Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de...

15
Cosmography and Data Visualization Daniel Pomar` ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit´ e Paris-Saclay, 91191 Gif-sur-Yvette, France el` ene M. Courtois Universit´ e Claude Bernard Lyon I/CNRS/IN2P3, Institut de Physique Nucl´ eaire, Lyon, France Yehuda Hoffman Racah Institute of Physics, Hebrew University, Jerusalem 91904, Israel R. Brent Tully, Institute for Astronomy, University of Hawaii, 2680 Woodlawn Drive, Honolulu, HI 96822, USA ABSTRACT Cosmography, the study and making of maps of the universe or cosmos, is a field where vi- sual representation benefits from modern three-dimensional visualization techniques and media. At the extragalactic distance scales, visualization is contributing in understanding the complex structure of the local universe, in terms of spatial distribution and flows of galaxies and dark mat- ter. In this paper, we report advances in the field of extragalactic cosmography obtained using the SDvision visualization software in the context of the Cosmicflows Project. Here, multiple vi- sualization techniques are applied to a variety of data products: catalogs of galaxy positions and galaxy peculiar velocities, reconstructed velocity field, density field, gravitational potential field, velocity shear tensor viewed in terms of its eigenvalues and eigenvectors, envelope surfaces en- closing basins of attraction. These visualizations, implemented as high-resolution images, videos, and interactive viewers, have contributed to a number of studies: the cosmography of the local part of the universe, the nature of the Great Attractor, the discovery of the boundaries of our home supercluster of galaxies Laniakea, the mapping of the cosmic web, the study of attractors and repellers. 1. Introduction Throughout the ages, astronomers have strived to materialize their discoveries and understanding of the cosmos by the means of visualizations. The oldest known depiction of celestial objects, the Ne- bra sky disc, dates back from the Bronze age, 3600 years ago (Benson 2014). While most of the astro- nomical representations are projections to two di- mensional sketches and images, the introduction of the third dimension in depictive apparatuses has been sought by astronomers as an essential mean to promote understanding. Such objects as the armillary spheres, dating back from the Hel- lenistic world, or the modern era orreries used to mechanically model the solar system, played an important role in the history of astronomy. To- day, computer-based interactive three-dimensional visualization techniques have become a fruitful re- search tool. Here, we present the impact of vi- sualization on cosmography in the context of the Cosmicflows Project. 1 arXiv:1702.01941v1 [astro-ph.CO] 7 Feb 2017

Transcript of Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de...

Page 1: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

Cosmography and Data Visualization

Daniel PomaredeInstitut de Recherche sur les Lois Fondamentales de l’Univers

CEA, Universite Paris-Saclay, 91191 Gif-sur-Yvette, France

Helene M. CourtoisUniversite Claude Bernard Lyon I/CNRS/IN2P3, Institut de Physique Nucleaire, Lyon, France

Yehuda HoffmanRacah Institute of Physics, Hebrew University, Jerusalem 91904, Israel

R. Brent Tully,Institute for Astronomy, University of Hawaii, 2680 Woodlawn Drive, Honolulu, HI 96822, USA

ABSTRACT

Cosmography, the study and making of maps of the universe or cosmos, is a field where vi-sual representation benefits from modern three-dimensional visualization techniques and media.At the extragalactic distance scales, visualization is contributing in understanding the complexstructure of the local universe, in terms of spatial distribution and flows of galaxies and dark mat-ter. In this paper, we report advances in the field of extragalactic cosmography obtained usingthe SDvision visualization software in the context of the Cosmicflows Project. Here, multiple vi-sualization techniques are applied to a variety of data products: catalogs of galaxy positions andgalaxy peculiar velocities, reconstructed velocity field, density field, gravitational potential field,velocity shear tensor viewed in terms of its eigenvalues and eigenvectors, envelope surfaces en-closing basins of attraction. These visualizations, implemented as high-resolution images, videos,and interactive viewers, have contributed to a number of studies: the cosmography of the localpart of the universe, the nature of the Great Attractor, the discovery of the boundaries of ourhome supercluster of galaxies Laniakea, the mapping of the cosmic web, the study of attractorsand repellers.

1. Introduction

Throughout the ages, astronomers have strivedto materialize their discoveries and understandingof the cosmos by the means of visualizations. Theoldest known depiction of celestial objects, the Ne-bra sky disc, dates back from the Bronze age, 3600years ago (Benson 2014). While most of the astro-nomical representations are projections to two di-mensional sketches and images, the introductionof the third dimension in depictive apparatuses

has been sought by astronomers as an essentialmean to promote understanding. Such objects asthe armillary spheres, dating back from the Hel-lenistic world, or the modern era orreries used tomechanically model the solar system, played animportant role in the history of astronomy. To-day, computer-based interactive three-dimensionalvisualization techniques have become a fruitful re-search tool. Here, we present the impact of vi-sualization on cosmography in the context of theCosmicflows Project.

1

arX

iv:1

702.

0194

1v1

[as

tro-

ph.C

O]

7 F

eb 2

017

Page 2: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

The discipline of cosmography is both very oldand very young. Since ancient times, societieshave tried to understand their place in the cos-mos and have created representations to expresstheir ideas. Today we have learned that we livein an expanding universe, 13.7 billion years old,filled with mysterious dark matter and dark en-ergy, with galaxies in lattices of the cosmic web.However our representations of the structure ofthe universe, cosmography, are remarkably primi-tive. We only have good knowledge of a tiny frac-tion of the potentially visible universe. The questof the present research is to substantially improveour understanding of our local neighborhood of theuniverse. Inevitably, what we know degrades withdistance. However, nearby our emerging picture isrich in detail and, farther away, the outer bound-aries where our maps blur into the backgroundconfusion are being pushed back dramatically.

The Cosmicflows Project aims at the recon-struction and mapping of structure of the LocalUniverse using peculiar velocities of galaxies astracers of the source density field; it has releasedthree catalogs, Cosmicflows-1 with 1791 galax-ies (Tully et al. 2008), Cosmicflows-2 with 8161galaxies (Tully et al. 2013), Cosmicflows-3 with17669 galaxies (Tully et al. 2016). This projecthas its foundations built on earlier works; TheNearby Galaxies Atlas (Tully & Fisher 1987) is afirst cartography of the three dimensonal struc-ture of the Local Universe within redshifts of3000 km/s, in association with the data docu-mented in the Nearby Galaxies Catalog (Tully1988). This Atlas highlights the location of ourgalaxy on the periphery of the Local Void. Thedistribution of galaxies is dominated by the pres-ence of the Virgo cluster. Then, using the unionof Abell and Abell-Corwin-Olowin catalogs of richclusters, a cartography of structures extending to30,000 km/s was published by Tully et al. (1992).Maps of the distribution of superclusters are pro-vided in the plane of the Supergalactic Equatorthat host the Great Attractor, the Great Wall,the Shapley Concentration, the Pisces-Cetus Su-percluster (Figure 4 in Tully et al. 1992). Mapsof orthogonal slices display the structure at higherSGZ altitudes with the Hercules, Corona-Borealisand Aquarius-Capricornus superclusters, and atlower SGZ altitudes with the Lepus, Horologium-Reticulum and Sextans superclusters (Figures 3

and 5). This early cartography was severely lim-ited by the Zone of Avoidance associated with theobscuration caused by the dust and stars of ourown galactic disk. This Zone of Avoidance createsa three-dimensional wedge inside which direct ob-servations are not possible. It is a strength ofthe Cosmicflows program that it provides a recon-struction of the structures lying within the Zone ofAvoidance using the gravitational influence theyexert on galaxies lying nearby. Three-dimensionaldensity contour maps within 8000 km/s were pre-sented by Hudson (1993), see Figure 10 therein;these show from two complementary viewpointshow filaments and high-density blobs such asVirgo-Centaurus-Hydra, Perseus-Pisces, Pavo,Coma are organized. Using the 2M++ galaxyredshift compilation based on 2MRS, 6dFGRS-DR3 and SDSS-DR7, Lavaux and Hudson (2011)presented maps of the density field within 15,000km/s. Using the Cosmicflows-1 Catalog, Courtoiset al (2012) presented maps of both density andvelocity fields within 3000 km/s with reconstruc-tion of the Local Void and the Great Attractor.

2. The Cosmicflows program

The objective of the Cosmicflows program is tomap the distribution of matter and flows in the lo-cal universe, understand the motion of our galaxywith respect to the CMB, and provide new insightsin cosmology by providing near-field measurementof basic parameters such as the Hubble Constant.The fundamental ingredients of this program arethe measurements of peculiar velocities of galaxies,that is their deviation with respect to the Hubbleexpansion. These peculiar velocities of galaxiesserve as sensors of the source gravitational field,including dark matter. The Cosmicflows programhas three critical constituents: observations, theo-retical modeling, and visualizations. Each of theseis state-of-the art.

On the observational side, the measurement ofaccurate distances to galaxies permits the separa-tion of deviant motions from the cosmic expansion(Tully et al. 2008, Tully et al. 2013, Tully et al.2016). Our program has accumulated, by far, thelargest and most coherent assembly of galaxy dis-tances.

The translation of measured radial velocitieswith substantial errors into a three-dimensional

2

Page 3: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

map of galaxy motions is accomplished with aWiener Filter technique averaging multiple real-izations constrained in a Bayesian analysis by thedata and an assumed power spectrum of initialdensity fluctuations (Zaroubi et al. 1999, Hoff-man 2009, Courtois et al. 2012, Doumler et al.2013). The analysis gives, in addition to the 3-Dvelocity field, the density and potential fields, andthe velocity field can be separated into local andtidal components. The shear of the velocity fieldat each position defines eigenvectors and eigenval-ues of the V-web, descriptors of whether the loca-tion is in a knot, filament, sheet, or void (Hoffmanet al. 2012). The analysis leads to the identifi-cation of all the important basins of gravitationalattraction and repulsion.

It is the role of the visualizations to clarifythe interplay between these various componentsin quite complicated circumstances. Cartographyhas played a seminal role in the development ofour current understanding. With new data sets,already arriving and projected, the visualizationchallenges are multiplying.

3. The SDvision visualization software

The SDvision (Saclay Data Visualization) soft-ware is deployed in the IDL “Interactive Data Lan-guage” platform (Pomarede et al. 2008). Origi-nally developed in the context of the COAST“Computational Astrophysics” Project for the vi-sualization of astrophysical simulations (Audit etal. 2006), it was realized that this software couldalso be used to visualize cosmographic data (Po-marede & Pierre 2011, Pomarede et al. 2013).IDL was chosen for its widespread and long-termuse in the Astronomy community, the profes-sional support and development plan offered byits owner company, the extensive astronomicallibraries it offers, and for the high-performancevisualization techniques it provides: IDL givesaccess to hardware-accelerated rendering tech-niques through its “Object Graphics” interfacesto OpenGL, as well as multi-threaded usage ofmultiple-core processors. The graphics objectscan be coupled with GLSL (OpenGL ShadingLanguage) shader algorithms to perform both sci-entific computation and visualization by graphicscard. The SDvision visualization software consiststoday in 100,000 lines of code addressing the issues

of strategic importance in the field of cosmogra-phy: visualization of scalar fields, vector fields andclouds of points.

The SDvision widget interface provides a multi-tude of tools to generate visual objects and act ontheir properties. Built around a main view win-dow where rotations, translations and scaling canbe obtained by mouse click-and-drag actions, thewidget provides facilities relevant to cosmographysuch as sliders allowing on-the-fly geometrical cutson the galaxy catalog on display. The widget in-terface can be seen in Figure 1 in the context ofthe visualization of scalar fields. Here, the useris presented with a histogram. The user can clickinteractively on this histogram, an action that willresult in the computation and updated visualiza-tion of the corresponding isosurface. This inter-active facility is most useful for the exploration ofcomplex scalar fields, especially when coupled tothe 3D navigation possibilities. The widget inter-face can also be seen in Figure 2 in the contextof the visualization of a catalog of galaxies. Ageometrical filtering is obtained by action on thesliders labelled xmin, xmax, ymin, ymax, zmin,zmax. Among the ∼ 315, 000 galaxies of the cat-alog, ∼ 33, 000 are selected after applying thesecuts.

Scalar fields, such as density or temperaturefields, are visualized by the means of three com-plementary techniques: 1) Ray-casting volumerendering. The ray-casting is a CPU-intensivetechnique that propagates a ray through the vol-ume under scrutiny and builds-up on contributionsfrom crossed cells, thus requiring a complete newcomputation at every change in the viewpoint.In this technique, several composite functions areavailable to measure the value of a pixel on theviewing plane by analyzing the voxels falling alongthe corresponding ray: the most basic functionis the Maximum Intensity Projection where thecolor of each pixel on the viewing plane is de-termined by the voxel with the highest opacityvalue along the corresponding ray, the color ofthe voxel being obtained against some lookup ta-bles providing the three R, G, B colors and opac-ity functions. A more sophisticated compositingfunction is the Alpha blending, where a recur-sive equation assumes that the color tables havebeen pre-multiplied by the opacity table to obtainsemi-transparent volumes. This technique is pre-

3

Page 4: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

ferred where multiple layers of structures can hideeach other. Finally the ray-casting algorithm canbe engaged to produce RGBA renderings wherered, green, blue, and alpha channels are associatedto four different physical fields. The ray-castingalgorithm is multithreaded, exploiting all avail-able computing cores on shared-memory comput-ing systems. 2) Isosurfaces reconstruction is usedto display surfaces of constant value taken by thescalar field. The reconstruction is performed byIDL’s SHADE VOLUME procedure, that is simi-lar to the Marching-cube algorithm. The resultingsurface is visualized as a Gouraud-shaded polygon.The interactive visualization of this polygon bene-fits from the hardware acceleration by the graphicscard. Figure 1 shows an example of surfaces re-constructed with this technique. 3) Slicing is usedto map a texture on a simple slice of the volumeunder scrutiny, at any position and orientation.

Vector fields, such as velocity fields or mag-netic fields, are visualized by two techniques: 1)Streamlines reconstruction using IDL’s PARTI-CLE TRACE procedure. This procedure tracesthe path of a massless particle through a vectorfield, given a set of starting points (or seed points).Particles are tracked by treating the vector field asa velocity field and integrating. Each particle istracked from the seed point until the path leavesthe input volume or a maximum number of itera-tions is reached. The vertices generated along thepaths are returned packed into a single array alongwith a polyline connectivity array used to feed apolyline object. The collection of seeds can obeysome predefined distribution (2D or 3D uniformgrid, spherical grid) or seeds can be selected in-teractively by clicking on any graphics object ondisplay, a most useful feature to study the struc-ture of flows. 2) Hedgehog display where three-dimensional arrows are anchored on the same col-lection of points or seeds as in the streamlines pro-cedure.

Point Clouds can be used to materialize the po-sitions of galaxies obtained from catalogs. Theyare visualized using two techniques: 1) usingmarkers of definite sizes (polylines or polygons).This is useful to get varying apparent sizes of themarkers versus perspective and zooming. 2) Us-ing sprites. In this technique a sprite shader forcesthe adjacent pixels of the projected position of thepoints to participate in the rendering, resulting in

markers of fixed apparent sizes, independent ofany perspective or zooming. This hardware- ac-celerated technique is extremely fast and efficientfor the rendering of larger numbers of cloud points.

The SDvision software can be controlled eitherinteractively or through scripts describing a se-quence of commands. This latter option is usedmainly to produce videos exposing the evolutionin a simulation or to explore a volume followingpredefined selected routes. It has facilities to pro-duce Stereo3D outputs and 8-cam outputs for au-tostereoscopic screens. The favored data structurefor scalar and vector fields is the uniform grid.More sophisticated data structures such as AMR(Adaptive Mesh Refinement) are handled by pro-jection onto a uniform grid of adequate resolution.The algorithms implemented in SDvision favor theuse of shared-memory architectures equipped bymultiple-core processors (to benefit from the ray-casting rendering technique) and massive avail-able RAM (to manage large datasets, in particularlarge 3D datagrids) and with high-range graphicsunits (for the acceleration of polygon renderingand use of GLSL Shaders).

4. Cosmography use cases

In this section we illustrate the cosmographicpotential of our software in four use cases: 1)the visualization of galaxy catalogs, 2) the map-ping of cosmic flows, 3) the visualization of three-dimensional basins of attraction, and 4) the car-tography of the cosmic web.

4.1. Visualization of catalogs of galaxies

Catalogs of galaxies may be produced that con-tain any number of galaxies, up to several mil-lions. The visualization of their distribution isa key ingredient in cosmography. An example ofrunning a visualization of the XSCz catalog of red-shifts (Jarrett et al. 2000) is displayed in Fig-ure 2. In this example, the galaxies are mapped inthe Supergalactic Coordinate System in units ofredshift (km/s). A thin slice -1000 km/s < SGX< +1000 km/s is selected. The markers used tomaterialize the positions of the galaxies can ei-ther be spheres or sprites. The rendering of manyspheres is time-consuming, however it has the ad-vantage of giving a physical size to the galaxymarkers; individual galaxies, or groups, or clusters

4

Page 5: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

can be approched during the exploration of thedata, with their individual object apparent sizeincreasing during zoom-ins. On the other hand,sprites are rendered in real-time, but they occupya fixed number of pixels whatever the position ofthe eye; their apparent size does not change whenzooming in or out. In terms of cosmography, theSDvision interface provides useful tools: for in-stance the possibility to select rectangular slices(using interactive sliders and fields visible in Fig-ure 2 on the left-hand side of the widget). Slicescan be also oriented at any angle as a functionof supergalactic longitute and latitude. Spher-ical shells and cylindrical cuts can also be per-formed. The reference coordinate system can bedisplayed. A cosmography-dedicated interface isproposed through which several tools are avail-able: for example a galaxy finder, upon activation,will find all the galaxies within a given search ra-dius from any clicked point, and print their IDs,names and positions. The visualization proposedin Figure 2 is typical of the maps obtained usingredshift surveys: it reveals a web of filaments con-necting clusters of galaxies and separating emptyvoids. This is the cosmic web. The first such mapswere published 30 years ago (de Lapparent et al.1986). The 3D visualization of galaxy catalogs al-lows the exploration and the mapping of the struc-tures in galaxy distribution (see e.g. Courtois etal. 2013). It is also useful to perform compar-isons with reconstructed products that have com-plex three-dimensional architectures, such as thedensity field, basins of attraction, or the CosmicV-web (see sections below).

In terms of visualization of galaxy catalogs, theCosmicflows Project brings in an additional mat-ter: the need to visualize a radial velocity mea-surement tied to each galaxy position. Such “pe-culiar velocities” are materialized as 3D vector ar-rows, as exemplified in Figure 5. These vectorsare the primary data input to the Wiener Filteralgorithm that reconstructs the fully three dimen-sional velocity field.

4.2. Mapping cosmic flows

Cosmic flows are revealed in the visualization ofthe velocity vector field. Our favored means of vi-sualizing a three-dimensional velocity vector fieldis to use streamlines. Streamlines are polylines ac-counting for the reconstruction of a path from an

initial “seed” position and proceeding by steps ofconstant size along the direction dictated by thelocal velocity arrow. The SDvision widget inter-face allows interactive exploration of various stepsizes (for a given line length, shorter step sizesrequire more computing time) and various totalstreamline lengths. Short streamlines can be usedto explore the structures in local computations ofthe flow field, while long streamlines are useful tomaterialize the large scale structure manifested inthe computations of the full flow. The visualiza-tion of streamlines offers certain benefits over thestandard vector-based representation (that is alsoimplemented in SDvision). The streamlines depictclearly and robustly the geometry of the flow. Theflow field (presented in Figure 3) is characterizedby a convergence and by a divergent point, that wecall an attractor and a repeller. It further showsthe filamentary nature of the flow. The demon-stration is much more difficult to realize from thearrows visualization. It is the continuity of thestreamlines, and the clear representation of thesources and sinks of the streamlines that make thedifference. In addition, the coloring of the stream-lines provides a very clear visualization of the am-plitude of the local velocity field.

An important aspect of streamline visualizationis the seeding. The baseline configuration is to dis-tribute the seeds on a Cartesian 3D uniform grid.This is an efficient way to obtain a global mapof the cosmic flows. Such visualizations reveal es-sential structures in the flow: the convergence onattractors for instance, or the compression of thelines into filaments. In Figure 3, the seeds arelocated on the nodes of a 113 uniform Cartesiangrid, providing a reasonably nuanced view of theglobal structure of the flow. In Figure 4, left, alighter sample of 53 seeds is used. This seedingdoes a good job at finding an attractor but doesnot inform much on the rest of the volume. In Fig-ure 4, right, a heavier sample made of 213 seeds isused. Here the view is saturated with information.These examples illustrate how the seeding affectsthe visualization.

Since the seeding along three directions can bethe source of confusion between lines located atvarious depths, it is interesting to restrict the seed-ing to planes. The SDvision widget allows one tointeractively scan through the volume by seedingthe streamlines on planes normal to each cardi-

5

Page 6: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

nal direction. The range and number of seeds canbe tuned interactively to optimize the information,keeping in mind that too many streamlines can re-sult in confusion, while too few can result in miss-ing essential information. Another way to seed thestreamlines is to use an ancillary source. For in-stance, the position of galaxies from any catalogcan be used as seeds. This is what is achievedin Figure 5. This is especially interesting in thecontext of the Wiener Filter analysis, as we canvisualize together the input peculiar radial veloc-ity of the algorithm, and its output fully 3D recon-structed peculiar velocity. Another type of seedingis to take at random uniform distribution, for in-stance in a 3D volume, or confined to a 2D plane.This latter option is illustrated in Figure 6 withshort streamlines, providing a map of local cos-mic flows in the plane of the Supergalactic Equa-tor. Finally, it is also possible to combine random-ness and ancillary constraints, by e.g. distribut-ing seeds randomly inside or outside an isosurfaceof any given scalar field, or any other volume de-fined in any mathematical way. An example ofsuch seeding is presented in Figure 7 where seedsare distributed randomly and uniformly within thesurface of the Laniakea basin of attraction (see fol-lowing section).

4.3. Basins of attraction and the definitionof superclusters

A basin of attraction is a volume inside whichthe flow exhibits convergence onto a unique point.This notion emerged when this convergence wasfirst observed in the Wiener Filter reconstructionof the Cosmicflows-2 Catalog (Tully et al. 2013).It was proposed to use this notion as a definitionfor superclusters of galaxies, and the application ofthis idea resulted in the identification of the Lani-akea supercluster, our home supercluster (Tully etal. 2014). The visualization of basins of attractionis illustrated in Figure 7, where the Laniakea basinof attraction is displayed together with the Arrow-head basin of attraction (Pomarede et al 2015) assemi-transparent polygons. The concurrent visu-alization of streamlines of the velocity field seededrandomly within the Laniakea basin of attractiondemonstrates the convergence of the flow onto asingle attractor. The structure of the Laniakeabasin of attraction in the plane of the Supergalac-tic Equator can also be appreciated in Figure 6

where the tiling of the universe in adjacent basinsof attraction can be inferred.

4.4. The Cosmic V-Web

The analysis of the velocity field in terms ofexpansion and compression of the flow providesthe signatures of a web-like structure called theCosmic Velocity Web, or V-Web (Hoffman et al.2012). At each point in space, four possible situa-tions are identified:

• a compression of the flow observed in threeorthogonal directions is a signature of aknot. Such a knot is acting as an attractorwhere the flow is converging.

• a compression of the flow observed in twoorthogonal directions and an expansion ob-served in the third direction is a signatureof a filament. The filament collects the flowfrom its surroundings and direct it along asingle direction.

• an expansion in two directions and compres-sion in the third direction is the signaturefor a sheet. The flow runs freely along twodirections and experience compression in thethird.

• an expansion in all three orthogonal direc-tions is a signature of a void, a region thatcan be considered as a repeller from whichan expulsion is ongoing.

These signatures are expressed mathematically bythe velocity shear tensor. The velocity shear ten-sor is characterized in terms of its three eigenval-ues ordered as follows: λ1 > λ2 > λ3.

The mathematical prescriptions for the knots, fil-aments, sheets, and voids are the following:

λ3 > λ3threshold > 0 is signature of a knot.

λ2 > λ2threshold > 0 is signature of a filament.

λ2 < λ2threshold < 0 is signature of a sheet.

λ1 < λ1threshold < 0 is signature of a void.

Figure 1 presents a visualization of the V-Webin terms of its knots and filaments, using multi-ple surfaces of constant values of λ3 and λ2. Theemerging structure is that of a three-dimensionalweb of knots connected by filaments. This recon-struction includes uncharted territories where di-

6

Page 7: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

rect observations are not permitted, such as thezone of avoidance.

5. Video productions

If an image can be worth a thousand words,a video is worth many thousand (Pomarede et al2015). Through a series of rotations, translations,zooms, the viewer of the visualization video followsstructures in three dimensions and can grasp therelationships between features on different scaleswhile retaining a sense of orientation. Variousgraphics objects (polygons, polylines, clouds ofpoints) materializing various fields (density, ve-locity, V-web elements, gravitational potential,galaxies,...) and other entities (basins of attrac-tion) can be represented simultaneously to studytheir relationships. In case of complex configu-rations of objects where confusion can arise, thefading-in and out of certain objects in the courseof the video can promote deeper understanding.

Several visualization videos have been pro-duced as part of the peer-reviewed publicationslisted in the sections below. These videos are re-leased on Vimeo and YouTube and can also beaccessed through dedicated web pages hosted byCEA/IRFU. The videos can either be viewed oninternet browsers or downloaded in several resolu-tions fitted to most devices: HD 1080p, HD 720p,SD, Mobile SD. In some cases, Stereo3D versionsare available and comments are available as CC“Closed Captions” that can be also downloadedseparately.

5.1. Cosmography of the Local Universe

This 17min35s video with comments is pre-sented at:

http://irfu.cea.fr/cosmography

It explores the cosmography of the local universeby exploring the 22 maps presented in Courtoiset al. 2013. The reconstructions of the velocityfield are based on the Cosmicflows-1 Catalog ofpeculiar velocities. This video obtained 410,000views so far and was downloaded 10,000 times.

5.2. Laniakea Supercluster

This 7 minutes-long video with comments isavailable at:

http://irfu.cea.fr/laniakea

It presents the identification and characterizationof the Laniakea Supercluster of galaxies. It hasbeen published as part of Tully et al. 2014. Thereconstructions of the velocity field are based onthe Cosmicflows-2 Catalog of peculiar velocities.This video was viewed 170,000 times so far in itsvarious forms.

5.3. Laniakea Nature Video

By using extracts of the two previous videos,the Nature Publishing Group has produced thefollowing 4min11sec video:

https://youtu.be/rENyyRwxpHo

This video has been released on YouTube’s NatureVideo Channel:

www.youtube.com/user/NatureVideoChannel

As of september 2016, this channel has 150,000subscribers and harbors 316 videos produced overthe course of the past eight years, for a grand to-tal of 33.6 million views. The Laniakea Video ob-tained a record of 500,000 views in a single dayupon its release and became by far the most pop-ular video with a total of more than 4 millionsviews, 30,000 likes and 2,100 comments.

5.4. The Arrowhead Mini-Supercluster

This 4min18s video is available at:

http://irfu.cea.fr/arrowhead

It explores the Arrowhead supercluster, a basinof attraction of limited size located at the con-tact region of the Laniakea supercluster and itstwo giant neighbors Perseus-Pisces and Coma (Po-marede et al 2015). This video has been viewed600 times so far.

6. Interactive web-based visualizations

We are exploring the use of Sketchfab, a web-based service that enables the upload and sharingof 3D models. It is an efficient tool to share vi-sualizations among collaborators. The Sketchfabinteractive viewer is based on WebGL, and can beembedded like a video player in any html docu-ment. This new technology thus offers a new wayto visualize data interactively in scientific articles.

An example of such an interactive visualizationreproducing the Figures 8 and 9 of Courtois et al.2013 can be viewed here:

7

Page 8: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

https://skfb.ly/R9pN

This visualization shows the rendering of threeisosurfaces of the reconstructed density field ofthe Local Universe. Navigation basics include ro-tations (left click and drag), zoom (scroll mousewheel when available, or Ctrl+Left click and dragup and down), pan (right click and drag). Moresophisticated hints and controls are available byclicking on the question mark icon provided by theplayer. Annotations are used to mark the most rel-evant structures. The navigation capabilities allowan extensive exploration of the cosmography.

Technically, objects such as polygons, polylines,textures, can be uploaded in Sketchfab in a vari-ety of different formats. We use the WavefrontOBJ file format. The SDvision software has a fa-cility to dump polygon objects such as isosurfacesto OBJ files. More sophisticated writer codes aredeveloped in IDL.

The Sketchfab platform offers an advanced in-terface to fine-tune the 3D properties of the up-loaded objects. It also provides Virtual Reality(VR) in association with VR gears and VR hel-mets. The VR interface in action in the contextof a cosmography visualization is shown in Fig-ure 8; here the interface provides a series of inter-active widgets to set the initial viewing position,the world scale, and the floor level.

7. Conclusion

Data visualization is a key ingredient in thefield of extragalactic cosmography. It promotesunderstanding of the three-dimensional structureof the cosmos and can be part of the discovery pro-cess. The development of our visualization soft-ware has accompanied the developments in obser-vations and theory. Together they contributed toadvances in the field, as evidenced in this paper.

Cosmography is an ancient science that re-sponds to a deep desire within societies to answerthe philosophical question of our place in the Uni-verse. It perpetuates the effort mankind has madein mapping the cosmos. From a societal viewpoint,the releases of the previous generations of cosmo-graphical maps have demonstrated a profound en-thusiasm in society as demonstrated by the audi-ences reached by diverse media. The video “La-niakea: Our home supercluster” produced by theNature Publishing Group is ranked as number one

in popularity on the “Nature Video” channel onYouTube with over 4 million views. There havebeen an abundance of articles in news outlets, pop-ular science media and blogs, social networks anddirect contacts with both the public and scientists.

8. References

Audit, E., et al., 2006, ASPCS, 359, 9

Benson, M. 2014, Cosmigraphics - Picturing spacethrough time (Abrams, New York) 71

Courtois, H., et al. 2012, ApJ, 744, 43

Courtois, H., et al. 2013, AJ, 146, 69

de Lapparent, V., Geller, M.J., & Huchra, J.P.1986, ApJ, 302L, 1

Doumler, T., et al., 2013, MNRAS, 430, 888

Hoffman, Y., 2009, in Lecture Notes in Physics,Berlin Springer Verlag, Vol. 665, Data Analy-sis in Cosmology, ed. V.J. Martınez, E. Saar, E.Martınez-Gonzalez, & M.-J. Pons-Borderıa, 565-583

Hoffman, Y., et al., 2012, MNRAS, 425, 2049

Hudson, M.J., 1993, MNRAS, 265, 43

Jarrett, T.H., et al., 2000, AJ, 119, 2498

Lavaux, G., & Hudson, M.J., 2011, MNRAS, 416,2840

Pomarede, D., & Pierre, M., 2011, IAUS, 277, 154

Pomarede, et al., 2008, ASPCS, 385, 327

Pomarede, D., Courtois, H., Tully, R.B., 2013,IAUS, 289, 323

Pomarede, D., et al. 2015, ApJ, 812, 17

Tully, R.B., & Fisher, J.R., Nearby Galaxies Atlas,Cambridge University Press, 1987

Tully, R.B., Nearby Galaxies Catalog, CambridgeUniversity Press, 1988

Tully, R.B., et al. 1992, ApJ, 388, 9

Tully, R.B., et al. 2008, ApJ, 676, 184

Tully, R.B., et al. 2013, AJ, 146, 86

Tully, R.B., et al. 2014, Nature, 513, 71

Tully, R.B., et al. 2016, AJ, 152, 50

Zaroubi, S., Hoffman, Y., Dekel, A., 1999, ApJ,520, 413

8

Page 9: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

Fig. 1.— The SDvision widget engaged in the visualization of multiple isosurfaces of the cosmic V-web. Thevelocity web cartographied here is extracted from the analysis of the eigenvalues λi=1,2,3 of the velocity sheartensor. Five surfaces of constant values of the λ3 eigenvalue, shown here in several nuances of red, tracethe knots of the web. A surface of constant value of the λ2 eigenvalue, shown in gray, traces the filaments.The view is augmented with annotations and polylines indicating the most salient cosmographic structures.These annotations and polylines are added to the transformable model like the other graphics objects. Athree-dimensional signpost made of three colored arrows (red, green, blue) is anchored on the origin of theSupergalactic coordinate system, that is the location of the Milky Way, with each arrow associated to itsthree cardinal axes (SGX, SGY, SGZ).

9

Page 10: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

Fig. 2.— The SDvision widget engaged in the visualization of a thin slice of the XSCz catalog of redshifts.Our galaxy, the Milky Way, resides at the origin. The positions of galaxies are given in the Supergalacticcoordinates system in units of redshift. The butterfly like distribution is typical of the redshift surveysviewed in the SGY-SGZ and SGX-SGY projections, where the obscuration by our galactic plane creates athree-dimensional wedge in the data sample where no measurements are available, the so-called “Zone ofAvoidance”.

10

Page 11: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

Fig. 3.— Comparison of the visualization of a velocity field using arrows (left) and streamlines (right).

Fig. 4.— Comparison of the visualization of a velocity field using streamlines seeded on uniform Cartesiangrids of 53 seeds (left) and 213 seeds (right).

11

Page 12: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

Fig. 5.— Visualization of peculiar velocities and Wiener Filter products. The galaxies of the Cosmicflows-2Catalog selected in a thin slice -500 km/s < SGZ < +1500 km/s are shown as white speckles. Their peculiarvelocities are shown as three-dimensional arrows colored blue for inward motions and red for outward motions.The white polylines are streamlines of the velocity field reconstructed by the Wiener Filter, seeded at eachgalaxy location. These streamlines are “long” streamlines, in the sense that their reconstruction is pushedfar enough so that each of them reach an attractor. The grey Gouraud-shaded polygon is an isosurface ofthe reconstructed density field. This high-density surface reveal the presence of the most massive players ofthe local universe: the Great Attractor, Perseus-Pisces, and Coma. A more modest high-density patch, theArrowhead, is also identified. A thin slice of the density field is also rendered as a rainbow-color contourimage, showing nuances in the available densities, deep blue corresponding for example to voids in thecosmography. This whole visualization conveys much of the essence of the Cosmicflows Project: peculiarradial velocities are used as input to a Wiener Filter algorithm that produces a 3D velocity field, useditself to derive the density field and to search for structures in the distribution of the streamlines, thevisualization allowing the exploration of all these ingredients all together. Here the whole process results inthe identification of the Arrowhead basin of attraction inside which the streamlines converge on an attractor.

12

Page 13: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

Fig. 6.— Visualization of cosmic flows. This map of the Supergalactic equatorial plane (SGZ=0) is obtainedusing “short” streamlines: short in the sense that their reconstruction is halted at a given number of stepswithout pursuing the convergence onto an attractor. Here, each single streamline benefits from it ownlocal 3D velocity field, obtained by separation of the local (divergent) component of the velocity field fromthe external (tidal) component. This combination of using a local field and short streamlines results inthe mapping of very local flows (at the cost of a single arbitrary parameter, the radius of the sphere ofdivergent/tidal separation). This process reveals a structure of adjacent basins of attraction, whether fullyreconstructed (Laniakea, Arrowhead) or partially reconstructed (Perseus-Pisces, Coma, Shapley) due to theirlocation on the edges of the data sample. This map is also powerful in illustrating the concept of evacuationof the flow from the voids. Also shown is a rainbow-color contour image of the density field that providesadditional context: the basins of attraction are found to be associated with higher-density environments.Frontiers between adjacent basins of attractions are running through underdense regions.

13

Page 14: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

Fig. 7.— Visualization of basins of attraction. The two surfaces enclosing the basins of attraction of Laniakeaand Arrowhead are shown as blue and yellow semi-transparent Gouraud-shaded polygons. Streamlines seededrandomly within the Laniakea envelope are shown as black polylines. To provide more context, the V-webis also represented with a grey surface for the filaments and a red surface for the knots. In terms ofcosmography, it is particularly interesting to examine the Arch filament that connects the Perseus-Piscesknot to the Laniakea knot, crossing the Laniakea border at a normal incidence.

14

Page 15: Daniel Pomar ede - arXiv.org e-Print archive · 2017. 2. 8. · Daniel Pomar ede Institut de Recherche sur les Lois Fondamentales de l’Univers CEA, Universit e Paris-Saclay, 91191

Fig. 8.— The Sketchfab VR interface. The scene on display features isosurface objects reproducing theFigures 8 and 9 of Courtois et al. 2013. The corresponding interactive web-based visualization can be viewedhere: https://skfb.ly/R9pN. This VR interface provides tools to modify scale, floor level, and initial viewinglocation. Other 3D properties such as lighting, camera field of view, materials, animations, annotations canalso be fine-tuned using dedicated interfaces.

15