The deposition process was visually accompanied by gradual color change of the surface of PS from black to red which is typical for copper. Gas bubbles released from the surface of the sample were also observed. The activity of the gas evolution was weakened with the increase of deposition time. According to
], the released gas is hydrogen which is a product of the redox reactions. The decrease of its evolution means a slowing of the process.
shows top-view SEM images of PS1 (a, b) and PS2 (c, d), both of which were immersed into the basic solution of copper sulphate for 4 s (a, c) and 180 s (b, d). The maximum deposition time (180 s) was chosen because at that moment, hydrogen release almost stopped for both types of PS, i.e., the redox process was too weak for the valuable amount of copper deposition or finished. It is seen that the porous surface is covered with copper particles of various dimensions. The phase composition of copper deposited by displacement method on PS1 and PS2 was assessed in earlier research from their XRD patterns
]. It was found that all copper deposits had a polycrystalline nature with prevalent growth of (111)-oriented crystals. Here, we analyzed the XRD patterns (presented in
]) and, following Scherrer's formula, calculated that Cu particles consist of 2-nm nanocrystals. That is in good agreement with the known data of EBSD analysis
presents particle size distribution histograms counted from the SEM images (Figure
). The accuracy of size evaluation did not exceed 2 nm. Early stages of deposition for both types of PS resulted in the growth of separated copper particles (Figure
a,c) with diameters which varied in the range of 20 to 100 nm (Figure
a,c). However, higher porosity of PS template (PS2) inhibits the process of particle size increase (Figure
c) as the average diameter of Cu particles on PS2 is slightly less than that on PS1 (Figure
a,c). It is probably caused by less number of electrons from the smaller Si elements of the PS2 skeleton (see the structure of PS in
]). Further deposition of Cu on PS1 led to a slight increase in the diameter of copper particles (Figure
b), and their coalescence (Figure
b) resulted in a tightly packed film formation. On the other hand, Cu deposition for 180 s significantly shifted the particle size range from 20 to 100 nm to 80 to 280 nm (Figure
c,d), while the morphology of copper deposit still represents the separated particles. At the same time, the view of the underlying porous material differs in comparison with Figure
c: the sponge converted into a grainy porous structure. To understand the changes, we recognized the paper
] that reported AES analysis of Cu/PS1 and Cu/PS2 formed for 180 s of Cu deposition. The first one represented a nanocomposite with the amount of Cu decreasing from 95% to 15% at pore deepening, while the second structure contained almost no Si traces. Combining those data with Figure
b,d, we suppose that 180-s processing leads to (1) Cu/PS1 nanocomposite formation with a prevalent location of copper deposit as a film in the near surface region of the porous layer and (2) PS2 conversion into a porous copper layer which is partially covered with separated Cu particles of 160-nm average diameter. So, the Cu deposit structure greatly depends on the type of PS template. The porosity and thickness of PS are managed by anodic current density and time of anodization, respectively
]. The distance between pore centers is a constant parameter
]. An increase of current density leads to the increase of pore channel diameter. As a result, the porosity of PS increases simultaneously with the thinning of the pore walls. It is very likely that the complete displacement of PS2 is caused by a better reagent exchange in conditions of wider pore channels and smaller elements of Si skeleton in PS.
SEM top views of Cu NPs. Cu NPs were grown on PS with a thickness of 1 μm and porosity of 50% to 55% (a, b) and 80% to 85% (c, d) by displacement deposition for 4 s (a, c) and 180 s (b, d).
Size distribution histograms of copper NPs. Histograms were calculated for Cu NPs grown on PS with a thickness of 1 μm and porosity of 50% to 55% (a, b) and 80% to 85% (c, d) by displacement deposition for 4 s (a, c) and 180 s (b, d).
Formation of copper particles of the nanoscale range on the outer PS surface requires the use of PS of only 1-μm thickness. A thin porous layer allows minimizing the amount of reagents and deposition time needed for the growth of NPs
]. However, that limited the thickness of the converted porous copper film just to 1 μm. In trying to study the properties of such porous copper, we separated it from the Si substrate, but the metallic film had too weak mechanical strength and, in free form, represented pieces of about 25-mm2
area. Thus, to further work with the free porous copper, the increase of its thickness was highly required.
Supposing the formation of a thicker layer of porous copper, we used PS3 (see Table
) in connection with prolonged copper displacement deposition. The porosity of PS3 was the same as that of PS2, but the pores deepened up to 7 μm with increased anodization time. Visual monitoring of Cu deposition process showed the formation of copper deposit on the outer surface of PS3. Starting from 900 to 1020 s, we did not observe the evolution of hydrogen bubbles, so the time of PS3 immersion was limited to 1020 s. Figure
shows SEM images in top (a) and cross-sectional (b) views of PS3 immersed into the basic solution for copper deposition for 1020 s. The same sample was analyzed by XRD in
] which revealed polycrystalline copper presence in its composition as well as small amounts of Cu2
O. The top of the porous layer is covered with a noncontinuous copper film which consists of coalesced particles. The correct evaluation of particle diameters is difficult because their boundaries are unclear, but some of them might be measured as 40 to 50 nm in diameter (Figure
a). The thickness of the film does not exceed 450 to 500 nm (Figure
b). The dissolution of PS3 took place as its thickness decreased from 7 to 5.8 to 6 μm. To find the depth of copper penetration into PS3, energy-dispersive X-ray spectroscopy (EDX) scanning was performed (Figure
). Despite metal penetration to the bottom of pores, the content of copper in the porous material was too small in comparison with silicon and decreased from the entrance to the bottom of pores. The presence of oxygen is probably caused by the formation of Cu2
, and Oy
as well as the small amount of penetrated carbon from the air during sample drying. We suppose that the distribution of copper in PS3 is caused by poor and slow exchange of reagents in the depth of pores coupled with the rapid growth of the copper film at the top of the porous layer, i.e., unevenness of the displacement process rate along the pore length. Finally, the upper metallic film closes the entrances of pore channels and prevents further redox reactions in porous volume. We carried out an additional experiment connected with the increase of the porosity of PS, but it led to the destruction of the PS skeleton. At the same time, decreasing the PS thickness to 3 μm did not provide a significant increase of copper amount as the upper copper film still formed faster than PS converted.
Top (a) and cross-sectional (b) SEM views of quasi-continuous Cu film. Cu film was grown on PS3 of 7-μm thickness and 80% to 85% porosity by displacement deposition for 1020 s.
EDX scan of the cross section of Cu/PS3. (a) SEM cross-sectional view, (b) EDX data obtained by scanning along the yellow line.
Therefore, to achieve a complete conversion of Si skeleton in copper, it is necessary to uniform the displacement rate along the pore channels. To meet the requirement, we increased the pore volume and improved the wettability of the surface of PS. The first change was connected with the use of macroporous silicon instead of mesoporous media. Macroporous Si has the same shape and order of pores as mesoporous materials, but the diameter of pores might be an order of magnitude greater
]. Widening of pore channels provides better and more rapid penetration of reagents into the pore channels because diffusion limitation attenuates. Figure
shows SEM images in cross-sectional (a) and top (b) views of macroporous Si (PS4) prepared for conversion into copper which represents an ordered array of parallel pore channels. Pillar-like pore walls have bases wider than their tops and ragged surface. The thickness of porous layer was limited to 2.5 to 3 μm. This limitation was caused not only by the time of anodic treatment as in the case of mesoporous silicon. SEM analysis of several macroporous PS4 samples grown for different time periods was performed. PS thickness increase resulted in gradual thinning of the tops of Si pillars. In that way, porous layer reaching the thickness of 3 μm began to dissolve. The gravimetrically measured porosity of PS4 was 60% to 65%. Pore density (number of pores per square centimeters) was found from the evaluation of SEM images and varied in the range of 2 to 2.5·
shows the distribution histogram of pore sizes of PS4. The common range of pore diameters is rather broad, but most part of the pores has a diameter of channel entrances varying in the range of 600 to 800 nm.
SEM cross-sectional (a) and top (b) views of PS4.
Pore size distribution histogram of PS4.
Wettability improvement was achieved by adding isopropanol (C3
OH) into the solution of copper salt. This alcohol significantly decreases the surface tension of water-based solutions
], providing better contact between liquid and solid surfaces
]. PS4 was left in the copper solution of improved wettability for 7200 s. Then, we observed the separation of the copper membrane from the Si substrate. The underlying Si had a surface without remains of porous layer. Figure
shows SEM images of the cross section (a), top side (b), and bottom side (c) of the separated membrane and related EDX point analysis which are considered in the next paragraph. The membrane represents a two-layered structure of 8-μm thickness. The top surface (Figure
b) was the sample/solution interface, while the bottom (Figure
c) was connected with the substrate. The top layer has a thickness of about 5 μm and represents a tightly packed array of parallel column-like agglomerates which are perpendicular to the substrate, i.e., columns grew along the pore direction of the original PS4. On the other hand, the bottom layer looks like a sponge of 3-μm thickness consisted of chains of small particles. Figure
presents the size distribution histograms of agglomerates and particles of the top and bottom surfaces of the membrane. Histograms were calculated from Figure
b,c. The diameters of upper agglomerates are an order of magnitude greater than those of bottom particles. At the spongy layer, the particles of 160- to 200-nm diameters dominate. The prevalent diameter range of the upper agglomerates is 2500 to 3500 nm, but elements of two times less in diameters (to 1,500 nm) were found. The density of agglomerates was about 9
, while the density of the bottom NPs was four orders of magnitude higher (9 to 16
Figure 8 SEM (a, b, c) and EDX point analyses (d, e, f) of the porous copper membrane. The porous copper membrane was formed by displacement deposition of copper on PS4 for 7200 s from the solution of improved wettability; porous copper membrane was analyzed in (more ...)
Size distribution histograms of agglomerates and particles. Histograms were calculated for (a) top and (b) bottom surfaces of the porous copper membrane.
To reveal the elemental composition of the membrane, EDX analysis of the cross section, top side, and bottom side were carried out (Figure
d,e,f). EDX scan of the cross section was attempted as well, but it was impossible to correctly focus the 1-μm electron beam on the non-flat surface of the agglomerates. To overcome doubts on the elemental composition, EDX analysis was performed in ten different points of the cross section, and each showed 97 to 99 at.% of Cu content. An example of point EDX in the cross section is presented in Figure
b,c confirms the copper nature of the obtained membrane. Overall, the membrane uniformly contains 95 to 99 at.% of copper with small amounts of oxygen and carbon. The maximum content of Si atoms was 0.1%, i.e., it might be declared that the obtained membrane represents the copper material. The gravimetrically determined porosity of the membrane was 60% to 65% in comparison with bulk copper.
Based on the results of SEM and EDX analyses, we propose the following phenomenological model of the formation of porous copper membrane. On the stage of full impregnation of PS with the solution, Cu NPs nucleate and grow on the surface of PS skeleton. As metal deposition was carried out simultaneously with dissolution of Si pillars, PS skeleton was converted into bottom spongy copper layer. The supposition might be proved by equality of the thickness of the original PS4 to that of the bottom copper layer (2.5 to 3 μm). In our opinion, new copper NPs grow and coalesce on the outer surface of the spongy copper layer. In that way, a layer of huge copper agglomerates is formed, whereas stresses on the Si/Cu membrane interface exceed over the interaction force between silicon and copper atoms when the copper membrane separates from the substrate as observed during the experiment. Detailed understanding of the porous copper membrane formation requires more careful in-depth study which is under the scope of the future paper.
The temperature variation of the Young modulus E||
measured for the porous copper membrane during the flexural vibration and that of E
measured during the extensional vibration are reported in Figure
. In both directions, E
increases at low temperatures, as usual in most solid samples. The measured values of the Young modulus (both E||
) are much smaller than the value of E
(110 to 128 GPa) for bulk polycrystalline copper, due to the high porosity and to the quasi-bidimensional feature of the membrane. It can also be noticed that the values of the Young modulus along the two directions differ by a factor of 300 at low temperature and 500 at high temperature, indicating a strong anisotropy of the sample, which is stiffer in the direction parallel to the pores, that is perpendicular to the plane of the membrane. Recently, a systematic experimental and theoretical investigation of the elastic constants and of the Young modulus of a block (approximately 10
) of polycrystalline copper containing elongated pores was reported
]. All crystallites had one crystallographic direction aligned along the  Cu axis and another two randomly oriented in the perpendicular plane. The pores were oriented along the  direction, and their diameters ranged between 15 and 380 μm. In such a system, the values of E||
strongly depend on the ratio of the axes of the ellipsoids associated to the pores. At low porosity (p
20%), for pores having a high ellipticity, E>E||
]. However, both values decrease with increasing porosity and virtually reach a null value for p
100%. However, while E||
decreases linearly with p
has a stronger dependence on porosity which leads to E||
] as in the case of the membrane investigated in the present paper. However, a quantitative comparison of the elastic modulus values between the membrane investigated in the present paper and the structure of
] is not possible because the typical dimensions of the samples and of the pores differ by various orders of magnitude. Figure
also shows a clear hysteresis between cooling and heating in both vibration modes, which is reproduced upon subsequent cycling (results not shown) and could be possibly due to the absorption and desorption of gases on the porous structure.
Temperature variation of the Young modulus. It was measured parallel to the pores or perpendicularly.
The measurements of Raman spectra with water-soluble CuTMpyP4 as an analyte were performed for both sides of the porous copper membrane. Since water did not wet the surface of the porous Cu film (the solution formed a ball-shaped drop on the surface), the CuTMpyP4 precipitated from an aqueous-alcohol solution in a 1:1 ratio by volume. Figure
shows Raman spectra of CuTMpyP4 deposited from a 10−6
M solution on the top (a) and bottom (b) sides of the porous copper membrane and on PS without copper coating for comparison (c). It can be seen that distinct vibrational bands are observed in all spectra which correspond to CuTMpyP4 features reported in the literature
]. The Raman intensity from the top side was three times of magnitude higher in comparison with that from the bottom side and from the porous surface of the PS sample. This observation reveals that the copper nanostructured surface fabricated on the top of PS by Cu displacement deposition exhibits some degree of SERS activity. In contrast, Cu porous layer from the bottom does not demonstrate enhancement of Raman signal. Comparing the 441.6-nm excited SERS spectrum of CuTMPyP4 (Figure
a) with ordinary Raman spectrum in solid (Figure
c), the close coincidence of the maxima wavenumbers and relative intensities of the bands can be observed. It means that no preferential orientation (geometry of binding
) exists for CuTMPyP4 molecules adsorbed on the Cu membrane. The structure of the top of copper film was similar to the Si nanopillar array covered with copper that demonstrated the SERS activity in the paper
]. So, the enhancement was likely to be caused by two reasons: (1) plasmon concentration on the tips of copper pillars and (2) the ‘hot spots’ in the copper pillar connections.
Figure 11 SERS spectra. (a) SERS spectrum of 10−6 M CuTMpyP4 adsorbed on the top side of porous copper membrane. (b, c) Raman spectra of 10−6 M CuTMpyP4 deposited from water and solution on bottom side of Cu membrane and porous silicon substrate, (more ...)