Innovative soft magnetic multilayers with enhanced in-plane anisotropy and ferromagnetic resonance frequency for integrated RF passive devices

The imperative for lighter, faster and “smarter” mobile devices, combined with the expected exponential growth in wireless data transfer toward 2020 and beyond, due to the rise of the Internet of Things (IoT) paradigm shift and 5G networks, puts device manufacturers under enormous pressure. Innovative technologies, new materials and device architectures that enable higher performance, increased functional diversity at the chip and system levels, and lower power consumption are therefore more vital now than ever before. For the time being, the Moore’s scaling law¹ continues to drive the conventional silicon complementary metal-oxide-semiconductor (Si-CMOS) technology deep into the nanoscale by breathtaking advances in integrated circuit (IC) miniaturization and novel system-on-chip (SoC) architectures.² 

Lately, however, users increasingly require portable systems that combine computing, communication, power and sensing functionalities in highly miniaturized packages, and this implies that heterogeneous technologies such as digital, radio frequency (RF), analog, power modules, microelectromechanical systems (MEMS), sensors, biochips and optics need to be integrated monolithically with the CMOS platform.³ While various sensors, transducers and actuators made using MEMS fabrication processes and high performance electronic and optoelectronic devices based on other semiconducting materials with electrical/optical properties beyond those of Si (e.g. Ge, GaAs, GaN, SiC, etc.) are progressively being integrated with Si CMOS,⁴ further miniaturization of power supplies and RF modules remains an important goal.

The obvious path to achieve these objectives would be to integrate these electronic circuits on a Si chip, reducing their size and weight, increasing yields and therefore reducing overall costs. In doing so, we could also improve reliability and performance, but since the scaling laws of the passive circuit elements (e.g. resistors, capacitors, inductors, transformers) cramped inside the power supply circuits and RF devices do not follow the traditional Moore’s law, reducing the overall size of these devices is not an easy task. For example, the major problem of on-chip inductors (typically planar, tightly wound, square shaped spiral inductors), used currently in various electronic devices, is their large chip area utilization (inductance densities < 100 nH/mm²).⁵ For instance, Intel states that the on-chip inductors in their DC-DC converters used for power management in multi-core processors occupy approximately a quarter of the total available chip area, making them rather costly components,⁶ illustrating the urgent need to shrink these components for the next generation mobile devices.

Due to their flux amplification properties and high operating frequencies, integrated thin film magnetic cores with high permeability promise further device miniaturization, lower energy loss and thus lower power operation. Figure 1 shows the schematics of an integrated micro-transformer with a magnetic core that is magnetically very efficient compared to other designs (e.g. planar).^7,8 Despite involving a relatively complicated process flow, e.g. precise alignment between various layers and formation of numerous vias for contacting the top and bottom electrode layers so that they wrap around the magnetic core, such a design can be realized by state-of-the-art MEMS and chip fabrication processes. The magnetic core material needs to have a high relative permeability (μ_r), since the inductance (L) of magnetic core inductors (and corresponding solenoids associated with magnetic core micro-transformers) is proportional to μ_r. Moreover, a high corresponding ferromagnetic resonance (FMR) frequency (f₀) is required for RF applications, since μ_r ∼ 1 for operation beyond f₀. On the other hand, since the product μ_r·f₀ is proportional to the saturation magnetization (4πM_s),⁹ keeping both high μ_r and f₀ at the same time is very difficult. Additionally, magnetic core materials should exhibit low coercivity (H_c) and have large electrical resistivity (ρ). Finally, the magnetostriction (i.e the property of ferromagnetic materials that causes them to change their shape or dimensions during the process of magnetization) should be kept as low as possible.

Owing to their large saturation magnetization (e.g. ∼1.5-2 T) and low hysteresis loss, soft magnetic layers based on amorphous alloys, e.g. FeCoB and CoTaZr, currently receive great attention for their potential for GHz frequencies applications.^10–12 While these amorphous materials have ρ in the range 100-130 μΩ·cm, thus larger than that of polycrystalline metals (e.g. for Permalloy ρ ∼ 20 μΩ·cm¹³), it is not large enough for high frequency operation due to the eddy current loss. We can overcome this impediment by laminating the magnetic material with intermediate, non-magnetic interlayers, such as Al₂O₃, AlN, SiO₂, Ta₂O₅, etc., which not only helps to reduce significant eddy currents, but also to improve the magneto-static coupling between adjacent magnetic layers that are mandatory for device performance at high frequencies.¹² 

In this work, we investigate the ferromagnetic properties of soft magnetic thin films based on the amorphous alloys FeCoB and CoTaZr, which are laminated with thin Al₂O₃ dielectric interlayers. The ferromagnetic layers may consist of single FeCoB or CoTaZr layers when a single target is used, or a fine structure comprising alternating FeCoB and CoTaZr nanolayers when two targets are employed (substrate cage rotates continuously such that the substrates face different targets alternatively). By changing the cage rotation speed and sputter power of the individual stations, we tune the thickness of individual FeCoB and CoTaZr nanolayers, which provides us with additional means to tailor the coercivity, magnetostriction, and anisotropy field (H_k) for high frequency applications. While this work is limited to the case of FeCoB and CoTaZr soft magnetic alloy materials, our method could be applied to more than two materials, and other material combinations, including magnetic/nonmagnetic ones.

We prepared FeCoB/Al₂O₃, CoTaZr/Al₂O₃ and (FeCoB + CoTaZr)/Al₂O₃ multilayers by dynamic sputtering onto continuously moving 8” Si/200nm-thermal-SiO2 wafers in a high-throughput Evatec LLS EVO II sputter system (base pressure < 10^-8 mbar).^11,12 The system is equipped with five cathodes (e.g. PM1-PM5), a moveable shutter and a rotating substrate cage that allows up to nine 8” wafers to be processed in the same batch. (Fig. 2a). Since the substrate cage rotates continuously around its axis, each time a wafer passes in front of a running cathode a very thin layer is deposited on the wafer when the shutter is open. Depending on the sputter power and cage rotation speed, the thickness of this layer can vary from a few tenths of nm up to several nm. Thus, a thin film gradually builds up from these nanolayers, the total thickness of which is controlled by choosing the number of cage rotations. When only a single cathode (e.g. FeCoB, CoTaZr, Al₂O₃) operates at a time during the cage rotation, the thin film has the same composition throughout its thickness. Alternatively, if more than one cathode (e.g. FeCoB and CoTaZr) is used, then the deposited thin film consists of a nano-layered structure (e.g. FeCoB + CoTaZr) (Fig. 2b).

The multilayers studied in this work consist of four period bilayer stacks with alternating 80 nm thick magnetic layers and 4 nm thick Al₂O₃ dielectric interlayers. The FeCoB, CoTaZr and (FeCoB + CoTaZr) magnetic layers were deposited by DC or pulsed DC sputtering (duty cycle 40%) at a pressure of 1.7×10^-3 mbar using Fe-28%Co-20%B (at.%) and Co-4.5%Ta-4%Zr long life (∼250 kW·h) targets, whereas the Al₂O₃ interlayers were deposited by RF sputtering from monoblock Al₂O₃ targets at a pressure of 5×10^-3 mbar. To deposit the 80 nm thick FeCoB and CoTaZr layers, only one cathode was operated at a time (e.g. either PM1 for FeCoB, or PM3 for CoTaZr), while the 80 nm thick FeCoB + CoTaZr layers were deposited using both PM1 and PM3 cathodes simultaneously (Fig. 3). Sputter powers and nominal layer thicknesses of the multilayers studied in this work are presented in Table I.

We introduced the in-plane magnetic anisotropy in these multilayered thin films during sputtering by a linear magnetic field parallel to the wafer plane, which is designed such that the magnetic field of the magnetron located behind the opposite target is not perturbed. Moreover, being located inside the motor housing, and therefore outside the vacuum chamber, this aligning field system can be easily placed or removed without breaking vacuum.

Layer thickness and surface/interface roughness of the sputtered multilayers were determined by means of the X-ray reflectometry (XRR) technique (see Fig. 4a). The XRR specular scans were measured using a laboratory-built diffractometer equipped with a Cu X-ray tube, a graded parabolic multilayer mirror with the angular aperture of 0.028° on the primary arm as collimator and monochromator, and a reversed parabolic multilayer mirror as a focusing element after the sample placed in front of the scintillation detector slit (aperture 0.08 mm) in order to reduce the background and increase the dynamic range. The degree of crystallinity was probed by a glancing angle X-ray diffraction (GAXRD) technique (Fig. 4b). The GAXRD spectra, consisting of detector two-theta (2θ) scans for an incidence angle ω = 6°, were measured with the laboratory diffractometer above, equipped with a Cu X-ray tube and a parabolic graded multilayer mirror with angular aperture of ∼0.028° on the primary arm, and a highly oriented pyrolitic graphite (HOPG) monochromator with a soller slit (angular aperture of 0.8°) in front of the scintillation detector. The size of the incidence beam on the sample was 10 mm × 1.5 mm. The Scherrer equation was used to estimate the grain size (D), D = 0.95λ/(wcosθ), where λ = 0.154 nm is the wavelength of the Kα line of the X-rays generated by the Cu tube, w is the width of the diffraction peak (i.e. full-width at half maximum, FWHM) in radians, corrected by the resolution function.¹⁴ 

Surface morphology and root-mean-square roughness (R_q) of the multilayers was investigated by means of atomic force microscopy (AFM) using a Park Systems NX20 microscope. The resistivity of the soft magnetic single layers was determined using a KLA-Tencor OmniMap RS100 four-point probe resistivity mapping system.

The distributions of magnetic properties (H_c, H_k, the angular dispersion of EA) on the 8” wafers were measured up to a radius of 70 mm by means of the magneto-optic Kerr effect (MOKE) method using a laboratory-built instrument with a laser diode (wavelength 632.8 nm). Moreover, the static magnetic properties (H_c, H_k, M_s) averaged over the entire 8” wafers were obtained from hysteresis (BH) loops using a Shb Instruments MESA 200 B-H looper measurement system. This set up was also used to gain insight on the magnetostriction state of the soft magnetic multilayers investigated in this work as follows. First, we measured the hysteresis loop of the as-deposited wafer from which we determined the corresponding anisotropy field (H_k). Then, we placed the wafer between a set of ceramic knives, and subsequently we slightly bent the wafer by applying an external force. At that point, we measured another (BH) loop in this stressed state, from which we determined the new anisotropy field (⁠$H k *$⁠). Consequently, we obtained the change of anisotropy field (⁠$Δ H k = H k * − H k$⁠) when the sample was strained, which was caused by the inverse magnetostriction effect originated from the in-plane anisotropic stress (ε) induced by the wafer bending.¹⁵ This change is a measure of the magnetostriction state of the soft magnetic multilayer, since the saturation magnetostriction (λ_s) is related to ΔH_k by the following expression: λ_s = μ₀M_sΔH_k(1 + ν)/3εY_m, where ν and Y_m are the Poisson’s ratio and Young’s modulus of the film, and M_s is the saturation magnetization.¹⁶ Finally, magnetic hysteresis loops were measured at room temperature by a vibrating sample magnetometer (Lakeshore 735 VSM) on 4 mm × 4 mm samples. The high frequency response of the soft magnetic multilayers was investigated from 100 MHz to 10 GHz with an RF permeameter based on the single coil technique.¹⁷ Dynamic permeability was measured along the hard axis of the film without any external field using the solenoid perturbation method.

Figure 5 compares the specular XRR scans for the soft magnetic multilayers I-IV, where the intensity oscillations with different frequencies stem from individual thickness periodicities within the samples. It can be seen that all reflectivity curves exhibit a fast oscillation with a period of 0.055°, corresponding to the thickest periodic sequence in the multilayer, which is the magnetic/non-magnetic bilayer stack. Moreover, all curves exhibit a slow oscillation with a period of 0.9°, corresponding to the thinnest layer in the magnetic/non-magnetic bilayer stack, e.g. Al₂O₃. The reflectivity curves corresponding to the multilayers III and IV exhibit in addition an even slower oscillation with a periodicity of ∼2.4°, corresponding to the shortest periodically repeated sequence in these multilayers, which is the FeCoB/CoTaZr bilayer.

We determined the individual layer thickness and the interface roughness by fitting the specular XRR scans with a model of a periodic multilayer that assumes the layer roughness to increase from the bottom to the top surface. The initial values of the refraction indexes of various layers were taken from Ref. 18, and the specular reflectivity were simulated using a matrix formalism.¹⁹ Multilayers I and II were modeled by FeCoB/Al₂O₃ and CoTaZr/Al₂O₃ bilayer stacks with four periods, whereas the multilayers III and IV were modeled by X/Al₂O₃ bilayer stacks with four periods, where the ferromagnetic layers X consist of FeCoB/CoTaZr bilayer stacks with 40 and 39 periods, respectively. The layer thickness and interface roughness obtained from the fits are presented in Tables II and III. Thus, the layer thicknesses from XRR are very close to the nominal values in Table I. Moreover, the interface roughness for all multilayers is very low, of the order of several tenths of nm. Thanks to the very low interface roughness, all periodic features within the multilayer stack could be detected, including the very thin magnetic FeCoB/CoTaZr nano-bilayers. As expected, the highest roughness values were detected in the complex multilayers (e.g. III and IV) exhibiting a nanolayered structure, since these samples have significantly more interfaces than the multilayers I and II owing to the large number of FeCoB/CoTaZr nano-bilayers in each of the four magnetic sublayers (see Table III). A similar trend was also revealed by the AFM measurements, as shown in Table III.

Since reflectivity on such thick, layered systems is practically insensitive to the deep buried substrate, in the fitting models above the roughness of the SiO₂ substrate was kept constant at 0.20 nm, as measured by AFM. Finally, the fitting analysis of the specular XRR scans revealed that the electron densities, which influence the refractive indices for the X-ray scattering, were ∼12 % higher for the FeCoB layers, and ∼5 % lower for the CoZrTa and Al₂O₃ layers with respect to the corresponding values of bulk FeCoB, CoZrTa and Al₂O₃ materials, respectively.

Figure 6 compares the in-plane hysteresis loops (along the EA and HA directions) of the soft magnetic multilayers I-IV, measured by both MOKE in the center of the processed 8” Si/SiO₂ wafers (see Fig. 6(a,b)) and a B-H looper that averages over the entire wafer area (see Fig. 6(c,d)). The shape of the hysteresis loops clearly reveals that all multilayers exhibit well-defined magnetic anisotropies. The corresponding values of coercivity along EA (H_c for MOKE, H_c,e for B-H looper) and HA (H_c,h for B-H looper) and anisotropy field (H_k) are presented in Table IV as function of the Fe content in the multilayers. The Fe content in the multilayers III and IV was estimated based on the ratio between the thickness of FeCoB nanolayers and that of the FeCoB/CoTaZr nano-bilayer, as determined from XRR. On one hand, we observe that H_c, H_c,e and H_c,h scales up when the Fe content increase. On the other hand, H_k gradually increases from 18.30 Oe to 34.71 Oe up to an Fe content of 25.8 %, and then decreases to 25.77 Oe up to the Fe content of 52 % corresponding to the FeCoB/Al₂O₃ multilayers. The difference between the coercivity values obtained with the MOKE and B-H looper could be due to the limited area measured by the former (85 points, 30 mm edge exclusion), whereas the latter averages over the entire wafer area. Moreover, MOKE is only sensitive to the top most magnetic layer, whereas the B-H looper is sensitive to the whole layer stack. The soft magnetic properties (H_c, H_k) together with the magnetic dispersion, i.e. local magnetic anisotropy skew angle (not shown here), were also mapped over the wafers (edge exclusion 30 mm) by MOKE as presented in Fig. 7. Thus, for all multilayers the magnetic dispersion was better than ± 2°, coercivity was very low over the whole wafer areas, and the uniformity of anisotropy field was better than 2 %.

To shed light on the dependence of multilayer coercivities with the Fe content, we performed GAXRD measurements with an incidence angle ω = 6° as depicted in Fig. 4(b). The GAXRD spectra for the as-deposited multilayers I-IV are shown in Fig. 8(a). Besides some weak diffraction peaks, which could be attributed to the Si/SiO₂ substrate, all multilayers exhibit a broad (110) diffraction peak at ∼ 44.5° that is due to the body-centered cubic (BCC) α-FeCo phase. The exact angular position (2θ₁₁₀) and full-width at half maximum (FWHM, w₁₁₀) of the (110) diffraction peaks together with the corresponding lattice parameter (a) and average grain size (D) using the Scherrer equation¹⁴ were determined from the GAXRD spectra by fitting the curves with multiple Gaussians (see Table V). Thus, while the lattice parameter of the nanocrystalline grains in the magnetic layers is between 2.868 Å and 2.889 Å, and it does not seem to exhibit a dependence with the Fe content, D increases from 1.83 nm for 0 % Fe (CoTaZr) to 2.81 nm for 52 % Fe (FeCoB) (see also Fig. 8(b)). Interestingly, the electrical resistivity, obtained from sheet resistance measurements performed for 80 nm thick magnetic films with the same composition as the magnetic sublayers in the multilayers I-IV grown on additional 8” Si/SiO₂ wafers, also increases with the Fe content as shown in Fig. 8(b).

Having determined the average grain size in the nanocrystalline magnetic sublayers of the multilayers I-IV, we can now represent H_c vs. D as shown in Fig. 8(c). Consequently, H_c seems to exhibit an increase with the grain size, and this is more evident for the B-H looper measurements. By representing the H_c vs. D dependence in a log-log plot as shown in Fig. 8(d), we see that even though our magnetic layers lie more in the amorphous alloys territory (i.e. D ∼ 2 nm), the increase of coercivity with the grain size could still be described by the random anisotropy model for nanocrystalline ferromagnets, according to which H_c ∼ D⁶.^20–22 However, we should point out that, since the grain size in our samples extends over a very small range (e.g. 1.8 – 2.8 nm), this behavior cannot be confirmed with complete certainty. Moreover, since in uniaxial magnetic films coercivity scales with the effective magnetic anisotropy (K_e) like H_c ∼ K_e/M_s, where M_s is the saturation magnetization, the increase of H_c with the Fe content could also caused by the higher magnetostriction state in the Fe enriched layers that enhances the magneto-elastic anisotropy contribution to K_e.²³ 

In order to check this hypothesis, both the (BH) loops of the as-deposited multilayered samples and those when the layers were stressed by bending the wafers were measured as described in the experimental section (see Fig. 9(a)). Figure 9(b), which for clarity presents just the first quadrant (i.e. H, B ≥ 0) of the (BH) loops in Fig. 9(a), reveals that some of the samples exhibit significant magnetostriction (e.g. I, FeCoB/Al₂O₃), whereas others have virtually none (e.g. II, CoTaZr). In fact, by measuring the change of the anisotropy field due to the applied external stress for every sample, and by representing it versus the corresponding Fe content, as shown in Fig. 9(c), we can see that ΔH_k (and thus the magnetostriction) increases when the Fe content in the sample increases. This suggests that the increase of coercivity with the Fe content in Fig. 8(b) might be due to an enhancement of magnetostriction. Moreover, this shows that our method of producing magnetic thin films with nano-layered structure allows us to tailor the magnetostriction by simply selecting the thickness of the individual nano-layers.

While the gradual increase of coercivity with Fe content (i.e. with the fraction of FeCoB layers in the FeCoB/CoTaZr nano-bilayers) could be explained through influences of grain size and magnetostriction, the reasons for the behaviour of the anisotropy field in Fig 6(f) remain unresolved. One possible cause for the anisotropy field enhancement in the nanostructured multilayers (e.g. III and IV) could be the higher interface roughness in these samples as was revealed by XRR measurements in the previous section.^24,25

However, a more plausible cause for this enhancement is the modification of magnetocrystalline anisotropy due to the many interfaces present in the nanostructured multilayers that introduce symmetry breaking, since it is known that the interface contribution to magnetic anisotropy becomes important, especially for ultrathin magnetic films. Thus, a spin reorientation transition from perpendicular to in-plane magnetization orientation could occur when the thickness of magnetic nanolayers is reduced.²⁶ In order to test this last assumption, we prepared additional X/Al₂O₃ multilayers with four periods and 4-nm-thick Al₂O₃ interlayers, where the 80-nm-thick ferromagnetic layers X consist of FeCoB/CoTaZr multilayers with FeCoB and CoTaZr nanolayers having the same thickness (i.e. average Fe content of 26%) and increasing FeCoB/CoTaZr nano-bilayer thickness (from 0.4 nm to 4 nm). For that, we kept the sputter powers for FeCoB and CoTaZr targets constant and we increased the cage rotation speed. Consequently, for nano-bilayer thicknesses of 0.4 and 4 nm, the magnetic layers X consist of 400 and 40 FeCoB and CoTaZr nanolayers, respectively. The corresponding H_k values of these multilayers measured with the Kerr magnetometer are shown in Fig. 9(d). We observe a gradual increase of H_k when the FeCoB/CoTaZr nano-bilayer thickness is reduced, e.g. from 26 Oe for 4 nm to 48 Oe for 0.4 nm, thus almost a factor 2, which demonstrates that the interface contribution to magnetic anisotropy is indeed the cause for the enhancement of in-plane anisotropy. While this enhancement might appear unexceptional, one should not forget that we achieved this by keeping the coercivity extremely low (<0.1-0.2 Oe), which is mandatory for soft magnetic multilayers required by ultra-low loss RF passive devices. Previously, we obtained H_k values much higher than 48 Oe in FeCoB-based soft magnetic multilayers sputtered by using horizontal collimators (135 Oe) or higher pressures (82 Oe), but the corresponding coercivities were much higher (∼1 Oe).^11,12 Moreover, we found out that the magnetic properties of nanostructured multilayers are very stable upon successive post-annealing treatments up to 300°C (not shown here), whereas those of the multilayers sputtered with horizontal collimators are not (depending on the annealing temperature H_k can decrease by at least 30%), which is a big advantage for device integration.

The broadband (100 MHz - 10 GHz) RF spectra of real (μ’) and imaginary (μ”) parts of magnetic permeability (μ) were measured to reveal the high frequency performance of the soft magnetic multilayers I-IV. Examples of such RF spectra for the multilayer II (e.g. CoTaZr/Al₂O₃) are shown in Fig. 10(a). The experimental spectra for all samples could be described by the classical Landau-Lifschitz-Gilbert (LLG) equations in the macro-spin approximation (i.e. assuming an anisotropic film with uniform magnetization) with a damping factor (α).^27,28 Hence, all soft magnetic multilayers should exhibit a uniform magnetization throughout their structure, and this was already evident from the hysteresis (BH) loops in Fig. 6d. The corresponding values of the FMR frequency (f₀), line width (⁠$Δf$⁠) and α are presented in Fig. 10(b) and Table VI as a function of the Fe content in the multilayers. Thus, the evolution of FMR frequency with the Fe content (see Fig. 10(b)) is analogous to the dependency of anisotropy field versus the Fe content (see Fig. 6(b)), in conformity with Kittel’s ferromagnetic equation $f 0 ∼ γ 0 2 π 4 π M s . H k$⁠.²⁹ In the similar manner, the dependency of the permeability with the Fe content in Fig. 10(c) can be explained, since μ ∼ 4πM_s/H_k.

In summary, we presented an innovative, economic method for manufacturing soft magnetic materials with improved properties (e.g. in-plane anisotropy, FMR frequency) in an industrial, high-throughput Evatec LLS EVO II magnetron sputtering system. The method consist of sputtering multilayered thin films on 8” bare Si and Si/200nm-thermal-SiO2 wafers by simultaenously using two or more cathodes. Due to the continuous rotation of the substrate cage, such that the substrates face different targets alternatively, the thin film materials deposited in this manner exhibited a nanolayered structure with very sharp interfaces as was demonstrated by XRR measurements for the case of FeCoB and CoTaZr materials. The thickness of the individual nanolayers was varied between 0.5 and 1.5 nm by changing the cage rotation speed and the power of each gun, providing us with excellent control to tailor new, composite ferromagnetic materials. Kerr magnetometry and B-H looper measurements revealed that the coercivity of these thin films is very low, 0.2-0.3 Oe, and gradually scales up with the thickness of FeCoB nanolayers. On the other hand, the in-plane anisotropy field of the nanolayered films exhibited a noticeable enhancement that was explained by the interface contributions to magnetic anisotropy in the nanostructured multilayers. These novel soft magnetic multilayers, with enhanced in-plane anisotropy that was stable upon successive post-annealing treatments up to 300°C, allowed operation at higher frequencies, as revealed by RF measurements. This deposition method can be applied to more than two materials at the same time, and other material combinations, including magnetic/nonmagnetic ones, potentially paving the way for integrated thin film magnetic cores with dramatically improved properties.

We are grateful to Marco Padrun, Maurus Tschirky, Silvan Wüthrich and Thomas Nadig for continued support, Stephan Voser and Walter Fasnacht for generating the schematics of the 3D MEMS micro-transformer in Fig. 1, Alfred Badertscher, Daniel Schneider, and Heinz Felzer for technical assistance, Hartmut Rohrmann for scientific discussions, and Allan Jaunzens for proofreading the manuscript. Mojmír Meduňa acknowledges the MEYS of the Czech Republic under the project CEITEC 2020 (Project No. LQ1601).