This project at Energy BioSystems is a classic example of how biotechnology projects have been sold to investors who don’t understand the limitations of bioprocessing. Based on initial laboratory results, and without a full understanding of the pathway, the company rushed to build a $1 million pilot plant and lined up investors. Then the reality of the limitations of the technology set in. By the time I joined the company, it was struggling to figure out how to make a process work that required a supply of organic carbon to generate the reducing equivalents necessary for the biotransformation pathway. They had also learned that converting dibenzothiophenes into phenols was not conducive to maintaining the health of expensive biocatalyst. Ultimately, the company spent $80 million with, by the end, only a well-elucidated biodesulfurization pathway and a handful of patents to show for it.

This was my first startup experience and this, plus the hands-on experience with a variety of process technologies were valuable lessons. This was also my first introduction to the SuperPro Designer^® process simulation modeling software that I now use regularly in my consulting work.

Microbiology, Molecular Biology & Biochemistry

Energy BioSystems had an excellent research group and by the time I joined the company, they had cloned and sequenced all of the pathway genes and characterized all of the enzymes. As a newcomer to Texas, I took every opportunity to get out and explore the state on weekends. I always carried a supply of sterile sample tubes with me and whenever I came across an oil well, I gathered a sample of the surrounding soil. Handed to the molecular biologists on Monday, they had usually found genes from the pathway in the sample by Friday.

Although the model compound, dibenzothiophene, was readily desulfurized by our biocatalysts, crude oils and their fractions contain a wide variety of alkylated dibenzothiophenes (and alkylated benzothiophenes as well). These proved to be much harder to desulfurize. This was one of the technological limitations of the process. We had an excellent chemistry group that were experts in characterizing these compounds. However, we were never able to determine if the slower rate of conversion was the result of enzyme substrate specificity, slower mass transfer from the oil to inside the bacterial cells, or a combination of the two.

The microbiology was another limitation of the process. The optimum growth conditions for the Rhodococcus erythropolis biocatalyst that the company had developed were near neutral pH and around 30°C. Unfortunately, these are also the optimum growth conditions for many other microorganisms as well. To be economical, a biodesulfurization process must operate continuously for long periods of time. Doing so with this organism in an oil refinery was effectively going to be impossible. Realizing this, I began to catalog successful continuous fermentation processes (there are not many) and the combinations of selective pressures that they use. I then worked to educate the company on these realities and recommended a path forward based on developing a practical biocatalyst and process with multiple selective pressures.

Biodesulfurization Pilot Plant

My work at Energy BioSystems started with managing the first pilot plant and designing a second generation pilot plant based on best practices in both bioprocessing and oil refining. This work included developing the design basis for the pilot plant, preliminary drawings and a total installed cost estimate. However, as the reality of the situation set in, it soon became apparent that there needed to be a better understanding of the process before the second generation pilot plant was built.

Hydroxyphenylbenzene Sulfonates Kilo Lab

One means of improving the economics of any process is to maximize the value of all by-products of the process. Both the traditional oil refining and the biorefining industries have learned this lesson and modern refineries of both types let very little go to waste. In the case of biodesulfurization, the problem of accumulating toxic phenols was solved by the scientists by blocking the final step in the pathway that removed the sulfinate group from the alkylated hydroxyphenylbenzene sulfinate intermediates. The process then accumulated these water-soluble sulfinates. Sulfinates, in turn, are readily oxidized to sulfonates and alkyl sulfonates have value as anionic detergents.

While the chemists worked on derivatizing the hydroxyphenylbenzene sulfonates and identifying potential markets for these by-products, my group set up a kilo lab to make enough product for trials. This process performed the biotransformation in a 250 liter fermentor using diesel fuel as the starting material. We then used a stacked disk centrifuge to separate the spent biomass and desulfurized oil from the aqueous phase containing the hydroxyphenylbenzene sulfinate (HPBS). This was cleaned up through a hollow-fiber membrane filter and then passed over an anion-exchange resin to remove the HPBS. This was eluted as a concentrate in the sodium salt form. To convert the sulfinates to sulfonates (HPBSO₃), we contracted with a local specialty chemical manufacturer to perform a hydrogen peroxide oxidation.

We then developed a multistage liquid-liquid extraction system using a series of bench-top fermentors as CSTRs to convert the HPBSO₃ to the acid form, extract this into an organic solvent, and then neutralize the concentrated acid with concentrated NaOH to produce a highly concentrated sodium salt solution. This was then dried to produce the final product.

Process Simulation Modeling

Eventually, as the company downsized, I became the keeper and developer of the SuperPro Designer^® process simulation models. These models were initially developed by one of my colleagues, and the approach to developing the models by the group is an excellent example of how to develop and calibrate simulation models. The SuperPro software has built-in equipment and installation cost models, but these appear to be based on pharmaceutical stainless steel equipment purchased in New Jersey. As such, the prices were not representative of equipment costs in oil refineries in Houston. To calibrate the base model, we worked with our engineering company partner to develop a traditional total installed cost estimate. These costs were then used to adjust the SuperPro equipment and installation cost models so that the base cost model from SuperPro matched the engineering cost estimate. This gave us a cost model that could be used to accurately predict capital costs as we developed the process.

As the development effort continued, I used the SuperPro model to quickly evaluate alternative designs. In a matter of just a few minutes, I could insert or remove a piece of equipment and recalculate the entire plant capital and operating costs. This is where process simulation models have the most value: Used early in the process development process to direct the research and development effort. Unfortunately, in this case, the models also made it very clear that, combined with the technical limitations of the process, biodesulfurization as we understood it was not going to be economically viable.

More Reading

Pacheco, M. A., Lange, E. A., Pienkos, P. T., Yu, L. Q., Rouse, M. P., Lin, Q., & Linguist, L. K. (1999). Recent advances in desulfurization of diesel fuel, p. 1–26. In National Petroleum and Refiners Association, Annual Meeting, NPRA AM-99–27. National Petroleum and Refiners Association, San Antonio, Tex.