Supercomputing our energy (Day 261)

Supercomputing our energy (Day 261)

12th February 2015

But processing and modelling large amounts of data to help our understanding of complex and mammoth tasks like the formation of the universe, predicting weather patterns, or large and complex engineering problems require more than the average desktop computer.

Hence, the growth of supercomputers in recent times. But they don't come cheap.

Later this year the UK's Met Office £97 million (US$ 146 million) supercomputer will come online.

Eventually, its processing power will be 16 petaflops - meaning it can perform 16 quadrillion calculations every second.

The "Cray XC40" machine will have 480,000 central processing units or CPUs, which is 12 times as many as the current Met Office supercomputer, made by IBM.

At 140 tonnes, it will also be three times heavier - more a 'floortop' than a desktop.

There are lots of other eye-catching projects using supercomputers including Durham University's Eagle Project, which is powerful enough to re-create the entire universe.

Other universities are also investing in supercomputers with the potential to help the chemical and process industries. One recent project includes research to help improve the efficiency of biofuel and petroleum processing.

The work has been led by researchers at the University of Minnesota, who have identified potential materials that could improve the production of ethanol and petroleum products.

Petrochemical and biofuel refineries use materials called zeolites that act as molecular sieves to sort, filter, and trap chemical compounds.

They also catalyze chemical reactions necessary to produce and upgrade fuel and chemical feedstock from petroleum-based and renewable resources.

There are more than 200 known zeolites and hundreds of thousands predicted zeolite variations. The key to improving biofuel and petrochemical processes is to find which zeolites work best.

However, one of the problems is that synthesizing novel zeolites in the lab is a long, complicated process that can take many months each.

Researchers from the University of Minnesota and Rice University developed a complex computational screening process that can look at thousands of zeolites in the virtual world and identify their performance for specific applications.

To analyse all the known and predicted structures would take decades.

Instead, researchers from the University of Minnesota and Rice University have used a supercomputer to develop a complex computational screening process that can look at thousands of zeolites in the virtual world and identify their performance for specific applications.

This reduces the need for trial and error experimentation in the lab.

The team have used Mira - a supercomputer with nearly 800,000 processors - which has enabled them to run in one day the equivalent computations requiring about ten million hours on a single-processor computer.

As a result, the researchers found a few all-silica zeolites with superior performance that contain pores and channels with the ability to accommodate ethanol molecules but to shun hydrogen bonding with water molecules.

One of these zeolites - which was synthesized and tested by chemical engineering and materials science professor Michael Tsapatsis - was found to be so effective that it could change the ethanol/water separation process from a multi-step distillation process to a single-step adsorptive process.

Similar zeolitic materials could also have possible applications for separations in the biofuels and petrochemical industry.

I'm sure we will see more breakthroughs like this as a result of the huge processing power of modern computers. Congratulations to everyone involved.


The full research paper: “Discovery of optimal zeolites for challenging separations and chemical transformations using predictive materials modeling” is available on the Nature Communications website.