1 Introduction

Nanomultilayers (NMLs) are thin layered structures formed by the sequential repetition of two or more single-material layers with individual thicknesses in the nanometric range. They are the subject of interest of many research studies because they can display unique properties and lead to outstanding performances, which could not be obtained by the employment of the relative monolithic components (Ref 1). For example, multilayer structures can show a very high reflectivity, hence being adopted as mirrors tailored for specific wavelengths (Ref 2,3,4,5). Other multilayers exhibit unique magnetic properties (Ref 6,7,8), or superior yield strength and hardness compared to the one predicted by the rule of mixture of constituent materials, as it was observed in Mo/Pt (Ref 9), Ag/Nb (Ref 10), Cu/Ni (Ref 11), and Cu/W (Ref 12, 13).

Because of the high density of interfaces in NMLs, they generally possess a high excess energy in the form of interfacial energy. Moreover, the fabrication techniques normally employed to produce them, e.g., DC magnetron sputtering, induce the build-up of internal stress (Ref 14), which contributes to the overall energy of the system with an elastic energy term in first approximation. In addition, structural defects can occur, which further increase the excess energy. For these reasons, NMLs are often in a highly metastable state and can undergo a structural transformation driven by stress relaxation and interface energy minimization if sufficient activation energy is supplied to the system, e.g., as thermal energy. For example, Cu/W NMLs transform into an isotropic Cu/W nanocomposite when annealed at 800 °C, with the final structure consisting of W particles embedded in a Cu matrix (Ref 15, 16). Depending on the temperature range, the transformation takes place following specific steps. In the low-temperature range (300-400 °C), the mobility of Cu becomes significant and it starts to diffuse. However, the diffusion of Cu atoms is not only intra-layer, but also inter-layer with W-W grain boundaries representing a preferential path for Cu inter-layer mobility (Ref 17). Eventually, Cu was also observed to diffuse to the surface, i.e., outside of the multilayer structure (Ref 15). In successive stages of the transformation at higher temperature, Cu was diffusing back inside the NML system (500-600 °C), but the real NML-to-NC transformation was taking place only when the mobility of W was also starting to become significant (~700 °C). Similar behavior was observed also for other NML systems, e.g., Ag/AlN (Ref 18,19,20).

Cu diffusion in the multilayers is driven by stress relaxation and structural defects annihilation (Ref 17, 21). Hence, stress gradients and defects concentrations can be intentionally induced in the structure to create diffusional pathways for a patterned and controlled Cu diffusion and outflow. For example, Lin et al. (Ref 21) used laser radiation to locally create defects and stress fields in Cu/AlN multilayers, achieving in this way a preferential Cu outflow along the drawn structures at temperatures as low as 360 °C.

In this work, we assess at first the feasibility of inducing a patterned Cu outflow in Cu/W NMLs. To create preferential paths for the out-diffusion of Cu, we modified the pristine surface of multilayers by focused ion beam (FIB) bombardment, which is a commonly adopted technique to nanomanufacture thin films and coatings (Ref 22). Subsequently, we investigated the effect of structural parameters of multilayers (i.e., ordered and compact structure with compressive stress vs. disordered structure with open grain boundaries and tensile stress) and of processing parameters (i.e., annealing temperature, FIB current, and scanning time) for tailoring the Cu outflow.

2 Methods

Cu/W nanomultilayers (NMLs) were deposited by DC magnetron sputtering with an Orion 5 system from AJA International, Inc. The multilayer structure consists of 10 repetitions of 10 nm of Cu and 10 nm of W, up to a total thickness of 200 nm, plus an additional 25 nm of W as a buffer layer between the substrate and the multilayer (hence, the total thickness is 225 nm). Substrates are Si (001) wafer squares with side length of 1 cm and thickness of 0.5 mm, with a 90-nm-thick a-SixN coating on top. Prior to deposition, the substrates underwent four cleaning steps: The first three steps are 4-min-long ultrasonic baths in acetone, ethanol, and isopropanol, respectively; the fourth cleaning step is a plasma cleaning inside the deposition chamber obtained by applying to the substrate an RF bias for 2 minutes at 50 W of power and 2 Pa of Ar-pressure. Two types of NMLs were deposited by choosing two different Ar-pressures, while the gun power was kept at 80 W both for Cu and for W in every deposition. The two chosen Ar-pressures of 0.27 and 2 Pa lead to multilayers with different microstructures and residual stress states (Ref 16). Indeed, Cu/W NMLs grown at 0.27 Pa have an ordered and compact microstructure and an in-plane compressive stress, whereas Cu/W NMLs grown at 2 Pa have a more disordered microstructure characterized by open grain boundaries and an in-plane tensile stress.

A FEI Helios NanoLab 660 dual-beam scanning electron microscopy (SEM) and Ga FIB system was used to generate the defect patterns on the surface of as-deposited NMLs (from now on called trenches) and to acquire SEM images of the surface and of the cross section of selected samples. A scheme of the FIB-processing procedure is given in Fig. 1. The trenches were generated with a length of 20 μm and with different depths, which depend on the set of FIB-processing parameters and stress/microstructure of samples. The chosen set of parameters and their relation with the structure of trenches are discussed in Section 3. Annealing treatments at 500 °C were performed in a tube furnace in reducing atmosphere (Ar + 5 vol.% H2) for a fixed period of 100 min.

Fig. 1
figure 1

Experimental procedure: (a) A FIB is scanned on top of the multilayer to damage the surface; (b) resulting trench after the FIB-processing step; (c) Cu outflow over the trench after annealing at 500 °C for 100 min in an atmosphere of Ar + 5 vol.% H2

Full size image

3 Experimental Results and Discussion

To showcase the influence of FIB-processing parameters on the Cu outflow, three combinations of Ga ion-beam current and scanning time are used in this work. They are labeled A, B, and C and are reported in Table 1. The main effect of the FIB-processing is the ablation of the first layers of the multilayer. In addition, impinging Ga ions introduce localized stress fields and increased defect concentrations around the scanned zone, either through atomic peening, and/or by implantation into the crystal lattice. The trenches generated by the FIB-processing on NMLs with tensile and compressive stress state for all sets of parameters are shown in Fig. 2. Images of the surface trenches at higher magnification (insets) highlight the differences of the surficial microstructure between compressive and tensile samples and of the respective trenches generated by FIB-processing. As discussed elsewhere (see Ref 16, 23), NMLs with tensile stress have a more disordered structure, characterized by rougher interfaces and open grain boundaries, compared to compressive NMLs. For the compressive Cu/W NMLs, stress values of − 3.8 and − 1.8 GPa were measured by XRD in W and Cu, respectively, whereas in tensile NMLs stress values of 1.5 and 0.7 GPa in W and Cu were measured (Ref 16). Notably, clear differences in appearance of the trenches can be observed for NMLs with different stress states. This may be attributed to a different interaction between FIB and NML surface due to different roughness, but also to a different volume density and structure of the ablated NML material.

Table 1 The three sets of FIB parameters used to create the trenches on the multilayers (trench length 20 μm)
Full size table
Fig. 2
figure 2

Trenches left by sets of parameters (A), (B), and (C) on compressive and tensile NMLs. Insets are zoomed-in images. For sets (A) and (B), there is also an image of the cross section

Full size image

Two exemplary cross sections for as-prepared trenches are presented in Fig. 3. They belong to NMLs with compressive stress, and the trenches are produced with the parameter sets A (left) and B (right). In the trench region for parameter set B, the thickness of the multilayer is reduced from its initial thickness (of ca. 225 nm) to ca. 170 nm, which indicates a removal of the first 5-6 monolayers by the FIB-processing. Parameter set A is less aggressive due to the shorter scanning time and consequently leads to a much less pronounced ablation.

Fig. 3
figure 3

Cross sections of NMLs with compressive stress with trenches left by (a) set of parameters A and (b) set of parameters B

Full size image

SEM images of the trenches after annealing are given in Fig. 4. The panels (a), (b) and (c) show FIB trenches produced with different set of parameters A, B and C (ion current and scanning time) (see Table 1) in the compressive NMLs; similarly, the panels (e), (f) and (g) show the case for tensile stress NMLs.

Fig. 4
figure 4

(a), (b), (c) SEM images of NMLs under compressive stress around the trenches produced with the sets of parameters A, B, and C, respectively; (d), (e), (f) SEM images analogous to (a), (b), and (c), but for NMLs under tensile stress. Insets show the marked regions at higher magnification

Full size image

As can be derived from Fig. 4(a) and (d), the ablation created by parameter set A is not enough to induce the Cu outflow at 500 °C, neither for the compressive sample nor for the tensile one. For set B, Cu outflow happens for both stress/microstructure conditions (Fig. 4b and e) and is more pronounced for the tensile sample (Fig. 4e). Indeed, in this case Cu particles cover the whole length of the trench, reaching in some cases considerable dimensions (over 2 μm in length, 1.5 μm in width, and 0.6 μm in height). The material outflow forming the Cu particles originates from within the multilayer structure. The Cu outflow can thus be viewed as a kind of thin-film dewetting (Ref 24) within the multilayers that reduces the excess energy of the NML and is facilitated and steered by the trenches.

The set C of FIB parameters is more favorable to induce the Cu outflow in compressive NMLs. Indeed, the number of Cu particles in Fig. 4(c) is higher than in Fig. 4(b). A massive Cu outflow is present also in the tensile NML. However, the number of big Cu particles is lower than the one obtained with parameters B (compare Fig. 4(f) with Fig. 4(e)). It is also worth to note that in Fig. 4(e), the small particles are placed at the edges of the trench, not on top of it as in the other figures.

To better investigate the effects of the annealing process, cross-sectional images were acquired, shown in Fig. 5 for the compressive NML and in Fig. 6 for the tensile NML. In Fig. 5(a), the effects of FIB-processing with parameters A and subsequent annealing are hardly discernible. In Fig. 5(b) and (d), cross-sectional images of compressive multilayers with trenches produced with parameter set B are reported. The surface ablation in Fig. 5(b) looks minimal, limited to only the first monolayers. Hence, in direct comparison with the cross section before the annealing reported in Fig. 3, it appears that the annealing procedure led to a local recovery of the multilayer structure. In Fig. 5(d), the Cu particle grown on top of the NML is as thick as the multilayer itself. Interestingly, the multilayer structure under the particle seems uneffected by the FIB-processing step, whereas a small difference can be perceived on the sides of the particle: The first two W monolayers on the top are sintered together, resulting in a surficial W layer with a thickness of 20 nm. Figure 5(c) and (e) shows the cross sections corresponding to set C of FIB-processing parameters. In this case, damage to the NML was apparently too substantial to achieve a recovery of the surface. As in case of parameter set B, particles of substatial size in comparison with the original NML thickness could be observed.

Fig. 5
figure 5

SEM cross-sectional images of NMLs under compressive stress after annealing. (a), (b), (c) show trenches created with sets of parameters A, B, and C, respectively, at locations without particles. (d) and (e) show selected particles ontop of the trenches parameter sets B and C. The apparent gap between NML and particle in (e) is caused by a non-centric cut through the particle. (Note: no particles were observed with the set A of parameters)

Full size image
Fig. 6
figure 6

SEM cross-sectional images of NMLs under tensile stress after annealing. (a), (b), (c) show a region without particles for trenches created with sets of parameters A, B, and C, respectively. (d) and (e) are images of a particle ontop of the trenches created with sets of parameters B and C. (Note: no particles were observed with the set A of parameters)

Full size image

In contrast to compressive NMLs, the structure of the tensile NML is porous and has open grain boundaries. This can be observed in the corresponding cross sections (Fig. 6). The cross section in Fig. 6(a) for parameters set A appears very similar to the as-deposited one (reported in Ref 16); hence, the annealing process did not affect the microstructure. As in case of the compressive NML, the trench obtained with parameters A is hardly visible. Interestingly, in this sample (and only in this sample), the onset of NML delamination from the substrate due to the high tensile stress can be observed: the bottom part of the NML is still in contact with the substrate only in a 200-nm-wide region in the middle of Fig. 6(a), whereas on the sides, a gap has already formed between the bottom part of the multilayer and the substrate.

In Fig. 6(b) and (d), cross sections of tensile multilayers with trenches obtained with parameter set B are shown. This parameter set again leads to formation of Cu particles with substantial size (Fig. 6d). It is generated by the Cu outflow of the multilayer, which reduces its thickness from the original 225 nm to the 160-170 nm measured under the particle. It is also noteworthy that the multilayer structure below the trench and at its sides does not show pores and open grain boundaries anymore. This suggests that the Cu diffusion and outflow during the annealing process also filled the voids and led to compacting of the structure. These structural changes also lower the excess energy and reduce the overall reactivity of the multilayers. Pores and open grain boundaries can still be observed on the lateral regions, which were less affected by the Cu outflow.

Analogous observations can be made for Fig. 6(c) and (e), where trenches were obtained through FIB-processing with parameters C. Here, the ablation of FIB-processing is very pronounced and destroyed the entire multilayer, reaching the substrate. The annealing step led to a Cu outflow that compacted the regions on the sides of the trench, but did not completely fill the trench itself (Fig. 6e).

4 Conclusions

In this work, the feasibility of achieving a patterned Cu outflow in Cu/W multilayers is demonstrated. The directional outflow is induced by locally ablating the multilayer with FIB and by subsequently annealing the NML for 100 min at 500 °C. The influence of FIB-processing parameters and stress/microstructure of multilayers on the Cu outflow was investigated in detail. In general, NMLs under tensile stress with a disordered structure are more favorable to achieve outflow. In this case, big Cu particles can be observed on the surface on top of the trenches, while the multilayer structure underneath compacts itself, associated with the annihilation of pores and open grain boundaries. FIB parameters were shown also to play a role in the amount of directional Cu outflow for a given microstructure and stress state of the multilayers. A more pronounced Cu outflow with a bigger ablation is observed only in NMLs under compressive stress, whereas in NMLs under tensile stress the most favorable set of parameters to foster the Cu outflow was the middle one (B).

Although further studies are needed to optimize the Cu directional outflow at even lower temperatures and shorter times, this work reveals a promising methodology to achieve a controlled Cu outflow and patterning with (sub-)micron dimensions at temperatures much lower than the melting point of Cu. Specially designed patterns to create structures and to join components with complex shapes in the micrometer-to-nanometer scale by the Cu outflow can thus be achieved.