Journal of the Ceramic Society of Japan 116 [6] 712-716 2008 Paper Corrosion behavior of reaction bonded Si3N4–SiC and SiAlON– SiC composites in simulated aluminum smelting conditions Mark I. JONES,† Ron ETZION,* Jim METSON,* You ZHOU,** Hideki HYUGA,** Yu-ichi YOSHIZAWA** and Kiyoshi HIRAO** Department of Chemical & Materials Engineering, University of Auckland, 20 Symonds St, Auckland, New Zealand *Light Metals Research Centre, University of Auckland, 23 Symonds St, Auckland, New Zealand **National Institute of Advanced Industrial Science & Technology, AIST, 2266-98, Shimo-Shidami, Moriyama-ku, Nagoya 463-8560, Japan Si3N4–SiC composites, which are widely used as sidewall refractories in aluminum smelting cells, have been produced by reaction bonding and their corrosion performance assessed in simulated aluminum electrochemical cell conditions. Additional samples were produced with the silicon nitride bonding phase replaced by β SiAlON with compositions ranging from Z = 1–4 in Si6–zAlzOzN8–z. The formation of the Si3N4 and SiAlON bonding phases were studied by reaction bonding of silicon powders in a nitrogen atmosphere at low temperatures to promote the formation of silicon nitride, followed by a higher heating step to produce β SiAlON composites of different composition. The corrosion performance was studied in a laboratory scale aluminum electrolysis cell where samples were exposed to both liquid attack from molten salt bath and corrosive gas attack. For the Si3N4 bonded samples, the corrosion resistance was shown to be strongly dependent on the environment during corrosion testing, with samples in the gas phase showing higher corrosion than those immersed in the bath. Samples that had been pre-soaked in the bath and then tested in the gas phase showed the highest corrosion due to the combined effects of bath penetration and gas attack. For the SiAlON bonded samples, the corrosion results showed similar trends but were complicated by the presence of a strongly adherent layer on the samples which influenced volume measurements. This layer is believed to be alumina and appears to form through an interaction with the SiAlON and the bath components, since no such layer was observed for the silicon nitride samples. ©2008 The Ceramic Society of Japan. All rights reserved. Key-words : Silicon nitride, SiAlON, Silicon carbide, Composites, Corrosion, Aluminum smelting [Received February 15, 2008; Accepted April 17, 2008] markedly.7) One of the most widespread uses of RBSN in molten 1. Introduction metal handling is in the Hall–Heroult cells used for the produc- Reaction bonding is a low cost, low temperature alternative tion of Al metal. In this process, alumina is electrolytically heating technique for producing silicon nitride ceramic materials reduced into molted aluminum, whereby the bound oxygen in the where the nitride phase is produced by reacting metallic silicon alumina reacts with carbon electrodes to form carbon-dioxide in a nitrogen atmosphere at temperatures of around 1400°C. SiA- gas and metallic aluminum. The electrolyte is mainly cryolite, lON materials can be produced in the same manner by first pro- Na3AlF6, with dissolved alumina and additions of other fluorides ducing the nitride phase and then subsequently increasing the such as AlF3 and CaF2 which serve to reduce the melting point temperature to where the nitride reacts with the Al and O usually of pure cryolite (1009°C) down to a cell operating temperature provided through additions of alumina and aluminum nitride. of between 920 and 980°C.8) The electrolytic cells are lined with Compared with conventional sintering of silicon nitride powders, a refractory material to protect the cell and maintain the correct Reaction Bonded Silicon Nitride (RBSN) has several advanta- heat balance. Originally made from carbon, these sidewall lin- geous features including low sintering shrinkage, low raw mate- ings have now been widely replaced by Si3N4–SiC composites rial cost and high strength retention at elevated temperature.1)–3) produced by reaction bonding. In these composite materials the However, the fact there is little or no shrinkage associated with RBSN, typically 20–30 mass% acts as the bonding phase holding reaction bonding means that unless sintering aids are used the the silicon carbide particles together. The use of these materials resultant materials tend to have porosity of around 20–30%.4) allows thinner sidewalls therefore giving increased cell capacity Whilst therefore not suitable for high performance applications, and the ability to accommodate larger anodes therefore improv- RBSN ceramics find widespread use as a refractory material in ing productivity.9) During cell operation, the sidewall is protected metal handling applications due to properties such as good ther- by the presence of a layer of solidified bath, but if this “frozen mal stability and thermal shock resistance, good chemical corro- ledge” becomes disrupted by anode effects or instability, the sion resistance and lack of wetting by many metals and alloys.5),6) refractory is directly exposed to aggressive chemical environ- Typical applications include blast furnace refractories, where ments both from liquid metal and the cryolitic electrolyte bath, RBSN materials have increased lifetimes of furnace linings and from the gas phase above the bath.10) In these environments, materials show different levels of degradation depending on † Corresponding author: M. I. Jones; E-mail: mark.jones@auckland. manufacturer but also on location within the cell.9) The objec- ac.nz tives of this work are to produce controlled compositions of both 712 ©2008 The Ceramic Society of Japan Journal of the Ceramic Society of Japan 116 [6] 712-716 2008 JCS-Japan Si3N4–SiC and β SiAlON–SiC composites by reaction bonding and study their corrosion resistance in an environment that sim- ulates that observed in Al smelting. 2. Experimental procedure 2.1 Sample preparation The starting materials for the Si3N4–SiC composites were an abrasive grade silicon carbide (< 40 μ m GC #400, Fujimi Kenmazai Kogyo Co., Ltd., Japan) and a fine, high purity silicon powder (< 1 μ m, > 99% Kojundo Chemical Lab. Co., Ltd., Japan). For the SiAlON–SiC composites, Al2O3 (AKP–50, Sumitomo Chemical Co., Ltd., Japan) and AlN (Grade E, Tokuyama, Corp., Japan) were added in appropriate amounts to produce Si6–zAlzOzN8–z with Z values ranging from 1 to 4. In these materials the oxygen content of the nitride powders was taken in to consideration when designing the compositions, and in all cases the compositions were designed so as to give com- Fig. 1. Experimental rig used for corrosion testing. posites with 25% by weight of the bonding phase (Si3N4 or SiAlON) assuming complete nitridation of the Si powder. The powders were ball milled in methanol using a silicon nitride pot and balls, dried in a vacuum evaporator and passed through a 250 2.3 Characterization μ m sieve. The powders were uniaxially pressed in to plates with The density of the reaction bonded samples was determined dimensions of 47 × 42 × 12 mm under a pressure of 40 MPa and using the Archimedes method in distilled water. The degree of then isostatically pressed at 200 MPa. Heating was carried out in nitridation was determined by changes in weight measured a graphite resistance furnace (High Multi–5000, Fujidempa before and after heating. Phase structure was assessed by X-ray Kogyo Co., Ltd., Japan) under a 0.5 MPa nitrogen atmosphere. diffraction (XRD) and microstructure was observed using both Pressed samples were located inside a double walled crucible optical microscopy and scanning electron microscopy (SEM). arrangement with an inner crucible of boron nitride and an outer Following corrosion tests, the samples were ultrasonically one of graphite. Nitridation was carried out by slowly heating the cleaned in an AlCl3 solution to remove loosely adherent bath and samples through the temperature range 1200–1450°C at 0.25°C the degree of corrosion was assessed as a change in volume, min–1. The SiAlON formation was studied at temperatures rang- determined by measurement of density and mass. ing from 1700–1900°C by subsequently raising the temperature 3. Results and discussion at 10°C min–1 to the desired temperature and then holding for a period of 2 h. 3.1 Characterization The phases present at different sintering temperatures, the 2.2 Corrosion testing degree of conversion of Si to Si3N4 and the relative density of the For the corrosion testing, samples of approximately 30 × 10 × sintered samples are all given in Table 1. The development of the 10 mm were cut from the sintered materials and subjected to final microstructure can be determined by following the change electrolysis conditions in a purpose built experimental cell. A in crystalline phases as a function of temperature. At 1380°C, α schematic of the experimental cell is shown in Fig. 1 and consists Si3N4 was observed for all samples but the remaining presence of a graphite crucible surrounded by a can made of inconel. The of crystalline Si indicates that the transformation was incom- electrolysis tests were carried out at a temperature of 1000°C for plete. At 1450°C the transformation was complete and the silicon 48 hours at a voltage of 4V and current of 15A. The molten bath nitride was present in both α and β crystalline forms. The relative was typical of that observed in industrial practice with a compo- amounts of the two Si3N4 polymorphs, determined from XRD 11) sition of 78% Na3AlF6, 10% AlF3, 7% Al2O3 and 5% CaF2. peak intensities using the method of Gazzara and Messier, Three different sets of sample were tested for each experiment; showed that the α phase was predominant with values typically the first set of samples were immersed directly in the bath and around 70% of the total Si3N4. For the Z = 1–4 samples at this subjected to attack by molten liquid.