Purification and utilization of a formerly incinerated sodium nitrite bearing wastewater stream

3.1 Isolation of sodium nitrite containing minimal organic contamination

The conventional workup of the batch reaction, shown in Scheme 1, involved extraction of the reaction mixture with dilute sodium hydroxide, to afford a dark aqueous phase with a pH of 13–14, containing approximately 22 wt% of NaNO₂ and 3.0–5.0 wt% organic material (~30,000 ppm TOC), the vast majority of the organics being the compounds shown in Figure 1.

It was then found that extraction of the reaction mixture with the requisite amount of deionized water afforded a 40 wt% aqueous solution of NaNO₂ with a pH of 8–11, containing 3500 ppm to 6000 ppm of TOC; the majority of the TOC being the PTC and the balance primarily being the compounds shown in Figure 1. A second extractive purification of the solution of bisimide in organic solvent with dilute caustic, produced a wastewater stream containing significantly less organic contamination, with only traces of NaNO₂ (<1000 ppm). It was was found that this secondary wastewater was amenable to inexpensive conventional wastewater treatment processes that did not involve incineration.

The comparison of the nitrite bearing wastewater from workup of the reaction with dilute caustic vs. water is shown in Table 1. The 40% NaNO₂ aqueous solution containing 3500 ppm to 6000 ppm TOC was chosen as the material for evaluating alternative remediation processes, primarily because of the much lower amount of organic contamination present [3]. A plant trial was done, wherein the reaction shown in Scheme 1 was extracted with water, instead of dilute caustic, to afford large amounts of the desired ~40 wt% sodium nitrite with ~5000 ppm TOC.

3.2 Biotreatment of contaminated ~40 wt% aqueous NaNO₂ contaminated with 5000 ppm of TOC

A two-stage, anoxic/aerobic, biotreatment process of this material was piloted. The contaminated 40 wt% nitrite solution was diluted with water to provide an aqueous solution that contained ~1000 ppm nitrite. This was then continuously fed to a tank containing activated sludge (biomass), held under anoxic conditions at a pH of 8.2–8.6, to reduce the nitrite to nitrogen. Methanol was also continuously fed to the anoxic tank as a food source for the biomass. The effluent from the anoxic section was then continuously fed to a tank that contained biomass held under aerobic conditions, to hopefully mineralize organic material to carbon dioxide. The effluent of the aerobic tank was continuously fed a clarifier to settle the biomass, and the clear water effluent of the clarifier was filtered through sand. Conditions were found to mineralize the nitrite to nitrogen, however, it was found that most of the organic compounds were recalcitrant and were not destroyed in the anoxic or the aerobic sections of the pilot plant. Simple biotreatment was not an option to treat the raw wastewater. However, biotreatment could be feasible if the organics could first be removed [3].

3.3 Nanofiltration purification of contaminated ~40 wt% aqueous NaNO₂ contaminated with 5000 ppm of TOC

We explored nanofiltration as a way to remove the organic impurities from the contaminated 40 wt% nitrite, to produce a filtrate that was amenable to the denitrification biotreatment process, or to produce a commercial grade product. Commercial sodium nitrite is typically sold as a 40 wt% aqueous solution (or as a dried solid), containing <100 ppm of chloride, with a pH of 8–10. Nanofiltration involves forcing liquid containing soluble contaminants across a membrane under pressure. The membrane rejects soluble organic contaminants, while allowing monovalent ions to pass, to hopefully afford a clean permeate. Typically, membranes reject high molecular weight organic material; we did not know if lower molecular weight organics would be rejected.

A lab membrane vessel was used to evaluate 10 commercially available nanofilters. It was found that a DESAL-5 DK 2540 nanofilter and the LCI MPS-34 nanofilter were the most efficient. The DESAL-5 membrane was constructed of a polymeric flat thin film composite in which a polyamide selective layer was supported on a polyethersulfone substrate [4]. The MPS-34 was of proprietary nature.

A pilot skid was constructed to evaluate both nanofilters (Figure 2). A 115 l tank was filled with the ~38 wt% aqueous nitrite containing ~5000 ppm TOC at pH 11. The material was pumped at 11 l/min at 28°C, with a low pressure pump to a high pressure pump, and thence to two in-series housings containing either DESAL-5 DK (GE Osomics, Minnetonka, MN, USA) or LCI MPS-34 (Koch Membrane Systems, Wilmington, MN, USA) filters, at 400 psig. The permeate was collected (100 l, ~87% of the initial feed) in a vessel and analyzed for TOC and nitrite content, leaving ~15 l (~13% of the initial feed) of concentrated membrane reject; the results are shown in Table 2. It was found that the LCI membrane produced permeate at 40 ml/min, whereas the DESAL-5 DK membrane produced permeate at 400 ml/min. It was also found that the TOC in the permeate increased with higher feed temperatures and higher feed pressure to the membranes. However, both filters failed to afford a product that could easily be biotreated, because of the residual low molecular organic material still present. No commercial outlet had been identified for the permeate product that contained the residual organic material and considerable amounts of chloride, resulting from the PTC used in the bisimide forming reaction (Scheme 1). Additionally, there was still a considerable amount of membrane rejected material that required some sort of treatment. This process option was not pursued [3].

Figure 2

Nanofiltration purification skid mounted pilot unit.

3.4 Supercritical treatment of ~40 wt% aqueous NaNO₂ contaminated with 5000 ppm of TOC

A high-temperature, high-pressure technology developed by SRI was investigated [5]. In the SRI process, sodium nitrite is added to a wastewater containing organic contaminants and then fed to a vessel at 2500 psig to 3000 psig, at 350°C to 400°C. At these temperatures and pressures, water exists in a supercritical state, and the nitrite effectively oxidizes the organic material in situ to sodium carbonate and sodium halide, if organic halides are present. The goal was to produce a material that could easily be biotreated to mineralize the nitrite without having to contend with the recalcitrant organic material, or to identify a market for the thermally treated product.

In the laboratory, the 40 wt% aqueous NaNO₂ containing ~5000 ppm TOC was diluted with water by a factor of 10 and then fed with a high pressure pump to a 1.6 mm inner diameter metal tube that was coiled with an overall length of 300 cm, contained in a high temperature oven maintained at 375°C. The exit of the column was connected to a back pressure control valve to maintain 400 psig in the column. The exit of the valve was connected to a cooling coil to condense the effluent to afford a liquid at room temperature. The waste feed flow rate was adjusted to provide a residence time of the material to be 2–10 min at high temperature and pressure. It was found that after treatment under these conditions, the effluent contained <20 ppm of unknown TOC with the formation of a stoichiometric amount of sodium carbonate.

The contemplated commercial process design for the process is shown in Figure 3. The process would involve pumping the raw contaminated 40 wt% nitrite to a heat exchanger, to preheat the material, which then enters a tube and a shell reactor, wherein the outside of the tubes are heated with molten salt. The molten salt would be heated electrically or in a direct fire heat exchanger (not shown in Figure 3). The feed may be diluted with water to better control the process; this aspect of the process requires further development. The residence time of the material in the reactor would be adjusted to achieve the desired destruction of the organic material in the feed. The effluent of the reactor would go to a separator, to allow any oxides of nitrogen (NOx) or organic vapor to separate from the liquid stream and then to be cooled in a cooler, before going across a back pressure control valve needed to keep the reactor at high pressure. The gases would then be destroyed in a vent incinerator. The liquid effluent of the separator would flow to the heat exchanger to preheat the feed to the reactor, then to a cooler and then across another back pressure control valve, needed to keep the reactor under high pressure. The effluent would go to another separator to allow any organic or NOx vapor to separate from the liquid stream, and this gas stream would also flow to the vent incinerator. Finally, the liquid effluent would go to an evaporator, to produce the sodium nitrite product of the desired wt%. The evaporator would be required should the feed to the reactor be diluted with water and it was intended that a 40 wt% NaNO₂ product was to be marketed.

Figure 3

Process flow diagram of a high temperature and pressure destruction unit.

The cost to build this organic destruction commercial unit, diagramed in Figure 3, to ultimately produce a product of ~40% NaNO₂ containing ~5 wt% Na₂CO₃, was estimated to be >$15,000,000 ($15 MM), not including a biotreatment facility to denitrify the product. A 40 wt% aqueous NaNO₂ commercial product, containing 5 wt% Na₂CO₃, was surmised to have limited applicability in the market place. This option was not pursued after consideration of the cost to build, safety aspects of the high pressure process, and perceived lack of market outlets [3].

3.5 Molten salt oxidation of ~40 wt% aqueous NaNO₂ contaminated with 5000 ppm of TOC

We explored the idea of removing the water from 40 wt% aqueous NaNO₂ containing ~5000 ppm TOC, and then melting the material (melting point of NaNO₂ is 271°C) whereupon the nitrite would oxidize the organics. Again, the goal was to produce a material that could be easily biotreated, by virtue of the absence of recalcitrant organics, or to identify a market for the treated product. Alkali metal salts of nitrite and nitrate, when mixed with organic materials, are explosive materials when heated, and it is the ratio of fuel to oxidant that dictates the reactive properties. Dry mixtures of sodium nitrite containing 1, 2, 3, and 9.5 wt% TOC were prepared. The TOC spiked to the samples was composed of the organic material present in the wastewater. These samples were then studied by a variety of methods; differential scanning calorimetry (DSC), accelerated rate calorimetry (ARC), thermogravimetric analysis (TGA), the Fauske Reactive System Screening Tool (ARSST) [6], dedicated tube propagation (DTP) test [7], and the hot block test for energetic events. Sodium nitrite containing 1 wt%, 2 wt%, and 3 wt% TOC showed self-heating at 260°C in the ARC, but did not decompose uncontrollably. All the other analyses showed that these mixtures were otherwise benign. ARC, DTP, the hot block test, and ARSST analysis of the NaNO₂ mixture containing 9.5 wt% TOC showed this composition to be highly reactive.

Having completed the safety evaluations, 40 wt% sodium nitrite containing 5000 ppm TOC was evaporated to dryness and then heated to 420°C for 3 h with stirring on a 50 g scale, in a 100 ml, 3-necked, round-bottom flask, equipped with the means to maintain a nitrogen atmosphere and a water cooled condenser open to the atmosphere. Oxides of nitrogen were evident in the outlet of the condenser. Some organics volatilized from the reaction mixture and condensed on the inner walls of the condenser. The heat treated material was cooled to afford a white solid. Analysis of the material by ion chromatography, TOC, and titration, showed that the product was composed of ~93 wt% sodium nitrite, ~1 wt% sodium nitrate, and ~6 wt% Na₂CO₃, and contained <20 ppm TOC. High pressure liquid chromatography (HPLC) showed the absence (limit of detection of 500 ppb) of any of the organic compounds present in the initial contaminated material.

A pilot plant was built to fully develop the technology (Figure 4). A 115 l electrically heated vessel was charged with ~23 kg of solid commercial grade sodium nitrite and melted at 420°C. The molten salt was agitated. The 40 wt% aqueous sodium nitrite containing 5000 ppm TOC was slowly added to the molten salt, whereupon the water and trace organic material distilled from the vessel and was condensed in a water cooled condenser. The gas from the condenser was sent to an NOx furnace, which converted the NOx to nitrogen. The condensate from the condenser was treated with carbon to remove the trace organic material present. The addition of the feed to the heated vessel continued until ~115 kg of product had accumulated in the vessel. The molten product was cooled to ~300°C and then slowly added to water to ultimately afford a 40 wt% solution of sodium nitrite containing ~5 wt% Na₂CO₃ and ~1 wt% NaNO₃ containing <20 ppm TOC. The pilot data was used to estimate the costs of building and operating a safe commercial molten salt unit. Ultimately, a commercial outlet for the material was not identified, the cost to build the facility was significant, and the safety aspects of the process were challenging. This approach was not pursued [3].

Figure 4

Molten salt oxidation pilot facility to treat contaminated ~40 wt% aqueous NaNO₂.

3.6 Carbon purification of the ~40 wt% aqueous NaNO₂ contaminated with 5000 ppm of TOC

Activated carbon treatment of the 40 wt% aqueous NaNO₂ containing 5000 ppm of TOC (organics shown in Figure 1) was investigated early on. A variety of carbons from Calgon Carbon Corporation, Tigg Corporation, and Cabot Norit Activated Carbon were evaluated. These carbons included coconut and wood-based carbon, lignite carbon, and coalbased carbon (virgin and reactivated). Each carbon was pulverized to afford finely divided powered material. Portions (100 g) of the 40 wt% NaNO₂ containing 5000 ppm TOC, described in Section 2.1., were shaken in plastic jars with 1 g, 3 g, 5 g, 10 g, and 15 g of carbon for 12 h. The carbon from each sample was filtered off, and the filtrates were then analyzed for TOC. The data obtained is referred to as carbon isotherm data.

The absorption capacity for the organic impurities of each carbon was determined by applying the Freundlich equation to the isotherm data [8]. This initial study showed that Calgon CPG carbon (virgin coal-based activated carbon) was superior, followed by DSR-A mesh size 40×5 (reactivated carbon), and Calgon Colorsorb G carbon (a wood-based carbon) performed the worst. Lignite carbon was also explored, but this type of carbon cannot be regenerated and must be landfilled after use, therefore it was removed from consideration [9]. The most cost effective carbon found was Calgon DSR-A reactivated coal-based carbon, the workhorse for water purification applications. Spent carbon returned to the carbon supplier is reactivated in a furnace at 900°C in the presence of a controlled amount of oxygen and steam. The carbon from the furnace is either fed to a cooling screw and screened to remove carbon fine particles (carbon fines), or water quenched to afford what is known as pool react carbon, which contains more carbon fines. Dry screened carbon is more expensive than pool react carbon. Carbon fines are an important consideration when handling carbon on a large scale. It was determined from the isotherm data that ~0.04 kg of Calgon reactivated DSR-A carbon was required to treat 1 kg of 40 wt% aqueous NaNO₂ containing 5000 ppm of TOC, to provide a filtrate that contained <50 ppm of TOC and ~1700 ppm of chloride. A pilot trial was designed to continuously feed contaminated 40 wt% nitrite bearing wastewater to a commercial carbon bed unit.

Additional material had to be collected from the plant for the pilot. Enough water was added to a batch of bisimide (Scheme 1) to afford an aqueous phase of 38% to 45 wt% sodium nitrite at 50°C to 80°C. The solubility limit of NaNO₂ in water at room temperature is ~45 wt%, and ~41 wt% at 0°C. The aqueous nitrite obtained contained ~4000 ppm TOC to ~5300 ppm TOC, the majority of the TOC being the PTC, with a pH of 11. The material had a viscosity of ~3 centipoise at 25°C. Analyses of the truck loads of material collected for the pilot are shown in Table 3.

Calgon advised that the contact time of the contaminated water with the carbon be at least 10 min/centipoise, meaning that that the contact time of the 40 wt% contaminated nitrite with the carbon should be at least 30 min.

The carbon purification process was piloted on truck load quantities to determine the quality of the material produced, to develop a market for the product, to determine the biotreatability of the treated material (should a market not be found for the product), and to determine the economic viability of the overall process. Dry Calgon reactivated carbon weighs ~465 kg per cubic meter; 20% of the space it occupies is taken up by the skeletal carbon, 40% of the space is between the carbon particles themselves, and the remaining 40% of the space is within the pores of the particles of carbon, important information to know when operating carbon beds. Two vessels that held 4536 kg of carbon were operated in series (Figure 5).

Figure 5

Simplified process diagram for the carbon purification of raw 40 wt% aqueous NaNO₂.

Calgon pooled reactivated carbon was slurry loaded into the carbon beds, and the clean water was drained from the carbon bed. The majority of the carbon fines were removed with this clean water, but some remained in the system. The raw contaminated 40 wt% aqueous nitrite was filtered as it was added to the carbon bed feed tank, with a low efficiency 10 micron bag filter to remove any suspended solids.

Each carbon bed was ~15,000 l in volume. The carbon bed level in each bed was approximately 0.6 m from the top tangent line of the tank. Each bed contained 4536 kg of carbon, 9.75 m3 of carbon, or about 9765 “liters” of carbon. The valve at the bottom of the lead bed (the first bed in series) was closed. The raw feed was pumped to the top of the first carbon bed at ~90 liters per minute (lpm).. As the lead bed was filled, a small amount of material was allowed to exit the vent of the bed into a tote, to ensure the bed was full. Once the first bed was full, the valve was opened at the bottom of the first bed to allow flow into the top of the second bed; the valve below the lag bed was closed. The second bed was allowed to fill, and a small amount of material was allowed to exit the top of the lag bed into a tote. The beds were then allowed to sit (soak) for about 2 h. The bottom valve on the lag bed (the second vessel in series) was opened and the raw contaminated 40 wt% aqueous NaNO₂ solution was pumped to the carbon beds at 225 lpm; purified effluent product was collected in a hold tank. Effluent samples from the first and second beds were taken at set intervals and analyzed for TOC, wt% NaNO₂, and signature organic compounds (HPLC), to monitor the capacity of the carbon.

The initial material from the beds was diluted by the water, initially on the carbon. After >50 ppm TOC was detected in the effluent of the lag carbon bed, the material in the beds was pressed out using nitrogen pressure and was recycled to the feed tank. The beds were then filled with water and allowed to soak; they were then drained to a separate tank (soak collection tank, Figure 5). This soak procedure was repeated. The combined soak material (~15 wt% NaNO₂) was concentrated to 40 wt% NaNO₂ in the concentrator (Figure 5) and sent to the raw carbon bed feed tank in subsequent piloting work. Carbon destined for regeneration must contain acceptable levels of residual sodium, and it was found that two water soaks of the carbon was necessary to meet Calgon’s acceptance criteria. The carbon was then transferred out of the bed with water into a truck and sent to Calgon for regeneration. The process is not economically viable or environmentally friendly if the carbon cannot be regenerated.

The water in the pores of the carbon diluted the first amount of product. Five trucks were fed to the beds and five trucks of product were collected. Approximately 94,800 kg (~37,330 kg of NaNO₂) of raw material was fed to the beds and 91,080 kg (33,035 kg of NaNO₂) were collected. The beds were run until known organics broke-through the lag bed. The trial showed that ~0.1 kg of carbon were necessary to purify ~1.0 kg of the feed, higher than the 0.04 kg of carbon per kg of feed determined in the lab. Analyses of the five trucks collected are shown in Table 4.

The last truck of material was analyzed by HPLC and did not contain any known organic constituent above its detection limit (detection limit of ~500 ppb). The color of the product was straw yellow, the last truck being a slightly deeper yellow; commercial grade material is straw yellow. The pH of the feed was ~10 to 11, yet the product had a bulk pH of 7. The contact pH of carbon was determined by stirring 25 g of carbon in 100 ml of water for 5 min and then determining the pH of the supernatant with a pH probe. The contact pH is a measure of acidic groups bound to the carbon. The contact pH of pooled reactivated carbon is typically 9 to 10.5, however periodically it can be as low as 6. It was speculated that the carbon used in this first carbon bed trial was 6 or 7. The vapor space of the carbon beds was tested for the presence of NOx but none could be detected. Our own experience has shown that sodium nitrite, when treated to pH 6 with an acid, generates nitrous acid, which decomposes to NOx when passed through carbon.

The product material was analyzed for trace metals by inductively couple plasma (ICP) spectroscopy, and common anionic impurities by ion chromatography (IC) (Table 5). Commercial material was also analyzed for comparison.

Calgon pooled reactivated carbon contains calcium and magnesium and it can be seen that this leached from the carbon beds into the product. A large amount of spent carbon that is reactivated, was formerly used at municipal wastewater treatment plants, where the carbon becomes contaminated with calcium and magnesium. The raw 40 wt% sodium nitrite feed to the beds contained <10 ppm of calcium and magnesium. The amount of chloride present in the product is considerably higher than commercial grade material, potentially limiting its use in the marketplace. Finally, the product contained much less nitrate anion than commercial grade material (~150 ppm vs. 5000 to 10,000 ppm).

The product material was sent to potential customers for evaluation. Unexpectedly, multiple applications were identified that could tolerate the presence of chloride in the product and trace levels of calcium. Within a short period of time, several customers were identified willing to purchase the entire amount of sodium nitrite produced as shown in Scheme 1 once carbon purified. The selling price is lower than the current suppliers. Customers are using the material for pigment and chemical intermediates manufacture (diazotization chemistry), and in wastewater treatment facilities.

The process was scaled up several more times with pooled Calgon React Carbon as described above, except that the concentrated carbon bed soaks were processed with the raw feed from the plant, and a couple of interesting observations were made. In the second (and subsequent 5 truck campaigns), the pH of the product, collected in two tanks, was higher and residual calcium was lower (Table 6).

The spent carbon regeneration process does not remove the calcium and other metals, and these metals can leach from the bed in the next life of the carbon. It was speculated that this is what happened in the first carbon bed trial. Subsequent experience has borne this out.

The carbon utilization for the second and third trials was better than the first; ~0.085 kg of carbon were necessary to purify 1.0 kg of feed, a 15% improvement in carbon utilization.

Carbon fines in the carbon-treated product are filtered out using one set of 4-micron nominal filter bags, followed by a set of Nylon 5u/1u/1u multilayer composite bag filters (Filtrations Systems, Florida, US). Removing all the carbon fines from pooled reactivated carbon when first loading them into a bed is difficult and one must be diligent to keep the fines out of the product.

Based on several pilot trials, the draft specification for the product was established (Table 7).

The process has been further optimized. The separation of the aqueous wastewater from the bisimide product phase (Scheme 1) has been improved, and has led to reducing the amount of TOC in the raw aqueous NaNO₂ from the plant. Raising the wt% of the raw contaminated nitrite isolated in the plant from 40 wt% to 45 wt% also resulted in less TOC being present in the wastewater. These improvements, and several others, have led to excellent carbon utilization (~0.035 kg of carbon per kg of feed). As of 2012, the entire nitrite bearing waste stream that was formerly incinerated, is now purified and sold commercially, and represents approximately 10% of the NaNO₂ demand in the US.

The cost savings to our business was in excess of $5 MM/year. The environmental impact of the carbon purification process was significant. Because of improvements of the process to harvest the sodium nitrite from the plant, not discussed here, the yield of the bisimide product was increased by 1%, saving $0.6 MM/year in raw material cost. The carbon footprint of the polyetherimide process has been reduced by 4500 metric tons/year of greenhouse emissions (since we no longer use natural gas to incinerate the waste), energy usage was reduced by 65,000 MM Btu/year (natural gas savings) and water usage was reduced by 72 MM liters/year (water formerly needed to cool the hot gas from the incinerator). The reduction of greenhouse emissions/year and Btu/year took into account the energy necessary to regenerate the spent carbon from the carbon purification process. The program was recently recognized by the National Pollution Prevention Roundtable (MVP2 Award 2011), and by the American Chemistry Council’s Responsible Care Program in 2012. In addition, the production of the polyetherimide resin, from which the nitrite wastewater emanates, continues to increase and will utilize the carbon purification process.