Tackling waste water with nanofiltration membranes

May 1, 2008
David Johnson, Kamla Jevons
Pharmaceutical Technology Europe
Volume 20, Issue 5

The pharmaceutical industry must address the release of nonbiodegradable APIs into the environment.

The pharmaceutical industry must address the release of nonbiodegradable APIs into the environment, and the objective of this study was to develop a solution for treating an aqueous process waste stream that contained APIs and an organic solvent. It was found that solvent-stable nanofiltration (NF) membranes displayed the potential to produce a filtered stream suitable for discharge to the conventional chemical treatment plant, as well as enabling a 70% waste stream volume reduction. The developed process resulted in the installation of a system that would enable future production expansion to take place without an increase in energy requirements.

The route for waste

Pharma's use of thermal oxidation is a well-established disposal route for waste streams containing organic solvents and APIs that cannot be inactivated by conventional chemical and biological waste water treatment techniques. Thermal oxidation is very effective for such waste streams, but it uses an uneconomical amount of energy for high water, low organic solvent and contaminant content streams. Because of high capital and operating costs, as well as environmental constraints, this study investigates and develops a solution based on membrane technology, which offers an energy efficient and 'green' technology process.

On the go...

A client's site that manufactures APIs by a multiple step chemical synthesis process, which results in waste streams of varying composition, was examined. The aqueous waste stream contains <2% organic solvent, mono and divalent salts, APIs and byproducts. The study objective was to reduce the waste stream volume by ~70% and permeate quality compliant with site discharge limits for the APIs.

Waste stream characterization dictated a requirement for a membrane that would be stable with the site process organic solvent and provide a high rejection of the ~650 MW API. Reverse osmosis (RO) membranes offer a significantly high rejection of the target molecule, but, unfortunately, current state-of-the-art technology for these membranes does not provide long-term membrane integrity in the presence of the organic solvent.

The characteristics of the SelRO1 NF MPS-44 hydrophilic membrane (Koch Membrane Systems; MA, USA) offered the potential for retention of the API and stability with the organic solvent.

NF separations with complex streams are difficult to predict or model without experimental data. A series of membrane filtration tests was required to establish proof of concept, and the results provided confidence to embark on pilot site trials to obtain the data required for production scale-up of the custom-designed waste treatment system. In this work, the approach to trials and some of the results are presented.

Methods and material Process stream characterization. The process waste stream from the API synthesis reactor sequences was collected in the aqueous waste tank (Figure 1). Initial proof-of-concept test work conducted trials on grab samples to establish API rejection. Pilot trials were conducted with composite samples for process development trials.

Figure 1: Crossflow flat sheet test cell.

Proof-of-concept trials. A flat sheet test cell was used to verify the initial rejection profile of the key components with the SelRO MPF-44 NF membrane at 10–35 bar operating pressure (Figure 2). The membrane sample was soaked in concentrated feed material for 12 months to establish membrane integrity.

Figure 2: Pilot trials plant.

Spiral wound MPS-44 2540 style membrane elements were operated on a crossflow pilot plant in a batch concentration mode of operation. Trials were conducted on grab samples taken from the process waste stream to acquire indicative membrane performance with respect to rate of permeation. Parameters that can impact the membrane performance, such as feed stream composition, test sample pH and operating pressure 10–35 bar, were investigated. API rejection at operating pressure 10–35 bar was also explored. On each process trial run conclusion, a clean in place (CIP) regime with alkaline and acid solutions was instigated to verify that the membrane had achieved an effective clean. Samples were collected during the trials for analysis of the key components.

Site pilot trials. Site pilot trials were performed on 24 h composite samples. Feed material of the composite sample was derived from a metered split stream from the site aqueous waste stream storage tank.

A total of 25 trial runs was performed. The trials programme was designed to establish flux variation with volume concentration at natural and modified pH, as previous trials had indicated flux improvement at slightly acidic conditions. From these trials, batch variability and run time implication on membrane performance was established.

Target molecule rejection had indicated a potential requirement to reprocess the permeate with the NF membrane to ensure 100% confidence that the site discharge consent limits could be met. Hence, trials were conducted to verify flux performance and API rejection with permeate derived from the primary NF process.

During the trials, permeate reprocessing and data collection were essential because online quality measurements of the API were not possible, and the analytical data of the collected samples would only be available some time after the conclusion of the trials.

All NF system CIP wash liquors were reprocessed to minimize the volume of material sent for thermal oxidation and trials were conducted to verify treat ability when blended with the main feed stream.

As the production-scale plant would operate on a 24/7 basis, the trials had to demonstrate a stable performance of the NF membrane during a 22-h processing period, and the remaining 2 h had to be adequate to conduct the CIP regime.

The exact organic solvent levels in the complex waste stream were difficult to analyse and, hence, quantitative evaluation of impact on membrane performance was not established. Visually, it was apparent that the stream contained a varying organic solvent concentration backed by site process operation information.

Membrane stability verification. Membrane elements used during the trials were exposed to the feed material and CIP solutions for up to 6 months. As stated previously, flat sheet membrane was soaked in a concentrated feed material for ~12 months. During the phases of the test programme, a salt rejection test was employed to establish any major variation in membrane rejection characteristics.

On conclusion of the trials, the test elements were autopsied to ascertain the condition of the membrane and element construction materials. The flat sheet membrane was tested to confirm its flux performance and sucrose rejection after the soak test.

Discussion

Data presented in this article are limited because of client confidentiality, but the NF membrane solute transport mechanism in aqueous and nonaqueous streams has been studied and eminently reported in the literature.2–5 However, the results presented in this article show that NF membrane characteristics, feed stream composition and operating parameters all contribute to the target molecules rejection and permeability rates. The pilot trials were devised to evaluate all anticipated NF operating parameters that could impact on permeability rate, molecule rejection and achievable volume reduction.

The <2 % solvent levels in the process stream witnessed during the pilot trials did not appear to majorly effect the membrane performance.

Figure 3: Flux versus volume concentration factor (VCF) with pH adjustment.

Initial trials had shown relatively low permeation rates of 2–4 L/m2/h (land feed stream natural pH of 9.4–9.7). Adjustment of the pH from 6 to 7 had a dramatic influence on the permeation rate and process volume concentration (Figure 3). Adjustment of the feed stream natural pH from 9.7 to 3.6 altered the API solubility in the aqueous/organic solvent mixture, displaying an improved flux. However, at pH values <6, API transmission through the membrane increased dramatically (Figure 4). Temperature and pressure was not major contributor to the API rejection.

Figure 4: Effect of pH on API rejection. Run 9 T24 decant.

Feed stream temperature does have a major implication on flux performance and the production-scale plant's required facility for stream temperature maintenance of ~30–35 °C, especially during winter months (Figure 5).

Figure 5: Flux versus temperature.

Waste stream limitation on volume reduction of 65–70% is caused by API and other dissolved solids level of the stream, limiting practical flux.

The double pass NF system provides a high level of security, ensuring that permeate APIs are at undetectable levels and remain below the site established standards (Figure 6).

Figure 6: Flux versus volume concentration factor (VCF) - high feed dissolved solids first pass NF.

The trial membrane autopsy highlighted the importance of waste stream prefiltration to 100 μ removal rating. This is essential to ensure that extraneous particulate material does not damage the NF membranes.

Conclusion

The key variables studied enabled the process modelling and design of the production-scale system. Operating hydrodynamics influence on membrane performance followed predictable behaviour for most of the parameters.

Rejection of the API and membrane permeability at a range of pH values is variable. One of the major contributing factors is API solubility and interaction in the mixed stream as the API's structure and concentration levels significantly limit achievable volume reduction.

Experience of NF membrane system design, operation and understanding membrane separation processes contributed to successful trials suite development, enabling relevant data collection. The optimization and monitoring of the production scale system operation will continue. Further studies are planned to determine the final system design and to evaluate the long-term operating experience of the NF system.

Kamla Jevons is European Business Development Manager at Koch Membrane Systems (UK). She has more than 20 years' experience in industrial application of membrane separations and new process developments in the biopharmaceutical industry.

David Johnson is Market Development Manager at Koch Membrane Systems. David has a broad experience in membrane technology and the application of membrane filtration to provide industrial process and wastewater treatment solutions.

References

1. SelRO is a registered trademark of Membrane Products Kiryat Weizman Ltd.

2. J.M.K. Timmer, "Properties of nanofiltraion membranes; model development and industrial application," PhD Thesis, Technische Universiteit Eindhoven, The Netherlands (2001).

3. www.ivt.uni-linz.ac.at/homeE/pages/research/publications/pdf/Behaviour.pdf

4. J. Geens et al., J. Membr. Sci., 281(1–2), 139–148 (2006).

5. M. Mänttäri, A. Pihlajamäki and M. Nyströ, J. Membr. Sci., 280(1–2), 311–320 (2006).