Major Theme Stories
Professor CHENG Ka-wai Eric
Professor
Department of Electrical Engineering
Time for transport to bet big on ammonia
Inexpensive zero-carbon emission ammonia-fuelled power is a step close to joining the wheel and combustion engine as a transformative transport technology thanks to a world-first ammonia fuel-cell electric vehicle developed in three months by an interdisciplinary team at the Hong Kong Polytechnic University (PolyU).
Despite being touted as the future of passenger cars, hydrogen combustion engines or hydrogen fuel cell-powered motors have a number of issues that make them an impractical choice for powering vehicles, and most of today’s electric vehicles rely on lithium-ion batteries, which are expensive, bulky and hard to recycle. The combined use of ammonia and fuel cells has several advantages over both approaches.
“Hydrogen is highly explosive and needs to be stored under high pressure, which makes it expensive to handle and raises safety concerns, and some sources are not emissions free. Whereas ammonia is not explosive, can be stored at much lower pressure, is abundant and, crucially, is totally carbon neutral,” says Eric CHENG, Power Electronics Research Center Director and professor in PolyU’s Electrical Engineering Department, who leads the project.
Vroom, vroom
The proof of concept ammonia-powered fuel cell hybrid electric golf cart uses technology and processes the team developed, such as the Ammonia Energy Management System, Ammonia to Electrical Energy Management System, ammonia purification technique, and Energy Optimisation System to convert ammonia to hydrogen, which is then consumed by a proton-exchange membrane fuel cell (PEMFC).
The PEMFC converts the chemical energy produced during the electrochemical reaction of hydrogen and oxygen into electrical energy, in contrast to the direct combustion of hydrogen and oxygen to produce thermal energy.
“There’s no need for a large hydrogen tank in the vehicle as all the hydrogen is consumed almost immediately,” says CHENG.
Ammonia is safer than hydrogen—it is not highly flammable and is much less volatile—and it’s more energy-dense, which means it can be easily and cheaply transported, making it more suitable for long-distance transport.
For instance, hydrogen must be compressed, cooled, and transported to fuelling stations. It’s usually stored in vehicles at a pressure of 700 bars, while ammonia only requires 8 bars of pressure for storage.
“Refilling ammonia is easier to operate—it’s similar to today’s gas stations—and liquid ammonia is more suitable for storage in electric vehicles,” says CHENG. “The refilling process is faster than refilling hydrogen or charging a battery.”
Furthermore, ammonia is cheaper than hydrogen, as it is already used in many industrial applications and therefore infrastructure and production processes are already in place. It can be a cleaner fuel than hydrogen, as it emits zero greenhouse gases and can be produced with renewable energy sources.
Ammonia-fuelled vehicles emit nitrogen and water vapour, which are naturally occurring components of the atmosphere and safe.
Innovation through collaboration
The team’s multidisciplinary approach covering physics, applied physics and electrical engineering was crucial to achieving the breakthroughs necessary to build a unified configuration for the cart.
“We had to solve the problems of how to place the cracker converter in the vehicle, how to manage the energy flow, how to convert the ammonia, and how to temporarily store the power,” says CHENG. “The key challenge was putting all the innovations together in a single technology.”
While some of the technologies involved are well-established, it was the team members’ different perspectives that enabled the project to move forward quickly.
“A chemical engineer may work on converting ammonia into hydrogen, but they wouldn’t be in a position to readily relate that to electrical engineering, for instance,” adds CHENG. “When our diverse team members talked with each other, we rapidly came up with multidisciplinary solutions that fit together as a whole.”
The next stage of the project, which is funded by the Innovation and Technology Fund and supported by academia and industry partners including EV Dynamics Ltd., Oxford University, Automotive Platforms and Application Systems R&D Center and HKOXGI Ltd, is to develop a long-distance ammonia-fuelled bus. Work is progressing at pace and completion is expected Mid 2023 .
Professor Mo YANG
Associate Head and Professor
Department of Biomedical Engineering
Dr Siu Hong Dexter WONG
Research Assistant Professor
Department of Biomedical Engineering
Humanity had a lucky escape with COVID-19 — a more deadly microbe, like a mutated Ebola, Avian influenza, Marburg, or MERS-CoV virus, all potential causes of the next pandemic, could have produced a far higher death toll.
But given recent advances like messenger RNA (mRNA) vaccines, and the development of nanomaterial-based optical techniques to speedily detect SARS-CoV-2 at an early stage by PolyU’s Professor Mo YANG, Associate Head and Professor of Biomedical Engineering, and Dr Siu Hong Dexter WONG, Research Assistant Professor of Biomedical Engineering, the world is better prepared to tackle future infectious threats.
Testing times
Yang and Wong’s work on advanced biosensors, completed over a year during the COVID-19 pandemic, applies nanomaterials to nucleic acid detection, and promises to improve upon the performance of today’s testing gold standard, the polymerase chain reaction (PCR) method.
“PCR tests for COVID-19 usually require very complex handling processes in terms of the RNA harvesting, which involves laboratory work of four to six hours,” says WONG. “Our solutions’ greatest advantages are they have fewer steps, reducing the wait to around one hour, and lowering the risk of false positive signals. Also, they can detect the level of infection.”
Testing is a vital tool for managing pandemics as it enables the identification of infected individuals and helps prevent the pathogen’s spread. It provides information for public health officials to determine the outbreak’s extent and make informed decisions on how to prioritize resources.
Also, quickly identifying individuals who have been in close contact with an infected person and providing them with the appropriate care can limit the severity and spread.
Conventional nucleic acid detection techniques like PCR often require Förster resonant energy transfer (FRET), which relies on the transfer of energy from a donor (fluorophore) to an acceptor (quencher). However, high background signals can lower the sensitivity of these techniques.
Yang and Wong’s AIEgen/graphene oxide nanocomposite (AIEgen@GO)-based two-stage “turn-on” nucleic acid biosensor for the rapid detection of SARS-CoV-2 viral sequence deploys graphene oxide nanosheets to reduce background signals, thereby increasing sensitivity. When the AIEgen@GO meets target viral sequences, the AIEgen detaches from GO for the fluorescent recovery and is lighted up to achieve a dual-on mechanism to maximize the detection sensitivity.
"We want to decrease the background by applying these nanomaterials to raise sensitivity,” says WONG. "The overall improvement of these technologies is better capacity in terms of sensitivity.”
A key challenge in creating the platform was determining the best ratio between the graphene and the agent, which took several months to overcome. The result is a platform that “can be modified to detect any type of nucleic acid,” adds WONG.
New horizon
The second testing solution developed by Yang and Wong is the CRISPR-Cas12a integrated SERS nanoplatform with chimeric DNA/RNA hairpin guide for ultra-sensitive nucleic acid detection.
CRISPR-Cas12a is a more complex technology than the graphene oxide nano composite biosensor.
“It’s a nucleic acid detection technique, like PCR, but unlike PCR, it doesn’t require pre-amplification, which can be time-consuming and increase the risk of false positive signals,” says WONG.
CRISPR-Cas12a uses the principle of surface enhanced Raman scattering, a powerful optical technique, to detect target nucleic acids without amplification.
The newly developed platform utilizes nano satellites made of many small nanoparticles surrounding a bigger nanoparticle. The strong Raman signal between the nanoparticles is initially strong, but weakens as the CRISPR-Cas12a protein is applied to detect target nucleic acids.
However, nanomaterials impose steric hindrance to limit the accessibility of Cas12a to the narrow gap (SERS hot spots) among nanoparticles (NPs) for producing a significant change in signals after nucleic acid detection.
The new platform shows that specifically designed chimeric DNA/RNA hairpins (displacers) can be destabilized by activated CRISPR-Cas12a in the presence of target DNA, liberating excessive RNA that can disintegrate a core–satellite nanocluster via toehold-mediated strand displacement for orchestrating a promising “on-off” nucleic acid biosensor.
It also proves that applying displacers was more effective in decreasing the SERS intensity of the system and attained a better limit of detection (LOD, 1 aM) than by directly using activated CRISPR-Cas12a, with high selectivity and stability for nucleic acid detection.
“This level of detection from a nano platform is only 10 times less than PCR, but the overall processing time is around two hours,” says WONG.
The knowledge and techniques gained from these endeavors chart a path to the development of more rapid, sensitive, and inexpensive diagnostic methods.
Professor Ajay Kumar
Professor
Department of Computing
In March, 2022, a video appeared on Ukrainian national news showing President Volodymyr Zelensky instructing the country’s troops to surrender to Russia’s invading forces. Quickly debunked as fake, the damage could have been catastrophic for Ukraine it if had been believed.
As artificial intelligence technology advances, threats like this posed by photorealistic fake images and videos will only increase, threats that recent research conducted at PolyU’s Biometrics Research and Innovation Centre (BRIC) lays bare.
“There are techniques out there that make it extremely difficult, perhaps impossible, to detect fake images,” says Professor Ajay Kumar at PolyU’s Department of Computing. '“Our research cautions the forensic community on the reliability of widely employed GAN-generated fake image detectors, and the need for significant detection advances.”
Make-believe
Deepfakes are AI-generated images and videos that are designed to look and sound like real people. The technology behind them is advancing rapidly, and the implications of deepfakes are far-reaching and dangerous.
One example is the deepfake porn videos that were created of female celebrities, including Scarlett Johansson, which caused distress, as the actress told The Washington Post.
Deepfakes can also be used to manipulate public opinion and spread misinformation. For instance, a deepfake video of Facebook CEO Mark Zuckerberg was created and spread online in a bid to influence the 2020 US presidential election.
They can also be used to falsify evidence and manipulate the outcome of court cases, as well as spread hate speech and propagate false news.
“These deepfake tools are became becoming increasingly common and easy to use, and that poses significant risks. We need to get ahead of the curve, to have the capabilities to detect deepfakes before they cause damage,” says Kumar.
But just how big is the problem? In short, it’s impossible to tell.
“Because of the challenges of detecting deepfakes, it’s very hard to quantify. But there is an abundance of examples,” says Kumar.
Call to action
The ability to detect deepfakes is critical to preserving trust in media and preventing the spread of misinformation.
To combat the issue, researchers have developed deepfake detection algorithms, and organisations such as Facebook, Microsoft, and the Partnership on AI’s Media Integrity Steering Committee have launched initiatives to spur innovation in deepfake detection.
But despite the help of cutting-edge deep learning models, deepfake detection technology is not full-proof and is one-step behind the most recent advances.
One of the commonest methods to generate fake images is through generative adversarial networks (GANs), which can manipulate pixels in a novel way.
Today’s detection techniques rely on identifying the fingerprints of this method, and had been thought to be reliable and accurate.
However, Kumar’s team created a highly sophisticated technique of producing deepfakes that beat state-of-the-art detectors, which analyse spectral power distribution discrepancy, and were not picked by support vector machine (SVM)-based or convolutional neural network-based classifiers that analyze such discrepancies.
“We developed cutting-edge techniques for generating highly sophisticated and high-resolution fingerprint images that you cannot differentiate,” says Kumar.
As well as sounding a warning, the results, published in a paper under the title Think Twice Before Detecting GAN-generated Fake Images from their Spectral Domain Imprints and presented at 2022’s Conference on Computer Vision and Pattern Recognition in New Orleans, also point a way forward.
By identifying how it’s possible to cheat the most deepfake detectors, the researchers have also highlighted where more work needs to be done.
“There is a need to invest in and develop deepfake detection technologies and techniques as well as educate the general public on trusted sources of information,” says Kumar.
