2.1. Oxidation apparatus

A schematic diagram of the experimental setup used in this study is presented in Fig. 1. It includes two main parts, namely a magnetic suspension balance and gas dosing system.

Fig. 1.Full-screen View

Schematic diagram of the experimental setup

The magnetic suspension balance allows changes in force and mass in controlled environments (pressure, temperature) to be measured with high accuracy. This system makes it possible to weigh samples with no contact over a large range of pressures and temperatures. Instead of hanging directly on the balance, a sample is linked to a "suspension magnet," which consists of a permanent magnet, sensor core, and device for decoupling the measured load (sample). An electromagnet maintains a freely suspended state for the suspension magnet based on an electronic control unit. Based on this magnet suspension coupling, the measured force is transmitted without any contact between the measurement chamber and microbalance, which is located outside the chamber under ambient atmospheric conditions. A controlled suspended state is achieved by using a direct analogous control circuit (proportional-integral-derivative (PID) controller and position transducer). This circuit modulates the voltage supplied to the electromagnet such that the suspension magnet is held in a constant vertical position. A microcontroller-driven digital set point controller superimposed over the direct PID controller facilitates various position settings for the suspension magnet.

The gas dosing system is used to regulate the gas pressure inside the balance volume. This is accomplished by injecting gas at specific pressures from external gas supplies into the balance to increase the pressure or removing gas from the balance via ventilation or vacuum to reduce the pressure. The pressure in the balance is continuously measured by pressure gauges (P1, P2). When combined with the appropriate controller software, this gas dosing system facilitates completely automated oxidation experiments. Table 1 lists the technical specifications of the magnetic suspension balance.

The graphite samples were produced by FangDa (China) Carbon New Material Co., Ltd. via the isotactic pressing of fine-grained materials to create near-isotropic structures. NG-CT-10 and NG-CT-20 are newly developed grades of nuclear-grade graphite with excellent performance. Table 2 lists the physical properties of NG-CT-10 and NG-CT-20 nuclear graphite, as well as IG-110 nuclear graphite [11], which is produced by Toyo Tanso from Japan, and is currently used in the HTR-PM project. NG-CT-10 and NG-CT-20 have superior mechanical properties and lower ash content compared to IG-110. The Young’s modulus, flexural strength and compressive strength of NG-CT-10 and NG-CT-20 are all greater than those of IG110. These materials are expected to be applied in future HTGR reactor projects.

Table 3 lists the impurity contents of NG-CT-10 and IG-110 nuclear graphite. The data for NG-CT-10 were provided by the manufacturer. Some impurity element contents of NG-CT-10, such as V, Ca, and Fe, are greater than those in IG-110, which could catalyze oxidation reactions and lead to a decrease in activation energy [12–13].

The graphite samples were machined into cylinders with a diameter of 6 mm and height of 8 mm. They were machined using a cemented carbide tool to avoid introducing any additional impurities. Before each test, the target specimen was ultrasonically washed in acetone for more than 30 min, then dried at room temperature.

2.2. Experimental procedure for oxidation

The oxidation experiments included four steps: pre-vacuum treatment under a vacuum pressure of approximately 0.01 Pa, heating, oxidation, and system recovery. The first step of the pre-vacuum treatment is to remove oxygen, moisture, and any other impurity gases from the system and from inside the graphite sample. Next, 99.999% pure argon is injected into the specimens under vacuum at a flow rate of 100 standard cubic centimeters per minute (sccm). The gas contained ≤1.5 ppm O₂, ≤0.5 ppm H₂, ≤1 ppm CO+CO₂+CH₄ and ≤3 ppm H₂O in addition to argon. In the second step, the sample is heated to the set oxidation temperature. The oxidation temperature ranges from 500 to 800 °C. In the third step of oxidation, dry air is injected at set volume flow rates. The magnetic suspension balance measures the weight and temperature every 5 s and adjusts the zero point every 3 min. Table 4 lists the key parameters for each step. The interfacial velocity of oxidation in this experiment was 1.28 m/min, which is comparable to most previous oxidation experiments [15–20].

2.3. Microstructural characterization

Sample morphologies were obtained using SEM and the sizes of exposed filler particles on the sample surfaces were derived from SEM micrographs. Porosity and pore size distributions were studied based on mercury intrusion using a Micromeritics Autopore IV 9510 Pore Size Analyzer. Raman spectra of the samples were captured by a HORIBA Jobin Yvon HR800 micro-Raman spectrometer to determine the sizes of nanocrystallites. For each sample, Raman Spectra were collected from three different spots on the sample surface. Each sample was oxidized for 2 h at the set temperature.

3.1. Weight loss

The weight-normalized oxidation rate (denoted OR_w, in g g⁻¹ min⁻¹) of the oxidized specimens versus time during oxidation in a temperature range of 550–800 °C is presented in Fig. 2. The oxidation rate accelerates significantly with an increase in temperature. The oxidation rate of NG-CT-10 at 800 °C is more than 100 times faster than that at 500 °C. The oxidation rate and weight loss rate of NG-CT-20 are both greater than those of NG-CT-10 in the range of 550–750 °C, while the opposite is true at 800 °C. This difference is caused by microstructural difference. The oxidation rate does not increase significantly with an increase in weight loss rate, especially for NG-CT-20 graphite. The curve of the weight loss rate for NG-CT-20 is almost a straight line, which differs from Wang's experimental results for IG-110 [21]. However, in Luo's experimental results [22], the variation curve for normalized weight with oxidation time is also nearly a straight line.

Fig. 2.Full-screen View

Weight loss rates of specimens oxidized at different temperatures

Such differences can be attributed to the effects of weight loss rate on oxidation rate at the same oxidation temperature. Based on the oxidation rate OR₀ of the initial oxidation stage with a weight loss rate of 1%, the relative oxidation rates OR/OR₀ at different weight loss rates can be calculated. The peak oxidation rate typically occurs in the range of 30%–40% weight loss [15]. This peak in relative oxidation rate is closely related to different nuclear graphite grades. The peak relative oxidation rate of IG110 graphite is 7.1, that of NBG18 graphite is 3.3, that of NBG10 graphite is 1.5 [15], and that of NG-CT-10 graphite peak is 2.1 with a peak weight loss rate of 34%. Compared to the peak of IG110 graphite, which is associated with ultrafine filler particles, the microporous structure of NG-CT-10 graphite is less changed and more stable following oxidation. This phenomenon can also be observed in the SEM micrographs.

3.2. Arrhenius plot and activation energy

An Arrhenius plot in a temperature range of 550–800 °C is presented in Fig. 3. The slopes and intercepts of the best-fit lines for two different regimes are presented in Fig. 3 as well. The slope is –E / R and the intercept is ln A, where E is the activation energy, R is the universal gas constant, and A is the pre-exponential factor. According to the relative resistance of chemical reaction kinetics and diffusion, the graphite oxidation process can be divided into three regimes: the chemical kinetics control regime (Regime I), inner diffusion control regime (Regime II), and outer diffusion control regime (Regime III). The logarithm of the oxidation rate in the range of 550−700 °C and 700−800 °C is linearly proportional to 1000 / T. Therefore, the temperature range of the chemical kinetic control regime is 550–700 °C and that of inner diffusion control regime is 700–800 °C for NG-CT-10. The transition temperature between the two regions for NG-CT-10 is 700 °C. The transition temperature between the two regimes for NG-CT-20 is 650 °C.

Fig. 3.Full-screen View

Arrhenius plots of graphite oxidation in air in a temperature range of 550–800 °C

The apparent activation energies are 161.4 kJ/mol for NG-CT-10 and 153.5 kJ/mol for NG-CT-20 with a linearity of approximately 0.99, which proves that the experimental data is accurate. Based on the influence of the internal diffusion process in graphite, the activation energy in Regime II is approximately half that in Regime I.

3.3. Morphology changes and the oxidation model

SEM micrographs are presented in Fig. 4. The insets presents SEM micrographs with higher magnifications for the corresponding samples, which reveal the oxidation surfaces of filler particles. Prior to oxidation, the surfaces are relatively smooth with few exposed filler particles and no very large or deep pores. Following oxidation, the surfaces are rougher and many filler particles are clearly exposed. However, no large new pores are visible on the surfaces.

Fig. 4.Full-screen View

SEM micrographs of NG-CT-10 and NG-CT-20 graphite NG-CT-10: (a) unoxidized and oxidized for 2 h at (b) 600 ºC, (c) 700 ºC, and (d) 800 ºC NG-CT-20: (e) unoxidized and oxidized for 2 h at (f) 600 ºC, (g) 700 ºC, and (h) 800 ºC

Some grades of nuclear graphite, including HSM-SC [23], NGB-18 [23], IG-11 [23], and IG-110 [24], exhibit new surface pores following oxidation. Additionally, original pores are often enlarged or even deepened [24, 25]. NG-CT-10 and NG-CT-20 behave differently following oxidation in that their surfaces become rougher with more exposed filler particles, but with no formation of large new pores. Although the grade dependencies of oxidation have been widely reported, the detailed microstructure dependencies of oxidation have rarely been reported, possibly because of the complexity of graphite microstructures. Chi et al. [26] reported that grade dependencies can be attributed to a number of differences in microstructure, including differences in pore microstructure, grain size or coke type. As newly developed grades of graphite, NG-CT-10 and NG-CT-20 are different from IG-110 in many aspects, including impurity content, density, uniformity, defects and porosity, any of which may be reasons for the grade dependencies of oxidation.

The sizes of exposed filler particles on the sample surfaces were measured from SEM micrographs and the average sizes of filler particles were calculated. As shown in Fig. 5, following oxidation at temperatures below 800 ºC, the average particle size seems to increase, but oxidation should make filler particles smaller. However, the apparent increase in filler particle size does not conflict with this concept because the SEM micrographs only show filler particles on the surfaces of the graphite specimens, where the removal of the binder phase was preferential. For unoxidized specimens, the binder phase occupies the space between filler particles, making it difficult to distinguish individual filler particles in SEM micrographs. Following oxidation at relatively low temperatures, the binder phase is removed before the filler particles, allowing the space between filler particles to be observed more clearly. Additionally, when the temperature reaches 800 ºC, the filler particles oxidize very rapidly, making surface filler particles look much smaller. This reasoning agrees with the fact that the micrographs of samples of both grades of graphite oxidized at 800 ºC are the roughest with the smallest filler particles on the surfaces, as shown in Fig. 4.

Fig. 5.Full-screen View

Filler particle sizes on graphite surfaces

In summary, based on the results of SEM observation, it is believed that on the surfaces of NG-CT-10 and NG-CT-20, the binder phase between filler particles is preferentially removed during oxidation. One can infer that at low oxidation temperatures, the binder is oxidized much faster than the filler particles, increasing surface roughness and exposing filler particles on the surface, as shown in the SEM micrographs. At high temperatures, the filler particles can be oxidized as quickly as the binder, reducing the average particle size. Additionally, at all temperatures, oxygen can only diffuse into the binder phase and filler particles can only be removed from the sample surfaces.

3.4. Pore size distribution

The average pore size and pore volume distribution information for NG-CT-10 and NG-CT-20 graphite before and after oxidation were measured based on mercury intrusion. The results are presented in Fig. 6. One can see that most of the pore volume is contributed by pores with a diameter of approximately 3 μm. These ~3 μm pores correspond to the spaces between filler particles on the sample surfaces. Because no new large pores are generated by oxidation, the pore volume distribution does not change significantly. The pore structures of NG-CT-10 and NG-CT-20 graphite exhibit good overall stability based on their dense structure and uniformity.

Fig. 6.Full-screen View

Pore volume distributions of (a) NG-CT-10 and (b) NG-CT-20

As shown in Fig. 7, the porosities of the graphite grades following oxidation were also measured using mercury intrusion. The unoxidized samples have a porosity of approximately 20%, which is similar to the values of other grades of nuclear graphite, such as HSM-SC and IG-11 [23]. Following oxidation, porosity increases significantly, indicating the development of new pore structures.

Fig. 7.Full-screen View

Porosities of NG-CT-10 and NG-CT-20 following oxidation

Graphite is known to contain multi-dimensional pore structures. In this paper, we refer to pores with a diameter greater than 1 μm as large pores and all other pores as small pores. To study the changes in different pore size regions separately, the average 4V / A pore diameters and volume proportions of large pores (pores > 1000 nm) and small pores (pores in 10-1000 nm) were calculated. The results are listed in Table 5. The proportion of small pores increases significantly following oxidation, while the proportion of large pores decreases. It is worth noting that the average pore size does not increase significantly with an increase in oxidation temperature or weight loss rate. It even decreases in some cases, which is different from the results for IG110 graphite [5]. These phenomena indicate that during oxidation, large amounts of small pores are generated in the binder phase. These small pores do not develop into large pores, which was confirmed by the SEM micrographs. Large or deep pores cannot be observed in the SEM micrographs following oxidation.

3.5. Raman spectroscopy

To measure the sizes of graphite nanocrystallite structures, Raman spectra of the graphite samples were acquired. It is well known that there are four typical Raman modes for graphite: G, D, D’ and 2D [27]. The G (graphite) mode at 1583 cm⁻¹ represents the E_2g optical mode of graphite crystals, which can indicate local structural order and is related to the nanocrystallite size of graphite. The D (disorder-induced) mode represents an inter-valley double-resonant Raman scattering process in which the participation of defects is required. The D’ mode represents an intra-valley double-resonant Raman scattering process. The second-order overtone (2D) results from two successive inelastic phonon scattering processes. Therefore, the participation of defects is not required for this mode. Fig. 8 presents the Raman spectra of unoxidized NG-CT-10 and NG-CT-20 and of samples oxidized at different temperatures. In the spectra ranging from 1000 to 3000 cm⁻¹, only three intense peaks can be observed. These peaks correspond to the D, G, and 2D modes. On the shoulder of the G peak, a small D’ peak can also be observed. Following oxidation, the positions, intensities, and widths of the peaks do not change significantly.

Fig. 8.Full-screen View

Raman spectra of (a) NG-CT-10 and (b) NG-CT-20

The sizes of nanocrystallites in graphite are closely related to the results of Raman spectroscopy. It has been suggested to use the intensity ratio of the D peak to the G peak (I_D / I_G) to determine nanocrystallite sizes [28] or defect numbers [29]. The relationships between I_D / I_G and defect numbers in the low-defect-density regime and high-defect-density regime have been discussed previously [27]. In this study, the Raman spectra of graphite samples oxidized at different temperatures were captured from three different locations on the samples. However, we found that changes in I_D / I_G following oxidation are erratic and exhibit no clear trends. Therefore we used the full weight at half maximum (FWHM) of the G peak to calculate the sizes of nanocrystallites. The FWHM of the G peak in each spectrum was calculated and the average FWHM(G) values are presented in Fig. 9. As suggested by Maslova et al. [28], nanocrystallite size can be calculated as: L_a = 430 / [FWHM(G) − 14] nm. Therefore FWHM(G) values of 15, 20, 25, and 30 cm⁻¹ correspond to nanocrystallite sizes of 430, 72, 39, and 27 nm, respectively. NG-CT-20 always exhibits a smaller FWHM(G) value than NG-CT-10, indicating the presence of larger nanocrystallites in NG-CT-20. Additionally, for both grades of nuclear graphite, FWHM(G) decreases following oxidation, implying that the observed nanocrystallite size decreases.

Fig. 9.Full-screen View

FWHM(G) values for NG-CT-10 and NG-CT-20

However, it was determined that the oxidized samples contain larger nanocrystallites than the unoxidized samples. As discussed previously, the binder phase is easier to remove compared to the filler particles. Wen et al. [30] investigated the complexity of the binder structures in Gilsocarbon and Pile Grade A. They observed well-graphitized structures, quinoline insoluble particles, nanoscale graphite structures, chaotic structures, and nongraphitizing carbon in the binders of these nuclear graphites. They found that the sizes of graphite crystallites in binders can vary from several nanometers to tens of nanometers. Therefore, the majority of nanocrystallites in binders are smaller than those in filler particles. When the binder phase is preferentially oxidized and removed, there are more exposed filler particles on graphite surfaces. Therefore, the observed average nanocrystallite size increases.

In summary, the results of Raman spectroscopy indicate that the degree of oxidative corrosion of nanocrystallites in filler particles is much less than that of nanocrystallites in the binder. However, filler particles are more graphitized and have fewer crystal defects and active sites, resulting in lower chemical reaction rates, but the interiors of filler particles are denser and the diffusion of oxygen preferentially follows the outer edges of filler particles and the interior of the binder. Therefore, the binder is preferentially oxidized and it is difficult for oxygen to diffuse between the nanocrystallites inside filler particles. This is why the size of the filler particle nanocrystallites does not decrease following oxidation. Filler particles are gradually oxidized and eroded, mainly on the outer graphite surface. This analysis is consistent with our intuitive judgment based on the SEM micrographs.