-
Less is More: Discovering Concise Network Explanations
Authors:
Neehar Kondapaneni,
Markus Marks,
Oisin Mac Aodha,
Pietro Perona
Abstract:
We introduce Discovering Conceptual Network Explanations (DCNE), a new approach for generating human-comprehensible visual explanations to enhance the interpretability of deep neural image classifiers. Our method automatically finds visual explanations that are critical for discriminating between classes. This is achieved by simultaneously optimizing three criteria: the explanations should be few,…
▽ More
We introduce Discovering Conceptual Network Explanations (DCNE), a new approach for generating human-comprehensible visual explanations to enhance the interpretability of deep neural image classifiers. Our method automatically finds visual explanations that are critical for discriminating between classes. This is achieved by simultaneously optimizing three criteria: the explanations should be few, diverse, and human-interpretable. Our approach builds on the recently introduced Concept Relevance Propagation (CRP) explainability method. While CRP is effective at describing individual neuronal activations, it generates too many concepts, which impacts human comprehension. Instead, DCNE selects the few most important explanations. We introduce a new evaluation dataset centered on the challenging task of classifying birds, enabling us to compare the alignment of DCNE's explanations to those of human expert-defined ones. Compared to existing eXplainable Artificial Intelligence (XAI) methods, DCNE has a desirable trade-off between conciseness and completeness when summarizing network explanations. It produces 1/30 of CRP's explanations while only resulting in a slight reduction in explanation quality. DCNE represents a step forward in making neural network decisions accessible and interpretable to humans, providing a valuable tool for both researchers and practitioners in XAI and model alignment.
△ Less
Submitted 8 July, 2024; v1 submitted 24 May, 2024;
originally announced May 2024.
-
Assessing the distribution of cancer stem cells in tumorspheres
Authors:
Jerónimo Fotinós,
María Paula Marks,
Lucas Barberis,
Luciano Vellón
Abstract:
In previous theoretical research, we inferred that cancer stem cells (CSCs), the cells that presumably drive tumor growth and resistance to conventional cancer treatments, are not uniformly distributed in the bulk of a tumorsphere. To confirm this theoretical prediction, we cultivated tumorspheres enriched in CSCs, and performed immunofluorecent detection of the stemness marker SOX2 using a confoc…
▽ More
In previous theoretical research, we inferred that cancer stem cells (CSCs), the cells that presumably drive tumor growth and resistance to conventional cancer treatments, are not uniformly distributed in the bulk of a tumorsphere. To confirm this theoretical prediction, we cultivated tumorspheres enriched in CSCs, and performed immunofluorecent detection of the stemness marker SOX2 using a confocal microscope.
In this article, we present a method developed to process the images that reconstruct the amount and location of the CSCs in the tumorspheres. Its advantage is the use of a statistical criterion to classify the cells in stem and differentiated instead of setting an arbitrary threshold. From the analysis of the results of the methods using graph theory and computational modeling, we concluded that the distribution of Cancer Stem Cells in an experimental tumorsphere is non-homogeneous. This method is independent of the tumorsphere assay being useful for analyzing images in which several different kinds of cells are stained with different markers.
△ Less
Submitted 5 March, 2024;
originally announced March 2024.
-
On the Engulfment of Antifreeze Proteins by Ice
Authors:
Aniket U. Thosar,
Yusheng Cai,
Sean M. Marks,
Zachariah Vicars,
Jeongmoon Choi,
Akash Pallath,
Amish J. Patel
Abstract:
Antifreeze proteins (AFPs) are remarkable biomolecules that suppress ice formation at trace concentrations. To inhibit ice growth, AFPs must not only bind to ice crystals, but also resist engulfment by ice. The highest supercooling, $ΔT^{*}$, for which AFPs are able to resist engulfment is widely believed to scale as the inverse of the separation, $L$, between bound AFPs, whereas its dependence on…
▽ More
Antifreeze proteins (AFPs) are remarkable biomolecules that suppress ice formation at trace concentrations. To inhibit ice growth, AFPs must not only bind to ice crystals, but also resist engulfment by ice. The highest supercooling, $ΔT^{*}$, for which AFPs are able to resist engulfment is widely believed to scale as the inverse of the separation, $L$, between bound AFPs, whereas its dependence on the molecular characteristics of the AFP remains poorly understood. By using specialized molecular simulations and interfacial thermodynamics, here we show that in contrast with conventional wisdom, $ΔT^{*}$ scales as $L^{-2}$ and not as $L^{-1}$. We further show that $ΔT^{*}$ is proportional to AFP size and that diverse naturally occurring AFPs are optimal at resisting engulfment by ice. By facilitating the development of AFP structure-function relationships, we hope that our findings will pave the way for the rational design of novel AFPs.
△ Less
Submitted 2 January, 2024;
originally announced January 2024.
-
A Foundation Model for Cell Segmentation
Authors:
Uriah Israel,
Markus Marks,
Rohit Dilip,
Qilin Li,
Morgan Schwartz,
Elora Pradhan,
Edward Pao,
Shenyi Li,
Alexander Pearson-Goulart,
Pietro Perona,
Georgia Gkioxari,
Ross Barnowski,
Yisong Yue,
David Van Valen
Abstract:
Cells are the fundamental unit of biological organization, and identifying them in imaging data - cell segmentation - is a critical task for various cellular imaging experiments. While deep learning methods have led to substantial progress on this problem, models that have seen wide use are specialist models that work well for specific domains. Methods that have learned the general notion of "what…
▽ More
Cells are the fundamental unit of biological organization, and identifying them in imaging data - cell segmentation - is a critical task for various cellular imaging experiments. While deep learning methods have led to substantial progress on this problem, models that have seen wide use are specialist models that work well for specific domains. Methods that have learned the general notion of "what is a cell" and can identify them across different domains of cellular imaging data have proven elusive. In this work, we present CellSAM, a foundation model for cell segmentation that generalizes across diverse cellular imaging data. CellSAM builds on top of the Segment Anything Model (SAM) by develo** a prompt engineering approach to mask generation. We train an object detector, CellFinder, to automatically detect cells and prompt SAM to generate segmentations. We show that this approach allows a single model to achieve state-of-the-art performance for segmenting images of mammalian cells (in tissues and cell culture), yeast, and bacteria collected with various imaging modalities. To enable accessibility, we integrate CellSAM into DeepCell Label to further accelerate human-in-the-loop labeling strategies for cellular imaging data. A deployed version of CellSAM is available at https://label-dev.deepcell.org/.
△ Less
Submitted 18 November, 2023;
originally announced November 2023.
-
Text-image Alignment for Diffusion-based Perception
Authors:
Neehar Kondapaneni,
Markus Marks,
Manuel Knott,
Rogerio Guimaraes,
Pietro Perona
Abstract:
Diffusion models are generative models with impressive text-to-image synthesis capabilities and have spurred a new wave of creative methods for classical machine learning tasks. However, the best way to harness the perceptual knowledge of these generative models for visual tasks is still an open question. Specifically, it is unclear how to use the prompting interface when applying diffusion backbo…
▽ More
Diffusion models are generative models with impressive text-to-image synthesis capabilities and have spurred a new wave of creative methods for classical machine learning tasks. However, the best way to harness the perceptual knowledge of these generative models for visual tasks is still an open question. Specifically, it is unclear how to use the prompting interface when applying diffusion backbones to vision tasks. We find that automatically generated captions can improve text-image alignment and significantly enhance a model's cross-attention maps, leading to better perceptual performance. Our approach improves upon the current state-of-the-art (SOTA) in diffusion-based semantic segmentation on ADE20K and the current overall SOTA for depth estimation on NYUv2. Furthermore, our method generalizes to the cross-domain setting. We use model personalization and caption modifications to align our model to the target domain and find improvements over unaligned baselines. Our cross-domain object detection model, trained on Pascal VOC, achieves SOTA results on Watercolor2K. Our cross-domain segmentation method, trained on Cityscapes, achieves SOTA results on Dark Zurich-val and Nighttime Driving. Project page: https://www.vision.caltech.edu/tadp/. Code: https://github.com/damaggu/TADP.
△ Less
Submitted 1 April, 2024; v1 submitted 29 September, 2023;
originally announced October 2023.
-
MABe22: A Multi-Species Multi-Task Benchmark for Learned Representations of Behavior
Authors:
Jennifer J. Sun,
Markus Marks,
Andrew Ulmer,
Dipam Chakraborty,
Brian Geuther,
Edward Hayes,
Heng Jia,
Vivek Kumar,
Sebastian Oleszko,
Zachary Partridge,
Milan Peelman,
Alice Robie,
Catherine E. Schretter,
Keith Sheppard,
Chao Sun,
Param Uttarwar,
Julian M. Wagner,
Eric Werner,
Joseph Parker,
Pietro Perona,
Yisong Yue,
Kristin Branson,
Ann Kennedy
Abstract:
We introduce MABe22, a large-scale, multi-agent video and trajectory benchmark to assess the quality of learned behavior representations. This dataset is collected from a variety of biology experiments, and includes triplets of interacting mice (4.7 million frames video+pose tracking data, 10 million frames pose only), symbiotic beetle-ant interactions (10 million frames video data), and groups of…
▽ More
We introduce MABe22, a large-scale, multi-agent video and trajectory benchmark to assess the quality of learned behavior representations. This dataset is collected from a variety of biology experiments, and includes triplets of interacting mice (4.7 million frames video+pose tracking data, 10 million frames pose only), symbiotic beetle-ant interactions (10 million frames video data), and groups of interacting flies (4.4 million frames of pose tracking data). Accompanying these data, we introduce a panel of real-life downstream analysis tasks to assess the quality of learned representations by evaluating how well they preserve information about the experimental conditions (e.g. strain, time of day, optogenetic stimulation) and animal behavior. We test multiple state-of-the-art self-supervised video and trajectory representation learning methods to demonstrate the use of our benchmark, revealing that methods developed using human action datasets do not fully translate to animal datasets. We hope that our benchmark and dataset encourage a broader exploration of behavior representation learning methods across species and settings.
△ Less
Submitted 30 June, 2023; v1 submitted 21 July, 2022;
originally announced July 2022.
-
A possible solution to the Milky Way's binary-deficient retrograde stellar population. Evidence that $ω$ Centauri has formed in an extreme starburst
Authors:
M. Marks,
P. Kroupa,
J. Dabringhausen
Abstract:
Context. The fraction of field binaries on retrograde orbits about the Milky Way is significantly lower compared to its prograde counterpart. Chemical and dynamical evidence suggests that the retrograde stellar population originates from $ω$ Centauri, which is either the most massive globular cluster (GC) of the Milky Way or the putative core of a former dwarf galaxy. Aims. Star formation conditio…
▽ More
Context. The fraction of field binaries on retrograde orbits about the Milky Way is significantly lower compared to its prograde counterpart. Chemical and dynamical evidence suggests that the retrograde stellar population originates from $ω$ Centauri, which is either the most massive globular cluster (GC) of the Milky Way or the putative core of a former dwarf galaxy. Aims. Star formation conditions required to produce the retrograde binary population are constrained assuming that the retrograde stellar population originates from $ω$ Centauri's progenitor. Methods. We match the observed low binary fraction with dynamical population synthesis models, including a universal initial binary population and dynamical processing in star clusters, making use of the publicly available binary population synthesis tool BiPoS1. Results. It is found that either the GC progenitor of $ω$ Centauri must have formed with a stellar density of $\approx 10^8 \; M_{sun} \; pc^{-3}$ or the $ω$ Centauri dwarf galaxy's progenitor star cluster population must have formed in an extreme starburst with a star formation rate exceeding $1000 \; M_{sun} \; yr^{-1}$ and probably a top-heavy embedded-cluster mass function with suppressed low-mass cluster formation. The separation and mass ratio distribution for retrograde field binaries are predicted for comparison with future observations. Conclusions. A viable solution for the deficiency of binaries on retrograde orbits is presented, and star formation conditions for $ω$ Centauri as well as orbital parameter distributions for the Milky Way's retrograde binary population are predicted. The dwarf galaxy origin for $ω$ Centauri is tentatively preferred within the present context.
△ Less
Submitted 1 December, 2021;
originally announced December 2021.
-
BiPoS1 -- a computer programme for the dynamical processing of the initial binary star population
Authors:
J. Dabringhausen,
M. Marks,
P. Kroupa
Abstract:
The first version of the Binary Population Synthesizer (BiPoS1) is made publicly available. It allows to efficiently calculate binary distribution functions after the dynamical processing of a realistic population of binary stars during the first few Myr in the hosting embedded star cluster. Instead of time-consuming N-body simulations, BiPoS1 uses a stellar dynamical operator which determines the…
▽ More
The first version of the Binary Population Synthesizer (BiPoS1) is made publicly available. It allows to efficiently calculate binary distribution functions after the dynamical processing of a realistic population of binary stars during the first few Myr in the hosting embedded star cluster. Instead of time-consuming N-body simulations, BiPoS1 uses a stellar dynamical operator which determines the fraction of surviving binaries depending on the binding energy of the binaries. The stellar dynamical operator depends on the initial star cluster density as well as the time until the residual gas of the star cluster is expelled. BiPoS1 has also a galactic-field mode, in order to synthesize the stellar population of a whole galaxy. At the time of gas expulsion, the dynamical processing of the binary population is assumed to efficiently end due to the subsequent expansion of the star cluster. While BiPoS1 $has been used previously unpublished, here we demonstrate its use in the modelling of the binary populations in the Orion Nebula Cluster, in OB associations and as an input for simulations of globular clusters.
△ Less
Submitted 10 November, 2021;
originally announced November 2021.
-
Robust Disentanglement of a Few Factors at a Time
Authors:
Benjamin Estermann,
Markus Marks,
Mehmet Fatih Yanik
Abstract:
Disentanglement is at the forefront of unsupervised learning, as disentangled representations of data improve generalization, interpretability, and performance in downstream tasks. Current unsupervised approaches remain inapplicable for real-world datasets since they are highly variable in their performance and fail to reach levels of disentanglement of (semi-)supervised approaches. We introduce p…
▽ More
Disentanglement is at the forefront of unsupervised learning, as disentangled representations of data improve generalization, interpretability, and performance in downstream tasks. Current unsupervised approaches remain inapplicable for real-world datasets since they are highly variable in their performance and fail to reach levels of disentanglement of (semi-)supervised approaches. We introduce population-based training (PBT) for improving consistency in training variational autoencoders (VAEs) and demonstrate the validity of this approach in a supervised setting (PBT-VAE). We then use Unsupervised Disentanglement Ranking (UDR) as an unsupervised heuristic to score models in our PBT-VAE training and show how models trained this way tend to consistently disentangle only a subset of the generative factors. Building on top of this observation we introduce the recursive rPU-VAE approach. We train the model until convergence, remove the learned factors from the dataset and reiterate. In doing so, we can label subsets of the dataset with the learned factors and consecutively use these labels to train one model that fully disentangles the whole dataset. With this approach, we show striking improvement in state-of-the-art unsupervised disentanglement performance and robustness across multiple datasets and metrics.
△ Less
Submitted 26 October, 2020;
originally announced October 2020.
-
Sparse Sampling of Water Density Fluctuations near Liquid-Vapor Coexistence
Authors:
Erte Xi,
Sean M. Marks,
Suruchi Fialoke,
Amish J. Patel
Abstract:
The free energetics of water density fluctuations in bulk water, at interfaces, and in hydrophobic confinement inform the hydration of hydrophobic solutes as well as their interactions and assembly. The characterization of such free energetics is typically performed using enhanced sampling techniques such as umbrella sampling. In umbrella sampling, order parameter distributions obtained from adjac…
▽ More
The free energetics of water density fluctuations in bulk water, at interfaces, and in hydrophobic confinement inform the hydration of hydrophobic solutes as well as their interactions and assembly. The characterization of such free energetics is typically performed using enhanced sampling techniques such as umbrella sampling. In umbrella sampling, order parameter distributions obtained from adjacent biased simulations must overlap in order to estimate free energy differences between biased ensembles. Many biased simulations are typically required to ensure such overlap, which exacts a steep computational cost. We recently introduced a sparse sampling method, which circumvents the overlap requirement by using thermodynamic integration to estimate free energy differences between biased ensembles. Here we build upon and generalize sparse sampling for characterizing the free energetics of water density fluctuations in systems near liquid-vapor coexistence. We also introduce sensible heuristics for choosing the biasing potential parameters and strategies for adaptively refining them, which facilitate the estimation of such free energetics accurately and efficiently. We illustrate the method by characterizing the free energetics of cavitation in a large volume in bulk water. We also use sparse sampling to characterize the free energetics of capillary evaporation for water confined between two hydrophobic plates. In both cases, sparse sampling is nearly two orders of magnitude faster than umbrella sampling. Given its efficiency, the sparse sampling method is particularly well suited for characterizing free energy landscapes for systems wherein umbrella sampling is prohibitively expensive.
△ Less
Submitted 14 March, 2018;
originally announced March 2018.
-
Using binary statistics in Taurus-Auriga to distinguish between brown dwarf formation processes
Authors:
M. Marks,
E. L. Martín,
V. J. S. Béjar,
N. Lodieu,
P. Kroupa,
M. E. Manjavacas,
I. Thies,
R. Rebolo López,
S. Velasco
Abstract:
Whether BDs form as stars through gravitational collapse ("star-like") or BDs and some very low-mass stars constitute a separate population which form alongside stars comparable to the population of planets, e.g. through circumstellar disk ("peripheral") fragmentation, is one of the key questions of the star-formation problem. For young stars in Taurus-Auriga the binary fraction is large with litt…
▽ More
Whether BDs form as stars through gravitational collapse ("star-like") or BDs and some very low-mass stars constitute a separate population which form alongside stars comparable to the population of planets, e.g. through circumstellar disk ("peripheral") fragmentation, is one of the key questions of the star-formation problem. For young stars in Taurus-Auriga the binary fraction is large with little dependence on primary mass above ~0.2Msun, while for BDs it is <10%. We investigate a case in which BDs in Taurus formed dominantly through peripheral fragmentation. The decline of the binary frequency in the transition region between star-like and peripheral formation is modelled. A dynamical population synthesis model is employed in which stellar binary formation is universal. Peripheral objects form separately in circumstellar disks with a distinctive initial mass function (IMF), own orbital parameter distributions for binaries and a low binary fraction. A small amount of dynamical processing of the stellar component is accounted for as appropriate for the low-density Taurus-Auriga embedded clusters. The binary fraction declines strongly between the mass-limits for star-like and peripheral formation. The location of characteristic features and the steepness depend on these mass-limits. Such a trend might be unique to low density regions hosting dynamically unprocessed binary populations. The existence of a strong decline in the binary fraction -- primary mass diagram will become verifiable in future surveys on BD and VLMS binarity in the Taurus-Auriga star forming region. It is a test of the (non-)continuity of star formation along the mass-scale, the separateness of the stellar and BD populations and the dominant formation channel for BDs and BD binaries in regions of low stellar density hosting dynamically unprocessed populations.
△ Less
Submitted 5 June, 2017;
originally announced June 2017.
-
M-dwarf binaries as tracers of star and brown dwarf formation
Authors:
Michael Marks,
Markus Janson,
Pavel Kroupa,
Nathan Leigh,
Ingo Thies
Abstract:
The separation distribution for M-dwarf binaries in the ASTRALUX survey is narrower and peaking at smaller separations than the distribution for solar-type binaries. This is often interpreted to mean that M-dwarfs constitute a continuous transition from brown dwarfs (BDs) to stars. Here a prediction for the M-dwarf separation distribution is presented, using a dynamical population synthesis (DPS)…
▽ More
The separation distribution for M-dwarf binaries in the ASTRALUX survey is narrower and peaking at smaller separations than the distribution for solar-type binaries. This is often interpreted to mean that M-dwarfs constitute a continuous transition from brown dwarfs (BDs) to stars. Here a prediction for the M-dwarf separation distribution is presented, using a dynamical population synthesis (DPS) model in which "star-like" binaries with late-type primaries ($\lesssim1.5 M_{\rm sun}$) follow universal initial distribution functions and are dynamically processed in their birth embedded clusters. A separate "BD-like" population has both its own distribution functions for binaries and initial mass function (IMF), which overlaps in mass with the IMF for stars. Combining these two formation modes results in a peak on top of a wider separation distribution for late M-dwarfs consistent with the late ASTRALUX sample. The DPS separation distribution for early M-dwarfs shows no such peak and is in agreement with the M-dwarfs in Multiples (MinMS) data. We note that the latter survey is potentially in tension with the early ASTRALUX data. Concluding, the ASTRALUX and MinMS data are unable to unambiguously distinguish whether or not BDs are a continuous extension of the stellar IMF. Future observational efforts are needed to fully answer this interesting question. The DPS model predicts that binaries outside the sensitivity range of the ASTRALUX survey remain to be detected. For application to future data, we present a means to observationally measure the overlap of the putative BD-like branch and the stellar branch. We discuss the meaning of universal star formation and distribution functions.
△ Less
Submitted 16 June, 2015;
originally announced June 2015.
-
Characterizing the Brown Dwarf Formation Channels from the Initial Mass Function and Binary-star Dynamics
Authors:
Ingo Thies,
Jan Pflamm-Altenburg,
Pavel Kroupa,
Michael Marks
Abstract:
The stellar initial mass function (IMF) is a key property of stellar populations. There is growing evidence that the classical star-formation mechanism by the direct cloud fragmentation process has difficulties reproducing the observed abundance and binary properties of brown dwarfs and very-low-mass stars. In particular, recent analytical derivations of the stellar IMF exhibit a deficit of brown…
▽ More
The stellar initial mass function (IMF) is a key property of stellar populations. There is growing evidence that the classical star-formation mechanism by the direct cloud fragmentation process has difficulties reproducing the observed abundance and binary properties of brown dwarfs and very-low-mass stars. In particular, recent analytical derivations of the stellar IMF exhibit a deficit of brown dwarfs compared to observational data. Here we derive the residual mass function of brown dwarfs as an empirical measure of the brown dwarf deficiency in recent star-formation models with respect to observations and show that it is compatible with the substellar part of the Thies-Kroupa IMF and the mass function obtained by numerical simulations. We conclude that the existing models may be further improved by including a substellar correction term that accounts for additional formation channels like disk or filament fragmentation. The term "peripheral fragmentation" is introduced here for such additional formation channels. In addition, we present an updated analytical model of stellar and substellar binarity. The resulting binary fraction and the dynamically evolved companion mass-ratio distribution are in good agreement with observational data on stellar and very-low-mass binaries in the Galactic field, in clusters, and in dynamically unprocessed groups of stars if all stars form as binaries with stellar companions. Cautionary notes are given on the proper analysis of mass functions and the companion mass-ratio distribution and the interpretation of the results. The existence of accretion disks around young brown dwarfs does not imply that these form just like stars in direct fragmentation.
△ Less
Submitted 18 February, 2015; v1 submitted 7 January, 2015;
originally announced January 2015.
-
Monte Carlo modelling of globular star clusters - many primordial binaries, IMBH formation
Authors:
Mirek Giersz,
Nathan Leigh,
Michael Marks,
Arkadiusz Hypki,
Abbas Askar
Abstract:
We will discuss the evolution of star clusters with an large initial binary fraction, up to 95%. The initial binary population is chosen to follow the invariant orbital-parameter distributions suggested by Kroupa (1995). The Monte Carlo MOCCA simulations of star cluster evolution are compared to the observations of Milone et al. (2012) for photometric binaries. It is demonstrated that the observed…
▽ More
We will discuss the evolution of star clusters with an large initial binary fraction, up to 95%. The initial binary population is chosen to follow the invariant orbital-parameter distributions suggested by Kroupa (1995). The Monte Carlo MOCCA simulations of star cluster evolution are compared to the observations of Milone et al. (2012) for photometric binaries. It is demonstrated that the observed dependence on cluster mass of both the binary fraction and the ratio of the binary fractions inside and outside of the half mass radius are well recovered by the MOCCA simulations. This is due to a rapid decrease in the initial binary fraction due to the strong density-dependent destruction of wide binaries described by Marks, Kroupa & Oh (2011). We also discuss a new scenario for the formation of intermediate mass black holes in dense star clusters. In this scenario, intermediate mass black holes are formed as a result of dynamical interactions of hard binaries containing a stellar mass black hole, with other stars and binaries. We will discuss the necessary conditions to initiate the process of intermediate mass black hole formation and the dependence of its mass accretion rate on the global cluster properties.
△ Less
Submitted 27 November, 2014;
originally announced November 2014.
-
The state of globular clusters at birth II: primordial binaries
Authors:
Nathan W. C. Leigh,
Mirek Giersz,
Michael Marks,
Jeremy J. Webb,
Arkadiusz Hypki,
Craig O. Heinke,
Pavel Kroupa,
Alison Sills
Abstract:
(abridged) In this paper, we constrain the properties of primordial binary populations in Galactic globular clusters using the MOCCA Monte Carlo code for cluster evolution. Our results are compared to the observations of Milone et al. (2012) using the photometric binary populations as proxies for the true underlying distributions, in order to test the hypothesis that the data are consistent with a…
▽ More
(abridged) In this paper, we constrain the properties of primordial binary populations in Galactic globular clusters using the MOCCA Monte Carlo code for cluster evolution. Our results are compared to the observations of Milone et al. (2012) using the photometric binary populations as proxies for the true underlying distributions, in order to test the hypothesis that the data are consistent with an universal initial binary fraction near unity and the binary orbital parameter distributions of Kroupa (1995). With the exception of a few possible outliers, we find that the data are to first-order consistent with the universality hypothesis. Specifically, the present-day binary fractions inside the half-mass radius r$_{\rm h}$ can be reproduced assuming either high initial binary fractions near unity with a dominant soft binary component as in the Kroupa distribution combined with high initial densities (10$^4$-10$^6$ M$_{\odot}$ pc$^{-3}$), or low initial binary fractions ($\sim$ 5-10%) with a dominant hard binary component combined with moderate initial densities near their present-day values (10$^2$-10$^3$ M$_{\odot}$ pc$^{-3}$). This apparent degeneracy can be broken using the binary fractions outside r$_{\rm h}$- only high initial binary fractions with a significant soft component combined with high initial densities can contribute to reproducing the observed anti-correlation between the binary fractions outside r$_{\rm h}$ and the total cluster mass. We further illustrate using the simulated present-day binary orbital parameter distributions and the technique introduced in Leigh et al. (2012) that the relative fractions of hard and soft binaries can be used to further constrain the initial cluster density and mass-density relation. Our results favour an initial mass-density relation of the form r$_{\rm h} \propto$ M$_{\rm clus}^α$ with $α<$ 1/3.
△ Less
Submitted 8 October, 2014;
originally announced October 2014.
-
Revisiting the universality of (multiple) star formation in present-day star formation regions
Authors:
Michael Marks,
Nathan Leigh,
Mirek Giersz,
Susanne Pfalzner,
Jan Pflamm-Altenburg,
Seungkyung Oh
Abstract:
Populations of multiple stars inside clustered regions are known to change through dynamical interactions. The efficiency of binary disruption is thought to be determined by stellar density. King and collaborators recently investigated the multiplicity properties in young star forming regions and in the Galactic field. They concluded that stellar density-dependent modification of a universal initi…
▽ More
Populations of multiple stars inside clustered regions are known to change through dynamical interactions. The efficiency of binary disruption is thought to be determined by stellar density. King and collaborators recently investigated the multiplicity properties in young star forming regions and in the Galactic field. They concluded that stellar density-dependent modification of a universal initial binary population (the standard or null hypothesis model) cannot explain the observations. We re-visit their results, analyzing the data within the framework of different model assumptions, namely non-universality without dynamical modification and universality with dynamics. We illustrate that the standard model does account for all known populations if regions were significantly denser in the past. Some of the effects of using present-day cluster properties as proxies for their past values are emphasized and that the degeneracy between age and density of a star forming region can not be omitted when interpreting multiplicity data. A new analysis of the Corona Australis region is performed within the standard model. It is found that this region is likely as unevolved as Taurus and an initial density of $\approx190M_{\odot}\;pc^{-3}$ is required to produce the presently observed binary population, which is close to its present-day density.
△ Less
Submitted 15 May, 2014;
originally announced May 2014.
-
Inverse dynamical population synthesis: Constraining the initial conditions of young stellar clusters by studying their binary populations
Authors:
Michael Marks,
Pavel Kroupa
Abstract:
Binary populations in young star clusters show multiplicity fractions both lower and up to twice as high as those observed in the Galactic field. We follow the evolution of a population of binary stars in dense and loose star clusters starting with an invariant initial binary population and a formal multiplicity fraction of unity, and demonstrate that these models can explain the observed binary p…
▽ More
Binary populations in young star clusters show multiplicity fractions both lower and up to twice as high as those observed in the Galactic field. We follow the evolution of a population of binary stars in dense and loose star clusters starting with an invariant initial binary population and a formal multiplicity fraction of unity, and demonstrate that these models can explain the observed binary properties in Taurus, Rho-Ophiuchus, Chamaeleon, Orion, IC 348, Upper Scorpius A, Praesepe, and the Pleiades. The model needs to consider solely different birth densities for these regions. The evolved theoretical orbital-parameter distributions are highly probable parent distributions for the observed ones. We constrain the birth conditions (stellar mass, M_ecl, and half-mass radius, r_h) for the derived progenitors of the star clusters and the overall present-day binary fractions allowed by the present model. The results compare very well with properties of molecular cloud clumps on the verge of star formation. Combining these with previously and independently obtained constraints on the birth densities of globular clusters, we identify a weak stellar mass -- half-mass radius correlation for cluster-forming cloud clumps, r_h / pc ~ (M_ecl / M_sun)^(0.13+-0.04). The ability of the model to reproduce the binary properties in all the investigated young objects, covering present-day densities from 1-10 stars pc^-3 (Taurus) to 2x10^4 stars pc^-3 (Orion), suggests that environment-dependent dynamical evolution plays an important role in sha** the present-day properties of binary populations in star clusters, and that the initial binary properties may not vary dramatically between different environments.
△ Less
Submitted 7 May, 2012;
originally announced May 2012.
-
Evidence for top-heavy stellar initial mass functions with increasing density and decreasing metallicity
Authors:
Michael Marks,
Pavel Kroupa,
Jörg Dabringhausen,
Marcel S. Pawlowski
Abstract:
Residual-gas expulsion after cluster formation has recently been shown to leave an imprint in the low-mass present-day stellar mass function (PDMF) which allowed the estimation of birth conditions of some Galactic globular clusters (GCs) such as mass, radius and star formation efficiency. We show that in order to explain their characteristics (masses, radii, metallicity, PDMF) their stellar initia…
▽ More
Residual-gas expulsion after cluster formation has recently been shown to leave an imprint in the low-mass present-day stellar mass function (PDMF) which allowed the estimation of birth conditions of some Galactic globular clusters (GCs) such as mass, radius and star formation efficiency. We show that in order to explain their characteristics (masses, radii, metallicity, PDMF) their stellar initial mass function (IMF) must have been top-heavy. It is found that the IMF is required to become more top-heavy the lower the cluster metallicity and the larger the pre-GC cloud-core density are. The deduced trends are in qualitative agreement with theoretical expectation. The results are consistent with estimates of the shape of the high-mass end of the IMF in the Arches cluster, Westerlund 1, R136 and NGC 3603, as well as with the IMF independently constrained for ultra-compact dwarf galaxies (UCDs). The latter suggests that GCs and UCDs might have formed along the same channel or that UCDs formed via mergers of GCs. A fundamental plane is found which describes the variation of the IMF with density and metallicity of the pre-GC cloud-cores simultaneously. The implications for the evolution of galaxies and chemical enrichment over cosmological times are expected to be major.
△ Less
Submitted 21 February, 2012;
originally announced February 2012.
-
The stellar and sub-stellar IMF of simple and composite populations
Authors:
Pavel Kroupa,
Carsten Weidner,
Jan Pflamm-Altenburg,
Ingo Thies,
Joerg Dabringhausen,
Michael Marks,
Thomas Maschberger
Abstract:
The current knowledge on the stellar IMF is documented. It appears to become top-heavy when the star-formation rate density surpasses about 0.1Msun/(yr pc^3) on a pc scale and it may become increasingly bottom-heavy with increasing metallicity and in increasingly massive early-type galaxies. It declines quite steeply below about 0.07Msun with brown dwarfs (BDs) and very low mass stars having their…
▽ More
The current knowledge on the stellar IMF is documented. It appears to become top-heavy when the star-formation rate density surpasses about 0.1Msun/(yr pc^3) on a pc scale and it may become increasingly bottom-heavy with increasing metallicity and in increasingly massive early-type galaxies. It declines quite steeply below about 0.07Msun with brown dwarfs (BDs) and very low mass stars having their own IMF. The most massive star of mass mmax formed in an embedded cluster with stellar mass Mecl correlates strongly with Mecl being a result of gravitation-driven but resource-limited growth and fragmentation induced starvation. There is no convincing evidence whatsoever that massive stars do form in isolation. Various methods of discretising a stellar population are introduced: optimal sampling leads to a mass distribution that perfectly represents the exact form of the desired IMF and the mmax-to-Mecl relation, while random sampling results in statistical variations of the shape of the IMF. The observed mmax-to-Mecl correlation and the small spread of IMF power-law indices together suggest that optimally sampling the IMF may be the more realistic description of star formation than random sampling from a universal IMF with a constant upper mass limit. Composite populations on galaxy scales, which are formed from many pc scale star formation events, need to be described by the integrated galactic IMF. This IGIMF varies systematically from top-light to top-heavy in dependence of galaxy type and star formation rate, with dramatic implications for theories of galaxy formation and evolution.
△ Less
Submitted 4 November, 2012; v1 submitted 14 December, 2011;
originally announced December 2011.
-
The dynamical fingerprint of gas-expulsion: Insights into the assembly of the Milky Ways' old GC system
Authors:
Michael Marks,
Pavel Kroupa
Abstract:
Since the oldest globular clusters (GCs) are early residuals from the formation of the Milky Way (MW), GCs were exposed to the likely evolving potential of our Galaxy. The expulsion of the residual-gas from the GC's embedded progenitors is sensitive to the conditions in the pre-MW gas cloud. By means of N-body computations it is shown that gas throw-out from initially mass-segregated GCs affect th…
▽ More
Since the oldest globular clusters (GCs) are early residuals from the formation of the Milky Way (MW), GCs were exposed to the likely evolving potential of our Galaxy. The expulsion of the residual-gas from the GC's embedded progenitors is sensitive to the conditions in the pre-MW gas cloud. By means of N-body computations it is shown that gas throw-out from initially mass-segregated GCs affect the shape of the low-mass stellar mass function (MF) and that its imprint might still be visible in the present-day MF (PDMF). The strength of the tidal-field at birth influences the degree of gas-expulsion driven low-mass-star depletion and therefore the PDMF probes the MW potential at the time of GC formation. It is argued that among the old GC population in the MW, younger GCs show stronger low-mass-star loss than older GCs. This is shown to be consistent with a contracting and self-gravitating cloud in which fluctuations in the pre-MW potential grow with time. An initially relatively smooth tidal field evolved into a grainy potential within a dynamical time-scale of the collapsing cloud (based on Marks & Kroupa 2010, MNRAS, 406, 2000).
△ Less
Submitted 21 October, 2011;
originally announced October 2011.
-
Dynamical population synthesis: Constructing the stellar single and binary contents of galactic field populations
Authors:
Michael Marks,
Pavel Kroupa
Abstract:
[abridged] The galactic field's late-type stellar single and binary population is calculated on the supposition that all stars form as binaries in embedded star clusters. A recently developed tool (Marks, Kroupa & Oh) is used to evolve the binary star distributions in star clusters for a few Myr so that a particular mixture of single and binary stars is achieved. On cluster dissolution the populat…
▽ More
[abridged] The galactic field's late-type stellar single and binary population is calculated on the supposition that all stars form as binaries in embedded star clusters. A recently developed tool (Marks, Kroupa & Oh) is used to evolve the binary star distributions in star clusters for a few Myr so that a particular mixture of single and binary stars is achieved. On cluster dissolution the population enters the galactic field with these characteristics. The different contributions of single stars and binaries from individual star clusters which are selected from a power-law embedded star cluster mass function are then added up. This gives rise to integrated galactic field binary distribution functions (IGBDFs) resembling a galactic field's stellar content (Dynamical Population Synthesis). It is found that the binary proportion in the galactic field of a galaxy is larger the lower the minimum cluster mass, the lower the star formation rate, the steeper the embedded star cluster mass function and the larger the typical size of forming star clusters in the considered galaxy. In particular, period-, mass-ratio- and eccentricity IGBDFs for the Milky Way are modelled. The afore mentioned theoretical IGBDFs agree with independently observed distributions. Of all late-type binaries, 50% stem from M<300M_sun clusters, while 50% of all single stars were born in M>10^4M_sun clusters. Comparison of the G-dwarf and M-dwarf binary population indicates that the stars formed in mass-segregated clusters. In particular it is pointed out that although in the present model all M-dwarfs are born in binary systems, in the Milky Way's Galactic field the majority ends up being single stars. This work predicts that today's binary frequency in elliptical galaxies is lower than in spiral and in dwarf-galaxies. The period and mass-ratio distributions in these galaxies are explicitly predicted.
△ Less
Submitted 13 September, 2011;
originally announced September 2011.
-
Energy Shift and Wavefunction Overlap of Metal-Organic Interface-States
Authors:
M. Marks,
N. L. Zaitsev,
B. Schmidt,
C. H. Schwalb,
A. Schöll,
I. A. Nechaev,
P. M. Echenique,
E. V. Chulkov,
U. Höfer
Abstract:
The properties of Shockley-type interface states between $π$-conjugated organic molecular layers and metal surfaces are investigated by time-resolved two-photon photoemission experiments and density functional theory. For perylene- and naphthalene-tetracarboxylic acid dianhydride (PTCDA and NTCDA) adsorbed on Ag(111), a common mechanism of formation of the interface state from the partly occupied…
▽ More
The properties of Shockley-type interface states between $π$-conjugated organic molecular layers and metal surfaces are investigated by time-resolved two-photon photoemission experiments and density functional theory. For perylene- and naphthalene-tetracarboxylic acid dianhydride (PTCDA and NTCDA) adsorbed on Ag(111), a common mechanism of formation of the interface state from the partly occupied surface state of the bare Ag(111) is revealed. The energy position is found to be strongly dependent on the distance of the molecular carbon rings from the metal and their surface density. Bending of the carboxyl groups enhances the molecular overlap of the interface state.
△ Less
Submitted 19 July, 2011;
originally announced July 2011.
-
An analytical description of the evolution of binary orbital-parameter distributions in N-body computations of star clusters
Authors:
Michael Marks,
Pavel Kroupa,
Seungkyung Oh
Abstract:
A new method is presented to describe the evolution of the orbital-parameter distributions for an initially universal binary population in star clusters by means of the currently largest existing library of N-body models. It is demonstrated that a stellar-dynamical operator exists, which uniquely transforms an initial orbital parameter distribution function for binaries into a new distribution dep…
▽ More
A new method is presented to describe the evolution of the orbital-parameter distributions for an initially universal binary population in star clusters by means of the currently largest existing library of N-body models. It is demonstrated that a stellar-dynamical operator exists, which uniquely transforms an initial orbital parameter distribution function for binaries into a new distribution depending on the initial cluster mass and half-mass radius, after some time of dynamical evolution. For the initial distribution the distribution functions derived by Kroupa (1995a,b) are used, which are consistent with constraints for pre-main sequence and Class I binary populations. Binaries with a lower energy and a higher reduced-mass are dissolved preferentially. The stellar-dynamical operator can be used to efficiently calculate and predict binary properties in clusters and whole galaxies without the need for further N-body computations. For the present set of N-body models it is found that the binary populations change their properties on a crossing time-scale such that the stellar dynamical operator can be well parametrized as a function of the initial cluster density. Furthermore it is shown that the binary-fraction in clusters with similar initial velocity dispersions follows the same evolutionary tracks as a function of the passed number of relaxation-times. Present-day observed binary populations in star clusters put constraints on their initial stellar densities which are found to be in the range 10^2 - 2x10^5 M_sun pc^-3 for open clusters and a few x 10^3 - 10^8 M_sun pc^-3 for globular clusters, respectively.
△ Less
Submitted 24 June, 2011;
originally announced June 2011.
-
Initial conditions for globular clusters and assembly of the old globular cluster population of the Milky Way
Authors:
Michael Marks,
Pavel Kroupa
Abstract:
Comparing N-body calculations that include primordial residual-gas expulsion with the observed properties of 20 Galactic globular clusters (GCs) for which the stellar mass function (MF) has been measured, we constrain the time-scale over which the gas of their embedded counterparts must have been removed, the star formation efficiency the progenitor cloud must have had and the strength of the tida…
▽ More
Comparing N-body calculations that include primordial residual-gas expulsion with the observed properties of 20 Galactic globular clusters (GCs) for which the stellar mass function (MF) has been measured, we constrain the time-scale over which the gas of their embedded counterparts must have been removed, the star formation efficiency the progenitor cloud must have had and the strength of the tidal-field the clusters must have formed in. The three parameters determine the expansion and mass-loss during residual-gas expulsion. After applying corrections for stellar and dynamical evolution we find birth cluster masses, sizes and densities for the GC sample and the same quantities for the progenitor gas clouds. The pre-cluster cloud core masses were between 10^5-10^7 Msun and half-mass radii were typically below 1 pc and reach down to 0.2 pc. We show that the low-mass present day MF slope, initial half-mass radius and initial density of clusters correlates with cluster metallicity, unmasking metallicity as an important parameter driving cluster formation and the gas expulsion process. This work predicts that PD low-concentration clusters should have a higher binary fraction than PD high-concentration clusters.
Since the oldest GCs are early residuals from the formation of the Milky Way (MW) and the derived initial conditions probe the environment in which the clusters formed, we use the results as a new tool to study the formation of the inner GC system of the Galaxy. We achieve time-resolved insight into the evolution of the pre-MW gas cloud on short time-scales (a few hundred Myr) via cluster metallicities. The results are shown to be consistent with a contracting and self-gravitating cloud in which fluctuations in the pre-MW potential grow with time. An initially relatively smooth tidal-field evolved into a grainy potential within a dynamical time-scale of the collapsing cloud.
△ Less
Submitted 2 June, 2010; v1 submitted 13 April, 2010;
originally announced April 2010.
-
The influence of gas expulsion and initial mass-segregation on the stellar mass-function of globular star clusters
Authors:
Michael Marks,
Pavel Kroupa,
Holger Baumgardt
Abstract:
Recently de Marchi, Paresce & Pulone (2007) studied a sample of twenty globular clusters and found that all clusters with high concentrations have steep stellar mass-functions while clusters with low concentration have comparatively shallow mass-functions. No globular clusters were found with a flat mass-function and high concentration. This seems curious since more concentrated star clusters ar…
▽ More
Recently de Marchi, Paresce & Pulone (2007) studied a sample of twenty globular clusters and found that all clusters with high concentrations have steep stellar mass-functions while clusters with low concentration have comparatively shallow mass-functions. No globular clusters were found with a flat mass-function and high concentration. This seems curious since more concentrated star clusters are believed to be dynamically more evolved and should have lost more low-mass stars via evaporation, which would result in a shallower mass-function in the low-mass part.
We show that this effect can be explained by residual-gas expulsion from initially mass-segregated star clusters, and is enhanced further through unresolved binaries. If gas expulsion is the correct mechanism to produce the observed trend, then observation of these parameters would allow to constrain cluster starting conditions such as star formation efficiency and the time-scale of gas expulsion.
△ Less
Submitted 4 March, 2008;
originally announced March 2008.
