Emergent properties from CuPd alloy films under near-infrared excitation

The photoexcitation of metals offers a route to the ultrafast generation of charge carriers (e.g., hot electrons and holes) with technological applications in ultrafast photodetection¹ and energy conversion.² Over the past several decades, the hot-carrier dynamics of Au- and Pt-group single-element metals have been widely studied.^2,3 Au group metals have deep lying d-band states (e.g., >2 eV below the Fermi level) and thus require visible light to excite interband (e.g., d → sp) driven hot carriers. More recently, Landau damping from plasmonic excitations on Au group metals has allowed hot-carrier generation below the interband energy threshold (IET) using near-infrared (NIR) light.¹ While nanostructured plasmonic materials allow for tunable hot-carrier energy distribution, the resulting devices still suffer from very low efficiency.⁴ Recent work has demonstrated that the excitation of abundant d-band states can aid in improving the efficiency of hot carrier devices.⁵ Thus, materials that offer low-lying d-band states would allow these improvements to transfer to NIR hot carrier applications (e.g., ultrafast NIR photodetection). Traditional Pt-group catalytic materials satisfy this requirement but have extremely short hot carrier lifetimes due to extensive e–e scattering from their large electronic density of states (EDOS) near the Fermi level.⁶ Short carrier lifetimes are one of the major roadblocks to efficient charge extraction from hot carrier devices. Metal alloys with NIR-accessible d-band states provide a route to overcoming this by allowing facile EDOS tuning through band hybridization effects that yield emergent optical and electronic properties.⁶ 

Our group recently reported that Au₅₀Pd₅₀ noble-transition metal alloy films excited at 1550 nm provided 20-fold more hot holes than Au films and three-times longer lifetimes than pure Pd.⁶ Hybridization between the Au 5d and the Pd 4d states results in a new band structure responsible for these emergent electronic properties.⁶ The 5d states in pure Au are split into 3/2 and 5/2 bands due to spin–orbit coupling effects. When alloyed with Pd, only the Au 5d^5/2 band hybridizes with the Pd 4d states.^6,7 Furthermore, under dilute Pd conditions, deleterious thermalization effects are minimized, and Au–Pd hybridization effects are weak.⁸ Thus, other noble-Pd metal alloy systems, which exhibit stronger band hybridization under dilute Pd conditions, are expected to lead to improved hot-carrier performance.

In the current study, we deposit disordered Cu_xPd_1−x alloy thin films by magnetron co-sputtering and investigate their stoichiometry-dependent optical, structural, electronic, and transient near-infrared absorption properties. We show experimentally, through ultraviolet photoelectron spectroscopy (UPS), and computationally, through the projected density of states calculations, that strong hybridization between the Cu 3d and Pd 4d bands exists in the alloy films. Unlike the Au 5d band, the Cu 3d band does not exhibit strong spin–orbit coupling.⁹ Therefore, as previously reported, we observe the formation of a common band between the full Cu 3d and Pd 4d bands.⁸ This strong d-band hybridization in the alloy films drives the emergent behavior we observe across our various measurement techniques and makes Cu_xPd_1−x alloys a promising material for NIR hot carrier optoelectronic devices.

All films were deposited by magnetron sputtering in a Lesker CMS-18 dual-chamber system. The alloys were co-sputtered with the guns in a confocal geometry. Before sputtering, the chamber was evacuated to a base pressure of 2 × 10⁻⁷ Torr. The working pressure for sputtering was 1–3 mTorr and was maintained by downstream pressure control with a constant Ar flow of 20 SCCM. Cu and Pd targets (3 in. diameter; 99.99 wt. %) were used with DC and RF power, respectively. The source substrate distance was ∼100 mm. We varied the power from 0 to 200 W on the Cu and Pd targets to achieve the desired alloy stoichiometry while maintaining a nearly constant film deposition rate of 30 nm/min. Samples on Si(100) substrates with a metal film thickness of 50 nm were used for UPS, x-ray photoelectron spectroscopy (XPS), ellipsometry, and x-ray diffraction measurements. Double-side polished Z-cut quartz substrates with metal film thicknesses of 5 nm were used for time-resolved terahertz spectroscopy (TRTS).

All photoelectron measurements were carried out at the 5 m toroidal grating monochromator (5 m-TGM) beamline at the Center for Advanced Microstructures and Devices (CAMD) at Louisiana State University. The synchrotron beamline and endstation are described in detail elsewhere.¹⁰ The beamline was equipped with a photoemission endstation utilizing an Omicron EA125 hemispherical electron energy analyzer with a five-channel detector and dual Mg/Al x-ray source. UPS spectra were collected with a constant pass energy of 10 eV at a photon energy of 85 eV in an ultra-high vacuum with a pressure of 10⁻¹⁰ Torr. UPS spectra were collected in normal emission geometry with a 45° incident angle to the surface normal. All UPS spectra were normalized to a point on the background whose intensity solely emerges from inelastic secondary electrons. The binding energies were referenced with respect to the Fermi level of the Cu sample. The surface of the samples was cleaned with ion sputtering (1 keV, Ne⁺) for 10 min before photoemission measurements to remove any near-surface contaminants. XPS was conducted on the same sample in the same chamber following the UPS measurement. The XPS spectra were collected in constant pass energy mode with a pass energy of 30 eV. The near-surface composition of the 50 nm alloy films was determined by XPS using CASA software to model the data qualitatively.

A Horiba SmartSPM atomic force microscope in the tapping mode was used to measure the topology of the 5 nm-thick alloy films on fused quartz substrates. The tip used for the scans was a HQ:NSC15/Al BS from MikroMasch. The scan area size was 2.5 × 2.5 μm².

A Woollam RC2 variable-angle spectroscopic ellipsometer was used to measure the Ψ and Δ values for the optically thick 50 nm pure Cu, Pd, and Cu_xPd_1−x alloy films. Ψ and Δ values were collected over a 210–2500 nm wavelength range with 60°, 65°, and 70° incidence angles. The complex permittivity was determined for Cu, Pd, and the alloys by fitting the data to a two-layer (air-metal) Kramers–Kronig valid B-splines¹¹ model using the CompleteEASE software package.

A PANalytical EMPYREAN x-ray diffraction (XRD) with a Chi Phi XYZ 5 axis cradle stage equipped with Cu Kα radiation was used to evaluate the phase composition of the films. The samples were pre-aligned at maximum Si(100) intensity, and the scans were performed at a 0.5° incidence angle omega.

The change in conductivity was investigated in 5-nm thick Cu_xPd_1−x alloys upon 1550 nm excitation using time-resolved terahertz (THz) spectroscopy (TRTS) in transmission geometry. All films were deposited on Z-cut quartz to avoid THz absorption by the substrate. TRTS was performed with a regeneratively amplified Ti:Sapphire laser (Coherent Libra, 800 nm, 1 kHz, 50 fs pulse duration) coupled to a home-built spectrometer. The sample was excited with a 1550 nm pump using an optical parametric amplifier (Coherent TOPAS). Broadband terahertz radiation (0.2–2 THz) was generated and detected using ZnTe nonlinear crystals. The change in transmission upon chopped photoexcitation was normalized by the non-photoexcited signal to determine ΔT/T. More details on the THz setup can be found elsewhere.^12–14 The 1550 nm pump and THz probe spot size were measured using a translating razor blade by recording the transmitted power as a function of translated distance. An error function was used to extract the pump and probe spot size at the sample position. The pump and probe spot sizes were 3.2 and 2.3 mm in diameter, respectively. For all the THz measurements, the fluence was kept at 90 μJ cm⁻².

The simulation results were based on the electronic-structure method of Korringa, Kohn, and Rostoker (KKR)^15–19 in combination with the Coherent-Potential Approximation (CPA)^17–19 to handle the chemical and magnetic disorder effects on an equal footing with the electronic structure. The multi-atom KKR-CPA software was used, which was originally developed at Oak Ridge National Laboratory/University of Cincinnati,²⁰ to determine the electronic structure of complex alloys. The code is available by request from the corresponding author. For these calculations, the potentials were modeled within the Atomic Sphere Approximation (ASA), including a maximum angular momentum of L− = 6 for the spherical basis,²¹ and the Perdew, Burke, and Ernzerhof for solids generalized gradient functional²² was employed for the density functional theory (DFT) functional. For the k-space integration, the special k-point integration technique^23,24 was used on a 12 × 12 × 12 and a 18 × 18 × 18 Monkhorst–Pack mesh.²⁵ Because Green's function for the system was determined by a semi-circular energy contour, two K-meshes were used. The semicircular energy contour contained 20 energy points. The first starting point was in a region with no electronic states (an energy gap exists between core states and valence) and used the 12 × 12 × 12 mesh. As the energy grid was traversed, the energy points were within 0.25 Ry. From the top of the contour, a denser 18 × 18 × 18 k-point mesh was used since there were states near the top of the contour (Fermi level), increasing the structure in the Green’s function, where it was becoming more singular. The ground-state volumes were determined by calculating the total energy at nine different lattice constants and fitting them to the Birch–Murnaghan equation of state to obtain the ground-state volume.^26,27 Once the ground-state volume was obtained, the EDOS was determined by calculating the Green’s function on an energy grid parallel to the real energy axis but with a 0.003 Ry in the complex energy plane. The joint electronic density of states (JDOS) can be directly obtained from the EDOS. Equation (1) was applied for the hot carrier distribution, replacing the product of the densities of states at the initial and final energies by the JDOS and coupling with their respective distribution functions.

We report the real [Fig. 1(a)] and imaginary [Fig. 1(b)] permittivity values for sputtered 50 nm Cu_xPd_1−x alloy films and their constituent pure metals on silicon chips using spectroscopic ellipsometry spanning the ultraviolet to NIR wavelengths. Interband and intraband transitions in metal films strongly dictate the imaginary permittivity, i.e., dielectric losses. As shown, Cu exhibits the lowest losses of all the films in the NIR because the excitation energy is below the interband energy threshold (IET) for Cu, which is ∼2.1 eV (∼590 nm).²⁸ Pd, on the other hand, has interband transitions close to the Fermi level and thus exhibits significantly higher losses since both interband and intraband transitions are accessible in the NIR. The inset in Fig. 2(b) shows the trend in the real (ε₁) and imaginary (ε₂) permittivity at 1550 nm (dashed line) as an increasing fraction of Pd is added to the alloy films. As Pd is added to the alloys, the ε₁ and ε₂ values increase, with most of the change in the permittivity values occurring before 50 at. % Pd fraction is reached. Beyond 50 at. % Pd, the real and imaginary permittivity trend flips, with both values decreasing slightly as pure Pd approaches. These results clearly show that the optical properties of the Cu_xPd_1−x alloys are not simple ratios of the pure films. Our previous study of Au_xPd_1−x alloy films also observed non-simple mixing of the pure optical properties, but Cu_xPd_1−x films show even more non-linear behavior indicative of stronger hybridization effects, as we initially predicted.

Unlike the Au–Pd system, which is miscible for all compositions in the solid phase, the Cu–Pd thermodynamic phase diagram exhibits several regions of immiscibility at room temperature.²⁹ To avoid phase segregation in our Cu_xPd_1−x films, we sputtered them at high rates to lock in a single-phase structure. To verify the phase behavior of the Cu_xPd_1−x films, we used grazing incidence x-ray diffraction (GIXRD). The GIXRD patterns in Fig. 1(c) confirm that our thin-film Cu_xPd_1−x alloys are single-phase alloys. Figure 1(c) also shows that all samples are polycrystalline FCC, and from top to bottom, we see how the Bragg peaks shift to larger spacings as we add the larger Pd atom into the Cu lattice. As predicted by Vegard’s law, the lattice constants of the alloys falls between the lattice constant of the pure metals with some deviation from linearity, as shown in Fig. S1.

We acquired synchrotron-based UPS data from the thin films to better understand the electronic structure changes upon alloying pure Cu and Pd metals. In addition to examining the valence states of the alloys, we used XPS to investigate subsequent changes in the core levels of each of these metals to probe their chemical states. Because these two surface sensitive techniques were employed in our earlier work on Au_xPd_1−x alloy films,⁶ electronic/chemical structural trends revealed in that work can be directly compared and contrasted with those in the current study of Cu_xPd_1−x alloy films.

Figure 1(d) shows the valence EDOS of pure Cu and Pd metals and alloy films of various compositions. Only the EDOS down to 3 eV binding energy (BE) is shown. While the pure Cu film is characterized by filled 3d band states stretching from 2 to 5 eV BE and, smaller unfilled density of s–p states up to E_F, the Pd film is characterized by a strong, partially filled 4d spectral weight, which increases from 5 eV BE up to E_F. Although both comprise different band structures and spectral weights for their densities of states, it is apparent that the Pd valence bands directly overlap in energy with the Cu d-band below 2 eV, akin to the Au_xPd_1−x alloy film system.⁶ Upon alloying these metals, notable changes in the d-band structure of both Cu and Pd are revealed with UPS. As the Pd content of the alloy increases, the centroid of the d-band of Cu moves toward E_F, while the Pd 3d band center shifts away from E_F (toward high BE), see Fig. 1(d). Furthermore, accompanying this shift in the Pd d-bands with increasing Cu concentration, there is a concomitant attenuation of the Pd density of states at or near E_F (within 0.9 eV). For the 50–50 alloy, there remain appreciable Pd states near E_F, but they rapidly decrease at higher Cu concentrations.

With the knowledge that d-states close to the Fermi level are a requisite for NIR applications, the UPS data allow us to directly compare the EDOS as a function of alloy composition. As shown in the shaded region in Fig. 1(d), upon alloying Cu with only 4% Pd, the measured EDOS at 0.8 eV (1550 nm) already appreciably increases compared to Cu. Further addition of Pd into the alloy, for example, at 32%, increases the measured photoemission feature by ∼eightfold, which is 62% of the value of pure Pd at the same energy.

In general, s₋p₋d hybridization is proposed for the alloys within the framework of electroneutrality and electronegativity considerations.^30,31 Moreover, as revealed in previous alloy studies, and also this study, the alloying introduces a strong hybridization of Cu 3d states with surrounding Pd 4d states and causes deformation and spectral shape changes on the valence band structure of the pure metals.^31–33 These studies observe an overall smearing of the valence band features upon alloying. Similarly, our recent work on the Au–Pd alloy system also exhibited this type of hybridization between Au and Pd and consequent modification in the d-band structure, primarily in the region where there is strong d-band overlap.⁶ Owing to the complete overlap of the Pd and Cu d-bands in this study, a stronger hybridization is also expected and confirmed with UPS. Moreover, the binding energy shifts in UPS and XPS spectra cannot be explained by simple charge transfer between the constituents of the alloys based on the electronegativity arguments described above.^30,31 Our XPS measurements reveal (Fig. S2) core level shifts consistent with the changes observed in the valence band of the Cu–Pd alloys, wherein Cu shifts upward and Pd shifts downward upon dilution. We attribute this to the strong s₋p₋d hybridization upon alloying and the concomitant changes in the atomic screening compared to the pure metals. In addition, our XPS results also confirm that the metallic binding energy of Cu 2p (∼932.8 eV) and Pd 3d (∼355.6 eV) indicates no oxide formation.

To further investigate the Cu and Pd band hybridization upon alloying, we calculated the partial EDOS using the KKR-CPA code. Figures 2(a) and 2(b) show the calculated partial EDOS for the pure metals and four alloy compositions, e.g., 17%, 32%, 50%, and 75% Pd. In pure Pd, the 4d band (grey-solid line) is the predominant source of electronic states [Fig. 2(a)] and intersects E_F due to the 4d–sp hybridization.³⁴ Cu exhibits strong 3d–4sp hybridization about 1–2 eV below E_F. Alloying Cu (orange solid-line) with Pd (solid-grey line) Fig. 2(a), one can identify apparent common band behavior, in particular with Cu d-states completely under the Pd states. Although Cu, Ag, and Au are all in the same column of the periodic table, both AuPd and AgPd alloys exhibit split behavior^6,8 where, for example, the center of the d-bands of Cu and Pd have shifted away from one another, producing a much smaller overlap of their respective EDOS. Thus, Cu forms a more complete hybridization with Pd than Au and Ag. This leads to CuPd alloys having very different excitation properties than Pd alloys of Ag and Au. Furthermore, as the composition of Pd increases, the Cu d-state band center shifts, and the d-band broadens due to disorder, pushing more Cu d-states closer to E_F. This leads to more d-states available for excitation and a corresponding increase in the joint density of states.

The pure Cu 3d band (orange-solid line) shown in Fig. 2(a) provides a negligible amount of d-states for interband transitions near E_F. However, as the Pd composition increases, there is a corresponding increase in Pd d-states near E_F and a broadening of Cu d-states toward E_F. Due to the decrease in Cu composition, the available Cu d-states for possible band transitions attenuates. Figure 2(b) shows the s and p states of the alloys, which, as expected, are very broad with little weight in the EDOS. The shifts of the Cu d-band center and the smearing of its states are supported by XPS data (see Fig. S3), where the features in the valence bands are exclusively due to Cu and Pd metallic contributions in the XPS spectra. The binding energy shifts in UPS and XPS spectra cannot be explained by a simple charge transfer between the constituents of the alloys. This is consistent with the broadening of the states due to disorder and the complete overlap of Cu with Pd d-states that suggests increased hopping between sites.

The increase in the EDOS near E_F for Pd rich Cu_xPd_1−x alloys, compared to pure Cu, translates into more hot carriers under 1550 nm excitation, as shown in Fig. 2(c). Equation (1) describes when a hot carrier is photon excited with energy E_ph, is raised from an initial state at E − E_ph to a higher energy state E. The transition probability for this event is proportional to the joint density of states (JDOS), which is the product of two electronic densities of states at the initial and final energies multiplied by their distribution functions³⁵ 

where $fE$ is the Fermi–Dirac distribution function. Thus, the hot-carrier generation probability is proportional to the product of the EDOS at the initial and final energies and their respective Fermi–Dirac distribution functions. We calculated the hot carrier distribution using Eq. (1) with the Fermi distribution function (f) at 300 K and photon energy (E_ph). We determined the joint density of states from the calculated EDOS.

The excitation generates a basically flat, near-zero distribution of both hot electrons and holes for pure Cu. This nearly zero response for pure Cu is because the d-states for Cu are shifted 1–2 eV below E_F. As mentioned above, as the Pd composition is increased, the Cu d-states broaden toward E_F with a corresponding increase in Pd d-states near E_F, which leads to the gradual but uniform increase in hot carriers. For a Pd composition greater than 50%, the carrier distribution starts to become more appreciable due to the combination of having some Cu and Pd d-states near E_F, which favors hotter holes than electrons. Hotter holes are beneficial for hot-carrier devices with metal/p-type semiconductor Schottky junctions.^36,37 We calculate an ∼14-fold increase in the number of hot-holes 0.78 eV below E_F and an ∼threefold increase in the number of hot-electrons 0.78 eV above E_F, for an Cu₅₀Pd₅₀ alloy, as compared to Cu. This increase is similar to Au₅₀Pd₅₀ alloy system⁶ where a 20-fold increase was found for hot-holes and a sevenfold increase for hot-electrons, demonstrating analogous processes leading to the physical properties between these alloy systems. We attribute the larger number of 0.78 eV hot-electron in our 1550 nm excited Cu₅₀Pd₅₀ alloy films to Cu–Pd strong d-band hybridization and resulting interband transitions near E_F where both Cu and Pd have d-states. Hybridization of the Cu 3d and Pd 4d bands shifts the d-band center closer to the Fermi level, allowing direct interband transitions.

Studying the electron dynamics in the Cu_xPd_1−x films upon NIR excitation is critical to developing a more complete understanding of the electronic structure of the alloys. We measured the relaxation dynamics of the hot carriers as a function of alloy composition using TRTS in a transmission configuration. TRTS is an established technique for probing the different pathways that contribute to the energy relaxation of photoexcited carriers.³⁸ The 50-nm-thick films studied in Fig. 1 were too thick for transmission work; therefore, co-sputtered 5-nm-thick films on Z-cut quartz substrates were used to allow transmission measurements. The 5 nm-thick films are below the skin depth for Cu and Pd and therefore allow us to neglect the ballistic diffusion of hot electrons.³⁹ 

We studied the carrier dynamics of the films after a 1550 nm pump by probing the temporal change in broadband terahertz (THz) transmission as shown in Fig. 3(a). ΔT/T is the differential peak electric field of the terahertz pulse transmitted through the sample upon photoexcitation, normalized by the peak electric field without photoexcitation. In general, an increase in terahertz transmission corresponds to a decrease in conductivity. Figure 3(b) shows a barely measurable change in ΔT/T upon photoexcitation of pure Cu but a sharp negative peak in ΔT/T at early times for Pd fractions up to ∼50 at. %. Alloys beyond 50 at. % Pd display positive contributions to ΔT/T persisting for >5 ps. These positive features are due to thermalization of hot carriers into the lattice. The negligible change in ΔT/T for the pure Cu films is a result of the 1550 nm (∼0.8 eV) excitation being below the IET of Cu (∼2.1 eV).²⁸ Interestingly, the dilute Pd alloys show a sharp decrease in ΔT/T at early times. We attribute this to an increase in photoconductivity due to the generation of non-thermalized hot carriers. As the Pd fraction in the alloy increases, up to ∼50 at. %, the amplitude of the negative dip in ΔT/T at early times also increases. This clearly shows that increasing the Pd fraction in the alloy provides more accessible d-band states for NIR (i.e., 1550 nm) excited interband transitions and is consistent with the composition-dependent trends in the measured EDOS in Fig. 1(d) and the calculated alloy hot-carrier distributions in Fig. 2(c), above. Notably, the diluted Pd alloys do not display a long-lived rise in ΔT/T due to thermalization of hot carriers. Electron emission is one possible explanation for the lack of thermalization in the dilute Pd case, i.e., carriers emitted into the vacuum do not return to heat the lattice. Two possible processes leading to electron emission in this case are multiphoton absorption and field emission.⁴⁰ In multiphoton absorption, several photons are simultaneously absorbed, raising the electron above the work function barrier of the metal. Higher-energy photons would be more efficient at this process since fewer of them have to be absorbed to lift the electron over the barrier. For field-driven electron emission, surface roughness is required to enhance the electric field strength. Atomic force microscopy topology scans of the 5 nm-thick films on quartz show that the Cu₄₈Pd₅₂ films are rough with 2 nm rms, while other alloy compositions display lower roughness values (see Fig. S3). We calculate an electric field of ∼600 kV/cm for the 4 μJ NIR pump laser with ∼100 fs pulse width. This high field strength coupled with the roughness of the films could possibly enable field-driven electron emission, but more studies are necessary to confirm the precise mechanism for the lack of thermalization in the dilute Pd alloy films.

Pd-rich alloys, on the other hand, show positive contributions to ΔT/T at early times with additional long-lived positive features. Heating of the lattice decreases the conductivity of the film, which is observed as an increase in the overall terahertz transmission. Alberding et al. reported a similar TRTS response from thin Ti and Au films excited at 800 nm.⁴¹ As more Pd is added to the alloy the EDOS near the Fermi level increases thus increasing e–e, and e–ph scattering, reducing the lifetime of the hot-carriers and resulting in rapid thermalization.⁶ We observed similar behavior in 5 nm-thick Au_xPd_1−x alloy films, where compositions near 50–50 provided optimal EDOS and carrier lifetimes but Pd-rich films resulted in rapid thermalization.⁶ This clearly demonstrates the power of band hybridization effects in noble-transition metal alloys for developing tunable hot carrier materials.

In summary, we have shown that Cu_xPd_1−x alloy films exhibit emergent electronic and optical properties. Both photoelectron spectroscopy and density functional theory calculations attribute this emergent behavior to strong d-band hybridization between the Cu 3d and Pd 4d bands. Furthermore, ultrafast spectroscopy reveals composition-tunable electron dynamics upon NIR pump, with dilute Pd alloys exhibiting non-thermalized hot carriers at early times. NIR interband transitions in the dilute Pd alloys provide a route to non-thermalized hot holes that could have important applications in optoelectronic devices.

See the supplementary material for calculated lattice constants, XPS spectra, AFM images, and additional TRTS spectra for the CuPd alloy films.

The authors acknowledge funding from the National Science Foundation under collaborative research Award Nos. DMR-2114304 (LSU) and DMR-2114312 (Drexel). K.M.M. acknowledges funding from the Army Research Office through Award No. W911NF2110129. The authors thank the LSU Center for Advanced Microstructures and Devices and the LSU Nanofabrication Facility for assistance in characterizing and fabricating the alloy films. The authors thank B. T. Diroll for helpful discussions on field-driven electron emission and Jens Neu for helpful discussions on terahertz spectroscopy.

The authors have no conflicts to disclose.

Gregory A. Manoukian, Orhan Kizilkaya, and Sergi Lendinez contributed equally to this work.

Gregory Manoukian: Formal analysis (supporting); Investigation (equal); Visualization (supporting). Orhan Kizilkaya: Conceptualization (lead); Funding acquisition (lead); Investigation (equal); Project administration (lead); Resources (equal); Supervision (equal); Writing – original draft (supporting); Writing – review & editing (lead). Sergi Lendinez: Formal analysis (equal); Investigation (equal); Writing – review & editing (equal). Luis D. B. Manuel: Formal analysis (supporting); Funding acquisition (equal); Investigation (supporting); Supervision (equal); Visualization (lead); Writing – review & editing (equal). Tiago R. Leite: Formal analysis (supporting); Investigation (supporting); Visualization (lead). Karunya S. Shirali: Investigation (supporting); Visualization (supporting). William A. Shelton: Formal analysis (supporting); Investigation (equal); Supervision (equal); Visualization (supporting); Writing – review & editing (supporting). Phillip T. Sprunger: Investigation (supporting); Writing – review & editing (equal). Jason B. Baxter: Funding acquisition (equal); Investigation (equal); Supervision (equal); Writing – original draft (supporting); Writing – review & editing (equal). Kevin M. McPeak: Conceptualization (lead); Funding acquisition (lead); Project administration (lead); Resources (equal); Supervision (equal); Visualization (supporting); Writing – original draft; (lead); Writing – review & editing (lead).

The data that support the findings of this study are available within the article and from the corresponding author upon reasonable request.