Introduction

Hierarchical porous materials are an emerging field of study for their utility in applications where the capacity to have a range of feature sizes within the same component is advantageous [1,2,3]. Hierarchical or gradient nanoporous materials, which allow for multiple pore sizes in the same network of material and possess one or more nanoscale features, are of high interest for their enhanced Raman scattering, as well as their applications as multifunctional sensors or catalysts due to their ability to combine ion transport via large pores with the increased surface area provided by smaller pores [4,5,6]. Within the realm of nanoporous (NP) materials, NP metal foams are of particular interest for properties such as low relative density and high surface area to volume ratio, making them good candidates for battery materials [7,8,9,10]. While much of the work on dealloyed nanoporous metal foams to date has focused on noble elements such as Au, Ag, or Pt, cheaper and more available alternatives are also being explored [11,12,13,14]. Dealloyed NP Cu is of particular interest for its applications in areas of green energy research such as CO2 conversion, where it increases both the reaction efficiency and selectivity for specific hydrocarbon reaction products [13].

While there are a variety of processing methods able to synthesize gradient NP structures including additive manufacturing, templating, and sintering, dealloying is the most direct method for developing nanoporous metallic materials [6]. Dealloying, the selective corrosion of one or more elements in an alloy to generate a continuous porous structure, is notable because it can be performed on a variety of precursor shapes (e.g., films, ribbons, rods) and can be used to produce pore sizes on both the micro- and nano-scales, enabling tunability in both connectivity and specific surface area [15]. Nevertheless, synthesizing graded NP materials via dealloying is challenging due to factors such as the alloy composition, the initial microstructure, and the changing reaction kinetics. For example, precursor alloys previously used for graded dealloyed samples contained phases such as intermetallics, which dealloyed in unpredictable ways [6, 16, 17]. Specifically, when a less noble phase preferentially corrodes and the atoms enter solution, the composition and electrochemical potential of the dealloying medium change, inducing changes in the reaction kinetics over time [17, 18]. The reaction kinetics, and the subsequent progression of the dealloying reaction front, can be further altered through changes to the reaction time or temperature [19, 20]. Additional compositional factors such as the parting limit of the alloy, or minimum quantity of the more noble component for which the reaction will occur, and the potential difference between all alloy constituents can influence both the evolution of the dealloying front and the resulting porous microstructure [21, 22]. Approaches such as combining dealloying with sintered materials and various multi-step dealloying processes have also been investigated in attempts to synthesize graded nanoporous materials from less complex precursors [5, 23, 24]. Preliminary synthesis of NP Cu foams with a pore size gradient through the thickness has been demonstrated for a bilayer film, where the layer order at the compositional interface influenced the resulting foam morphology [25]. However, the effects of the layer order and the type of interface on the dealloying reaction front have yet to be addressed. The fundamental mechanisms controlling the synthesis of graded NP materials are not understood, and further work is needed to better define the evolution of the dealloying front for precursor materials with multiple alloy compositions [13, 26].

In this study, we investigate the evolution of dealloying fronts in similar and dissimilar alloy compositional interfaces in gradient nanoporous copper foams with precursor materials of various Cu–Al and Cu–Zn compositions. Specifically, we examined two distinct cases: a change in relative density (similar interface) and a change in chemical potential (dissimilar interface). Four configurations of bilayer copper alloy films were synthesized via magnetron sputtering due to its flexibility in precursor materials and control over composition [27]. The as-sputtered films were imaged with scanning electron microscopy (SEM) before and after dealloying at a range of time intervals. Subsequently, SEM and energy-dispersive x-ray spectroscopy (EDS) were performed in order to analyze the evolution of the dealloying front and the foam morphology over time. Vickers indentation was performed to further analyze the effects of the layer order and type of interface on foam morphology. Overall, this work addresses critical dealloying parameters and the effect of interfaces for developing gradient nanoporous materials.

Results and discussion

As-sputtered films

For the dealloying reaction to occur, alloys must have a large electrochemical potential difference between constituent elements such that one component preferentially dissolves into solution [28, 29]. Additionally, there must be a high enough atomic fraction (the parting limit) of the less noble constituent to prevent the more noble atoms from forming a protective passive layer during the initial stages of corrosion [21, 22]. Thus, three binary copper alloy precursors were selected for this work: Cu20Al80, Cu35Al65, and Cu20Zn80. Cu, Al, and Zn have reversible reduction potentials of 0.340 V, − 1.66 V, and − 0.762 V, respectively, with potential differences between any two of the three elements above the threshold necessary for dealloying [30]. These individual compositions have previously been shown to form different ligament sizes with relative densities of 20%, 35%, and 20%, respectively, in the final dealloyed NP foam [31]. Therefore, the selected alloys were chosen to produce a gradient in relative density after dealloying or a gradient in chemical potential.

As mentioned in the experimental procedures, bilayer samples in four configurations were sputtered as solid films, which are shown in the cross-sectional SEM micrographs in Fig. 1. Two distinct columnar layers are visible in each micrograph, corresponding to the two layers in each as-sputtered film, where Figs. 1(a–b) are configurations 1 (Cu20Al80/Cu35Al65) and 2 (Cu35Al65/Cu20Al80) and the columnar grains in both layers are packed tightly together. In contrast, for Figs. 1(c–d), configurations 3 (Cu20Al80/Cu20Zn80) and 4 (Cu20Zn80/Cu20Al80), there are noticeable columnar voids between grains. Previous sputtering research on Cu–Zn alloys observed that alloys with a high fraction of Zn atoms had reduced adatom mobility during sputtering, inhibiting the growth of a more uniform film and producing a voided columnar structure [32,33,34,35].

Figure 1
figure 1

SEM micrographs of the fractured as-sputtered films of (a) configuration 1, (b) configuration 2, and cut as-sputtered films of (c) configuration 3 and (d) configuration 4. The orange labels in the top right corner indicate sample configuration. The black arrow on the right marks the film growth direction (Color figure online).

Full size image

The dealloying front

In the dealloying process, the dealloying front is the region where the reaction is actively occurring [25]. Along this boundary, the less noble component preferentially corrodes into solution while atoms of the more noble component reorganize into a bicontinuous-connected structure and expose the unreacted solid surface. For a reaction in an electrolytic solution, this begins at the sample/solution interface, and progresses into the sample as the planar interface between the solution and the solid and porous zones of the sample [11, 36]. In order to capture the progression of the dealloying front, both HCl (electrochemical dealloying) and H3PO4 (free corrosion) were tested for the film configurations in this study. Electrochemical dealloying in 5 wt.% HCl yielded a reaction time of less than five seconds, which made it unfeasible to retain analyzable samples for films of this thickness (2 µm) and precluded examination of the dealloying front. The accelerated reaction time can be attributed to the chloride ions, which have been shown to augment surface diffusivity in both Cu–Al and Cu–Zn alloys, increasing the speed of the reaction front [17, 37]. In contrast, the presence of phosphate ions in the dealloying solution for Cu alloys has been shown to inhibit surface diffusion of Cu, resulting in both slower progression of the dealloying front and smaller, smoother ligaments [37]. Consequently, free corrosion in concentrated H3PO4 was selected as the method of choice for this study as the reaction time was significantly slower, and samples could be obtained at intervals to examine the reaction progression.

The evolution of the dealloying front is shown in Figs. 2 and 3, which exhibit cross-sectional SEM micrographs and their corresponding EDS composition maps across three different time intervals: 30 min, 45 min, and “final step.” “Final step” was defined as the last time interval before a sample of a given configuration dissolved in the electrolyte. For all four configurations, the main planar dealloying front fully progressed through the films before either layer had fully reacted, yielding higher levels of either residual Al or Zn as shown in Table 1. The relative density \({p}_{relative}\) for each layer at the last time step was calculated with the following equation:

$${p}_{relative}=\frac{{n}_{Cu}*\left(1+R\right)}{100}$$

where nCu was the expected number of Cu atoms per 100 atoms of precursor alloy, and R was the residual fraction of Al or Zn for that layer. The red dashed lines on each SEM micrograph indicate the position of the dealloying front at that time point. On the EDS maps, Cu and Al are marked with yellow and red, respectively, and Zn is indicated with blue. In order to isolate the effects of porosity, composition, and chemical potential, the dealloying front is analyzed for two cases. Figure 2 presents configurations 1–2 (C1 and C2), bilayer films of different compositions but the same elements and an expected difference in relative density between layers, while Fig. 3 displays configurations 3–4 (C3 and C4), with an expected difference in potential and pore/ligament size between layers.

Figure 2
figure 2

SEM micrographs of Cu–Al bilayer films dealloyed in H3PO4 via free corrosion and corresponding EDS maps for configuration 1 (Cu20Al80/Cu35Al65) and configuration 2 (Cu35Al65/Cu20Al80) at various time intervals. The red dotted line marks the position of the dealloying front after 30 min (a–d), 45 min (e–h), and “final step” at 105 min (i and j), and 120 min (k and l). EDS map indicates both Al and Cu content (Color figure online).

Full size image
Figure 3
figure 3

SEM micrographs of Cu–Al/Cu–Zn bilayer films dealloyed in H3PO4 via free corrosion and corresponding EDS maps for configuration 3 (Cu20Al80/Cu20Zn80) and configuration 4 (Cu20Zn80/Cu20Al80) at various time intervals. The red dotted line marks the position of the dealloying front after 30 min (a–d), 45 min (e–h), and “final step” at 60 min (i and j). EDS map indicates Al, Cu, and Zn content (Color figure online).

Full size image
TABLE 1 Sample composition, ligament size, and relative density.
Full size table

Effect of relative density: Cu–Al/Cu–Al compositions

Porous materials (foams) can be categorized by both their pore/ligament size and their relative density (foam density divided by the bulk density of the material), which is inversely related to the porosity [38]. The expected relative densities based on compositions for this study, when fully dealloyed, are either 20% (80% porosity) or 35% (65% porosity), with ligament sizes after free corrosion dealloying in the range of 50–60 nm for a single-composition sample [31, 37]. The calculated relative densities for each layer after the final dealloying time step are in Table 1 and are greater than the expected relative densities as the objective of this study was to fully observe the reaction over time, not to have the reaction run to true completion. Thus, the “final step” dealloying time point was defined as the last time point at which there was a viable sample that could be retained for further testing, which does not necessarily coincide with when the dissolution reaction would have run to completion. Dealloying is a diffusion-based process, which is generally slow, and dealloying reactions may take multiple days to reach completion [39, 40]. However, due to the fast reaction times, there was not enough time for the residual Al or Zn atoms to exit the structures, and longer dealloying times did not produce usable samples that could be analyzed.

For simplicity of discussion, samples will be referred to as per the expected relative density. In Fig. 2, configuration 1 has an expected 80% porosity in the top layer and 65% porosity for the bottom layer, while configuration 2 has the inverse setup. This provides two distinct scenarios for understanding the reaction front as a function of the porosity, facilitating or restricting the flow of the electrolyte.

The dealloying reaction front between Cu20Al80 and Cu35Al65 proceeded fastest when the lower-Cu fraction was the top layer, defined here as the layer that was in contact with the solution first. This can be seen after 30 min where the dealloying front progressed ~ 440 nm for Figs. 2(a–b) (C1), compared to ~ 240 nm for Figs. 2(c–d) (C2). As the quantity of Cu atoms in the surface in contact with the electrolyte in Figs. 2(a–b) was closer to the parting limit, there was less reorganization of Cu atoms during dealloying, which increased the degree of etching along the columnar GBs [41]. After 45 min, the reaction front in Figs. 2(e–f) slowed as it entered the Cu-rich layer (~ 640 nm traveled in total), while the speed of the reaction front in Figs. 2(g–h) increased as it entered the Cu20Al80 layer (~ 970 nm in total). Here, the decrease in the concentration of Cu atoms in the bottom layer of configuration 2 facilitated the advancement of the H3PO4 electrolyte into the film, as the increased porosity provided more channels for the Al ions to leave the electrolyte/alloy interface. Configuration 1 displayed the opposite [11, 41]. After 105 and 120 min, the dealloying front advanced to the bottom of the film for the two configurations in Figs. 2i, j, k, l, respectively, though at a slower pace than the earlier time intervals. At this late stage of the reaction, the presence of Al atoms in solution changed the electrochemical interactions between the H3PO4 electrolyte and the remaining portion of the film [6]. Additionally, the foam morphology was expressed differently depending on the layer order. For configuration 1, the reduced reassembly of Cu atoms during dealloying in the top layer preserved the macroscopic columnar structure [37]. While for both configurations, the Cu35Al65, layer [Figs. 2i, k] showed a more continuous connected porous structure as there was a greater degree of rearrangement for the Cu atoms. For configuration 2 specifically, the higher fraction of Cu atoms in the top layer enabled rearrangement to occur earlier in the reaction, producing a more isotropic nanoporous structure through both layers [41].

Effect of chemical potential: Cu–Al/Cu–Zn compositions

In configurations 3 and 4, the expected relative density (20%) was the same for all layers but there was a difference in chemical potential between the selected Cu–Al and Cu–Zn compositions of 2.0 V (Cu/Al) and 1.102 V (Cu/Zn) [30]. This difference in reduction potential between constituents is related to the dissolution rate and is large enough that the Al and Zn are preferentially soluble and the Cu can diffuse along the electrolyte/film boundary layer [6, 29].

After 30 min in solution, the reaction front through configuration 3[Figs. 3a, b] had progressed halfway through the Al-rich layer (~ 530 nm), while the front for configuration 4 [Figs. 3c, d] progressed through the Zn-rich and into the Al-rich layer (~ 1400 nm). After 45 min [Figs. 3e–h], the reaction front in configuration 3 [Figs. 3e–f] had traveled ~ 650 nm in total, while the reaction front in configuration 4 (Figs. 3g–h) had advanced through most of the Al-rich layer, ~ 1700 nm total. For configuration 4, dealloying times of greater than 45 min led to sample disintegration, whereas for configuration 3, the reaction front progressed almost to the layer interface after 60 min, with no viable samples at longer time intervals. In Figs. 3(i, g), both configurations displayed etching along and within the columnar grains of the top layers, as the spaces between the columnar grains provided a pathway for the H3PO4 etchant to penetrate the film vertically through the thickness before lateral etching at the surface in contact with the electrolyte was completed [41]. This was most pronounced in configuration 4 (Fig. 3g), and these results are corroborated by the original as-sputtered microstructure in Fig. 1d, which is strongly columnar with voids between individual columns. For configuration 3, which had more voids in the bottom layer of the as-sputtered film [Fig. 1c], the increase in available etchant pathways once the reaction front reached the compositional interface facilitated the progression through the bottom layer [41]. Furthermore, the Cu20Zn80 alloy (bottom layer of C3 and top layer of C4) had particularly fast dealloying kinetics and therefore a low critical potential threshold to initiate the reaction [22].

An additional factor in the reaction progression is the reduction potential difference between the Al and Zn atoms, which was − 0.0898 V [30]. Work on ternary alloys for dealloying has shown that for a one-step reaction, the presence of a third element that is electrochemically active in the selected dealloying medium will change both the dealloying process and the resulting porous microstructure [42]. While configurations 3–4 in this investigation were bilayers of two distinct compositions and not ternary alloys, the presence of both Al and Zn atoms in solution once the reaction front reached the alloy interface changed the electrochemical potential of the unreacted film in that medium, increasing the speed of the front through the remaining film [6, 42]. Thus, the reaction mechanism after the dealloying front reached the compositional interface was analogous to that of ternary alloys and can be applied to complex dealloying precursors. In the next section, we will discuss the effects of both the chemical potential interface and the relative density interface on the resulting dealloyed foam morphology.

Foam morphology

To illustrate the effects of relative density and chemical potential on the resulting foam morphology, top-surface and cross-sectional SEM micrographs were taken of each configuration after dealloying, at the “final step” time point. Figures in the left column (4a, 4c, 4e, 4 g) show the representative top-surface morphology while figures in the right column (4b, 4d, 4f, 4 h) show the corresponding cross-sectional morphology for configurations 1 thru 4. For configurations with a composition of 20 at.% Cu in the top layer (C1, C3, and C4), the limited reorganizational capacity of the Cu atoms promoted dealloying along the grain boundaries in addition to within the grains. For configuration 2, with Cu35Al65 as the top layer, the grain structure at the top surface is not visible, and the characteristic isotropic porous structure is not constrained by the original grains as there was more reorganization during the reaction [41]. This isotropic porous structure is also visible in the cross-section of configuration 2 [Fig. 4d]. In contrast, the top surface of configurations 1, 3, and 4 [Figs. 4b, f, h] exhibited islands of a more isotropic structure surrounded by voids, corroborating that etching proceeded both along the GBs and within the grains [13, 41]. The expected ligament size for all samples after free corrosion was in the range of ~ 50–60 nm but in actuality, the final foam ligament size between the four sample configurations ranged from 21 to 28 nm [31, 37]. While the ligament and pore size of the final nanoporous structure generally change with composition, it can also be tuned with the choice of electrolyte [28, 43,44,45]. In this case, a combination of a fast reaction time and phosphate-inhibited surface diffusion for the Cu atoms prevented ligament coarsening, as there was minimal time for the Cu to aggregate, producing smaller ligament sizes [13, 37].

Figure 4
figure 4

Representative SEM micrographs of foam morphology for top surface (left column) and cross-section (right column), taken at the final (“final step”) dealloying time interval. Configuration 1 (a–b), configuration 2 (c–d), configuration 3 (e–f), and configuration 4 (g–h). The orange labels in the top left corner indicate sample configuration (Color figure online).

Full size image

In order to further assess the distinctive surface morphologies, Vickers indentation was performed on the top surfaces as shown in Fig. 5. Representative top-surface SEM micrographs of the indented samples for configurations 1, 2, and 3, respectively, are shown in the left column and zoomed-in micrographs depicted in the right column. Both configurations 1 [Figs. 5a–b] and 3 [Figs. 5e–f], which displayed top-surface intergranular cracking in Fig. 4, showed deformation that was contained to the indent. The deformation did not propagate, as the material was not continuous. However, there were slight differences after indentation, and the most visible cracking was in configuration 1, which had both the largest voids and a mismatch in relative density between the top and bottom layers [46]. Thus, the bottom layer has a qualitative effect on the deformation behavior of the bilayer film, and configuration 3, which did not have a relative density gradient, appears less brittle. There was no comparison for configuration 2 [Fig. 5c–d], which was a better-connected porous material [Fig. 4c] with a more isotropic foam morphology that showed no intergranular cracking and instead displayed cracking along the corners and sides of the indent [47]. Configuration 4 is not shown, as the sample was too brittle for testing. Future work should include heat treatments prior to dealloying in order to induce a less sharp compositional boundary between layers and to reduce the effects of the original columnar grain structure on the morphology of the dealloyed foams.

Figure 5
figure 5

Representative SEM micrographs of top-surface Vickers indents performed at the “final step” dealloying time interval (left column) where the blue boxes in the left column denote the location of respective zoom-ins (right column) of the edge of the indent. The orange labels in the top left corner indicate sample configuration as follows: configuration #1 (a–b), configuration #2 (c–d), and configuration #3 (e–f). Configuration #4 not shown as sample was too brittle for testing (Color figure online).

Full size image

Conclusions

This work explores the effect of relative density and chemical potential interfaces on the dealloying front and morphology of gradient NP copper foams. Specifically, bilayer nanoporous films were synthesized via free corrosion dealloying of Cu–Al/Cu–Al (C1, C2) and Cu–Al/Cu–Zn (C3, C4) precursors in concentrated H3PO4 and analyzed across three different time points. All configurations with a lower at% of Cu in the top layer yielded faster initial progression of the dealloying reaction front and etched voids along the columnar GBs, due to reduced rearrangement of Cu atoms. The configuration containing the Cu-rich top layer produced an isotropic NP structure. In contrast, for chemical potential interfaces, the inherently fast dealloying kinetics of the Cu20Zn80 alloy and the electrochemical interactions between Al and Zn atoms in solution facilitated faster reaction times while the large voids between grains in the precursor films increased etching along the GBs. Slower reaction times produced a more uniform morphology, which should be considered when selecting both precursor alloys and layer order. By examining both the effects of relative density and chemical potential, this study highlights the critical factors for the synthesis and development of these graded structures.

Materials and methods

Three copper alloys with a reduction potential difference of 0.5 V or greater were selected in this work, including Cu20Al80, Cu35Al65, and Cu20Zn80. Bilayer films with individual 1 µm layer thicknesses (total film thickness of 2 µm) were synthesized via DC magnetron sputtering. Custom alloy targets were sputtered at 50 W, onto Si < 110 > wafers at a base pressure of 3.1 × 10–6 to 3.3 × 10–6 Pa from 5.08 cm (2 in) sources. Four layer configurations were synthesized: Cu20Al80/Cu35Al65, Cu35Al65/Cu20Al80, Cu20Al80/Cu20Zn80, and Cu20Zn80/Cu20Al80. These are referred to in this work as configurations 1–4, respectively (see Table 1).

The cross-sections of the as-sputtered films were prepared by wire saw (Princeton Scientific WS-25) or by manual fracture and were characterized with SEM (Helios G4 PFIB UXe DualBeam FIB/SEM). The as-sputtered films were dealloyed by free corrosion in concentrated H3PO4 at room temperature for 30 min, 45 min, and until the reaction reached the “final step,” defined here as the last time interval before sample disintegration. All dealloyed samples were cleaned three times with DI water and dried for a minimum of 12 h. The films were not removed from the substrate, and cross-sections of the dealloyed samples were prepared by fracture. SEM micrographs were obtained for both the NP foam top surface and cross-sections, while EDS (Oxford UltimMax 170 Silicon Drift Detector) composition maps were performed solely on the cross-sections. Cross-sectional ligament sizes after the final time interval were processed by ImageJ.

Vickers indentation was performed with a Leco LM-100 Vickers indenter at 10 g of force and for a dwell time of 10 s. Indentation testing was performed on the top surface of the four dealloyed films at the “final step” time for each film. Post-fracture morphology of the indents was analyzed with top-surface SEM imaging.