2017 Volume 57 Issue 6 Pages 970-977
Mechanical activation study was carried out to enhance the oxidation roasting of the LD converter slag. XRD, SEM/EDS, particle size, specific surface area and thermal gravimetric analysis methods were used to determine the effect of mechanical activation. It has been shown that mechanical activation leads to the decrease of particle size, the increase of surface area, and the disorder of the crystal structure. TGA results reveals that mechanical activation not only can decrease the final oxidation temperature by 50 to 100°C, but also significantly decreased the apparent activation energy.
Mechanochemistry is a branch of chemistry that is concerned with chemical and physicochemical changes of substances of all stages of aggregation due to the influence of mechanical energy.1) It is an effective method to improve the reaction rate and low down the temperature in metallurgical industry. It has been reported to accelerate the oxidation of ilmenite concentrated ore,2,3) stone coal4) and molybdenite.5)
Vanadium bearing slag is the dominant sources (accounts for 58%) in the production of vanadium.6,7) Separating vanadium form vanadium bearing slag contain a series of processes, such as oxidation roasting, leaching and purification. Oxidation roasting is regarded as the key process under the processes presently employed to extract vanadium from vanadium bearing slag, especially in sodium salt roasting process8) and calcium roasting process.9) The calcium roasting process has an advantage over sodium salt roasting process because it forms less corrosive gases and little environment pollution. The calcium roasting process has a direct relationship with the subsequent leaching efficiency of vanadium.10,11)
In the past decades, a considerable number of investigations have focused on the optimization of calcium roasting process,8,12,13,14) including heating rates, roasting temperatures, roasting time, and the addition of additives. The roasting temperature has significant influence on the oxidation of vanadium bearing slag. The increase of roasting temperature can improve the oxidation efficiency and shorten the oxidation time. However, high roasting temperature leads to the sinter ring problem in the rotary kiln that seriously affects the production efficiency.15,16)
The mechanical activation of minerals is a non-equilibrium process, in which physical energy is transferred into a powder by crystalline damage, makes it possible to reduce their reaction temperature and apparent activation energy. Therefore, in the present study the utilization of high-energy ball mill for enhancement of oxidation properties of vanadium bearing LD converter slag and its influence on the particle size, specific surface area, XRD pattern and morphology. The transformation mechanisms and non-isothermal kinetics of the slag oxidation were also investigated.
The slag sample was got from the basic oxygen furnace process of Panzhihua Iron and Steel Group Corporation (Sichuan, China). It was ground with the size finer than 125 μm. The metallic iron was separated from the fine slag by using a magnetic separator and the non-magnetic part of the slag was employed in this study. The chemical analysis of the non-magnetic part is given in Table 1.
A Retsch PM 100 planetary ball mill (Retsch, Germany) equipped with zirconium oxide container (100 mm inner diameter, 72 mm height, 500 mL volume) and balls (10 mm diameter) were employed for the mechanical activation of the slag. For each grinding trying, about 100 g slag samples and 500 g balls were subjected to dry milling in air atmosphere. Milling was carried out at 400 rpm for all the batch experiments. The samples were activated in the planetary ball mill for 10, 20, 40, and 80 min individual.
The oxidation experiments were conducted on a SETARAM SETSYS Evolution TG-DTA thermal analyzer (SETARAM Instrumentation, France). Each samples (about 100 mg) were heated in a dynamic air atmosphere (20 mL/min) from room temperature to 1050°C. The heating rate was maintained at 5, 10, 15, 20 K/min in the separate experiments.
The phases of samples were identified through X-ray diffractometry (XRD; Rigaku D/max 2500 PC, Japan) with Cu Kα radiation (λ = 0.154 nm, 40 kV, 150 mA) at a scan rate of 0.3°/s. The line broadening analysis of the X-ray diffractograms and determination of structure and lattice parameter were carried out using MDI Jade 6.0 software. The particle size was analyzed with a laser diffraction particle size analyzer (Mastersizer 2000, Malvern Instruments Ltd., UK), and BET surface was determined by a micropore physisorption analyzer (ASAP 2020M, Micromeritics Instrument Corporation, USA) using N2 as analysis gas. The morphology and micro-constitution of the samples was carried out using a field emission scanning electron microscope (JSM-7800 FEG SEM, JEOL) equipped with an Oxford EDS.
The effect of milling time on the particle size distribution of the slag is represented in Fig. 1 and the characteristic value is summarized in Table 2. The unmilled slag has a large particle size in the range of 1.7–125 μm with a d50 of 29.42 μm, d90 of 123.39 μm and volume mean diameter D[4, 3] of 46.23 μm. With 10 minutes milling time, the particle size slightly decreased with a d50 of 11.43 μm, d90 of 51.37 μm and volume mean diameter of 26.59 μm. Longer milling time shows a narrowing of the particle size distribution in the range of 0.04–20 μm and volume mean diameter significantly decreased to 3.67 and 0.93 μm for 20 and 40 minutes milled slag, respectively. Further extending milling time to 80 minutes, the particle size distribution dramatically widened in the range of 0.04–45 μm and the volume mean particle size increased to 5.27 μm as the agglomeration of the fines.
Particle size analysis of slag with different milling time.
The BET surface area of the slag is shown in Table 2; the unmilled slag has a relatively low surface area about 0.2 m2/g. While the surface area sharply increased to 1.64 m2/g with 10 minutes milling time, then gradually increased to 3.29 m2/g with the extending milling time to 80 minutes.
The SEM micrographs of samples after different times of activation are shown in Fig. 2. The unmilled sample shows irregular shape particles with a more angular texture and a large range of particle size. The particles become fine and uniform as the time of activation increased, as evidenced in the SEM image of the sample with 40 minutes milling time.
SEM micrographs of samples (A) unmilled (B) 40 min milled (C) 80 min milled.
The XRD patterns and the major minerals present in the samples with different time of milling are presented in Fig. 3. The main phase in unmilled sample is ulvospinel (Fe2TiO4), vanadium spinel ((Mn, Fe)(V, Cr)2O4), and fayalite (Fe2SiO4). There are no new peaks observed during activation indicating that no phase changes occurred during activation, whereas the intensity of XRD peaks significantly reduced, the line broadened and shifted. The decrease in intensity of diffraction lines is the result of decrease in crystalline phase. The shift in diffraction lines is due to the uniform strain.
XRD of the unmilled (0 min), 10 min, 20 min, 40 min, and 80 min activated samples.
The variation of crystallite size and strain with time of activation was calculated from the line broadening of the vanadium spinel ((Mn, Fe)(V, Cr)2O4) using Scherrer’s formula:
Where Bt represents the full width at half-maximum intensity of the peak, λ is the wavelength of the radiation used, θ is the diffraction angle, D is the average crystallite size, ε is the strain, and B0 is the instrumental line broadening. The variation of crystallite size, strain and lattice parameters with time of milling are shown in Table 2. The crystallite size decreased from about 125 to 71 nm with the increasing milling time to 80 minutes whereas the strain increased from 0.1 to about 0.58%. The unit cell parameter (cubic structure) decreases from 0.850 to 0.845 nm with 10 minutes milling time, then increases to 0.848 nm with the extending milling time to 80 minutes.
The TG thermograms for the samples with different times of activation are given in Fig. 4. It shows that, in unmilled sample, the value of weight nearly remain stable below 300°C, then gain at a relatively low rate in 300–600°C, and then gain at a fast rate in 600–900°C. However, the weight gain always at a fast rate for all the milled samples in 300–900°C. Moreover, all the samples reach the maximum value of weight gain around 1000°C. Characteristic temperature and weight gain in the TG curves are summarized in Table 3.
Plots of TG versus temperature: (A) 5 K/min, (B) 10 K/min, (C) 15 K/min and (D) 20 K/min.
T1, T2 and T3 (°C), first, second, and third peak temperatures in the DTG curve; Tf (°C), final temperature in the TG curve, Wt (%), the maximum weight gain.
For unmilled sample, the temperature of maximum weight gain increased from 984 to 1036°C with the increase of heating rate from 5 to 20 K/min. However, the maximum weight gain occur nearly at the same temperature, around 934°C in samples with 20, 40 and 80 min of activation. The average maximum weight gain is 7.03 Wt.% in the whole roasting process for the unmilled sample. The maximum value of weight gain decreasing with the increase of the time of activation: 6.88 Wt.% gain in sample with 20 min of activation, 6.82 Wt.% in samples with 40 min of activation and 6.79 Wt.% in sample with 80 min of activation. This may be in due to the slightly oxidation of the samples as the local overheating during the long time activation. Furthermore, mechanical activation can improve the reaction activity and low down the reaction temperature. As shown in Fig. 4, the maximum oxidation rate also decreasing with the increase of the time of activation as the changes of crystallite size and strain of the milled samples.
Figure 5 shows the DTG thermograms for the samples with different times of activation. It was observed that the oxidation of samples contains several reactions and these reactions overlapped together. The DTG curves for unmilled sample only have one obvious peak and the maximum weight gain centered at 826, 837, 836 and 846°C at 5, 10, 15 and 20 K/min, respectively (as shown in Table 3). An interesting phenomenon is observed for milled samples. The DTG curves for milled slag consists of two or three peaks and the maximum weight gain centered around 600°C. The DTG peaks become more obvious with the increase of activation time. The peak temperatures of the DTG curve increasing with the increase of heating rate. This behavior can be attributed to heat-transfer problems between the sample and instrument.
Plots of DTG versus temperature: (A) 5 K/min, (B) 10 K/min, (C) 15 K/min and (D) 20 K/min.
As shown in Fig. 6, taking heating rate of 10 K/min as an example, mechanical activation can significantly low down the oxidation temperature in the same conversion degree. The oxidation temperature directly decreased when activation time increased from 0 min to 20 min. However, the temperature slightly decreased when activation time further increased to 40 min and 80 min. These non-linear changes of temperature and activation time were related to the non-linear changes of the surface area of the samples.
Conversion degree isograms versus activation time and oxidation temperature at heating rate of 10 K/min.
For a liner heating, the heating rate β is usually a function of time.18)
The original mass gain versus temperature curves obtained at constant heating rate were transformed into the degree of conversion (α) versus temperature curves by means of the following equation:21)
Both model fitting and isoconversional methods can obtain kinetic parameters. In contrast, the isoconversional methods are quite extensively employed to obtain a reliable kinetic description of the investigated process. The isoconversional methods avoid indistinguishable fits or mathematical expressions with high correlation and compensation effect made by model-fitting methods.22,23,24)
Kissinger-Akahira-Sunose (KAS) method25) is quite extensively employed to determine the activation energy in non-isothermal kinetic analysis. This method gives more accurate apparent activation energy (Ea) as compared to Ozawa-Flynn-Wall method.21,26) Thus, in this work KAS method is used which is originally recommended by Coats and Redfern.27) The equation is expressed as follows:
Where Tα is the absolute temperature (K) at a particular α, Eα is the activation energy (KJ mol−1) at a particular α, G(α) is the integral form of the kinetic function. At a particular α with different heating rates, the plots of ln(β/Tα2) against 1/Tα gives a straight line. The slope of the corresponding line is equal to −Eα/R that gives the value of apparent activation energy at corresponding α value. This method is rather accurate because it does not include any mathematical approximations.23)
The plots of ln(β/Tα2) against 1/Tα according to Kissinger method (Eq. (7)) were shown in Fig. 7. We have calculated the Eα values for α values varying 0.1–0.9 with a step of 0.1 and found the dependencies of Eα vs. α. The variation of activation energy with conversion degree using this method is shown in Fig. 8. The values of activation energy of unmilled sample has a fast increment from144 kJ/mol (at α=0.1) to 191 kJ/mol (at α=0.2) then slightly increased to 222 kJ/mol (at α=0.5) and then dramatically increased to 360 kJ/mol (α=0.8), finally decreased to 337 kJ/mol (at α=0.9). However, the variation of activation energy is almost same in all the activated samples. It was observed that the overall curve of activation energy shifts towards a lower value with the increasing time of activation. The values of activation energy are increased from 101 kJ/mol, 95 kJ/mol and 89 kJ/mol (at α=0.1) to 219 kJ/mol, 200 kJ/mol and 197 kJ/mol (at α=0.7) and finally decreased to 207 kJ/mol, 182 kJ/mol and 155 kJ/mol for samples with 20 min, 40 min and 80 min of activation, respectively.
Determination of activation energy by Kissinger-Akahira-Sunose method: (A) 0 min activated, (B) 20 min activated, (C) 40 min activated, (D) 80 min activated.
Apparent activation energy versus conversion degree by Kissinger-Akahira-Sunose method.
As shown in Fig. 8, the unmilled sample shows maximum activation energy (360 kJ/mol) required for the oxidation. The activation energy for the oxidation of activated samples decrease to much lower values, viz. 219 kJ/mol for 20 min activated sample, 200 kJ/mol for 40 min activated sample and 197 kJ/mol for 80 min activated sample. The nonlinear relationship of activation energy with the conversion rate indicates that oxidation of LD converter slag is a multi-step kinetics process in all the cases.28) Moreover, the multi peaks in the DTG curves also demonstrated that. For this multistep reaction kinetics, it may consist of many reactions and has several plausible reasons that are discussed later.
The XRD patterns of the unmilled sample and 80 min activated sample roasted from room temperature to different temperatures at a same heating rate of 10 K•min−1 are shown in Figs. 9 and 10, respectively.
XRD of the unmilled sample after roasted from room temperature to 400, 500, 600, 700, 800 and 900°C at a heating rate of 10 K/min.
XRD of the 80 min milled sample after roasted from room temperature to 400, 500, 600, 700, 800 and 900°C at a heating rate of 10 K/min.
After heating to 400°C, the main phase in both samples are still fayalite (Fe2SiO4), vanadium spinel ((Mn, Fe)(V, Cr)2O4) and ulvospinel (Fe2TiO4), and only ferrous oxide (FeO) in these phases oxidized to magnetite (Fe3O4) and hematite (Fe2O3). The intensity of fayalite, vanadium spinel and ulvospinel peaks are significantly decreased with the raising of temperature and hematite increased. The peaks of hematite become apparent at 600°C for unmilled sample while at 500°C for milled sample. A large amount of pseudobrookite (Fe2TiO5), small amount of manganese vanadium oxide (Mn2V2O7), rutile (TiO2) and anorthite (CaAl2Si2O8) occurred as new phases when the temperature increases to 700°C for the unmilled sample while only 600°C for the milled sample. The appearance of pseudobrookite is due to the following reaction:
The appearance of rutile can be explained by the decomposition of pseudobrookite as follow reaction:
The vanadium pentoxide reacts with manganese oxide generates manganese vanadium oxide as follow:
Further increased to 900°C, the peak intensity of hematite and pseudobrookite become stronger due to the complete oxidation of vanadium spinel, ulvospinel and fayalite. Meanwhile, the peak intensity of the silicon oxide significant increased as a result of reaction:
However, V2O5, CaV2O6, Ca2V2O7 and Ca3V2O8 were not found in all of the XRD patterns due to their low content or poor crystallinity.
The plausible pathways of the oxidation of vanadium bearing LD converter slag are shown in Fig. 11. The particle size of the milled slag is much smaller than the unmilled slag, as described in the previous section. The SEM and EDS analysis of the unmilled and 80 min milled sample are shown in Fig. 12 and Table 4, respectively. As shown in Fig. 12, the unmilled sample consists of several different kinds of mineral phase with different colors. These mineral phases are cohered together. According to the EDS analysis results (Table 4) it is known that the light color area at points a, c and d in Fig. 12 is vanadium spinel, the dark color at point b is the olivine. The mean particle size of vanadium spinel varied between about 10 and 40 μm. This also can be gotten from the micrographs of the vanadium slag made by other researchers.14,29)
Plausible pathways of the oxidation of unmilled slag and milled slag.
The micrograph of the unmilled and 80 min milled sample.
However, the particle size distribution of unmilled slag varied between 1.7 and 125 μm. Undoubtedly, most of vanadium spinel will be surrounded by the olivine and this will brings about barriers and difficulties in the oxidation of the wrapped vanadium spinel. As shown in Fig. 11, the outmost olivine will be oxidized first; the coating will be decomposed and exposed the surface of vanadium spinel, then is the oxidation of the internal vanadium spinel.
However, the mean particle size of milled slag (about 1 μm) is much finer than the size of vanadium spinel in the unmilled slag. The vanadium spinel and olivine are separated and rarely cohered together (as shown in Fig. 12). The unreacted-core shrinking model can describe the oxidation of the olivine and vanadium spinel. Most likely, the oxidation process will be controlled by the surface chemical reaction and internal diffusion as follow equations:30,31)
(1) Mechanical activation effectively contributed to decreasing the mean particle size of converter slag from 46.2 to 0.93 μm and increasing the specific surface area from 0.2 to 3.3 m2/g.
(2) Mechanical activation causes the decrease of diffraction lines intensity, broaden and shift of diffraction lines.
(3) Mechanical activation provides a relatively faster oxidation process and prompts the overlapped reactions to be more independent, which can approximately decrease the final oxidation temperature from 1000 to 930°C.
(4) The maximum apparent activation energy of oxidization process for unmilled, 20 min, 40 min and 80 min activated slag was evaluated to be 360 kJ/mol, 219 kJ/mol, 200 kJ/mol and 197 kJ/mol, respectively.
(5) The main phases of primary slag were fayalite, vanadium spinel and ulvospinel, the main phases changed into hematite, pseudobrookite, silicon dioxide and vanadium compounds after roasting.
(6) The mechanical activation pretreatment can destroy the structure of vanadium slag, drop the oxidation temperature, improve the oxidation rate, and shorten the oxidation time.
The authors are indebted to Mr. Zhiming Yan for his assistance in SEM testing of this project, as well as to the Natural Science Foundation of China (Grant No. 51404047 and Grant No. 51234010) and Basic and Frontier Research Program of Chongqing (Grant No. cstc2014jcyjA50011) for sponsoring part of this work.