W.Li, Z.H. Chen/Ceramics Intemational 35 (2009)747-753 inside, whose maximum sizes are about 10-20 um From the point of configuration and sizes, these channels both resemble the special pores in 20-0. 1 um range that are deduced from the MIP results. Therefore. it is reasor channel network plays an important role in 3D-CPSic Like MiP, bubble point method is also an intrusion method, while it can only characterize those pores which are always open throughout the samples, and its data reflect the narrowest es of the 5 shows the cumulative volume curve and Psd of 3D-C Sic obtained via this method It can be seen that the pores sizes distribute concentratively. Most of the open pores(>95 vol %)are 0.3-0.4 um-sized, and the single peak on PSD curve appears at 0.35 um, while the first bubble Fig. 4. The paths between the carbon bundles. The ar ndicate the path: emerged at 2.7 um, the maximum throats size. By comparison, that are not fully blocked by matrix. Area b refers to the same location in Fig 3. the range of PSD by bubble point method is only a small part of the one by MIP, which implies that minority pores of total are full-open actually. Obviously, the full-open pores size is in the extrusion starts, the Hg flow retreats from the chambers via the range of the inter-bundle channels (20-0. 1 um), and is hroats, and tends to snap-off for the great disparity of their especially close to the critical pore sizes of the mercury sizes from hundreds of microns to 20 um or less, thus some Hg extrusion(0. 1-0. 2 um). This result proves that the channels can retained and entrapment occurs, according to the relevant provide the passage through which the fluids flow across the theories[ 14]. Therefore, the real sites of entrapment among 20- sample on percolation mechanism, and their throat size 0. 1 um is the inside large chambers between bundles, while determines the difficulty of flowing, e. g, the type C channel they are shielded by narrower throats, which is different from in Fig. 3 controls the threshold pressure at which the fuids can the exposed ones. Hence, the left of the lst extrusion curve can penetrate into the samples be deduced as the dash line in Fig. 1, providing the pressures are reduced continuously from the atmospheric value 3.3. Evolution of porosity Also in 20-0. 1 um range, the Ist extrusion, 2nd intrusion and 2nd extrusion overlap each other on the whole, without any To investigate the formation and evolution of the hysteresis or entrapment. This phenomenon implies that the Hg complicated porosity of 3D-CfSiC, the partially densified flow can move reversibly in these pores of 20-0. 1 um, which specimens at intermediate stages, i.e., enduring various PIP are free from the effects of shielding or heterogeneity of sizes. recycles, were also characterized by MIP, the results are plotted Below 0.1 um in Fig. 2, though very small, the hysteresis in Figs. 6 and 7. does exist between the intrusion and extrusion both in lst and In Fig. 6, the total intrusion volume of 3PlP 3D-CuSiC is 2nd cycle, while the entrapment is not observed. The starting much higher than the finished, and the profile of the former's point of 2nd extrusion is 0. I um, which is similar to the lst PSD curve is also different from the latter. The first peak lies one s from above 100-40 um, which is caused by the exposed inter- According to all the analysis above, it is clear that there are bundle chambers, the second peak reflects the summation of the some special pores with multiple functions in 3D-C / SiC, volumes of the inter-bundle channels and the chambers whose sizes are in the range of 20-0. 1 um. On one hand, these shielded by them, while it has much higher increments than pores act as the throats at the ends of the convergent large chambers of hundreds of microns, limiting and joining them, on 20 the other hand, they connect and shield the small pores under 0 I um. Meanwhile, these pores lead to the sample surface,.16 allowing the Hg flowing unrestrictedly and reversely. Thus comes into being inside the 3D-CfSiC From the micrograph of 5 of adjacent bundles, some connect the inter-bundle chambers, e.g., type B, and others join the chambers indirectly via else 5 6 paths, e.g., type C. Especially, type C paths reside near the surface and lead to the outside These paths connect each other, 2 spread inside and make a path tree. As the SEM images shown n Fig. 4, these paths are not fully blocked by the filled matrix ike the area B, but include mass openings and micro-channels Radius(um) owing to the incomplete infiltration or micro-cracks derived Fig. 5. Cumulative volume curve and PSD of 3D-CSic by bubble point from the pyrolysis of precursor. Hence, they are called channels methodextrusion starts, the Hg flow retreats from the chambers via the throats, and tends to snap-off for the great disparity of their sizes from hundreds of microns to 20 mm or less, thus some Hg retained and entrapment occurs, according to the relevant theories [14]. Therefore, the real sites of entrapment among 20– 0.1 mm is the inside large chambers between bundles, while they are shielded by narrower throats, which is different from the exposed ones. Hence, the left of the 1st extrusion curve can be deduced as the dash line in Fig. 1, providing the pressures are reduced continuously from the atmospheric value. Also in 20–0.1 mm range, the 1st extrusion, 2nd intrusion and 2nd extrusion overlap each other on the whole, without any hysteresis or entrapment. This phenomenon implies that the Hg flow can move reversibly in these pores of 20–0.1 mm, which are free from the effects of shielding or heterogeneity of sizes. Below 0.1 mm in Fig. 2, though very small, the hysteresis does exist between the intrusion and extrusion both in 1st and 2nd cycle, while the entrapment is not observed. The starting point of 2nd extrusion is 0.1 mm, which is similar to the 1st one’s. According to all the analysis above, it is clear that there are some special pores with multiple functions in 3D-Cf/SiC, whose sizes are in the range of 20–0.1 mm. On one hand, these pores act as the throats at the ends of the convergent large chambers of hundreds of microns, limiting and joining them, on the other hand, they connect and shield the small pores under 0.1 mm. Meanwhile, these pores lead to the sample surface, allowing the Hg flowing unrestrictedly and reversely. Thus, these pores integrate the others, and a 3D porous network comes into being inside the 3D-Cf/SiC. From the micrograph of the morphology (Fig. 3), there are apparent paths at the borders of adjacent bundles, some connect the inter-bundle chambers, e.g., type B, and others join the chambers indirectly via else paths, e.g., type C. Especially, type C paths reside near the surface and lead to the outside. These paths connect each other, spread inside and make a path tree. As the SEM images shown in Fig. 4, these paths are not fully blocked by the filled matrix like the area B, but include mass openings and micro-channels, owing to the incomplete infiltration or micro-cracks derived from the pyrolysis of precursor. Hence, they are called channels inside, whose maximum sizes are about 10–20 mm. From the point of configuration and sizes, these channels both resemble the special pores in 20–0.1 mm range that are deduced from the MIP results. Therefore, it is reasonable to suppose that the channel network plays an important role in 3D-Cf/SiC. Like MIP, bubble point method is also an intrusion method, while it can only characterize those pores which are always open throughout the samples, and its data reflect the narrowest parts’ sizes of the pores. Fig. 5 shows the cumulative volume curve and PSD of 3D-Cf/SiC obtained via this method. It can be seen that the pores’ sizes distribute concentratively. Most of the open pores (>95 vol.%) are 0.3–0.4 mm-sized, and the single peak on PSD curve appears at 0.35 mm, while the first bubble emerged at 2.7 mm, the maximum throat’s size. By comparison, the range of PSD by bubble point method is only a small part of the one by MIP, which implies that minority pores of total are full-open actually. Obviously, the full-open pores’ size is in the range of the inter-bundle channels (20–0.1 mm), and is especially close to the critical pore sizes of the mercury extrusion (0.1–0.2 mm). This result proves that the channels can provide the passage through which the fluids flow across the sample on percolation mechanism, and their throat size determines the difficulty of flowing, e.g., the type C channel in Fig. 3 controls the threshold pressure at which the fluids can penetrate into the samples. 3.3. Evolution of porosity To investigate the formation and evolution of the complicated porosity of 3D-Cf/SiC, the partially densified specimens at intermediate stages, i.e., enduring various PIP recycles, were also characterized by MIP, the results are plotted in Figs. 6 and 7. In Fig. 6, the total intrusion volume of 3PIP 3D-Cf/SiC is much higher than the finished, and the profile of the former’s PSD curve is also different from the latter. The first peak lies from above 100–40 mm, which is caused by the exposed inter￾bundle chambers, the second peak reflects the summation of the volumes of the inter-bundle channels and the chambers shielded by them, while it has much higher increments than Fig. 4. The paths between the carbon bundles. The arrows indicate the paths that are not fully blocked by matrix. Area B refers to the same location in Fig. 3. Fig. 5. Cumulative volume curve and PSD of 3D-Cf/SiC by bubble point method. 750 W. Li, Z.H. Chen / Ceramics International 35 (2009) 747–753