Open-cell tungsten nanofoams: Chloride ion induced structure modification and mechanical behavior

In this work, a nanoporous structure composed entirely of tungsten ligaments was synthesized by selective phase dissolution of a nanocrystalline tungsten-copper composite in ferric chloride aqueous solution at room temperature. Observation of the tungsten nanoligament modifications in both ferric chloride and hydrochloric acid solution illustrated that the chloride ions accelerate the surface diffusivity of tungsten atoms by two orders of magnitude, thus causing an evolution of the tungsten nanoligaments upon reconstruction and growth with the increase of dissolution time. Using nanoindentation and Vickers’ microhardness measurements, we discovered that the created tungsten foam deforms via a fast densification in combination with the formation of cracks due to a low ligament strength.


Introduction
Metallic foams, i.e., porous metals, are presently the focus of very active research and development activities in numerous industrial applications such as light-weight optics, biomedical implants, high temperature filters, electrodes, electromagnetic shields, sound absorption materials, catalysts and heat exchangers [1]. Generally, in comparison to other porous materials (including plastic, glass and ceramic), porous metals have an integrative advantage in their physical and mechanical properties, including a high electrical or thermal conductivity, superior structural stability at high temperatures, good thermal shock resistance, high mechanical strength and toughness [1]. In general, there are three different types of metallic foams based on the pore structures: open cell, closed cell, and a combination of the two [2]. Closed-cell metallic foams, which do not allow access to their internal surface, are mostly used in structural, load-bearing applications, as they have good mechanical properties. In contrast, open-cell metallic foams are permeable and have very high surface areas as required for flow-through applications or when surface exchange is involved.
Tungsten (W) owns the highest melting temperature among all metals, a low coefficient of thermal expansion, a high density and superior strength [3]. Hence, W foams are dominantly applied as high current density cathodes in high power lamps [4], radiation shields against gamma and X-rays [5], or engineering and structural materials with excellent dynamic compressive properties [6]. For the function of a porous W cathode, homogeneous distribution of open pore channels is essential, as it will be impregnated with electron emissive material [7,8]. Upon emission, the emissive material, which evaporates during the formation of the current, can be constantly supplied to the surface through the open pore channels so that the cathode functions continuously. In addition, by infiltrating the empty open pores with elements of low atomic number, such as compounds of hydrogen and boron, which are preferable for stopping neutron irradiation, the opencell porous W has potential for application in radiation shielding against neutrons [9].
Porous W is mainly processed by powder metallurgy (PM). However, in PM processing, it is difficult to control the porosity and the ratio of opening pores [10]. Chemical vapor deposition (CVD) is another common preparation method for W [11,12]. It has the advantage of forming porous W with controllable pore distribution and pore sizes across several hundreds of micrometers. Tape-casting (TC) is generally used to form thin and flat metallic sheets on a large scale [13]. By applying this method, bulk porous W with an open, biporous structure and pore sizes down to 3-6 μm was produced [13]. The aim of our work is to tailor open-cell nanoporous (np) W, as np metals possess exceptionally high mechanical strength [14,15] and radiation tolerance [16][17][18] compared to conventional metal foams. In addition, the nanopores can limit the evaporation rate of the emissive materials, thus possibly prolonging the lifetime of porous W cathodes [8]. As described above, it is challenging to tailor W foams with pore sizes in a nanometer T range by using PM, CVD or TC. Therefore, new fabrication techniques are required.
For the fabrication of open-cell np materials, the most common method is electrochemical dealloying (ECD), by which a less-noble component is selectively dissolved from a precursor alloy due to the standard electrode potential difference between the constituent elements [19,20]. This typically restricts the application of the method to noble metals and similarly corrosion resistant metals (e.g. Au [19][20][21], Ag [22] and Cu [23]). The liquid-metal dealloying (LMD) is a recently developed method, which utilizes the miscibility difference between alloy components and metallic melts to produce np metals, e.g., np Ti [24,25], Nb [26] and Si [27]. However, coarsening of the nanoporous strucutres probably can take place because of the very high dealloying temperatures. Additionally, chemical etching is required to remove the melts that are solidified in the pores. Another possibility is the vaporphase dealloying (VPD) method. Depending on the vapor pressure difference between solid elements, nanoporous structure can be formed by evaporating one component from alloy [28]. This method is suitable for refractory metals (e.g. Nb, Ta, Mo, V, etc) that may be susceptible to oxidation during ECD and is also capable to yield small ligament sizes that are difficult to achieve via liquid metal dealloying. However, the produced porous film layers are limited to hundreds of nanometers and the sacrificial elements are not fully removed [28]. In our previous work, a unique route involving severe plastic deformation (SPD) of a coarse-grained immiscible tungsten-copper (W-Cu) composite followed by selective dissolution of the nobler Cu was developed [29]. When using ammounium persulfate ((NH 4 ) 2 S 2 O 8 ) as etchant, this technique allows for producing W nanofoams with entirely interconnected open pores, which would not be possible for ECD, LMD and CVD techniques. Furthermore, this approach can be applied to bulk composite systems and yields pure porous structures with a thickness that is challenging to achieve via VPD. Similar to ammonium persulphate ((NH 4 ) 2 S 2 O 8 ) etchant (see Ref. [29]), iron chloride (FeCl 3 ) can dissolve the nobler Cu phase while leaving the less noble W unattacked as well. In addition, FeCl 3 has several advantages over (NH 4 ) 2 S 2 O 8 , such as controlled process parameters (e.g. pH value), stable effectiveness over a long term storage, low product price and low harmfulness to human health. Therefore, in this work we tailor np W by selective dissolution of the nobler Cu phase in FeCl 3 etching solution from a W-Cu nanocomposite. Based on characterization of the resulting porous structures by scanning electron microscopy (SEM), and mechanical tests by depth-sensing nanoindentation and Vicker's microhardness measurements, the following two open questions are addressed: (i) How does the porous structure evolve during the selective dissolution in FeCl 3 solution? (ii) What is the mechanical behavior of the created nanofoams under compressive stress? By taking the np W created by using (NH 4 ) 2 S 2 O 8 etchant [29] as a reference, the influence of the two etchants on the microstructure modifications and the corresponding evolution mechanisms of the porous structure are elucidated.

Experimental procedures
Starting material is a nanocrystalline (nc) W-Cu composite, which was acquired via high pressure torsion (HPT) of a coarse-grained W-20 wt% Cu composite (W 65 Cu 35 in volume ratio and W 58 Cu 42 in atomic ratio) followed by heat treatment at 300℃ in a vacuum furnace for 1 h to reduce the amount of forced mechanical mixing between W and Cu. We have demonstrated that such a W:Cu ratio allows the creation of a network of entirely open pores [29]. A detailed introduction to the HPT procedure of coarse-grained W-Cu composite can be found in Ref. [30,31]. The starting material is comprised of equiaxed W grains with an average size of 15 nm and interconnected Cu phases with 20 nm on average [29]. High resolution transmission electron microscopy (HR-TEM) analysis in Ref. [29] revealed that the starting material contains three different types of interfaces, i.e., oxygen-free and low defective W grain boundaries (GBs), W GBs decorated by Cu and O atoms, and W/ Cu phase boundaries (PBs) with a trace amount of O. W moving entities (including adatoms, step adatoms and kink atoms described in the terrace ledge kink model [32]) will be formed once Cu and O atoms at the latter two interfaces are removed [29].
Nanoporous W was developed by selective phase dissolution of the starting material in a concentrated FeCl 3 aqueous solution (Table 1, FeCl 3 , 5.5 mol/L) at room temperature (RT). To suppress the hydrolysis of Fe 3+ ions, hydrochloric acid (HCl) was added into the FeCl 3 solution to maintain a pH below 2. After a scheduled etching time (see Table 1), samples were taken out from the beaker and transferred into distilled water for 30 min to remove the etchants. The samples were further immersed in 2 wt% sodium hydroxide (NaOH) for 30 min to remove the tungsten oxides (WO 3 ), which is possibly formed due to the hydrogen evolutional corrosion. At the end, the samples were cleaned in distilled water for several times and dried by the warm air. Visual inspection shows that the as-etched samples lose their metallic luster and exhibit noticeable cracks, blisters and delaminations.
For observing the microstructure, a scanning electron microscope (SEM, LEO 1525, Carl Zeiss GmbH, Germany) equipped with an energy dispersive X-ray spectrometer (EDX) was utilized. Here, the surface of porous metals for SEM observation was obtained by mirror polishing before etching. To unveil the microstructure beneath the surface, a femtosecond laser system [33] was used to produce cross-sections of the specimen. The laser system contains an integrated femtosecond laser (OneFive Origami 10XP) to allow a fast processing of arbitrary structures and a focused ion beam (FIB) column (Orsay Physics Ga + ion FIB) to offer a precise rework. The overall composition of the investigated materials was determined by EDX analysis.
The mechanical response of the created foams under compressive stress was investigated by depth-sensing nanoindentation and Vicker's microhardness (Buehler MacroMet® 5100 Series, Japan) measurements. For the nanoindentation, a G200 Nanoindenter (KLA Corporation, USA) equipped with a Berkovich tip (Synton-MDP, Switzerland) is utilized. All indentations were carried out at RT on the surfaces of the specimens that had been polished before etching. Nine nanoindentation tests performed with a constant strain rate of 0.05 s −1 were conducted on each specimen. The continuous stiffness measurement (CSM) method was applied to track hardness as a function of the indentation depth. Vickers' microhardness performed with 10 s holding was used as an indication for the deformability and bonding strength of the W ligaments.  (Fig. 1a). As the etching time prolongs to 4 h, the surface is covered by flaky ligaments with a length of 160 nm on average. The length of these flaky ligaments increases to 280 nm when further prolonging the etching time to 18 h. Together with the ligament growth, the number of cracks, blisters and delaminations or even fragmentation increases (see insert image in Fig. 1c). As shown in Fig. 1d, only W is identified by EDX for each as- To unveil the possible influence of HCl solution on the resulting peculiar porous structure, the specimens etched for 1 and 2 h in the FeCl 3 solution were further immersed in diluted HCl solution (pH = 2) for 18 h. Note that the 1 h etched sample exhibits a surface of cellular ligaments similar to the 2 h etched one (see Fig. 1a). After 18 h immersing in the diluted HCl solution, the cellular ligaments on 2 h etched specimen evolved into flaky ones and the average ligament length increased to~480 nm (Fig. 2a), which is much bigger than the ligaments on the surface of 18 h FeCl 3 etched specimen. On the 1 h etched specimen, aggregates of flaky ligaments are observed after the further immersing in the diluted HCl solution ( Fig. 2b and 2c). It reveals that the ligament structure evolves with nucleation and aggregation of new W grains. The aggregation process of new grains causes the depletion of W in certain areas. From the surface morphology of the 2 h etched specimen, it is deduced that the depletion areas can be further filled by other aggregates of W ligaments, once there are more W atoms participating in the movement along the metal/solution interfaces.

Microstructure evolution
As the goal is to create a np bulk material, the depth-depended evolution of porosity is analyzed in Fig. 3. Fig. 3a and b show a femtosecond laser processed and a FIB repolished cross-sections, respectively. The position of the observed cross-sections on the 1/6 HPT specimen is schematically displayed in Fig. 3c. Two layers with different laser machining morphologies are well distinguishable in Fig. 3a, which are probably caused by different ablation characteristics of nc W-Cu and porous W. The etching front located in a depth of~15 μm is parallel to the specimen surface, demonstrating a uniform dissolution of Cu and permeation of the etching solution over the observed area. EDX analysis shows that Cu is absent in the upper layer (i.e., as-etched region), revealing that the created W foam is entirely open cell. After reworking the cross-section with FIB, the pores (dark contrast) which are uniformly distributed in the W matrix (bright contrast) are clearly observed. A quantitative measurement shows that these pores have a size of~10-30 nm, which is comparable to the size of Cu particles (20 nm) in the nanocrystalline W-Cu precursor, but much smaller than the pores observed on the sample surface (see Fig. 1c). Such small pores are probably correlated with the depth-depended Cu etching and W grain reconstruction procedures. To unveil the nanoporous structure along the cross-section, microstructure observations in the transmission electron microscopy (TEM) are required. However, the fabrication of TEM specimens turned out to be very challenging under the present conditions of a very thin foam layer (~15 μm) and depth-depended porous structures. Therefore, a detailed characterization of the nanoporous structure is not presented here.
Porosity is strongly correlated to the foam mechanical behavior as discussed in the following. However, due to the W grain reconstruction and the peculiar foam morphology, it is problematical to make a direct calculation of the porosity from the amount of Cu in the W matrix (35 vol%), or to measure the porosity simply from a two-dimensional SEM image. Given the formation and gradual evolution of cracks, blisters and delaminations at the sample surface (see Fig. 1), it is deduced that the porosity is higher than 35 vol%. At larger sample depth, where Fig. 3b represents for, the foam seems to have a porosity  comparable to the Cu volume percentage (i.e., 35 vol%).

Mechanical behavior
The nanoindentation response of a 4 h FeCl 3 etched specimen is displayed in Fig. 4. Hardness (H) and reduced Young's modulus (E r ), calculated by analyzing the load-displacement indentations using the Oliver-Pharr method [34], are given in Fig. 4a and b versus displacement. At the applied load of 670 mN, a maximum depth of around of 2 μm is reached, allowing us to study the mechanical response of an ensemble of ligaments and pores instead of individual ligaments [35]. The mechanical behavior of the created foam can be revealed from the indentation depths smaller than 1 μm, as the influence from the unetched substrate layer is no longer negligible at larger indentation depths. At indentation depths ranging from 30 to 100 nm, the hardness is as low as~1 GPa, corresponding to a foam yield strength (σ) of 380 MPa via the constraint factor C* by H = C*·σ. Note that the value of C* (0.26, Ref. [29]) for the reference nanofoam was taken for the calculation of σ. In comparison to the reference nanofoam, which has a yield strength value about 2 GPa, the nanofoam tailored by using FeCl 3 solution has a much lower strength. The poor mechanical strength of the foam created by FeCl 3 etching should be dominantly controlled by the reconstructed ligament structures. The porosity, which is generally higher than the reference foam (35 vol%, Ref. [29]) may to some extent   also contribute to its poor strength. At the indentation depths below 30 nm, due to the indentation size effect or contact imperfections, the foam tends to be harder than 1 GPa. With the increase of the indentation depths from 100 nm, the hardness of the created porous W increases initially fast and then saturates at indentation depths above 1 μm. The reduced Young's modulus shown in Fig. 4b exhibits a changing trend similar to the hardness, i.e., E r increases rapidly from~60 GPa at indentation depths~100 nm and saturates at a value of 180 GPa. Note that 180 GPa is much smaller than the E r of solid nanocrystalline W (335.8 GPa, Ref. [29]), revealing its porous nature during the indentation process. Residual indentation impression in Fig. 4c shows that deformation of the created foam is contained in the contact area without material pileup or deformation adjacent to the residual impression. Research of Leitner et al. [36] suggests that the created foam deformed via a collapsing and compacting of the evolved peculiar ligaments (i.e. densification), thus leading to the increase of hardness and Young's modulus for indentation depths between 100 nm and 1 μm.  [29]. At the load of 10 N, indentation depths around 8.8 and 10.6 μm were calculated for the FeCl 3 and the (NH 4 ) 2 S 2 O 8 etched specimens, which correspond to approximately 60 and 70% of the total thickness of the as-etched layers, respectively. Such deep indentation depths give rise to the issue of substrate effects. Thus, the Vicker's hardness values are no longer reliable and will not be discussed. As seen from Fig. 5, severe cracking and fracturing are observed around the indentation impression of the FeCl 3 etched specimen, while only minor cracks are noticed in the indentation impression of np W tailored by using (NH 4 ) 2 S 2 O 8 . This behavior is in good agreement with the nanoindentation tests, which unveils that the ligament strength in the nanofoam created by using FeCl 3 solution is much weaker than the nanofoam created in (NH 4 ) 2 S 2 O 8 solution.

Discussion
As described above, when using FeCl 3 as etchant, equiaxed W grains in the nc W-Cu starting material evolved into peculiar W ligaments due to grain reconstruction, including grain nucleation, aggregation and growth. This grain reconstruction behavior can be described by the solid/liquid interface diffusion mechanism, which has been widely proposed to explain the nucleation and growth processes of ligaments of np metals during dealloying of crystalline precursors [19,[37][38][39][40]. As described by the Arrhenius equation given below, the diffusion of metal atoms along the metal/solution interfaces is a thermally activated process [41]: where D s is the surface diffusion coefficient, D 0 is the pre-exponential factor taken as a constant, R is the gas constant, T is the etching temperature, and H Δ S * is the activation enthalpy for the surface diffusion process. The activation enthalpy H Δ S * is composed of two parts, H Δ F * and H Δ D * , representing the activation enthalpies for forming the moving entities and their diffusion process, respectively [41]. A list of intrinsic diffusion parameters for W on three different W crystal planes at RT is given in Table 2. Note that Table 2 does not contain all data available in the literature. The variation in D s values for W along different crystal planes gives an explanation for the formation of anisometric ligaments, i.e., flaky ligaments. In experiments where the D s was determined, the calculation of H Δ S * and D 0 was derived from the slope and intercept (extrapolation to infinite temperature) of an Arrhenius plot. Therefore, the H Δ S * and D 0 listed in Table 2 are with certain errors. To simplify the calculation as well as to minimize the errors, the average D s and H Δ S * for W intrinsic diffusion are determined from the listed intrinsic diffusion systems by neglecting their changes with crystal planes.
Referring to the isothermal grain growth in polycrystalline materials [46], the interplay between ligament radius r and the dealloying conditions can be modeled using the following diffusion equation [47]: where γ is the surface free energy, a is the lattice parameter, and t is the dealloying time. Integration of Eq. (2) between r 0 and r, for t = 0 and t = t, respectively, leads to where r 0 represents the radius of the coalescing particle at t = 0. In the case of our work, r 0 is taken as the ligament length measured for 4 h etched specimen (Fig. 1b). Eq. (3) allows a quantitative determination of D s from the experimentally measured ligament lengths. Values for parameters in Eq. (3) are listed in Table 3. Note that 2500 ergs/cm 2 is an average energy value for γ, which neglects the fact that its magnitude changes with crystal planes [48] or the etching solution. A D s value of 5.7 × 10 -16 cm 2 /s is therefore calculated by using Eq. (3). In comparison to the intrinsic diffusion coefficient (4.8 × 10 -18 cm 2 /s), the surface diffusivity of W atoms in the used FeCl 3 solution is thereby accelerated by two orders of magnitude. According to Eqs. (1) and (3), the changes of H Δ S * and r as a function of the varying D s are displayed in Fig. 6. It is clearly seen such a prominent increase of D s corresponds to a H Δ S * reduction of about 15%, i.e., from 84.5 to 72 kJ/mol, which should be due to the interaction between solvent ions and the surface W atoms [41,49]. Moreover, it is noticed from Fig. 6 that no ligament growth is noticeable when D s is lower than 1 × 10 -17 cm 2 /s, or when H Δ S * decreases by less than 2%. From the perspective of thermodynamics, this finding provides a good explanation for the case of using (NH 4 ) 2 S 2 O 8 as etchant [29], in which no grain modifications were observed. It unveils that the influence of (NH 4 ) 2 S 2 O 8 solution on the ligament structure is negligible because the interaction of solvent ions in (NH 4 ) 2 S 2 O 8 solution with the surface W atoms is extremely weak or even absent. The question remains as to which ions exactly interact with the surface W atoms. As described in Section 3.1, cellular ligaments on the 2 h etched surface evolved into flaky ligaments in both acidified FeCl 3 and diluted HCl solution. It shows that H + or Clions, which exist in both solutions, are possibly the interacting ions. But H + ions are excluded, as in (NH 4 ) 2 S 2 O 8 solution, where H + with a concentration   [29], surface moving entities diffuse with a speed probably smaller than 1 × 10 -17 cm 2 /s according to Fig. 6, as no ligament evolution was observed. Therefore, it is concluded that W surface diffusion is drastically facilitated by the interaction between Clions and surface W atoms. This conclusion is also in good accordance with reports showing that Cl --containing solutions (e.g. neutral NaCl solution [21,50], or acidic HCl solution [51,52]) have an acceleration effect on the surface diffusion of the noble Au during dealloying. A common feature accompanied with dealloying is the formation of cracks within the np metals [53][54][55]. The cracking phenomenon during dealloying is generally regarded as a result of the rate of stress increase resulting from the removal of active atoms overweighing the rate of stress relief from surface diffusion of remaining atoms [56]. When using FeCl 3 as the etchant, the stress increase rate seems to be faster than the stress relief rate. Unveiled by the indentation tests, the created ligaments are not strong enough to withstand the residual internal stress, causing ultimately the formation of cracks in the as-etched layer. Accompanying with the cracks, blisters and delaminations are observed in the as-etched layer as well. Formation of the blisters is probably correlated with the grain nucleation and aggregation processes, while appearance of the delaminations can be interpreted by the mechanical mismatch between different ligament structures.

Conclusions
We aimed at creating a tungsten nanofoam by selective phase dissolution from nanocrystalline W-Cu in an FeCl 3 solution, exploring its porous structure evolution behavior during the selective dissolution process and its mechanical response under compressive stress. The Cu phase turned out to be selectively removed in the used FeCl 3 solution, leading to a successful formation of nanoporous tungsten. However, the formed porous structure was unstable. Its ligaments evolved from a cellular to a flaky structure with the prolonging of etching time.
Observation of cellular ligament evolution in the acidified FeCl 3 solution and the diluted HCl solution showed that Clions can reduce the activation enthalpy for W surface diffusion by interacting with W atoms, thus accelerating the mobility of W atoms. Based on the surfacediffusion controlled coarsening mechanism, the surface diffusion coefficient of W atoms in the used FeCl 3 solution was calculated as 5.7 × 10 -16 cm 2 /s, which is around two orders of magnitude faster than the intrinsic W diffusion value. Nanoindentation tests revealed that the created foam has a ligament strength as low as 380 MPa and deforms via a fast collapsing and compacting of the ligaments. The low ligament strength ultimately led to the formation of cracks during the etching process.
In conclusion, this work provides a novel methodology for creating nanoporous W. Combining our previous work [29], which focused on tailoring W nanofoam by using (NH 4 ) 2 S 2 O 8 solution, with the current results on using FeCl 3 as etchant, it is shown that the type of etching solution significantly influences the resulting W porous structures. Due to the nanocrystalline nature of the W-Cu precursors, it is hardly to produce W foam layers above a thickness of tens of microns. The low ligament strength in the present case also limits the use of the created foam in structural applications. However, the created W foams are nanostructured and entirely open cell, thus having a potential to be used as high current density cathodes. In addition, the fast surface diffusion characteristics of W atoms in Cl --containing solutions provide an unique way to tailor porous W materials with various structural parameters, such as ligament shapes and sizes, for engineering applications requiring different physical, mechanical and other properties. For instance, due to the exceptional resistance of W to melting, W foam foils can be used to filter high temperature liquid and gas in aerospace, power electronics and metallurgical industries. The filtering speed meanwhile can be adjusted by increasing/decreasing the layers of the thin foam foils, variable ligament shapes and sizes.

Author contributions
M.Z. designed the experiment, fabricated the materials and conducted the microstructure characterization and data analysis. M.J.P. performed the cutting and SEM imaging of the cross-section (Fig. 4). This work was carried out under the supervision of D.K. M.Z. wrote the original paper draft. D.K. and M.J.P. reviewed the draft paper and offered many valuable suggestions to modify the draft paper.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.