I have worked on simulations of accretion flows on scales up to 10,000 r_g (approximately 0.5 pc for a 10⁹ M_⊙ SMBH). These simulations, performed using the codes PLUTO and H-AMR, target the production of outflows in radiatively inefficient systems with Eddington ratios below 0.01. In such regimes, radiation pressure is negligible, and the resulting outflows consist of relativistic jets and hot, magnetized winds. Focusing on these winds—which become particularly relevant for systems with low black hole spin—we find that they carry a substantial amount of energy from subparsec scales, with wind luminosities of L_wind ∼ 0.01–0.08 Ḿc², corresponding to approximately 10^39–40 erg/s in physical units. Given winds this powerful, our estimates suggest that SMBHs with M_BH ≳ 10⁸ M_⊙ and accretion rates ˙M ≳ 10^-3 ˙M_Edd can reduce the star formation rate by more than 10% over a period of 10⁷ years of activity.
🔗 Paper 1 Paper 2

I use hydrodynamic simulations performed with the moving-mesh code AREPO to study the interaction between small-scale AGN winds, with speeds of approximately 10,000 km/s, and galactic discs containing an idealised, clumpy interstellar medium (ISM).
In my ongoing work (Almeida et al. 2025, in preparation), I use high-resolution simulations to explore how AGN-driven winds interact with a multiphase interstellar medium. By implementing a refinement scheme that captures rapidly cooling, fast-moving gas, the simulations achieve sub-parsec resolution in outflowing regions with temperatures below 20000 K.
This project investigates how cool clouds form and evolve within AGN outflows, and how their properties may depend on the AGN’s power and the structure of the surrounding ISM. The results have important implications for interpreting observed outflow rates and for developing more accurate models that connect small-scale AGN physics with galaxy-scale feedback.
Stay tuned — the paper will be published soon!

Jets are ubiquitous in AGN systems, carrying away enormous amounts of energy and depositing it far from their origin. However, the plasma composition of AGN jets remains a long-standing open question in the field. To investigate this, I am performing numerical simulations of jets at parsec scales, predicting their multi-wavelength emission, and exploring the contributions of different particle species to the observed signatures, including variability and polarization.
Our approach relies on relativistic MHD simulations using PLUTO, evolving the gas while also tracking Lagrangian particles that represent ensembles of different species (e.g., electrons, positrons, hadrons). We compute the physical properties of these particles and analyze how they contribute to emission across various wavelengths. The study explores a range of effects, from redshift-dependent emission features to advanced physical models that incorporate resistivity and detailed treatments of particle acceleration at shocks.

To characterize the emission from LLAGNs, I use a combination of radiatively inefficient accretion flow (RIAF), relativistic jet, and hot wind models to build theoretical SEDs that span from radio to gamma rays. These models allow us to test how different components dominate at different wavelengths and help constrain physical properties such as black hole spin, mass accretion rate, and jet power. Through this approach, I investigate how even weakly accreting black holes may impact their surroundings.
Recognizing the computational demands of fitting complex SED models to large datasets, I developed a deep learning approach capable of reproducing theoretical LLAGN SEDs with a speed-up of more than 1,000× compared to traditional methods. This tool not only enables faster and more scalable analysis but also opens the door to fitting LLAGN SEDs in large surveys. The model was successfully applied in Almeida et al. (2022), demonstrating its potential as a practical and accurate tool for interpreting low-luminosity black hole emission in a variety of astrophysical environments.