting in pressures over 20 MPa in the nip. The contact length in the nip is roughly 4 mm in the direction of motion of the foil. Thus, the web is heated under pressure in the nip in a short time (from 8 ms up to 80 ms). The final unit in the machine is a rewinding unit that maintains tension control and releases the web from the surface of the shim and makes a roll of the processed material.
2.3 Scatterometry
In scatterometry, an optical fingerprint of a surface is used to determine its morphology and/or chemical composition. In our case, this optical fingerprint, denoted $\eta (\lambda )$, is given by
It is defined as the ratio between the light transmitted through a transparent sample, $S(\lambda )$, with respect to the incoming light, $R(\lambda )$. A dark measurement, $D(\lambda )$, is taken with the light source switched off to account for electrical dark counts and background light. To reconstruct the surface structure from the optical fingerprint, inverse modelling is used. Here we numerically generate a library of theoretical optical fingerprints, $\boldsymbol{f}(\alpha ,\Omega _{i})$, for a range of sample parameters, $\alpha$ (topological parameters and sample material), and experimental conditions, $\Omega$ (wavelength, polarization, angle of incidence of the light, and number of Fourier terms retained in the calculations). The library of optical signals is generated using an implementation of RCWA as described in detail in [11]. From the generated library, an optimal solution is found by minimizing
Based on preliminary simulations of transmission signals with varying height, width and period, an inter-correlation between these parameters was found, meaning that different permutations of the parameters generate similar transmission signals $\boldsymbol{f}(\alpha ,\Omega _{i})$. This inter-correlation could be removed using complementary measurement techniques such as AFM, SEM or ellipsometry, or using more information (additional measurement wavelengths, polarizations or angles of incidence), but it was decided to simplify the modelling to only reconstruct the parameter most prone to change during fabrication, namely the height of the surface structures. In previous work we have seen that once a sufficient height has been achieved, the width and period will be defined by the shim [15]. Therefore, the embossed periodic gratings are described as simple rectangular lines with a given period, a width equal to half the period, and a height ranging from 0 to 250 nm in steps of 1 nm. This height range is chosen to cover the extremes: No replication ($h = {0}\;\textrm {nm}$), perfect replication ($h = h_{\textrm {shim}} = {230}\;\textrm {nm}$), as well as every height in between. During the reconstruction, the period, width, and the directionality of the gratings are locked to the corresponding values of the shim.
2.4 Measurement setup
A schematic of the system is shown in Fig. 2. The light from a white LED is polarized (along the direction of motion of the foil) before impinging on the sample. The polarizer increases the image contrast between fields with grating orientations parallel- and perpendicular to the light polarization, respectively. The transmitted light is focused onto an aperture to eliminate uncollimated light and light originating from higher diffraction orders; thus, only the 0th order transmission from the sample is retained. After the aperture, the light is underfocused slightly using a lens to match the NA of the spectrograph (Kymera-193i, Andor). The spectrograph has an entrance slit of ${10}\;\mathrm{\mu} \textrm {m}$ and is equipped with a grating optimized for first order diffraction (blaze of 500 nm with 150 lines/mm). A Dove prism can be inserted in the beam path before the spectrograph to rotate the image relative to the entrance slit. The diffracted light is finally collected by a camera (Zyla 5.5, Andor).
Fig. 2. Schematic of the measurement setup. The inset in the upper right corner shows the hyperspectral data cube created by continuously measuring the tracks in the $y$-direction as the sample is moving along the $x$-direction.
2.5 Hyperspectral camera
Using the spectrograph and camera as a hyperspectral linescan camera, a spectrum is acquired for each pixel in a line perpendicular to the direction of movement of the sample film. This results in a two-dimensional image with a spatial axis $y$ (referred to as tracks) and a wavelength-resolved axis. Acquiring a series of images as the sample moves past the camera, and stitching these images together along the $x$-axis (referred to as frames), a hyperspectral image can be formed where each pixel contains a spectrum suitable for scatterometric analysis; see the inset in the top right corner of Fig. 2. The resolution in the $x$-direction is defined by the acquisition rate and the web speed. Using the entire camera area of 2160 x 2560 pixels when producing hyperspectral images, the system is capped at an acquisition rate of 100 Hz. This rate is restricted by the integration time and internal hardware of the camera. In order to improve the acquisition rate, one can restrict the readout to a subset of tracks or use binning to reduce the amount of data, and in turn achieve acquisition rates of up to 1 kHz. Alternatively, the system can be used as a conventional camera by opening the slit and using the spectrograph grating as a mirror by imaging the 0th order reflection onto the camera, which is referred to as imaging configuration (see Fig. S1, Supplement 1). The imaging configuration makes it simple to align the system with different areas on the sample before hyperspectral images are acquired. For the images acquired at 100 Hz, the tracks were binned from 2160 to 540 tracks and for the images acquired at 1 kHz only the data for 4 tracks was read out. For both cases, the maximum integration time allowed by the hardware was used (9 ms and 0.9 ms, respectively) to maximize the signal-to-noise ratio, and the wavelengths were binned from 2560 to 134 bins.
3. Results
In Fig. 3(A) and (B), a hyperspectral image sliced at a wavelength of 450 nm is shown. Fig. 3(B) corresponds to the unit cell in Fig. 1(C). The different fields can be clearly identified, and the operator can simply select the pixels where the analysis should be performed. For the data shown in Fig. 3, the $x$-resolution is $\simeq {1.7}\;\textrm {mm}$. With a field size of 4 mm we ensure that we always have at least one pixel fully contained in each field.
Fig. 3. (A) Hyperspectral image sliced at a wavelength of 450 nm. Five unit cells are labeled using Roman numerals and the roll direction is indicated by the arrow. (B) Zoomed-in image of the area marked in (A). The six labeled fields are further analyzed. (C) $\chi ^{2}$ (Fig. 2) as a function of the modeled structure height. The best fitting height and the corresponding confidence interval can be seen in the insert. (D) Average spectral response of 30 tracks in a single frame for the different unit fields (dots) and best fitting simulation (dashed line). (E) Extracted heights for the labeled fields for the different unit cells shown in (A). The dashed line denotes the average found over the five unit cells for a particular field. The $\sigma$ values and error bars in (C) and (E) respectively, denote the $k=2$ expanded confidence intervals.
Since the measurements are performed during fabrication, vibrations are expected. The effect of vibrations in the system can be estimated by looking at the “sharpness” of the edges of our unit cells along the $y$-axis in Fig. 2 (horizontal in Fig. 3(A) and (B)). This edge is sharper than 10 pixels (corresponding to 500 micron) over all frames. To eliminate vibrational effects in the analysis, we select areas that are at least 20 pixels away from the edges for analysis.
By numerical modelling, it has been found that changing the angle of incidence of the probing light by a few degrees has only a sub nanometer effect on the reconstructed height.
The results from the six fields labelled in Fig. 3(B) are analyzed further here. Data from the other fields can be found in Supplement 1, Fig. S2. The grating periods are 1200 nm for fields 1 and 2, 1300 nm for fields 3 and 4, and 1400 nm for fields 5 and 6. In Fig. 3(C), $\chi ^{2}$ is plotted as a function of the height in the fitted model for the six fields. We see that all structures have unique solutions, but at different heights. The $k=2$ expanded confidence intervals are calculated as described in [16] (page 716). This method is selected based on speed and simplicity, although various other methods to estimate the uncertainty of the reconstructed parameters exist [17–20]. Following this approach, a confidence interval of roughly 10 nm is found for the different fields. In Fig. 3(D), the spectral response from the fields labelled in Fig. 3(B) can be seen together with the best fitting models. The fits are not “spot on”, which is to be expected when imperfect structures are approximated as perfect rectangles with only the height as a fit parameter. A general trend is observed: Fields with lines perpendicular to the roll direction (field 1, 4 and 5) are found to have a lower height (and therefore a worse replication) compared to lines parallel to the roll direction (field 2, 3 and 6). Furthermore, it is found that high-period lines yield a better replication fidelity, as also observed for injection molded structures [21]. This trend is also observed for the other structures, as can be seen in Supplement 1, Fig. S2. In Fig. 3(E), the same fields are measured for all five unit cells in Fig. 3(A). Here we see that the height estimated by the system is consistent for the different unit cells.
As a reference, traceable AFM measurements were performed on the six fields labelled in Fig. 3(B); the measured heights can be seen in Fig. 4 together with the values found by the scatterometer. The AFM used is a Park Systems NX20 equipped with a Point Probe Plus tip, operated in tapping mode. The measurements were analyzed in SPIP using the ISO 5436 step height module. We see a good agreement between the two instruments, although the AFM measures slightly larger heights. We believe that this is caused by small local defects in the sample, where lines have a lower height due to an imperfect imbossing. This results in a lower average height over the large area measured by the scatterometer, compared to the localized AFM measurements, where defective areas were avoided. We can not completely rule out the presence of universal artefacts in the structure, causing the observed difference in height, however, such artefacts are not noticaeable in the AFM data, and the height differences are small compared to measurements uncertainties.
Fig. 4. (A) AFM line profile measured from field 3. (B). Heights measured by the scatterometer (black crosses) and AFM (red circles) for the different fields labeled in Fig. 3(B). The error bars denote the $k=2$ expanded confidence intervals.
The scatterometry data presented in Figs. 3 and 4 were acquired using an acquisition rate of 100 Hz. To demonstrate high speed acquisition, an example of data acquired using an acquisition rate of 1 kHz can be seen in Fig. 5, where the number of tracks was limited to four. We see that this measurement finds the same height with a slightly larger confidence interval, arising from a lower signal to noise ratio. In the end it becomes a trade-off between measurement area and speed. Exploring alternative data transfer methods could potentially mitigate this trade-off.
Fig. 5. $\chi ^{2}$ (Fig. 2) as a function of the modeled structure height for field 6 (as shown in the upper right insert) measured at an acquisition rate of 1 kHz. The measured data and best fitting model can be seen in the lower left insert. The reconstruction finds a height of 199 ± 10 nm.
The instrument makes it possible to characterize the fabricated structures in-line, without presenting a bottleneck in the production system. This can be used to create or optimize a fabrication protocol from a large parameter space. Indeed, the optimal fabrication parameters can vary for different nanostructures and materials, and therefore the system presented makes a production significantly more flexible. Once the fabrication parameters are set, the system can furthermore be used for quality control in a running production line.
4. Conclusion
We have presented an instrument based on a hyperspectral linescan camera for performing in-line scatterometry analysis on roll-to-roll fabricated nanostructures during production. The acquired data can be displayed as an image, making it very intuitive compared to other diffraction based solutions. Acqusition rates of up to 1 kHz have been demonstrated. The measurements are found to be accurate with a confidence interval on the scale of nanometers. The system allows for direct feedback to an operator at the fabrication line, facilitating real-time optimization of embossing parameters as well as in-line quality control.
Funding
Danish Agency for Institutions and Educational Grants; Eurostars (E113231 - R2RMON).
Disclosures
The authors declare no conflicts of interest.
Supplemental document
See Supplement 1 for supporting content.
References
1. N. C. Panoiu, W. E. I. Sha, D. Y. Lei, and G.-C. Li, “Nonlinear optics in plasmonic nanostructures,” J. Opt. 20(8), 083001 (2018). [CrossRef]
2. I. S. Kwon and C. J. Bettinger, “Polydopamine nanostructures as biomaterials for medical applications,” J. Mater. Chem. B 6(43), 6895–6903 (2018). [CrossRef]
3. H. Liu, X. Liu, J. Meng, P. Zhang, G. Yang, B. Su, K. Sun, L. Chen, D. Han, S. Wang, and L. Jiang, “Hydrophobic Interaction-Mediated Capture and Release of Cancer Cells on Thermoresponsive Nanostructured Surfaces,” Adv. Mater. 25(6), 922–927 (2013). [CrossRef]
4. H. Jiang and B. Kaminska, “Scalable Inkjet-Based Structural Color Printing by Molding Transparent Gratings on Multilayer Nanostructured Surfaces,” ACS Nano 12(4), 3112–3125 (2018). [CrossRef]
5. J. S. Clausen, E. Højlund-Nielsen, A. B. Christiansen, S. Yazdi, M. Grajower, H. Taha, U. Levy, A. Kristensen, and N. A. Mortensen, “Plasmonic Metasurfaces for Coloration of Plastic Consumer Products,” Nano Lett. 14(8), 4499–4504 (2014). [CrossRef]
6. M. Qanbarian, A. Qasemian, and B. Arab, “Molecular dynamics simulation of enhanced heat transfer through conical Al/Cu nanostructures,” Comput. Mater. Sci. 180, 109710 (2020). [CrossRef]
7. T. Shao, F. Tang, L. Sun, X. Ye, J. He, L. Yang, and W. Zheng, “Fabrication of Antireflective Nanostructures on a Transmission Grating Surface Using a One-Step Self-Masking Method,” Nanomaterials 9(2), 180 (2019). [CrossRef]
8. K. Maghsoudi, R. Jafari, G. Momen, and M. Farzaneh, “Micro-nanostructured polymer surfaces using injection molding: A review,” Mater. Today Commun. 13, 126–143 (2017). [CrossRef]
9. S. Murthy, M. Matschuk, Q. Huang, N. K. Mandsberg, N. A. Feidenhans’l, P. Johansen, L. Christensen, H. Pranov, G. Kofod, H. C. Pedersen, O. Hassager, and R. Taboryski, “Fabrication of Nanostructures by Roll-to-Roll Extrusion Coating: Roll-to-Roll Fabrication of Nanostructures,” Adv. Eng. Mater. 18(4), 484–489 (2016). [CrossRef]
10. Y. Chen, “Applications of nanoimprint lithography/hot embossing: a review,” Appl. Phys. A 121(2), 451–465 (2015). [CrossRef]
11. M. H. Madsen and P.-E. Hansen, “Scatterometry-fast and robust measurements of nano-textured surfaces,” Surf. Topogr.: Metrol. Prop. 4(2), 023003 (2016). [CrossRef]
12. M. Kreuzer, G. L. Whitworth, A. Francone, J. Gomis-Bresco, N. Kehagias, and C. M. Sotomayor-Torres, “In-line metrology for roll-to-roll UV assisted nanoimprint lithography using diffractometry,” APL Mater. 6(5), 058502 (2018). [CrossRef]
13. G. Alexe, A. Tausendfreund, D. Stöbener, and A. Fischer, “Model-assisted measuring method for periodical sub-wavelength nanostructures,” Appl. Opt. 57(1), 92 (2018). [CrossRef]
14. G. L. Whitworth, A. Francone, C. M. Sotomayor-Torres, and N. Kehagias, “Real-time Optical Dimensional Metrology via Diffractometry for Nanofabrication,” Sci. Rep. 10(1), 5371 (2020). [CrossRef]
15. J. S. Madsen, L. H. Thamdrup, I. Czolkos, P.-E. Hansen, A. Johansson, J. Garnaes, J. Nygard, and M. H. Madsen, “In-line Characterization of Nanostructured Mass-produced Polymer Components Using Scatterometry,” J. Micromech. Microeng. 27(8), 085004 (2017). [CrossRef]
16. W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, eds., Numerical recipes in C++: the art of scientific computing (Cambridge University, 2002), 2nd ed.
17. T. A. Germer, H. J. Patrick, R. M. Silver, and B. Bunday, “Developing an uncertainty analysis for optical scatterometry,” in Metrology, Inspection, and Process Control for Microlithography XXIII, vol. 7272J. A. Allgair and C. J. Raymond, eds., International Society for Optics and Photonics (SPIE, 2009), pp. 255–265.
18. J. Endres, A. Diener, M. Wurm, and B. Bodermann, “Investigations of the influence of common approximations in scatterometry for dimensional nanometrology,” Meas. Sci. Technol. 25(4), 044004 (2014). [CrossRef]
19. M. Hammerschmidt, M. Weiser, X. G. Santiago, L. Zschiedrich, B. Bodermann, and S. Burger, “Quantifying parameter uncertainties in optical scatterometry using Bayesian inversion,” in Modeling Aspects in Optical Metrology VI, vol. 10330B. Bodermann, K. Frenner, and R. M. Silver, eds., International Society for Optics and Photonics (SPIE, 2017), pp. 8–17.
20. M. Wurm, T. Grunewald, S. Teichert, B. Bodermann, J. Reck, and U. Richter, “Some aspects on the uncertainty calculation in Mueller ellipsometry,” Opt. Express 28(6), 8108–8131 (2020). [CrossRef]
21. J. S. Madsen, S. A. Jensen, L. Nakotte, A. Vogelsang, L. H. Thamdrup, I. Czolkos, A. Johansson, J. Garnæs, T. Nielsen, J. Nygård, and P. E. Hansen, “Scatterometry for optimization of injection molded nanostructures at the fabrication line,” Int. J. Adv. Manuf. Technol. 99(9-12), 2669–2676 (2018). [CrossRef]
