Open calls in doctoral studies 2026

Motivation letter

Please write a brief motivation letter (in English, maximum of 6000 characters with spaces) based on the following points:

1. Why are you interested in this PhD project, explain your choice.
2. What is your previous experience in this field? Explain how your educational and professional background relates to the project you are applying to.
3. What are the analytical/scientific methods you have practiced.
4. Describe briefly the methods and main results of your MSc thesis.
5. Decribe your earlier research activities, including research publications and conference presentations, if available.
Assessment criteria for motivation letter:

- motivation and argumentation of skills and the choice of the project
- relevant study and work experience and other relevant activities (publications, project management etc.) as required to present in the motivation letter.

Interview

The applicant must describe the wider scientific background of the doctoral project and possible applicability of the results, also their motivation to be admitted to PhD studies with particular project. The admissions interview is conducted by the admissions committee. Only applicants whose motivation letter is assessed positively will be invited to the interview (minimum positive result is 35 points out of 50).

The entrance interview is used to assess the following:

- knowledge of the wider scientific background of the project and possible application of the expected results
- applicant’s motivation to pursue doctoral studies in the relevant field of science and to work in this field
- wider analytical and generalization skills regarding the research and study topics.
The entrance interview takes place most probably in June 2026.
International applicants who cannot be present at the interview in Tartu, may conduct an online interview. Applicants will be informed of their interview date and time by the respective faculty.

Both the motivation letter and entrance interview are assessed on a scale of 0 to 50 points, minimum positive score is at least 35 points. To be invited to an interview, the applicant must earn at least 35 points for the motivation letter.

Supervisor(s): Arun Kumar Singh, Karl Kruusamäe

Multi-arm robotic systems hold the potential to revolutionize unstructured environments—such as logistics and retail—by autonomously stocking shelves or collaboratively manipulating heavy payloads. However, these complex behaviors remain out of reach for classical control stacks, which rely on rigid orchestration, and current generative AI approaches, which struggle with safety in obstacle-heavy settings. This project proposes an end-to-end planning and control framework designed to bridge this gap by mapping multi-modal inputs directly to robot actuation.

Our research introduces two primary innovations: Safety-Embedded Neural Architectures, which integrate safety constraints (e.g., Control Barrier Functions) directly into the generative model design rather than as post-hoc optimizations; and a Sample-Efficient Learning paradigm that utilizes sparse, high-level sub-goals to master tasks without dense human demonstrations. The research plan progresses from establishing high-fidelity simulation benchmarks (Year 1) to developing constrained architectures (Year 2), and finally validating the system through sim-to-real transfer and few-shot learning on physical hardware (Years 3-4). The outcome will be a robust, safety-critical system for collaborative manipulation and a suite of open-source benchmarks for the robotics community.

Supervisor(s): Indrek Must

Physical interactions still resist broad automation despite progress in AI. Due to difficulties in cross-platform use of physical training data, physical interactions remain fragile and unsuitable for critical operations, such as major infrastructure maintenance. This project implements a robotic embodiment (especially, an effector) that achieves physical grounding by in-situ development. Blow spinning enables the development and modification of effectors unique to each interaction type or even instance. Our pipeline requires minimal data on the operational environment, as the embodiment forms through direct physical interaction after robot deployment. This rethinks the robot embodiment from a static, factory-fixed form to an adaptive, gradually developing structure. The project takes in situ spun embodiments from the first feasibility demonstration to use case validation, embodying the next generation of embodied AI that truly engages the physical presence of robots.

Supervisor(s): Tauno Tiirats, Veronika Zadin

We aim to advance the computational modelling of electric-field-induced surface modifications in metals for high-gradient particle accelerators at CERN. When high electric fields (~100 MV/m) are applied to metal surfaces, atomic-scale rearrangements occur that can be exploited for surface engineering or lead to material damage as vacuum breakdown (VBD), a critical limitation for accelerator performance. Recent observations at the Future Circular Collider (FCC) indicate enhanced VBD activity in beam steering regions at lower-than-expected fields, suggesting the involvement of radiation-assisted surface VBD mechanism.

Our hypothesis is that synchrotron radiation generates mobile surface species (adatoms), which undergo electric-field-biased diffusion and aggregate into nanoscale protrusions acting as field emitters that trigger VBD. To test this hypothesis, we combine multiscale modelling and experimental validation. Atomistic simulations will quantify adatom production rates and radiation-induced surface damage, upscaling with a continuum surface diffusion model predicts macroscopic roughening rates while integration into Monte Carlo surface conditioning (MC-SC) framework leads to breakdown statistics.

Bayesian calibration will be applied to infer uncertain physical parameters from CERN VBD datasets and to quantify predictive uncertainty. The outcome will be a physics-based, experimentally validated predictive model for breakdown estimation and mitigation in high-electric-field systems, with broader relevance to electron sources, X-ray technologies, and advanced accelerator applications.

Supervisor(s): Maria Põldma, Arno Põllumäe, Georg Martin

Marine ecosystems are in a state of continuous change, influenced by climate change, eutrophication, and the spread of non-native species. Zooplankton, as a key mediator of energy and nutrient flows and a sensitive indicator of ecosystem health, provides important information on ecosystem status and dynamics. This doctoral project focuses on the diversity and dynamics of Baltic Sea zooplankton using modern molecular methods. First, the study addresses the detection of non-native species using environmental DNA (eDNA), which enables early identification of low-abundance species and complements traditional plankton monitoring. Second, the project aims to enhance the genetic reference database necessary for accurate zooplankton identification through eDNA analyses, improving taxonomic resolution and reliability. Third, changes in zooplankton and community dynamics in Estonian coastal waters over the past decades are analysed, linking observed patterns to climatic and environmental drivers. Finally, the diversity and spatial and seasonal dynamics of microzooplankton are investigated using eDNA, highlighting the previously underappreciated role of this component in the microbial food web. Overall, this work offers a comprehensive and innovative approach to assessing the state of the Baltic Sea ecosystem, supporting scientific understanding and informing marine management and policy decisions.

Supervisor(s): Vallo Tilgar

The vertebrate gut microbiome is a diverse microbial community influenced by external factors, shaping host physiology and behavior via the gut-brain axis and immune system. This has led to the novel hypothesis that animals and their microbiota function as a unit of selection—the holobiont. While gut microbiota has been well-studied in lab models, their dynamic interactions with wild hosts remain poorly understood. In nature, animal microbiomes are more variable and diverse due to broader environmental exposure and greater genetic diversity among wild hosts. This knowledge opens up a promising avenue for exploring the evolutionary role of gut microbiota in free-living populations.

This project investigates bidirectional interactions between the gut microbiome and wild great tits across different environments. First, we will explore geographic variation in gut microbiota (GM), expecting adaptive shifts in microbial traits along latitudinal gradients and habitat heterogeneity. We also predict GM composition will correlate with host personality traits. Second, we will study the potential role of GM in coping with local environmental stressors within populations in the wild. We will conduct experiments to manipulate the stress levels of animals in ways that simulate the effects of both natural and anthropogenic stressors. We will also explore how experimental alterations in the microbiota and its metabolites affect the ability of the host to cope with prevailing local environmental conditions. Combining innovative experiments with molecular and biochemical methods, we aim to uncover evolutionary mechanisms driving phenotypic variation and adaptation in free-living populations.

Supervisor(s): Jeffrey Carbillet

Human‑induced environmental changes are rapidly increasing the loss of natural breeding habitats for seabirds, forcing many species to relocate into increasingly dense breeding colonies, or into urban habitats. Increasing colony densities can intensify competition for high‑quality nesting sites and increase energetic trade‑offs, potentially reshaping selective pressures and population dynamics. At the same time, a growing number of seabirds opt for a different strategy, exploiting urban habitats. However, the ecological and fitness consequences of this strategy remain poorly understood. These two parallel strategies highlight a need to clarify how seabirds, one of the most endangered vertebrate groups, respond to accelerating anthropogenic changes. This project combines complementary approaches at global, regional, and local scales to improve our understanding of seabird ecology in a rapidly changing world. At the global scale, we will conduct the first systematic synthesis of existing data on urban‑breeding seabirds to assess how urban environments influence behaviour, health, and reproductive outcomes. At the regional scale, we will evaluate the potential of seabirds as sentinels of zoonotic diseases within a One Health framework, focusing on the role of Common gulls (Larus canus) in the transmission of Toxoplasma gondii across terrestrial and marine ecosystems. At the local scale, using a 50‑year dataset from a Common gull colony in Matsalu National Park, we will examine how increasing breeding density is shaping evolutionary processes such as sexual selection, testing for long‑term changes in white wing patch size as an honest signal of individual quality. Together, these provide a multi‑scale assessment of how habitat loss, urbanisation, and interactions with humans are reshaping seabird ecology and evolution. The project advances our understanding of the mechanisms underlying seabird population change and offers valuable insights for conservation, public health, and the broader study of wildlife responses to human-induced environmental changes.

Supervisor(s): Lõhmus, Marko Kohv

This project investigates current trajectories and rates of peatland change. It is framed as an input to peatland landscape simulation modeling by refining the conceptual structure and parameterization of the SooSim model (Lõhmus et al. 2024). Such models are essential for spatial mid-term forecasts supporting biodiversity conservation, ecosystem restoration, ecosystem service assessments, and land-use planning. The thesis will focus on three key post-restoration processes: (i) transitions among mire vegetation types following ditch blocking, (ii) woody vegetation dynamics in restored and drained mires, and (iii) passive recovery in abandoned drainage systems. By quantifying these processes, the project will strengthen the empirical foundation of peatland simulations while improving theoretical understanding of restoration trajectories, probabilities and rates, including hysteresis between degradation and recovery pathways, and trade-offs between restoration and other land-use values.

Empirical work combines permanent vegetation plots, national-scale forest-structure surveys, and remote-sensing analyses (LiDAR, orthophotos, Sentinel-2). Vegetation transition rates and tree-species-specific woodland dynamics will be modeled using statistical and state-space approaches, while passive recovery will be assessed through spatial analyses calibrated by field surveys. The results will be integrated into SooSim 2.0, introducing improved vegetation-transition, woody-dynamics, and passive-recovery modules. National-scale simulations under restoration plans to 2050 will provide updated forecasts of mire-type distribution, woodland dynamics, and conservation-value trajectories across Estonia.

Supervisors: Toomas Tammaru, Erki Õunap, Robert Davis

Ongoing climate change poses a major challenge to biodiversity, yet our understanding of species’ adaptability, particularly in thermal traits, remains limited. While thermal traits are often assumed to evolve slowly, empirical evidence for this claim is scarce. The proposed PhD project will contribute to an ongoing research initiative investigating the evolutionary conservatism of thermal traits in insects. Specifically, the project will analyse insect distribution patterns to assess thermal adaptations, leveraging most recent phylogenetic reconstructions and advanced phylogenetic comparative analyses. The study will include three main components: (1) a phylogenetic comparative study of latitudinal distribution limits, focusing on European moths, with the aim to quantify the rate of evolution in thermal niche parameters; (2) a similar study on altitudinal distribution limits, using Ecuadorian Andean moths as a model system; and (3) an investigation of evolutionary transitions between tropical and temperate distributions across all insects globally.

Supervisor(s): Kessy Abarenkov

This doctoral thesis develops an AI-driven National Biodiversity Digital Twin (EstBioDT) for Estonia, integrating eElurikkus data, eDNA observations, and environmental variables. The system transforms national biodiversity data into a dynamic, continuously updated digital representation to support research, conservation, and policy decision-making.

Machine‑learning models that integrate species occurrence records, environmental variables, and eDNA‑based observations originating from national biodiversity data flows, will be developed. These models will be used to estimate biodiversity status, detect changes over time, and simulate future scenarios under land‑use, climate, and management pressures. The Digital Twin also incorporates data from managed landscapes in forestry and agriculture, integrating DNA and image-based observations to provide early-warning signals, support scenario-based decision-making, and apply new environmental DNA technologies and machine learning methods for the detection of agricultural pests.

Overall, the thesis operationalizes a self-learning, adaptive Biodiversity Digital Twin, delivering repeatable, transparent tools for biodiversity assessment, conservation planning, and landscape management.

Supervisor(s): John Davison, Marina Semtšenko

Droughts are becoming more frequent and severe under climate change, with major consequences for grassland productivity and biodiversity. Plant and soil microbial responses to environmental stress are underpinned by complex interaction networks that remain poorly understood. Improving knowledge of these plant–soil linkages is essential for predicting drought impacts and developing management strategies that reduce drought vulnerability. This PhD project aims to address these gaps through a combination of experimental, field, and comparative approaches. First, it will evaluate commonly used experimental methods for simulating drought, assessing how differences in soil moisture dynamics in pots and mesocosms influence plant performance and microbial communities relative to natural field conditions. These results will be integrated with a comprehensive literature review to provide guidance for future drought research. Second, the project will exploit a unique network of study sites in the Pyrenees, Spain, spanning a steep aridity gradient (from <400 mm to >2000 mm annual precipitation) within a small geographic area. Using soil sampling and fungal metabarcoding, the project will examine how fungal guilds respond to increasing aridity and how microbial niche characteristics vary along this gradient. Greenhouse experiments will further test the role of microbial communities and local plant–soil adaptation in modulation plant drought resilience. Finally, the project will investigate how plant diversity and local water availability shape microbial strategies, particularly contrasting biotrophic and saprotrophic fungi. Overall, the research will advance understanding of climate–plantmicrobe interactions and inform grassland management under future drought scenarios.

Supervisor(s): Meelis Pärtel, Carlos P. Carmona

This doctoral project develops a framework for quantifying uncertainty in dark and functional diversity metrics. Contemporary ecology increasingly relies on metrics such as biodiversity potential, community completeness, and functional trait space to move beyond simple species counts. However, these metrics are based on indirect inference, incomplete trait information, uneven sampling, and probabilistic assumptions. Without explicit treatment of uncertainty, it is difficult to determine whether observed patterns reflect ecological processes or methodological artefacts. The project aims to identify, quantify, and analyse the main sources of uncertainty affecting estimates of biodiversity potential, its realisation, and functional characteristics. It builds on the dark diversity framework and on state-of-the-art functional diversity approaches. The thesis has three components: (1) assessing uncertainty in dark diversity estimation, including sampling design, co-occurrence structure, and scaling; (2) evaluating how uncertainty in trait data and functional space construction influences functional diversity indices; and (3) integrating both sources within a unified framework to examine how uncertainty propagates to derived metrics such as species pool size and community completeness. By shifting biodiversity research from point estimates to confidence-aware metrics, the project strengthens ecological inference and improves the robustness and applicability of biodiversity assessments across basic and applied ecology.

Supervisor(s): Tuul Sepp

The Baltic Sea is one of the most polluted seas in the world. Pollution from shipping and industry, including chemicals and plastics has placed heavy stress on marine life. One outcome of these environmental pressure is the appearance of diseases in sea organisms. A recently discovered disease is bivalve transmissible neoplasia (BTN) — a contagious cancer that affects clams and mussels. Unlike most cancers, BTN spreads between individuals when living cancer cells move from one animal to another. Although it was discovered only about ten years ago, it is already considered one of the most serious and overlooked diseases in marine animals.

BTN has been found in Baltic clams in the Gulf of Gdańsk, but no large-scale study has examined whether it is widespread across the Baltic Sea. It is also unknown whether other important species, such as soft-shell clams and blue mussels, are affected. In addition, we do not yet understand whether human-driven environmental pressures — such as pollution, dense shipping traffic, or microplastics — increase the spread of this cancer.

Our group’s preliminary findings from the Estonian coast suggest that BTN may already be present here. To investigate further, we will sample clams and mussels from about 20 locations across the Baltic Sea, covering areas with different levels of pollution and shipping activity. At each site, 100300 individuals per species will be examined. We will first screen them under the microscope for signs of cancer, and then use advanced genetic (PCR-based) methods to confirm whether the cancer is transmissible.

With controlled laboratory experiments, we will test whether microplastics and polluted sediments increase cancer transmission or make clams more vulnerable. By exposing healthy clams to infected individuals under different environmental conditions, we will measure transmission rates, disease progression, and immune responses.

Overall, this research will provide the first comprehensive overview of transmissible cancer in the Baltic Sea and help determine whether pollution is driving its spread. The results will improve understanding of marine ecosystem health and may inform future environmental protection strategies.

Supervisor(s): Georg Martin, Jonne Kotta

This thesis addresses the challenges of assessing blue economy impacts in the Baltic Sea by combining in situ experiments, large-scale habitat mapping, and spatialdynamic modelling to evaluate ecosystem responses under different management scenarios. A data-driven algorithm will quantify the cumulative impacts of blue bioeconomy activities and their interaction with existing stressors. Findings will be published in peer-reviewed journals, focusing on artificial substrate colonization, the ecological footprint of blue economy installations, and cumulative environmental impacts of renewable energy and aquaculture in the NE Baltic Sea.

Supervisor(s): Jonne Kotta

European countries face the pressing commitment to protect 30% of their seas by 2030 and ensure that conservation measures deliver the intended ecological benefits by integrating the best available ecological knowledge into their design. Although research increasingly calls for conservation measures to reflect ecological complexity, implemented actions still rely largely on structural descriptors such as species and habitat occurrences, and focus on single or a few species. This overlooks trophic interactions and biogeochemical processes that sustain ecosystem functioning, weakening conservation effectiveness. Food-web and bioenergetic modelling provide promising solutions for quantifying community-wide fluxes of energy and matter, and, when combined with ecological stoichiometry, can translate these fluxes into robust indicators of ecosystem functioning. Yet their use in decision-making remains limited due to their methodological complexity, high data demands, and the lack of end-to-end workflows to produce planning-ready outputs. To bridge these gaps, the PhD project will build on recent advances in bioenergetic food-web and species distribution modelling to quantify and map communitywide and ecosystem functioning indicators, and integrate them into key decision-support approaches, particularly cumulative effects assessment and the prioritization of conservation measures. The thesis will develop a decision-making-ready operational workflow and digital tools, and will implement them in the Baltic Sea to produce planning-ready outputs addressing the following questions: (1) Which areas disproportionately support key ecosystem functioning processes (e.g., nutrient cycling, carbon sequestration)? (2) How well are these areas covered by marine protected areas (MPAs), and how exposed are they to human pressures? (3) How should MPAs be adapted to safeguard these areas?

Supervisor(s): Aveliina Helm

Floodplain meadows, concentrated in the Baltic region, are semi-natural systems whose ecological value derives both from their high biodiversity and from their crucial role in flood regulation, nutrient retention, and shaping biogeochemical cycles at the landscape scale. Despite their priority status within the European Union’s conservation framework, the condition of floodplain meadows in the Baltic region remains unfavorable, and ecologically effective management covers only part of their existing area.

In management and restoration practice, floodplain meadows have largely been treated as botanical units, although their structure and dynamics – and consequently their management and restoration – are clearly influenced by measurable biophysical processes such as flood duration and volume, nutrient availability in soil and water, and microtopography. When these factors are not systematically integrated into management decisions, restoration and maintenance remain largely practice-based and difficult to generalize. The aim of this doctoral thesis is to develop a process-based ecological framework linking the hydrological and physical characteristics of floodplain meadows to their plant and bird communities and to management practices. The central hypothesis is that analyzing biophysical gradients makes it possible to delineate functionally distinct areas, which in turn allow the effects of management and interannual hydrological variability on biodiversity to be described and understood.

The study will quantify the main environmental gradients of Estonian floodplain meadows using high-resolution remote sensing data (including LiDAR), field surveys, and long-term monitoring data on plant communities and birds. The project will enable a shift from a descriptive and static classification of floodplain meadows towards a dynamic generalization that integrates environmental conditions and biodiversity status. The resulting relationships will provide a scientifically grounded basis for evaluating restoration and management measures and will contribute to the development of landscape and land-use models both in Estonia and across the Baltic region.

Supervisor(s): Mehis Rohtla, Markus Vetemaa

Pikeperch (Sander lucioperca) is an important link in an aquatic food chain and one of the most actively managed predatory fish species throughout its range. This is also the case in Estonia, where pikeperch is under considerable fishing pressure in both inland waters and coastal waters. So far, catches have been regulated based on various annual monitoring surveys and information obtained from commercial catch data, but despite this, the status of pikeperch remains poor. The reasons for the poor status are both biotic (human-induced and natural mortality) and abiotic (unfavorable environmental conditions), but another problem is the lack of knowledge about the species' ecology, especially its migration, the impact of recreational fishing, and the formation of strong generations.

To improve the situation, this project will conduct a study of the spawning and feeding migrations between Lake Peipsi and the Emajõgi River, and between Pärnu Bay and the Pärnu River. The extent of migration is being studied using acoustic telemetry and external fish tagging, and spawning ground selection is being described using sonar to create a 3D map of the bottom and the characteristics of the water body. Mapping the most important spawning grounds enables the implementation of targeted fishing restrictions, which are essential for increasing the species' abundance. Tagged fishes are used to estimate fishing mortality in the Emajõgi and Pärnu rivers, helping to better manage fishing in those areas. With the cooperation of recreational fishermen, the birth biome (river or sea) of individuals caught in the Pärnu River is determined using microchemical analyses of otoliths in order to clarify the importance of the Pärnu River in the context of the Pärnu Bay fish stock.

In addition to the above, the abundance of fish may also be affected by the sharp increase in the popularity and efficiency of recreational fishing over the past few years. At present, there is no data on recreational fishing from a fisheries management perspective, and its actual impact on fish stocks remains unknown. Therefore, the second sub-topic of this doctoral thesis is the impact of recreational fishing on the stock in the Peipsi-Pihkva-Lämmijärve and Pärnu Bay and river system. To describe the actual situation, multi-level studies are conducted, which also involve significant application of citizen science. The data obtained will enable the use of statistical models to estimate recreational fishermen's catches during a single fishing season. Furthermore, the mortality of recreational fishermen's discards will also be assessed, as this can also have a significant impact on fish stocks.

Supervisor(s): Mehis Rohtla, Markus Vetemaa, Aleksander Klauson

The increasingly intensive human use of marine areas has led to a substantial rise in underwater noise levels in both offshore and coastal waters. In Estonia, underwater noise and marine acoustics have become increasingly relevant due to the extensive development of offshore wind farms. Sound is one of the primary channels for communication and orientation for aquatic organisms. Depending on species-specific hearing capabilities and the physical properties of the noise, anthropogenic underwater sound may cause auditory damage, induce stress, or alter fish behaviour. Noise associated with offshore wind farms may mask biologically important acoustic cues required for locating conspecifics, prey, and suitable habitats, as well as for predator detection and intraspecific communication. Elevated noise levels may also trigger avoidance behaviour, particularly if affected areas constitute important feeding or spawning habitats or restrict access to such areas. Ultimately, these impacts may negatively affect fish population abundance and the overall state of the marine environment.

The aim of this doctoral project is to assess the effects of low-frequency noise on Baltic herring (Clupea harengus membras) using field experiments and acoustic measurements conducted in both natural and controlled semi-natural environments. As offshore wind farms are not yet operational in Estonian waters, offshore wind turbine operational noise will be simulated using a state-of-the-art underwater sound projector. The project will: (1) evaluate the effects of low-frequency sound simulating wind farm operational noise on spawningmigrating and spawning Baltic herring; (2) assess the effects of such sound on feeding Baltic herring; (3) determine behavioural response thresholds in relation to low-frequency sound pressure; and (4) quantify behavioural response thresholds to particle motion, a critically important yet poorly studied component of underwater noise for fish sensory systems.

Supervisor(s): Marina Semchenko, Siim-Kaarel Sepp, Tsipe Aavik

In a process known as plant-soil feedback (PSF), plants modify soil microbial communities in ways that differentially affect the survival and growth of subsequent generations of plants. These feedbacks range from positive (mutualistic) to strongly negative (pathogenic). Several studies have demonstrated the importance of negative PSFs in explaining plant species co-existence. However, while plant-microbial interactions also vary between genotypes within species, the role of PSF in plant population dynamics and local adaptation is rarely assessed. Intraspecific plant-soil feedback can become particularly important under global change, where genetic variation determines adaptive potential. The aim of this PhD project is to enhance our understanding of how global change affects intraspecific plant-soil feedback via modulation of mutualistic and pathogenic interactions with soil fungi. The project will focus on agricultural, semi-natural grasslands undergoing land use change and experiencing increasing pressures from extreme weather events such as droughts and fluctuating winter conditions. This aim will be achieved by i) examining plant-soil feedback along the gradient of land use abandonment where grassland species experience declines in population size and functional diversity; ii) assessing the interactive effects of fertilisation and drought on plant-soil feedback in plant populations and microbial communities originating from historically unfertilised or intensively managed, fertilised grasslands; and iii) investigating the mechanistic basis of changes in plant-soil feedback in response to summer and overwintering conditions in agricultural soils that are either dominated by pathogenic or mycorrhizal fungi. By linking plant population dynamics with plant–soil interactions under land-use and climate change, this project will clarify how soil microbes influence plant adaptation, ecosystem stability, conservation, and sustainable agriculture.

Supervisor(s): Siim Pikker, Veikko Pentti Linko, Arvi Freiberg

Optical cavities, particularly planar Fabry–Perot (metal–dielectric–mirror) resonators, are key platforms in molecular strong coupling, polaritonics, and quantum optics. By enabling hybridization of light with excitonic, plasmonic, and vibrational excitations, these systems provide access to the strong-coupling regime, where optical and energetic properties can be profoundly modified. Such cavity-mediated effects underpin emerging applications in polaritonic light sources, cavity-controlled energy transfer, and polaritonic chemistry, including reaction-rate engineering.

This project builds on our group’s established expertise in cavity optics, polariton physics, strong lightmatter coupling, and DNA origami–based nanosystems. We will further develop a grayscale UVlithography technique for the rapid, parallel fabrication of Fabry–Perot microresonators. The method enables high-throughput production of dye-doped cavities, increasing fabrication yield from a few devices per day to thousands, thereby substantially accelerating experimental studies.

The work will focus on improving fabrication accuracy and defect control, optimizing photoresistemitter combinations (including quantum dots, molecular dyes, photosynthetic proteins, and DNA origami FRET systems), and extending the approach to hollow, variable-thickness cavities for liquidphase measurements. Advanced Fourier-plane spectroscopy will be employed to characterize the resonators and evaluate their performance in optical and vibrational strong-coupling experiments, with emphasis on spectral modification and cavity-controlled energy-transfer processes. As a result of this doctoral project, at least three scientific publications are planned. The present doctoral project is associated with two PRG projects that commenced in 2026.

Supervisor(s): Veiko Palge, Juhan Matthias Kahk, Dirk Oliver Theis

Quantum computation is a novel paradigm of computing which is capable of solving problems that remain intractable for conventional computing. In particular, quantum computers are able to efficiently simulate large quantum systems, a task beyond the reach of conventional computers, regardless of their scale. This is of great interest for quantum chemistry and materials science, which underlie chemical and pharmaceutical industries and development of new materials and technologies. While fault tolerant quantum computing is predicted to arrive earliest in a decade, near term quantum devices which involve a few hundred qubits are becoming a reality presently. The project focuses on simulation of quantum systems that are of interest in quantum chemistry and solid state physics using the paradigm of variational quantum algorithms, as well as fault tolerant quantum computation. Analysis, benchmarking, testing and development of the range of methods that are used for simulation and computation of electronic structure and energy surfaces need to be carried out in order to make progress in quantum simulation on near term quantum devices.

Supervisor(s): Velle Toll

Do anthropogenic aerosols affect Earth’s climate by acting as ice-seeding particles? Our recent work revealed that certain industrial aerosols seed ice, suggesting that an overlooked climate-forcing mechanism may exist. This project develops the first-ever global database of anthropogenic aerosol sources seeding ice and calculates the associated impact on Earth’s climate. For this, the work combines long-term satellite records, atmospheric dispersion modelling, and laboratory experiments. Aerosol dispersion modelling will allow to automatically identify ice-seeding at 30 thousand industrial facilities.

The ice seeding at isolated localised industrial sites will be compared with ice seeding around megacities and larger industrial regions at climate-relevant scales: over multiple decades at spatial scales spanning hundreds-by-hundreds of kilometres. To quantify the climate impact by ice-seeding aerosols, we calculate it relative to the climate impact by aerosols serving as seeds for liquid clouds. By this, we address a major open question in climate science: are current climate predictions missing an important anthropogenic climate-forcing mechanism?

Supervisor(s): Laur Järv, Mercè Guerrero Román

The first direct image of a supermassive black hole in 2019 has been one of the most impressive scientific achievements not only for the gravitational research community but also for the general public, appearing in newspapers and on TV news. Understanding the physics behind this image is crucial to gathering more information about our universe and the theories governing it.

The equations of gravity describe how matter curves spacetime and how particles move within it. When these equations fail to explain observations fully, we can question either our assumptions about matter or the theory of gravity. The absence of direct evidence for dark matter and energy in explaining cosmic dynamics, for example, motivates exploring theories beyond Einstein’s general relativity. These lead to different equations and, therefore, also different black hole solutions, which leave distinct imprints on accretion dynamics and photon trajectories. In this way, black holes serve as natural laboratories for testing gravitational theories through their observable images. However, using toy models to reduce the complexity of the problem may introduce degeneracies between the effects of spacetime geometry and accretion disk physics, potentially leading to misinterpretation of the results.

Establishing a solid foundation for rigorously testing general relativity and theories beyond that against the current and future black hole observations requires a deep understanding of how gravity influences all physical processes encoded in the image. The objective of this PhD project is to make systematic steps to assess the impact of theories beyond general relativity on (1) the geometry of spacetime as manifested in the “shadow” of the image, (2) the matter dynamics and the feasibility to develop semianalytic models of the accretion disk, and (3) the potential signatures in the polarisation patterns of the observational data.

Supervisor(s): Mari Sepp

Autism spectrum disorders (ASD) rank among the most common neurodevelopmental disorders globally. Hundreds of ASD risk gene have been identified, but interpreting their functional significance remains challenging. Cumulative evidence implicates deficits in the cerebellum in ASD aetiology, but a systematic characterisation of the disease mechanisms in the cerebellum is missing. In the planned PhD thesis projects, we will build a comprehensive map of cell-level ASD phenotypes in the developing cerebellum using single-cell genomics, spatial mapping, and sparse labelling of cellular morphologies. We will characterise the molecular, cellular, anatomical and morphological phenotypes in the cerebella of ASD mouse models, and map the phenotypes elicited by perturbations of ASD risk genes in human induced pluripotent stem cell-derived cerebellar organoids. These approaches will enable the identification of points of convergence and phenotype-driven ASD subgroups, and improve our understanding of disease mechanisms.

Supervisor(s): Kelli Lehto, Reedik Mägi, Jaanika Kronberg

This PhD project is part of a larger interdisciplinary ERC-funded programme to understand the causal mechanisms underlying adult attention deficit and hyperactivity disorder (ADHD) mechanisms. While ADHD is a childhood-onset, highly heritable neurodevelopmental disorder, recent years have seen a marked increase in adult diagnoses, especially following the COVID-19 pandemic. Notably, adult ADHD features — such as inattention, impulsivity, and emotional dysregulation — may also be influenced by environmental factors like stress, sleep deprivation, substance use, screen time, and post-viral effects. The current understanding of the environmental influences on adult ADHD remains limited, particularly due to challenges in disentangling causality from confounding.

This project will systematically assess the role of environmental exposures (e.g. childhood adversity, life stress, substance use, screen use, COVID-19 severity) on adult ADHD features, leveraging innovative genomics-informed approaches. The research will employ polygenic scores and longitudinal analyses, as well as parent-child trio designs, to disentangle genetic and familial confounding and true causal effects. The PhD candidate will have access to the extensive registry-linked Estonian Biobank (over 211,000 participants, including 86,000 with detailed mental health data and 11,500 parent-child trios), the UK Biobank (500,000 participants) and opportunities for collaboration with major international cohorts and consortia. The project involves three core themes: mapping environmental risks for adult ADHD across age and gender; quantifying genetic confounding and mediation effects; and investigating familial confounding using advanced genetic epidemiological methods. This work aims to clarify the interplay between genes and environment in adult ADHD representation, with broad implications for research and practice.

Supervisors: Burak Yelmen, Flora Jay

Despite major advancements in genome-wide association studies and the increasing availability of diverse datasets, the underlying genetic mechanisms of complex traits remain largely elusive. Recent deep learning applications enable predictive and generative modeling of high-dimensional genomic data, yet the black-box nature of these models limits their biological utility. In this project, we propose to bridge the gap between interpretability and generative neural networks by introducing a comprehensive framework for modeling the interactive genomic landscape of complex traits. By developing domain-specific architectures and interpretability methodologies, we aim to capture multi-locus and multi-phenotype structure, generate realistic synthetic cohorts, and identify key genomic positions. This framework will provide a holistic approach to complex trait genetics, offering novel insights and ultimately advancing precision medicine.

Supervisor(s): Urmo Võsa, Priit Palta

The Estonian Biobank (EstBB) links genomic, metabolomic, electronic health records (EHR), and drug-usage data for over 200,000 participants, yet the full potential of these data modalities for disease prediction remains underused. This PhD project will improve trait and disease prediction by integrating machine learning and modern AI with complementary data layers - molecular QTL resources, multi-omics measurements, and longitudinal clinical histories. The work will also combine large-scale QTL datasets (eQTL, pQTL, metabolite and splicing QTL) with high-powered GWAS summary statistics to prioritise trait-relevant molecular features via colocalisation, Mendelian randomisation, and genetic correlation analyses. These functional signals will inform both refined PRS and extended models that incorporate QTL-derived genome-wide annotations and QTL-based polygenic scores for intermediate molecular traits. Finally, the project will develop a multimodal generative AI model that integrates PRS and omics with longitudinal EHR event sequences to forecast future health states as trajectories rather than static risk estimates. These models will be trained in EstBB and externally validated in UK Biobank data, enabling more personalised, context-aware prediction of disease onset, progression milestones, readmissions, and projected healthcare costs.

Supervisors: Alena Kushniarevich (Institute of Genomics, UT (UTIG), Lehti Saag (Institute of Genomics, UT (UTIG), Kristiina Tambets (Institute of Genomics, UT (UTIG), Georgi Hudjashov (Institute of Genomics, UT (UTIG)

Recent advances in genomics and bioinformatics have enabled the integration of large-scale modern biobank data with time-stratified ancient human genomes, providing powerful tools for studying human evolutionary history and the genetics of complex traits. Improvements in genotype imputation for ancient genomes have substantially increased analytical power, facilitating more robust investigation of the evolutionary origins of modern genetic variation and disease susceptibility. By combining ancient genomic data with modern genome-wide association studies and polygenic scores, this doctoral project examines how the genetic architecture of complex traits has changed over time and aims to characterise the evolution of complex traits in the Eastern Baltic region over the past two thousand years through joint analysis of ancient and contemporary human genomes.

Supervisors: Erik Abner, Prof. Elin Org

The aim of this doctoral project is to investigate how exposure to infectious diseases, immune system function, and the human microbiome interact with host genetics to influence the risk of chronic diseases. Using genetic data from the Estonian Biobank, microbiome DNA sequencing data, serological profiles from biobank participants, and electronic health records, the project seeks to identify biological links between virome composition, antibody profiles in blood, and a wide range of chronic health traits. The project will apply bioinformatic and statistical genetics approaches to characterise the gut virome, building on existing microbiome DNA sequencing data. Virome features, including their presence, abundance, and diversity, will be quantified and analysed in relation to bacterial community structure, host genetic variation, and chronic disease phenotypes derived from electronic health records. This will enable the identification of virome patterns associated with long term health outcomes, as well as shared genetic factors influencing host-microbiome interactions.

In parallel, phage-based immunoprecipitation sequencing (PhIP-seq) will be established to detect antibodies against a broad range of infectious agents in plasma samples from the Estonian Biobank. These data will allow population level assessment of past infectious exposure and immune responses and enable analyses of seroprevalence patterns and their associations with chronic health traits. Integrating serological data with microbiome and virome features in overlapping samples will make it possible to assess how immune history, microbial ecosystems, and host genetics jointly shape the risk of complex diseases.

Combining these different layers of health-related data will provide a population-based framework for studying disease mechanisms linked to infections. The results of this project will improve our understanding of the biological links between infectious exposure and chronic diseases, and support future research in risk assessment, prevention, and the development of personalized medicine approaches.

Supervisor(s): Anneli Kährik

The PhD project explores the eco-social challenge of deep housing, how to achieve housing renovation without reducing housing affordability. Estonia needs to significantly reduce the greenhouse gas emissions, with buildings accounting for 53% of final energy use and requiring a 60% emission reduction by 2030. However, rising housing costs and financial barriers create inequalities in access to renovation support, often favoring well-organized housing associations with low share of vulnerable households. The research will capitalize on Estonia’s comprehensive housing and population data infrastructure developed within the Infotechnological Mobility Observatory. It aims to assess the spatial and social dynamics of the need and actual housing renovation, examine perceptions of key stakeholders through survey and qualitative studies to shed new light on the social and psychological factors that may facilitate residents’ and communities’ engagement in the renovation process and develop an innovative framework to facilitate buildings’ renovation. The results of the PhD project will serve as input for the work of the Ministry of Climate's housing policy and building renovation steering group. Three publications will emerge from this research: (1) an analysis of spatial and demographic changes linked to renovation activities, (2) an evaluation of residents’ and stakeholder perspectives on equitable access to renovation opportunities, and (3) an innovative framework for pathways to renovation. By addressing these issues, the PhD work will help shape policies that balance environmental goals with housing affordability in Estonia.

Supervisor(s): Kadri Leetmaa, Ingmar Pastak

This PhD project investigates digital transformation from the perspective of ageing rural communities. Although digitalisation research has predominantly focused on urban “smart city” agendas, significantly less attention has been paid to rural regions and older populations. The project aims to address this gap by examining how digital transformation unfolds in contexts marked by demographic ageing, population decline, and spatial peripherality. It seeks to further develop the concept of “double exclusion”, conceptualising the intersecting conditions of exclusion while remaining open to diverse local experiences and outcomes. Empirically, the research explores how older rural residents encounter, interpret, and engage with digital change, and how local social infrastructures—such as libraries and other community-based institutions—mediate these processes. Particular attention is given to the gendered dimensions of digital engagement. By shifting the analytical focus from metropolitan innovation narratives to peripheral settings, the project aims to generate empirically grounded insights into the varied trajectories of digital transformation in rural societies.

Supervisor(s): Taavi Pae, Raivo Aunap, Holger Virro

The cartographic collection of the National Archives of Estonia is exceptionally rich, comprising over 150,000 maps, approximately one third of which have been digitised. The 17th- to 20th-century cartographic record in Estonia — from the Great Swedish Cadastre through manor plans to national topographic series — provides detailed depictions of fields, forests, wetlands, infrastructure and place names. Historical large-scale maps are a key source for reconstructing past land use, settlement patterns and cultural landscapes; however, most archival holdings exist only as scanned images, which limits their use in quantitative spatial analysis and long-term environmental research. Systematic examination of historical maps for information extraction is labour-intensive and demands considerable expert effort. In recent years, artificial intelligence and machine learning methods have advanced rapidly, offering new opportunities for automated analysis. The aim of this doctoral thesis is to develop and evaluate an automated GeoAI-based workflow capable of recognising historical map content, learning its hierarchical cartographic structure, and classifying mapped objects into machine-readable, georeferenced spatial datasets. These outputs will enable the integration of historical spatial information with modern spatial databases, supporting a wide range of GIS applications, particularly landscape-change and palaeogeographical analyses. The project addresses methodological challenges arising from stylistic variability, symbol overlap, sparse graphical representations of area features, and the physical degradation of historical source materials. Special emphasis will be placed on quantifying classification uncertainty and analysing error patterns to ensure scientific robustness. By transforming complex historical cartographic imagery into structured semantic spatial data, the project significantly enhances the usability of archival map collections and contributes to long-term landscape and environmental history research.

Supervisor(s): Jaan Pärn, Ülo Mander

Organic‑rich wetlands store vast global reserves of carbon and nitrogen, yet their long‑term stability depends on the molecular composition of soil organic matter (SOM) and its response to hydrology, climate, and land use. Despite this importance, global understanding of how dry‑season water‑table minima shape SOM chemistry, how hydrology interacts with trophic status to influence N/S/P‑bearing compounds, and which molecular markers indicate restoration progress remains limited. This doctoral project addresses these gaps using a uniquely coherent global dataset of organic soils collected from 80 sites across tropical, temperate, and boreal regions between 2011 and 2024. Each site includes natural and drained conditions and represents major hydrotopes from fens and swamps to ombrotrophic bogs. A subset of 156 samples was analysed at the Max‑Planck‑Institute for Biogeochemistry using Orbitrap ultrahigh‑resolution mass spectrometry, enabling detailed characterization of SOM molecular composition.

The study has three objectives: (1) determine how dry‑season hydrology influences SOM oxidation state, aromaticity, and thermodynamic properties; (2) identify global patterns linking hydrology, trophic gradients, and heteroatom‑rich compounds; and (3) develop early‑warning molecular indicators of wetland restoration.

The outcomes will advance fundamental understanding of SOM persistence and provide practical tools for assessing degradation, guiding rewetting efforts, and informing global conservation strategies.

Supervisor(s): Kaido Soosaar, Alisa Krasnova

Forests offset nearly one third of anthropogenic CO₂ emissions, yet the climate impact of forest management remains insufficiently constrained. Although managed forests are central to LULUCF accounting, forest C assessments globally still rely largely on either ecosystem-scale eddy covariance (EC) flux measurements or stock-based inventories, with limited methodological integration. These approaches are seldom systematically reconciled, leading to persistent discrepancies in net ecosystem C balance (NECB) estimates and limiting confidence in emission factors used for climate policy.

The aim of this doctoral study is to advance understanding of C cycling in managed boreal and hemiboreal forests by integrating biometeorological and biometric approaches within a unified NECB framework. Continuous EC measurements of CO₂ exchange (NEE, GPP, Reco) will be analysed together with soil chamber measurements of CO₂, CH₄ and N₂O fluxes and detailed assessments of above- and belowground C stocks derived from forest inventories, litterfall monitoring and soil C analyses. By systematically linking flux-based and stock-based estimates across contrasting management regimes, the study seeks to reduce structural uncertainty in ecosystem C budgeting.

Supervisor(s): Evelyn Uuemaa, Triin Reitalu

Satellite data are widely used to monitor ecosystems at regional and national scales. In grasslands, they support vegetation productivity assessment, habitat evaluation, and land-use planning. Most current approaches rely on vegetation indices such as NDVI or statistical models linked to field data. While useful, these methods often reduce complex ecological systems to a few summary indicators and may not fully capture biodiversity patterns, species composition, or ecological dynamics such as seasonal variability and management effects. Recent advances in artificial intelligence have introduced new geospatial foundation models that learn directly from large volumes of multi-temporal satellite imagery. By analysing repeated observations over time, these models can detect spatial structure and seasonal patterns without requiring extensive labeled training data. This makes them promising tools for large-scale environmental monitoring. However, research so far has mainly focused on how accurately these models predict specific variables, rather than on understanding what ecological information their internal representations (embeddings) actually contain. This PhD project investigates what modern artificial intelligence models capture about grassland ecosystems in Estonia. Instead of focusing only on prediction performance, it examines whether the learned representations reflect meaningful ecological gradients, such as biomass, seasonal variation, biodiversity, and management intensity. The project also explores how these relationships vary across spatial scales, from individual plots to entire landscapes, providing new information on how AI-based models represent ecological structure.

Supervisor(s): Mikk Espenberg

Peatlands are among the most important natural carbon stores, yet drained and agriculturally managed peatlands have become significant sources of greenhouse gases (GHGs). Although peatlands cover only about 3% of the global land surface, nearly one‑fifth are degraded, contributing more than 5–30% of global anthropogenic GHG emissions. The microbial processes that regulate the production and consumption of key GHGs (e.g., CH4 and N2O) remain insufficiently quantified. In addition to soil‑based processes, the aboveground plant‑associated microbiome (phyllosphere) plays an important role but is not accounted for in current emission estimation methods. National inventories rely largely on the IPCC Tier 1 methodology, which uses default emission factors and leads to high uncertainty. With climate change expected to increase microbial activity through rising temperatures, more accurate modelling becomes crucial. This creates a need for higher‑tier IPCC approaches: Tier 2, which provides country‑ or region‑specific emission factors, and Tier 3, which relies on process‑based models. This doctoral project develops a hybrid modelling framework combining a mechanistic ecological model with machine learning. The mechanistic component captures the influence of environmental conditions (e.g., temperature, moisture), while the data‑driven component learns complex GHG dynamics based on microbial functional gene ratios and microbiome structure. The outcome is a Tier 3 hybrid model and corresponding simplified Tier 2 equations suitable for use in national GHG inventories.

Supervisors: Ülo Mander, Jaan Pärn

Peatlands cover only 3% of the land area, but contain up to 1/3 of the soil carbon and 20% of the nitrogen (N). Drainage and climate warming cause significant C and N losses both as greenhouse gas (GHG) fluxes and leaching. In addition to carbon dioxide (CO2) and methane (CH4), another important GHG is nitrous oxide (N2O), which has a radiative effect 273 times greater than CO2 and is the main depleter of the stratospheric ozone layer. Drained wetlands, together with agriculture, are the main sources of N2O. Due to the complex relationships between N2O fluxes and environmental factors and the irregular temporal dynamics, there are still no models that adequately describe N2O fluxes that would allow for the prediction of the emission and the design of measures to reduce it. The University of Tartu has a unique database that contains in-situ measured GHG (including N2O) fluxes and soil microbiome data and important environmental parameters from 68 regions around the world from 2011-2025 by the working group. In addition, we integrate data collected in the working group’s experimental areas and similar measurement data from the temperate zone (15 areas in Estonia) and the tropics (Malaysia, Peru, Congo, Uganda and Réunion, a total of 14 areas). Literature data is also added. As a result of the integration of these data, an empirical PeatN2O model will be developed, taking into account similar models for modeling CO2 and CH4 fluxes (Landscape-DNDC, Forest-DNDC; ECOSSE, JULES, etc.). We will identify the process chains of N2O balance by combining microbiological, isotope analysis and reactomics methods. A global peatland N2O emission model will be constructed by integrating the process-based PeatN2O model with MODIS and SENTINEL satellite data (GPP, NDVI, soil N, soil water regime). The work expands the understanding of peatland N cycles biogeochemistry, which is the basis for their protection and environmentally sustainable management.

Supervisor(s): Ain Kull

Permanently wet and carbon-rich (functioning) mires are key ecosystems for climate and water regulation, and sound land use planning, national greenhouse has (GHG) reporting and meeting national objectives (LULUCF, ecosystem accounting, nature restoration, etc.). Precise mapping of peatlands functioning as wetlands is crucial, as these ecosystems contribute significantly to GHG regulation, nutrient cycling and flood buffering. Peatland mapping has traditionally been based on categorizing land cover based on the appearance of communities. Although modern remote sensing and machine learning methodologies are also used, this approach ignores or misclassifies many peatlands that actually function as mires hydrologically and biogeochemically, especially specific transitional or successional communities, peatlands affected by drainage or certain land use, etc. Thus, the impact of their actual distribution and status on climate, land use planning decisions, and GHG reporting is also not considered with sufficient accuracy.

The aim of this PhD study is to develop a remote sensing-based framework for mapping the actual distribution and status of peatland ecosystems, distinguishing those that function as permanently wet peatlands (mires), as an alternative to the conventional land cover categorization based on the appearance of communities. Rather than treating peatlands as static land cover classes, the PhD project conceptualizes them as dynamic systems whose responses to disturbances reflect their ecological status, hydrological functioning, and exposure to external pressures. The PhD project explores the possibilities of remote sensing-based identification of functioning peatlands based on responses in soil moisture, hydrological and vegetation patterns in response to climatic and anthropogenic disturbances (e.g. extreme precipitation, drought, drainage, rewetting).

Supervisor(s): Kalle Kirsimäe

This PhD project addresses the growing challenge of securing supplies of critical raw materials (CRMs) essential for modern, technology-driven economies, with a particular focus on cobalt (Co) and rare earth elements (REEs). Demand for these metals is rapidly increasing in the European Union and globally due to their key role in the green and digital transition, while primary production is highly geographically concentrated, exposing supply chains to geopolitical and economic risks. In response, recycling and valorization of secondary raw materials have emerged as strategic tools for improving CRM supply security, in line with EU policy frameworks. The project investigates the potential of secondary raw materials, including mining and ore-enrichment tailings, industrial slags, combustion residues, and other industrial by-products, as alternative sources of Co and REEs. Particular attention is given to cobalt associated with manganese ores and Fe–Mn oxyhydroxide phases, where Co can become enriched through coprecipitation and sorption processes, and to REEs occurring in industrial wastes such as coal fly ash, red mud, and phosphate-processing residues.

Supervisor(s): Raul Paat, Andres Marandi

Climate change and long-term land use modification increasingly influence hydrological regimes, surface water groundwater interactions, and water balance in catchments across Estonia. Peatlands are among the most sensitive landscape elements in this context, as drainage and land use change have altered water storage and flow pathways. At the same time, climate driven changes in precipitation patterns and the increasing occurrence of hydrological extremes affect the functioning of entire catchments.

The aim of this doctoral thesis is to improve the understanding of hydrological processes in pilot catchments representing diverse geological, hydrological, and land use conditions in Estonia. The research focuses on changes in water regime and water balance under climate and land use change, while treating peatlands as integral components of catchment scale hydrology rather than isolated systems.

The study develops process-based surface water models with different levels of complexity for three pilot catchments and complements them with geospatial machine learning approaches to estimate water table dynamics and hydrological responses. A key objective is the comparative evaluation of modelling approaches to identify robust and transferable methods suitable for climate change and land use change impact assessments.

The results of the doctoral research provide a scientific basis for selecting appropriate hydrological modelling approaches and support evidence-based decision making in water management, climate adaptation, and peatland restoration planning. The outcomes contribute to national climate adaptation strategies, water management planning, and the achievement of LULUCF related objectives in Estonia.

Supervisor(s): prof Alar Rosentau

During the doctoral project, novel datasets on Holocene (the last 11,700 years) sea-level changes and coastal paleoenvironments will be collected from the seabed and coastal areas of Pärnu Bay, allowing refinement of the ages and amplitudes of Baltic Sea transgression events. Unique fragments of coastal landscapes preserved on the floor of Pärnu Bay—such as rivers, estuaries, lagoons, and submerged forests—provide an opportunity to investigate lower-than-present Holocene sea levels and the subsequent Baltic transgressions.The sea-level change data collected within the project will be incorporated into the HOLSEA global sea-level database, and the submerged-landscape data will be included in the EMODnet Geology pan-European dataset. Together, these integrations enable more regionally detailed syntheses concerning coastal landscapes drowned during major transgressions.

Supervisor(s): Helger Lipmaa

The doctoral project “Novel techniques in zk-SNARKs” focuses on researching and developing new methods for modern zero-knowledge proof systems (zk-SNARKs). The goal is to study and improve state-of-the-art zk-SNARK techniques that are relevant both for leading academic venues (e.g., EUROCRYPT, CRYPTO, ZKProof) and for real-world deployments in the ZK industry. The research addresses key contemporary SNARK paradigms, including Plonkish protocols and the design space of polynomial commitment schemes (KZG, IPA, FRI and related approaches), as well as recursion, proof composition, and advanced security notions such as knowledge soundness and simulation extractability. While the project is naturally connected to the supervisor Helger Lipmaa’s strong research background in cryptography and zero knowledge, the research directions are not restricted to the supervisor’s prior topics. The expected outcome is the development of novel theoretical and practical techniques that improve the security, efficiency, and scalability of zk-SNARKs, enabling publications in top-tier cryptography and security conferences.

Supervisor(s): Dirk Oliver Theis, Veiko Palge

This PhD project investigates fault-tolerant quantum computing (FTQC) using large-scale stabilizer simulations to analyze protocol performance under realistic noise assumptions. Leveraging the candidate's existing expertise in tools like Stim and Sinter, the project skips the standard literature review to maximize the 48-month window for high-impact research. The work focuses on three pillars: logical memory scaling, the stability of QEC protocols under hardware perturbations, and the impact of lattice surgery on error rates. The project is structured to ensure that three peer-reviewed journal articles are submitted and accepted by Month 42, providing actionable insights into the design of scalable, fault-tolerant quantum architectures.

Supervisor(s): Kairit Sirts

Large language models are increasingly used in text generation-based applications such as question answering, summarization, content creation, and decision support. At the same time, it remains difficult to assess how well these models perform in low-resource languages such as Estonian, particularly in open-ended text generation tasks. Existing evaluation methods often focus on narrow task-specific benchmarks or multiple-choice tests, which provide only a limited view of model behavior in realistic usage scenarios. This doctoral project focuses on developing a systematic evaluation framework for open-ended text generation in large language models. Rather than treating generative quality as a single score or as the outcome of isolated tasks, the project conceptualizes it as a multi-dimensional phenomenon that encompasses linguistic quality, task-level capabilities, and contextual and cultural appropriateness. In addition to proposing a conceptual framework, the research aims to develop concrete evaluation resources and methodologies grounded in this framework. The evaluation framework will be developed and empirically validated using Estonian as a representative low-resource language. The work includes adapting existing evaluation resources, developing new benchmark datasets, and conducting structured human evaluations based on clearly defined criteria. Furthermore, the project examines how results from automated evaluations and human judgments relate to real-world model usability, such as decisions concerning model comparison, selection and deployment. While the empirical focus is on Estonian, the proposed framework is intended to be generalizable and applicable to other languages as well.

Supervisor(s): Marek Oja and Kerli Mooses

Treatment guidelines aim to improve quality of care, patient safety, and efficient use of healthcare resources. However, guidelines are typically published as narrative or semi-structured documents that are not machine-readable, limiting their use in large-scale analytics and real-world evaluation. As a result, guideline adherence assessment relies mainly on aggregated statistics, which fail to capture the timing, sequence, and complexity of real patient treatment trajectories. This doctoral research aims to develop and evaluate AI-based methods for transforming treatment guidelines into structured, computable treatment pathways and applying them to guideline adherence assessment and risk modeling. Large language models and AI agents will be tested to interpret guideline text, align guideline-defined pathways with real-world patient trajectories, and help to explain adherence to treatment and deviations from recommended treatment. In addition, polygenic risk scores will be integrated to assess the added value of genetic information in treatment pathway-based risk prediction.

Supervisor(s): Dmytro Fishman

This PhD project aims to develop and validate an anatomy-aware foundational AI model for CT that enables much faster creation of multi-organ tumour detection models. The key contribution is a reusable “base model” trained on large-scale, heterogeneous CT data to reduce both training time and the need for expert annotations while maintaining clinically relevant performance. The work is innovative because it moves beyond organ-by-organ development by introducing CT-specific pretraining that captures 3D anatomy and lesion patterns, efficient adaptation strategies (fine-tuning/prompting), and evaluation aligned with clinical decision-making. The project will (1) benchmark existing foundation models against strong supervised baselines, (2) develop and test a COMPASS-based anatomy-guided pretraining strategy, and (3) demonstrate speed-up and annotation savings when building new organ models from the foundation model. The candidate, Salme Ussanov, already has strong CT experience through her Master’s thesis and an ongoing publication, making her well-positioned to execute the project successfully.

Supervisor(s): Amnir Hadachi

The transition toward sustainable urban mobility requires not only infrastructure changes but also a deeper understanding of how travel choices are shaped by daily routines, social context, and behavioral constraints. Although large volumes of mobility data are available, existing approaches often estimate mode-shift potential using simplified assumptions that overlook these factors. This PhD project develops a context-aware machine learning framework to model the realistic potential for shifting from car-based travel to sustainable mobility modes. By learning from high-resolution smartphone trajectories, survey-derived mobility profiles, and contextual urban data. Moreover, the project jointly models structural feasibility and behavioral adoption likelihood, enabling data-driven, explainable estimates of sustainable mobility potential across individuals and social groups.

Supervisor(s): Naveed Muhammad, Arun Kumar Singh

Autonomous vehicles and robots must anticipate how nearby agents—such as pedestrians and other vehicles—are likely to move so they can plan safe and comfortable trajectories. Yet a central scientific question remains unanswered: how accurate do these predictions actually need to be for effective motion planning? Humans navigate complex environments successfully without computing precise trajectories; instead, they infer others’ general intent and react smoothly to changes. This project explores whether autonomous systems can adopt a similar strategy—placing greater emphasis on understanding intent rather than on minimizing geometric prediction errors—thereby improving robustness, safety, and interpretability.

This project investigates a framework in which prediction models are optimized together with the motion planner that uses them. The research will quantify how prediction fidelity influences planning quality, then develop intent centric, end to end models designed explicitly for downstream navigation tasks. The methodology spans controlled simulation studies, high precision indoor robot trials, and real-world deployment on an autonomous test vehicle. Expected outcomes include new theoretical insights into prediction–planning coupling, improved navigation performance for robots and autonomous vehicles, and open-source tools that benefit the broader research community.

Supervisor(s): Vesal Vojdani

Many important software systems are multilingual: user-facing logic is often written in a high-level language, while performance-critical parts (libraries, runtimes, glue code) are implemented in C. The correctness of the C code is crucial for the safety guarantees of higher-level languages. For instance, OCaml integrates C libraries via a foreign function interface (FFI). The required C stubs must respect subtle rules for interacting with OCaml's garbage collector. Violations can lead to rare and hard-to-debug defects that survive for years and only manifest under specific calling contexts, heap shapes, or thread schedules.

This PhD project develops sound static analysis methods for verifying such critical C components in context: the analyzed component may be called from partially unanalyzed client code, potentially written in another language. The work (1) advances analysis of OCaml-C stubs by incorporating OCaml-side type information and developing methods to model calls from unknown OCaml code; (2) designs new abstract domains for critical concurrency idioms that rely on space separation (per-thread data, thread pools) and require symbolic, parametric reasoning about unbounded numbers of threads and their heaps; and (3) combines these results for the multi-core setting, where parallel OCaml domains may invoke C stubs concurrently, introducing data races on both the OCaml and C heaps

The expected outcome is new abstract interpretation theory and an open-source prototype evaluated on real-world libraries.

Supervisor(s): Piret Luik Marili Rõõm

Database courses are core courses of computer science education, yet large student cohorts make assessment time-consuming and limit the quality of feedback provided to learners. Although automated assessment (AA) systems are widely used, they predominantly focus on checking the final correctness of SQL queries and provide limited support for more complex database objects such as views, functions, procedures, and triggers. In addition, the development and maintenance of AA often require programming expertise, which restricts their broader adoption by instructors.

The aim of this doctoral project is to design, implement, and empirically evaluate an integrated AA system for database courses that combines formal correctness checking with AI-assisted feedback and structured troubleshooting. The project builds on the existing Silmused AA system used at the University of Tartu and extends it with an AI-assisted feedback module and an NLP-supported instructor authoring interface. AI is employed to generate explanatory and pedagogically meaningful feedback, while correctness evaluation remains grounded in deterministic database execution.

The research follows a design-based research methodology and is conducted in authentic educational settings. The proposed system integrates automated assessment, adaptive troubleshooting through question–hint sequences, and a no code authoring workflow for instructors. The anticipated outcomes include improved student learning through timely and actionable feedback, reduced instructor workload, and transferable design principles for AI-enhanced automated assessment in computer science education.

Supervisor(s): Mubashar Iqbal, Raimundas Matulevičius

This research proposes a Cognitive Digital Twin (CDT) framework for explainable and causal anomaly detection in complex, dynamic, and safety-critical systems. The conventional data-driven anomaly detection methods often lack interpretability and causal insight. Thus, the objective of this work is to overcome these limitations by embedding cognitive capabilities and domain knowledge into digital twin architectures. The proposed framework integrates real-time system data with interpretable models of system behaviour, enabling the detection of anomalies while providing transparent explanations of abnormal conditions and their underlying causes. The framework adopts a hybrid methodology that combines machine learning, symbolic reasoning, and causal inference to support both predictive accuracy and explainability. By explicitly modelling causal relationships among system components, the CDT framework enables root-cause analysis, counterfactual reasoning, and actionable decision support. The expected contributions of this research include a unified framework for causal and explainable anomaly detection, novel techniques for anomaly attribution and reasoning, and empirical validation in representative industrial or safety-critical application domains.

Supervisor(s): Faiz Ali Shah, Kallol Roy, Vesal Vojdani

Software vulnerabilities are defects that cause security failures, enabling attacks on confidentiality, integrity, or availability. Poor source code quality is a major contributor to such vulnerabilities. Static analysis tools are widely used for early vulnerability detection because they do not require code execution, but they suffer from false negatives and false positives, limiting their practical effectiveness. To address these limitations, the work proposes using autonomous, agentic AI and outlines key research questions comparing its performance to static tools, its ability to analyze unfamiliar proprietary code, and its effectiveness in generating safe, context-aware patches. The PhD project introduces a novel Agentic AI pipeline based on the hypothesis that geometric voids in code embedding spaces correspond to vulnerabilities. The pipeline transforms code into an Abstract Syntax Tree, generates embeddings, detects topological voids, and uses a reinforcement learning agent to locate them. To handle sparse rewards and complex search spaces, a “supermoves” technique accelerates exploration. Detected vulnerabilities are mapped back to code and explained through executable test cases.

Supervisor(s): Anastasija Nikiforova

Artificial Intelligence is rapidly transforming organisational and public sector practices, yet existing approaches to Responsible AI and Sustainable AI remain largely aspirational and fragmented. Ethical principles, sustainability goals, and governance requirements are frequently articulated at a high level, while their integration into day-to-day AI development, deployment, and management practices remains limited.

This doctoral project aims to develop and validate a governance and organisational capability framework for Responsible and Sustainable AI across the AI lifecycle. The research conceptualises AI responsibility and sustainability not as inherent properties of AI systems, but as outcomes of organisational capabilities, governance structures, and decision-making practices. The AI lifecycle serves as the primary analytical focus, encompassing data sourcing, model development, deployment, monitoring, and decommissioning, while the broader AI value chain is used to capture environmental, social, and organisational impacts beyond individual systems.

Using a Design Science Research methodology, the project combines empirical studies with the design of governance artefacts such as capability models and decision-support frameworks. These artefacts aim to translate technical sustainability and responsibility indicators into actionable organisational and policy-level instruments, with their empirical validation to be conducted real-world (organisational and/or public sector) contexts.

The expected outcomes include theoretical contributions to Information Systems and AI governance research, as well as practical tools supporting responsible, sustainable, and trustworthy AI adoption.

Supervisor(s): Orlenys López Pintado

Existing approaches to Business Process Simulation are largely focused on simulating individual business processes in isolation. Their purpose is to help analysts estimate process performance and evaluate improvement options. However, real business processes rarely run in isolation. They depend on other processes and external actors, which often leads to delays and additional variability. Data-driven simulation can automatically discover models from event logs, but it usually assumes that these interactions are visible within a single log. In practice, cross-process dependencies and external interactions are often not explicitly captured, especially across organizational boundaries. This motivates the need for simulation methods that explicitly account for process context.

This doctoral project will develop techniques for context-aware business process simulation. To achieve this goal, the first objective is to discover simulation models of interrelated business processes by inferring dependencies such as causal relations, synchronization points, and shared resources or data objects. The second objective is to model interactions with external actors by capturing response variability and unexplained delays in event data. Finally, the third objective is to use the resulting models for what-if analysis to assess business process optimization options.

Supervisor(s): Sven Laur, Raivo Kolde

Large language models (LLMs) have created significant new opportunities for clinical research by enabling the automated extraction of information from unstructured medical texts, such as patient complaints, diagnostic findings, and treatment outcomes. However, practical challenges remain: medical datasets are often hundreds of gigabytes in size, and extracting thousands of different data points would require many complete passes through the data, making direct LLM application impractical.

This doctoral project proposes a multi-stage approach to address these challenges. First, high-recall filters combining keyword searches with lightweight classifiers trained on LLM-annotated data can reduce input volume by over 100-fold. Second, a two-phase extraction strategy improves model grounding by having the model first identify relevant phrases before producing structured output, making verification easier. Third, LLMs can contextualize extracted facts, determining whether information describes a patient's current condition, medical history, or relates to family members.

The project will develop and validate these tools through small-scale studies before scaling to the Estonian Biobank and EstHealth30 datasets, covering over 600,000 patients. Conducted in collaboration with the University of Tartu Medical department, the Estonian Biobank, and the OHDSI network, this work extends ongoing research using machine learning to detect adverse drug reactions and drug discontinuation events. The expected results include scalable extraction solutions for medical facts, methods for distilling LLM knowledge into simpler models, and the creation of standardized databases from unstructured Estonian medical data to support more comprehensive clinical research and healthcare applications.

Supervisor(s): Meelis Kull, Raul Vicente Zafra

Spatial AI is modelling where something is, how it changes over time, and how different observations fit together. This is critically important for achieving safety in technologies such as autonomous driving, drone navigation, and other areas of robotics. Measurements are noisy, information can be missing, and the same evidence can sometimes support several possible explanations. Because of this, a useful spatial model should not only produce a prediction, but also represent uncertainty and update it as new evidence arrives.

This PhD project studies the foundations behind reliable spatial uncertainty. A central idea is that uncertainty is not just a real-valued quantity, but depends on how a model represents space internally. The project starts by investigating belief representations and updating rules for continuous space, including structured options inspired by neuroscience. In particular, it looks at grid cells, i.e. neurons with regular, repeating spatial patterns, as motivation for modular and periodic representations, and asks when such structure helps in accounting for uncertainty.

The project then develops methods to evaluate and calibrate spatial uncertainty, so that probability statements match real-world frequencies in a well-defined way, including for predicted regions of likely locations. Finally, it studies how uncertainty should change under shifts in context, such as changes in geometry or noise levels, aiming for methods that remain reliable when conditions differ from those seen during training. While space is the main testbed, the goal is to produce general concepts and tools that also apply to other structured prediction settings.

Supervisor(s): Raimundas Matulevičius

The research aims to develop a comprehensive architecture to ensure secure and privacy preserving supply chain workflows. Furthermore, as the supply chain workflow involves collaboration among multiple stakeholders, the architecture shall consider a federated approach. The research will focus on mapping supply chain workflows, considering current regulatory trends, and analysing use cases and associated risks. The collected inputs will be crucial in the design of the federated architecture, ensuring privacy, security, and forensic readiness, and primarily focusing on the assurance of the integrity of the supply chain workflow, both from the data and the activities perspective. A challenging aspect of the architecture shall be balancing the privacy of the involved parties through the design and use of PETs, while ensuring forensic readiness, accountability, and the ability to investigate supply chain attacks.

Supervisor(s): Dietmar Pfahl, Kristiina Rahkema

Software vulnerabilities continue to pose a major challenge to the security and reliability of modern software systems. Even when vulnerabilities are identified and patches are released, it is often unclear whether a particular product version is affected, whether a fix has been applied correctly, or whether similar weaknesses persist elsewhere in the codebase. These uncertainties complicate vulnerability management for developers, security analysts, and downstream users, and contribute to delayed remediation and recurring security incidents.

This PhD project aims to address these challenges by treating vulnerability fixes themselves as a primary source of knowledge for automated security analysis. By learning from how vulnerabilities are repaired in real-world software projects, it becomes possible to reason not only about the presence of known vulnerabilities, but also about the correctness of fixes and the existence of related, previously undiscovered flaws. The proposed research seeks to develop techniques that bridge the gap between vulnerability disclosure, patch analysis, and automated testing.

Supervisor(s): Anastasija Nikiforova, Piret Luik

The rapid diffusion of generative artificial intelligence (AI), particularly large language models, is transforming how academic knowledge is produced in universities. Students increasingly rely on AI systems as collaborators in research and writing, yet institutional governance structures have not evolved at the same pace. Current approaches to Responsible AI in education remain largely aspirational, offering limited guidance on accountability, transparency, and institutional responsibility.

This doctoral project investigates how generative AI becomes embedded in academic knowledge production and how universities can govern this transformation responsibly. Drawing on Information Systems, AI governance, and digital education research, the study examines student–AI co-production practices and their institutionalization within higher education. Empirically, the project employs digital ethnography, go-along studies, and model ethnography to observe AI-mediated writing practices, complemented by the development of a Provenance Scale that traces AI contribution across stages of academic work. The project follows a Design Science Research approach to develop and evaluate a governance framework for responsible AI use in universities. The expected outcomes include conceptual models, governance guidelines, and policy-relevant artefacts that support transparency, trust, and accountable AI adoption in knowledge institutions. By bridging theory and practice, the research offers actionable solutions for higher education governance.

Supervisor(s): Janno Siim

Zero-knowledge proofs (ZKPs) are cryptographic protocols that allow one to prove statements about private data without revealing any additional information. Zero-knowledge SNARKs are proofs that are particularly small in size and fast to verify. Verifying the proof often takes much less time than verifying the statement directly. This makes SNARKs interesting even when they do not preserve data privacy.

The security of SNARKs and many other ZKPs is often proven in idealized models or under poorly understood assumptions. This project aims to improve the security of SNARKs by basing them on more trustworthy assumptions, proving that they satisfy stronger security properties than previously believed, and potentially uncovering attacks against existing schemes. Besides the security, we also investigate other aspects of ZKPs, such as improving their efficiency, making them post-quantum secure, and adopting them in new applications such as zero-knowledge machine learning.

SNARKs secure billions of dollars in cryptocurrencies, and ZKPs are used in the Estonian e-voting scheme. This project deepens the understanding of the security of those protocols.

Supervisor(s): Maiara F. Bollauf

This PhD project is about developing new ways to keep digital information secure in the future. Today’s encryption methods protect information like online banking, private messages, and digital signatures, but many of them become vulnerable once powerful quantum computers are built. To prepare for that, researchers are working on post-quantum cryptography, which aims to design security systems that can resist even quantum attacks. One (and probably the most) promising area within post-quantum cryptography is latticebased cryptography. Lattices are highly symmetrical mathematical structures that can be used to build encryption and digital signature systems that are believed to be secure against quantum computers.

The student will investigate a new and challenging mathematical problem, known as the lattice isomorphism problem, and investigate its use in designing efficient digital signature schemes with rigorous security guarantees. The student will look at special types of lattices build from error-correcting and will use advanced techniques to provide strong mathematical proofs that the proposed systems are secure. The project may also explore connections to privacy-preserving technologies, such as zero-knowledge proofs, which allow someone to prove they know a secret without revealing it.

The project sits at the intersection of cryptography and coding theory and is wellsuited for students with a strong background or interest in mathematics, cryptography, or theoretical computer science.

Supervisor(s): Helger Lipmaa and Matteo Campanelli

The doctoral project “Modern Zero-Knowledge and SNARKs: Theory and applications” addresses a rapidly developing area of cryptography that enables one to prove the correctness of computations without revealing the underlying sensitive data. Such techniques are becoming a key component of modern digital infrastructure, supporting secure data processing, privacy-preserving services, and trustworthy distributed systems.

The goal of this project is to study and develop new methods in zero-knowledge proof systems (zk-SNARKs), with a focus on both their mathematical foundations and practical applicability. The research will investigate state-of-the-art approaches that shape the field, while addressing core challenges such as computational efficiency, scalability, and rigorous security guarantees.

The project builds on the strong and internationally recognized research background of the supervisor, Helger Lipmaa, in cryptography and zero-knowledge. At the same time, it is ambitious and forward-looking, encouraging exploration beyond existing frameworks. The doctoral student will develop new theoretical models and practical techniques aimed at advancing the reliability and usability of zk-SNARKs.

The importance of this research lies in its contribution to technologies that are essential for a secure and privacy-respecting digital society. The expected results will advance fundamental science while also enabling practical applications across domains where trustworthy and verifiable computation is required.

Supervisor(s): Sedat Akleylek

Post-quantum cryptography has become increasingly important because widely used public-key systems will become insecure once large-scale quantum computers are available, mainly due to Shor’s polynomial-time algorithm. For example, current internet security protocols that protect online banking, ecommerce transactions, and government communications rely on traditional public-key cryptography that could be broken in the near future. To address this risk, NIST launched a standardization project to replace traditional public-key cryptosystems with quantum-safe alternatives. This thesis focuses on two important research directions in post-quantum cryptography: improving latticebased algorithms to reduce computational complexity and developing efficient software implementations that resist implementation/side-channel attacks. These improvements are motivated by real-world deployment needs on platforms such as smartphones, smart cards, embedded devices, and cloud servers, where performance, memory, and energy consumption are critical. Core arithmetic operations like polynomial multiplication are optimized for specific parameter sets and hardware constraints while ensuring constant-time execution to prevent information leakage. By studying both traditional and post-quantum schemes, the research develops practical countermeasures against timing and cache-based attacks that have affected real systems in the past. The expected outcomes include efficient and secure lattice-based cryptographic primitives, threshold protocols for distributed trust (for example, protecting keys in financial systems or digital identity services), and open-source implementations suitable for integration into real applications. Overall, this thesis aims to narrow the gap between theoretical post-quantum designs and secure, attack-resistant implementations, delivering solutions that are practical for real-world use and future-proof against quantum threats.

Supervisor(s): Kallol Roy, Hannes Keernik, Velle Toll

This PhD project proposes a Continuous Space U-Net model (CSU-Net) for detecting aerosol-polluted cloud tracks from MODIS satellite images. These cloud tracks — bright linear features formed when industrial aerosol plumes modify cloud properties — are valuable natural experiments for understanding how anthropogenic air pollution affects Earth’s climate. While ML models have previously been used to detect ship‑polluted tracks over oceans, identifying similar tracks over land has been difficult due to more complex cloud patterns and limited training data. Cloud tracks are transformed into continuous embeddings, enabling learning in a symmetry-rich space rather than raw image space. This improves data efficiency, scalability, and reduces manual annotation through geometric invariance. CSU-Net employs a U-Net backbone with two encoders and one decoder, trained on hand-logged ship and industrial cloud track datasets to reconstruct polluted cloud tracks at scale. The project will produce a publicly available dataset of roughly one million industry track cases in clouds and enable improved estimates of anthropogenic aerosol impacts on clouds and Earth’s climate. This work supports reducing uncertainty in anthropogenic climate forcing and contributes to the ERC‑funded CloudTracker project.

Supervisor(s): prof. Jaak Nerut and assoc. prof. Rutha Jäger

Electrocatalysis is a key process in various energy conversion devices, including electrolysers. By using electricity from renewable energy sources, an electrolyser can produce hydrogen through water splitting. Currently, most hydrogen is still produced from fossil fuels, resulting in significant carbon dioxide emissions.

One of the most promising devices for water splitting is the proton exchange membrane electrolyser, which can operate at high current densities, respond quickly to fluctuating renewable electricity, and produce high-purity hydrogen. Their wider deployment is currently limited by high cost and the need for highly durable materials capable of operating under harsh acidic conditions. A key challenge is the oxygen evolution reaction (OER) at the anode, which is kinetically slow and requires precious-metal catalysts. The most effective catalysts currently rely on iridium, a scarce and expensive material.

This doctoral project aims to develop improved catalyst materials - rare earth elements doped iridium oxide - that retain the high stability of iridium oxide while enhancing efficiency, thereby reducing the required iridium loading. The work combines materials development with advanced physical and electrochemical characterisation to understand how changes in catalyst composition and structure affect OER activity and stability. By identifying key factors that determine the structure-activity-stability relationship and testing the most promising materials under real-world operating conditions, the project promotes more affordable and scalable green hydrogen production while reducing dependence on critical raw materials.

Supervisor(s): Ivo Leito, Märt Lõkov

Nitrogen‑containing organic bases, such as amines and heterocycles, are ubiquitous. Their basicity plays a central role in applications ranging from the development of pharmaceuticals and agrochemicals to synthesis and separation technologies. Despite their importance, basicity data (pKₐH values) in non‑aqueous media remain scarce and often unreliable, even though many relevant processes occur in solvents other than water. Furthermore, no experimental biphasic pKₐH (pKₐHow) values are currently available, despite their advantages in the case of phase transfer, membrane transport, and extraction processes.

This PhD project addresses these gaps by establishing the first comprehensive, selfconsistent dataset of pKₐH values for nitrogen bases in widely used non-aqueous solvents (e.g., acetonitrile, DMSO, DMF, propylene carbonate, ethanol, acetic acid). First, a critical survey of the existing literature will identify compounds and solvents where pKaH data is inconsistent or missing. A diverse set of bases (ca 50) and 5–7 solvents will be selected for systematic study. Using fundamentally different techniques – UV‑Vis spectrometry, nuclear magnetic resonance (NMR) spectrometry, and potentiometry – a thorough experimental basicity study will be conducted. The resulting data combined with critically evaluated literature values and supporting computations, will form a recommended reference set of non‑aqueous basicities of nitrogen bases.

The project will further deliver the first experimental pKₐHow values for selected nitrogen bases in an octanol–water biphasic system, enabling more accurate modelling of biphasic processes. Based on the accumulated dataset, a prediction model will be developed to estimate pKₐH values across solvents from known values in at least one medium, including water. Finally, a data presentation scheme for experimental pKₐH data will be established, offering the first broadly applicable standard for the presentation of acidity/basicity data in the scientific literature.

Supervisor(s): Koit Herodes, Ivo Leito

Deep eutectic solvents (DESs) form when two (sometimes four) solid components interact strongly upon mixing and depress the mixture’s melting point, producing a room-temperature liquid (a deep eutectic solvent). They are emerging as environmentally friendly and versatile alternatives to conventional solvents used in analytical chemistry. They can improve both targeted and nontargeted analyses by selectively extracting specific compound groups, thereby reducing the co-extraction of other sample components, and simplifying the complexity of large datasets. Yet, the fundamental chemical factors influencing the extraction performance of DESs, like the roles of the components’ pKa values, remain poorly understood.

This project aims to systematically link DES composition to extraction efficiency and analyte properties and demonstrate the value of this understanding through real-world analytical analyses, including extracting phenolic compounds from human samples, improving pesticide analysis in citruses, and mapping the chemical space of Estonian peat and wood-processing residues. Additionally, the project will explore novel strategies to combat the practical limitations of using DESs, such as their high viscosity.

Supervisor(s): Ave Sarapuu, Kaido Tammeveski

Low-temperature fuel cells are a promising technology for renewable energy-based economy. Anion exchange membrane fuel cells (AEMFCs) represent an attractive alternative to conventional proton exchange membrane fuel cells, as they enable the use of non-precious metal catalysts for the oxygen reduction reaction at the cathode. Transition metal–nitrogencarbon (M–N–C) catalysts containing atomically dispersed active sites (MNₓ) are particularly promising; however, their performance in fuel cells is often limited by insufficient active site density and suboptimal porous structure.

This project aims to develop efficient and stable M–N–C catalysts for AEMFCs from biomass-based precursors. Ionothermal carbonisation in molten salts provides a versatile route to tailor the porous structure of carbon nanomaterials, while cation exchange enables the formation of high density of atomically dispersed MNₓ sites. The project will focus on systematically tuning porosity and increasing active-site density through variation of synthesis parameters, precursors, and transition metals. Catalysts’ structure–performance relationships will be investigated using physicochemical characterization and electrochemical testing, progressing from rotating disk electrode measurements to gas diffusion electrode experiments and single-cell AEMFC tests.

Supervisor(s): Enn Lust, Silvester Jürjo

Rare earth elements (REE) play a crucial role in modern technology. Since production of REE is mainly dominated by China, alternative REE production methods should be developed by European Union, to overcome possible market fluctuations. Estonian phosphorite ore might be a potential source of REE, which are obtained as a by-product of fertiliser production.

Novel combined chemical separation method will be developed in this project. In the first step, heavy and light rare earths will be separated from each other by liquid extraction. Other contaminating chemical elements (like Ca and Fe) will be collected and removed to obtain high-purity REE. Single REE will be isolated using combined ionic liquid and electrochemical electroreduction methods. Molecular mechanisms of liquid extraction and optimal conditions for liquid extraction, electroredox processes conditions and kinetics will be established and discussed.

Supervisor: Nadežda Kongi

Electrochemical synthesis of urea from carbon dioxide and nitrate derived from waste streams offers a promising alternative to conventional energy-intensive processes with substantial carbon emissions, enabling simultaneous carbon and nitrogen recycling while reducing environmental impact.

This doctoral project is founded on the hypothesis that optimally positioning the carbon- and nitrogen-intermediates by varying the interatomic distance on the dual-atom site catalysts minimizes the energy barrier for carbon–nitrogen coupling. To test this hypothesis, the project integrates simulations with data-driven modelling to accelerate the discovery of efficient catalysts for electrochemical urea synthesis.

DFT calculations will be employed to elucidate reaction mechanisms, identify the steps that control reaction rate and selectivity, and quantify the geometric and electronic factors governing carbon-nitrogen bond formation. The resulting dataset will be used to train machine learning models capable of screening a broad catalyst space and identifying promising catalyst candidates.

By combining atomistic modelling with machine learning, the project aims to establish transferable catalyst design principles and predictive tools that support the experimental development of high-performance electrocatalysts. The expected outcomes will contribute to sustainable chemical manufacturing and advance computational methodologies for rational catalyst design.

Supervisor: Nadežda Kongi

This doctoral project focuses on the development and experimental investigation of dual-atom-site catalysts (DASCs) for the electrochemical co-reduction of carbon dioxide and nitrate to urea. The research addresses key limitations of current urea electrosynthesis technologies, namely the lack of catalysts capable of promoting selective carbon–nitrogen (C–N) coupling while suppressing competing reactions, particularly hydrogen evolution. The central hypothesis is that catalytic activity and selectivity can be systematically tuned by controlling the composition, coordination environment, and spatial arrangement of adjacent metal centres in DASCs. Such materials are expected to enable independent stabilisation of critical reaction intermediates (e.g., *CO and *NO2), thereby lowering kinetic barriers for C–N bond formation. The research includes catalyst design, synthesis, advanced physicochemical characterisation, and electrochemical performance evaluation. Multiple synthetic approaches will be employed, including dual-atomic interfaces derived from metal nanoparticles, metal–nitrogen-doped carbon (M–N–C) systems, and binuclear molecular complexes. Structural, electronic, and surface properties of the catalysts will be analysed using complementary techniques such as XPS, SEM/TEM, XRD, and IR spectroscopy. Electrochemical assessment will involve systematic testing under progressively relevant configurations (single-cell, H-type, and flow-cell systems), combined with quantitative product analysis and performance optimisation. The project aims to establish experimentally validated structure–performance relationships governing C–N coupling in DASCs, providing mechanistic insights into activity, selectivity, and stability. The expected outcomes will contribute to the rational design of advanced electrocatalysts for multi-reactant systems, support sustainable nitrogen and carbon utilisation strategies, and advance electrochemical routes for lowcarbon chemical manufacturing.

Supervisor(s): Marek Mooste, Kaido Tammeveski

High-temperature proton exchange membrane fuel cell (HT-PEMFC) is an electrochemical energy conversion device, the development of which is necessary to promote green hydrogen energy and economy. The advantages of HT-PEMFC, over its already commercialised low-temperature version, are suitability for lower purity fuel, better thermoregulation, absence of liquid water, and kinetically more favourable electrode reactions. These circumstances make HTPEMFC technology promising for the heavy transport and aviation sectors.

The main shortcoming for bringing HT-PEMFC to the market is the lack of a suitable cathode catalyst. A highly durable and affordable catalyst is needed to catalyse oxygen reduction at the cathode of the fuel cell. Here, we propose nanocarbon materials doped with several transition metals and nitrogen for this purpose. To increase the materials' oxygen reduction activity, additional heteroatoms are introduced into their composition, and nanocarbon supports are optimised for greater durability. In addition, the possibility of adding small amounts of Pt to catalysts and novel durable catalyst support materials will be investigated.

Supervisor(s): Pilleriin Peets, Ivo Leito

Targeted screening captures only a small fraction of chemical information in complex samples, limiting insights into biological and environmental systems. Non-targeted metabolomics, introduced about two decades ago, offers a broader view but faces challenges such as high-confidence structural annotation and variability caused by instrument and data processing choices.

This PhD aims to advance metabolomics by embedding computational mass‑spectrometry-based chemical descriptors like molecular fingerprints and compound class vectors within metabolome characterization. These approaches allow characterization of chemical space without full structural annotation, enabling better interpretation of metabolite relationships. Using non-targeted LC-HRMS for expanded MS2 coverage, the workflow will apply advanced tools (SIRIUS, MetFrag, MS2Quant) and statistical methods alongside innovative visualization and prediction platforms (GNPS, CCV, MS2Tox). By leveraging computational strategies, this work reduces experimental costs, supports retrospective analyses, and provides deeper insights into metabolite diversity and connectivity.

Supervisor(s): Srinu Akula, Kaido Tammeveski

Proton exchange membrane fuel cells (PEMFCs) and anion exchange membranes fuel cells (AEMFCs), are popular electrochemical energy conversion devices due to their high efficiency. Platinum-based catalysts and their alloys supported on nanocarbons are typically used to catalyse the oxygen reduction reaction (ORR). However, scarcity, poor stability and high cost are main barriers towards the commercialisation of fuel cells. Therefore, several approaches are employing to improve the electrocatalytic activity and performance of non-precious metal catalysts. The M-N-C (M = Fe, Co, Ni, Mn, Cu, etc.) single-atom catalysts (SACs) are new frontiers in the field of electrocatalysis for fuel cells due to the maximum metal atomic use and high electrocatalytic activity. Hence, suitably engineered single-atom catalysts, porous metalorganic frameworks play crucial roles in improving the performance of these materials via synergy between heterogeneous catalysts sites. Heteroatom (N, P, S) doping of the carbon nanomaterials influence the electronic properties, which is favourable for ORR electrocatalysis. This electrocatalytic effect can be enlarged further by creating the M-Nx coordination environment with the additional metal atom sites. A thorough electrochemical and physical characterisation of the catalyst materials will be carried out and accelerated durability testing will be conducted, including polymer electrolyte fuel cell performance evaluation.

Supervisor(s): Signe Vahur, Anu Teearu-Ojakäär, Ivo Leito

Chemical analysis of materials from cultural heritage (CH) objects is essential for historians, archaeologists, and conservators seeking information on the origin, authenticity, and age of artefacts, or for selecting suitable conservation materials. Analysis of paints containing organic pigments is challenging due to their complex multicomponent composition, which can change significantly with ageing via oxidation, degradation, etc. For the analysis of CH objects, non-invasive and non- or minimally destructive direct surface analysis methods are preferred. Analysis of these materials with common chromatographic and mass spectrometric (MS) techniques is complicated because they require a sample piece, specific sample preparation, and measurement conditions. Currently, no reliable laser ablation and mass spectrometry-based analytical technique is available that allows direct analysis of organic materials on the solid surface of an artefact.

Under the chair of analytical chemistry, the CH research group is currently developing a novel laser ablation sampling probe (LASP) that can be coupled to the atmospheric pressure chemical ionisation (APCI) high-resolution (HR)MS system. This quick, controllable, precise analytical system enables direct MS analysis of organic materials on the artefact surface under ambient conditions, without sample removal and with minimal surface damage.

The main aim of this PhD project is to develop measurement methodologies for the novel LASP-APCIHRMS system for the analysis of the composition of organic pigment-containing paints and to assess the capabilities of the novel analytical system.

The PhD student will be part of a very interdisciplinary research group where chemists, physicists, engineers, conservation scientists, and IT specialists work. Within this PhD project, the possibilities for analysing CH objects will be expanded.

Supervisor(s): prof. Enn Lust, Thomas Thomberg, Andres Lust

The SARS-CoV-2 and human influenza viruses spread between people through respiratory droplets with a diameter of less than 10 μm and by direct contact. The spread of these viruses can be significantly reduced through the use of various filter materials, as they are capable of capturing airborne viral particles of different sizes. Novel virucidal filter materials activated with two different metal (Cu and Zn, Zn-Ag and Cu-Ag) nanoclusters will be prepared by electrospinning a solution of poly(vinylidene difluoride) in N,N-dimethylacetamide without and with the addition of corresponding metal salts as virucidal (killing) agents. Detailed physical, chemical and porous structure analysis will be conducted. Synthesized materials will be tested for their virucidal properties against SARS-CoV-2 and human influenza viruses. The optimal concentrations of metallic nanoclusters in the initial solution will be varied and determined and their concentration in the synthesized fibre materials will be established. Hg porosimetry and gas sorption measurements will be performed for materials porous structure characterization. The chemical composition and concentration of the polymer solution, solvents or mixtures of solvents will also be varied for complete analysis. The electromagnetron sputtering method will be used for deposition of some metal nanoclusters in order to achieve activated virucidal materials. The introducing different metal nanoclusters into a nanofibrous filter material will allow the synthesis of a material that broader virucidal effect.

Supervisor(s): Jaanus Harro, Mati Karelson, Margus Kanarik

Recent advances in epitranscriptomics have revealed RNA methylation as a novel and critically important regulator of neuronal function and behaviour. N6-methyladenosine (m⁶A), the most abundant internal modification in eukaryotic RNA, influences RNA splicing, stability, nuclear export, and translation and is dynamically regulated by methyltransferase writer complexes, demethylase eraser enzymes, and m⁶A-binding reader proteins. Increasing evidence links changes in RNA m⁶A methylation to neurodegenerative and psychiatric disorders, including anxiety, depression, substance use disorders and suicidality. Recent studies further suggest that modulation of RNA methylation may have therapeutic potential, as pharmacological manipulation of RNA methylation reduces anxiety- and depression-like behaviour in rodents, while genetic variation in RNA demethylases is associated with personality traits conferring psychiatric vulnerability.

Building on these novel findings, this doctoral project aims to systematically and innovatively evaluate RNA methyltransferase activators and demethylase inhibitors in validated animal models of anxiety, depression, and addiction. The project will characterise the behavioural effects of these compounds and elucidate the underlying neurobiological and molecular mechanisms, with particular focus on monoaminergic systems. Sex- and dose-dependent behavioural outcomes will be assessed alongside measurements of monoamine neurochemistry, both in brain tissue and in extracellular fluid (measured by HPLC and in vivo microdialysis).

At the molecular level, the project will examine drug-induced changes in global and region-specific RNA m⁶A methylation, enzymatic activity of methyltransferases and demethylases, gene expression, and transcript-specific methylation (measured by UHPLC, mass spectrometry, ELISA, RT-PCR, MeRIP-PCR). Transcriptome- and epitranscriptome-wide analyses will be used to identify novel molecular networks associated with behavioural responses (measured by RNA-seq, MeRIP-seq). Finally, the impact of compounds on neuronal dendritic morphology will be evaluated to link molecular alterations with structural plasticity (analysed by histochemistry). We expect that pharmacological modulation of RNA m⁶A methylation will alleviate anxiety- and depression-like behaviour through circuit-specific molecular and structural changes.

Supervisor(s): Siim Salmar, Kaija Põhako-Esko

This doctoral project aims to transform lignin–cellulose mixtures (LCM), an intermediate of birch biorefining, into a tunable platform for advanced bio-based materials. Instead of fully separating lignin and cellulose, the project exploits their natural synergy to develop nanostructured hydro- and aerogels with tailored properties. Controlled fractionation using water–organic solvents and ionic liquids, combined with ultrasound-assisted nanostructuring, will enable selective modification of lignin content and in situ formation of CNF/CNC networks.

By integrating detailed chemical characterization with FTIR, semi-solid-state NMR, and rheological analysis, and molecular dynamics simulations, the project will establish quantitative multiscale structure–property relationships in LCM-based hydrogels and aerogels. This combined experimental–computational approach will clarify how lignin content, network architecture, and intermolecular interactions (such as H-bonding and hydrophobic aromatic interactions) govern mechanical, structural, and transport properties. The resulting knowledge will enable the rational design of high-performance LCM nanocomposite materials e.g. for applications in biomedicine, cosmetics and in smart materials for energy applications, separation technologies, functional coatings etc. The project supports the development of more flexible, sustainable, and value-added biorefinery pathways.

Supervisor(s): Veljo Kisand; Kadri Runnel

Building on earlier sedaDNA studies and recent pilot work, the project advances the use of lake-sediment DNA to understand long-term changes in forest biodiversity, with a particular focus on fungi—an ecologically crucial but conservation-wise underrepresented group. Fungi are indispensable components of forest ecosystems, functioning as decomposers, mutualists and pathogens. Yet their conservation is hindered by limited knowledge of how present-day fungal diversity patterns have developed over time. SedaDNA preserved in lake sediments provides a promising way to reconstruct past fungal communities, but its use remains methodologically and ecologically underdeveloped. This PhD aims to fill that gap by (1) establishing robust laboratory and analytical protocols for fungal sedaDNA work at the University of Tartu, (2) examining how reliably fungal taxa are detected across different sites, sediment types and time periods, and (3) evaluating how well sedimentary fungal DNA reflects contemporary forest fungal communities.

The thesis consists of three integrated parts. The first focuses on methods development— optimising DNA extraction, purification, and metabarcoding approaches for low-abundance and ancient fungal DNA, and creating standardised workflows for future research. The second part uses Holocene lake sediment cores from Estonia–Latvia, Sweden and Lithuania to study detection probabilities, time-dependent DNA decay, taxonomic biases, and cross-site comparability. The third part links sedaDNA results with present-day fungal diversity data from forest soils and surface sediments, allowing assessment of the spatial and ecological representativeness of fungal sedaDNA and its relevance for biodiversity monitoring. By integrating methodological refinement with ecological calibration, this project provides essential baseline knowledge for applying sedaDNA in fungal conservation and long-term biodiversity assessment.

Supervisor(s): Boris Deshev

This doctoral project focuses on the development and scientific application of LEIDMA, a machine-learning framework designed to detect extremely weak signals in radio astronomical observations. The work combines methodological innovation with astrophysical research, addressing both the technical challenges of analysing modern large datasets and a fundamental open question in galaxy evolution: how galaxies acquire gas from their surroundings.

The first part of the PhD will concentrate on designing, training, and optimising neural-network models capable of identifying faint spectral signatures that remain undetected with conventional analysis techniques. Using curated training datasets and archival observations, the student will develop robust and reproducible workflows for automated signal extraction, contributing to the preparation of analysis tools required for next-generation facilities such as the Square Kilometre Array.

The second part of the thesis will apply these methods to deep radio surveys, including archival Arecibo observations and complementary HI datasets, to investigate the presence and properties of diffuse neutral hydrogen in the circumgalactic and intergalactic medium. Through statistical signal detection and stacking analyses, the project aims to place new constraints on gas accretion processes across different galactic environments.

By combining advanced data science with observational astrophysics within an international collaboration, the project will produce both a publicly available analysis tool and new scientific results, while providing comprehensive interdisciplinary training for the doctoral candidate.

Supervisor(s): Mait Lang, PhD

The study uses Sentinel-2 MSI, Landsat-8/9 OLI data, and the airborne laser scanning (ALS) database provided by the Estonian Land board. To prepare, the study analyses also last 40 years records of landscape fires to collect information of fire behaviour. Time series of multispectral satellite images and ALS data are used to construct input maps for landscape fire behaviour models (starting with https://cwfis.cfs.nrcan.gc.ca/background/summary/fbp) and the models will be tested on the observation data from known vents. Finally, scenarios are simulated for test sites and geographic locations in Estonia where high forest fire risks occur. The results will establish basis for the integration of remote sensing data analysis into strategic, tactical and operational planning.

Supervisor(s): Mihkel Pajusalu

The goal of this PhD project is to research, benchmark and compare various machine learning algorithms and their deployment strategies, focusing on FPGA (Field Programmable Gate Array) SoCs (System on Chips) as computing platforms and considering both ground and space segments. An important aspect of this will be remote reconfigurability and the mission control operations needed to enable it.

The main deployment target in this project will be the FPGA SoC based Command Module payload built by the Estonian Student Satellite Foundation, planned to be launched on the ESTCube-3 CubeSat (development began in 2025, launch planned for 2028). As the payload will be built to be reconfigurable for multiple experiments in different data processing domains, it will provide an excellent opportunity to conduct this project and to test the developed framework on various use cases.

In addition, it is intended for this research to result in a new framework for testing, benchmarking and deploying machine learning algorithms independent of the target hardware and utilizing different sensors and instruments depending on the use case. The resulting product will speed up the development of machine learning algorithms for satellite autonomy applications and simplify the mission operations.

Supervisor(s): Heleri Ramler, Eike W Günther

This PhD project investigates the connection between Galactic evolution and the formation and occurrence of exoplanets around early-type stars. Different regions of the Milky Way have experienced distinct star-formation histories, chemical enrichment, and dynamical evolution, which are expected to influence the conditions under which stars and planetary systems form.

The project focuses on early-type A stars, which remain underrepresented in exoplanet studies compared to Sun-like stars. Due to their higher masses, different internal structures, and stronger radiation environments, early-type stars provide a unique opportunity to explore planet formation and evolution under physical conditions that differ from those of late-type stars.

By combining precise Gaia astrometry and kinematics with high-resolution stellar spectroscopy, the project will compare early-type stars with and without detected planets across different Galactic environments. The analysis will examine how stellar kinematics, chemical composition, and Galactic context relate to the presence and properties of planetary systems.

The results of this work will extend planet–star–Galaxy studies into a previously unexplored regime, providing new constraints on how Galactic-scale processes shape planet formation around intermediate-mass stars and informing future exoplanet surveys and space missions.

Supervisor(s): Antti Tamm, Rien van de Weygaert

In recent years, cosmology has achieved high precision in determining the general properties of the Universe, yet we remain far from a complete understanding of galaxy formation and evolution. The unexpectedly massive galaxies recently discovered in the early Universe with the James Webb Space Telescope demonstrate the gaps in our knowledge regarding matter assembly and star formation processes. Cosmic voids provide a unique opportunity to study galaxy formation and evolution in an environment where the large-scale matter density is very low, galaxy formation proceeds with significant delay (thus occurring closer to us and being more easily observable), and external influences are weaker.

The doctoral project investigates the role played in the evolution of galaxies located in voids by their immediate surroundings — the galaxy group — and the role of the general largescale environment — the cosmological void. The results will contribute to a better understanding of the physical processes shaping galaxy formation and evolution, including the interplay between dark matter and dark energy in different environments and at different stages of cosmic evolution.

Central to the research is the unique dataset provided by the ongoing J-PAS survey, containing information on the properties and spatial distribution of millions of galaxies. The results will be compared with cosmological simulations in order to verify the correctness of the methods and to test the validity of cosmological models.

Supervisor(s): Krista Alikas, Riho Vendt, Viktor Vabson

Water leaving radiance is a key parameter for ocean colour (OC) satellite radiometry. It is the basis for higher order products (e.g. chlorophyll a) and subsequent spatiotemporal analyses. Measurement schemes for in situ above-water radiometry are already well addressed, but in-water measurements, despite considered more accurate, still need attention. Optical laboratory facilities at Tartu observatory will be advanced to allow the characterization and calibration of in-water radiometers, assuring the traceability of a measurement and uncertainty budget derivation when moving from controlled laboratory to variable outdoor conditions. Outdoor comparisons of common radiometers in various deployment strategies, together with the development of new sensor prototype will help the community to optimize the in-water measurements. This allows producing traceable in situ measurements required for every OC satellite mission for validation, vicarious calibration and algorithm development.

Supervisor(s): Miia Rannikmäe, Moonika Teppo

Upper secondary education is a pivotal period for shaping students’ academic trajectories, career aspirations, and long-term engagement in STEM (science, technology, engineering, and mathematics). Research shows that cognitive achievement alone does not adequately predict STEM persistence. Instead, science identity—encompassing self-concept, perceived competence, sense of belonging, and social recognition in scientific practices—plays a central mediating role in sustaining engagement.

This longitudinal project examines how STEM achievement and science identity co-develop during upper secondary education and identifies the individual, social, and instructional factors shaping these trajectories. The study combines secondary analysis of national science e-assessment data (2026) with newly collected data from the same cohort in Grade 12 (2029), enabling analysis of developmental change over time.

A multidimensional science identity instrument will be developed and psychometrically validated, including tests of longitudinal measurement invariance. Advanced person-centered longitudinal modeling approaches (e.g., latent profile and transition analyses) will identify heterogeneous achievement–identity profiles and developmental pathways.

The project advances theory by proposing an integrative framework linking cognitive achievement, social experience, and instructional practice, while providing evidence-based guidance to promote equitable and sustainable STEM participation.

Supervisors: Katrin Vaino, Helin Semilarski, Konstantinos Korfiatis

This doctoral project aims to develop and validate a theoretically grounded assessment framework for fostering and measuring systems thinking in lower secondary design-based STEAM education (Grades 7–9). Although systems thinking is widely recognised as essential for addressing complex global challenges such as climate change, sustainable energy transitions, and urban resilience, it remains conceptually fragmented and methodologically under-operationalised in school practice.

The project begins with a systematic literature review to clarify conceptual dimensions and assessment approaches to systems thinking in science education. Based on these findings, a multi-method assessment framework will be developed, combining analysis of iterative design artefacts, structured systems-oriented tasks, and students’ explanatory, metacognitive reflections on their systemic reasoning and design decisions. The framework will first be tested in a pilot study, allowing refinement of assessment tools.

The main study will implement an intervention comprising teacher professional development and the classroom enactment of systems thinking–oriented design-based learning scenarios. In these interdisciplinary contexts, students will engage in iterative prototyping, using structured data collection and computational analysis to investigate system dynamics, feedback mechanisms, and non-linear relationships.

The study also examines teachers’ pedagogical beliefs and their role in fostering systems thinking. By linking students’ developmental trajectories with classroom practices, the project delivers a validated framework for measuring systems thinking and provides empirically grounded guidance for strengthening the pedagogical foundations of STEAM education in Estonia and internationally.

Supervisor: Viktor Abramov

The research of this project lies at the intersection of two areas of mathematics: geometry and algebra. In addition, the structures and methods developed within this project may be applied in theoretical physics. More precisely, the project is carried out in the field of geometric structures based on the theory of Lie and Poisson algebras. This area is actively developing, and in recent years important generalizations of the concepts of Lie and Poisson algebras have emerged. One such generalization is the transposed Poisson algebra. In this project, we propose an extension of this structure to superalgebras and introduce the notion of a transposed Poisson superalgebra. We study the properties of this structure, find identities, propose methods for constructing such structures using Jordan superalgebras and within differential geometry, and also investigate possible applications in theoretical physics.

Supervisors: Tiina Kraav, Meelis Käärik

Statistical literacy is essential in a data-driven society, both in everyday decision-making and in higher education across disciplines. However, research consistently shows that many university students – particularly non-statistics majors – struggle with fundamental statistical ideas. Difficulties commonly concern variability, randomness, causal reasoning, p-values, and the interpretation of confidence intervals. These challenges often become especially visible during the transition from secondary to tertiary education.

At the same time, artificial intelligence tools, including large language models, are rapidly entering school practice. In Estonia, the national AI Leap initiative promotes the systematic integration of AI into education. AI can provide explanations, examples, and feedback, but it may also create an illusion of understanding, where correct answers mask shallow conceptual grasp.

The aim of this doctoral research is to investigate how statistics education in upper secondary school can be designed so that AI supports deep conceptual understanding rather than replacing students’ reasoning. The study examines how students use AI when solving statistics tasks, which forms of AI interaction promote meaningful learning, and how AI-supported instruction can reduce common misconceptions.

Furthermore, the project explores how well-designed AI-supported learning environments can strengthen students’ readiness for tertiary-level statistics, particularly in disciplines where statistical reasoning is required but not the main focus of study.

The results will provide evidence-based guidelines for integrating AI into statistics education and contribute to broader discussions on how to support students’ transition from secondary school to higher education in a data-rich world.

Supervisor(s): Sirje Pihlap, Bjarnheiður Kristinsdóttir

This doctoral thesis aims to design and theoretically ground a pedagogical model for integrating artificial intelligence (AI) and digital technologies into lower secondary mathematics education in ways that enhance deep mathematical thinking, particularly in solving word problems.

Research shows that students’ difficulties in word-problem solving primarily arise from challenges in comprehension, situational modelling, and metacognitive regulation rather than from computational skills alone. At the same time, AI tools such as ChatGPT provide personalised explanations and interactive feedback, but also pose risks, including inaccurate reasoning and overreliance that may weaken critical thinking. Moreover, students’ attitudes toward mathematics are closely linked to their learning outcomes, making it essential to develop instructional approaches that support both cognitive and affective dimensions of learning.

The study employs a design-based research (DBR) methodology in Grades 7–9. It develops and iteratively refines a pedagogical model that combines a flipped classroom structure with Liljedahl’s Thinking Classroom approach. In this model, pre-lesson activities include teacher-created videos, structured AI-supported tasks, GeoGebra-based modelling, and formative digital assessment, while classroom instruction focuses on collaborative problem solving and collective reasoning.

The thesis consists of three interconnected studies: the development of the theoretical and pedagogical design framework; (2) an empirical investigation of how structured AI-supported preparation influences students’ modelling, reasoning, discourse, and attitudes; and (3) the formulation of transferable design principles for responsible AI and technology integration. The outcome is an empirically validated pedagogical model that aligns technological innovation with learning goals in mathematics education.

The innovativeness of the research lies in its holistic integration of technology, AI, and a flipped learning framework within a coherent pedagogical design. Unlike prior studies that examine these elements separately, this research investigates their structured interaction and theoretical alignment.

Supervisor(s): Vambola Kisand, Alexander Vanetsev, Angela Ivask, Wei Cao

This project focuses on developing nanostructured photocatalytic materials that are active under visible light to create antimicrobial surface coatings for reducing microbial transmission on frequently touched surfaces in healthcare facilities and public spaces. Photocatalytic antimicrobial action is driven by the generation of reactive oxygen species (ROS), which oxidize and decompose organic contaminants and microbial cell components. Current commercial photocatalytic coatings, predominantly based on TiO₂, require UVA illumination, limiting their usefulness indoors where visible light dominates. A key challenge in the field is producing visible‑light‑active photocatalysts that generate ROS as efficiently as UVA‑activated materials.

To address this gap, during this project several promising visible-light responsive photocatalysts with known antimicrobial properties will be synthesized and evaluated: WO₃, ZnIn₂S₄, ZnIn₂S₄/g‑C₃N₄, and ZnIn₂S₄/WO₃. These compounds will serve as the active components in coating formulations. Another important challenge is ensuring strong, durable adhesion of the coatings to different surfaces. Commercial acrylic paints and ethyl‑methacrylate polymer will be explored as matrix materials and to achieving uniform distribution of the synthesized photocatalysts within these matrices remains one of the central tasks of this project.

Overall, this project aims to close the technological gap that currently limits photocatalytic coatings to UVA‑dependent systems. By integrating material synthesis, surface engineering, and antimicrobial testing, it will establish a robust basis for practical, indoor‑applicable antimicrobial surfaces.

Supervisor(s): Tarmo Tamm

Hydrogels have emerged as a central class of biomaterials because their high water content, biocompatibility, and tunable mechanical and chemical properties allow them to mimic key features of native soft tissues. Their open and compliant architecture supports cell viability by enabling nutrient transport and waste removal, while their chemistry can be engineered to deliver bioactive signals or undergo controlled degradation. These characteristics have positioned hydrogels as indispensable materials in tissue engineering, regenerative medicine, drug delivery, and bioengineering interfaces. However, despite recent advances— like self‑healing, high toughness, stimuli‑responsiveness, and conductive formulations — current hydrogel systems still struggle to simultaneously meet the mechanical robustness, long‑term biocompatibility, and translational reliability required for biomedical use. Many state‑of‑the‑art designs optimize successfully on a single or at best a subset of desired properties, while achieving predictable biointegration, controllable degradation kinetics, and multi‑functional performance in a single platform remains a major challenge.

This project addresses the aforementioned limitations by developing next‑generation hydrogel‑forming polymer composite networks with precisely controlled copolymer architecture, allowing independent tuning of mechanics, degradation behaviour, and biological interactions. The central hypothesis is that adjusting crosslinking density and spatial organization of hydrophilic and hydrophobic domains will enable decoupled optimization of cell–matrix interactions and mechanical performance, ultimately supporting sustained and spatially defined biointegration of diverse cell types. The work will involve designing and synthesizing new (bio)polymer backbones, constructing hybrid and interpenetrating networks, reinforcing systems with particles or fibers, and mapping structure–property–function relationships through detailed physicochemical and mechanical characterization. Promising candidates will undergo extensive biological evaluation—adhesion, viability, migration, and phenotype‑relevant assays—to establish cell‑instructive functionality. The project will then incorporate programmable degradation mechanisms (hydrolytic, enzymatic, or stimuli‑responsive), culminating in the demonstration of a selected material platform in medically relevant applications such as complex tissue regeneration, advanced wound care with antimicrobial and pro‑healing functions, or delivery systems for therapeutic cells, probiotics, or drugs.

Supervisor(s): Taivo Jõgiaas, Jekaterina Kozlova, Helle-Mai Piirsoo

This doctoral research proposes an innovative approach to enhance the mechanical properties of carbide-based ceramics for applications in advanced tools and devices through grain boundary engineering and field-assisted sintering. Atomic layer deposition and MXenes will be used to design and modify grain boundary architecture by introducing intergranular films and reaction-derived phases, such as MAX phases and secondary carbide phases, to significantly enhance strength, fracture toughness, and thermal shock resistance.

Controlled grain-boundary modification will be supported by current-assisted and field-related effects of spark plasma sintering (SPS). A systematic investigation of these effects will reveal the still-poorly understood kinetic mechanisms of SPS densification. Particular emphasis will be placed on crystallographic and chemical characterisation of grain boundaries down to the atomic level to understand how their structure and composition influence the mechanical performance of the obtained ceramics.

Supervisor(s): Angela Ivask, Merilin Rosenberg

Antimicrobial products are intended to rapidly inactivate microbes and must meet minimum efficacy criteria. However, their performance in real-world applications is highly condition-dependent, and much less is known about the post-exposure biology of the cells that survive brief, intense biocidal contact. Unlike classical studies of prolonged sublethal or subinhibitory exposure in liquid media, short semi-lethal exposures leave limited time for inducible defenses to engage. Survivors are therefore expected to arise from pre-existing physiological heterogeneity and biocide-induced injury states in specific exposure conditions. These survivor states may reshape subsequent antibiotic tolerance during recovery in host-relevant environments - an underexplored interface between biocide use and clinically relevant antimicrobial tolerance.

The proposed PhD project will (i) characterize survivor states after short, semi-lethal exposures to disinfectants and antimicrobial surfaces under end-use application-relevant multi-stress scenarios, and (ii) quantify how these states influence antibiotic tolerance during recovery in host-relevant conditions. The project integrates time-kill assays with multiparameter microscopy/flow cytometry, early-recovery transcriptomics, and targeted genetics. Outcomes will support application-specific mechanism-based interventions, including synergistic active-agent combinations and coordinated use of antimicrobial surfaces with cleaning/disinfectant formulations.

Supervisor(s): Tiina Tamm

All viruses, regardless of genome type and size, utilize the host cell’s translation machinery, specifically ribosomes and translation factors, for the synthesis of viral proteins. Certain viruses employ specialised RNA elements, called internal ribosome entry sites (IRESs) to recruit ribosomes directly to their RNA. The use of such internal initiation allows specifically decrease the cellular protein synthesis and increase the translation of viral proteins.

The aim of this project is to determine how modified ribosomes facilitate IRES-mediated protein synthesis. Our approach involves using an IGR-type IRES element derived from the cricket paralysis virus and investigating a budding yeast ribosome mutant capable of initiating IRES-mediated translation more efficiently than wild-type ribosomes. We hypothesise that during viral infection, viruses specifically modify ribosomal proteins, creating a pool of modified ribosomes. This allows the viruses to translate their own RNA more efficiently than cellular mRNAs. We use budding yeast as an experimental system because it allows for rapid genome modification and the use of well-developed biochemical and functional analysis methods. We will also monitor IRES-mediated translation in Drosophila cell lines, the natural host of this virus. The project results in a deeper understanding of virus-host interactions, enabling the development of therapeutic strategies to effectively control the viral diseases.

Supervisor(s): Jaak Truu, Carlos Carmona

This doctoral project examines how microplastic-associated biofilms assemble and change across connected aquatic ecosystems, and whether a trait-based representation yields more transferable insight than taxonomy lists. Microbial communities are characterised as distributions of measurable trait proxies (continuous and categorical), summarised as community-weighted means and dispersions, to test environmental filtering, successional trade-offs, and links to antibiotic resistance outcomes.

Field experiments in Estonia deploy PET and PP as standardised substrates across a freshwater-brackish continuum and an upstream–downstream wastewater treatment plant (WWTP) exposure gradient. Each polymer is incubated as new and mildly weathered material, with glass as an inert surface control. Incubations run for three months, with weekly sampling to capture longitudinal successional trajectories.

A two-layer profiling strategy combines long-read amplicon sequencing for complete time-series coverage with hybrid short- and long-read metagenomics at selected time points to reconstruct MAGs, infer traits, and quantify antibiotic resistance genes. Trait trajectories are modelled as functions of ecosystem context, WWTP exposure, polymer type, ageing, and time using hierarchical models and variance partitioning. Predictive performance is compared across taxonomy-only, trait-only, and combined models via cross-validation across sites, and published plastisphere datasets are used to benchmark generality.

Supervisor(s): Arnold Kristjuhan, Riinu Kiiker

Plant diseases in agriculture are mainly controlled with synthetic pesticides. Although effective, their long-term and intensive use has led to environmental contamination, biodiversity loss, and the development of resistant pathogen populations. Biological control offers a safer and more sustainable alternative. It relies on naturally occurring microorganisms that suppress plant pathogens without harming the surrounding ecosystem.

This doctoral project aims to develop yeast-based biological control methods against two important cereal diseases in Estonia: septoria tritici blotch in wheat and net blotch in barley, caused by Zymoseptoria tritici and Pyrenophora teres f. teres. The project uses the Estonian Yeast Stock Collection of wild yeast strains isolated from Estonian environments. Because these yeasts are native species, their potential use as biocontrol agents minimizes ecological risks and avoids introducing non-native organisms.

The research will (1) characterize natural yeast communities on wheat and barley across growing seasons; (2) identify strains capable of inhibiting plant pathogens and tolerating environmental stresses such as temperature changes and UV radiation; and (3) test the most promising candidates in growth chamber and field trials, including assessing their effectiveness and impact on native microbial communities.

By developing locally adapted yeast-based solutions, the project accelerates the development of safer and sustainable crop protection in Estonia.

Supervisor(s): Margus Varjak

Mosquito-borne alphaviruses pose a significant threat, including chikungunya virus and Eastern equine encephalitis virus, which cause severe diseases such as arthritis and encephalitis, respectively. To date, no specific antiviral drugs are available against these viruses, and a vaccine is available only for one member of the alphavirus genus. Therefore, it is essential to develop new antiviral strategies.

This PhD project focuses on the alphavirus capsid protein (CP), which is crucial for virion assembly and host interactions. However, its interactome in cells remains poorly understood. Using a proteomics pipeline, the study aims to identify host proteins that interact with CPs of different alphaviruses, allowing the identification of shared and unique host factors across viruses.

Subsequently, the impact of these identified host factors on viral replication will be assessed to determine proviral factors. The proviral proteins are of significant interest, as they could serve as drug targets to inhibit virus growth. Targeting host proviral factors rather than the pathogen itself offers a potential alternative strategy to suppress multiple viruses with a single drug.

By the end of the PhD project, a comprehensive alphavirus CP interactome in host cells will be established, and novel therapeutic targets will be identified. Furthermore, mechanistic insights into alphavirus replication will enhance our understanding of fundamental biology.

Supervisor(s): Ilona Faustova

Bio-based adhesives derived from natural proteins offer a sustainable alternative to petroleumbased glues. Many organisms produce highly effective adhesive proteins that function in challenging conditions, such as underwater or on rough surfaces, providing inspiration for advanced bio-based bonding systems.

The Tem-TA124 project aims to develop a modular library of recombinant protein building blocks for creating efficient and customizable bio-based adhesive systems. This PhD project focuses on the final stage of protein production and testing, particularly on developing yeast-based adhesive solutions for sustainable construction materials.

During the project, novel adhesive formulations will be developed, optimized, and tested both independently and as components of composite materials. These materials will be used for prototyping architectural joints and components, 3D-printable inks, and other construction applications.

The research promotes circular resource use, reduces carbon emissions, and strengthens Estonia’s wood sector. By integrating architectural expertise with biotechnology and materials science, the project ensures the practical application of these bio-based adhesives in sustainable construction.

Supervisor(s): Veikko Linko, Olavi Reinsalu, Sven Oras

The goal of this PhD project is to advance genomic engineering in living systems by integrating emerging DNA nanotechnology with the groundbreaking CRISPR/Cas platform. The research will develop tailor-made gene-encoding DNA nanostructures designed to support efficient genomic insertion and controlled translation and expression in live cells. To enhance their performance in complex biological environments, these origami-based assemblies will be further optimized through modular molecular coatings. The resulting toolkit will open new scientific and bionanotechnological possibilities, supporting applications such as gene editing. Ultimately, these innovations will help address major global sustainability challenges in health, agriculture, and environmental protection.

Supervisor(s): Reet Kurg, Margit Mutso

Despite great advances in cancer treatment it remains a leading cause of death worldwide. Therefore, understanding the fundamental processes that promote tumour formation and cause tumour recurrence is crucial for the development of better and more effective new drugs. The aim of the project is to understand how cancer-testis antigens function during tumorigenesis. We focus on melanoma-associated antigens MAGEA proteins and try to elucidate their interactome in the cell. The second aim is to clarify the role of MAGEAs in inducing genomic instability. The proteomics, biochemical and microscopy approaches will be used to address the questions. Together, these results contribute to understanding how and whether MAGEAs function as oncogenic drivers.

Supervisor(s): Hanna Hõrak, Ljudmilla Timofejeva, Kristiina Laanemets

Stomata are small pores in leaves that mediate CO2 entry for photosynthesis and water loss via transpiration. Different plant groups have different stomata: mostly a stoma is made up of two kidney-shaped guard cells, but in cereal crops a stoma is formed by two dumbbell-shaped guard cells surrounded by two subsidiary cells. Most of the knowledge on the genetics of stomatal development and physiology originates from studies in the model plant Arabidopsis thaliana (Arabidopsis), less is known of stomatal developmental mechanisms in grasses, including cereal crops. Both Arabidopsis and barley produce stomata on both the upper and lower leaf surface, but it is not known how the distribution of stomata between leaf surfaces is controlled and whether same mechanisms are responsible for this process in Arabidopsis and barley. The aim of the project is to understand whether stomatal distribution between leaf surfaces is controlled by the same genes in Arabidopsis and barley and how mutations in respective genes affect leaf physiology, growth, and yield in barley. The project helps to understand how well results from studies of stomatal development and distribution in the model plant Arabidopsis translate to cereal crops. If they translate well, the project results will also show how stomatal distribution between leaf surfaces affects growth and yield in barley. The obtained knowledge can be used in breeding for barley varieties with more efficient stomatal distributions.

Supervisor(s): Hannes Kollist, Yuh-Shuh Wang

Photosynthetic CO2 assimilation takes place in plant chloroplasts. Guard cells that form stomatal pores can sense changes inside the plant as well as in the surrounding environment, and use this information to balance uptake of atmospheric CO2 for photosynthesis and water lost by transpiration. Within the cell, CO2 is hydrated and CO2/HCO3- equilibrium is formed. We have recently shown that CO2/HCO3--dependent interaction between two protein kinases, MPK12 and HT1, is the primary CO2-sensing mechanism in stomatal guard cells; however, the role of chloroplasts in CO2 sensing is not known. Based on our preliminary results, we hypothesize that HT1 can shuttle between plasma membrane and chloroplasts, thereby coordinating CO2 responses across the cell. In this project, we will verify this hypothesis and explore novel roles of the HT1 in guard cell signalling. We have already generated plant lines in with HT1 is either tethered to plasma membrane or chloroplasts and analysing their stomatal responses to CO2 changes will allow dissecting the involvement of specific organelle. Detailed assessment of the HT1 localization and interaction with partner proteins under changing CO2/HCO3- conditions will provide insights into its specific roles in different cellular compartments. Precise regulation of stomatal opening and closing is vital for plant growth because it provides means to regulate plant water use efficiency at given point of time. Completion of this project will advance our understanding in plant CO2 sensing and regulation, and ultimately provide knowledge basis on the future crop improvement in the future world with elevated CO2.

Supervisor(s): Hanna Hõrak, Pirko Jalakas

Stomata are small openings in the leaf surface that mediate carbon dioxide entry for photosynthesis and water loss via transpiration. Most plants have stomata only on the lower (abaxial) leaf surface (hypostomaty), and therefore most of the information on the development and opening-closing mechanisms of stomata originates from studying abaxial stomata. Some plants also form stomata on the upper (adaxial) leaf surface; such amphistomatous species include the model plant Arabidopsis thaliana (Arabidopsis). How adaxial stomata are formed and how stomatal distribution between leaf surfaces is regulated in amphistomatous plants is not known. During the project, hypostomatous Arabidopsis variants will be isolated from a mutagenized population, the mutations causing hypostomaty will be mapped, and the functions of the candidate genes found will be studied to understand their role in the formation of adaxial stomata. In addition, the PhD student will identify promoters that drive gene expression specifically in the abaxial or adaxial stomatal lineage cells and investigate how elevated air temperature affects the distribution of stomata between leaf surfaces. The project will help to understand how adaxial stomata are made and how stomatal distribution between leaf surfaces is controlled in plants.