Filename
stringlengths 22
64
| Paragraph
stringlengths 8
5.57k
|
|---|---|
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | We follow Euclid Collaboration: Blanchard et al. (2020) and assume four non-overlapping redshift shells, centered at z = (0.9, 1.2, 1.5, 1.8), and with a depth of ∆z = (0.2, 0.2, 0.2, 0.3), respectively, over a total projected area of 15 000 square degrees. With these values, we derive volume factors η for each of the considered redshift bins, shown in the last column of Table 2. We note that the mean values of the four redshift shells used in Euclid Collaboration: Blanchard et al. (2020) do not match per- fectly the redshifts of the four comoving snapshots used in this work. However, this is only marginally relevant, since we do not carry out a proper comparison to the Fisher forecasts obtained in that analysis. In fact, this will be a more suited aspect of inves- tigation when considering the same observables, i.e., the Leg- endre multipoles of the anisotropic galaxy power spectrum, and especially when considering more realistic number densities, as pointed out in Sect. 2.1. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | A proper comparison between the results obtained using the full-box volume and the rescaled ones is presented in Sect. 5.5. In addition to the Euclid-like shells, we consider three additional volume rescalings, by dividing the range between Vbox and Vshell into four evenly sized intervals. This leads to a total of five dif- ferent sets of covariances, based on the volumes defined above. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | In this section we describe the theoretical framework of pertur- bation theory (PT), which is essential to understand the evolution of post-inflationary fluctuations in the matter density field δ into the current large-scale distribution of galaxies via gravitational instability. This description is expected to be accurate only down to the mildly nonlinear regime, where the amplitude of the den- sity contrast δ is small enough to be perturbatively expanded. In the strong nonlinear regime we expect this model to fail, as gravitational collapse leads to the formation of bound structures beyond the regime of validity of perturbative approaches. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | For the sake of completeness, we report the expanded ex- pressions for the second- and third-order symmetrised kernels, F2(q1, q2) and F3(q1, q2, q3) in Appendix B. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | The one-loop model in SPT, however, fails to accurately de- scribe the nonlinear damping of the acoustic oscillations due to bulk flow displacements (Eisenstein et al. 2007; Crocce & Scoc- cimarro 2008; Baldauf et al. 2015b). At first order, this effect can be reproduced in the theoretical model for Pmm(k) by a proper re- summation of all infrared (IR) modes q < k, i.e., of comoving separations larger than the one under consideration (see Crocce & Scoccimarro 2006 and Crocce & Scoccimarro 2008 for a de- scription of the BAO smearing in the context of renormalised perturbation theory). |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | The three-dimensional Dirac function is represented with the standard notation δ (3) D . Finally, the power spectrum PXX(k, z) of any component, matter or biased tracer, is defined as the auto- correlation of the corresponding density field δX, such that (cid:10)δX(k) δX(k′)(cid:11) ≡ (2π)3 PXX(k) δ (3) D where the presence of the Dirac function and the independence of the power spectrum from the orientation of the wave mode k reflect the underlying assumption of homogeneity and isotropy. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | A more standard procedure to include these corrections is based on the split of the linear power spectrum PL as the sum of a smooth Pnw and wiggly Pw component (Seo et al. 2008; Baldauf et al. 2015b; Blas et al. 2016), that is PL(k) = Pnw(k) + Pw(k) . |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | of the nonlinear matter power spectrum as Pmm(k) = P IR-NLO which contains one free parameter, cs, that must be treated as a nuisance parameter to be fitted against real or, in our case, simu- lated measurements. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | The general perturbative expansion of the galaxy density field δg is based on the sum of all the individual operators that are a func- tion of properties of the environment in which galaxies reside, such as the underlying matter density field and the large-scale tidal field. More precisely, this sum includes all those operators that are sourced by second derivatives of the gravitational poten- tial Φ and the velocity potential Φv (see Desjacques et al. 2018, for a detailed review on the subject). |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | where the square brackets of the last term mean that the evalu- ation of the one-loop correction is carried out using the leading order IR-resummed power spectrum in place of the linear one. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | where each operator is multiplied by a free bias parameter that determines its overall amplitude.12 The different terms in Eq. (18) can be summarised as follows. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | usually denoted as counterterm in the EFTofLSS framework. Here, the parameter cs can be interpreted as an effective speed of sound (Baumann et al. 2012; Carrasco et al. 2014; Baldauf et al. 2015a), reflecting the influence of short-wavelength per- turbations, but accounts as well for the complex physics behind galaxy formation (when considering biased tracers of the matter density field). |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | where G2(k1, k2) is the standard one-loop kernel for the nonlin- ear evolution of the velocity divergence field, and Eq. (29) has to be symmetrised with respect to its arguments (k1, k2, k3). |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | Inside Eq. (27) , αP,1 is a free nuisance parameter that ac- counts for deviations from a purely Poissonian shot-noise.13 In addition, it is also required as a way to reabsorb the otherwise non-zero low-k limit of one of the individual one-loop contribu- tions, as explained in Appendix B. Similarly, αP,2 parametrises the next-to-leading order correction, which scale as k2. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | where, once again, the square brackets of the second and third terms in Eq. (32) reflect how the evaluation of the one-loop and counterterm contributions is carried out sourcing the leading or- der IR-resummed matter power spectrum, P IR-LO mm , in place of the linear power spectrum, PL(k). |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | A significant fraction of the bias parameters that have been introduced in this section enters in the expression for Pgg(k) only at higher-order, as clearly pointed out by the presence of only the linear bias b1 in the expression for the leading-order galaxy power spectrum Eq. (31). This is significantly different from higher-order correlators of the galaxy density field, such as the galaxy bispectrum, for which both the local and non-local quadratic biases, b2 and bG2 , appear also in the expression for the leading-order term, and can therefore be constrained with much better accuracy (Oddo et al. 2021; Eggemeier et al. 2021). |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | Eggemeier et al. 2020; Pezzotta et al. 2021, for recent applica- tions). Nonetheless, their use in this analysis is well justified, since we focus on HOD samples for which the assignment of a galaxy into a host halo is only determined by the mass of the latter. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | In Sect. 5 we carry out tests to determine whether the pre- viously defined relations can be employed to analyse clustering measurements adopting Euclid requirements. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | Given the poor constraining power of the galaxy power spec- trum alone, it has become standard practice in real-data analyses to fix some of them to some physically motivated values or rela- tions. This is important not only to obtain a larger constraining power for the remaining parameters, but also to ensure that none of them experiences strong degeneracies such as the one exhib- ited by the (cid:0)bG2 , bΓ3 (cid:1) pair (see Appendix B). In this work, we test two different relations, which are briefly summarised in the next paragraphs. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | where the bias parameters with a superscript L stand for the corresponding Lagrangian quantities, i.e., at the time of forma- tion. The previous relations are commonly referred to as coevo- lution, or local Lagrangian relations when setting to zero the Lagrangian bias, and have been extensively used in most real- data analyses to fix one or both non-local parameters (see, e.g., Feldman et al. 2001; Gil-Marín et al. 2015; Sánchez et al. 2016; Grieb et al. 2017). However, recent results (Lazeyras & Schmidt 2018; Abidi & Baldauf 2018) have indicated that measurements from numerical simulations seem to suggest lower values for bG2 with respect to its local Lagrangian relation. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | In the previous sections, the relationship between the galaxy and the matter density field has been described through an Eulerian- based framework. However, this is not the only description of the galaxy power spectrum in the quasi-linear regime. Other ap- proaches are possible, often based to various degrees on results from numerical simulations. We consider here the so-called hy- brid Lagrangian models. They draw from Lagrangian perturba- tion theory for the bias expression connecting galaxy and matter overdensities, but rely on simulations to capture the development of nonlinearities when converting Lagrangian quantities to the observable Eulerian quantities. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | before, the expansion in Eq. (37) does not include the next-to- leading-order correction to the tidal field, captured by the opera- tor Γ3. This implies that the two bases are only equivalent under the assumption of co-evolution for the Eulerian parameter bΓ3 (see Eq. 34). |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | ters. The parameters controlling bias, shot-noise, and countert- erms can be effectively treated as free nuisance parameters, to be marginalised over after sampling the joint posterior distribu- tion. The set of parameters of interest is therefore restricted to the cosmological parameters, in our case θ ≡ {h, ωc}. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | The final model depends on four free parameters, the linear 2 , the tidal quadratic bias b L 1 , the local quadratic bias b L bias b L s2 , and the higher-derivative bias b L ∇2δ, to which we add the extra free parameter αP,1 to account (at first order) for non-Poissonian shot-noise, in the same way as it is done in the Eulerian PT model. We use a different notation for the quadratic tidal bias, since the definition of the tidal field operator is slightly different from the one presented in Sect. 3.1. Similarly for the Laplacian bias, which in this case only models higher-derivative correc- tions, but could also (partially) absorb unmodelled nonlocal ef- fects coming from higher orders, extra physics, such as baryonic effects, or the smoothing of the density field performed in La- grangian space. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | where Pi j(k) are the 15 cross-spectra of the five previously de- fined advected operators. To compute the Pi j terms, baccoemu has been trained with high-resolution Pi j measurements from 800 combinations of cosmologies and redshifts, obtained apply- ing the cosmology-rescaling technique to four main N-body sim- ulations (Angulo & White 2010; Zennaro et al. 2019; Contreras et al. 2020). |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | S −1(θ) (cid:0) ⟨θ⟩ − θ fid where ⟨θ⟩ and θ fid represent the mean of the posterior distribu- tion of the selected parameters and their fiducial values, respec- tively, and S (θ) is a square matrix containing the auto- and cross- covariance among all the entries of the vector θ.16 The meaning of Eq. (39) is straightforward: we are quantifying the deviation of the posterior distribution from the fiducial values of the corre- sponding parameters, and expressing this information in terms of the intrinsic error of those parameters. In the case where θ con- sists of only one parameter, the FoB simply expresses how far the posterior is from the fiducial value in units of the standard devi- ation of the parameter, with the 68% and 95% percentiles cor- responding to values of FoB of 1 and 2, respectively. Note that when considering more than one parameter these values change, as they need to be computed by directly integrating a multivari- ate normal distribution with the corresponding number of dimen- sions. For n = 2, we have that the new thresholds for the 68% and 95% percentiles are 1.52 and 2.49, respectively. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | As a final remark, notice that, even if it is possible to find a relation between the Lagrangian and Eulerian bias parameters, the two sets do not correspond to the same physical quantities. This happens because in the Eulerian framework they properly represent the response of galaxy formation to large-scale pertur- bations, while in the Lagrangian one this physical meaning is lost due to the advection of the operators to Eulerian coordinates. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | In this section we describe the methodology used to determine the best combination between different models, scale cuts, and bias configurations. In addition we list the details of the fitting procedure and the priors of the selected parameter spaces. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | In the context of model selection, the most relevant aspects to take into consideration are the range of validity of a given model and the precision and accuracy of the constraints on the parame- ters of interest. The procedure that we adopt is based on the se- lection of the maximum wave mode kmax up to which the model is still capable of providing a good description of the data vec- tors, while still recovering the correct input parameters. This can be quantified by means of three different performance metrics (employed in, e.g., Osato et al. 2019; Eggemeier et al. 2020; Pez- zotta et al. 2021; Eggemeier et al. 2021), which are described in the next subsections. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | where S (θ) is once again the covariance matrix of the parame- ters θ, and det(S ) its determinant. The meaning of this quantity can be more clearly understood assuming a flat posterior distri- bution with null correlation between the entries of θ. In this case, det(S ) represents the volume of the hyper-rectangle over which the posterior distribution of θ is distributed. Similarly, for non- zero parameter correlations, det(S ) represents the hyper-volume contained in the hyper-surface defined by the covariance matrix S . Therefore, a high value of the FoM corresponds to a more statistically significant constraint of the model parameters. |
Processed_Euclid_preparation._TBD._Galaxy_power_spectrum_mod.txt | Table 3. List of model parameters, split into cosmological and nui- sance ones, with the latter further divided into the two bias models de- scribed in Sect. 3. The nuisance parameters consist of bias parameters, EFTofLSS counterterm, and shot-noise terms. For each parameter, the imposed prior is specified in the last column of the table. The letter U stands for a uniform distribution, with edges identified by the first and second element of the pair, respectively. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | In the era of precision physics measurements at the LHC, efficient and exhaustive estimations of theoretical uncertainties play an increasingly crucial role. In the context of Monte Carlo (MC) event generators, the estima- tion of such uncertainties traditionally requires independent MC runs for each variation, for a linear increase in total run time. In this work, we report on an automated evaluation of the dominant (renormalization-scale and non-singular) perturbative uncertainties in the PYTHIA 8 event generator, with only a modest computational overhead. Each generated event is accompanied by a vector of alternative weights (one for each uncertainty variation), with each set separately preserving the total cross section. Explicit scale-compensating terms can be included, reflecting known coefficients of higher-order splitting terms and reducing the effect of the variations. The formalism also allows for the enhancement of rare partonic splittings, such as g → b¯b and q → qγ, to obtain weighted samples enriched in these splittings while preserving the correct physical Sudakov factors. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | particle-level prediction can be recycled after applying the weight correction for that event. This is particularly important if the entire event has been folded with a (time-consuming) detector simulation. The uncertainty on the other components of the prediction, such as parton showering, multiparton interactions, and hadronization, is more challenging to estimate. This is because of the algorithms applied to sample probability distributions and the iterative nature of the algorithms. The state-of-the- art is to select a (small) number of event-generator parameters and make entirely new predictions based on them, as e.g. in the “Perugia” tune variations [2] and/or “eigentune” variations [3–6]. Since each of these new predictions makes different particle-level predictions, each generated event must be passed through a detector simulation as part of a realistic analysis. This fact greatly reduces the number of parameter variations than can be performed. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | In this paper, we present a method to estimate the effect of parameter variations in the parton shower for a given kinematic configuration. This is similar to what was done previously in VINCIA for final-state radiation (FSR) [7], but is here extended to initial state radiation (ISR) and adapted to PYTHIA’s parton-shower framework [8–10]. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | We also show how to use the same method to generate a weighted sample enhanced in the oc- currence of specific shower branchings, such as g → b¯b, with correctly calculated weights (including correct physical Sudakov form factors). This could be useful, e.g., for the B physics community. (We note that equivalent proposals for “biasing” or “boosting” specific shower splitting probabilities were also made in [11,12].) The two methods can be combined, so one can also get uncertainties on a biased sample, although this latter capability has not yet been implemented in the current PYTHIA code. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | The probability for a branching in the parton shower is encapsulated in the Sudakov form factor which, for realistic applications, must be evaluated numerically. A practical numerical method for this is the veto algorithm, known in computer science as the ”thinning algorithm” [13, 14], which involves re- jecting (or thinning out) trial branchings. We start with a brief review of this method as applied to parton showers in section 2.1. We then turn to the main focus of this paper: incorporating systematic variations of the branching probabilities. For a unitary (probability-conserving) shower, such varia- tions necessarily imply opposite variations in the non-branching probabilities through the rejections. The specific form these variations must have to preserve the unitarity of the shower are derived in section 2.2. A further interesting application of the same framework is presented in section 2.3, allow- ing to generate correctly weighted showers with biased kernels, as was already proposed for q → qγ splittings in [11]. With the general formalism now in hand, sections 2.4 and 2.5 describe the specific application of the framework to renormalisation-scale and non-singular term variations in the shower, respectively. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | At the technical level, in the original VINCIA implementation [7], the set of variations that could be performed were defined by the authors (hardcoded), with limited options for users to modify e.g. by which factor to vary the renormalisation scale up and down. The PYTHIA implementation has been made significantly more general, allowing users considerable flexibility to define any number of simultaneous or separate variations, as documented in detail in PYTHIA’s online HTML documenta- tion, with a set of default variations chosen by the authors. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | As in VINCIA, the modification to the rejection and acceptance probabilities are accumulated during the shower evolution and presented after the shower has finished as (a set of) alternative global event weights; one for each variation. The relative probability for each event to occur under different showering assumptions (represented by the variations) is given by the weight calculated for the given variation relative to the nominal (unvaried) event weight. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | the limit of infinitely many events. Depending on the magnitude of each variation and how “long” the shower evolutions are (bigger phase spaces imply more room for changes to accumulate), the variation weights will fluctuate around their mean values. This will reduce the statistical precision on the uncertainty variations relative to the nominal sample. To exemplify, take a sample of 100 identical hard 4-jet events, and say that one of them experienced a very unlikely branching somewhere deep in the shower, say at the 20th branching (i.e. with minimal impact on 4-jet distributions). These 100 events would all enter with the same weight in the nominal sample. But the event that happened to contain the unlikely 20th branching can acquire a much larger weight in one of the variations if the probability for that branching to occur is much larger for that variation. The 4-jet cross section computed from the variation weights would then be dominated by the single event with large weight, corresponding to a much worse statistical precision, in spite of the fact that the actual weight change occurred not at the 4-jet level but much deeper in the shower. This is a simple consequence of accumulating the variations through the shower history, which — depending on future uses of the algorithm — may make it desirable to introduce further options for controlling the amount of variation performed at each stage of the shower. For the time being, for practical applications, we advise to monitor the variations of the uncertainty weights in each histogram bin so that any issues due to very rare events with very large variation weights do not go unnoticed. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | with t = p2 ⊥evol [8], and P (z) a DGLAP splitting kernel [15–17]. We emphasize, however, that the for- malism presented here is valid for arbitrary P (t, z) and could be applied equally well to dipole/antenna- showers. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | with P (t, z) = 0 outside the boundaries of the physical phase space, and Pacc < 1 guaranteed by ˆP > P . If the trial is accepted, physical momenta are generated corresponding to the chosen values of t and z, and the pre-branching partons are replaced by the post-branching ones, including the effects of recoils etc. If the trial is rejected (with probability Prej = 1 − Pacc), the parton system remains in its original state. In either case, the scale of the (accepted or rejected) trial becomes the new value for t0, from which the evolution is restarted to find the next (lower) trial scale. The procedure ends when t < tmin. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | and similarly for the higher-n terms which yield hyper-triangle integrals that can always be written on product form, prefaced by a factor 1/n! which gives the fractional volume occupied by a single ordered slice t0 > t1 > t2 > . . . > tn > t of the full n-hypercube. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | where we inserted the definition of the trial Sudakov, ˆ∆, in the second line, cancelled it against the ˆP term from the failed-branching exponential, and finally used the definition of the physical Sudakov, eq. (1). The last line is the desired expression, which now gives the physical resummed branching probability, independently of the trial function. This expression is identical to eq. (4), proving the correctness of the veto algorithm and in particular that the final result is independent of the choice of trial function, as long as ˆP > P . |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | The main part of our paper consists of the proof, to all orders in perturbation theory, of a conjecture developed by one of us in Ref. [7] in the context of the VINCIA shower generator [22]. According to this proposal, the veto algorithm discussed above can be modified to simultaneously compute several alternative sets of weights for each event, answering roughly: what would the weight of this event have been, if we had used, for instance, an alternative value for the strong coupling or an alternative splitting function? The number of variations that can be included is in principle infinite (each requiring very little computing and memory resources), hence several alternative definitions of the same source 2, of uncertainty can be evaluated simultaneously (e.g., renormalisation-scale variations by factors 2, and 4 can all be included) and final plots can be made using only a subset of these. We here prove the validity of the algorithm to all orders in perturbation theory, and implement it in the PYTHIA 8 event generator [10]. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | where the difference between the alternative radiation kernel P (cid:48) and the original P can be, for in- stance, different αs scale choices, different non-singular terms in the splitting kernels, and/or different effective higher-order contributions to the splitting kernels. Note however that we assume that the t and z definitions remain the same. Translations between different t choices are discussed in [23] (and the resulting equations are used in VINCIA to provide an uncertainty variation corresponding to the difference between virtuality-ordered and p⊥-ordered showers) while exploring different z definitions (and more generally, different recoil strategies) would require a future generalisation of the algorithm presented here. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | where the apostrophes on both P (cid:48) and ∆(cid:48) emphasize that the modified radiation probability enters in both places. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | hence exactly the same structure emerges for the reweighted sample as for the underlying veto al- gorithm above, just with P replaced by P (cid:48). The proof that eq. (18) results from the sum over all possibilities is therefore identical to the proof of the original (unweighted) veto algorithm above. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Two remarks are in order. Firstly, we emphasize that the relative reject-ratio, eq. (17), contains the difference ˆP − P in the denominator. This means that, if the trial overestimate, ˆP , is “too perfect” (meaning it is very close to P ), the denominator can become close to singular, resulting in large and possibly numerically unstable weights. Algorithmically, what happens is that there are very few failed trials, hence the modifications to the Sudakov factor are not mapped out very well; each failed trial will have a very large job to do. Technically, we address this by applying a “headroom factor” to the trial functions when automated uncertainty-variations are requested, ensuring that there is always a non-negligible probability for trials to be rejected at the cost of computational speed. By default, we choose a headroom factor of 2. For the representative example of hadronic Z decays, this results in a slowdown of the code of only about 20%. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Given enough phase space for evolution, this factor can become arbitrarily different from unity, repre- senting that, e.g., a very active shower history is exponentially more likely to occur in a shower with a large value of αs than in one with a small value. In principle, this is both physically and mathemat- ically correct. In practice, however, it is not desirable that branchings at low evolution scales in the shower should significantly alter the modified event weights. Technically, we treat this by imposing a few limiting factors on the variations, as detailed below. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | A second important use case for shower algorithms is to evaluate the fragmentation contributions to processes like photon and B hadron production, via splittings like q → qγ and g → b¯b respectively. (π0 → γγ and similar hadron decay processes obviously contribute substantially to the former as well; our focus here is on the perturbative contributions only.) Since these processes are relatively rare (αem (cid:28) αs and Pg→b¯b (cid:28) Pg→gg), the generation of adequate event samples featuring these processes can suffer from substantial inefficiencies. A complementary case is the generation of high-multiplicity minimum-bias samples in pp collisions, for which events enriched in the number of perturbative MPI could help to improve the generation efficiency (though of course there is also a contribution from events with few MPI but very active hadronization steps). |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | A similar line of argument as above allows us to construct weighted samples enriched in these processes, while preserving the exact Sudakov factors. We note that this method is formally identical to the one presented for q → qγ branchings in Ref. [11]; we include its definition and all-orders proof here mostly for completeness, and to have it presented in the same notation as above. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | where the last asymptotic shows that the reweighting factor becomes independent of the bias in the limit that the bias factor is very large. Nonetheless, the difference is important since, as we shall see below, this is what allows us to recover the physical Sudakov factor. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | We note that if one is interested only in enhancing a single branching of the given type, all events featuring the branching will be accompanied by a single power of the constant inverse-bias factor, eq. (25), hence that weight can alternatively just be applied to the event sample as a whole, and will cancel in any normalized distributions. The important part is thus the application of eq. (26) to each rejected trial branching, in order to recover the physical Sudakov factor. Similarly to above, this is a procedure that will only work well when there is at least a minimal number of rejected trial branchings, ensured e.g., by choosing ˆP > 1.2P . |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | As required, the nested integrals translate between ˆPbiased and the physical branching probability, P , such that the produced Sudakov factors will depend only upon P , not Pbiased. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | The first major class of variations we include are variations of the shower renormalization scales. This can be done for both QED and QCD, with the latter normally dominating the overall uncertainty. It is worth noting, however, that for a coherent shower algorithm, a scale choice of p⊥ accompanied by the so-called CMW scale factor [24, 25] absorbs the leading second-order corrections to the splitting functions for soft-gluon emission. A brute-force scale variation would destroy this agreement. We therefore provide an option to allow an explicit O(α2 s) compensating term to accompany each scale variation, driving the effective scale choice back towards p⊥ at the NLO level, while leaving the higher- order components of the scale variation untouched. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | with β0 = (11NC − 2nF )/3, NC = 3, and nF the number of active flavours at the scale µ = p⊥. Note that, if there are any quark-mass thresholds in-between p⊥ and kp⊥, then αs(p⊥) and αs(kp⊥) will not be evaluated with the same nF . Matching conditions are applied in PYTHIA to make the running continuous across thresholds, so this effect should be small for reasonable values of k. Nonetheless one could in principle add an additional term αs/(2π) ln(mq/(kp⊥))/3 to compensate for the differ- ent β0 coefficients used in the region between the threshold and kp⊥; however since the variation is numerically larger without that term, and since the ambiguities associated with thresholds are anyway among the uncertainties one could wish to explore, for the time being we consider it more conservative to not include any such terms. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | We emphasize that the compensation term in the expressions above is only included for gluon emissions, not for g → q ¯q splittings. The latter are subjected to the full (uncompensated) variation, αs(kp⊥)/αs(p⊥). |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | This does not significantly restrict the range of variation for perturbative branchings (even when αs ∼ 0.5, a full 40% amount of variation is still allowed), but it does prevent branchings very near the cutoff from generating large changes to the event weights. Removing this bound would not significantly affect the perturbative physics uncertainties, but would cause much larger weight fluctuations (between events with and without some very soft branching near the end of the evolution), mandating much longer run times for the same statistical precision. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Figure 1: Illustration of the default renormalisation-scale variations for FSR, by a factor of 2 in each direction. The central (default, unweighted) shower calculation is shown in blue, with /// hashing indicating the range spanned by the variation weights. The dashed (red) and solid (yellow) lines represent the results of standalone runs with µR = 0.5p⊥ and µR = 2p⊥ respectively. Left: without the NLO scale-compensation term. Right: with the NLO scale-compensation term (the default setting). Distribution of 1-Thrust for e+e− → hadrons at the Z pole, excluding b-tagged events; ISR switched off; data from the L3 experiment [26]. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | include both types of variations (independent and correlated), and compare the results obtained at the end of the run. From a practical point of view, the FSR αs choice mainly influences the amount of broadening of the jets, while the ISR αs choice influences resummed aspects such as the combined re- coil given to a hard system (e.g., a Z, W , or H boson, or a t¯t, dijet, or γ + jet system) by ISR radiation and also how many extra jets are created from ISR. The latter of course also depends on whether and how corrections from higher-order matrix elements are being accounted for. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | An illustration and validation of the automated renormalisation-scale variations is given in fig. 1, for the case of FSR and the distribution of 1-Thrust in e+e− → hadrons events at the Z pole, compared (QED ISR is switched off and b-tagged events are to a measurement by the L3 experiment [26]. excluded in this comparison.) First, we perform three separate dedicated runs, using µR = 2p⊥ (solid yellow lines with square symbols), µR = p⊥ (the default choice, solid blue lines with dot symbols), and µR = 0.5p⊥ (dashed red lines with open + symbols). For the central run, we also included the automated weight variations presented here, for the same factor-2 µR variations. The range spanned by the reweighted central distribution is shown by the blue /// hashed areas. On the left-hand side of fig. 1, the NLO scale-compensation term is switched off, and we see that the results of the independent runs are faithfully reproduced by the reweighted central-run distributions. (The small difference in the first bin is due to the absolute limit of |∆αs| ≤ 0.2 which we impose in the reweighting framework.) On the right-hand side of fig. 1, the same distributions are shown, but now with the NLO scale-compensation term switched on. The difference between the standalone runs (where no compensation is applied) and the reweighted distributions illustrates the effect of the compensation term. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Figure 2: Illustration of the default renormalisation-scale variations for ISR, by a factor of 2 in each direction. The central (default, unweighted) shower calculation is shown in blue, with /// hashing indicating the range spanned by the variation weights. The dashed (red) and solid (yellow) lines represent the results of standalone runs with µR = 0.5p⊥ and µR = 2p⊥ respectively. Left: without the NLO scale-compensation term. Right: with the NLO scale-compensation term (the default setting). Distribution of the p⊥ spectrum of the lepton pair in pp → Z → e+e−/µ+µ− at the Z pole (66 < m(cid:96)(cid:96)/GeV < 116), for leptons in the phase-space window |η(cid:96)| < 2.4, p⊥(cid:96) > 20 GeV; data from the ATLAS experiment [27]. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | test observable. The peak region below p⊥Z = 40 GeV is shown in the top row of plots (on a linear scale) while the bottom row shows the tail of the spectrum (on a log-log scale). As in fig. 1, the hashed area in the plots in the left-hand (right-hand) column shows the uncertainty band with the NLO scale- compensation term switched off (on). The effect is here less than in the FSR case, cf. fig. 1, presumably due to the compensation term being proportional to αs(mdip) where mdip can be very large in the ISR case. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Note to experimentalists: rather than performing dedicated runs for µR variations, we recommend using the uncertainty weights instead, since the renormalisation-scale compensation term is only avail- able for the latter and allows slightly more aggressive (smaller) uncertainty estimates. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | All shower formalisms are based upon the universal nature of the singular infrared (soft and/or collinear) limits of QCD. In these limits, the exact form of the splitting functions are known (to a given order), regardless of whether we express them as DGLAP kernels, dipole/antenna functions, or by any other means. Away from these limits, however, in the physical phase space on which the kernels will be ap- plied as approximations, there are in principle infinitely many different radiation functions to choose from, sharing the same singular terms but having different nonsingular ones. This represents a fun- damental ambiguity for shower algorithms which cannot be evaded by, e.g., setting the non-singular terms to zero. Firstly, any such (arbitrary) choice would not address the underlying issue. Secondly, it would not be stable against reparametrisations of the radiation functions themselves. For example, zero in one dipole parametrisation does not correspond to zero in another, see e.g. [7, 28]. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Moreover, varying the splitting kernels by nonsingular (a.k.a. “finite”) terms produces uncertainty envelopes which are quite complementary to those produced by renormalisation-scale variations [7]. The reason is that renormalisation-scale variations are by construction proportional to the (default) shower radiation functions, while nonsingular terms vary the radiation functions themselves. In re- gions far from the singular limits, the pole terms are highly suppressed and the default shower radia- tion functions may not bear much resemblance to the matrix elements for the process at hand. In such regions, process-dependent nonsingular terms dominate, and corresponding nonsingular-term varia- tions in the shower radiation functions can therefore easily produce much larger (and more realistic) uncertainty estimates than renormalisation-scale changes. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | We therefore believe that an exhaustive exploration of parton-shower uncertainties should at least grant the capability to perform nonsingular variations of the shower kernels, while the final decision whether and how to use them can still be left up to the user. An observation of large nonsingular-term uncertainties in the context of a physics study would be a direct indication of a need to incorporate fur- ther corrections from matrix elements, e.g. via one of the many matching/merging strategies available in PYTHIA 8. This is because the matrix elements contain the correct (process-dependent) nonsin- gular terms for the process at hand, thus nullifying the nonsingular-term uncertainties at least in any phase-space regions populated by the matrix elements. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Figure 3: Illustration of the default nonsingular variations for FSR splitting kernels, corresponding to cNS = ±2 (shown in red with \\\ hashing), compared with the default renormalisation-scale variations by a factor of 2 with the NLO compensation term switched on (shown in blue with /// hashing). Left: matrix-element corrections OFF. Right: matrix-element corrections ON. Note that the range of the ratio plot is greater than in fig. 1 Distribution of 1-Thrust for e+e− → hadrons at the Z pole, excluding b-tagged events; ISR switched off; data from the L3 experiment [26]. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | b = 2pb · pg [29], with pb the 4-momentum of the massive quark and pg that of the emitted gluon. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | from which we also immediately confirm that the relative variation explicitly vanishes when Q2 → 0 or P (z) → ∞. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | To motivate a reasonable range of variations, we take the nonsingular terms that different physical matrix elements exhibit as a first indicator, and supplement that by considering the terms that are induced by PYTHIA’s matrix-element corrections (MECs) for Z boson decays [30]. In particular, the study in [28] found order-unity differences (in dimensionless units) between different physical processes and three different antenna-shower formalisms: Lund dipoles a la ARIADNE [31,32], GGG antennae a la VINCIA [7, 33, 34], and Sector antennae a la Kosower [28, 35]. Therefore, here we also take variations of order unity as the baseline for our recommendations. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | in red with \\\ hashing, roughly spans the range between PYTHIA with and without matrix-element corrections. In the right-hand pane, where PYTHIA’s internal MECs for Z → 3 jets [30] are switched on, the splitting-kernel uncertainty is essentially zero in the 3-jet region 1 − T ≤ 0.33, since the nonsingular terms are there provided by the matrix elements. There are in principle still nonsingular- term uncertainties starting from the 4-jet level, beyond 0.33. Note that the ratio panes in fig. 3 have a larger range than those of fig. 1 and that, for comparison, the renormalisation-scale uncertainty, with the scale-compensation term switched on, is still shown in blue with /// hashing. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | The case of nonsingular-term variations for the ISR splitting kernels is shown in fig. 4, again compared to the renormalisation-scale variations (with the NLO compensation term switched on), for the same p⊥Z distributions as were shown in fig. 2. For this specific case, PYTHIA’s matrix-element corrections [36] do not have as dramatic an effect on the central prediction as they did for FSR, as can be seen by comparing the central lines of the plots in the left-hand column of the figure (MECs OFF) to the ones on the right (MECs ON). The variation of nonsingular terms, however, is completely cancelled when MECs are switched on, as expected for a distribution dominated by a single emission. For completeness, we remark that the reweighting strategies presented here, and parton showers in general, are based on exact cancellation between real and virtual corrections. This is called detailed balance and is also referred to as unitarity in the parton-shower context. However, the KLN theo- rem [37, 38] allows for violations of this balance by non-singular terms. Hence a realistic assessment of the full uncertainties of parton-shower calculations should take into account that non-singular terms can contribute not only in the radiation functions, as above, but also at the level of breaking detailed balance. This would amount to an estimate of the possible size of NLO (and higher) K-factors. To accomplish this consistently, however, several further aspects would need to be addressed, including variations already at the Born level and ensuring that weight modifications at the n-th branching in the shower don’t change the total cross-section by more than factors proportional to αBorn+n . We deem these considerations to be beyond the scope of this work, but emphasize that they should be investigated. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | We have described the mathematical formalism and practical validation for a new way of calculating perturbative uncertainty estimates in the PYTHIA 8 Monte Carlo event generator, following the pro- posal made in [7]. Instead of performing independent Monte Carlo runs for each (set of) parameter variation(s), we effectively recycle the vetoed trials of the Sudakov veto algorithm to provide a numer- ical mapping of the probability-density changes resulting from different choices of renormalization scales and non-singular terms. The result is cast as a vector of weights for each event whose zero element corresponds to the nominal (user) settings, with the uncertainty variations telling how much the probability to obtain that event would have changed under different showering assumptions. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Each set of weights is separately unitary, in the sense that they integrate to the same total inclusive cross section for the process at hand. It is therefore important to note that non-unitary changes, such as “K-factor” variations, are not accessed by this framework, but would have to be estimated sepa- rately. The same is true for PDF variations and for variations of the non-perturbative fragmentation parameters. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Figure 4: Illustration of the default nonsingular variations for ISR splitting kernels, corresponding to cNS = ±2 (shown in red with \\\ hashing), compared with the default renormalisation-scale variations by a factor of 2 with the NLO compensation term switched on (shown in blue with /// hashing). Left: matrix-element corrections OFF. Right: matrix-element corrections ON. Distribution of the p⊥ spectrum of the lepton pair in pp → Z → e+e−/µ+µ− at the Z pole (66 < m(cid:96)(cid:96)/GeV < 116), for leptons in the phase-space window |η(cid:96)| < 2.4, p⊥(cid:96) > 20 GeV; data from the ATLAS experiment [27]. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | or matching the underlying shower formalism to higher-logarithmic accuracy. Several approaches for the former are now emerging (and are available in PYTHIA, the most advanced being UNLOPS [39]), while the latter remains a long-standing and highly non-trivial problem. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Our approach is based on a proposal first made in Ref. [7], which we have here proved to be valid to all orders in perturbation theory. We have also included several validations illustrating that the automated weight variations produced by our implementation do indeed reproduce the results of independent runs with the corresponding parameter changes. The formalism shares qualitative features with the proposal for ”boosting” splitting probabilities in [12] and with the proposal for biasing photon emissions made in Ref. [11], and indeed for those purposes our approach reduces to those of [11, 12]. Recently, two techniques for fast uncertainty variations for NLO calculations were presented [40] in the context of the SHERPA event generator [41], one based on interpolation grids and another based on analytically calculable weights. Our approach differs in several respects from both of these. Most importantly, our formalism applies to all orders rather than just to the first order of corrections, hence variations are performed all throughout the shower. Secondly, as we have shown, our strategy is formally exact (in the limit of infinitely many generated events), while the weights computed in [40] are only approximate in the shower context. An elaborate study of parton-shower uncertainties was also recently performed in the HERWIG context, using conventional methods (independent runs) [42]2. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | We end by remarking on possible pathologies that can arise, and how best to deal with them. If an event is very rare in the baseline sample but much more likely in a variation, the result will necessarily be a very large weight for that variation. Especially after cuts the statistical precision of the weighted samples can therefore be much lower than for the nominal ones. To address this, we recommend biasing the nominal sample to make the relevant rare occurrences more frequent. This has the additional benefit of improving the statistical precision also of the nominal weights in the tails of the distributions. Note however that the technology for combining the uncertainty variations with biases has not yet been implemented at the time of writing; we eagerly await feedback from the community on issues encountered in practical studies, on which to base the development of future capabilities and recommendations. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Details of how to switch on the new automated framework in PYTHIA and how to define the list of uncertainty variations to be performed in an actual run have been included in a new HTML documentation file in the online set of PYTHIA documentation files. These technical specifications may change as the code evolves. The current version of these descriptions can be consulted at: http://home.thep.lu.se/˜torbjorn/pythia82html/Welcome.html, under the in- dex heading “Automated Shower Variations”. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Note added in proof: during the completion of this work, we became aware of two complementary projects which allow to perform renormalisation-scale and PDF variations in a manner analogous to ours, implemented in the HERWIG and SHERPA generators respectively; see [44, 45] for details. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | Acknowledgements: The motivation to undertake this work arose from discussions at the Physics at TeV Colliders Workshop, Les Houches, June 2015. PS is supported in part by the Australian Research Council, contract FT130100744. Fermilab is operated by the Fermi Research Alliance, LLC under Contract No. De-AC02-07CH11359 with the United States Department of Energy. |
Processed_Automated_Parton-Shower_Variations_in_Pythia_8.txt | event and multiparton scattering measurements,” Eur. Phys. J. C76 no. 3, (2016) 155, arXiv:1512.00815 [hep-ex]. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | The filament has been studied in detail at several radio frequencies by Perley et al. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | the hot-spot and the ridge line of the bar shows a “pedestal” of lower surface brightness. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | An advantage of PPM for this type of simulation is its excellent resolution of shocks. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | Table 1. Simulation physical parameters (η, M ), time coverage (tinit ≤ frame intervals (∆t). Times are expressed in terms of the scale t0. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | pair of parameters for a relativistic jet with equivalent thrust or power, e.g. Rosen et al. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | our choices of the parameters (η, M ). These are non-relativistic evaluations, assuming a number density of 10−2 cm−2 and a temperature of 107 K in the ambient medium. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | We revolve the hydrodynamic data frames of ϕ and p to form 3D cylindrical structures. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | image (Figure 2) also suggests faint arcs connecting the ends of the bar to the hot-spot. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | the line of sight. The displayed subregions are the same size as those in Figure 8. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | Fig. 10. A selection of morphological matches to Pictor A, from ray-traced renderings 300 pixel sub-region about a jet with parameters (η, M ) = (10−4, 5). |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | Fig. 11. Morphological matches as in Figure 10 but with jet parameters (η, M ) = (10−4, 10). |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | Fig. 12. Morphological matches as in Figure 10 but with jet parameters (η, M ) = (10−4, 50). |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | Fig. 13. Morphological matches as in Figure 10 but with jet parameters (η, M ) = (10−2, 5). |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | Fig. 14. Morphological matches as in Figure 10 but with jet parameters (η, M ) = (10−2, 10). |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | Fig. 15. Morphological matches as in Figure 10 but with jet parameters (η, M ) = (10−2, 50). |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | The observed filament is extended more to the south of the inferred jet than the north. |
Processed_Complex_shock_structure_in_the_western_hot-spot_of.txt | ends. This is the result of the annular shocks arising in the region of the hot-spot. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.