University of Pretoria · Clean Energy Research Group

CERG – FLUX

Fluids, Learning, and Uncertainty in compleX systems

Pronounced "SURGE-FLUX", CERG-FLUX Lab is a research subgroup within the Clean Energy Research Group (CERG) at the University of Pretoria, developing physics-informed computational methods to understand transport phenomena — from turbulent multiphase flows and lattice Boltzmann dynamics to scientific machine learning, high-energy particle physics at CERN, and quantum technologies.

Principal Investigator Assoc. Prof. Muaaz Bhamjee
Parent Group CERG · EBIT
Collaborations ATLAS · UPQuST
PI Website Lab GitHub Org

01 — Research Pillars

Transport phenomena across scales

Our work spans five interconnected pillars, unified by a common thread: physics-informed computational simulation of transport phenomena.

∇

Computational Fluid Dynamics & Multiphase Flow

Modelling complex transport in industrial and environmental systems — from air-core vortex formation in hydrocyclones to turbulent multiphase separation.

VOFEulerianASMANSYS FluentOpenFOAM
λ

Lattice Boltzmann Methods

Mesoscale simulation of fluid dynamics through kinetic theory — coupling LBM with VOF and DEM for particle-laden and free-surface flows.

PalabosLBM-VOFLBM-DEMKinetic Theory
∂

Scientific Machine Learning

Physics-informed neural networks, neural operators, and hybrid architectures that embed governing equations as inductive bias for forward and inverse problems.

PINNsPyTorchNeural OperatorsNavier-Stokes
⊕

High-Energy Particle Physics

Contributing to the ATLAS experiment at CERN's Large Hadron Collider — UP's institutional Representative/Team Leader, advancing detector physics and data analysis.

ATLASCERN LHCSA-ATLASDetector Physics
ψ

Quantum Technologies

Exploring quantum computing, quantum machine learning, and quantum sensing applications through the University of Pretoria Quantum Science and Technology group (UPQuST).

UPQuSTQuantum ComputingQuantum Machine LearningQuantum Sensing

02 — Vision

Why FLUX?

Physics doesn't respect disciplinary boundaries. Neither should the methods we use to understand it.

CERG-FLUX Lab exists at the intersection of fundamental physics, computational mathematics, and modern machine learning. We believe the most impactful advances in simulation science emerge when classical methods — rooted in conservation laws and kinetic theory — are augmented by data-driven approaches that respect the physics they serve.

As a subgroup within the Clean Energy Research Group, CERG-FLUX Lab brings a focused computational identity to CERG's broader mission. Our research programme draws on industrial CFD, mesoscale LBM, deep learning, particle physics, and quantum information — not as parallel silos, but as a coherent portfolio connected by the computational simulation of transport phenomena.

We value depth over breadth, rigour over hype, and building over buying.

03 — People

The team

CERG-FLUX Lab is a research subgroup within the Clean Energy Research Group at the University of Pretoria.

MB

Muaaz Bhamjee

Principal Investigator · Associate Professor

Muaaz's career spans industrial CFD at Hatch Africa, geospatial data science and climate foundation models at IBM Research, and academic appointments at UJ and UP. His PhD work on hydrocyclone CFD modelling benchmarked VOF, ASM, Eulerian, and LBM-VOF approaches. He serves as UP's institutional representative in the ATLAS Collaboration, vice-president of the South African Association for Theoretical and Applied Mechanics (SAAM), and vice-president of the South African National IUTAM Committee.

Website Scholar LinkedIn GitHub Substack

Postdoctoral Research Fellows

Positions in pipeline

PhD Students

Saphira Mawe Takoutchouang
PhD · UP
An integrated study of fluid–structure interactions in pathological arteries: development of numerical and experimental models for understanding pathophysiological mechanisms and optimizing methods to improve haemodynamic modelling accuracy.
Masodi Ramokali
PhD · UP
Hybrid Quantum-Classical-ML based Lattice Boltzmann Computational Fluid Dynamics Solver.
Martin SW Potgieter
PhD · UP
Hybrid Deep Learning and Lattice Boltzmann Method for Multiphase Flow Simulation.
Abiola Yetunde Akinbowale
PhD · UP
Quantum Machine Learning for Rare Particle Signatures in High-Dimensional Physics Data
TBC
PhD · UP
—
1–2 registrations pending

MEng Students

Mwalimo M Maila
MEng · UP
Modeling of Multiphase Flow in Cyclone Separators
Brighton Terence Kazembe
MEng · UP
Modeling of a heated gas solid fluidized bed using a Eulerian Eulerian granular model and a dense discrete phase model
TBC
MSc · UP
—

Collaborators

Prof Simon Connell

University of Johannesburg

CFD modelling in the ATLAS ITk

Exotic Higgs decay in BSM searches at ATLAS

Dr Loan Truong

University of Johannesburg

Exotic Higgs decay in BSM searches at ATLAS

Prof Lerothodi Leeuw

University of Pretoria

CFD modelling in the ATLAS ITk

Exotic Higgs decay in BSM searches at ATLAS

Quantum Machine Learning for Rare Particle Signatures in High-Dimensional Physics Data

Coming soon

04 — Publications

Recent publications

Dynamically loaded from Semantic Scholar. For a full list, see the PI's Google Scholar.

05 — In the News

Latest updates

06 — Infrastructure

How we compute

We build and operate local HPC infrastructure and leverage national supercomputing facilities — automation-first, purpose-built for research workloads.

CHPC Lengau

South Africa's national supercomputing facility — petascale CPU and GPU clusters for large-scale simulation and ML workloads. chpc.ac.za

Homelab HPC Cluster

Multi-node cluster (mjolnir, legion) with full monitoring stack, SSH tunnels, systemd services, and a FastAPI dashboard.

ANSYS RSM

Remote solve management for Fluent and CFX workloads across networked nodes with NFS-backed job submission.

IBM QPUs via the CHPC Quantum Innovation Center

Leverage IBM Quantum Processing Units for quantum computing/machine learning research and development via the CHPC QIC

PyTorch + SciML

Custom PINN training pipelines with physics-constrained loss functions, chain-rule normalisation, and PDE loss scheduling.

Palabos LBM

Lattice Boltzmann simulations with DEM coupling for particle-laden flows — unified runtime-flag-controlled source.

Future Roadmap

Tailscale mesh networking, Slurm job scheduling, dedicated head node, self-hosted RustDesk relay, Rocky Linux migration.

Open Research Topics — CERG-FlUX Lab

07 - Open Research Topics

Problems worth working on

The topics below are actively available for postgraduate and postdoctoral research within the CERG-FlUX Lab. They span our core pillars — CFD & LBM, high-energy physics, and quantum computing & machine learning — and often sit at the intersections between them. If a topic interests you, get in touch.

▦
CFD & Lattice Boltzmann Methods Pillar I
MSc / MEng Self-funded

Multiphase LBM with VOF for high-density-ratio flows — hydrocyclone validation

Incorporate a true multiphase or two-phase VOF model in the Lattice Boltzmann Method capable of handling high-Reynolds-number flow with high density ratios and capturing reverse flow at outlet boundaries. Validated through a case study on air-core formation in a hydrocyclone using existing experimental data.

LBM VOF Multiphase Hydrocyclone

Supervisor: Prof M. Bhamjee

MSc / MEng Self-funded

LBM for multiphase flows in gas-solid cyclone separators

Apply and refine the Lattice Boltzmann Method for simulating multiphase flows in gas-solid cyclone separators used in mineral processing. Assess strengths and limitations in handling turbulence, phase interactions, and particle separation efficiency, with validation against experimental data.

LBM Gas-Solid Cyclone Mineral Processing

Supervisor: Prof M. Bhamjee

PhD Self-funded

GPU-accelerated LBM for multiphase heat transfer on HPC infrastructure

Develop a high-performance lattice Boltzmann solver targeting multiphase flow and conjugate heat transfer, leveraging GPU parallelism on CHPC Lengau/Sebowa clusters. Target clean energy and industrial applications requiring large-scale, time-resolved thermal-fluid simulation.

LBM GPU / CUDA HPC Heat Transfer

Supervisor: Prof M. Bhamjee

MSc / MEng Self-funded

Pore-scale LBM simulation of thermal energy storage in porous media

Use thermal lattice Boltzmann methods to simulate heat transfer and phase change at the pore scale in metal foam structures saturated with phase change materials. Investigate the interplay between foam morphology, porosity, and charging/discharging rates for compact energy storage.

LBM Porous Media PCM Thermal Storage

Supervisor: Prof M. Bhamjee

⊕
High-Energy Physics — ATLAS Experiment Pillar II
PhD Self-funded

Exotic Higgs boson decay searches in BSM scenarios at ATLAS

Search for non-standard Higgs boson decay channels using LHC Run 3 and HL-LHC datasets. Develop novel analysis strategies for rare decay topologies with complex final states, contributing to global BSM physics searches through the ATLAS experiment.

ATLAS Higgs BSM LHC Run 3 HL-LHC

Supervisor: Prof M. Bhamjee

MSc Self-funded

Physics-informed ML for signal–background discrimination in ATLAS

Develop and benchmark classical and ML-based classification approaches for separating rare physics signals from Standard Model backgrounds in ATLAS data. Focus on interpretability, physics-informed feature engineering, and robustness to systematic uncertainties.

ATLAS Machine Learning Classification Systematics

Supervisor: Prof M. Bhamjee

MSc Self-funded

Fast surrogate models for ATLAS detector simulation

Investigate generative deep learning architectures (GANs, normalising flows, diffusion models) as fast surrogate replacements for full Geant4 detector simulation in the ATLAS calorimeter, targeting orders-of-magnitude speedup while preserving physics fidelity.

ATLAS Generative Models Fast Sim Geant4

Supervisor: Prof M. Bhamjee

◈
Quantum Computing & Machine Learning Pillar III
PhD Funding available

Quantum sensing for mineral processing

Investigate quantum sensing modalities — including nitrogen-vacancy (NV) centre diamond magnetometry and superconducting quantum interference devices — for non-invasive, high-resolution characterisation of flow, temperature, and particle dynamics in mineral processing systems. Benchmark against conventional measurement techniques and assess feasibility for in-situ process monitoring in cyclone separators, flotation cells, and comminution circuits.

Quantum Sensing NV Centres Mineral Processing Process Monitoring

Supervisor: Prof M. Bhamjee

PhD Self-funded

Quantum machine learning for rare particle signatures in high-dimensional physics data

Investigate quantum-enhanced classification and anomaly detection for identifying rare particle signatures in ATLAS datasets. Benchmark variational quantum classifiers and quantum kernel methods against classical ML baselines on NISQ hardware and simulators.

QML ATLAS Anomaly Detection NISQ

Supervisor: Prof M. Bhamjee

PhD Self-funded

Hybrid LBM–deep learning for multiphase flow simulation

Develop a novel hybrid modelling approach integrating the Lattice Boltzmann Method with Physics-Informed Neural Networks (PINNs), Graph Neural Networks (GNNs), and Generative Adversarial Networks (GANs) to enhance accuracy and computational efficiency of multiphase flow simulations. Test against traditional solvers and experimental data.

LBM PINNs GNNs GANs

Supervisor: Prof M. Bhamjee

MSc / MEng Self-funded

ML-enhanced multiphase flow modelling for industrial separators

Integrate deep learning methods (PINNs, GNNs) with traditional numerical approaches to improve phase interaction predictions in CFD. Apply to industrial processes including cyclone separators and fluidised beds, evaluating performance against conventional models.

PINNs GNNs CFD Cyclone Separators

Supervisor: Prof M. Bhamjee

MSc / MEng Self-funded

Deep learning surrogate models for high-Reynolds-number CFD

Investigate deep learning-based surrogate models for accelerating CFD simulations of high-Reynolds-number flows. Analyse trade-offs between accuracy and computational cost, testing against traditional numerical solvers and experimental datasets.

Deep Learning Surrogate Models High-Re CFD

Supervisor: Prof M. Bhamjee

PhD Self-funded

Quantum algorithms for lattice Boltzmann fluid simulation

Explore quantum computing approaches to the collision and streaming steps of the Lattice Boltzmann Method, assessing potential speedups on gate-based quantum hardware. Benchmark quantum-classical hybrid LBM workflows against GPU-accelerated classical solvers for canonical flow problems.

Quantum Computing LBM Hybrid Algorithms

Supervisor: Prof M. Bhamjee

MSc Self-funded

Quantum sensing for high-resolution thermal-fluid measurement

Investigate nitrogen-vacancy (NV) centre diamond magnetometry and other quantum sensing modalities for non-invasive, high-spatial-resolution measurement of temperature and flow fields in compact heat transfer systems, benchmarking against conventional thermometry.

Quantum Sensing NV Centres Thermometry

Supervisor: Prof M. Bhamjee

⬡
Cross-Cutting & Emerging Bridging Pillars
PhD Self-funded

LBM-based digital twin framework for real-time thermal management

Develop a digital twin coupling GPU-accelerated LBM simulation with sensor data streams for real-time thermal monitoring and predictive control. Combine reduced-order LBM with ML-based state estimation for fast inference in operational environments.

Digital Twin LBM Real-Time ML

Supervisor: Prof M. Bhamjee

MSc Self-funded

Uncertainty quantification for LBM and ML-augmented CFD

Develop Bayesian and ensemble-based uncertainty quantification methods for lattice Boltzmann and ML-augmented CFD predictions, quantifying epistemic and aleatoric uncertainty in multiphase flow simulations and physics-informed surrogate models.

UQ Bayesian LBM SciML

Supervisor: Prof M. Bhamjee

MSc Self-funded

Reinforcement learning for active flow control in LBM environments

Train reinforcement learning agents to discover active flow control strategies (blowing, suction, surface actuation) within LBM simulation environments, targeting drag reduction or heat transfer enhancement. Assess transferability of learned policies to higher-fidelity solvers.

RL Flow Control LBM Heat Transfer

Supervisor: Prof M. Bhamjee

Topics can be adapted to your background and interests. We welcome proposals that bridge our pillars in unexpected ways. If none of the above fits exactly but you're drawn to our research areas, reach out anyway. Several topics require self-funding — see the UP funding page, DSI-NRF Masters/PhD programme, or DSI-CSIR programme for options.

Get in touch →

08 — Join Us

Build the tools to find out

We are actively seeking motivated postgraduate students (MEng, MSc & PhD) and postdoctoral researchers with interests in CFD, LBM, scientific machine learning, high-energy physics, or quantum computing/machine learning and sensing. Funded positions are available through NRF, SA QuTI, UPQuST, and the ATLAS Collaboration.

Who we're looking for

Strong foundations in engineering, applied mathematics, computer science, software/computer engineering, or physics. Comfort with rigorous mathematics and writing code in Python, C++, or Fortran.

How we work

We read papers critically, ask good questions, document our work, and take ownership of our projects. We'd rather understand a method deeply than apply it blindly.

self-funded

Positions are funded through the NRF, SA QuTI, UPQuST, and the ATLAS Collaboration. Multiple openings at MEng, MSc, PhD, and postdoctoral level.

You don't need to arrive as an expert in all of our pillars — nobody is — but you should bring curiosity about how things work at a fundamental level and the patience to work through hard problems carefully rather than rushing to results.

Get in touch →

Please reach out to Prof Muaaz Bhamjee at muaaz.bhamjee@up.ac.za with a brief introduction, your CV, and any relevant work or project samples you're willing to share.