Seminář se koná každé úterý v 10:40 v posluchárně ÚTF MFF UK
v 10. patře katedrové budovy v Tróji, V Holešovičkách 2, Praha 8
Evolution has provided a remarkably elegant theory underpinning the diversity of life on Earth. The field of Cultural Evolution has since the 1970s been working towards a similar theory of principles that govern change across diverse domains of culture. We apply CE to learn more about a large European medieval musical tradition: Gregorian chant. Chant was a central part of the soundscape in Medieval Latin Europe — any Latin Christian liturgy, such as the weekly Sunday mass, involved prescribed chants as a major part of the ritual. Its melodies are sacred objects, obligating practitioners to conserve them exactly. Nevertheless, a single chant rarely has the exact same melody in any two surviving manuscripts. What is the structure of this diversity? Do „melodic dialects“ exist, and if yes, along what lines?
Choosing a set of analogies between biological evolution and historical processes of Gregorian chant transmission, we recover these “melodic dialects” using phylogenetic methods. We show that phylogenies recover historically plausible patterns of chant melody evolution. Some, but not all, institutional networks play a more prominent role than geographic proximity; despite adjacent musical traditions developing over centuries, time plays little role. Additionally, extreme phenomena in the inferred phylogenies draw attention to exceptional historical circumstance. By careful application of CE theory, we gain large-scale insight into how a layer of shared European cultural identity was constructed.
Predictive rendering originated as a technology stack centred around Monte Carlo image synthesis, and which was initially mainly intended for virtual prototyping of appearance critical objects. However, such technologies also turned out to be surprisingly useful for high quality movie VFX: and are now further evolving in the direction of thermal engineering. After giving an overview of the research portfolio of our group, I will give an overview of the differences between predictive rendering and "normal" graphics, of what makes it appealing to the VFX industry, and outline future research directions based on the same underlying technologies.
(seminář v češtině)
Přednáška se zaměřuje na uplatnění základních fyzikálních principů v kriminalistice a forenzních vědách, kde fyzika představuje klíčový nástroj pro objektivní analýzu a interpretaci kriminalisticky relevantních dějů. V úvodu bude podán přehled vybraných fyzikálních veličin a zákonů, které se v kriminalistické praxi nejčastěji uplatňují, a jejich vazby na reálné případy. Hlavní pozornost bude věnována problematice měření rychlosti, energie a síly. V oblasti balistiky bude diskutováno stanovení rychlosti střely a kinetické energie střely a jejich význam pro posouzení účinku zásahu. Dále bude rozebrána role rychlosti při řešení konkrétních kriminalistických úloh, zejména při rekonstrukci pohybu osob a událostí. Samostatná část bude věnována analýze rychlosti pohybu lidského těla při pádech z výšky a možnostem fyzikálního modelování těchto dějů. V závěru se přednáška zaměří na měření sil a přetížení působících na lidský organismus, zejména na otázky odolnosti a tolerance mozku vůči mechanickému zatížení. Diskutovány budou metody odhadu sil působících při úderech do hlavy a jejich význam pro forenzní hodnocení mechanismu vzniku poranění. Přednáška propojí teoretické fyzikální principy s praktickými příklady z kriminalistické a znalecké praxe.
Laser-ultrasonic (LU) methods are experimental techniques utilizing laser beams for generation and detection of ultrasonic waves and vibrations. These methods, thus, enable characterization of mechanical properties of solids without mechanically touching them. During the past two decades, several unique LU experimental arrangements have been developed at the Institute of Thermomechanics, among which LU measurements of anisotropic acoustic properties of single crystals have achieved particular recognition. The lecture will present two of our recent advancements in this direction: reaching Weyl’s asymptotic behavior in resonant ultrasound spectroscopy, and detecting ultra-transient oscillations in transient-grating experiments.
The talk will explain how mud and water behave under low-pressure conditions in the Solar System. It will explain why mud extruded onto the surface of Mars can be behaving the same way as lava, and how cryolavas on icy moons can simultaneously boil and freeze, drawing on experimental results obtained from low-pressure chamber studies of metastable liquids.
We present general formulations of the phase-equilibrium and phase-stability problems for multicomponent mixtures and verify that these formulations generalize the problems of phase equilibrium and phase-stability at constant volume, temperature, and mole numbers (VTN-flash), at constant internal energy, volume, and mole numbers (UVN-flash), and at constant pressure, temperature, and mole numbers (PTN-flash). Furthermore, we develop a numerical method for solving the general formulation of phase-equilibrium problems. This algorithm is based on the direct minimization of the objective function with respect to the constraints. The algorithm uses a modified Newton-Raphson method, along with a modified Cholesky decomposition of the Hessian matrix to generate a sequence of states with decreasing values of the objective function. Properties of the algorithm are shown on phase-equilibria problems of multicomponent mixtures in different specifications and with different levels of difficulty. Complexities and numerical performance of the individual flash formulations are discussed, and the application of these methods in the compositional reservoir simulation is presented.
Positrons — the antiparticles of electrons — appear in medical imaging (PET scans), materials diagnostics, astrophysics, and molecular spectroscopy, and are building blocks for exotic antimatter like positronium and antihydrogen, used to test fundamental symmetries of nature. When a low-energy positron encounters a molecule, the physics gets rich and subtle: the positron distorts the electron cloud, and a molecular electron can temporarily tunnel out to the positron — a process called virtual positronium formation. These many-body effects enhance annihilation rates by orders of magnitude and can allow the positron to bind to a neutral molecule. Getting the theory right is a genuinely hard computational problem, requiring strong electron-positron correlations to be accounted for from first principles. We tackle this using diagrammatic many-body theory, implemented in a parallelized C++ code, which has given the first ab initio predictions of positron-molecule binding energies in agreement with experiment, and enabled a framework that can treat positron scattering and annihilation on molecules with unprecedented accuracy.
Real analysis provides many counterintuitive results concerning such basic concepts as real functions of a single real variable. Before Bolzano, Cauchy, Riemann, and Weierstrass (and other nineteenth-century mathematicians), there was not even a general agreement on the definition of a function, let alone on more advanced concepts such as limits, derivatives, differentials, and integrals. The emergence of more complicated functions, together with the study of their finer properties, both motivated and facilitated the more precise development of the foundations of analysis and of mathematics in general. One way of understanding the motivation for the formalist approach is that the much higher degree of rigour we now possess allows us to accept not only those mathematical results that were already expected, but also unexpected or even counterintuitive ones. In this talk, I will briefly mention a small part of this fascinating history, and then move on to present a couple of classical examples of "monsters" (that is, continuous functions with surprising properties, such as non-differentiability or differentiability combined with nowhere monotonicity). Finally, I will prove one or two results that feel even more surprising than the mere existence of such monsters: namely, that in some sense, monsters are more "typical" than smooth functions.
The full quantum theoretical treatment of molecular systems is in general based on the Born-Oppenheimer approximation, separating the electronic and nuclear motions. This gives rise to potential energy surfaces (PES) on which the nuclei move, while the electronic states (called adiabatic states) depend parametrically on the nuclear coordinates. The approximation holds as long as the remaining non-adiabatic couplings between the electronic states, induced by the nuclear kinetic energy, are sufficiently small. However, in the vicinity of conical intersections or avoided crossings, it breaks down and requires a different (e.g. diabatic) treatment. Conical intersections, in particular, cause singular non-adiabatic couplings and give rise to a geometric phase (Berry phase). Over the past few years, we have been developing various methods to treat systems with strong non-adiabatic and relativistic couplings (spin-orbit and hyperfine). The resulting models were used to study the complex spectroscopic signatures of non-adiabatic coupling and the Berry phase, providing a theoretical framework for interpreting experimental results.
The development of optical microscopy has been driven by the quest to improve its spatial resolution, fundamentally limited by diffraction. Starting from the reminder of basic concepts of optical imaging, we will embark on the journey to overcome this limit and discuss the strategies used to this end, including stochastic methods and tip-based approaches. In tip-based approaches, resolution can be pushed to the ultimate limit by combining scanning tunneling microscopy (STM) with light. In this configuration, an atomically sharp tip forms a nanoscale optical cavity, where electromagnetic fields are confined to the scale of a single atom. Light-assisted STM (STML) enables new imaging modalities and allows probing molecular excitations in real space and imaging features associated with the molecular electronic structure. I will discuss the imaging principles and clarify the interpretation of light-assisted STM images across different STML imaging modes. Finally, I will illustrate these capabilities using concrete examples from our combined theoretical and experimental work.
The concept of a neural network once linked neuroscience and artificial intelligence quite closely, but the two fields have since followed increasingly separate paths. Biological neural networks are shaped by specific mechanisms of signaling, plasticity, and recurrent circuit dynamics, while modern artificial networks are typically defined by abstract architectures and training methods. This talk will explore that divergence from the perspective of computational neuroscience and theoretical physics. The emphasis will be on emergent properties of networks: how rich collective behavior arises from basic biophysical mechanisms, and how such emergent dynamics connect to the more top-down frameworks for computation and learning used in artificial neural networks.
Most of our understanding of matter is based on near-equilibrium probes, i.e., within the linear response. This has started to change radically with the development of strong and ultrashort laser pulses, which enable to induce and observe the non-equilibrium dynamics of electrons in systems ranging from atoms to molecules and solids on their natural timescale of a few attoseconds or femtoseconds. Harnessing the nonlinear response of strongly driven quantum many-body systems offers exciting opportunities to create dynamical states with novel properties and potential technological applications. At the same time, these experimental advances have revealed a significant gap in the available theoretical methods to describe driven quantum many-body systems, as solving the time-dependent many-body Schrödinger equation with sufficient accuracy remains challenging. In my talk, I will demonstrate how time-dependent quantum many-body systems can be tackled using reduced density matrices. I will show how reduced density matrices can be propagated without the knowledge of the full wavefunction, and how they can be used to quantify electronic correlations. As a concrete example, I will discuss high-harmonic generation in multi-electron atoms and demonstrate that electronic correlations strongly influence the harmonic yield.
Jiří Horáček David Heyrovský