Optical analysis of 3D-printed terahertz waveplates from common thermoplastics Authors: Gabriel F Freitas, Lucas T Roque, Nicolas M Kawahala, Felix G G Hernandez Published: 2025 ABSTRACT AbstractWe report the fabrication and characterization of birefringent terahertz (THz) waveplates produced via fused deposition modeling (FDM) 3D printing using three widely available thermoplastic filaments: polylactic acid (PLA), acrylonitrile butadiene styrene (ABS), and high-impact polystyrene (HIPS). Form-birefringent subwavelength gratings were printed using a 0.2 mm nozzle to tailor the phase retardance. The intrinsic THz optical properties of each polymer were first assessed through terahertz time-domain spectroscopy (THz-TDS) on unstructured slabs, revealing material-dependent absorption. Polarization-resolved THz-TDS was then employed to extract the effective birefringence, retardance, and polarization modulation of the fabricated waveplates. HIPS exhibited the lowest absorption and the most stable birefringent response, followed by ABS, while PLA showed strong absorption and limited polarimetric performance. Measured birefringence spectra agreed with effective medium predictions below the diffraction threshold (0.6 THz); beyond this range, retardance saturation enabled continued phase shifting. Full polarimetric analysis using Stokes parameters confirmed consistent modulation of ellipticity and polarization rotation, particularly in HIPS-based devices. These results provide a comparative framework for selecting printable materials and optimizing designs of low-cost, customizable THz polarization components.Keywords: Terahertz time-domain spectroscopy (THz-TDS), 3D printing, Birefringent waveplates, Form-birefringence, Polarization control, Fused deposition modeling (FDM)Subject terms: Materials science, Optics and photonics Introduction Fused-deposition modeling (FDM) 3D printing1 is a widely used additive manufacturing technique that has been increasingly employed for the rapid and cost-effective fabrication of optical components in the terahertz (THz) spectral range2. Common polymers compatible with FDM exhibit low absorption and moderate refractive indices below 1 THz3–6, while consumer-grade FDM printers typically offer spatial resolutions on the order of hundreds of microns7. This level of precision enables the production of subwavelength structures essential for manipulating THz waves. These advantages have led to the realization of a wide range of 3D-printed photonic devices—including waveguides8–10, diffraction gratings11–13, prisms14–16, lenses17–19, and waveplates20–22—capable of tailoring THz transmission, polarization, and wavefronts. Among these, polarization control of THz waves has gained particular attention due to its importance in polarization-sensitive terahertz time-domain spectroscopy (THz-TDS)23–25, imaging26–28, and emerging communication technologies29. While many THz-TDS setups rely on combinations of wire-grid polarizers (WGP) to decompose and analyze the polarization state30–32, the development of broadband THz waveplates remains a technological challenge33,34. Various materials—including quartz35,36, sapphire37, and silicon38,39—and diverse fabrication methods have been explored. However, such approaches often involve costly or time-consuming processes. In this context, FDM 3D printing offers a low-cost and versatile alternative for fabricating form-birefringent THz waveplates, with recent demonstrations achieving both quarter- and half-wave retardance using periodic subwavelength structures20–22. Nonetheless, evaluating the optical performance of waveplates produced from different filament materials remains essential to assess their suitability, printability, and structural fidelity. While previous demonstrations of 3D-printed THz waveplates have focused on single-material implementations or alternative grating architectures—such as super-wavelength metagratings in polylactic acid (PLA)20, subwavelength cyclic olefin copolymer (COC) structures fabricated near the resolution limit of FDM21, and stacked high-impact polystyrene (HIPS) gratings designed for achromatic performance22—a systematic assessment across multiple common FDM thermoplastics remains unavailable. Our contribution aims at filling this gap by establishing a direct, material-resolved comparison within a uniform subwavelength geometry, clarifying how differences in absorption, birefringence, and fabrication fidelity impact THz waveplate performance. In addition, we employ polarimetry mapping to visualize polarization modulation across frequency and orientation, providing a compact framework to identify operational regimes and deviations from effective-medium behavior. In this work, we demonstrate the fabrication and characterization of THz waveplates using FDM 3D printing with three commonly available thermoplastic filaments: PLA, HIPS, and acrylonitrile butadiene styrene (ABS). By combining subwavelength structural design with polarization-resolved THz-TDS, we evaluate the effective birefringence, retardance, and polarimetric response of each printed waveplate. The results reveal material-dependent performance trends, with HIPS exhibiting consistently low absorption and stable polarization modulation over a broad frequency range, supporting its suitability for THz applications in agreement with prior studies. This comparative study provides a practical framework for selecting and implementing 3D-printed birefringent optics in low-cost, customizable THz devices. Methods Waveplate design We adopt a form-birefringent waveplate design based on a one-dimensional subwavelength grating composed of alternating low- () and high-refractive-index () regions40, with respective widths and . The grating period is , and the corresponding fill factors are and . For wavelengths much larger than the period (), the structure behaves as a uniaxial anisotropic medium, and its optical response can be described using an effective medium approximation41. In this regime, light polarized parallel and perpendicular to the grating lines experiences distinct effective refractive indices, given by Here, the refractive indices are treated as real-valued (lossless), following the standard effective-medium approximation used for low-loss form-birefringent gratings41. This assumption is justified because, for the polymers studied here, the imaginary part of the refractive index is small compared to the real part over the sub-THz range (< 1 THz), as reported in Ref.3. To account for non-idealities at finite frequencies, second-order corrections are introduced, yielding the extraordinary () and ordinary () refractive indices38,41: which correspond to the principal axes of the effective birefringent medium. The resulting birefringence is then defined as . For a given phase retardance at frequency , the required waveplate thickness h is given by21: where c is the speed of light in vacuum. To maintain the validity of the approximation, the grating period must remain subwavelength. A practical upper frequency limit, above which diffraction becomes significant and the model breaks down, can be estimated by: For , corresponding to the minimum period attainable with 0.2 mm-resolution FDM printers, and a typical filament refractive index of , the resulting cutoff frequency is approximately 0.5 THz. 3D printing and materials To fabricate the waveplates, we employed a consumer-grade FDM 3D printer (Creality Ender 3 S1 Pro), known for its affordability and widespread use in desktop manufacturing. Three thermoplastic filaments were investigated (PLA, ABS, and HIPS) based on their availability, ease of printing, and distinct terahertz optical properties. These materials are summarized as follows: PLA42–44 is a biodegradable polyester derived from renewable sources such as corn starch or sugarcane. It is widely used in consumer-grade 3D printing due to its ease of use, relatively high thermal capacity, and non-toxic emissions during extrusion. However, it tends to be brittle, has limited heat resistance, and may be unsuitable for applications demanding long-term mechanical or thermal stability. PLA42–44 is a biodegradable polyester derived from renewable sources such as corn starch or sugarcane. It is widely used in consumer-grade 3D printing due to its ease of use, relatively high thermal capacity, and non-toxic emissions during extrusion. However, it tends to be brittle, has limited heat resistance, and may be unsuitable for applications demanding long-term mechanical or thermal stability. ABS45–47 is a petroleum-based thermoplastic composed of acrylonitrile, butadiene, and styrene. It is popular in FDM for its mechanical strength, impact resistance, and relatively high thermal stability. Nevertheless, it is prone to warping during cooling due to thermal contraction, which can cause dimensional inaccuracies and interlayer delamination. Mitigating these effects typically requires a heated bed, an enclosed build chamber, and proper surface adhesion. ABS45–47 is a petroleum-based thermoplastic composed of acrylonitrile, butadiene, and styrene. It is popular in FDM for its mechanical strength, impact resistance, and relatively high thermal stability. Nevertheless, it is prone to warping during cooling due to thermal contraction, which can cause dimensional inaccuracies and interlayer delamination. Mitigating these effects typically requires a heated bed, an enclosed build chamber, and proper surface adhesion. HIPS48,49 is a copolymer of polystyrene and polybutadiene rubber, characterized by low water absorption and high impact resistance. Notably, it exhibits markedly lower THz absorption in the THz range compared to other common FDM filaments4, making it a promising candidate for terahertz optics. Similar to ABS, HIPS may require a controlled printing environment to minimize warping and cracking during cooling. HIPS48,49 is a copolymer of polystyrene and polybutadiene rubber, characterized by low water absorption and high impact resistance. Notably, it exhibits markedly lower THz absorption in the THz range compared to other common FDM filaments4, making it a promising candidate for terahertz optics. Similar to ABS, HIPS may require a controlled printing environment to minimize warping and cracking during cooling. To in-house assess the intrinsic terahertz optical response of the filaments, we first printed structure-less bulk slabs (25.4 mm diameter, 2 mm thickness) using the default 0.4 mm nozzle. The refractive indices extracted from these samples guided the waveplate design. Subsequently, form-birefringent waveplates were fabricated by printing the subwavelength grating structures directly onto identical base slabs as a single continuous piece, employing a finer 0.2 mm nozzle to achieve a grating period of with equal fill factors () of alternating air and polymer regions. All printed structures used full rectilinear infill to minimize porosity and ensure optical uniformity. Printing parameters were carefully optimized for each material. PLA was the simplest to process, showing reliable extrusion and strong layer adhesion at moderate temperatures (/, nozzle/bed) without requiring special handling. ABS required a more controlled environment to prevent delamination, so printing was performed inside a simple enclosure to maintain ambient heat, with elevated temperatures (/) to reduce warping during cooling. Although HIPS did not demand temperatures as high as ABS (/), it presented the greatest challenge, requiring meticulous bed leveling and the application of an adhesive spray to ensure first-layer adhesion and prevent print detachment. All samples were printed at a uniform speed of 25 mm/s, and grating heights were tailored according to design equations to introduce nominal quarter-wave retardance around 0.15 THz, yielding , , and . This frequency was chosen so that the quarter-wave condition would not lie too close to the lower band edge, while the corresponding half-wave operation remained well within the usable range and below half of the cutoff frequency. A photograph of the fabricated waveplates is shown in Fig. 1a. Additionally, Fig. 1b and c show cross-sectional and top-view micrographs of the fabricated PLA waveplate, respectively, highlighting the printed grating geometry and its key parameters. a Photograph of the fabricated waveplates. b Cross-section micrograph of the PLA waveplate; h denotes the grating height, and is the period composed of the filament width and the air gap . c Top-view micrograph of the same waveplate. d Schematic diagram of the THz-TDS setup in transmission geometry with a configuration of wire grid polarizers (WGP) enabling polarization sensitive measurements. THz polarimetry measurements Polarization-resolved THz-TDS32 was performed in transmission geometry using photoconductive antennas (PCA) as both emitter and detector. A mode-locked Ti:sapphire laser provides 100 fs pulses at 780 nm and 80 MHz repetition rate to pump the emission and gate the detection of broadband THz pulses spanning range 0.1 THz to 2 THz. The THz beam is routed using off-axis parabolic mirrors, and a set of three wire-grid polarizers (WGP) was employed to resolve the polarization state of the transmitted THz electric field waveform E(t)30. A simplified schematic of the setup is shown in Fig. 1d. To extract the orthogonal polarization components of the transmitted field, WGP1 fixes the incident THz polarization along the laboratory x-axis, while WGP2 acts as a rotatable analyzer oriented at relative to WGP1. WGP3 is aligned parallel to WGP1 to maintain consistent detection polarization. Two time-domain signals, and , are recorded with WGP2 set to and , respectively. The horizontal and vertical components of the transmitted field are then reconstructed as and 31. These relations arise exactly from the successive projections implemented by WGP2 and WGP332. The sample was mounted on a rotational stage and measurements were taken at azimuthal angles from 0 to 360 in 10 increments. All measurements were acquired using a common time-zero reference, i.e., the temporal origin was not reset between rotation angles , ensuring consistent relative phase between the reconstructed field components. Fourier transformation of these time-domain signals yields the complex spectra and , enabling full characterization of the waveplate’s polarimetric response, including frequency-dependent polarization rotation and ellipticity31. Results THz optical characterization of 3D-printed polymers To evaluate the THz optical response of the materials employed in our waveplate fabrication, we performed THz-TDS measurements on 2 mm-thick, unstructured slabs of PLA, ABS, and HIPS. Although the THz properties of these polymers have been previously reported2–6, refractive index and absorption spectra can vary substantially between filament brands and batches. We therefore characterized the specific filaments used in our fabrication to ensure accurate modeling and enable a direct comparison of their optical performance. Figure 2 shows the optical coefficients extracted by comparing the THz-TDS response of each slab to a sample-free reference, using standard thick-sample transmission analysis50. Measured waveforms are presented at Supplementary Information51. In this procedure, Fabry–Perot reflections were removed by truncating the time-domain waveform prior to the first internal echo (without apodization), consistent with the windowing routine discussed in Ref.52. The phase was unwrapped using the standard cumulative unwrapping algorithm, and refractive index and absorption spectra were obtained from the truncated sample/reference ratio through the bulk slab approximation50. Optical properties of 2 mm-thick 3D-printed polymer slabs measured by THz-TDS. a Refractive index, and b absorption coefficient spectra for PLA (blue triangles), ABS (red squares), and HIPS (green circles) over the frequency range 0.1–1.0 THz. Error bars in a reflect the propagated thickness uncertainty; uncertainties in b are not shown because they are dominated by amplitude noise and are much smaller relative to the plotted values. Dashed lines are guides to the eye. Panel (a) displays the refractive index spectra below 1 THz for PLA (blue triangles), ABS (red squares), and HIPS (green circles), with dashed lines added to guide the eye. All three materials exhibit nearly flat dispersion, with average values of , , and . The uncertainties reported for the refractive indices arise from propagating the micrometer thickness error of the slabs, which is the dominant source of phase uncertainty in thick-sample THz-TDS extraction. Panel (b) presents the corresponding absorption coefficient spectra . PLA shows the highest absorption, rising from near zero to approximately 38 ; ABS displays intermediate losses, reaching about 13 ; and HIPS maintains low, nearly flat absorption across the band, with , except for an anomalous rise above 0.9 THz, which is outside the main operating frequency range considered here. These results highlight HIPS as the most favorable material among those tested for THz optical components, combining a relatively low refractive index with minimal absorption over the frequency range of interest. Although PLA and ABS are more commonly used in 3D printing, their significantly higher THz losses limit their applicability in transmissive optics across a broad frequency range. A comparison with literature data further emphasizes the variability in reported optical parameters: for instance, Brodie et al.3 report refractive indices of 1.54 for PLA, 1.47 for ABS, and 1.48 for HIPS, which differ slightly but meaningfully from the values measured here, and further deviate from the data compiled by Sun and Hu in Ref.2. Table S1 in the Supplementary Information51 summarizes these literature values alongside our measurements. Such discrepancies, though seemingly minor, can affect the design and performance of form-birefringent waveplates. Therefore, in-house characterization of the specific filament used is crucial to ensure accurate device modeling and reliable optical performance. Retardance and effective birefringence of fabricated waveplates Building on the refractive index data obtained from bulk slabs, we now evaluate the birefringent response of the waveplates fabricated from each material. By performing THz-TDS measurements with the electric field aligned along the principal axes of the subwavelength grating, we extract the frequency-dependent birefringence and resulting retardance. These quantities provide a direct measure of the effectiveness of each waveplate in introducing controlled phase shifts between orthogonal polarization components. The phase spectra used in this analysis were obtained using the same procedure applied to the bulk slabs, namely from time-domain signals truncated before the first Fabry–Pérot echo. The birefringence was computed from the unwrapped phase difference between the principal axes using . Figure 3 summarizes these results. Panels (a–c) show the frequency-dependent birefringence for waveplates made from HIPS, ABS, and PLA, extracted as the difference between the refractive indices retrieved along the two principal axes. The solid lines represent predictions from the effective medium model used during the design stage (Eqs. 3, 4). Good agreement is observed at low frequencies, while deviations emerge above 0.6 THz, marking the cutoff beyond which the subwavelength approximation is no longer valid. Birefringence and retardance of 3D-printed THz waveplates. a–c THz-TDS measured effective birefringence spectra for HIPS, ABS, and PLA waveplates, compared with the effective medium model (solid lines). d–f Corresponding retardance calculated from measured birefringence and grating height. Solid lines show linear fits up to the 0.6 THz cutoff. Quarter-waveplate operation occurs near 0.15 THz (HIPS, PLA) and 0.1 THz (ABS), with half-waveplate behavior at roughly twice these frequencies. Panels (d–f) show the corresponding retardance spectra , calculated from the thickness and extracted birefringence of each waveplate using Eq. (5). The curves follow a linear trend up to 0.6 THz, beyond which deviations emerge. Linear regressions fitted up to this cutoff frequency are overlaid for reference. The results indicate that quarter-waveplate operation () is achieved at 0.15 THz for both HIPS and PLA, and at 0.1 THz for ABS. As expected, half-waveplate behavior () occurs at approximately twice these frequencies. Remarkably, HIPS and ABS waveplates exhibit retardance saturation near above the cutoff, suggesting a degree of broadband quarter-waveplate-like functionality. This saturation is maintained up to 0.8 THz for ABS and extends to 1 THz for HIPS. Polarimetric response of the fabricated waveplates We now assess the polarization-modifying performance of the fabricated waveplates through THz polarimetry. By measuring the transmitted THz electric field at multiple waveplate orientations , we applied the Stokes formalism21,31,53 to extract the output polarization ellipticity and rotation angle spectra. These quantities are presented as false-color maps in Fig. 4. Further details on the Stokes analysis, along with the measured time-domain waveforms used in this evaluation, are provided in the Supplementary Information51. Polarimetric response of 3D-printed THz waveplates. a–c Ellipticity angle as a function of frequency and sample orientation for HIPS, ABS, and PLA waveplates, extracted from polarization-resolved THz-TDS using Stokes analysis. Periodic modulation with indicates birefringent behavior. d–f Corresponding polarization rotation angle , defined as the angle between the input polarization and the major axis of the transmitted polarization ellipse. The data have been unwrapped for continuity. Ellipticity maps are shown in the top row of Fig. 4 (panels a–c) for HIPS, ABS, and PLA, respectively. This parameter quantifies the degree of circular polarization imparted by the waveplates. Ideal waveplate behavior is characterized by a periodic modulation of as a function of , with the amplitude and symmetry of the pattern determined by the frequency-dependent retardance. In the maps, the white contour lines indicate , corresponding to linear polarization states. Along the frequency axis, these contours align with the half- and full-wave retardance conditions identified in Fig. 3 (panels d–f). Along the axis, they appear when the incident THz polarization is aligned with the waveplate’s principal axes. Among the three materials, HIPS exhibits the most regular and symmetric modulation pattern, consistent with a well-defined birefringent response. Alternating red and blue bands appear at fixed intervals across a broad frequency range, indicative of consistent retardance and a stable phase relationship between orthogonal polarization components. ABS shows qualitatively similar ellipticity modulation patterns but with increased distortion at higher frequencies, likely due to absorption-driven reductions in signal-to-noise ratio. In contrast, PLA displays irregular and noisy ellipticity maps, with distorted and less pronounced periodic modulation even at lower frequencies, reflecting its comparatively poorer optical performance in the THz range. Likewise, rotation angle maps are shown in the bottom row of Fig. 4 (panels d–f), depicting the polarization ellipse rotation as a function of frequency and waveplate orientation angle . The data were numerically unwrapped to enhance visualization, particularly to highlight the linear dependence characteristic of half-waveplate operation. At lower frequencies, all materials exhibit smooth and continuous variation of with , indicative of stable polarization rotation. Distinct bands appear along the frequency axis, reflecting transitions between different retardance regimes. Below the quarter-waveplate frequency, remains near zero across all , consistent with minimal polarization rotation. Between the quarter- and three-quarter-waveplate regimes, the maps exhibit a pronounced linear trend, indicative of effective half-waveplate operation. At higher retardances (above , including the full-waveplate condition) once again approaches zero, as the polarization returns to its initial orientation. Similar modulation patterns recur above the cutoff frequency, particularly for HIPS and ABS, though with increased distortions and reduced smoothness, reflecting a breakdown of ideal waveplate behavior in this regime. Discussion We conducted a comparative study of FDM 3D-printed THz waveplates fabricated from PLA, ABS, and HIPS, using form-birefringent subwavelength gratings and polarization-resolved THz-TDS to assess their birefringence, retardance, and overall polarimetric response. Ellipticity and rotation angle maps, plotted as functions of frequency and waveplate orientation, offered a compact and intuitive visualization of polarization modulation, clearly revealing narrowband regimes of quarter- and half-waveplate behavior within the 0.1 THz to 0.3 THz interval. These operational bands reflect the specific grating heights used in fabrication, which directly govern the phase delay and can be systematically tuned by controlling the printed thickness. Despite similar refractive indices near 1.5, the three materials exhibited markedly different behavior due to variations in THz absorption, as expected from literature. PLA suffered from strong losses that degraded signal quality and compromised reliable polarization modulation above approximately 0.4 THz. ABS performed better, sustaining effective polarization control up to around 0.8 THz. HIPS showed the most consistent performance, combining low absorption with stable birefringence and well-defined polarization modulation throughout the measured bandwidth from 0.1 THz to 1 THz. Above 0.6 THz, all waveplates deviated from effective medium predictions, and both ABS and HIPS exhibited retardance saturation near , indicating continued phase shifting beyond the design range that may be relevant for extended applications. This deviation is consistent with the known limits of the effective-medium approximation, which requires subwavelength grating period (), moderate index contrast, and negligible diffraction; as leads to approaching unity near 0.6 THz, the accuracy of the approximation is reduced. We note that the birefringence spectra extracted from the waveplates closely follow the effective-medium predictions, with no additional frequency-dependent terms that would suggest infill-induced form birefringence. Likewise, the optical characterization of the bulk slabs showed no excess loss or spectral irregularities beyond the expected material absorption, indicating that neither the rectilinear infill pattern nor printing-related issues such as warping or layer gaps produced measurable scattering or coherence degradation in the THz range below the designed cutoff. Alongside performance, fabrication considerations offer valuable insights. While PLA shows poorer THz performance, it remains the easiest to print reliably and can function adequately at the lower end of the frequency band. For instance, Rohrbach et al.21 and Dong et al.20 have demonstrated promising PLA waveplates using both sub- and super-wavelength designs. Although waveplates fabricated from ABS have not been reported to our knowledge, ABS presents better THz performance than PLA despite printing challenges, making it a potential candidate for specific applications. Among the materials tested, HIPS stands out for its low absorption and superior performance, enabling broadband designs as shown by Jäckel et al.22 using larger structures printed with a 0.4 mm nozzle. Building on this, our work demonstrates finer gratings enabled by a 0.2 mm nozzle—despite increased printing difficulties such as extrusion clogging—which extend the usable frequency range. A concise summary of the material-dependent performance is provided in Table 1 to support direct comparison among PLA, ABS, and HIPS. Table 1Summary of material-dependent performance of 3D-printed THz waveplates.MaterialTHz absorption (Fig. 2)Retardance & birefringence (Fig. 3)Polarimetry stability (Fig. 4)Practical considerationsPLAHighest: rises to Largest deviations; limited useful regimeIrregular modulation; noise at higher frequenciesVery easy to print; non-toxic; biodegradable (unsuitable for long-term devices)ABSIntermediate: up to Linear up to ; saturation to Moderate distortions at higher frequenciesHigh strength; impact-resistant; prone to warping; emits toxic fumes when printingHIPSLowest: less than Most consistent; retardance stable up to 1 THzMost regular, symmetric modulationLow water absorption; mechanically stable; can warp without a controlled enclosure Summary of material-dependent performance of 3D-printed THz waveplates. Together, these findings provide a practical framework for selecting and fabricating 3D-printed birefringent THz optics with tailored spectral and polarimetric characteristics, highlighting pathways to optimize performance through material choice, structural design, and printing parameters. Supplementary Information Supplementary Information. Author contributions N.M.K. and F.G.G.H. conceived and supervised the project. L.T.R. developed the procedure and fabricated the samples. G.F.F. performed the data analysis and generated all figures, with input from N.M.K. G.F.F. and N.M.K. contributed to data interpretation and contextualization, and wrote the manuscript with input from F.G.G.H. All authors discussed the results and reviewed the manuscript. Funding We acknowledge financial support from the São Paulo Research Foundation (FAPESP), Grants Nos. 2021/12470-8, 2023/04245-0 and 2023/11158-6, and from Grant Nos. 409245/2022-4 and 306550/2023-7 of the National Council for Scientific and Technological Development (CNPq). L.T.R. acknowledges support from the Programa Institucional de Bolsas de Iniciação Científica (PIBIC) of the CNPq. Data availability All data that support the conclusions of this work are contained within the article. Correspondence and requests for materials should be addressed to Felix G. G. Hernandez. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-32267-5. References Associated Data Supplementary Materials Supplementary Information. Data Availability Statement All data that support the conclusions of this work are contained within the article. Correspondence and requests for materials should be addressed to Felix G. G. Hernandez.