The Universal Eccentricity Distribution for Dynamical Gravitational-Wave Merger Channels Preprint: APS/123-QED Mor Rozner, Teagan A. Clarke, Isobel M. Romero-Shaw, Johan Samsing Abstract We argue that all dynamical astrophysical black hole merger channels are expected to result in a common eccentricity distribution at gravitational wave (GW) frequencies relevant for LIGO/Virgo/KAGRA (LVK) in the high eccentricity limit. This follows from the large separation of scales between the GW regime required for creating eccentric mergers in LVK, and the underlying astrophysical formation environment. Our analytical solution shows exceptional agreement with numerical studies. This finding has important implications for both theoretical studies and ongoing searches for eccentric GW sources. I Introduction The first detection of gravitational waves (GWs) marked the beginning of a new era in astrophysics ( 6 ; 5 ) . A decade later, more than 200 BBH (binary black hole) mergers have been confirmed by the LIGO-Virgo-KAGRA (LVK) collaboration ( 1 ; 14 ; 15 ) with a variety of properties ( 7 ; 8 ; 9 ; 13 ; 11 ; 12 ; 98 ; 97 , e.g.,) , which not only have enabled statistical studies, but also revealed physical systems that were once thought to be peculiarities. Many GW observables of merging BHs, such as mass and spin distributions, are somewhat degenerate across the proposed formation channels, which makes them hard to disentangle. However, a few recent outliers might have provided some hints, including GW200208_222617 and GW200105 that show signs of eccentricity ( 71 ; 36 ; 73 ; 55 ; 61 ) , GW190521 and GW231123 that are characterised by high masses well above the theoretically expected pair instability mass cut-off ( 10 ; 4 ) , and GW241110 that shows a mass ratio \sim{1:2} with the second spinning around \sim 0.7 , indicating that the most massive BH itself could be a product of a previous BBH merger ( 3 ) . Such hierarchical mergers are especially useful for providing constraints on various cluster channels ( 44 ; 23 ; 30 ; 100 , e.g.,) . Eccentric mergers are a smoking gun for dynamical formation mechanisms in general, including single-single encounters ( 58 ; 83 ; 35 , e.g.,) , binary-single encounters ( 82 ; 66 ; 89 ; 79 ; 33 ; 34 , e.g.,) , binary-binary encounters ( 106 , e.g.,) , hierarchical interactions ( 104 ; 21 ; 19 ; 56 ; 41 ; 65 ; 94 ; 27 , e.g.,) , gas-assisted mergers ( 95 ; 75 ; 74 , e.g.,) , and close encounters in discs ( 48 ; 78 ; 68 ; 29 , e.g.,) . Their distribution and occurrence rates are expected to provide one of the best and well-defined ways of distinguishing between dynamical and isolated binary formation channels both on a population level ( 81 ; 82 ; 86 ; 88 ; 89 ; 77 ; 80 ; 67 ; 50 ; 106 ; 105 ; 71 ; 70 ; 92 , e.g.,) , as well as through individual sources ( 85 ; 96 , e.g.) . This promise has sparked a major effort into creating accurate eccentric waveform models ( 26 ; 32 ; 62 , e.g.,) , GW template banks ( 103 ; 60 ) , and optimised search techniques for such eccentric GW sources ( 2 ) . For these searches, which are still in a very initial stage, it is highly useful to have a handle on the expected eccentricity distribution. Currently, observational analyses typical use a uniform or a log-uniform eccentricity distribution, P(e)\propto 1/e ( 69 ; 71 ; 36 ; 73 ; 55 ; 61 , e.g.,) ; however, we show here that this is not a natural prior. In this Letter , we illustrate that in the high eccentricity regime at frequencies relevant for LVK ( f\geq 10\ \rm{Hz} ), all dynamical merger channels are expected to follow the same eccentricity distribution. At the post-Newtonian (PN) level, this distribution takes a closed analytical form that is unique and clearly different from the typically assumed distributions, including P(e)\propto 1/e . Our findings are based on the argument that the astrophysical scales associated with BBHs at their assembly are much bigger than the ones required for eccentric mergers, which essentially makes the formation of eccentric mergers a stochastic process rather than fine-tuned with memory from the underlying astrophysical scales. We denote this as the ‘pinhole regime’ , as further illustrated in Fig. 1 . We compare with a few examples from fully dynamical N-body simulations, including chaotic binary-single scatterings and hierarchical triples, and indeed, find that our analytical solution correctly predicts the eccentricity distribution. Our universal form should serve as the best benchmark for eccentricity searches and analyses in the high eccentricity regime that currently defines the resolvable limit by LVK. II Assembly of Black Hole Binaries The assembly of a high-eccentricity BBH merger generally results from two BHs that encounter each other at a pericentre distance, r , after which the evolution becomes strongly GW-driven until merger. During the inspiral, the BBH will gradually circularise, but can still, in some cases, maintain residual eccentricity when it enters the detectable GW frequency regime, typically \geq 10\ \rm{Hz} ( 40 ) for LVK. In this Letter , we operate with two different frequencies: the peak frequency, f_{p}=\pi^{-1}\sqrt{GM/r_{f}^{3}} , where M is the total mass of the BBH, r_{f} is the pericentre distance; and the orbital (Keplerian) frequency, f_{T}=\pi^{-1}\sqrt{GM/a_{f}^{3}} , where a_{f} is the BBH semi-major axis (SMA). The question that we are aiming to solve is: what is the orbital eccentricity distribution at a given frequency f , and how does it depend on the underlying formation mechanism? As the orbital evolution of highly eccentric BBHs is essentially dictated by the change of angular momentum and energy over the first pericentre passage, the relevant initial conditions to start with are the distribution of pericentre distances r . The calculation then boils down to deriving the distribution of r and mapping it to the distribution of eccentricity at a given frequency f . We will proceed with this calculation in the following sections. II.1 Initial Conditions: The Pinhole Regime Figure 1: The Pinhole Regime. Top: For a given encounter to result in an eccentric merger, the two interacting BHs must pass each other at a relatively small pericentre distance. This restriction can be mapped to a corresponding small area at the encounter surface, which we refer to as the ‘pinhole’. As the size of this area is generally much smaller than any of the underlying astrophysical scales, the distribution of BHs passing through this area is approximately random, which maps to our proposed universal eccentricity distribution, as described in Sec. III . Bottom: At high eccentricity, all of the channels are expected to converge towards the same functional form, which is a result of the separation of scales. At lower eccentricity, the astrophysical environment will start to show up, however, LVK is not expected to resolve this limit yet with the current sensitivity. The scales associated with most astrophysical channels are much larger than the typical pericentre distance required to form a BBH with measurable eccentricity in LVK ( 89 , e.g.,) . This separation of scales leads us to introduce the ‘pinhole regime’ , which describes our setup where the incoming BH has to go through a ‘hole’ with radius b , i.e. the pinhole, to create the eccentric BBH merger, as illustrated in Fig. 1 . Because of this separation of scales, the spatial distribution of the BHs entering the pinhole will, to leading order, be stochastic, i.e. randomly distributed across the pinhole surface. To see how this propagates to observables, we start by relating the impact parameter of the incoming BH, b , to the corresponding pericentre distance, r , using Kepler’s equation, \displaystyle b=r\sqrt{1+{2GM}/{(rv_{\infty}^{2})}}\propto\sqrt{r}, (1) where M is the mass of the binary, and v_{\infty} is the velocity of the incoming BH at infinity, assuming that the velocity at pericentre is \gg v_{\infty} , which holds in the close encounter regime. With this relation, we can relate the distribution of incoming BHs, P(b) , to the corresponding distribution of pericentre distances, P(r) , as \displaystyle P(r)=P(b)\left|{db}/{dr}\right|, (2) where we throughout the Letter refer to the distribution of variable \mathcal{A} by P(\mathcal{A}) . The distribution for P(b) follows from the random encounter regime that describes the pinhole regime, which implies that the number of BHs entering within a radius b is simply proportional to the enclosed area, \propto b^{2} . From this follows that P(b)\propto b , and |db/dr|\propto r^{-1/2} . Substituting these terms into Eq. ( 2 ) above, we find that \displaystyle P(r)\propto\rm{constant}. (3) This distribution is generally valid, even outside the pinhole, for all processes where the single incoming BH is isotropically distributed at infinity, such as in most spherical-symmetric dense environments. To illustrate that our framework is also expected to describe the outcomes from few-body systems, we now consider two cases: The first is the assembly of BBHs in dense stellar clusters, where high eccentricity is reached through strong and chaotic binary-binary and binary-single interactions ( 87 ; 89 ; 106 , e.g.,) . In this case, the general finding is that the constant perturbation of the BBH angular momentum leads to a distribution in orbital eccentricity, P(e)=2e , also referred to as a thermal distribution ( 42 ; 16 ) . Here the distribution of pericentre distance can be calculated as P(r)=P(e)|de/dr|\approx 2/a\propto\rm{const.} , where we have assumed that the pericentre distance leading to a high eccentricity merger is much smaller than the initial BBH SMA. The second case is BBHs merging in hierarchical triples undergoing Zeipel-Lidov-Kozai (ZLK) oscillations ( 102 ; 49 ; 45 ) , where the high eccentricity tail has been argued to mainly originate from non-secular configurations ( 20 , e.g.,) . In this regime, the non-adiabatic change in angular momentum leads again to a somewhat random encounter distribution, which has a similar probability, P(r)\approx 2/a\propto\rm{const.} , as also argued in ( 43 , e.g.,) . We therefore conclude that as long as there is a clear separation of scales, this should be true. Figure 2: Analytical Eccentricity Distributions. The probability density function of the log of eccentricity, P(\log e)\propto eP(e) , as calculated in Eqs. ( 7 ) and ( 8 ) for the peak and orbital frequencies correspondingly (in dashed and dashed-dotted lines), compared to the asymptotic solution (solid line) and log-flat distribution (dotted line). The distributions will break below a characteristic eccentricity, e_{c} , set by the astrophysical environment, and at high eccentricity when finite effects are included (see Sec. III ). III A Universal Eccentricity Distribution The distribution of orbital eccentricity, e_{f} , evaluated at GW frequency f , is related to the distribution in initial pericentre distance, as follows, \displaystyle P(e_{f})=P(r)\left|{dr}/{de_{f}}\right|\propto\left|{dr}/{de_{f}}\right| (4) where the last step follows from Eq. ( 3 ). The relation between r and e_{f} can be found through the relation, a(e)\propto e^{12/19}(1+121e^{2}/304)^{870/2299}/(1-e^{2}) , between the binary SMA, a , and the eccentricity, e , given by ( 59 ) . From this, one can now write the initial pericentre distance, r , as a function of the BBH eccentricity e_{f} , at GW frequency f , as ( 73 , e.g.,) r\approx C\frac{1+e_{f}}{e_{f}^{12/19}}\left(1+\frac{121}{304}e_{f}^{2}\right)^{-\frac{870}{2299}}M^{1/3}f_{p}^{-\frac{2}{3}}, (5) and, r\approx C\frac{1-e_{f}^{2}}{e_{f}^{12/19}}\left(1+\frac{121}{304}e_{f}^{2}\right)^{-\frac{870}{2299}}M^{1/3}f_{T}^{-\frac{2}{3}}, (6) where C=\frac{1}{2}(G/\pi^{2})^{1/3} , f_{p} refers to the GW peak frequency, and f_{T} refers to the orbital frequency, as defined in Sec. II . Both of these definitions are currently being used in the literature, among other similar definitions ( 104 ; 46 ; 84 ; 38 , e.g.,) . With these relations, we can now write out the eccentricity distribution evaluated at GW frequency f , using Eqs. ( 4 ), ( 5 ) and ( 6 ), from which follows \displaystyle P_{p}(e_{f})\propto\frac{192-112e_{f}+168e_{f}^{2}+47e_{f}^{3}}{e_{f}^{31/19}(304+121e_{f}^{2})^{3169/2299}}, (7) and \displaystyle P_{T}(e_{f})\propto\frac{96+292e_{f}^{2}+37e_{f}^{4}}{e_{f}^{31/19}(304+121e_{f}^{2})^{3169/2299}}, (8) where P_{p}(e_{f}) and P_{T}(e_{f}) denote the distribution evaluated at the GW peak frequency ( f_{p} ) and the orbital frequency ( f_{T} ), respectively. As seem, common to the two distributions is that they both asymptote to \displaystyle P(e_{f})\propto e_{f}^{-31/19}. (9) Or equivalently, P(\log e_{f})\propto e_{f}^{-12/19} , where we have used P(\log e)\propto eP(e) . In addition, we see that they do not depend on the actual value for the frequency f or the BH masses, as long as our assumptions hold and finite size effects are not included. Fig. 2 shows the analytical solutions from Eq. ( 7 ) and ( 8 ) together with the often assumed prior, P(e)\propto 1/e (log-flat). While P_{p} gradually decreases with e_{f} , one notes that P_{T} surprisingly has a significant upturn towards higher eccentricities. This behaviour is correct in the orbit-averaged point-mass adiabatic 2.5\rm{PN} framework we consider, but likely requires numerical-relativity (NR) simulations to correctly describe. This can, e.g., be seen by considering Eq. ( 6 ) (see also Appendix B ), where r/(2GM/c^{2})\lesssim 10 for e_{f}\gtrsim 0.2 at M=40\ M_{\odot},f_{T}=20\ \rm{Hz} , which illustrates that even modest eccentricities maps to highly relativistic encounters. This leads to concerns using f_{T} as a reference, as otherwise strongly proposed in 101 . Finally, our solutions are valid down to eccentricity, e_{c} , corresponding roughly to the pericentre value for which GW capture is possible, r_{c} , which is where the energy radiated at pericentre is comparable to the energy of the surrounding astrophysical environment ( 39 ; 87 , e.g.) . For example, for an environment characterised by velocity v_{env} , the pericentre capture distance is approximately ( 63 ; 89 , e.g.) r_{c}\approx R_{s}(c/v_{env})^{4/7}, (10) where R_{s}=2GM/c^{2} , and c is the speed of light. By substituting this expression for r into Eq. ( 5 ) or ( 6 ), we can now isolate for the corresponding eccentricity, e_{c}\approx 0.01\times\left(\frac{M}{50M_{\odot}}\right)^{-\frac{19}{18}}\left(\frac{v_{env}}{50\ kms^{-1}}\right)^{\frac{19}{21}}\left(\frac{f}{10\ Hz}\right)^{-\frac{19}{18}} (11) where we have assumed that e_{c}\ll 1 . This relation also holds for few-body systems, e.g., for binary-single interactions, where v_{env} is then replaced by the binary orbital velocity \propto\sqrt{M/a} , which then also implies that for such scatterings e_{c}\propto a^{-19/42} (see also Sec. III.1 ). Connecting our work to the low eccentricity regime, which also has major implications for the Einstein Telescope and Cosmic Explorer, will be reserved for future studies. Figure 3: Astrophysical Eccentricity Distributions. Stellar cluster: Outcomes from 2.5PN binary-single interactions evaluated at f_{p}=50~\text{Hz} between equal mass BHs with m=20\ M_{\odot} , and initial SMA a_{0}=0.1\ \rm{AU},\ 0.01\ \rm{AU} , are shown in solid red and blue , respectively. Open red in the Zoom-in figure shows the a_{0}=0.1\ \rm{AU} case at f_{p}=20~\text{Hz} . Triples: Outcomes from hierarchical triple evolution resulting in BBH mergers, as adopted and described in 20 , are shown in grey for f_{p}=10~\text{Hz} . Our derived general scaling P(\log e)\propto e^{-12/19} is shown with a black dashed line. A log-flat uniform prior is shown with a black dotted line. Exceptional agreement is seen between the different channels and our predicted scaling. Note that each distribution is scaled to match at high eccentricity to ease comparison. III.1 Astrophysical Examples Here we compare our analytical results with distributions from two distinct astrophysical formation channels of eccentric mergers, with results summarised in Fig. 3 . The first is through chaotic binary-single interactions, which can take place in various dense environments, such as globular clusters ( 87 ) . We perform scatterings between a BBH with SMA a_{0} and an incoming single BH, for two values of a_{0}=0.1\ AU,\ 0.01\ AU , assuming equal masses of m=20\ M_{\odot} . Each of these configurations are simulated 250.000 times, assuming an isotropic scattering environment. The second channel are binaries driven to merger by an interaction with a tertiary companion ( 102 ; 49 ; 45 ; 25 ; 57 ; 47 ; 18 ; 17 ; 90 ; 64 ; 37 ; 53 ; 51 ; 99 ) . We use for the comparison results from 20 (N-body results, Fig. 5). As seen in Fig. 3 , the set of distributions we consider shows excellent agreement with our solution from Eq. ( 7 ) and asymptotic scaling P(e)\propto e^{-31/19} . Challenges in limited access, resolution, and small number statistics, prevented us from including more datasets for this comparison. We did also confirm our prediction for P_{T} , but as argued in Sec. II.1 , a different framework is ultimately needed in this limit. Lastly, we note that the often considered single-single GW capture channel ( 58 ; 83 , e.g.) will, by definition, give rise to the distributions we derive, as such systems generally are born from isotropically encounters. We therefore leave out simulations of this process. IV Conclusions We argue in this Letter that the majority, if not all, dynamical evolution channels, from chaotic few-body systems to hierarchical triples, are expected to result in a common eccentricity distribution in the high eccentricity limit at frequencies relevant for LVK observations. We argue this is because the distance required for two BHs to merge with residual eccentricity is typically orders of magnitude smaller than the underlying astrophysical scales during the initial stages. Therefore, the eccentricity distribution stemming from BBHs perturbed into this high-eccentricity regime will retain no memory of their birth environment or the formation process, which characterises the final stages of a merger. We term this the ‘pinhole regime’. Our universal eccentricity distribution has excellent agreement with results from simulations describing different astrophysical channels, including binary-single interactions in dense globular clusters and field triples. This finding can be used as a prior for LVK for optimised eccentricity searches, parameter estimation, and for injection campaigns to establish eccentricity-related observational selection effects. Our findings highlight for the first time that the exact distribution in eccentricity and frequency commonly resolvable by LVK ( e_{10}\gtrsim 0.05 ; ( 52 ) ) should to leading-order follow a unified form regardless of the details of the underlying dynamical formation channel. The obvious consequence is that, by definition, these different astrophysical formation mechanisms are fundamentally indistinguishable in this range via their eccentricity alone . Information about the formation mechanism and environment is encapsulated in (a), the mass distribution and escape velocity of the environment (see Eq. 11 ), which set the critical eccentricity at which the pinhole regime dominates; and (b), the fraction of eccentric sources, i.e., the probability of entering the pinhole regime. However, neither of these environmental influences can be measured if the multiple formation channels potentially producing non-eccentric sources cannot be observationally separated. There are two options to rectify this problem. The first is simply to measure lower eccentricities, which may already be achievable for lower-mass binaries and binaries with significant higher-mode content ( 54 ; 72 ) , and will be routine with future detectors ( 52 ; 76 ) . At lower eccentricities, the distributions predicted for different dynamical or triple formation scenarios are distinct ( 28 , e.g.,) . The second is to study eccentricity only in correlation with other parameters. Different formation scenarios lead to different correlations between eccentricity and redshift ( 28 ) , spin ( 93 ) , and mass. Massive sources, above the pair instability gap ( \gtrsim 50\ M_{\odot} ), are thought to originate either from hierarchical ( 31 ; 22 , e.g.,) or gas-assisted channels ( 24 , e.g.,) , while the population below the transition mass is likely to be produced by a mixture of channels—or even be dominated by triples ( 91 ) . The difference in the formation channels could also be reflected in the fraction of eccentric mergers, i.e., we expect a step-change in the probability of entering the pinhole regime around the transition mass. We leave the application of such a population model for future work. Finally, we acknowledge that our calculation assumes orbit-averaged 2.5\ \rm{PN} level dynamics, and is based solely on the equations in 59 . This limit is known to be inaccurate in the extremely high eccentricity regime, in which the Keplerian orbital elements are also ill-defined. In addition, we note that defining e_{f} at the orbital frequency, f_{T} , as recently proposed by 101 , maps to highly relativistic encounters for even modest values of e_{f} , which therefore requires full NR simulations to explore further. The question of how and where to define eccentricity therefore still seems to be challenging and needs a revisit. Regardless of these issues, the resulting distribution in eccentricity should still follow a common form. Acknowledgements. MR gratefully acknowledges support from the Institute for Advanced Study, and the generous support of the Kovner Member Fund. I.M.R-S acknowledges support from the Science and Technology Facilities Council Ernest Rutherford Fellowship grant number UKRI2423. JS is supported by the Villum Fonden grant No. 29466, and by the ERC Starting Grant no. 101043143 – BlackHoleMergs. The Center of Gravity is a Center of Excellence funded by the Danish National Research Foundation under grant No. 184. Appendix A Frequency dependence and general assembly distributions In this appendix, we demonstrate another method to calculate the eccentricity distribution and its dependence on the frequency (orbital frequency in particular) using the continuity equation. Let p_{0}(e_{0}) be the eccentricity distribution at assembly, in a set-up in which we assume that all the binaries are formed at a given small fixed orbital frequency f_{T}=f_{0} . We let the eccentricity propagate following the continuity equation, and measure it again after time t , \displaystyle\frac{\partial P(e,t)}{\partial t}+\frac{\partial}{\partial e}\left[\frac{de}{dt}P(e,t)\right]=0. (12) The continuity equation leads us to the conclusion that segments in the probability distribution function P(e) conserve their weight at different frequencies, and hence for any given orbital frequency f , which could be uniquely identified with a given corresponding time, \displaystyle P(e)=P_{0}(e_{0})\left|\frac{de_{0}}{de}\right|. (13) To calculate the Jacobian |de_{0}/de| , we define a function, \displaystyle\Phi(e,e_{0})=\frac{f}{f_{0}}=\left(\frac{a_{0}(e_{0})}{a(e)}\right)^{3/2} (14) and use the relation between a and e as given in eq. 5.11 59 , and also discussed in Section III . This function formulates an implicit relation between e and e_{0} , and in the general case, an expression for de_{0}/de could be extracted numerically. This statement is general and holds for any initial eccentricity distribution, and allows us to calculate the new eccentricity distribution. Figure 4: The probability density function of eccentricity, as calculated in Appendix A , at different orbital frequencies (solid blue lines), compared to log-flat distribution (dotted-dashed line), approximated solution (Eq. 8 , dotted) and asymptotic (solid red line). Our results and comparisons between different probability density functions are illustrated in Fig. 4 . In solid blue lines, we illustrate the distribution function as calculated at different orbital frequencies, assembled with thermal distribution at an initial frequency 10^{-4}\ \rm{Hz} , derived based on the calculation in Appendix A , compared to the distribution described in Eq. 8 (dotted), log-flat (dotted-dashed) and asymptotic (solid red). As shown, our closed-form solution in Eq. 8 compares well with the results from the calculation in Appendix A , and as the frequency increases, the latter converges to the former. The asymptotic solution describes the slope well, while the thermal and log-flat distributions differ substantially from our solution as described by the two methods above. It should be noted that in Fig. 4 , we show a normalised probability density function as a function of the eccentricity e_{f} . Appendix B Different definitions for frequency Fig. 5 shows the initial pericenter distance scaled by R_{s}=2GM/c^{2} , r/R_{s} , as a function of eccentricity, e_{f} , plotted using Eqs. ( 5 ) and ( 6 ) for f_{p} and f_{T} , respectively. For this we have assumed M=40\ M_{\odot} and f=20\ \rm{Hz} . As seen, when using f_{T} as reference frequency, r/R_{s} drops steeply towards 0 for increasing eccentricity, which calls for a full NR treatment even for modest eccentricities. Finite-size effects, breakdown of the PN formalism, as well as ill-defined orbital elements, make this an extremely challenging mapping that needs further work. We note that 101 recently proposed f_{T} as the optimal choice for comparison, however, our work seems to suggest this problem has to be revisited. Figure 5: Initial pericenter distance scaled by R_{s}=2GM/c^{2} , r/R_{s} , as a function of eccentricity, e_{f} , derived using Eqs. ( 5 ) and ( 6 ). Results for f_{p} and f_{T} are shown in red and blue , respectively. For this figure we have assumed M=40M_{\odot} and f=20~\text{Hz} .