NEC Semiconductors: microelectronics wastewater reclamation

Afterwards the wafers are placed in a constantly overflowing deionised (DI) water bath and then onto a second rinse stage before moving onto the next stage of production. The DI water can be either hot or cold, depending on the acid bath temperature, and is drained separately from the bench before being pumped to the treatment facility.

Transistor gate sizes can be smaller than 2 pm and as such any ions or the particles remaining on the silicon wafers can cause short circuits. Consequently the DI water quality standards need to ensure high-purity water (Table 5.16). Raw water supply to the production facility is high in organics due to the moorland intake and highly variable due to limited treatment at the local water treatment works. Consequently the potable water intake requires a large number of treatment processes prior to entering the DI production facility (Fig. 5.24). In fact, the front end of the works is similar to an advanced potable water works with coagulation-DAF and dual media depth filters to remove solids and activated carbon to remove organics and chlorine. Following the GAC is an anionic organic scavenger resin bed, a cartridge filter and then finally a reverse osmosis plant. The efficacy of the GAC in removing chlorine is essential to protect the resin and RO membrane from oxidation. The organic scavenger resin is regenerated on site with brine and caustic solution and contains two streams allowing for maintenance and regeneration.

The water then enters the DI plant which is split into primary and polishing stages. In the primary stage the water passes through a sequence of cationanion- cation ion exchange beds, a 10 pm cartridge filter to remove resin and precipitated organics and a 254 nm UV plant. Following the UV stage the water is filtered through a 3-stage RO plant arranged in a 7:3:2 array. The permeate is de-aerated prior to being pumped to the polishing stage of production. In the final stage the water passes through a cooler, 185 nm UV and IJF membrane filtration plant before being put into the supply ring main. Total production of ultra pure water is 200 m3 h-’ at an overall recovery rate of 71%. In comparison to the intake raw water the DI waste water is very low in organics and solids and so requires relatively less treatment to produce a water of sufficient quality to enter the primary DI production stage. The main difference between the water sources is the very low cation and bacteria count in the reclaim water (Table 5.17). This is to be expected as the reclaim water is generated from the DI baths, which follow high-purity acid baths. Importantly, as the baths are situated in the clean rooms no contamination is picked up from the atmosphere. The reclaim water does, however, contain much higher concentrations of sulphate from sulphuric acid remaining on the wafers. An important aspect of the scheme is that the use of reclaimed water does not affect product quality in any way. To ensure this, reclaimed water is monitored in terms of TOC and conductivity prior to being pumped to the primary DI production stage. Any waste above the allowable levels is rejected and passes to the effluent treatment plant. The plant was originally designed to accept water below 300 pS cm-l but has had to increase this to 600 pS cm-l to increase the amount of water through the reclaim plant.

Description of plant

The reclaim water is essentially deionised water with sulphate anions added: no cation removal is therefore required and the process flow sheet reflects this (Fig. 5.25). Reclaim water is pumped at a rate of 140 m3 d-l through a 14 m3 activated carbon bed (diameter = 2.8 m, height = 3.05 m) for the removal of hydrogen peroxide which can be present in some of the baths. This is necessary to protect the downstream processes from oxidation. After the adsorption bed the water passes through a 2.2 m diameter weak anion exchange bed with a WBV of 5.2 m3. The anion bed removes free ions of sulphate, phosphate and fluoride. The anion resin is regenerated with a caustic solution after a preset number of bed volumes have been treated or the pH falls below pH 6. The water then passes through a 10 pm filter to remove any resin beads or activated carbon particles. An UV lamp operating at a wavelength of 18 5 nm then sterilises the water and removes any trace organics remaining in the flow. The final stage of the treatment is a reverse osmosis plant consisting of 8 inch x 8 inch modules each fitted with four elements and arranged in a 5:2:1 array. The plant operates at a 90% recovery generating 130 m3 h-' of treated reclaim water.

Performance

Water enters the reclaim plant with a conductivity of 600 pS cm-l. The weak anion bed reduces the conductivity to 2 5 pS cm-l which is then further reduced through the RO plant to a final value of 3 pS cm-l. The water then mixes with the treated raw water before being passed onto the DI production facility where water is ultimately produced with a resistivity of 18.2 MQ cmP2. Some 58% of the high-purity water is reclaimed and recycled to the ultrapure water plant. The remaining water is sent to the effluent treatment facility with the RO concentrate and the brine regenerate. Under these conditions 54% of the production flow is generated from recycled flow.

Included in the reclamation strategy is the utilisation of waste heat through a heat exchange network. Hot reclaim drains are used to preheat the supply of high purity DI. This represents a huge energy saving by increasing the cold DI from 2 5 to 5 5°C and has a dual benefit as it cools the wastewater prior to being treated in the reclaim plant.

There is a significant financial saving from using reclaim water as 83% of the treatment cost is associated with bringing raw water on site and then disposing of it (Table 5.18). The reclaim water costs only 12% that of treating raw water generating a saving of 20% on the total operating and maintenance budget for the DI water treatment plant. The main economic decision is concerned with determining the amount of reclaim water to recycle. Zero reclaim incurs high operational costs due to raw water price but 100% reclaim incurs excessive additional up front capital investment. Between the two extremes is an economic optimum, which depends largely on the utility cost of the local region (Fig. 5.26). In order for the reclaim percentage to increase the cost curve needs to change such that the capital cost in relation to the raw water/discharge costs will decrease. The major barrier to this is the rapid changes that can occur in microprocessor production which means new chemical pollutant can be generated which the original plant is incapable of treating.

Discussion

The case studies featured have not only demonstrated that reuse of industrial waste water is technically feasible but economically viable. In all cases the installation of a membrane system for recycling water has resulted in an overall saving to the company involved. Annual savings have ranged from $102 282 to $642240 (Table 5.19) and have paid back the initial capital from within 8 months (automotive) to 6 years (power). The most cost-effective of the schemes has been the automotive plant (internal loop) which generates an annual return on investment of 147% equating to a profit of $2259 m-3 d-l. The least profitable of the schemes has been the power station at Eraring (external loop) which generates an annual return on investment of 19.2% equating to a profit rate of $1 71 m-3 d-l.

The savings have been generated from a number of sources in addition to the direct benefit of using less external (potable) water. In the case of the power stations, recycling has decreased the number of regenerations required in the demineraliser plant either reducing operating costs (Eraring) and/or increasing capacity (Flag Fen). In other cases recovery of raw materials has increased the profit margins, such as paint (Germany) and pulp solids (Chirk). Perhaps a less obvious saving has been in energy savings by recycling hot water streams and so reducing the heat requirement at the site (Livingston, South Wigston, Apeldoorn).

However, the driver for reuse is not always directly a financial one. In the majority of case studies outlined here the original interest in reclamation was generated from an indirect financial driver such as changes to legislation (textile, food) or a need to secure sufficient water supplies. For instance, in the case of the textile plant, legislation imposed a treatment requirement on the plant. Once money had to be spent to comply with the legislation, the benefits of ensuring the water could be recycled became important as otherwise the capital expenditure could not be recovered. The regulation requirements can go as far as zero liquid discharge, in which case reuse is a necessity rather than an option. In such cases, the driver is to reduce overall treatment costs (Doswell). However, in a number of the schemes the driver has been directly one of reducing operating costs of which water supply can be a major component (Eraring, Livingston).

The selection of membranes in the reclamation process train has occurred in a number of ways. In some industries either membrane technology is already used (Germany, Livingston) or the plant involves similar levels of technology (Eraring). At other schemes the use of membranes has been a radically new development (South Wigston). The familiarity with the technology appears to be in part linked to the need for indirect financial drivers to exist before reuse is considered. This is probably because a high degree of confidence is required in the technology designed for reclaiming thc water, since it cannot be allowed to adversely affect core production quality.

As expected it is difficult to draw commonalities from a broad range of industries. However, a number of points can be concluded. The key facet of the reclamation system in all cases has been its ability to withstand variations in the wastewater quality, whilst producing a water quality suitable for reuse. EMuent quality from membranes is usually very good, such that the main concern is achieving sufficient throughput without incurring excessive cost. This is reflected in the level of pretreatment required from the different schemes. In cases where simple closed loops are being generated the pretreatment requirement is minimal (as in the automotive industry). However, in situations such as the reuse of secondary effluent and other wastes with high fouling propensities, more involved pretreatment is required. In some cases this involves two membrane stages and in other cases more traditional pretreatment (such as coagulation followed by depth filtration).

Overall, the case studies have shown the suitability of membrane technologies in particular for industrial effluent recovery and reuse. The ability to produce reclaimed water of sufficient quality is dear. However, the throughputs are quite different between the schemes. For instance, comparing the specific fluxes of the four RO schemes described reveals a range between 0.56 and 3.63 LMH bar-' reflecting the differences in the RO feed water matrices. This demonstrates that each scheme is in part unique, potentially involving problems that have not been encountered in other industries. Moreover a common problem with potential industrial reuse schemes is a paucity of data describing the water quality and hence the design limits, Table 5.20 clearly illustrates this point where in some cases little or no water quality data are known in actual operating schemes. Although easily remedied, data paucity remains a major barrier to uptake of not only membranes for reuse but any treatment technology.