Research
My interdisciplinary background in mathematics, physics and computer science has directed my research in accelerator physics and astrophysics. In recent years, I have been using state-of-the-art mathematical techniques and computational architectures to advance our ability to simulate and optimize particle accelerators.
Inverse Compton Sources of X-Ray Radiation
Inverse Compton sources of electromagnetic radiation utilizing relativistic electrons have seen increased use in fundamental physics research in the recent years. The small frequency bandwidth of the scattered radiation is highly desirable for applications in nuclear physics, medicine and homeland security. As the intensity of the incident laser pulse increases, the bandwidth of the emitted radiation increases. We have recently shown that a judicious frequency modulation of the laser pulse can completely counteract this increase in bandwidth in Thomson sources (Terzić, Deitrick, Hofler & Krafft 2014; Terzić & Krafft 2016).
In studying the interplay between the intrinsic bandwidth of the single-electron scattering with the energy spread of the electron beam, we demonstrate that the laser chirping in inverse Compton sources at high laser intensities: (1) enables the use of higher order harmonics, thereby reducing the required electron beam energies; and (2) increases the photon yield in a small frequency band beyond that possible with the fundamental without chirping. This combination of chirping and higher harmonics can lead to substantial savings and design, construction and operational costs of the new inverse Compton sources. This is of particular importance to the widely popular laser-plasma accelerator-based inverse Compton sources, as the improvement in their beam quality enters the regime where chirping is most effective (Terzić, Reeves & Krafft 2016).
Our current efforts focus on:
Generalizing the frequency modulation to Compton regime when electron recoil may not be neglected (Krafft et al. 2016; Ranjan et al. 2018; Terzić et al. 2021);
Generalizing the 1D plane-wave to the 3D pulse model for the laser pulse (Maroli et al. 2018; Terzić et al. 2019; Johnson et al. 2022);
Improving performance of inverse Compton sources by avoiding non-linearities (Terzić et al. 2020).
Simulation of existing and future light sources (Deitrick et al 2018; Deitrick et al. 2021).
High-Performance Simulations of Coherent Synchrotron Radiation
When the electron bunches are forced to traverse a curved trajectory, they emit bright ultraviolet or x-ray radiation. The electron machines which produce such radiation, called synchrotron light sources, are powerful tools for cutting edge research in physics, biology, chemistry, energy, medicine and other fields. As the brightness and energy of these synchrotron light sources is extended beyond present levels, it becomes necessary to develop new computational modeling capabilities. One of the most critical needs is to develop efficient computer codes for simulating collective effects that severely degrade beam quality, such as coherent synchrotron radiation (CSR) and CSR-driven micro-bunching instability (Terzić & Bassi 2011).
We have developed an innovative code for high-performance simulations of synchrotron light sources which uses massively parallel computation on graphical processing units (GPUs) (Arumugam, Godunov, Ranjan, Terzić & Zubair 2013a, 2013b, 2016; Arumugam, Godunov, Islam, Ranjan, Terzić & Zubair 2017) and wavelet methodology (Terzić, Pogorelov & Bohn 2007; Bohn et al. 2006).
We also designed and deployed an algorithm for adaptive, multidimensional numerical integration optimized to run on GPUs (Sakiotis, Arumugam, Paterno, Ranjan, Terzić & Zubair 2021, 2022).
Using Genetic Algorithms for Optimization in Accelerator Physics
Accelerator physics deals with intricate systems which depend on many interrelated specifications/variables and physical quantities. One of its main goals is to design and operate accelerators so as to achieve an efficient interplay between these many quantities, thereby optimizing their performance. This is why genetic algorithms (GAs)—efficient, robust, multidimensional nonlinear optimization tools—are crucially important.
We implemented a parallel GA paradigm and applied it to a number of problems in accelerator physics, demonstrating that many previously intractable problems are now well within reach of GA optimization (Hofler et al. 2013; Terzić et al. 2014).
In a recent study, we also showed that solving a large subset of multidimensional nonlinear optimization problems can be significantly improved by decoupling their intrinsically linear and nonlinear parts (Cotnoir & Terzić 2016).
Simulations of Long-Term Collective Effects in Particle Colliders
Long-term simulation of the beam dynamics for millions or billions of turns in a storage ring or a collider are a direct way to validate the dynamical stability of the ring's design. This is highly desirable for the design and optimization of existing and figure storage rings and colliders, such as LHC, RHIC, LHeC and electron-ion colliders. However, such long-term simulations have been prohibitive due to their heavy computational load. Various techniques of estimating the long-term dynamical stability based on relatively short-term simulations may not provide the necessary level of confidence. One particular complication of simulating collider beam dynamics is the necessity to account for the beam-beam interaction, which is an integral part of the collider dynamics and must be solved for each bunch crossing, with the solution coming at the high computational cost.
We are developing algorithms specifically crafted to exploit fully the massive parallelism offered by the modern GPUs, thereby rendering the long-term beam-beam simulations tractable (Arumugam, Majeti, Ranjan, Zubair, Godunov & Terzić 2017).
Astrophysics & Cosmology
The discrepancy between the visible mass in galaxies or galaxy clusters, and that inferred from their dynamics is well known. The prevailing solution to this problem is dark matter. In a recent study, we showed that a different approach, one that conforms to both the current Standard Model of Particle Physics and General Relativity, explains the recently observed tight correlation between the galactic baryonic mass and its observed acceleration. Using direct calculations based on General Relativity's Lagrangian, and parameter-free galactic models, we show that the nonlinear effects of General Relativity make baryonic matter alone sufficient to explain this observation (Deur, Sargent & Terzić 2020).
In the past, I have worked on a number of other astrophysical problems. Here are some of them:
Structure of dark matter galactic halos (Merritt et al. 2006; Graham et al. 2006a; Graham et al. 2006b)
Destabilizing effects of supermassive black holes at the centers of galaxies (Terzić 2002; Kandrup, Sideris, Terzić & Bohn 2003; Terzić & Kandrup 2004)
Realistic elliptical galaxy models (Terzić & Graham 2006; Terzić & Sprague 2007)
Role of chaos in galaxy evolution (Terzić 2002; Kandrup, Sideris, Terzić & Bohn 2003; Terzić & Kandrup 2003; Terzić & Kandrup 2004)