Characterization of Individual Chemical Reactions Consuming Acid During Nuclear Waste Processing at the Savannah River Site

Legacy nuclear waste generated at the Savannah River Site (SRS) during production of enriched uranium and plutonium during the Cold War is currently being processed in the Defense Waste Processing Facility (DWPF) into a stable borosilicate glass waste form for long term storage. The majority of the legacy waste is stored either as mixtures of hydroxides and hydrous oxide insoluble solids or as precipitated salt cakes in large cylindrical storage tanks at SRS. These 3.5-4.5 thousand cubic meter (900,000-1,200,000 gallon) carbon steel storage tanks also contain 5-7M sodium solutions rich in hydroxide, nitrate, and nitrite anions. Batch feed preparation for DWPF involves washing the aqueous phase of slurries obtained from one or two individual waste tanks to about a 1M sodium concentration.

DWPF brings approximately 25 cubic meters of fresh washed sludge slurry into the Sludge Receipt and Adjustment Tank (SRAT) at a time. DWPF employs batch processing. Acids are added to adjust the rheology, dissolve some of the alkaline earth and transition metal compounds, convert HgO to elemental mercury for steam stripping, destroy some of the existing anions such as carbonate and nitrite, and chemically reduce a portion of the MnO₂ to Mn²⁺. (Thermal reduction of Mn in the DWPF melter leads to reduced melter throughput and can also lead to foam formation.) Optimal performance of the DWPF waste glass melter requires a balance of chemical oxidizers and reductants in the feed in order to avoid foam formation and metal precipitation. The balance is achieved by dividing the SRAT acid demand between nitric acid (oxidizer) and formic acid (reductant).

The original concept for DWPF involved a dual feed capability that would process insoluble radioactive waste sludges in parallel with soluble radioactive salt waste cake material. The DWPF flowsheet was subsequently modified to substitute 90 wt% formic acid for the dilute formic acid stream coming from salt waste processing. An acid addition strategy was derived based on what was currently known:

moles acid/L slurry = (titrated base) + 2*total carbonate + 0.75*(nitrite ion) + 1.2*Mn + Hg

All right-hand-side quantities are measured for the slurry (aqueous phase plus insoluble solids) and converted to moles per liter slurry units. The base equivalents term represents the equivalent hydroxide molarity of the slurry determined by titration to pH 7.

This approach to predicting acid requirements was initially seen as an interim measure that would not be needed once “short-term” issues with salt waste processing were resolved. DWPF, however, continues to operate in the “interim” mode fourteen years later. Supporting experimental work done during that time has shown that the above equation predicts the acid demand lower than observed by 20-30%.

Changes in the preparation strategy of the legacy waste sludges over the past decade were made to reduce the volume of decanted wash water, since storage space and evaporator capacity are limited. The resulting DWPF feed sludge compositions require proportionally more acid to process. This doctrinal evolution has constrained the quantity of acid that can be added relative to the minimum required to accomplish various processing goals that were part of the original design concept. Tightened constraints have renewed interest in understanding the nature and timing of the chemical reactions that consume acid. Predicting and controlling the quantity of acid added in the SRAT is an essential part of the DWPF operational strategy for minimizing noble metal catalyzed hydrogen generation from excess formic acid.

Another issue with the acid equation (despite the fact the equation predicted low) was the potential for double counting certain species. For example, since measurement of hydroxide ion concentration was problematic in the system, a slurry titration to pH 7 was substituted. A total inorganic carbon measurement was used to get the carbonate concentration of the slurry. Soluble carbonate, however, was partially titrated to bicarbonate during the titration measurement and was being counted twice for acid demand. Experiments showed that it was only soluble carbonate, and not insoluble carbonate, that was being titrated and counted twice in the current equation. There was also a concern that HgO might be dissolved during titration. Experiments indicate that mercury was not appreciated dissolved during pH 7 titrations, although titrations to pH 5.5 were being impacted.

Two other findings emerged from the preliminary experiments. Different titration techniques produced different equivalent base molarities for the slurry. DWPF used 20:1 to 25:1 diluted slurry in their automatic titrator, while some work within the Savannah River National Laboratory used direct titrations. Results by the two methods were typically within 5-15%, however, and this has been deemed acceptable. Obtaining a reliable measurement for slurry total inorganic carbon, however, has been a bigger issue. Uncertainties of reported quantities have been high. Therefore, finding methods to predict the stoichiometric acid requirement that did not require the measurement of slurry TIC or the titration to pH 7 were promoted over a strategy that only sought to refine some of the coefficients in the existing acid equation.

A matrix of large lab-scale tests were completed in duplicate with an integrated sampling plan that permitted the chemistry of the SRAT during acid addition to be elucidated and better quantified. Data from these tests filled gaps in the historical database. Reactions were identified that were actively consuming acid. The dissolutions of Mg(OH)₂ and CaCO₃ were successfully tracked as functions of pH during processing using simulated wastes. These reactions were found to occur below pH 7 and at rates comparable to half of the acid addition rate (expected, since Mg²⁺ and Ca²⁺ need two H⁺ to neutralize their anions). They also were contributing significantly to the total demand. Other reactions were found to consume acid, but the concentrations were too small to have a significant impact on the calculated acid consumption. Equilibrium models predict most of the insoluble carbonate will be found associated with calcium.

Careful analysis of the by-products of nitrite destruction led to an improved average coefficient (three parallel reactions) that was 33% higher than in the original equation. The following second generation stoichiometric acid equation was proposed to explain the data:

moles acid/L slurry = (titrated base) + supernate carbonate + (nitrite ion) + 1.5*Mn + 1.5*Ca + 1.5*Mg + Hg

Soluble carbonate was given one mole acid per mole, since titration accounts for the second mole needed to convert carbonate to carbon dioxide. Calcium and magnesium were brought in at 75% of the stoichiometric requirement for Mg(OH)₂ and CaCO₃, since there exists a certain fraction of non-acid reacting compounds, such as oxalate, phosphate, sulfate, etc. associated especially with calcium and perhaps magnesium, but ultimately these terms are accounting for the insoluble carbonate contribution so that no measurement of slurry TIC is required. Preliminary validation experiments show that this equation fits the historical database much better than the original equation, while still retaining the first-principles premise that the chemistry can be understood in detail as a sum of individual reactions.

A second, less direct, approach was taken toward evaluating the stoichiometric acid demand during SRAT processing. This approach led to an equation that contained neither titration nor TIC measurements. The premise was that a sum could be made over the soluble cations present after acid addition/consumption that would represent the cations that had had a basic anion neutralized by H⁺. It was recognized immediately that this sum would overstate the acid demand, since not all of the associated anions were bases. Therefore, a credit for non-acid reacting anions was taken to reduce the result obtained by summing the soluble cations. This equation and the second generation equation track each other well while responding more appropriately to compositional changes than the original stoichiometric acid equation.

The new data on the chemistry of the major species present during SRAT processing has led to a greatly improved understanding of the sequence of reactions occurring and their impact on acid consumption. This has led to the development of two new equations for predicting the stoichiometric acid requirement for SRAT processing. The two new equations have already been implemented for lab-scale testing in parallel with the original equation. The plan is to attempt to implement a change in DWPF at the time of the next major sludge batch (within six months).