My name is Dylan Nelson. I am currently a fifth year graduate student in astrophysics at the Center for Astronomy (CfA), Harvard University, in Cambridge, MA. I work with Lars Hernquist, modeling cosmological gas accretion, spiral structure and star formation on galactic scales. We run numerical simulations using Arepo, a finite volume hydrodynamics code based on a moving unstructured mesh.
Harvard-Smithsonian Center for Astrophysics
60 Garden Street, MS #10
Cambridge, MA 02138
All movies are h.264 encoded MKV/MP4 video files. If you need a compatible player I recommend VLC.
Moving mesh cosmology: Tracing cosmological gas accretion
Check out the: [Gadget/Arepo Halo Comparison Project]
We investigate the nature of gas accretion onto haloes and galaxies at z=2 using cosmological hydrodynamic simulations run with the moving mesh code AREPO. Implementing a Monte Carlo tracer particle scheme to determine the origin and thermodynamic history of accreting gas, we make quantitative comparisons to an otherwise identical simulation run with the smoothed particle hydrodynamics (SPH) code GADGET-3. Contrasting these two numerical approaches, we find significant physical differences in the thermodynamic history of accreted gas in massive haloes above 1010.5 solar masses. In agreement with previous work, GADGET simulations show a cold fraction near unity for galaxies forming in massive haloes, implying that only a small percentage of accreted gas heats to an appreciable fraction of the virial temperature during accretion. The same galaxies in AREPO show a much lower cold fraction, ‹20% in haloes of ~1011 solar masses. This results from a hot gas accretion rate which, at this same halo mass, is an order of magnitude larger than with GADGET, together with a cold accretion rate which is lower by a factor of two. These discrepancies increase for more massive systems, and we explain both trends in terms of numerical inaccuracies with the standard formulation of SPH. We note, however, that changes in the treatment of ISM physics -- feedback, in particular -- could modify the observed differences between codes as well as the relative importance of different accretion modes. We explore these differences by evaluating several ways of measuring a cold mode of accretion. As in previous work, the maximum past temperature of gas is compared to either a constant threshold value or some fraction of the virial temperature of each parent halo. We find that the relatively sharp transition from cold to hot mode dominated accretion at halo masses of ~1011, is a consequence of the constant temperature criterion, which can only separate virialised gas above some minimum halo mass. Examining the spatial distribution of accreting gas, we find that the filamentary geometry of accreting gas near the virial radius is a common feature in massive haloes above 1011.5 solar masses. Gas filaments in GADGET, however, tend to remain collimated and flow coherently to small radii, or artificially fragment and form a large number of purely numerical "blobs". These same filamentary gas streams in AREPO show increased heating and disruption at 0.25-0.5 virial radii and contribute to the hot gas accretion rate in a manner distinct from classical cooling flows. [arXiv]
This short video Compares two N-body (gravity only) simulations of a disk galaxy forming spiral arms due to the gravitational perturbations from massive "giant molecular clouds" (GMCs) which are represented as dots. The GMCs have a finite lifetime of 10Myr after which time they are reborn. On the left the formation is biased towards existing overdensities (i.e., GMCs are reborn within arms), while on the right the formation is unbiased (random). See D'Onghia et al. (2012) for more details.
Despite the several recognized methods for generating galactic spiral structure through interactions with external bodies, invoking a similar response in a secularly evolving galaxy is notably more difficult. Using high resolution (100 million particle) N-body simulations we model the evolution of an isolated stellar disk embedded in a Hernquist dark matter halo. We extend the work of D'Onghia et al. (2011) and consider the spiral structure arising from the gravitational influence of massive perturbers (e.g., giant molecular clouds) corotating in the stellar disk. Within a single rotation period we can develop a prominent, large-scale response in an initially smooth disk. This response represents the incoherent sum of small amplitude "wakes" generated by the swing amplifier acting locally in the neighborhood of each perturber. We systematically investigate this effect over a range of galaxy models.
By varying the critical wavelength with respect to axisymmetric instability, and requiring the disk to be everywhere stable by insuring Toomres Q>1, we explore the spectrum of spiral morphologies generated by the collective swing amplification mechanism. We make quantitative correlations between spiral structure and host galaxy and halo properties. We also predict the number of arms as a function of the strutural properties of the galaxy, and construct a classification catalog of arm morphologies to compare to spatially resolved observations of nearby galaxies. We measure radial variation of the spiral pattern speed using a modified Tremaine-Weinberg method, and find taht the pattern speed both decreases with radius and closely tracks the circular velocity of the disk, in excellent agreement with several recent observations. [Cefalu Poster] [ads]
The following four snapshots link to movies showing the evolution of projected surface density for the LC-2, LC-4, LC-6, and LC-8 models (from left to right) from t=0 to 1 Gyr. Each is a 35mb video file.
The barred galaxy verification simulation, steadily centered and rotating with a constant pattern speed of ~45 km/s/kpc at t > 600 Myr. 150mb video file.
MWA Correlator FoV Weighting (MIT Haystack)
Feasibility study of a correlator field of view weighting technique to address data volume and processing requirements of next generation radio telescope arrays (MWA and SKA), including mitigating the impact of excised frequency bands due to radio-frequency interference. [paper] [presentation]
Cloud Structure and the Origin of the IMF in rho-Ophiuchus (IfA Hawaii)
An in-depth, multi-wavelength comparison between the populations of dense, pre-stellar cores, and young, pre-main sequence stars in the rho-Ophiuchus region. We examined cloud structure and stellar content in order to probe the idea of a one-to-one mapping between core and stellar mass distributions.[paper] [presentation]
Weak Lensing Survey of Nearby SDSS Galaxy Clusters (Fermilab)
Investigation of galaxy clusters as probes of cosmology and the physics of structure formation. Cluster masses are derived through weak gravitational lensing measurements, using the Sloan Digital Sky Survey (SDSS) data. We presented weak lensing measurements of a sample of high-mass, low redshift (z < 0.1) clusters and found good agreement when compared with dynamical and X-ray estimates. [paper] [presentation]
LHC Tier2 Monitoring (INFN Roma)
Extension a monitoring/diagnostic system for the high-performance Tier2 grid designed for LHC data storage and analysis. Our implementation is a scalable, distributed system with a dynamic web-based interface provides status monitoring, diagnostic data visualization, and automatic fault notification and resolution. [presentation]
Some older writeups which occasionally get requested:
Brief descriptions of some of the more technical projects I am working on.
Arepo Visualization Toolkit (ArepoVTK)
ArepoVTK is designed to produce high quality, presentation-ready visualizations of hydrodynamic simulations run with Arepo. It performs volumetric ray tracing in 3D through linearly reconstructed scalar and vector fields defined on an unstructured Voronoi tessellation of space. It also includes higher order spatial interpolation techniques such as natural neighbor interpolation. Time interpolation between discrete snapshots is currently under investigation. The framework supports multi-dimensional transfer functions to investigate fluid quantities, and explores novel visualization techniques for combining such a volume rendering approach with coincident point particle datasets (both luminous and dark).
[Website coming soon.]
Delaunay Triangulation using Parallel Incremental Extrapolation on GPUs
We develop a method for constructing the Delaunay triangulation of a point set which is massively parallel and designed for the many-core architecture of graphical processing units (GPUs). We implement a "parallel incremental extrapolation" algorithm on the plane (2D) under the general position assumption and measure promising speedup with respect to our naive serial implementation.
Monte Carlo Tracer Particles on a Moving Mesh
Tracing the origin and (thermo)dynamical history of accreting gas in Eulerian grid codes requires Lagrangian “tracer particles”. The typical approach, whereby massless particles are passively advected by interpolating the local fluid velocity field, is found to exhibit systematic bias in its ability to trace the mass flow. An alternative, probabilistic “Monte Carlo” method associates tracers with parent gas cells and exchanges them based on mass fluxes through each face. The Poisson noise inherent in this approach is minimized with the ALE moving mesh scheme but may be intractable for strictly Eulerian AMR codes.
Front-Tracking Techniques for Multiphase Viscous Flow
We investigate the numerical simulation of multiphase fluid flow problems in two dimensions. In particular, we implement a front-tracking approach where a number of discrete points represent the free interface between two fluid phases. This boundary is advected in time, and at each timestep we calculate the surface curvature and include a model for surface tension effects. The Lagrangian surface is coupled to a fixed, Cartesian grid mapped to a square domain. The incompressible Navier Stokes equations are used to model continuum fluid flow of both phases, which have different densities and physical viscosities. We use the projection technique to split the second order time update into an advection-diffusion step following by a pressure correction step to enforce the divergence free constraint, while the spatial discretization uses the finite volume approach with staggered, rectangular control volumes for pressure and velocity. We investigate the numerical accuracy and convergence of the curvature calculation, area conservation of a high density drop surrounded by low density air, and the generation of spurious numerical velocities. The approach is then used to simulate the bounce of a water drop off of a rigid boundary. A proof of concept boundary merger algorithm is presented to handle the topological change of two colliding water drops, and extensions to more accurate numerical methods and physical models are discussed.
[Also related: Continuous Galerkin Navier-Stokes in 2D Writeup.]
These experiments are almost all works in progress. Comments welcome!
The Universe in Gas
"The Universe in Gas" shows a large volume of a simulated universe, 20 Mpc/h on a side, at redshift zero (the present time). Volume rendering highlights iso-surfaces of gas density, temperature, and their relation. Bright peaks in the density reveal galaxies, which are surrounded by their hot halo atmospheres, and interconnected with filaments arising from the large scale structure of the universe. This animation visualizes results from a numerical hydrodynamical simulation of a cosmological volume run with the moving mesh code AREPO. Included are the effects of gas (baryons), dark matter (not shown), as well as stars and black holes (also not shown) and their energetic feedback processes. Made with ArepoVTK. [HD download]
Coffee Cup Problem (3D)
An impressive early animation made by V. Springel for the AREPO code was a 2D box within which a moving, curved solid boundary moves in a circular motion, meant to represent a spoon stirring a cup of coffee. We extend this to toy problem to 3D with an importer which creates an initial condition from any STL surface mesh - in this case, a spoon. Volume rendered with ArepoVTK, illuminating the spoon and highlighting density features due to mixing and the development of KH-like instabilities. [HD download]
Gadget/Arepo Halo Comparison Project (v1)
We present a large catalog of several thousand halos extracted from cosmological simulations. Each is shown from three orthogonal views and with different rendering techniques - velocity field scatter plots, SPH kernel projection maps, and a large scale comparison with the dark matter field. A range of halo masses is rendered at each z=0,1,2,3 and individual halos are matched between the two simulations. [Visit Website]
Cosmological Gas Accretion Trajectories
This animation shows the trajectories of individual gas elements (on the left) and dark matter particles (on the right) which are bound to the same halo at redshift zero, evolving in time from redshift four.
When things go wrong...
ArepoVTK Development Gallery
A scrapbook of sorts of the ongoing development of ArepoVTK. Some extremely simple test meshes exploring different rendering techniques and approaches for transfer functions, as well as mesh visualization. Some galactic disks, face-on and edge-on. Some whole box and zoom-ins from cosmological hydro simulations.
These WebGL experiments are all works in progress and may frequently change. Any modern version of Firefox or Chrome should have no problem with these examples. Comments welcome!
Interactive Gadget/Arepo Halo Comparison Project (v2)
The interactive Gadget/Arepo halo comparison project is a WebGL experiment that compares the gas distribution around two dark matter halos. Part of the Moving Mesh Cosmology simulations, one halo is taken from a cosmological simulation run with the well-known SPH code GADGET3. The other is a matched object run with the new moving mesh code AREPO. The experiment allows the user to manipulate the view, the fluid quantities which are displayed, and the rendering method employed. [Launch Now]
Tracing Cosmological Gas Accretion... Through Time
This second experiment visualizes the time evolution of tracer particles as they accrete into a galaxy at low redshift. Their trajectories relative to the evolving halo are animated, while the maximum temperature they obtain between each point is represented by a color mapping. Catmull-Rom splines interpolate between tracer positions at discrete snapshots. The evolving radius and virial temperature of the parent halo are represented by the changing virial sphere. In addition to viewing one instant at time, the radial mode also allows us to move all tracers to the same radius and investigate the time-collapsed geometry and thermal heating. [Launch Now]
Structure of 3D Voronoi Tesellations
We implement the single-pass, shader-based wireframe rendering technique of Bærentzen et al. (2008) in WebGL (no geometry shaders), for the case of arbitrary polygonal faces. A Voronoi mesh is exported as its constituent faces, each having N vertices and requiring N+2 triangles in our approach. The edges and interiors of each face are simultaneously rendered by the fragment shader, using at each pixel the window space distance to the edge of the face. For large meshes, we can "illuminate" only a slab or a radial shell. The geometry can also be "exploded" by radially displacing each cell center. Note: The last two data sets (diego_disk and halo314) are extremely large (>2GB card required, e.g. GTX 670) and may make your browser unstable. With this technique we can render at most ~100k cells, corresponding to ~1mil faces and ~5mil triangles. Stereoscopic 3D support for side-by-side type systems (e.g. Oculus Rift, or see the CfA dual polarized projector setup). [Launch Now]
Dylan Nelson :: Curriculum Vitae
Mail Stop 10, 60 Garden St, Cambridge, MA 02138
- Cosmological Gas Accretion and Disk Formation
- Galactic Scale Star Formation, Spiral Structure
- Hydrodynamic/N-body Numerical Simulations, Methods, and Visualization Techniques
Harvard University, Cambridge, MA
- September 2009 - Present
- Astrophysics PhD Program
- Secondary Field: Computational Science and Engineering (CSE)
University of California Berkeley, Berkeley, CA - Graduated 2008
- Triple Major: Physics, Astrophysics, and Mathematics
- Member: National Society of Collegiate Scholars
- Junior Member: American Astronomical Society
- Honors Standing: September 2004 - May 2008
Montgomery High School, Santa Rosa, CA - Graduated 2004
- Full International Baccalaureate (IB) Diploma
- Member: California Scholarship Association
AWARDS AND RECOGNITION:
- Harvard Institute for Applied Computational Science (IACS) Student Fellowship: 2014
- National Science Foundation (NSF) Graduate Research Fellowship: 2009-2011 - Harvard
- John P. Merrill Graduate Fellowship: 2009-2010 - Harvard
- Pomerantz Physics Scholarship Recipient: 2007-08 - UC Berkeley
- Dean's Honor List: Fall 2005 - UC Berkeley
PUBLICATIONS AND PROCEEDINGS:
Vogelsberger, M., Genel, S., Springel, V., Torrey, P., Sijacki, S., Xu, D., Snyder, G., Bird, S., Nelson, D. R., Hernquist, L. (2014). Simulating the coupled evolution of dark and visible matter in the Universe, in prep.
Nelson, D.R., D'Onghia, E., Hernquist, L. (2011). The Morphology and Pattern Speed of Spiral Structure. Advances in Computational Astrophysics Conference Proceeding, cefalu 2011. [ads]
Kubo, J. M., Annis, J., Hardin, F. M., Kubik, D., Lawhorn, K., Lin, H., Nicklaus, L., Nelson, D. R., Reis, R. R., Seo, H-J., Soares-Santos, M., Stebbins, A., Yunker, T. (2009). The Sloan Nearby Cluster Weak Lensing Survey, ApJ, 702, 110. [ads] [arXiv]
Nelson, D. R., Swift, J. J., Williams, J. P. (2007). Cloud Structure and the Origins of the Stellar Initial Mass Function in rho-Ophiuchus, AAS Meeting 211, Austin, Session #89.12. [ads]
Nelson, D. R., Doeleman, S. S., Lonsdale, C. J., Oberoi, D., Cappallo, R. J. (2006). Effectiveness of the Correlator Field of View Weighting Technique in Source Attenuation, AAS Meeting 209, Seattle, Session #85.10. [ads]
Harvard University Teaching Fellow:
- Astronomy 16 - Stellar and Planetary Astronomy (Prof. D. Charbonneau)
- Astronomy 17 - Galactic and Extragalactic Astronomy (Prof. J. Lee)
- Applied Computation 274 - Computational Fluid Dynamics (Prof. D. Knezevic)
Development Languages and Environments:
- C/C++ (inc. MPI, pThreads, and CUDA)
- Python, IDL
- Mathematica, Matlab
PREVIOUS RESEARCH EXPERIENCE: