An Optical Gamma-Ray Burst Catalogue with Measured Redshift PART I: Data Release of 535 Gamma-Ray Bursts and Colour Evolution

We present the largest optical photometry compilation of Gamma-Ray Bursts (GRBs) with redshifts ($z$). We include 64813 observations of 535 events (including upper limits) from 28 February 1997 up to 18 August 2023. We also present a user-friendly web tool \textit{grbLC} which allows users the visualization of photometry, coordinates, redshift, host galaxy extinction, and spectral indices for each event in our database. Furthermore, we have added a Gamma Ray Coordinate Network (GCN) scraper that can be used to collect data by gathering magnitudes from the GCNs. The web tool also includes a package for uniformly investigating colour evolution. We compute the optical spectral indices for 138 GRBs for which we have at least 4 filters at the same epoch in our sample and craft a procedure to distinguish between GRBs with and without colour evolution. By providing a uniform format and repository for the optical catalogue, this web-based archive is the first step towards unifying several community efforts to gather the photometric information for all GRBs with known redshifts. This catalogue will enable population studies by providing light curves (LCs) with better coverage since we have gathered data from different ground-based locations. Consequently, these LCs can be used to train future LC reconstructions for an extended inference of the redshift. The data gathering also allows us to fill some of the orbital gaps from Swift in crucial points of the LCs, e.g., at the end of the plateau emission or where a jet break is identified.


INTRODUCTION
Gamma-ray bursts (GRBs) are among the most intense explosions in the universe, releasing from 10 46 to 10 54 ergs in -rays in time scales ranging from a fraction of seconds up to a few hours.GRBs have two main phases: the prompt emission, which is the initial explosion typically observed from hard X-rays to ≥ 100 MeV -rays, though sometimes in optical as well (Vestrand et al. 2005;Blake et al. 2005;Beskin et al. 2010); the afterglow (e.g., Costa et al. 1997;van Paradĳs et al. 1997;Piro et al. 1998;Gehrels et al. 2009;Wang et al. 2015), which is the long-lasting emission in X-ray, optical, and occasionally radio wavelengths following the prompt phase.GRBs are typically categorized as either Short GRBs (SGRBs) or Long GRBs (LGRBs), based on their prompt duration: SGRBs have  90 ≤ 2 s and LGRBs have  90 ≥ 2 s, where  90 represents the duration over which a burst emits between 5% and 95% of its prompt emission total measured counts (Mazets et al. 1981;Kouveliotou et al. 1993).
LGRBs originate from the deaths of massive stars (Woosley 1993;Paczyński 1998;Woosley & Bloom 2006;Cano et al. 2017), while SGRBs are produced by the combination of two compact objects, such as two neutron stars (NSs) (NSs, Eichler et al. 1989;Kochanek & Piran 1993;Metzger et al. 2011), or a NS and a black hole (BH) (NS-BH, Narayan et al. 1992).The central engines of these models are thought to be either hyper-accreting BHs or fast-spinning newborn NSs, called magnetars.
We provide details of the current challenges in GRB studies and the importance of creating a uniform optical catalogue for a large sample of LCs with good data coverage: (1) The transient nature of GRBs requires decisions about followup observations to be made swiftly, especially with unusual bursts such as high-redshift candidates, under-luminous GRBs, or SGRBs that could show an associated kilonova (KN, see e.g.Rossi et al. 2020;Rastinejad et al. 2022;Becerra et al. 2023a).
In addition, GRB classification is challenged by the emerging presence of many sub-classes besides the LGRBs and SGRBs, which account for the diversity of GRB features.Because of this complexity ★ E-mail: maria.dainotti@nao.ac.jp (NAOJ,SOKENDAI) † The second and third author have contributed equally ‡ The fifth and sixth author have contributed equally in GRB classification, a comprehensive collection of their complete optical LCs would aid in their categorization.
(2) Many GRB LCs show interesting features, such as flares, bumps, or plateaus, the understanding of which may give insight into the behaviours and progenitors of these GRBs.It is, therefore, crucial to have as complete coverage of the LCs as possible to better identify and characterize these features.In addition, complete optical LCs also allow for the estimation of the optical jet break time and comparison with jet breaks in other electromagnetic bands (-rays, X-rays, and radio), thus enhancing the investigation of chromatic or achromatic breaks in multiwavelengths (Becerra et al. 2023b).
Particularly the "plateau" phase, found in X-ray, optical, and radio LCs, following the prompt emission and preceding the steep decay phase (O'Brien et al. 2006;Sakamoto et al. 2007;Dainotti et al. 2013Dainotti et al. , 2020a,b;,b;Fraĳa et al. 2020;Levine et al. 2022;Fraĳa et al. 2022Fraĳa et al. , 2023;;Dainotti et al. 2022a;Becerra et al. 2023b) has been observed using data from the Neil Gehrels Swift Observatory (Swift, Gehrels et al. 2004), which has allowed for earlier observations of GRB afterglows.Many explanations have been proposed for the plateau, however, temporal gaps due to satellites orbital period, the lack of quick follow-up observations, and the process of stitching together multiple epochs make it difficult to draw a definitive conclusion.A catalogue that compiles data from different instruments and locations may allow some of these issues to be resolved.
Indeed, a large sample of GRB optical LCs will also allow testing of the standard fireball model (Piran 1999), which is currently the most accredited description of the GRB emission.Moreover, a collection of limiting magnitudes would help the community put new constraints on the GRB progenitor physics.
One key issue in addressing the challenges listed above is retrieving many optical data points, along with the colour evolution analysis of GRBs.Although several catalogues of optical observations exist, data from these and other databases, such as the GCN or the Swift catalogue, do not report data in a uniform format and do not always include host galaxy and spectral information.Having essential properties such as the coordinates, redshift, host extinction, the GRB classification reported in the literature, and the spectral index reported alongside the magnitudes in a uniform format would be helpful for the aforementioned future GRB studies.In this sense, the main advantage of this GRB catalogue is the uniformity in the data presentation.For each event the following are included: the time in seconds after the trigger, magnitude, magnitude error (1 ), filter, magnitude system (AB, Vega), observatory+telescope+instrument, a flag for specifying if the Galactic extinction correction has been performed, the source of data (GCN, paper, etc.), and a flag for the quality of the photometry reported.Since this catalogue will be released to the community as an open project, we encourage the astronomical community to contribute to the catalogue by submitting their data to our system following this convention.
Indeed, a critical advantage of this catalog is its ability to report data points from several telescopes that, when combined, can highlight features that otherwise would remain hidden.This allows important information to be revealed even if one telescope carries only a few data points.A remarkable example of this is the Subaru data for the GRB 010222A, which are critically positioned at the end time of the plateau, providing a more explicit determination of the plateau's end time, (see Dainotti et al. 2022b).A similar situation also applies to other LC segments.
The data from this catalogue will be available for download, reducing the time-consuming data collection and analysis of the LCs for the community.This work follows what has been initiated by several authors in the past, like the catalogues of Oates et al. (2009) and Roming et al. (2009Roming et al. ( , 2017) ) who focus on a specific instrument, i.e. the Swift Ultraviolet and Optical Telescope (UVOT).Here, we have expanded the analysis to a more extensive collection of different telescopes as done by Kann et al. (2006Kann et al. ( , 2010Kann et al. ( , 2011Kann et al. ( , 2024)).
Having clarified the urgent need for this catalogue, we present the compilation of 535 optical LCs included in a new web-based repository.In this paper, we also discuss the analysis of the LC colour behaviour and present two new pieces of software that aid in developing this catalogue and will be available to users -a new python package for processing GRB LCs and colour evolution and a Julia-based web scraper for the data gathering.
The paper is organized as follows: In Section 2, we describe the selection process for obtaining the LCs.In Section 3, we discuss our methodology for converting the data to a uniform format and determining the optical spectral index together with the colour behaviour for each GRB.In Section 4, we discuss our spectral analysis, our colour evolution analysis, and the division of GRBs into groups depending on the presence or lack of colour evolution based on our analysis.We also compare our results to those presented in the literature.In Section 5, we outline the functionality of the new web-based optical LC repository.This web-based repository is created with a newly developed, open-access python package designed for displaying the LCs, collecting information from NASA ADS sources (https://ui.adsabs.harvard.edu/)and GCNs (https://gcn.gsfc.nasa.gov/),converting the magnitudes into AB system, and correcting them for Galactic extinction, -correction effect, and investigating the color evolution following the method-ology of Section 4. We present our conclusions in Section 6.In Appendix A, we describe the instruments from which data were detected.In Appendix B, we discuss 5 peculiar GRBs for which a different spectral analysis is required.In Appendix C, we discuss the spectral fitting approaches of Kann et al. (2006) and Zaninoni (2013).In Appendix D, we discuss some examples of cases of disagreement between our analysis and the literature.

DATA SAMPLE
We compile a sample of 535 optical LCs from those published in the literature between February 28, 1997, and August 18, 2023, taking a part of the GRBs from Dainotti et al. (2022a).We have included all the GRBs with measured and confirmed redshifts.The redshift is either photometric or spectroscopic or from the host galaxy identification and it is taken either from (https: //www.mpe.mpg.de/~jcg/grbgen.html)or GRBOX (https: //sites.astro.caltech.edu/grbox/grbox.php)with a confirmed optical counterpart according to Greiner's table.Generally, we do not report optically dark GRBs, although 28 cases considered dark bursts in the literature have been included in this catalogue.This sample includes data privately shared by co-authors or by private communication.In particular, our sample includes 76 GRBs with data points from RATIR, 43 from Oates et al. (2009Oates et al. ( , 2012)), 25 from BOOTES (Alberto Castro-Tirado, 2023, private comm.),34 from Li et al. (2012Li et al. ( , 2015Li et al. ( , 2018Li et al. ( , 2022, private comm.), private comm.),19 from Zaninoni et al. (2013), 16 from Michael Hiram Siegel (2023, private comm.), 16 from Si et al. (2018) and references within, 15 from Alexei Pozanenko (2023, private comm.), 12 from the Kiso telescope (one of which elaborated by us), 11 from Cenko et al. (2009), six from VIRT (Brice Orange, 2023, private comm.), and one from Rossi et al. (2020).The rest of the information is taken from the literature such as GCNs, The Astronomer's Telegram (ATel), International Astronomical Union Circulars (IAUC), arXiv, and papers.The repository contains 64813 data points in total.Of these, 54300 are observed magnitudes, and 10513 are limiting magnitudes.We formatted the collected data as shown in Table 1 and stored them in space-separated .txtfiles.For each GRB, we collect the following information: (1) midtime of the observation (in seconds) after the satellite trigger; (2) magnitude; (3) magnitude error (1 ); (4) filter; (5) magnitude system; (6) observatory, telescope, and instrument; (7) a flag for Galactic extinction correction, where we assign 'y' if this correction has been applied in the literature, and 'n' when not; (8) source of the data (GCN, paper, arXiv, ATel, IAUC, or private communication); (9) a flag marking the points that are considered outliers (the discussion will be presented in Section 4.2.4), (10) the photometric/spectroscopic/host galaxy redshift, and (11) the GRB classification.While the information from (1) to ( 9) are stored only in the single GRB files, the redshift (10) and the GRB classification (11) are collected also in Table 1 of the Online Materials.All the magnitudes presented in this paper are reported in the AB system and corrected for the Galactic extinction.Magnitudes without errors and limiting magnitudes have been reported with zero magnitude error.
The Greiner and GRBOX web pages have been used to search for the references, the respective literature, and the GCNs reported in the Table 1 of the Online Materials.There, along with the redshift, we collect and present the following information in Table 4: the coordinates (right ascension = RA, declination = Dec), from the XRT and UVOT; the optical spectral index,  opt from the literature; the host extinction from the literature, including the magnitude extinction in the -band, and the best fit dust model for the host galaxy.In addition, we record the GRB classes.The convention adopted for the spectral index in this catalogue is given by the following flux expression:  (, ) ∼  −   − opt ,  being the temporal decay index (Sari et al. 1998;Jakobsson et al. 2004).

Sample homogenization
The data in magnitudes are collected from GCNs, literature (refereed papers, arXiv, ATel, IAUC), and private communications.As a first step to homogenize the sample, the magnitudes are converted to the AB system and corrected for the Galactic extinction, as follows: where, mag is the apparent magnitude value collected from the literature (in the observer frame),  ,gal =   *  B−V is the Galactic extinction,  B−V is the colour excess, and   is the extinction coefficient related to a given  band.To correct the magnitudes for the Galactic extinction, the dust maps of Schlafly & Finkbeiner (2011) or the Asiago Database on Photometric Systems (ADPS, Fiorucci, M. & Munari, U. 2003) are used.The shift Vega→AB is the shift from Vega to AB system; it is given for a generic filter in Blanton & Roweis (2007).

Spectral data analysis, 𝑘-correction, and host extinction
After converting the magnitudes to AB system and correcting for the Galactic extinction, the next step is to compute the optical spectral index ( opt ) and the host galaxy extinction in the -band (  ) together with the dust model.
To compute  opt and   , we need at least four magnitudes, all in different filters, observed at coincident epochs (times).Given two filters,  and  observed at different epochs, the observations are considered to be coincident if the difference in time of the two observations is less than 0.025 of the first epoch, as expressed in Equation 2. This percentage has been estimated by taking the peak of the distribution of the magnitude errors.The typical GRB magnitudes are reported with errors of ±0.04 − 0.05 mag, or equivalently ∼ 4 − 5%.This suggests that the simultaneity requirement can be relaxed up to half of the magnitude error (∼ 0.025) without significantly impacting the colour evolution.This criterion reads as follows: where   and   are the midtimes of the observations in seconds (after the satellite trigger) in the  and  bands, respectively.To estimate  opt , the spectral energy distribution (SED) of each GRB is fitted with a simple power law model according to the formulation   =  0  − opt shown in Sari et al. (1998); Jakobsson et al. (2004), where  0 is the normalization of the spectrum and  is the frequency.We remark that, in the current analysis, the host galaxy extinction is included following the approach reported in Zaninoni (2013).The formulation of Kann et al. (2006) can be proven to be equivalent to the one in Zaninoni (2013).This calculation is reported in Appendix C. The SED model   can be rewritten in terms of the wavelength (∝ 1/): where mag 0 is the magnitude value correspondent to  0 ,   is the host galaxy extinction in -band,   /  is taken from the extinction map reported in Pei (1992) for the three dust models: Milky Way (MW), Large Magellanic Cloud (LMC), and Small Magellanic Cloud (SMC), with   = 3.08, 3.16, 2.93 respectively, being   =    − a fixed number that corresponds to the ratio between the total extinction and the selective extinction in the -band.The subscript  refers to the wavelength  in the observer frame.The subscript 1 +  refers to the wavelength  in the host galaxy frame.
Using the Python package lmfit, the SEDs are fitted through the weighted Levemberg-Marquardt method, where the inverse square of the magnitude errors are taken as weights.We often encounter multiple spectral indices from the different epochs of a particular GRB, but not all provide a reliable fitting.Thus, we exclude the cases where |  opt | > | opt |.The prior values for   are constrained in the interval (0, 10), thus the SED fitting does not explore the parameter space region for which   < 0, which is indeed non physical.We do not remove any outliers from the SED fitting.However, we have flagged 601 outliers across 45 GRBs with the notation SED in the 9th flag column to indicate that the corresponding data points are outliers for the spectral fitting.Some GRBs are very well sampled in the optical wavelengths.Thus, the 0.025 simultaneity criterion of Equation 2 allows multiple  opt values at very close epochs.
Then, we choose the fitting with a probability value of  > 0.05, of this fit being computed not by chance, considering as the null hypothesis that the fit is not drawn by chance, which is accepted for  > 0.05.To estimate this probability, we integrate the inverse Gamma distribution from  to ∞, where  is the product between the reduced  2 value and the number of data points involved in the fitting, see Equation 4.
After estimating the  opt > 0 values, we also explore the possible  opt < 0 cases.The  opt < 0 condition is rarely expected if compared with the rest of the cases: indeed, we find such a condition only for 298 spectral indices across 68 GRBs.One remarkable example is GRB 090510A for which we find  opt = −0.853± 0.535, compatibly with the negative spectral index observed in -rays (Ackermann et al. 2010).The selected  opt values are reported in the Table 2 of the Online Materials.The -correction and the host correction are applied to the AB magnitude corrected for Galactic extinction as below: where according to the formulation in Oke & Sandage (1968); Peterson (1997); Greiner et al. (2015); Li et al. (2018).This treatment considers that the spectrum could show signs of evolution and the  opt may depend on the epoch.To highlight which value of  opt must be taken at each time, we have considered the following procedure.We first fit the spectral indices in our   / = 0.025 condition with the three extinction models (SMC, LMC, and MW).As dust model and   value, we consider the ones given by the best fit identified through the maximum probability .Then, concerning the spectral indices, we check which values among all the obtained  opt agree with each other within 2  with the iterative procedure discussed below.We first order the   opt values of each -th interval of a given GRB in the epoch  and then we compare each spectral index with all the previous ones with a 2  agreement criterion.If there is compatibility, the intervals are merged for that epoch and the procedure continues.If not, then a new epoch  ′ is defined, and the procedure continues with that new epoch.All the successive epochs for which the  opt value agrees within 2  are joined in groups to form one or multiple larger epochs.For each enlarged epoch, the  opt are then averaged to obtain only one value of the spectral index to be applied in the -correction in the enlarged epoch using the prescriptions of the Central Limit Theorem (CLT), without refitting the SED.Instead, we have used several constraints from the   / condition, and we have used several measures of the spectral indices at several different epochs, and we have constructed a smaller error bar based on the CLT.
To double-check that our procedure is reliable, we have considered several GRB cases randomly selected in which the literature agrees with our results in terms of the absence or presence of the colour evolution.We have checked with these GRBs that while performing compatibility of the spectral index in 1  leads to some discrepancy among our results and the literature, with the 2  we almost have full compatibility.The GRB cases on which we have tested our method are 071010A, 071025A, and 090313A for no colour evolution, and 060218A, 071031A, 100418A, and 130702A for colour evolution.For 5 GRBs (the 3% of the total 143), the 2  compatibility approach described above does not allow a reliable -correction process and these cases have been discussed as separate cases in the Appendix B and have not been included in our analysis.These GRBs are: 021004A, 110205A, 161219B, 171010A, and 171205A.For this reason, the number of GRBs with a reliable optical SED analysis reduces from 143 to 138.The error on the -correction is given by Δ () = | − 2.5 log 10 (1 + ) ×   opt ( ) |.The -correction is applied to all the GRBs that have the  opt ±  opt for which we have at least 4 data points available in different filters.For some GRBs, we only have limiting magnitudes in our catalogue, since the authors' data in the published literature are private, and thus for those, we are not able to compute  and perform the -correction.We calculate the error on the final magnitude value, mag corr , by using the error on the -correction and propagating them with their uncertainties from Equation 1 and the error given by the host galaxy extinction, namely, Δ  =     ×    ,    being the 1  uncertainty on   .

The optical and X-ray spectral indices comparison
When performing the comparison between optical and the XRT spectral indices (  ), we apply the same procedure described above to obtain the epochs in which the   values are computed.Indeed, the larger epoch with the 2  condition will allow the XRT spectral index epoch to be computed.The comparison of the   values extracted from the Swift-XRT Spectrum repository with our  opt is reported in Figure 1: there are 56 pairs compatible in 1  with the   =  opt relation, 48 pairs of ( opt ,   ) that are compatible in 1  with the   =  opt − 0.5 relation, 34 that are compatible with both in 1 , 79 are the cases in which either of the two, but not both, are incompatible with this scenario.Here, we discuss the 43 cases where both the relations are not compatible with the optical and X-ray spectral indices.For these 43 cases, we estimate the difference Δ   , = |   −   |: the values of Δ   , are normally distributed with a mean value  Δ  , = 1.26 and a standard deviation   Δ  , = 0.78.The z-scores, instead, computed as , follow a uniform distribution with mean    = 1.60 and standard deviation    = 0.93.Thus, overall they differ in less than 2 .The choice of the 2  ensures a good compromise on the final selected epoch: it avoids spurious and artificial evolution in the spectral index due to the tiny epoch.On the other hand, this epoch will still be short enough to appreciate if there is a small undergoing evolution.

Description of the rescaling factor
To study how the colour of GRBs changes over time in our dataset, we consider the measurements taken with different filters rather than considering the results from the spectral index evolution.Such a choice is justified by the fact that the colour evolution in the optical filters could be observed even with only two filters with at least 3 data points at a coincident epoch, while our SED analysis requires at least four different bands at coincident epochs, given that we need the slope, the intercept, and the   .Nevertheless, in the discussion of the colour evolution results we have provided a comparison between the SED evolution and the colour evolution results in our analysis.If a GRB does not display colour evolution, magnitudes in different filters can be rescaled to the most numerous filter.Performing this rescaling allows us to form the most complete picture of the GRB's behaviour, as some bands may contain points at critical times (i.e. at very early or late times, or even during the appearance of interesting features such as the plateau or a flare) that are not present in other bands.These points may be crucial to subsequent GRB studies, especially involving LC morphology -therefore, when possible, we aim to perform this rescaling.However, when a GRB does display colour evolution, rescaling colours might obscure unique characteristics of the surrounding environment that drive this colour change.
The origins of such features include the common causes of late-time chromaticities, like in the case of dust destruction that leads to a decrease in the optical-NIR extinction (Hoang et al. 2020), or even an associated Supernova (SN) that, when present at the end of the LC, reddens the colour of the GRB afterglow (Cano et al. 2017).Other late-time colour evolution cases are caused solely by the fading of the GRB afterglow with the subsequent emerging contribution of the host galaxy to the luminosity of the transient.Conversely, early-time colour variations typically result from specific phenomena, such as the cooling break energy frequency crossing the observed band or the shift from a reverse-shock regime to a forward-shock regime (Li et al. 2018).Host galaxy extinction presents another potential reason for colour changes: this occurs when dust reappears early on, following its initial destruction by intense GRB radiation, causing a shift to redder colours.However, the time scale for such a process to be set is  ≥ 6 * 10 8 / , where  is the density of the dust region (Perna et al. 2003).This is a process much slower than the time scale for the GRB afterglows to become no longer visible at optical wavelengths.To determine whether GRBs exhibit colour evolution, we start by calculating rescaling factors for each data point and then assess the behaviour of each filter.After determining the colour behaviour for each GRB, we perform the rescaling for the GRBs that do not undergo colour evolution.We developed a customized python script to calculate these rescaling factors, which will be included in the package.Here, we stress that the colour evolution analysis is propaedeutic and necessary for rescaling.The colour evolution analysis is described below.Then, the rescaling factors are calculated as follows: given a GRB with magnitudes in the filters  and , the rescaling factor is defined as the difference between the magnitude of the  filter and  filter at coincident epochs.Thus, the rescaling factor definition is: where   and   follow the condition expressed in Equation 2. The rescaling factors of the  filter to the most numerous inside a GRB LC (we use the  label to mark the most numerous filter) are fitted through the lmfit package that uses the weighted Levenberg-Marquardt algorithm (Levenberg 1944) with a linear model in the form: where  is the slope and  is the normalization.Figure 2 shows an example of magnitude rescaling in both the cases of duplicate and coincident epochs.GRB 000911A is shown before (upper left panel) and after (upper right panel) rescaling.The  band is the most numerous, with 23 occurrences.Lower panels show a case (GRB 030323A) for magnitudes that are not time duplicates, but obey the criteria of 0.025 of Equation 2. Here, the  band is the most numerous, with 16 occurrences.The error on the rescaling factor is defined as Δ  ,  =

√︃
Δmag 2  + Δmag 2  .At least three data points for the rescaling factors among the most numerous filter and the other filters are needed for the fitting.The colour evolution analysis is performed with two approaches: the  = 0 fitting and the variable slope fitting.In the  = 0 fitting, we fix the slope  in Equation 7 to zero and we fit the normalization , estimating then the probability  according to Equation 4, the reduced  2 , and the Bayesian Information Criterion (BIC) value.In this approach, if for the fitting we have  ≥ 0.05 (namely, the probability for the fitting to not be drawn by chance is greater than 0.05), then we consider the fitting as a no colour evolution case, otherwise for  < 0.05, we consider the fitting as a colour evolution case.In the variable slope approach, the slope  of Equation 6 is fitted together with the normalization  and two fitting parameters are free to vary.Also in this case we collect the values of , the reduced  2 , and the BIC.The definition adopted in this approach for the absence of colour evolution is that  should be compatible with zero to within 3 .If this condition is not satisfied, the filter has colour evolution.We consider as undetermined cases (neither colour nor no colour evolution) the ones where the absolute value of the uncertainty of the slope   is larger than the slope  itself.Concerning the rescaling procedure, when the magnitudes in different filters at a given time are within 1  of each other, the rescaling factor is not needed for those magnitudes.Even if no rescaling is needed in certain filters, some individual magnitude measurements in those filters might stand out as unusually different.If these outliers are not caused by flares or bumps, then they might be due to problems like observation difficulties or calibration errors.We have checked the agreement of both the  = 0 and the variable slope methods with already published results in the literature.The cases of disagreement with the literature are presented in Section 4. Furthermore, in the cases where most of the filters do not agree with the literature, but at least two filters agree, these are considered as "apparent" disagreement.In the estimation of the rescaling factors, we apply the following approximations:  ′ ≈ ,  ′ ≈ ,  ′ ≈ ,  ′ ≈ ,  ′ ≈ ,  ′ ≈ , and   ≈ .These approximations guarantee that we can have better coverage of the LCs for the colour evolution investigation.The same assumption does not hold between the Johnson and Cousins filters, since approximating  ≈   and  ≈   would induce systematic effects in the estimation of the colour evolution.The main difference between our analysis and prior approaches for estimating the colour behaviour is our use of the most numerous filter as a reference for all other filters, while in the literature, they consider a pre-determined set of colour indices.Indeed, in the literature, it is very common to find the magnitude differences among the following pairs of filters:  − ,  − ,  − ,  − ,  − , 2 −  2,  2 − 1, 1 − ,  − ,  − ,  − , and  −  (Simon, V. et al. 2001;Li et al. 2018).Nevertheless, with the script in our package, it is possible to change the reference filter for rescaling without necessarily taking the most numerous filter, allowing users to perform the colour evolution analysis using a different filter.

The definition of the Gold and Diamond samples
We define the "Gold" sample and the "Diamond" sample among our total GRBs.To be part of the Gold sample, a GRB must have a computed  opt in the current analysis and fulfill one of two conditions: A) the achromatic behaviour, characterized by colour evolution or the absence thereof, aligns with findings from existing literature.
B) if the chromatic behaviour is determined in this work due to an increased data set compared to previous analyses.
To be part of the Diamond sample, the GRB must have a computed  opt and exhibit no colour evolution.It is crucial to define the Gold sample since it encompasses all the cases where we either find a confirmation of the chromatic behaviour found in the literature or we show for the first time their colour evolution analysis through the unprecedented collection of data of the same GRBs.The Diamond sample, which also includes a subset of the Gold one, is the one characterized by the possibility of rescaling the GRBs to the most numerous filter, which we remind is a necessary step to improve the fitting quality for the LCs.When  is set to zero, the Gold sample comprises 92 GRBs, and the Diamond sample comprises 100 GRBs.When  is variable, 92 GRBs fall in the Gold sample and 67 GRBs in the Diamond sample.To show LCs with the same filters, since the  is on average the most numerous, in Figure 4, we present all the GRBs'  band (345 cases) together with the Gold (shown in yellow) and Diamond (shown in blue) samples for the  = 0. Furthermore, Figure 4, which we name the "Kann plot" in memory of D. A. Kann.

RESULTS ON THE SPECTRAL ANALYSIS AND THE COLOUR EVOLUTION
In the following Subsections, we detail the spectral analysis and the colour evolution results, respectively.

The spectral analysis fitting results
Following the procedure described in Section 3.2, we were able to estimate the spectral index for each epoch of the LC and obtain a total of 924 good fits (namely, with a probability of not being drawn by chance ≥ 0.05) for 138 GRBs.As an example, in Figure 5   The magnitudes and the wavelengths are in AB system, corrected for the Galactic extinction.0.95 (yellow empty diamonds) at the redshifts from 0.1 to 10.The most relevant contribution given by the -correction may arise from spectral indices that are ∼ 0.6 at a redshift  ∼ 10 and impact the observations by 1 .Given that our sample does not extend to such high redshifts, the GRBs in our sample where the -correction is not computed do not suffer from major systematic effects.

The colour evolution analysis results
In the following three Subsections, we show the results of the colour evolution analysis performed by fixing  = 0, then letting  be free to vary, and their comparison with the literature.In both treatments, we define three major groups listed below: • GROUP 1: GRBs without colour evolution; • GROUP 2: GRBs with colour evolution; • GROUP 3: GRBs undetermined by us.
All 535 GRBs and their classification into groups are reported in the Table 4 of the Online Materials.In order to understand if these groups have a trend in their redshifts, we have plotted histograms of redshifts for these three groups in Figure 7 based on the analysis with the slope  = 0.There appears to be no clear dependence of the groups' population on the redshift, with the distributions being clustered at  < 4. The same holds for the variable slope case (reported in the Online Materials).The flowchart in Figure 8 summarizes the number of GRBs retained at each step, from the -correction and the  host extinction correction, to the colour evolution analysis.Due to the restrictive criteria on the epoch, we retain fewer GRBs than the literature, but we highlight that our procedure is uniform across all cases.

4.2.1
The colour evolution: fitting with the fixed  = 0 Following the methodology described in Subsection 3.3, we fit the rescaling factors with a slope  = 0 and determined the colour evolution for a given GRB on the compatibility in majority of the filters.We determined this compatibility for 173 GRBs, excluding the 5 GRBs discussed in 3.2.There are only 4 GRBs (030329A, 100814A, 130427A, and 141220A) in both methods ( = 0 and variable slope) that, despite having enough data points, do not have a constrained value of the probability for the fitting.The results of this analysis are schematized in the left panel of Figure 10.The comparison with the literature of the  = 0 cases is detailed below: • agreement with the literature: 84 cases (75 cases with no colour evolution, and 9 with colour evolution).
• disagreement with the literature: 20 cases (6 cases reported colour evolution in our analysis, but the literature suggests no colour evolution, and 14 cases where our analysis indicates no coour evolution, but the literature suggests colour evolution).
• apparent disagreement with the literature: 3 cases.As defined in Section 3.3, the apparent disagreement cases are the ones where most filters do not agree with the literature, but at least two filters do agree (1 case with colour evolution in our analysis, but no colour evolution in the literature, and 2 cases with no colour evolution in our analysis, but colour evolution in the literature).The disagreement and apparent disagreement GRBs are listed in Table 2.
• determined in our analysis, but not in the literature: 66 cases.9 of these have colour evolution, while 57 show no colour evolution.
• determined in the literature, but not by us: 70 cases (46 cases with no colour evolution, and 24 with colour evolution).We stress that the cases that are undetermined by us are the GRBs for which we do not find at least 3 rescaling factors in at least one filter.
According to our colour evolution analysis with  = 0, we define the following groups (summarized in the upper panel of Figure 13): • GROUP 1: GRBs without colour evolution (148, among which 100 have a measured value of  opt ).In this Group, 75 agree with the literature (61 of which have  opt ), 14 disagree with the literature (10 of which have  opt ), 2 have an apparent disagreement with literature and have  opt , and 57 are determined by us, but undetermined in the literature (27 have  opt ).
• GROUP 2: GRBs with colour evolution (25, and 4 of them with  opt ).In this Group, 9 agree with the literature (3 with  opt ), 6 disagree with the literature and none of these has  opt , 1 has an apparent disagreement (without  opt ), and 9 are determined by us, but undetermined in the literature (1 with  opt ).
• GROUP 3: GRBs undetermined by us (358 with  opt available in 34 cases).In this Group, 45 are undetermined for us but determined in the literature as cases with no colour evolution, 8 them with  opt .21 GRBs are undetermined by us, but are determined in the literature as cases with colour evolution, 4 with  opt .Finally, 292 are both undetermined by us and in the literature (for those, 22 have  opt ).
The advantage of fixing the slope  = 0 is that it reduces the number of GRBs for which we have the same number of filters that exhibit colour evolution and not.Thus, we have a reduced number of undetermined cases.As a consequence, this increases the number of GRBs for which there is an agreement with the literature.In Figure D1, we report the colour evolution analysis for the 3 apparent disagreement cases for  = 0 fitting (080109A, 080319B, and 130702A).

The colour evolution: fitting with the linear model (variable slope)
We also fit the linear model of Equation 7allowing the slope  to vary.Thus, the slope obtained describes the time evolution of the rescaling factors, if present.We determined the chromatic behaviour of 141 GRBs, 18.5% less than the slope  = 0 analysis.This variation is due to the low probability for a linear fitting with variable .We note that the 4 GRBs for which the fitting probability  is not constrained in the  = 0 fitting (030329A, 100814A, 130427A, and 141220A) have an unconstrained  also in the variable slope fitting and consequently, they are excluded from the groups and the following discussion.The results of the comparison between the variable slope case and the literature are the following: • agreement with the literature: 66 cases (57 cases with no colour evolution, and 9 with colour evolution).
• disagreement with the literature: 20 cases (10 cases reported colour evolution in the literature, but our analysis suggests no colour evolution and 10 cases where the literature suggests no colour evolution, but our analysis indicates colour evolution).
• apparent disagreement with the literature: 4 cases (3 cases with colour evolution in our analysis, but no colour evolution in the literature, and 1 case with no colour evolution in our analysis, but colour evolution reported in the literature).
• determined in our analysis, but not in the literature: 51 cases.16 of these have colour evolution, while 35 show no colour evolution.
• determined in the literature, but not by us: 87 cases (58 cases with no colour evolution, and 29 with colour evolution).In the variable  fitting, the cases that are undetermined by us are the GRBs for which either we do not find at least 3 rescaling factors in at least one filter or all the filters with at least 3 rescaling factors have   > .
According to our colour evolution analysis with variable slope instead, we define the following groups (reported in the lower panel of Figure 13): • GROUP 1: GRBs without colour evolution (103, among which 68 have a measured value of  opt ).In this Group, 56 agree with the literature (43 of which have  opt ), 10 disagree with the literature (6 of which have  opt ), 1 has an apparent disagreement with literature and has  opt , and 36 are determined by us, but undetermined in the literature (18 have  opt ).
• GROUP 2: GRBs with colour evolution (38, and 19 of them with  opt ).In this Group, 9 agree with the literature (5 with  opt ), 11 disagree with the literature and 7 of them have  opt , 2 have an apparent disagreement (both with  opt ), and 16 are determined by us, but undetermined in the literature (5  opt ).
• GROUP 3: GRBs undetermined by us (389, with  opt available in 51 cases).In this Group, 57 are undetermined for us but determined in the literature as cases with no colour evolution, 16 of them with  opt .27 are undetermined by us but are determined in the literature as cases with colour evolution, 7 with  opt .Finally, 305 are both undetermined by us and in the literature (for those, 28 have  opt ).

Comparison between the 𝑎 = 0 and the variable slope methods
We compare the results of the slope  = 0 and the variable slope methods.Setting the slope  = 0, the number of GRBs with determined chromatic behaviour is increased by 23% (173 versus 141) compared to the variable slope approach;  = 0 fitting also somewhat reduced the disagreement cases with the literature (17% versus 13%) and slightly increased agreement (47% versus 49%) compared to the variable  fitting.Overall, while there is a slight increase in agreement with the fixed slope model, the overall level of agreement between the two models remains similar and thus, leads to a similar comparison with the literature.As an example, in Figure 9, we compare the  = 0 fit with the variable slope fit for GRB 050319A.We see that for the filters ,   ,  the variable slope value  is compatible with zero in less than 3  (  = −0.298±0.128,   = −0.049±0.134,  = −0.090±0.096), in agreement with what we observe in the  = 0 fitting.The cases of disagreement with the literature and the 5  outliers detailed in the Sec.4.2.4 for both methods are listed on the Online Materials.We evaluated BIC for 136 GRBs for which both fitting were determined.To compare the two fitting methods, we follow the procedure described by Dainotti et al. (2021).For each filter, we find the minimum of the BIC between the variable  and the  = 0 fittings, namely   = (  .,  =0 ).Then, for the two models, we evaluate the relative likelihood   =   /(    ), where   is the weight of the -th model, defined as   =  ((  −   )/2).If, for the -th model, the   > 0.95 then that model is preferred to the other.If both of the   values are < 0.95 then there is no statistical difference between the two models.When the majority of the filters in a given GRB reveal a model that is favoured compared to the other according to its   , then the  model is preferred.
Of the 136 GRBs, for 119 GRBs (87.5%), the   is < 95% for both models, making them statistically indistinguishable.In the remaining 17 GRBs, the   favours the variable slope over the fixed  = 0 slope.Furthermore, all these 17 cases have colour evolution according to the variable slope analysis, showing the model's effectiveness in capturing colour evolution.The pie chart in lower right panel of Figure 10 provides a summary of the BIC analysis' results.
To summarize, comparing the results of the spectral index evolution with the colour evolution results through the two fitting methods, we find the following cases: 6 GRBs for which both  = 0 and variable  show no colour evolution similarly to the results obtained with the SED; 6 GRBs for which  = 0 shows no colour evolution, variable  has colour evolution, and the SED shows in 3 cases evolution and in other 3 cases no evolution; 15 GRBs for which both  = 0 and variable  show no colour evolution but the SED shows evolution; 3 GRBs for which either the  = 0 or variable  are undetermined but the SED shows evolution.

The outliers
In this Section, we discuss the outliers found in the catalogue and their impact on the colour evolution analysis.We classify the outliers as follows: the bad photometry points for which the photometry is not reliable according to their sources (for example, in the case of some GCNs), marked with ; the points whose magnitudes are separated by at least 5  from the other points at coincident epochs marked with  ; the non-simultaneous outliers, for which it is not possible to apply the criterion of coincident epochs defined in Equation 2and their magnitudes deviate more than 5  from the other data points that are the closest according to their time epoch, marked with .According to our definition, there are many outliers with mixed properties, for example, they can be  and  or  and .In total, there are 1492 outliers in this catalogue: 1022 are of the -type, 801 are  -type, and 656 are -type.These outliers are in total a fraction of 1492/64813 = 2%.We have also added to clarify that the points marked  are outliers, which will provide issues during the fits, while the points marked  may be outliers.A conservative approach, similar to the one we used, would reject both the  and  points, while a less cautious approach focused on looking into small effects may require the inclusion of the  points in the fits.In the following, we focused only on the conservative approach and thus rejected points labeled ,  , and .However, we here highlight that the users can also investigate and use these data points for their own data analysis.To evaluate the impact of the outliers with respect to the linear fit of the rescaling factors, we also performed the fitting of the rescaling factors after removing these outliers when their photometry is not fully reliable (like in some cases of the GCNs).This analysis has been performed for both the  = 0 and the variable slope fitting.In the  = 0 approach, there are 49 GRBs for which there are one or more outliers.Among these 49, 12 have outliers caused by non-reliable photometry of GCNs.In the cases of 060218A, 100621A, 120422A, and 210619B, the removal of the outliers does not change the colour behaviour of the GRBs.For 080430A, 081028A, 110503A, 141028A, 150323A, 160121A, 181010A, and 210610B, instead, the removal of the outliers from GCNs prevents the analysis from finding a determined colour behaviour given the lack of sufficient rescaling factors.Concerning GRB 160131A, before the removal, the  band has no colour evolution, but it shows colour evolution after the removal of GCN 18960 (Hentunen et al. 2016).In the variable slope approach, among all the cases that show outliers (41 in total), in 10 GRBs (060218A, 110503A, 111228A, 120422A, 150323A, 160131A, 180325A, 181010A, 210619B, and 230812B) the photometry may not be reliable since the data points that constitute outliers are taken from GCNs.In detail, in the case of 060218A and 120422A, the removal of the outliers from GCN 4831 (Rodgers et al. 2006) and GCN 13279 (Perley et al. 2012) respectively, does not change the results.Concerning 160131A, removing data from GCN 18960 (Hentunen et al. 2016) in the  filter changes the behaviour from no colour evolution to colour evolution.In all the 7 remaining cases, the removal of the outliers led to undetermined colour evolution due to the lack of sufficient data.The case below (in Figure 11) shows GRB 120422A where there is no difference between removing and leaving in the outliers.In the left upper panel, the LC shows the fit highlighting the outlier in the J band with the magenta circles, with the respective rescaling factors in the lower panel.The J filter has rescaling factors slope  = −1.818± 1.091 (shown in the bottom panel), which indicates no colour evolution.In the right panel, the LC is presented without the outlier in the J band: the J filter has rescaling factors slope  = −1.143± 0.470, which still indicates no colour evolution.

Disagreement of our results with the literature
The main reasons for disagreement are summarized below with the label corresponding to Table 4 of the Online Materials: (i) GRBs may exhibit achromatic (labeled as A.1) or chromatic behaviour (labeled as A.2) across different wavelengths, but the comparison between us and the literature is performed in different bands.
(ii) Different methodologies in data analysis (labeled with B).
(iii) Inconsistent data availability at coincident times with certain filters (C.1) or lack of sufficient data (labeled with C.2).
(iv) rescaling in different filters between us and the literature (D.1 labels cases of agreement with the literature when changing the filter for rescaling; D.2 indicates GRBs when only a few filters agree with the literature after changing the reference filter; D.3 labels GRBs when we have the same number of filters with and without colour evolution).
For the  = 0 fitting, there are 20 GRBs whose colour evolution is in disagreement with the ones reported in the literature and 3 GRBs for which there is an apparent disagreement behaviour.These cases are listed in the left half of Table 2.There are additional 70 cases for which we are not able to discern any colour behaviour but they are discussed in the literature.For the variable slope analysis, we find 21 GRBs whose colour evolution is in disagreement with the ones reported in previous work and 4 GRBs for which there is an apparent disagreement.For 87 cases, we are not able to discern any colour behaviour while the literature discusses them.These events are listed in the right half of Table 2.The results of this scheme are reported in the 4th column of Table 4 on the Online Materials.

The approach of the monochromatic light curve fitting
Finally, we briefly discuss the monochromatic LC fitting approach to estimate the colour evolution.It might be argued that our approach could be less accurate than the monochromatic LC fitting.We do infer the evolutionary behaviour by comparing the time-decay power-law index parameter () estimated in each filter for a given GRB.The advantage of this method is that we can determine it for multiple segments of the LCs assuming our   / = 0.025.If we fit all LCs we can have evolution in some segments which would be masked out if we had not considered this   / condition.The two-step process is now automated and the code will be made publicly available so anyone can change the condition on   / from 0.025 to the desired value.In such a process, the LCs features will not be washed out since the criterion for   / is pretty restrictive, so all features, if they exist, must be preserved.
Fitting the monochromatic LCs and estimating their slopes is a less restrictive assumption than our method where we investigate the colour evolution only in the time ranges where multiple filters are  available.For completeness of the treatment, we tested the monochromatic method.We consider here as an example the case of GRB 221009A, shown in Figure 12, where we can observe the fitting through a power-law model defined below.We fitted the monochromatic magnitude values versus log 10 () with a simple power-law model following Li et al. (2018).In this model,  is the slope value and all the slopes are compatible with the ones in  filter to 1, indicating no colour evolution.The power-law model is the following: mag  () = 2.5  log 10 () + mag 0 , where the corrections for the Galactic and host extinctions are already included in the mag  value on the left side, the subscript  is the filter, and mag 0 is the intercept of the fitting.Applying this model to GRB 221009A and fitting only the filters for which at least 3 data points are available, we obtain the following values of :   = 1.09 ± 0.18,   = 1.04 ± 0.11,   = 0.94 ± 0.12,   = 1.15 ± 0.03,   = 0.99 ± 0.09,   = 1.13 ± 0.03,   = 1.13 ± 0.22,    = 1.06 ± 0.13,   = 1.09 ± 0.40,   = 0.67±0.10,  = 1.15±0.03,   = 1.10±0.14,   = 1.04±0.15,and  625 = 1.43 ± 0.07.These values are all compatible in 1 and this is the same outcome that we obtain with our approach of fitting the rescaling factors as a function of time.We note that the average value of the slope of GRB 221009 is different in some filters Table 2.The disagreement and apparent disagreement cases for both the  = 0 and the variable slope fittings.For each of the subtables the structure is the following: first column = the GRB is specified; second column = the colour evolution (Yes) or no colour evolution (No) behaviour from literature; third column = the colour or no colour evolution of the GRB from our analysis.due to the different number of data points, however, the results are compatible within the uncertainties.

OPTICAL REPOSITORY WEBPAGE
Here, we introduce the framework for the dedicated webpage developed to facilitate the analysis of optical LCs.Following a similar approach to the repositories established by Evans et al. (2009), our webpage is designed as a powerful and efficient resource for the community.It enables users to view and interact with LCs and provides valuable insights into their colour evolution by utilizing modules developed as integral components of the grbLC package (see Section 5.1).Currently, the repository comprises 535 LCs, encompassing the magnitude files corrected for Galactic extinction and transformed into the AB system, and the initially gathered raw data, which users can conveniently download.The webpage infrastructure is constructed using streamlit, a python-based framework for generating web-based data visualization applications with minimal reliance on web technologies.The seamless integration of streamlit with other python packages has streamlined the development of an interactive web application based on grbLC, allowing a fully functional online interactive user experience.Furthermore, the expandable and customizable nature of streamlit enables us to iterate between improved versions of the webpage readily.Shortly, our webpage is set to expand its scope by incorporating additional observed optical LCs from the literature and private sources that may not be present currently.Interested users can contact the corresponding author and contribute their materials -be they codes or data -as part of a potential follow-up to this catalogue.

The grbLC package
The grbLC package is an open-source python-based program designed to handle photometric data from various telescopes worldwide.The package associates the telescope information in the GRB magnitude files with our comprehensive table of telescope calibration data.This data encompasses details such as effective wavelengths, zero point flux densities of filters, and conversion values for magnitude system transformations.The package is comprised of three modules for scraping GCNs, conversion in uniform units, and colour evolution, respectively.Currently, grbLC can be manually built from the source, accessible on GitHub at the link https://github.com/SLAC-Gamma-Rays/grbLC,by Mac, Linux and Windows users.We provide the users with detailed documentation on GitHub to walk them through the package, including the installation instructions.
In addition to hosting our catalogue, the webpage provides an alternative user-friendly way to interact with the grbLC package.The primary page provides essential information about a given GRB, including its RA, DEC, redshift (), optical spectral index ( opt ), host extinction correction ( ℎ  ) along with the dust model, and plots of magnitudes.Users can toggle between plots of the original magnitudes collected from literature and the homogenised version where all magnitudes are shifted to the AB system and corrected for Galactic extinction.By using the user-friendly interface menu, users can explore the colour evolution behaviour of a GRB and download all magnitude values in a machine-readable .txtfile.Figure 14 depicts an example of this interface.A demonstration of the web repository can be viewed at https://youtu.be/3n1jtJH4uGU

The GRB data scraper
To streamline the compilation of our data sample, we created a web scraper that automates the GCN search process for pertinent information about a specific GRB.A major part of this module is developed with the Julia Programming Language-based for efficient data scrapping.This web scraper employs reading techniques to parse and segregate data for our sample automatically.Like a "scavenger hunt," the process encompasses two primary phases: manual data collection and coding automation for future endeavours.The manual data collection phase involves crawling GCNs and published data sources to gather relevant information.This phase helps us find GRB identifiers and keywords related to each telescope to help automation.Based on this information, our data scraper can conduct searches by employing regular expression (regex) packages to filter GCNs containing the specified keyword.By inputting the GRB name, the web scraper compiles all associated GCNs and stores them in a user-accessible .txtfile.Furthermore, the data scraper includes a feature that extends its search to the NASA Astrophysics Data System (ADS) catalogue.This functionality allows users to search for papers related to a designated GRB and keyword.The GCN scraping process is conducted on the local machine, working on already downloaded .txtfiles, while the relevant papers are efficiently downloaded in .pdfand compressed into .gzformat to the user's computer, facilitating a swifter and more streamlined access to published data.In the current version, the data scraper can identify more than 32500 optical data points from the following telescopes and networks: UVOT, MASTER, RATIR, GROND, KAIT, IKI-FuN GRBs.We stress that the development of this tool is an ongoing project and the plan is to include all the optical telescopes reported in the literature.

The magnitude conversion
Standardization becomes crucial when dealing with a diverse data sample from various sources to accommodate the different reporting formats.The primary variables requiring attention are the correction for Galactic extinction and the magnitude system.This standardization process is the core function of the conversion module.The Galactic extinction correction is implemented based on Equation 1.The module utilizes the python package dustmaps to retrieve the colour excess ( − ) from Schlegel et al. (1998).The conversion coefficient (  ) for a specific band is sourced from Schlafly & Finkbeiner (2011) or the Asiago Database on Photometric Systems (ADPS, Fiorucci, M. & Munari, U. 2003) if the former lacks the band information.The package also facilitates the conversion between Vega and AB magnitude systems.After the Vega→AB shift, the module allows for the application of -correction (utilizing the optical spectral index from literature) and the host galaxy extinction correction, following Equation 5.The extinction maps presented in Pei (1992) for the three dust models (SMC, LMC, and MW) are used alongside the measured  ℎ  value from the literature.These conversions can be executed using the convertGRB function.

The colour evolution
The evolution module is dedicated to facilitating our colour evolution studies.Once a Lightcurve object is defined, users can visually explore the LC using the displayGRB function.This function generates an interactive plot using Plotly, featuring a colour map based on the observation bands.For a comprehensive colour evolution analysis, the colorevolGRB function comes into play.As detailed in Section 3.3, this function utilizes a specific methodology and provides users with a Pandas data frame.The data frame encompasses crucial information about the fit results of rescaling factor slopes for each filter and indications of whether colour evolution is detected.Additionally, the function generates a Matplotlib plot illustrating the rescaling factor slope.A significant aspect of this module is its ability to list the names of GRBs where colour evolution is identified or absent.The GRB can be selected from the drop-down menu on the left.The GRBs' right ascension, declination, redshift, and spectral index are displayed on the left.In the Colour Evolution menu, it is possible to visualize the filter fitting parameters and GRB rescaling factors.GCN and ADS searches enable searches on the NASA ADS website, providing access to the results about the required GRB.The magnitude file for the shown GRB can be downloaded as a .txtfile and stored using Download.The magnitudes in the original format or those in AB corrected for galactic extinction can be switched between using the Choose format.The filters to be shown can be chosen by clicking on them.
We introduce a compilation of 535 optical LCs with known redshifts, including our publicly accessible web-based LC repository.The Online Materials show each singular LC and spectral indices.This compilation encompasses 64813 data points gathered from literature, circular notices, and private communications.We computed the optical spectral index for 138 GRBs and investigated the presence of colour evolution.This analysis lays the groundwork for the subsequent rescaling of GRB LCs, a procedure feasible only in cases without colour evolution.Our results reveal that the colour evolution scenario, whether present or absent, has been determined for 173 GRBs.To reach this conclusion, we have provided a new homogeneous analysis for all LCs in the repository, which comprises the analysis for the colour evolution and spectral indices.We also compared our analysis and results to the ones provided in the literature.
In particular, we identified 84 cases of agreement with previous studies and 23 instances of disagreement by fitting the rescaling factors with fixed slope  = 0.Among the disagreement cases, it is worth noting that the disparity could arise from differences in the choice of the reference filter for estimating rescaling factors, where the literature may not necessarily consider the most numerous filter within the GRB as the reference.Another potential difference lies in assuming a 0.025 precision level for two epochs to be considered coincident.Besides the comprehensive analysis of the colour evolution, we have developed a user-friendly software suite for analyzing GRB LCs, offering various capabilities.The software can efficiently convert magnitudes reported in the literature to the AB system, correct for Galactic extinction, and incorporate -correction and host galaxy correction, where applicable.Additionally, we have created a web scraper tool that can search through GCNs and the NASA ADS to extract relevant data related to specific GRBs or keywords.This powerful tool streamlines data gathering for studies involving extensive datasets.We also provide detailed documentation of the functions included in the grbLC package in this study.The web scraper tool is also available within this webpage, alongside the option to conveniently download available data for specific LCs.Given the wide variety of formats in the available data, this painstaking data gathering procedure started with this open-source project, and it is still ongoing as additional refinements need to be added.We hope the community will join the effort and contribute via the GitHub repository, enhancing this package and its future use.

DATA AVAILABILITY
The data gathered in this catalogue was obtained either by public sources, in which case the published work is cited; GCN circulars, which have their number cited; or private communications.The open-source compiled catalogue and its data are available in grbLC at https://github.com/SLAC-Gamma-Rays/grbLC.The grbLC package and web-based repository are designed to be open-access and available to all community members.The data will be available after the acceptance of this manuscript.All are welcome to use our software, though we ask that if the provided data or software is used in any publication, the authors cite this paper and include the following statement in the acknowledgements: "Data used in our work is taken from the catalogue Dainotti et al. (2024), and the original data sources are cited within."All the photometric data for the 535 GRBs weigh 5.64 MB, while the literature information about GRB properties is stored in a table of 100 KB.

Figure 1 .
Figure1.The comparison between the optical spectral indices (in the x-axis) and the X-ray ones (in the y-axis).Upper panel.The green data points are the  opt s that follow the  opt =   − 0.5 relation (48 cases), while the red ones respect the  opt =   relation (56).The red line is the relation  opt =   , while the green line represents  opt =   −0.5.Lower panel.The purple data are the GRBs that respect both the relations  opt =   − 0.5 and  opt =   (34), while the blue ones do not respect any of the two (43).The color-coding is the same in both panels.

Figure 2 .FrequencyFigure 3 .
Figure 2.An example of GRB rescaling when times are duplicated, namely magnitudes co-occurring in different filters.GRB 000911A is shown before (upper left panel) and after (upper right panel) rescaling.Lower panels show a case for magnitudes that do not have time duplicates, but obey the condition in Equation 2. In the example, GRB 030323A is shown before (lower left panel) and after (lower right panel) rescaling.

Figure 4 .
Figure 4. Magnitudes in the AB system, corrected for Galactic extinction, are shown for the Gold (in yellow), the Diamond samples (in cyan), and all the other GRBs (in grey), identified via the  = 0 fitting.We call this plot the "Kann plot" in memory of D. A. Kann.
we show the SED fitting for GRB 080928A.While comparing spectral indices from different epoch, we need to consider how the evolution in redshift plays a substantial role.In Figure 6, we show the behaviour of the -correction with redshift for possible spectral indices values in GRBs, color-coded and with different symbols: 0.5 (blue circle), 0.6 (blue filled circles), 0.65 (yellow filled squares), 0.70 (green filled diamonds), 0.80 (purple downward triangles), 0.85 (red empty circles), 0.90 (blue empty squares), and

Figure 5 .
Figure 5.The SED fitting for GRB 080928A.The regression line is estimated by the Levemberg-Marquardt method.The filters, the log-time range, the  2 , and the reduced  2 of the fit are shown in the title of the plot.The magnitudes and the wavelengths are in AB system, corrected for the Galactic extinction.

Figure 6 .
Figure 6.The plot of -correction as a function of  for different  opt values.

Figure 8 .
Figure 8.The flowchart of the GRB analysis.

Figure 9 .
Figure 9.A comparison of the  = 0 fitting (left panel) and the variable slope fitting (right panel) for GRB 050319A.For both panels, the upper half shows magnitudes versus times, while in the lower half we report the rescaling factors vs time with their fitting functions.

Figure 10 .
Figure 10.Visual representation of the results of our colour evolution analysis.Upper panel.Venn diagrams detailing agreement with literature for the two models: Right -A linear model with  = 0, Left -A linear model with variable slope.Lower left panel.The comparison between the two models.Note that in this last panel the disagreement and apparent disagreement cases have been counted together for simplicity and have been reported in the red cell.Lower right panel.The best model based on BIC.In the pie chart, Blue represents the linear model with  = 0, Orange is the linear model with variasble  and Grey represents the cases where both models yielded equal BIC, hence equal preference.

Figure 11 .
Figure 11.Left panel.The GRB 120422A where the outlier for the  band (in the magenta circles) is included and the variable slope fitting is performed.The most numerous filter is shown with black rhombuses, while the other filters are shown with coloured dots.Right panel.The same case of the upper panel but with the removal of the  band outlier.The same convention on the colours of left panel is applied here.

Figure 12 .
Figure 12.The monochromatic LCs fitting for the GRB 221009A: the plot shows the magnitudes vs the log 10 of time after the trigger (in seconds) together with the best fit curves estimated with the simple power-law model.In the legend, the  values for each filter together with the 1  values.

Figure 13 .
Figure 13.The flowchart of the Groups division.Upper panel.Groups division based on the colour evolution with a linear fit where  = 0. Lower panel.Same groups for a fit with variable slope.In round brackets, the number of GRBs with known  opt .

Figure 14 .
Figure 14.An example of the grbLC main page for GRB 970228A.The plot shows the LC in magnitudes versus the log 10 of the midtime of the observations (in seconds) after the trigger.The points with error bars indicate the magnitudes, while the points with downward grey arrows indicate the limiting magnitudes.The GRB can be selected from the drop-down menu on the left.The GRBs' right ascension, declination, redshift, and spectral index are displayed on the left.In the Colour Evolution menu, it is possible to visualize the filter fitting parameters and GRB rescaling factors.GCN and ADS searches enable searches on the NASA ADS website, providing access to the results about the required GRB.The magnitude file for the shown GRB can be downloaded as a .txtfile and stored using Download.The magnitudes in the original format or those in AB corrected for galactic extinction can be switched between using the Choose format.The filters to be shown can be chosen by clicking on them.

Table 1 .
In this table, we report an excerpt of GRB 970508A from the collected catalogue of 535 GRBs.