Home  |  NARL  | 中文 
Sun, September 24, 2017

Seawater Temp. (-43.6 m)

EHTW Earth System Model: A fully integrated OAGCM without traditional coupler

B.J. Tsuang, National Chung Hsing University

Taiwanese scientists have integrated an atmosphere-ocean coupled model (named as ECHAM/SIT/TIMCOM or EHTW Earth System Model), recently. This ESM model has been used for present climate simulation, future climate simulation, hindcast and 45-d forecast. This model does not use any specific coupler while the OGCM is directly coupled with AGCM through SIT and well simulated such as diurnal variation of SST, Madden-Julian Oscillation (MJO), and ENSO.

This research group comprises NCHU professors Ben- Jei Tsuang (PI), I-En Liao(Big Data), Kuen-Tsann Chen, Lu-Hung Chen, Shih Dong-Sin (Hydrological Watershed Model), NCAR professor Yu-Heng Tseng (TIMCOM), Academia Sinica professor Huang-Hsiung Hsu (Model), NTU Lin Ho (Model Verification), NCU professors Shu-Chih Yang (Data Assimilation), Fang-Yi Cheng (Land surface processes), Chinese Culture University professor C.-W. June Chang (Regional ocean-atmosphere interaction), and so on.

NCHC and this research group have formed a close partnership resulting in significant progress in this research, such as Chau-Yi Chou improving program efficiency in 20%, Dr. Whey-Fone Tsai, Lung-Cheng Lee, and Shyh-Ching Lin collaborating big data and 3D visualization, San-Liang Chu, Chin-Chen Chu, Ying-Yu Shih improving storage efficiency, and Dr. Wei-Cheng Huang for parallel render.


Supercomputers can help doctors to decode cancer tricks

C.Y. Huang, National Yang-Ming University

Cancer is the top one killer in Taiwan. It kills twice much people than the second ranked killer, heart failure. Cancer is complicated and has numerous causes. It can be influenced by personal genetic background, tumor progression, tumor micro-environment, carcinogen contamination, virus infection, radiation, and so on. Therefore, tumors from cancer patients will be different even if they are the same type of cancer. We might have ever heard someone said: "Earlier stage tumors have good prognosis, and later stage tumors have bad prognosis.” Actually, the saying is not all true to patient's outcomes. Some late stage breast cancer patients still have acceptable prognosis. The main reason is that tumor stage is not the only indicator. For example, estrogen receptor (ER, ESR1) is a biomarker frequently used in breast cancer. Most ER positive tumors (examined by particular procedures from biopsies) are sensitive to Tamoxifen because the drug inhibits biological functions of estrogen receptor and limits proliferation abilities of tumor cells. Proliferation is the key ability for tumor cells to replicate themselves. It is why Tamoxifen works particularly good against ER positive tumors and not effective against ER negative tumors. Therefore, biomarkers are necessary and can provide better biological characteristics of tumors for predicting the responses of cancer therapies. In fact, many cancer drugs were designed to target specific biomarkers.

Hepatocellular carcinoma (HCC) is the major type of liver cancer. It is also one of three fastest-growing cancers in the US and the most lethal cancer in Asia. Unfortunately, compared to other cancer types such as breast cancer, HCC has poor prognosis and very limited treatment options. Therefore, identifying new biomarkers for HCC is imperative. Professor Chi-Ying F. Huang and his team from National Yang-Ming University performed ultra-deep RNA-Seq analysis to explore a new path toward biomarker identification. RNA-Seq is an application of new generation sequencing technologies for reading RNA sequences in the cells. It offers single base resolution at the genome wide scale and has largely impacted cancer research. Recent studies using RNA-Seq has greatly improved our knowledge of both the quantitative and qualitative aspects of the human transcriptome. However, RNA-Seq also generates large amount of data. For example, one pair of HCC samples can have raw data nearly one hundred gigabytes and require terabytes of disk space for analyzing the data. It goes without saying that such task requires substantial computing resources. NCHC and NRPB provided 32 nodes with 1,536 cores for helping Professor Huang's team. Therefore, the preliminary results of RNA-Seq analysis can be generated in barely 1 month.

After extensive bioinformatics analysis, the team further confirmed the expression patterns of several biomarker candidates in 55 pairs of HCC samples from Taiwan Liver Cancer Network (TLCN) and cancer cell lines. The results were consistent with what were observed in their RNA-Seq data. Especially, Dr. Kuan-Ting, Lin and his colleague discovered a novel gene, DUNQU1, named after "遁去的一*". This gene has never been documented and has intriguing expression patterns in HCC samples. The approach indeed revealed new possibilities of biomarker identification. The new biomarkers could be used for improving prognosis and developing new therapies in the future. The remarkable results have been published on Oncogene (Nature publishing group) in 2013. (http://www.nature.com/onc/journal/vaop/ncurrent/full/onc2013424a.html)

*“遁去的ㄧ” is derived from The Tang Dynasty Sworn Brothers legends (大唐雙龍傳, 黃易出版社有限公司), a famous Chinese martial arts novel. The term means "nothing is doomed, and there is a latent variable that flips yin and yang". In Chinese, “遁去” sounds like DUNQU. For that reason, the authors named the novel gene as DUNQU1.


J.L. Kuo, Institute of Atomic and Molecular Sciences, Academia Sinica

Water is important in biological functions and receives much attention in the science community. Though being very much studied, its physical and chemical properties as well as the origins of many water anomalies are not yet fully understood. It is conjectured that the anomalies (such as density, structure, thermodynamic, and physical properties) may be explained through the two-fluid model, namely, two types of structurally different water, the high density liquid (HDL) and the low density liquid (LDL), coexist in the supercooled regime. However, experimental study to the existence of the two fluids is hindered since the system enters crystallization stage earlier than the supercooled regime when decreasing the temperature, making the study of the supercooled fluid not accessible.

Numerically, it is possible to investigate the aforementioned fluid states since crystallization requires a longer time scale. The very first step is to retrieve an accurate phase diagram. With the correct topologies at different states, it is then possible to collect the thermodynamic and dynamic properties. However, it is not easy even for the first step since the system quickly enters the glassy state when temperature is decreased. In the language of potential energy landscape, it means that the kinetic energy is too low to help the system to escape from some local potential wells. The relaxation for these trapped states may takes up to several seconds, which is obviously unattainable by the current computational ability. Hence, various accelerated molecular dynamics (MD) methods have been designed to tackle this problem. With the help of NCHC, Academia Sinica principal investigator Dr. Jer-Lai Kuo and collaborates used the volume-density replica exchange molecular dynamics (VTREMD) to compute the water phase diagram spanning very large scale phase space. The aforementioned method is one of the accelerated MD, which is realized by enhanced sampling of states. This study constructs a system of 720 replicas of TIP4P/2005 water molecules in each one. Thanks to the NCHC’s ALPS cluster, the large test phase space is made possible to span from 145 K to 363 K, where the low temperature end is hardly achievable via classical molecular dynamics. A typical VTREMD test required 720 cores running in parallel for at least 20 ns simulation time, namely, 144,000 CPU hours. With this large-scale test, system convergence is speeded up (from possibly several ms to a dozen of ns) and thus renders a possible investigation to the deeply supercooled water. Next, LDL-like and HDL-like molecules are labelled by its distance of the 5th nearest neighbor oxygen, where those with distance larger (smaller) than 0.35 nm are LDL-like (HDL-like) ones. This study reveals the distinct behaviour of the two-fluid mixture under different states and is helpful in explaining the origin of water anomalies. More information is available at the Molecular and Material Modelling Lab (https://sites.google.com/site/jlkiams/) of Academia Sinica, Taipei, Taiwan.


Climate Modeling Activities of the Consortium for Climate Change Study

H.H. Hsu, Research Center for Environmental Changes, Academia Sinica

The “Consortium for Climate Change Study (CCliCS)” project recruits experienced researchers with expertise in climate study, model development and simulation to work on capacity building in climate change modeling, simulation, and climate signal interpretation. The consortium core, Laboratory for Climate Change Research, is hosted by the Research Center for Environmental Change, Academia Sinica to serve as the research hub of the consortium to organize the research activities and model suit implementation. Principle investigators and researchers are from National Taiwan University, National Central University, and National Taiwan Normal University, and Research Center for Environmental Change. The project is a five-year project from August 2011 to July 2016.

Research objectives:

1. To develop Taiwan’s capability in model development, develop a Community Earth System Model Suite that can be further improve locally, and provide it for use by research community.

2. Use the models to evaluate and project the influence of climate variability and change on East Asian climate and monsoon and the high-impact weather and climate in Taiwan.

Cooperation with the NCHC:

CCliCS project needs huge computing resource and storage space to support carry out high-resolution, long-term climate simulation and projection. The computer resources provided by the NCHC allow the CCliCS to perform (for the first time in Taiwan) the climate projection calculation.

"ALPS" provides excellent computing environment and professional technical support. In the past, because of the limited computing power with only 768 cores, climate simulations can only be conducted at the spatial resolution of 100 km X 100 km. The new 1,536 cores in ALPS provided by NCHC allow us to perform the simulation at the 23 km X 23km resolution and save the computing time by 51 percent. Climate change simulation in Taiwan cannot be done without the generous offer by NCHC.

Research subjects:

1. Development of a community Earth system model suite including a global Earth system model, a global cloud-resolving atmospheric model, and a high-resolution regional model.

2. Refinement of physical parameterization package and ocean model of an existing Earth system model to make it a Taiwan Earth System Model.

3. Assessment of relative influence of natural climate variability and anthropogenic climate change on high-impact weather and climate using developed models.

4. Capability building in the assessment and projection of climate change impact on the extreme weather and climate.

5. Establishment of climate change data and modeling platform.

6. Contribution to the writing of Taiwan Climate Change Science Report.


QUANTUM FLUCTations IN the QCD Vacuum

T.W. Chiu, National Taiwan U.

Understanding the vacuum of quantum chromodynamics (QCD) has tremendous impact on today's most challenging questions in physics and astronomy. QCD is the fundamental theory for the interaction between quarks and gluons. It provides the theoretical framework to understand the nuclear force and energy from the first principles, and also plays an important role in the evolution of the early universe.

Using NCHC supercomputers, researchers have computed the most realistic simulations to date of the quantum fluctuations of the QCD vacuum. By comparing these simulations of the QCD vacuum with the experimental data from major laboratories, physicists can gain new insights into key features of the strong interaction at the subatomic scale, as well as guiding their search for new insights into how the early universe evolved from the quark-gluon plasma phase to the hadron phase.

To simulate the quantum fluctuations of the QCD vacuum, researchers have used the DWFQCD code developed by Professor Ting-Wai Chiu and his research group at National Taiwan University. The code simulates the quantum fluctuations of quarks and gluons from the first principles of QCD, using the lattice fermion with optimal chiral symmetry. DWFQCD is very computational intensive, pushing the limit of computational methods, and requiring the fastest supercomputers. Since 2009, Professor Chiu’s research group has been taking advantage of the enormous floating-point computing power delivered by the Nvidia GPU, and using GPU clusters

(at NTU and NCHC) to simulate QCD with dynamical quarks. The researchers ran the most advanced simulation ever of lattice QCD on NCHC’s Formosa-5 GPU cluster, which incorporates the quantum fluctuations of (u,d,s,c) quarks and gluons, and provides scientists dynamical gauge configurations for understand the hadron physics as well as the basic science of the early universe. The successes are mostly due to DWFQCD’s world-leading high-precision chiral symmetry, capturing the most essential feature of quarks.

With these dynamical gauge configurations, researchers can probe the physics of the QCD vacuum using the low-lying eigenmodes of the quark matrix, since they encapsulate the important information to answer the crucial questions - how the proton gains its mass through the spontaneously chiral symmetry breaking of the QCD vacuum, as well as at what temperature in the early universe the chiral symmetry is restored. To make this research possible, NCHC ALPS supercomputer plays an important role, providing a platform for computing the low-lying eigenmodes of the quark matrix, which requires very-large scale computations.


1/12 Reactor Core Computation

C.C. Chieng, National Tsing Hua University

The development of a gas-cooled very high temperature reactor (VHTR) has always been an important consideration with regard to next generation nuclear power plants, especially considering the present day high demand on nuclear safety. This highlights the importance of analyzing the flow and temperature distribution within the reactor’s core, or thermal hydraulic phenomena during accident scenarios.

This study investigated core bypass flow during normal operations and validated whether or not the computations were reasonable. It then analyzed loss of flow accidents (LOFA) too. The research team began computations from a single channel that was gradually expanded to the 1/12 core section, and used both global and local symmetries to gain a detailed estimate of the strength of the resulting natural circulation as well as the level of heat transfer. Finally, the team successfully computed a more accurate natural convection using the validated 1/12 core model.

The NCHC’s supercomputers contributed to the repeated validation of this result by significantly shortening the time required for each individual case scenario from one to two years to only two months and, at the same time, also enabled the simulation of such large scale computational fluid dynamics (CFD).

Multiscale Molecular Simulation of Solar Energy Harvesting

C.W. Pao,Research Center for Applied Sciences, Academia Sinica

Organic photovoltaic cells (OPVs) are one of the promising renewable energy resources because of their low production costs, mechanical flexibility, and light weight comparing with their inorganic counterparts.
The photoactive layer of OPV – namely, the bulk heterojunction (BHJ) layer - comprises an interpenetrating network of electron donor and acceptor materials. Electron donor materials are usually semiconducting polymers such as P3HT, whereas electron acceptor materials are usually materials with high electron affinity such as fullerenes or semiconducting nanocrystals such as TiO2 nanoparticles/nanorods . The nanomorphologies of the BHJ layer is critical for the efficiency of OPVs and are very sensitive to the device fabrication conditions; therefore, comprehensive insights into the correlations between the BHJ nanomorphologies and fabrication conditions are important for promoting device performance. However, experimental characterization of the nanomorhologies of the BHJ layer is never trivial; therefore, computer simulations can effectively help fill the gap between device fabrication conditions and BHJ nanomorphologies. With ALPS at NCHC, for the first time, we are able to reveal the nanomorphologies of BHJ OPVs with system sizes compatible with experiments (50x50x50 nm3) by carrying out large-scale coarse-grained molecular dynamics (CGMD) simulations. We have studied the nanomorphologies of polymer-fullerene, and polymer-inorganic nanocrystal hybrid blends with unprecedented level of details, potentially helpful for experimental teams in Taiwan to develop next-generation OPV devices with superior performances.


Investigating the suitable materials for photocatalytic water splitting

J.L. Kuo, Institute of Atomic and Molecular Sciences, Academia Sinica

Computational model of water splitting processes at GaN surface decorated with a Pt4 cluster

The crisis of upcoming global petroleum shortage stimulates the investigations of possible renewable energy resources for human’s future usage. Harvesting sunlight energy through photocatalytic water splitting is one of the methods believed to be feasible in mass production for getting hydrogen gas. Photocatalysts like TiO2, (Ga1-xZnx)N1-xOx, SrTiO3 … combining with various cocatalysts were studied experimentally and theoretically to look for the most efficient ways of transforming sunlight energy into chemical energy. However, there are still lots of rooms for improving the overall performance after many tries and errors in recent few decades. The understanding of its detail mechanism is essential for the design of a stable and efficient sunlight harvesting system. Nevertheless, in many cases the atomic level details become inaccessible for limited resolution in experiments. On the other hand, computational modeling could help the understanding of what actually happened in whole processes in terms of charge transfer, reaction energy barriers, density of states, and etc. With the power of parallel computing cross many processors, modeling more complicated systems which are closer to experiments becomes possible. One of the study interests in Dr. Jer-Lai Kuo’s group is to investigate various processes, which could contribute to water splitting and hydrogen gas production. At the same time, they also aim at exploring suitable materials for photocatalysts and cocatalysts. They have studied the water adsorption behaviors on GaN polar (0001), non-polar (101 ̅0) surfaces and also the effect of ZnO doping on surfaces. You can find more information at Dr. Jer-Lai Kuo’s group webpage (http://jlk.iams.sinica.edu.tw/).

With the support of National Center of High-performance Computing, Dr. Jer-Lai Kuo’s group used more than 20 million processor-hours for all of their simulations during 2012 to 2013. Their jobs were using up to 512 processors in parallel computing for sophisticated models. They will continue their study with the powerful facility in National Center of High-performance Computing

Home | Top | Back