CARBON
8 8 ( 2 0 1 5 ) 7 0 –8 5
Available at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/carbon
Carbon periodic cellular architectures A. Szczurek a, A. Ortona b,*, L. Ferrari b, E. Rezaei b, G. Medjahdi a, V. Fierro a, D. Bychanok c, P. Kuzhir c, A. Celzard a,* a
Institut Jean Lamour, UMR CNRS – Universite´ de Lorraine n 7198, ENSTIB, 27 rue Philippe Se´guin, CS 60036, 88026 Epinal Cedex, France The iCIMSI Research Institute – University of Applied Sciences (SUPSI), Department of Technology and Innovation, Strada Cantonale, Galleria 2, CH 6928 Manno, Switzerland c Research Institute for Nuclear Problems of Belarusian State University, Bobruiskaya Street 11, Minsk 220030, Belarus b
A R T I C L E I N F O
A B S T R A C T
Article history:
The first carbon periodic cellular architectures derived from 3D printing, in the form of new
Received 23 December 2014
tetrakaidecahedra meshes, are reported and investigated in this paper. They were prepared
Accepted 23 February 2015
in hydrothermal conditions by a template method based on polymer periodic structures of
Available online 28 February 2015
the same geometry, and fabricated by a 3D printer using photocurable resin. Several formulations based on resorcinol–formaldehyde were tested, and the best ones were those using low concentrations of resorcinol at 150 C in a pressurised solution of nickel nitrate. After pyrolysis at 1000 C, catalytic graphitisation was demonstrated by TEM, XRD and Raman studies. The higher was the amount of nickel, the higher was the resultant graphitisation level. Mechanical tests were also carried out on such extremely lightweight periodic carbon structures, showing that these new materials present a much higher modulus than carbon foams of similar bulk densities. 2015 Elsevier Ltd. All rights reserved.
1.
Introduction
Highly porous, cellular, carbon materials can be prepared by different techniques, e.g. by foaming [1–8], leaching of sacrificial template particles [9], emulsion- or bubble-templating [10–17] of carbon precursors, among others, and hard-templating using polymer preforms [18–23]. In most cases, a polymer is used as the carbon source, and once the cellular structure of the material is stabilised, a pyrolysis step has to be carried out. The polymer can be pyrolysed directly or not, depending on its chemical nature. Thermoplastics indeed lose their shape when heated, therefore need to be stabilised beforehand, and most of them have a nearly zero carbon yield. In contrast, thermoset resins, especially phenolic ones, can be directly converted into carbon with a very satisfactory
yield, close to 50% [24]. When the polymer has a too low carbon yield, whether thermoset or not, an impregnation is required for getting enough material after pyrolysis. This impregnation allows coating the cellular structure of the polymer used as template, thereby stabilising its shape, as well as providing the necessary source of non-volatile carbon. As far as the present authors know, 3D printing has never been used so far for obtaining new carbon structures whereas it has been already shown to be an excellent method for preparing new, ordered, ceramics [25–27]. Only one rapid prototyping method has been used so far for preparing 3D scaffolds based on the extrusion of a paste composed of graphite with a polymer binder [28], which is a completely different method than that reported here. Additive fabrication is indeed not directly applicable to carbon, as pure carbon can
* Corresponding authors. E-mail addresses:
[email protected] (A. Ortona),
[email protected] (A. Celzard). http://dx.doi.org/10.1016/j.carbon.2015.02.069 0008-6223/ 2015 Elsevier Ltd. All rights reserved.
CARBON
8 8 (2 0 15 ) 7 0–85
71
Fig. 1 – (a) Tetrakaidecahedron cell parameters: strut diameter Ds, edge length L, height H, cell width D, and strut inclination h = 45 for a Kelvin cell. (b) Schematisation of the 4 · 4 · 4 cells numerical model. obviously neither been molten, crosslinked nor polymerised after layer-by-layer deposition by a 3D printer. Therefore, either the 3D structure has to be pyrolysed or, better, used as a template for another carbon source of higher yield. The polymer used in the present work was indeed a photocurable resin which almost completely evolved into gases by direct pyrolysis. In order to prepare the carbon replica of printed 3D structures, a template synthesis using a hydrothermal route has been used. We show here how such a process was optimised, especially through the use of catalytic graphitisation, and we present a few properties of these new materials. Tetrakaidecahedra meshes have been chosen for the present studies, but any other complex structure, periodic or nor, might be treated in the same conditions.
2.
Experimental
2.1.
Materials
2.1.1.
Polymer tetrakaidecahedra periodic cellular structures
The lattice structures were generated by a MATLAB code (MathWorks, Natick, MA, USA). Each model consisted of a 3D distribution of points (nodes) knowingly connected. This tool can generate any cell type with desired cell sizes depending on the nodes distribution and their connection. Each connection was modelled by two vertices and a line between them (strut axis). Different parameters such as strut’s diameter and elongation can be chosen to design a numerical model. 3D-printed polymer templates were designed by computer-aid design (UGS NX 8.5, Siemens, Germany). The numerical model consists of a 4 · 4 · 4 array of tetrakaidecahedra as represented in Fig. 1. The tetrakaidecahedron structure is especially interesting as it is widely accepted to represent the random foam structure while it is parametric. On the other side, it is an engineered structure, therefore it can be placed as the border between random foams and designed lattices.
1
Templates with D = H = 3, 4, 5 and 7.5 mm were 3D-printed (Eden 260 V 3D, Objet, Rehovot, Israel) with photosensitive resins (Fullcure 705 and 850 VeroGray, Object Ltd.), which were composed of acrylics, urethanes and epoxies. Resins were compounded with a photo-initiator that triggered the polymerisation under UV light. The 3D printing was performed with the maximum resolutions of 16 lm and 42 lm for vertical and planar directions, respectively. Fig. 2 shows the corresponding periodic tetrakaidecahedra architectures as directly obtained from the 3D printer. The cell size and the struts thickness were modified according to the values gathered in Table 1. In the present work, only carbon materials based on mesh A (see again Table 1) are described, but experiments showed that the same kind of structures can be easily prepared in the same conditions from the three others structures reported in Table 1 (see below). Not only changing the strut thickness at constant cell size and geometry (and therefore changing the porosity, all other things being equal) was possible, see Fig. 3(a), but totally different periodic structures were also available for preparing highly ordered and porous carbons, see Fig. 3(b). However, as the present paper is a first report of carbon architectures prepared by 3D printing, only carbons prepared from lattice A (see again Table 1) were described here.
2.1.2.
Carbon tetrakaidecahedra periodic cellular structures
Different tests were carried out for determining the method leading to the most regular and flawless carbon structures. The first trial consisted in doing a simple and direct pyrolysis of the polymer mesh in a flow of very pure nitrogen up to 1000 C at a heating rate of 1 C min1. As expected from the nature of the photocurable resin, principally based on various acrylates and on acrylic monomers and oligomers1, the structure was completely lost and only a very low amount of carbon, around 1% of the initial mass, remained.
Fullcure 850 Verogray, Safety data sheet, version 2. Object Geometries Ltd., 2008.
72
CARBON
8 8 ( 2 0 1 5 ) 7 0 –8 5
10 mm
Fig. 2 – Numerical domains of Fig. 1 after 3D-printing, from which new carbon structures were prepared by templating in hydrothermal conditions. A–D are the names of the polymer meshes, whose characteristics are listed in Table 1.
Table 1 – Cell parameters of the samples shown in Fig. 2, and calculated according to [29]. Mesh parameters
A
B
C
D
H = 4L sin h (mm) p D = 2L cos h + 2 b (mm) L (mm) Ds (mm) b/L (mm) h () Porosity (%)
5 5 1.77 0.45 1 45 94
7.5 7.5 2.65 0.5 1 45 97
4 4 1.41 0.44 1 45 93
3 3 1.06 0.33 1 45 93
Attempts were therefore made to coat the polymer lattices with a phenolic resin in hydrothermal conditions. For that purpose, various amounts of resorcinol were dissolved in hydroalcoholic solutions in the presence of formaldehyde or furfuryl alcohol and acidic catalyst. The following procedure
was inspired by a paper dealing with low-temperature hydrothermal treatment, by which surfactant micelles were used as soft templates for preparing mesoporous ordered carbons [30]. The polymer meshes were therefore soaked in the aforementioned, resorcinol-based, solutions and left for impregnation in an autoclave. The latter was filled at 1/3 of its capacity and heated at 50 C for several hours, then at 80 C for some additional hours. Four formulations are detailed in Table 2. After hydrothermal treatment, the impregnated meshes were washed, dried, and pyrolysed at 1000 C as described above. In these conditions, carbon structures were obtained, but the remaining solution after hydrothermal treatment was visibly far from having allowed the complete deposition of the resin on the polymer template. In other words, most of the solution of resin was wasted. Besides, after pyrolysis, huge and uneven volume shrinkage, about 80%, was observed that produced some significant deformations of the initial geometry. Moreover, SEM studies
Fig. 3 – Various 3D-printed architectures with which other periodic carbon structures can be prepared using the same process: (a) tetrakaidecahedra meshes of identical cell size but different strut thickness; (b) meshes having different geometries: (i) octet, (ii) cubic, (iii) hexagonal, and (iv) foam [27] lattices. (A colour version of this figure can be viewed online.)
CARBON
73
8 8 (2 0 15 ) 7 0–85
Table 2 – Formerly tested formulations for the hydrothermal treatment of polymer meshes. Formulation #
1
2
3
4
Resorcinol (g) Formaldehyde (g of 37 wt.% solution) Furfuryl alcohol HCl (g of 37 wt.% solution) Ethanol (mL) Water (mL) Hydrothermal treatment
3.3 5.0 0 0.4 20 20 8 h at 50 C 48 h at 80 C
3.3 7.0 0 0.4 20 20 3 h at 50 C 48 h at 80 C
4.95 7.5 0 0.4 20 20 1.5 h at 50 C 48 h at 80 C
5 0 8 0.7 40 0 4 h at 50 C 48 h at 80 C
revealed that the as-formed carbon was very porous (see below). It was thus decided to proceed using a hydrothermal treatment at higher temperature. Doing so, the resin completely deposited on the polymer template, as the remaining solution was indeed transparent. Advantageously, higher hydrothermal temperature also made the use of acid catalyst unnecessary and, according to Xie et al. [31], is supposed to allow a more favourable subsequent catalytic graphitisation. The process was as follows. Samples of polymer mesh of known mass and dimensions were first soaked in absolute ethanol for 15 min for removing impurities and rests of monomer or additives. Next, the specimens were dried, weighed again, and placed in a glass flask filled with an aqueous solution of a resorcinol–formaldehyde–nickel nitrate mixture. The presence of nickel was indeed thought to catalyse the conversion of part of the disordered carbon from resorcinol into a more ordered, graphite-like and therefore less brittle, carbon. It is indeed known that, during pyrolysis, metal salts mostly (but not only) from the Fe, Co, Ni series are reduced by carbon into metal nanoparticles, and that the latter induce a catalytic graphitisation at moderate temperature [32–38]. The corresponding formulations were based on a mixture of 0.475 g of resorcinol and 0.7 g of a 37 wt.% aqueous solution formaldehyde in 30 mL of water, to which 0, 2, 4, 8 or 22 g of Ni(NO3)2, 6H2O were dissolved. The corresponding formulations were called A0Ni, A0.5Ni, A1Ni, A2Ni and A5Ni, respectively. The glass flask was introduced in a stainless steel autoclave installed in an oven heated at 150 C for 24 h. After this time, the autoclave was taken from the oven and let to cool down to room temperature. The samples were recovered, washed with water, dried at 105 C for 4 h, and pyrolysed at 1000 C with a heating rate of 1 C min1. The final temperature was held for 2 h before cooling down to room temperature. The whole process of heating and cooling was carried out in a flow of very pure nitrogen of 100 mL min1. The as-obtained periodic carbon structures were washed in a 37% aqueous solution of HCl for removing nickel. The washing step took three days during which the acid was replaced daily by fresh one. Finally, the samples were washed one day with water for removing the traces of chloride ions, and dried.
2.2.
Characterisation
Complete characterisation was not performed for all samples, but only for carbon architectures derived from the last formulations, i.e. for the A0Ni to A5Ni series based on A meshes.
Only SEM was carried out for the samples derived from the formulations given in Table 2 and using the same A meshes, suggesting the need of formulations less concentrated in resorcinol but deposited at higher temperature (150 C against 50–80 C in the former trials). As for the AxNi series, samples’ morphology was observed at the macroscopic scale and by SEM, microstructure was investigated through XRD, TEM and Raman studies, and mechanical tests were carried out. These characterisation methods were chosen among others because they are fully relevant to these new materials, which present extremely wide cells as they are only based on a periodic 3D lattice of struts. For this reason, thermal or porosity studies were deliberately discarded. The catalytic graphitisation was especially investigated, as it led to more homogeneous and isotropic carbon structures, as shown below.
2.2.1.
Electron microscopy
Scanning Electron Microscopy (SEM) observations were carried out with either FEI Quanta 600 or Hitachi TM3000 Tabletop microscopes. The carbon meshes were installed on a carbon-coated sample holder for ensuring a good electrical contact with the latter. Prior to investigation, the samples were covered with carbon in a metallisation system under vacuum. Two modes of observations, using either secondary electrons (SE) or backscattered electrons (BSE) detectors, were used with FEI and Hitachi microscopes, respectively. Energydispersive X-ray spectroscopy (EDX) mapping was performed for observing the residual nickel distribution in the carbon structure obtained from the most Ni-rich formulation (A5Ni).
Table 3 – Volume shrinkage and carbon yield of carbon meshes, depending on the formulation. Formulation
Volume shrinkage (%)
Carbon yield (wt.%)
#1 #2 #3 #4 A0Ni A0.5Ni A1Ni A2Ni A5Ni
79.7 n.d.a n.d.a n.d.a 84.4 78.2 75.3 73.1 72.3
17.8 16.5 16.7 22.3 9.6 15.2 15.6 18.0 19.7
a Not determined because of irregular shapes of the final carbon structure, but close to 80%.
74
CARBON
8 8 ( 2 0 1 5 ) 7 0 –8 5
The effect of catalytic graphitisation with Ni was observed using Transmission Electron Microscopy (TEM), using a FEI CM200 microscope. In this case, two samples with different contents of nickel (A2Ni and A5Ni) were ground in ethanol, and a few drops of the resultant suspension were put on a sample holder grid.
2.2.2.
X-ray diffraction
X-ray diffraction (XRD) was carried out with a Panalytical X’Pert Pro diffractometer. The latter was used in a Bragg– Brentano configuration in reflection, and was equipped with a Cu (Ka radiation) anticathode and a high-speed multichannel X’Celerator detector. The carbon meshes from the AxNi series were ground, and the resultant powders were placed
on a flat sample holder. The latter was installed on a rotating spinner for allowing the highest number of grain to be in diffraction position. The interplanar spacing d002 was calculating from the wellknown Bragg’s law, whereas the sizes of the coherent domains along a and c axes, La and Lc, respectively, were calculated from the 1 1 0 et 0 0 l reflections, respectively, using the Scherrer equation [39]: La ðnmÞ ¼
k b cos h
ð1Þ
where k (nm) is the radiation wavelength (here 0.15418 nm), h (rad) is the reflection position angle, and b is the integral width (i.e., the ratio of peak area to maximal intensity) of
Fig. 4 – Examples of periodic carbon structures derived: (a) and (b) from mesh A using formulation #1; (c) and (d) from mesh A using formulation A0.5Ni; (e) from meshes D, C, A and B (from left to right, respectively) using the formulation A5Ni. (A colour version of this figure can be viewed online.)
CARBON
8 8 (2 0 15 ) 7 0–85
75
Fig. 5 – SEM pictures of: (a)–(f) periodic carbon structures derived from formulation #1 at increasing magnifications; (g) and (h) hollow struts in carbon structures derived from formulations #1 and #3, respectively.
Fig. 6 – SEM pictures of a periodic carbon structure prepared from the A5Ni formulation at different magnifications: the struts are solid.
the relevant peak in 2h units. Eq. (1) was preferred to the other formula using the FWHM of the bands and the corresponding multiplication factor K = 1.85 for La and 0.9 for Lc, respectively, as it avoids any problem related to peak asymmetry.
2.2.3.
Raman spectroscopy
Raman spectra were obtained with a Horiba XploRa Raman spectrometer using small pieces of carbon periodic structures from the AxNi series without further sample preparation. The
spectra were collected under a microscope using a 50· objective, which was the most adapted to rough and opaque samples. The Raman-scattered light was dispersed by a holographic grating with 1200 lines/mm and detected by a CCD camera. A laser of wavelength 532 nm, filtered so that the incident power was very low, i.e. 1.82 mW, was used. The reproducibility of the spectra was checked in different zones and after very different acquisition times as well as with even lower laser powers, thereby proving the homogeneity of the
76
CARBON
8 8 ( 2 0 1 5 ) 7 0 –8 5
materials and the absence of local heating and graphitization effects [40]. Each spectrum was obtained by accumulation of 30 spectra recorded from 500 to 3500 cm1 over 10 s, and was fitted by using three or more mixed Gaussian–Lorentzian profiles in the first-order region for the bands D4, D1, D3, G and D2 appearing at increasingly high wavenumbers [41]. The inplane ‘‘crystallite’’ size La was calculated from the most recent and general, following, equation [42]: 1 560 ID La ðnmÞ ¼ 4 ð2Þ El IG
deformation and load were continuously recorded. Based on the obtained stress – strain curves, Young’s modulus and maximum compressive strength were estimated. As already detailed elsewhere [46], Young’s modulus was defined as the slope of the linear, initial, part of the curve presenting the steepest slope, whereas the compressive strength was defined as the maximum height of the stress – strain curve.
where El (eV) is the energy of the laser line (here 2.33 eV at 532 nm), and ID and IG are the intensities of D1 and G bands, respectively.
The formulation was found to have an effect both on volume shrinkage and on carbon yield, as evidenced in Table 3. Increasingly high amounts of nickel not only improved the yield, as already evidenced for carbon from phenol–formaldehyde resin catalytically graphitised by iron [47], but also allowed slightly lower shrinkage and virtually no deformation of the periodic mesh during pyrolysis. Higher concentrations of resorcinol, formaldehyde or acid in the formulations #1 to #4 led to higher carbon yields, but the presence of nickel in the AxNi series improved the carbon yield despite the correspondingly lower amount of resorcinol that was used. The main deformation observed with formulations #1 to #4 was an anisotropic, inhomogeneous, shrinkage in the direction along which the polymer mesh was grown in the 3D printer. A close examination of the struts (see SEM images below) indeed clearly revealed the successive layers of deposited photocurable resin, so that the direction of growth was therefore not strictly equivalent to the other two orthogonal directions. As a consequence, the carbon structures prepared from formulations #1 to #4 were anisotropic, see
2.2.4.
Mechanical tests
Samples of periodic carbon structures from the AxNi series were investigated by quasi-static compression using an Instron 5944 universal testing machine equipped with a 2 kN head. The samples were not submitted as such to compression because of expected problems related to boundary conditions already discussed elsewhere [[43] and refs. therein]. The mechanical behaviour of brittle foams was indeed shown to be strongly affected by the bonding – or not – of the specimen ends to the loading platens. The bonding of the specimen ends to the stiff loading platens leads to a more homogeneous distribution of the load through the section and, consequently, to a higher value of compressive strength [44,45]. Therefore, the opposite faces of the samples were first glued to PMMA plates. The compression was carried out at a constant load rate of 2.0 mm min1 during which
3.
Results and discussion
3.1.
Morphology of carbon periodic structures
Fig. 7 – (a)–(c): SEM pictures of a periodic carbon structure prepared from the A5Ni formulation before HCl washing, showing Ni nanoparticles all over the surface, and (d)–(f): EDX mapping of image (c) for elements C, Ni, and Ni and C together, respectively. (A colour version of this figure can be viewed online.)
CARBON
8 8 (2 0 15 ) 7 0–85
Fig. 4(a and b). Carbon meshes from the AxNi series were much more isotropic, though never completely, but were homogeneous, see Fig. 4(c and d). Fig. 4(e) shows that the AxNi formulations could be successfully used for preparing periodic carbon structures derived from B, C and D polymer templates. Their detailed investigation, with correlations between properties and lattice parameter, will be the object of future works. The carbon microstructure is now described below for the different materials reported here.
3.2.
Microstructure of carbon periodic structures
SEM pictures of periodic carbon structures obtained from the aforementioned formulations are presented in Fig. 5. It was observed that formulations #1 to #4 led to highly porous
77
carbon struts presenting a very rough surface. This finding is related to the method by which the polymer meshes used as templates were prepared, i.e., by deposition of successive layers of photocurable resin. The initial struts therefore had a smooth but strongly corrugated surface, and this effect was dramatically amplified by resorcinol–formaldehyde resin deposition and subsequent pyrolysis. SEM pictures of carbon structures derived from formulations #1 to #4 also show that the struts are hollow, due to the significant difference of carbon yield between photocurable and resorcinol–formaldehyde resin: 1 wt.% and 45–55 wt.% at 1000 C, depending on the operating conditions [24], respectively. In contrast, carbon meshes from the AxNi series presented solid struts, see Fig. 6. For this reason, such materials were the only ones whose mechanical properties were measured.
Fig. 8 – TEM pictures at different magnifications of periodic carbon structures prepared from: (a)–(c) A2Ni and (d)–(f) A5Ni formulations. (g) Encapsulated nickel nanoparticle in A2Ni carbon.
78
CARBON
8 8 ( 2 0 1 5 ) 7 0 –8 5
Fig. 9 – TEM images of: (a) non-graphitised and (c) graphitised carbon phase in A5Ni, (b) and (d) being the corresponding FFT diffractograms, respectively. (e) Line-scanned intensity profile of the square area in (c), showing an interplanar spacing close to 0.34 nm. (A colour version of this figure can be viewed online.) Before being washed by HCl, the nickel was highly dispersed, as seen in Fig. 7, and therefore the catalytic conversion of carbon into a more graphite-like structure occurred, as proved by
TEM pictures given in Fig. 8. Indeed, and as expected, the presence of Ni induced the occurrence of well-organised, lamellar, domains dispersed in a matrix made of much more
CARBON
disordered carbon. These domains appeared in the form of hollow capsules, suggesting that the carbon graphitised at the contact of nickel particles subsequently removed by the acid treatment. Higher amounts of nickel in the formulation produced more of such capsules. In very rare occasions, the dissolution of nickel was incomplete and one metal particle could be remain in a graphite-like carbon shell, see Fig. 8(g). The carbon nanotexture is now described below, further confirming the partial graphitisation, depending on the initial amount of nickel.
3.3.
79
8 8 (2 0 15 ) 7 0–85
The XRD pattern of the periodic carbon structure derived from the formulation A0Ni is given in Fig. 10(a). This pattern was subtracted from the other spectra of the series, and the corresponding results are given in Fig. 10(b). It can be seen that the pattern of A0.5Ni is very similar to that of the reference A0Ni. In contrast, for the other samples, the intensity of the (0 0 2) band increased with the amount of nickel, and the overall diffraction pattern tended towards that of graphite. However, the corresponding interplanar spacing d002 was around 0.341 nm, i.e. close to that of a coke, compared
Carbon nanotexture
a
2 x 10
Intensity (a.u.)
More TEM pictures of the periodic carbon structure from the A5Ni formulation are given in Fig. 9, showing non-graphitised and graphitised carbon phases, as well as the corresponding fast Fourier transform (FFT) diffractograms of the digitised images. Fig. 9(a–d) clearly shows that two very different kinds of carbon texture coexist in the same material, one being highly disordered without any visible orientation, the other one having a high degree of periodicity and orientation over domains of several nm. Examination of the pixel intensity of a cross-section of the square area in Fig. 9(c) evidenced an interplanar spacing close to 0.34 nm, see Fig. 9(e). The same was found from XRD studies, as detailed below.
1.5 x 10
Table 4 – Graphitisation degree, g, interplanar spacing, d002, and in-plane and out-of-plane crystallite sizes, La and Lc, respectively. Formulation #
A0.5Ni
A1Ni
A2Ni
A5Ni
g (%) d002 (nm)a La (nm)b Lc (nm)b La (nm)c
– – – – 16.9
22 0.342 18.5 6.7 17.6
22 0.342 20.6 6.5 17.7
33 0.341 18.8 7.0 17.9
a b c
Estimated from Eq. (3). Estimated from XRD patterns through Eq. (1). Estimated from Raman spectra through Eq. (2).
4
4
4
1 x 10
3
5 x 10
0 0
20
40
60
80
100
120
2θ (°)
b A0.5Ni A1Ni A2Ni A5Ni
Fig. 10 – XRD patterns of periodic carbon structures from the AxNi series: (a) reference A0Ni; (b) other samples after subtraction of (a). (A colour version of this figure can be viewed online.)
80
CARBON
8 8 ( 2 0 1 5 ) 7 0 –8 5
to the value of 0.335 nm expected for graphite. Nevertheless, the d100 values, (0.207 nm), were close to that of graphite (0.213 nm). For a typical coke, d002 is close to 0.340 nm, but an h k band can be seen. In the present carbon meshes, the h k 0 and h k l (1 0 0 and 1 0 1) bands are well formed. This is also the case for the 1 1 0 and 1 1 2 reflections which are small but clearly visible in the material derived from the A5Ni formulation, and which prevail in the model of Drits and Tchoubard for estimating the graphitisation degree [48].
The graphitisation degree, g, was calculated from the formula of Maire and Mering, which reads [49]: g¼
0:344 d002 0:344 0:335
ð3Þ
The corresponding values of g are listed in Table 4, as well as the values of d002 and the sizes of the coherent domains La and Lc. It can be seen that, as expected, both La and Lc increased, on average, with the amount of nickel in the initial formulation. The value of d002 decreased accordingly, in
Fig. 11 – Raman spectra of periodic carbon structures from the AxNi series. In (a), the spectra are shifted along intensity axis for clarity, whereas (b) is the detail of the 1st order region showing D and G bands. (A colour version of this figure can be viewed online.)
81
8 8 (2 0 15 ) 7 0–85
agreement with a higher graphitisation degree. Although the latter is just a rough estimation, the observed changes agree with larger coherent domains. The increasing graphitisation degree from A0Ni to A5Ni was confirmed by Raman spectroscopy. Raman spectra of periodic carbon structures from the AxNi series are presented in Fig. 11. The spectrum of the sample from the A0.5Ni formulation is not shown because merged with that of A0Ni. In Fig. 11, the intensity of the bands were normalised with regard to the highest one, corresponding to the D1-band at 1361 cm1. In the first-order region (1000–1800 cm1), the spectra displayed a D1 band at a remarkably stable position of 1361 cm1. The G band was observed at 1612 cm1 for all carbons, except for A5Ni, for which the maximum was located at 1590 cm1. In addition to these two main bands, components at around 1190 and 1500 cm1 were also present. The shoulder at 1190 cm1, called D4 and mainly seen for low nickel amounts, is known to be characteristic of very poorly organised carbons, whereas the component at 1500 cm1, called D3, is also present as a very wide band in poorly crystallised carbons [50]. A shoulder on the G band around 1620 cm1, called D2, was only observed in A5Ni. The signification of this band is not yet well understood, but is known to exist only if the D1 band is also present [50]. In the second-order region (2200– 3400 cm1), three broad to very broad bands are seen, around 2700 cm1 (called S1), 2900 cm1 (called S2) and 3200 cm1. When the amount of nickel increased in the initial formulation, the G band clearly increased until becoming almost as high as the D band in A5Ni. The ratio ID/IG, used as a graphitisation indicator since the 70’s [51–54], is presented in Fig. 12 as function of the initial nickel amount. The drop of ID/IG confirms the corresponding increase of the structural order degree suggested by TEM and XRD. In the meantime, the D4 shoulder progressively vanished until complete disappearance in the A5Ni carbon. In contrast, the contribution of the D3 band, attributed to defects outside the plane of aromatic layers such as tetrahedral carbons [55], remained important. The bands in the second-order region became more intense and more individualised when the nickel content increased, especially the S1 band in A5Ni. The latter, sometimes also called G 0 band, has been attributed to the formation of fewlayer structured graphitic carbon [56,57]. Such features were already reported for saccharose-based chars, known to be non graphitisable as they remain turbostratic even at very high temperature [41]. The ‘‘catalytic graphitisation’’ reported here is therefore not a true graphitisation, but an increasing ordering never reaching the usual triperiodic structure of graphite. This is further proved by the presence of D2 band in A5Ni, which indicates that a significant amount of defects remained. This bands is indeed supposed to vanish when the graphitisation becomes complete, which is obviously far from being the case here. From the ID/IG data shown in Fig. 12, the corresponding values of La were calculated from Eq. (2) and reported in Table 4. Such values are close, though never identical, to those obtained from XRD through Eq. (1). However, the agreement can be considered as quite good with respect to the huge discrepancies already reported in the literature, and which have been assigned to various factors, such as: (i) different scale of
1.14 1.13 1.12
ID/IG
CARBON
1.11 1.1 1.09 1.08 1.07 0
5
10
15
20
25
Amount of Ni nitrate in the formulation (g)
Fig. 12 – Ratio of D1 to G band intensities as a function of the amount of nickel nitrate in the formulation. (A colour version of this figure can be viewed online.)
investigation of XRD (bulk) and Raman (micrometre-scale); (ii) artificial widening of XRD peaks whose incomplete correction may lead to overestimated La, or averaging over the bulk sample, thereby biasing La to lower values; (iii) attribution of different kinds of defects to the D1 band alone, whose decrease of intensity cannot be only ascribed to the increase of La [41]. All in all, it can be concluded that, whatever the method by which La is measured, obtaining absolute, accurate, values is never possible [41,58,59].
3.4.
Mechanical properties
Fig. 13(a) shows the way a periodic carbon structure collapsed when submitted to compression. The corresponding stress–strain curve is given in Fig. 13(b), in which modulus and compressive strength are defined. The compressive strength was the highest a few seconds after the beginning of the test, as some debris were ejected progressively from the carbon lattice submitted to compression, as evidenced in Fig. 13(a). Therefore the cross-sectional surface area of the sample decreased during the test, hence the lower strength. However, a progressive failure of the material at the microscopic level might also account for the observed decrease of strength. Fig. 13(b) also shows that, after the sudden collapse of successive strut layers, the contact was briefly lost between platens and sample whose height abruptly decreased, as the stress recurrently went back to zero. The data of Young’s modulus and compressive strength of the whole AxNi series are gathered in Fig. 14. Fig. 14(a) clearly shows that both modulus and strength increased with bulk density, as expected, except for the material containing no nickel, for which the mechanical properties were higher. The corresponding A0Ni carbon was indeed stiffer, due to the absence of graphite-like domains, and more nickel led to materials having a gradually lower modulus, as seen in Fig. 14(b). Indeed, like all phenolic resins, resorcinol– formaldehyde resin leads to a very stiff, brittle and shiny, glasslike carbon. Values as high 20.4–20.8 GPa have indeed
82
CARBON
8 8 ( 2 0 1 5 ) 7 0 –8 5
been reported for the modulus of glassy carbon [60]. Graphitisation was therefore expected to lead to a less stiff material, and this is indeed what was observed, as the modulus of graphite is typically three times lower than that of glassy carbon [60]. The same trend was observed for compressive strength, although the highest strength was observed for the material containing the lowest amount of nickel. However, the latter was also the one having the highest density, hence the result. Generally speaking, the values of modulus given in Fig. 14 are much higher than those reported for carbon foams of similar density, whereas the values of strength are lower [[23,46,61,62] and refs. therein]. This finding may be interpreted in terms of straight, non-porous, and
thick struts in periodic carbon meshes, therefore much stiffer than those found in typical carbon foams, for which the struts are far much thinner and with a very slim central part. As for the lower strength, it is related to the very big equivalent cell size of periodic carbon structures, as it is known that, in foams of constant bulk density, bigger cells lead to lower compressive strengths [63]. Before concluding, it can be said that these new 3D periodic carbon structures are homogeneous, repeatable and their characteristics can be strictly controlled through the preliminary 3D-printing process. The latter is the only one we know allowing to prepare a structure based on tetrakaidecahedra or other perfectly defined lattices, from
0.035
compressive strength 0.03
Stress (MPa)
0.025 0.02 0.015 0.01
modulus 0.005 0 0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Strain (%) Fig. 13 – (a) Snapshots of the compression of the periodic carbon structure derived from the A1Ni formulation (each photo is separated by 5 s), and (b) corresponding stress–strain curve. (A colour version of this figure can be viewed online.)
CARBON
Modulus w/o Ni Modulus w/ Ni
Strength w/o Ni Strength w/ Ni
50
0.06
40
0.04 20 0.03 10
0 0.045
0.02
0.05
0.055
0.06
0.01 0.065
3
Density (g/cm )
0.07 0.06 0.05
30 0.04 20 0.03 10
Strength (MPa)
0.05 30
Strength (MPa)
Modulus (MPa)
40
Strength
Modulus 0.07
Modulus (MPa)
50
83
8 8 (2 0 15 ) 7 0–85
0.02 0.01
0 0
5
10
15
20
25
Amount of Ni nitrate in the formulation (g)
Fig. 14 – Young’s modulus and compressive strength of periodic carbon structures from the AxNi series, plotted as a function of: (a) bulk density; (b) amount of nickel nitrate in the formulation. The lines are just a guide for the eye. (A colour version of this figure can be viewed online.)
which rigorous studies will be carried out in the near future. For example, we are now building a device for measuring the electromagnetic properties of such materials in the terahertz range (200 GHz–3 THz), as we expect that they behave as photonic crystals in this range of wavelengths. The same will also be performed for other geometries of 3D carbon meshes. Tetrakaidecahedron is, however, of special interest, as it mimics the random structure of reticulated foam structures whereas still allowing rigorous calculations. These new carbon materials can also be intermediates of 3D ordered SiC architectures, which can be used as heat exchangers [25–27].
4.
Conclusion
In the present paper, the first periodic, tetrakaidecahedra, carbon structures prepared by templating of 3D polymer meshes made by rapid prototyping were presented and characterised. This method was proved to be suitable for obtaining any kind of morphology, whether complex or highly ordered, or both, and opens the route to the preparation of new metamaterials based on carbon. For getting 3D carbon lattices, an optimisation was required as the polymer periodic meshes themselves had an almost zero carbon yield. Impregnation was therefore carried out in hydrothermal conditions using resorcinol–formaldehyde resin, so that acceptable carbon yields close to 15% were obtained, along with volume shrinkages around 75%. The best results were obtained when nickel nitrate was used as a graphitisation catalyst, allowing obtaining much almost isotropic shrinkage, higher carbon yield and, more important, non-hollow struts. Increasingly higher amounts of nickel were found to lead to more graphitized carbon 3D architectures, having lower interplanar spacing and bigger crystallite sizes, on average. The new periodic carbon structures presented here exhibited
very high modulus with respect to their extremely low bulk density. More detailed studies of such kind of 3D carbon structures will be carried out in the near future.
Acknowledgements The authors gratefully acknowledge the financial support of the CPER 2007–2013 ‘‘Structuration du Poˆle de Compe´titivite´ Fibres Grand’Est’’ (Competitiveness Fibre Cluster), through local (Conseil Ge´ne´ral des Vosges), regional (Re´gion Lorraine), national (DRRT and FNADT) and European (FEDER) funds. This research was also partially supported by FP7PEOPLE-2013-IRSES-610875 NAmiceMC and Belarus-CNRS project BRFFI F13F-004 (Belorussian and French teams).
R E F E R E N C E S
[1] Chen C, Kennel EB, Stiller AH, Stansberry PG, Zondlo JW. Carbon foam derived from various precursors. Carbon 2006;44(8):1535–43. [2] Gallego NC, Klett JW. Carbon foams for thermal management. Carbon 2003;41(7):1461–6. [3] Gan L, Liu M, Wang X, Tian C. Synthesis of porous carbon foams blowing by pentane. J Tongji Univ 2008;36(11):1552–5. [4] Tondi G, Fierro V, Pizzi A, Celzard A. Tannin – based carbon foams. Carbon 2009;47(6):1480–92. [5] Li X, Basso MC, Braghiroli FL, Fierro V, Pizzi A, Celzard A. Tailoring the structure of reticulated vitreous carbon foams. Carbon 2012;50(5):2026–36. [6] Prabhakaran K, Singh PK, Gokhale NM, Sharma SC. Processing of sucrose to low density carbon foams. J Mater Sci 2007;42(11):3894–900. [7] Narasimman R, Prabhakaran K. Preparation of low density carbon foams by foaming molten sucrose using an aluminium nitrate blowing agent. Carbon 2012;50(5):1999–2009.
84
CARBON
8 8 ( 2 0 1 5 ) 7 0 –8 5
[8] Jana, Fierro V, Pizzi A, Celzard A. Biomass-derived, thermally conducting, carbon foams for seasonal thermal storage. Biomass Bioenergy 2014;67:312–8. [9] Seo J, Park H, Shin K, Baeck SH, Rhym Y, Shim SE. Ligninderived macroporous carbon foams prepared by using poly(methyl methacrylate) particles as the template. Carbon 2014;76:357–67. [10] Wang D, Smith NL, Budd PM. Polymerization and carbonization of high internal phase emulsions. Polymer Int 2005;54(2):297–303. [11] Cohen N, Silverstein MS. Synthesis of emulsion-templated porous poly acrylonitrile and its pyrolysis to porous carbon monoliths. Polymer 2011;52(2):282–7. [12] Gross AF, Nowak AP. Hierarchical carbon foams with independently tunable mesopore and macropore size distributions. Langmuir 2010;26(13):11378–83. [13] Liu M, Gan L, Zhao F, Xu H, Fan X, Tian C, et al. Carbon foams prepared by an oil-in-water emulsion method. Carbon 2007;45(14):2710–24. [14] Thongprachan N, Yamamoto T, Chaichanawong J, Ohmori T, Endo A. Preparation of macroporous carbon foam using emulsion templating method. Adsorption 2011;17(1):205–10. [15] Szczurek A, Fierro V, Pizzi A, Celzard A. Mayonnaise, whipped cream and meringue, a new carbon cuisine. Carbon 2013;58(1):245–8. [16] Szczurek A, Fierro V, Pizzi A, Celzard A. Emulsion-templated porous carbon monoliths derived from tannins. Carbon 2014;74:352–62. [17] Szczurek A, Fierro V, Pizzi A, Stauber M, Celzard A. Carbon meringues derived from flavonoid tannins. Carbon 2013;65:214–27. [18] Inagaki M, Morishita T, Kuno A, Kito T, Hirano M, Suwa T, et al. Carbon foams prepared from polyimide using urethane foam template. Carbon 2004;42(3):497–502. [19] Chen Y, Chen BZ, Shi XC, Xu H, Hu YJ, Yuan Y, et al. Preparation of pitch-based carbon foam using polyurethane foam template. Carbon 2007;45(10):2132–4. [20] Alvarez S, Fuertes AB. Synthesis of macro/mesoporous silica and carbon monoliths by using a commercial polyurethane foam as sacrificial template. Mater Lett 2007;61(11– 12):2378–81. [21] Manocha SM, Kalpesh P, Manocha LM. Development of carbon foam from phenolic resin via template route. Indian J Eng Mater Sci 2010;17(5):338–42. [22] Yadav A, Kumar R, Bhatia G, Verma GL. Development of mesophase pitch derived high thermal conductivity graphite foam using a template method. Carbon 2011;49(11):3622–30. [23] Jana P, Fierro V, Celzard A. Ultralow cost reticulated carbon foams from household cleaning pad wastes. Carbon 2013;62:517–20. [24] Moreno AH, Arenillas A, Calvo EG, Bermu´dez JM, Mene´ndez JA. Carbonisation of resorcinol formaldehyde organic xerogels: effect of temperature, particle size and heating rate on the porosity of carbon xerogels. J Anal Appl Pyrolysis 2013;100:111–6. [25] Ortona A, D’Angelo C, Gianella S, Gaia D. Cellular ceramics produced by rapid prototyping and replication. Mater Lett 2012;80:95–8. [26] D’Angelo C, Ortona A, Colombo P. Finite element analysis of reticulated ceramics under compression. Acta Mater 2012;60(19):6692–702. [27] Ortona A, Rezaei E. Modeling the properties of cellular ceramics: from foams to lattices and back to foams. Adv Sci Technol 2014;91:70–7. [28] Lu X, Chen L, Amini N, Yang S, Evans JRG, Guo ZX. Novel methods to fabricate macroporous 3D carbon scaffolds and ordered surface mesopores on carbon filaments. J Porous Mater 2012;19:529–36.
[29] Sullivan RM, Ghosn LJ, Lerch BA. A general tetrakaidecahedron model for open-celled foams. Int J Solids Struct 2008;45(6):1754–65. [30] Liu L, Wang F-Y, Shao G-S, Yuan Z-Y. A low-temperature autoclaving route to synthesize monolithic carbon materials with an ordered mesostructured. Carbon 2010;48(7):2089–99. [31] Xie M, Yang J, Liang J, Guo X, Ding W. In situ hydrothermal deposition as an efficient catalyst supporting method towards low-temperature graphitization of amorphous carbon. Carbon 2014;77:215–25. [32] Wang Z, Zhang X, Liu X, Lv M, Yang K, Meng J. Co-gelation synthesis of porous graphitic carbons with high surface area and their applications. Carbon 2011;49(1):161–9. [33] Sevilla M, Fuertes AB. Catalytic graphitization of templated mesoporous carbons. Carbon 2006;44(3):468–74. [34] Han S, Yun Y, Park K, Sung Y, Hyeon T. Simple solid-phase synthesis of hollow graphitic nanoparticles and their application to direct methanol fuel cell electrodes. Adv Mater 2003;15(22):1922–5. [35] Oya A, Marsh H. Phenomena of catalytic graphitization. J Mater Sci 1982;17(2):309–22. [36] Oya A, Otani S. Three phase catalytic graphitization of a phenolic resin carbon by use of nickel particles. Carbon 1978;16(2):153–4. [37] Yudasaka M, Tasaka K, Kikuchi R, Ohki Y, Yoshimura S. Influence of chemical bond of carbon on Ni catalyzed graphitization. J Appl Phys 1997;81(11):7623–9. [38] Marsh H, Crawford D, Taylor DW. Catalytic graphitization by iron of isotropic carbon from polyfurfuryl alcohol, 725–1090 K a high resolution electron microscope study. Carbon 1983;21(1):81–7. [39] Warren BE. X-ray diffraction study of carbon black. J Chem Phys 1934;2(9):551–6. [40] Everall NJ, Lumsdon J, Christopher DJ. The effect of laserinduced heating upon the vibrational Raman spectra of graphites and carbon fibres. Carbon 1991;29(2):133–7. [41] Bernard S, Beyssac O, Benzerara K, Findling N, Tzetkov, Brown GE. XANES, Raman and XRD study of anthracenebased cokes and saccharose-based chars submitted to hightemperature pyrolysis. Carbon 2010;48(9):2506–16. [42] Canc¸ado LG, Takai K, Enoki T, Endo M, Kim YA, Mizusaki H, et al. General equation for the determination of the crystallite size La of nanographite by Raman spectroscopy. Appl Phys Lett 2006;88(16):163106-1–3. [43] Chollon G, Delettrez S, Langlais F. Chemical vapour infiltration and mechanical properties of carbon open-cell foams. Carbon 2014;66:18–30. [44] Brezny R, Green DJ. Uniaxial strength behavior of brittle cellular materials. J Am Ceram Soc 1993;76(9):2183–90. [45] Mora RJ, Waas AM. Strength scaling of brittle graphitic foam. Proc R Soc Lond Ser A 2002;458(2023):1695–718. [46] Celzard A, Zhao W, Pizzi A, Fierro V. Mechanical properties of tannin-based rigid foams undergoing compression. Mater Sci Eng A 2010;527(16–17):4438–46. [47] Zhao M, Song H. Catalytic graphitization of phenolic resin. J Mater Sci Technol 2011;27(3):266–70. [48] Drits VA, Tchoubar C. X-Ray diffraction by disordered lamellar structures. Berlin Heidelberg: Springer; 1990. 199– 232. [49] Maire J, Mering J. Graphitization of soft carbons. In: Walker PL, editor. Chemistry and physics of carbon, vol. 6. New York: Dekker; 1970. p. 125–90. [50] Beyssac O, Goffe´ B, Petitet JP, Froigneux E, Moreau M, Rouzaud JN. On the characterization of disordered and heterogeneous carbonaceous materials by Raman spectroscopy. Spectrochim Acta A 2003;59(10):2267–76. [51] Tuinstra F, Koenig JL. Raman spectrum of graphite. J Chem Phys 1970;53(3):1126–30.
CARBON
8 8 (2 0 15 ) 7 0–85
[52] Knight DS, White WB. Characterization of diamond films by Raman spectroscopy. J Mater Res 1989;4(2):385–93. [53] Ferrari AC, Robertson. Robertson. Interpretation of Raman spectra of disordered and amorphous carbon. Phys Rev B 2000;61(20):14095–107. [54] Canc¸ado LG, Jorio A, Pimenta MA. Measuring the absolute Raman cross section of nanographites as a function of laser energy and crystallite size. Phys Rev B 2007;76(6):064304-1–7. [55] Be´ny-Bassez C, Rouzaud JN. In: O’Hare AMF, editor. Scanning Electron Microscopy. Chicago: SEM Inc.; 1985. p. 119. [56] Ferrari AC, Meyer JC, Scardaci V, Casiraghi C, Lazzeri M, Mauri F, et al. Raman spectrum of graphene and graphene layers. Phys Rev Lett 2006;97(18):187401–4. [57] Oya A, Otani S. Influences of particle-size of metal on catalytic graphitization of non-graphitizing carbons. Carbon 1981;19(5):391–400. [58] Rouzaud JN, Oberlin A. Structure, microtexture, and optical properties of anthracene and saccharose-based carbons. Carbon 1989;27(4):517–29.
85
[59] Iwashita N, Park CR, Fujimoto H, Shiraishi M, Inagaki M. Specification for a standard procedure of X-ray diffraction measurements on carbon materials. Carbon 2004;42(4):701–14. [60] Field JS, Swain MV. The indentation characterization of the mechanical properties of various carbon materials: glassy carbon, coke and pyrolytic graphite. Carbon 1996;34(11):1357–66. [61] Zhao W, Fierro V, Pizzi A, Du G, Celzard A. Effect of composition and processing parameters on the characteristics of tannin-based rigid foams. Part II: physical properties. Mater Chem Phys 2010;123(1):210–7. [62] Amaral-Labat G, Sahimi M, Pizzi A, Fierro V, Celzard A. Mechanical properties of heat-treated organic foams. Phys Rev E 2013;87(3):032156-–032156-7. [63] Gibson LJ, Ashby MF. Cellular solids. Structure and properties. 2nd ed. Cambridge (UK): Cambridge University Press; 1997.
