Qualitative and quantitative analysis of the process chain: wood polymer composite – C-template – C/Si/SiC-ceramic by means of X-ray computed tomography Bernhard Plank1, Sascha Senck1, Christian Fürst2, Hannes Scharleitner1, Johann Kastner1 1
University of Applied Sciences Upper Austria, Stelzhamerstraße 23, 4600 Wels, Austria, e-mail: [email protected]
Kompetenzzentrum Holz GmbH, Altenberger Straße 69, 4040 Linz, Austria, e-mail: [email protected]
Abstract The quality of a ceramic is defined by its homogenous and non-porous volume. These parameters determine the durability and failure probability of a component. By means of XCT, the structural changes within the samples over the whole manufacturing route were evaluated. In the qualitative evaluation at lower XCT resolutions (> (70 µm)³ voxel size) pores, cracks, and higher- and lower dense inclusions were compared within and between corresponding samples. Cracks in the wood polymer composite (WPC) tend to shrink in the C-template during carbonisation. In some WPCs no big cracks were visible while in the C-templates big cracks appeared due to carbonising and shrinkage – but these cracks are fully closed in the ceramic after the Si infiltration-step. For quantitative analysis of phase composition, high resolution XCT-measurements between (0.8 µm)³ and (3 µm)³ voxel size were applied. Due to the fact, that the absorption contrast between the wood and polymer matrix in the WPC and contrast between Si / SiC and C / Air in the ceramic is very limited, this quantification step is very challenging. For the C-templates, applying a simple threshold algorithm, the quantification by XCT leads to promising results and fits well to LOM and mercury porosimetry. With SEM/EDS higher dense inclusions, visible in XCT, were analyesed and mainly consist of the elements C-O and trace amounts of, for example, Fe, Zn, Si or Mg. Keywords: Ceramic, WPC, C-template, X-ray Computed Tomography, Mercury porosimetry, SEM
1. Introduction In material development the use of renewable resources is an on-going process. This paper presents the results of a project for the development of biogenic C/Si/SiC-ceramic based on wood polymer composites (WPC). Silicon carbide (SiC) is one of the most important technical ceramics due to its hardness, chemical resistance and thermal properties even at very high temperatures. For cost reduction, WPCs were used as green-bodies, which are processed by extrusion or injection molding making 3D structures easily accessible. After a carbonisation process a porous carbon-template is produced, which can be transferred to a biogenic SiC ceramic by liquid silicon infiltration. For understanding and follow-up of the manufacturing process of such a biogenic ceramic it is important to determine the parameters that are influencing the quality parameters following each manufacturing step.  In a first feasibility study X-ray computed tomography (XCT) was chosen, because it is a very powerful non-destructive test-method to gain qualitative and quantitative results within a large volume . The results gained with XCT were compared with light optical microscopy (LOM), scanning electron microscopy (SEM) equipped with energy-dispersive X-ray spectroscopy (EDS) and mercury porosimetry.
iCT Conference 2014 – www.3dct.at
2. Material and Methods 2.1 Materials For the investigated samples the manufacturing route consists of three main steps. In step (I) WPCs are manufactured by extrusion. The WPCs consist of 50 to 62 wt.% of chopped wood particles between 100 and 500 µm, a matrix of 10 to 12 wt.% thermoplastic modificator and 25 to 32 wt.% phenolic resin. Step (II) was the carbonisation of the WPC by pyrolysis in an inert atmosphere (Nitrogen, argon) to produce a C-template. WPC samples were pyrolysed at 900 °C (C-template Type 1) and at 1600 °C (C-Template Type 2) to close nanopores < 2 nm . In the last production step (III), liquid-siliconinfiltration at 1600 °C and vacuum (1-5 mbar) of the C-template took place to receive the final C/Si/SiC-ceramic. For this work, different samples in the range of ~ 1 mm³ up to > 80 mm³ from every production step were investigated.
2.2 X-ray computed tomography (XCT) Depending on the size and maximum penetration length of the samples, they were investigated on two different XCT devices. Samples with a maximum penetration length of > 40 mm were scanned on a RayScan 250 E device, which is equipped with a 2048² pixel flat panel detector, a 450 kV mini and a 225 kV micro focus X-ray tube. Smaller samples were scanned on a GE phoenix|x-ray nanotom 180 NF device, equipped with a 2304² pixel flat panel detector and a 180 kV nanofocus X-ray tube. The voxel size (VS) was between (0.8 µm)³ and (70 µm)³. An overview of applied measurement parameters is given in Table 1. Data evaluation was mainly done in VG Studio Max 2.2. Sample / Material
WPC (fig. 1)
C-template (fig. 1)
C/Si/SiC-ceramic (fig. 1)
WPC (fig. 2)
C-template (fig. 2)
C/Si/SiC-ceramic (fig. 2)
Ref-Si, SiC, C (fig. 4)
WPC (fig. 3)
WPC quantitative (fig. 5)
C-template (fig. 3)
C/Si/SiC-ceramic (fig. 3)
C-template (fig. 7)
C/Si/SiC-ceramic (fig. 9)
Table 1: XCT-Measurement parameters on RayScan 250 E (grey) and Nanotom 180 NF. The described values are voltage (U) and current (I) on the nano-focus tube, integration time (Tint) on the detector, the number of projection images (n) and the resulting measurement time (Tmeas) at the specific voxel size (VS)
2.3 Standard methods / Reference methods Sample preparation and grinding was done at a Struers LapPol5 (with a LaboForce3 specimen-holder) and for light optical microscopy (LOM) an Olympus device was used. Scanning electron microscopy (SEM) was done on a TESCAN VEGA LMU 2 equipped with Energy-dispersive X-ray spectroscopy (EDS) from Oxford. Mercury porosimetry investigations were done externally on a Pascal Mercury Intrusion Porosimeter from Thermo Scientific. The calculation of void content is based on a simple mathematical model (Washburn's equation ). Synchrotron computed tomography was done at DESY in Hamburg with energy of 18 keV. Tomograms were reconstructed with (1.2 µm)³ voxel size.
3. Results and Discussion 3.1 Qualitative XCT-evaluation As the WPC consists of resins, polymers and chipped wood, the density and composition vary within the volume. Additionally, the structures within a sample are quite small and the sample shrinks about 25 % during the carbonisation in step (II). These are the challenges for the manufacturing process compared to synthetic basic components (e.g. woven C-fibres)  for ceramic production. The quality of a ceramic is defined by its homogenous and non-porous volume. These parameters determine the durability and failure probability of a component. By means of XCT, the geometrical and the structural changes within the samples over the whole manufacturing route were evaluated. The shrinkage during the carbonisation step (II) is clearly seen between the WPC and the C-template. In the qualitative evaluation at lower XCT resolutions (> (45 µm)³ voxel size) for example pores, cracks, higher- and lower dense inclusions were compared within the samples, as depicted in Figure 1. (#1.1) shows a crack in the WPC, which shrinks in the Ctemplate (#2.1) and is still present in the ceramic (#3.1). In the WPC no big cracks are visible (#1.2). In the C-template a big crack appears (#2.2) due to carbonisation and shrinkage, which is fully closed in the ceramic (#3.2) after Si infiltration.
Figure 1: Overview and detail view of XCT slices of the same position within a sample at three different manufacturing steps: WPC (left), C-template Type 1 (middle) and C/Si/SiC-ceramic (right). f.l.t.r.: voxel size: (70 µm)³ / (45 µm)³ / (52 µm)³. Further structural changes within the different production steps are depicted in Figure 2. In the WPC (left) the inhomogeneity at (#1.3) is barely visible, after carbonisation this inhomogeneity (#2.3) has a significantly higher grey value compared to the C-template and in the last production step this inhomogeneity still remains (#3.3), but
iCT Conference 2014 – www.3dct.at
compared to the high absorbing SiC matrix it has a lower grey value. Another inhomogeneity (#1.4) is visible in the WPC but it disappears after carbonisation and leaves a pore (#2.4). Depending on the production parameters used and the size of inner structures, most of the cracks and pores disappear during Si infiltration (top row). In the case that the defects are too big (bottom row (#2.5)) they were not closed during Si infiltration (#3.4 and #3.5).
Figure 2: Detail view of XCT slices of the same position within a sample at three different manufacturing steps: WPC (left), C-template Type 1 (middle) and C/Si/SiC-ceramic (right). f.l.t.r.: VS (top row): (70 µm)³ / (45 µm)³ / (52 µm)³; (bottom row): (22 µm)³ / (16 µm)³ / (16 µm)³
The quality of a ceramic is defined by its homogenous and non-porous volume. These parameters are the determining factors for the durability or failure probability of a component. For analysis of phase composition, high resolution XCT-measurements (< (3 µm)³ voxel size) were applied to the different samples. Figure 3 shows high resolution XCT slices of the individual production steps of C/Si/SiCceramics. In the WPC (left), the chipped wood particles and pores could be distinguished from the polymer matrix. Additionally there are some higher dense inclusions that can be recognized. After carbonisation, the polymer matrix disappears in the C-template (middle) and only pores and the higher dense inclusions remain. In the final C/Si/SiC-ceramic (right) four different phases are visible, but the differences in grey values between air / C and Si / SiC are quite low.
Figure 3: Detailed view of XCT slices of a WPC (left) and a C-template Type 1 (middle) after carbonisation of the same sample. The different sizes of the images give an impression of the shrinkage during the carbonisation process. The C/Si/SiC-ceramic (right) obtained by liquid-Si-infiltration of a C-template mainly shows the higher dense Si/SiC and the lower dense C/air phases. VS f.l.t.r: (2.5 µm)³ / (2 µm)³ / (3 µm)³.
To determine the relative differences in the grey values of the XCT-Results, test samples of clean SiC beta-phase 98%, Si, and C were collectively scanned in the same measurement and the grey values of each material were compared in Figure 4. For our reference samples, a small (2.6*1.3*4 mm³), homogeneous region of interest without artefacts of each material were evaluated. Therefore we could clearly separate C from air and with some overlap Si from SiC (right diagram). These grey values are
more or less “ideal” grey values. Because of the small structures and the limited resolution partial volume effects are present and more overlapping grey values in the final C/Si/SiC-ceramic will occur.
Figure 4: Reference samples of clean SiC, Si, and C (left). Grey values of each material from a homogenous region of 2.6*1.3*4 mm³ were evaluated (right). VS: (3 µm)³
3.2 Quantitative XCT-evaluation Due to the fact that the absorption contrast between the wood and polymer matrix in the WPC and contrast between Si / SiC and C / air in the ceramic is limited, quantification is very challenging. Additionally, the structures for quantification in the C-template and in the ceramic are quite small. In the WPC, the content of wood is of major interest. In the C-template the void content and distribution, and in the ceramic the content of voids, C, Si, and SiC are relevant. In the following section, we try to find quantitative values due to simple and timesaving thresholding. Figure 5 displays the segmentation for the investigated WPC. Because the contrast between matrix (phenolic resin matrix, thermoplastic modificator) and wood is quite low (left), for some matrix – wood combination (phenolic resin matrix) there is almost no contrast at all (right) a manual threshold by visual assessment was chosen. The quantitative results displayed here are only very raw estimates and depend on the user, but fit well with the process parameters (55 & 62 wt.%). The WPC on the left consists of ~53 vol.% and the right one of ~59 vol. % of wood. A quantification of the void content in the WPC would be easier, but has no high significance for this production step of C/Si/SiC-ceramics.
Figure 5: Segmentation of WPCs: ~53 vol.% (left) and ~59 vol.% wood content (right). Gauss 5; VS: (2.5 µm)³.
Figure 6 shows the target preparation of a WPC after XCT investigation. Due to the mechanical grinding process, bigger particles break out (#1) of the matrix and smaller holes (#2) are smeared up (left). Additionally, for further image analysis with LOM the contrast between wood (#3) and matrix (#4) is lower than with XCT (right).
Figure 6: LOM image (magnification 50x) after target preparation of an WPC (left) and XCT cross section before grinding. VS: (2.5 µm)³.
As mentioned above, in the C-template the void content and distribution is the main characteristic for quantification. Because the contrast between the air and C is good (Fig. 4 (right), a global ISO50 threshold between the Air- and C-peak was chosen for further quantification of the void content. This simple threshold method can be used for relative comparison of different samples scanned with the
iCT Conference 2014 – www.3dct.at
same measurement parameters. For absolute values of void content, this threshold should be adapted (some kind of offset correction) so that these values fit well with a chosen (accurate) reference method, as done for example in [5, 6] for carbon fibre reinforced polymers (CFRP). In our analysis an ISO62 threshold was used, calculated by the mean grey values: ISO62 = air + (material – air)*0.62. Figure 7 shows two different C-templates, one is carbonised at 1600 °C (C-template Type 2 - top row and 3D-rendering) and the second is carbonised at 900 °C (C-template Type 1- bottom row). The pictures in the middle show the segmentation coloured by defect volume. The red selection shows a big pore network of several connected pores. Separated pores have a small volume and are coloured in blue or green as visible in the 3D image. The evaluated void content of C-template Type 1 was 26.5 vol.% and for Type 2 27.7 vol.%.
Figure 7: C-Template Type 1 and Type 2. Voids segmented with ISO50 threshold (middle). Gauss 5, VS: (2 µm)³.
In Table 2, grey values and the calculated threshold for the segmentation of voids in C-templates are shown. In addition the evaluated porosity of XCT compared to LOM evaluation and mercury porosimetry is shown. XCT ISO62 and LOM have very similar values and the same trends between Ctemplate Type 1 and Type 2. The determination of pore distribution by means of mercury porosity shows no significant differences (< 0.3 vol.%) between both types. A comparison of void distribution between our XCT results and mercury porosimetry is not meaningful, because with the chosen software tools, merely one big pore (99.97 % of total porosity) was segmented. For future work, and better evaluation of void distribution some kind of watershed algorithm should be applied. Pores smaller than 4 µm in diameter could not be resolved with the chosen XCT-resolution. For Siinfiltration voids between 1 - 50 µm are mainly relevant. C‐template Type 1 – 900 °C Type 2 – 1600 °C
grey value threshold XCT ISO50 air material (C) ISO 50 % ISO 62 % [vol.%] 21,856 35,491 28,674 30,310 26.51 25,560 41,097 33,329 35,193 27.78
XCT ISO62 [vol.%] 33.70 34.92
LOM [vol.%] 34.60 35.90
mercury porosimetry [vol.%] 33.96 33.74
Table 2: Grey values and threshold for XCT-evaluation of C-templates. Results of XCT and reference methods for porosity evaluation with LOM and mercury porosimetry.
In Figure 8 a binarized LOM image (left) of C-template Type 2, used for void quantification in Table 2, is displayed. Since some structures break out during the grinding process bigger voids (black) are visible with LOM compared to XCT. Additionally, two SEM images of C-template Type 2 are displayed. Using SEM the irregular shape of the voids could clearly observed. Many voids are quite small < 4 µm (middle) and in addition, some particles (right) could be observed. Due to charging effects (electrical insulator) these particles are bright in SEM. On several of these particles an EDS
analysis was done. Compared to the C-template which consists of 99 wt.% C and 1 wt.% Ca, the displayed particle consists of 71.85 wt.% C, 27.23 wt.% O, 0.58 wt.% Br and 0.34 wt.% Cl. Other analysed particles have trace amounts (> 1 wt.%) of the following additional elements: Na, K, Zn, Si, Fe, Br, Te, Ca, Mg, Rh, Cl. Due to the different absorption coefficients, these particles could also be observed in most of the XCT images as higher density inclusions as described in Figure 2 (#1.3, #2.3 & #3.3) or Figure 3.
Figure 8: Binarized LOM image (left) and SEM (middle, right) of C-template Type 2.
Figure 9: Final C/Si/SiC-ceramic: XCT slices generated on Nanotom XCT (0.8 µm)³ and synchrotron CT (1.2 µm)³ VS. LOM and SEM image after target preparation of C/Si/SiC-ceramic.
A full quantification of all different phases in the final C/Si/SiC-ceramic was not possible with XCT due to the limited contrast and the small structures which are on the detection limit of XCT (Figure 9). By visual thresholding only the SiC content could be estimated. For Nanotom XCT SiC content of 76.6 vol.% and for synchrotron 80.45 vol.% was estimated. With LOM, a SiC content of 85.6 vol.% and for Si of 8.4 vol.% could be observed. C and Air (~ 6 vol.%) could not be distinguished due to break out during the grinding process. The SEM images clearly show that most of the breakouts (black in LOM) consist of C with very small voids within.
iCT Conference 2014 – www.3dct.at
4. Conclusion An overview of the methods applied for materials characterisation of WPC, C-template and C/Si/SiCceramics is given in Table 3.
chemical analysis size distribution volume distribution evaluated structures [µm] tested volume/area quantitative analysis: WPC C-template SiC non-destructive method
yes yes 2D no > 0.5 very small
synchrotron CT no yes 3D yes >3 small
mercury porosimetry no yes (mathematical) yes 0.001 … 100 middle
/ no no no
/ / only SiC yes
no yes no no / contamination
no yes 3D yes >4 large to small
no yes 2D no >1 small
only estimation yes only SiC yes
no yes Si and SiC no
Table 3: Comparison of used methods applied on WPC, C-template and C/Si/SiC-ceramics; (“ / “ not applied)
XCT has big advantages for the qualitative non-destructive evaluation of different manufacturing steps. It could be clearly shown that there are some higher dense phase structures in the WPC, which can’t be eliminated by the carbonisation and Si-infiltration steps. In SEM/EDS these higher dense structures could be identified as C-O structures with a significant amount of metallic elements. Further qualitative investigations show that smaller pores and cracks are well closed by Si-infiltration. Quantification of different phases by XCT is possible, but a verification of these values is quite difficult because destructive methods as LOM only give 2D information of a relative small volume. In addition, during the grinding process some of the fragile structures break out and influence the segmentation results. Other methods like mercury porosimetry are based on simple mathematical models which can be inaccurate for such complex void networks. For WPC, an evaluation with LOM was not meaningful. With XCT, a raw estimation of wood content could be gained. For the C-templates the quantification by XCT with an ISO62 threshold algorithm leads to quite the same results as obtained with LOM and mercury porosimetry (max. deviation ~1.2 vol.%). Due to the low contrast differences and the small dimensions of the phase structures in C/Si/SiC-ceramics, only an estimation of SiC content with XCT was possible. For the C/Si/SiCceramics LOM is the better method to gain meaningful quantitative results of Si and SiC content. Acknowledgement This work was supported by the project HMV-3D financed by the programme “Regio13—Regionale Wettbewerbsfähigkeit” of the European commission, by the European Regional Development Fund ERDF (Europäischer Fonds für Regionale Entwicklung EFRE) and the Government of Upper Austria.” Further thanks go to Markus Gillich for supporting us with SEM/EDS at the University of Applied Sciences Upper Austria and Guillermo Requena from Vienna University of Technology for synchrotron computed tomography scans. References  Fürst C.; Katzenberger O; Porous carbon-templates for biogenic SiC-ceramics on the basis of extruded thermoset-based wood polymer composites; Proc. Cellular Materials, 7-9. Nov. 2012, Dresden, Germany, (2012).  J. Kastner – Editor, Proceedings: Conference on Industrial Computed Tomography (ICT) 2012, Wels, Austria, (2012).  Edward W. Washburn, The Dynamics of Capillary Flow. In: Physical Review. Band 17, Nr. 3, S. 273–283 (1921).  J.M. Hausherr, et al., Material characterisation of C/SiC: Comparison of computed-tomography and scanning electron microscopy. In Conference on Damage in Composite Materials, (2006)  B. Plank, et al., Porositätsbestimmung in der Flugzeugindustrie mittels Röntgen-Computertomografie - Proceedings Industrielle Computertomografie Fachtagung 2010, Wels, Österreich, pp. 25-34 (2010).  R. Stoessel, et al., μ-Computed Tomography for 3D Porosity Evaluation in Carbon Fiber Reinforced Plastics (CFRP), International Symposium on Digital Industrial Radiology and Computed Tomography, (2011).