Changes between Version 71 and Version 72 of DC/Winter2013


Ignore:
Timestamp:
09/11/2017 11:46:08 AM (3 months ago)
Author:
ktl
Comment:

Removed content (which is now in DMTN-035) at Tim's request.

Legend:

Unmodified
Added
Removed
Modified
  • DC/Winter2013

    v71 v72  
    22 
    33= Winter 2013 LSST DM Data Challenge Release Notes = 
    4  
    5 Previous Data Release: [wiki:DC/Summer2012 Summer2012]. 
    6  
    7 == Image Differencing == 
    8  
    9 A report is available [attachment:W13report_V1-02.pdf? here]. 
    10  
    11 == SDSS Stripe 82 Reprocessing (co-adds and forced photometry) == 
    12  
    13 LSST Data Management have recently finished building r-band SDSS Stripe 82 co-adds and performing forced photometry on individual epochs of g, r, and i-band Stripe 82 data. The primary goal of this effort was to add the capability to make background matched co-adds to the DM stack, and test it at large scale by reprocessing real survey data. The full description of the challenge is available in the [attachment:DC_Handbook_v2.1.2.pdf Handbook (draft)]. 
    14  
    15 Background matched co-adds preserve the diffuse astrophysical backgrounds in the stacked image. This increases its scientific usefulness. Furthermore, the thus constructed background in the co-add has higher S/N, making it easier to subtract it when needed. We plan to use this method to generate the co-adds in LSST production, and it was therefore important to have it built into the stack early. This will make upcoming tests of stackfit/multifit algorithms more realistic. 
    16  
    17 Our preliminary analysis indicates the quality of this dataset is comparable to the quality of Stripe 82 reprocessing we performed for [wiki:DC/Summer2012 Summer 2012], but with significantly more area. '''Nevertheless, we caution you that these data were processed "on a budget", with prototype code, and minimal quality assessment. They are a byproduct of an ongoing software development effort, and not a result of a concerted scientific investigation. The code is still incomplete both in terms of features and quality, and the same will be true of the reprocessed data. In particular, if you plan to use this data set for science, expect to have to devote time to perform additional QA, and to have to communicate with DM developers to understand the details of the dataset.''' 
    18  
    19 If you do notice issues, or have questions, please don't hesitate to contact us at dm-help --at-- lsst.org. 
    20  
    21 A report is available [attachment:"S12 W13 Stripe82 Report.pdf"]. 
    22  
    23 === Data Access Rules === 
    24  
    25 The data products provided by LSST DM are intended only for members of the LSST Science Collaborations unless noted otherwise. If their use results in a publication, their users will have to to abide by the [http://www.lsstcorp.org/ahm2012/Plenary/BurchatPubBoard.pdf LSST Publication Policy]. 
    26  
    27 The products are protected by a "well known" username and password: ''lsst / 3gigapix! '' . Using this combination to access the data implies you understand and accept the restrictions on their redistribution and usage. 
    28  
    29 === Data Locations === 
    30  
    31 Note !#1: Some of these websites require the standard DM user name and password; see the 'Data Access Rules' section for more information. 
    32  
    33 Note !#2: The catalog and image access tools in use here are temporary solutions for data distribution built with off-the-shelf open source components: in particular, they are ''NOT'' representative of the Science User Interface the LSST will ultimately have. 
    34  
    35 Where to get data/information: 
    36  * Contact: dm-help --at-- lsst.org 
    37  * Winter 2013 Data Challenge Handbook (Draft): [attachment:DC_Handbook_v2.1.pdf] 
    38  * Catalogs access (phpMyAdmin): https://lsst-web.ncsa.illinois.edu/mydb/  
    39    * Note !#1: if you don't have a DM mysql database account, use [http://lsst-web.ncsa.illinois.edu/dbaccount this link] to sign up. 
    40    * Note !#2: You will need the "well known" username/password to access the signup form and the database entry page (see the Data Access Rules section above) 
    41    * Note !#3: If you're accessing using the native mysql client, the database name is {{{DC_W13_Stripe82}}}. 
    42  * FITS image access (coadds): [wiki:W2013WebDataAccess] 
    43  * Co-add visualization tool: http://moe.astro.washington.edu/sdss/ 
    44  * Data Management Software v6_1 Release (an intermediate Winter 2013 release): [wiki:Installing/Winter2013] 
    45  
    46 If you decide to make use of this data, feel free to inquire about the details at <dm-help --at-- lsst.org>. 
    47  
    48 === Description of Data Products === 
    49  
    50 We used 298 runs imaged as a part of SDSS Stripe 82 (2 million fields) to create a deep co-add approximately covering -40 deg < R.A. < 55 deg, -1.25 < Dec < 1.25 (237 deg^2^). No PSF matching was performed on the co-adds, making them deeper but less suitable for photometry. The co-adds were used to detect 14.7 million sources, most of which would otherwise fall below the faint limit of individual exposures. Photometry was performed in individual epochs, at the location of each source detected in the co-adds, resulting in 3.9 billion ''g'', ''r'' and ''i'' band measurements ("forced photometry"). ''u'' and ''z'' bands were not processed. We also produced catalogs of averaged forced photometry, both across the duration of the whole survey (~10 years), and on a yearly basis. 
    51  
    52 The co-adds are available as a series of 5126 FITS images, each spanning 2060 x 1937 pixels (see [wiki:W2013WebDataAccess] for how to access them). The catalogs are kept in a MySQL database and available through phpMyAdmin web interface, or (for power users), through the command-line mysql client. See the [wiki:DC/Winter2013#DataLocations prior section] for instructions on how to access the database, and how to open an account if you don't already have one. 
    53  
    54 The database contains 20 tables and views; however, only a few are of general interest (listed below). We do not have at this time a detailed description of the schema of each of these tables; however, the [http://lsst1.ncsa.uiuc.edu/schema/index.php?sVer=S12_sdss Summer2012 schema] should provide sufficient information to understand the meaning of most of these columns even though the exact names may have changed. 
    55  
    56 Most frequently used tables: 
    57  
    58 '''!AvgForcedPhot''' 
    59     A table with percentiles of photometry in each band (5th, 25th, 50th (median), 75th, 95th). Columns: 
    60  
    61 || '''Column''' || '''Description''' || 
    62 || deepSourceId || object identifier || 
    63 || ra || Right Ascension (degrees) || 
    64 || decl || Declination (degrees) || 
    65 || nMag_[gri] || number of measurements for the band || 
    66 || magFaint_[gri] || magnitude of the faintest measurement in the band || 
    67 || medMag_[gri] || magnitude of the median measurement in the band || 
    68 || magBright_[gri] || magnitude of the brightest measurement  in the band || 
    69 || q1Mag_[gri] || magnitude of the first (faintest) quartile measurement in the band || 
    70 || q3Mag_[gri] || magnitude of the third (brightest) quartile measurement in the band || 
    71 || faint5perMag_[gri] || 5th percentile magnitude in the band || 
    72 || bright5perMag_[gri] || 95th percentile magnitude in the band || 
    73  
    74     All percentiles were calculated on the fluxes and converted back to magnitude for convenience. 
    75  
    76 '''!AvgForcedPhotYearly''': 
    77     Same as the !AvgForcedPhot table, except the percentiles are computed for each year of the survey. Therefore there are typically ~10 rows per object. Compared to !AvgForcedPhot, this table has one extra column (''year'', running from 1 to 10), and no 5th and 95th percentile columns. 
    78  
    79 '''!DeepForcedSource''' 
    80     Table with forced photometry measurements in individual epochs. Use this table if you're interested in querying for complete light curves. 
    81  
    82 '''!DeepSource''' 
    83     A table of sources detected on co-adds. This is in effect the master "object catalog". Note however that because the co-adds were not PSF-matched, the photometry in this table will be relatively poor; use !AvgForcedPhot table instead. 
    84  
    85 '''!RefObject''' 
    86     A containing SDSS DR7 Stripe82 co-add ([http://arxiv.org/abs/1111.6619 Annis et al. 2011]) catalog. It's been matched to !DeepSource via !RefDeepSrcMatch table. 
    87  
    88 '''Science_Ccd_Exposure''' 
    89     A table with metadata for all SDSS Stripe82 [http://skyserver.sdss.org/dr7/en/sdss/data/data.asp fields]. 
    90  
    91 '''Science_Ccd_Exposure_coadd_r''' 
    92     A table with metadata for all co-add "patches" (when producing the co-add, we divided the sky into large "tracts", and each tract has been subdivided into "patches"). The patches are stored as FITS files on the [wiki:W2013WebDataAccess image server]. 
    93  
    94  
    95 === Example Queries === 
    96  
    97 '''Retrieve median g, r, i magnitudes for all objects in a (ra, dec) box''' 
    98  
    99 {{{ 
    100 SELECT 
    101         ra, decl, 
    102         medMag_g, medMag_r, medMag_i 
    103 FROM 
    104         `AvgForcedPhot` 
    105 WHERE 
    106               ra BETWEEN 0.01 and 0.02 
    107         AND decl BETWEEN 0.03 and 0.04 
    108 }}} 
    109  
    110 Alternatively, you can use [http://nebel.rc.fas.harvard.edu/mjuric/std/w12/external/scisql/0.1/scisql-0.1/doc/ scisql] geometry functions; this should speed up queries over large area: 
    111  
    112 {{{ 
    113 SET @poly = scisql_s2CPolyToBin(0.01, 0.03, 0.02, 0.03, 0.03, 0.04, 0.01, 0.04); 
    114 CALL scisql.scisql_s2CPolyRegion(@poly, 20); 
    115  
    116 SELECT 
    117         ra, decl, 
    118         medMag_g, medMag_r, medMag_i 
    119 FROM 
    120         `AvgForcedPhot` 
    121 WHERE 
    122         scisql_s2PtInCPoly(ra, decl, @poly) = 1 
    123 }}} 
    124  
    125 '''Retrieve a g-band light curve for object 1398579058966639.''' 
    126  
    127 {{{ 
    128 SELECT 
    129     deepSourceId, 
    130     deepForcedSourceId, 
    131     exp.run, 
    132     fsrc.timeMid, 
    133     scisql_dnToAbMag(fsrc.psfFlux, exp.fluxMag0) as g, 
    134     scisql_dnToAbMagSigma(fsrc.psfFlux, fsrc.psfFluxSigma, exp.fluxMag0, exp.fluxMag0Sigma) as gErr 
    135 FROM 
    136     DeepForcedSource AS fsrc, 
    137     Science_Ccd_Exposure AS exp 
    138 WHERE 
    139         exp.scienceCcdExposureId = fsrc.scienceCcdExposureId 
    140     AND fsrc.filterId = 1 
    141     AND NOT (fsrc.flagPixEdge     | fsrc.flagPixSaturAny | 
    142              fsrc.flagPixSaturCen | fsrc.flagBadApFlux   | 
    143              fsrc.flagBadPsfFlux) 
    144     AND deepSourceId = 1398579058966639 
    145 ORDER BY 
    146     fsrc.timeMid 
    147 }}} 
    148  
    149 Notes: 
    150  * The times (timeMid column) denote the mid-points of exposure each SDSS frame. Since SDSS took data in TDI mode, these have to be corrected to the effective time of observation of each object. 
    151  * No effort has been made to remove objects doubly-detected in overlap regions of SDSS frames. You may therefore get more than one measurement per run. 
    152  
    153 '''Retrieve a g-band light curves for all objects with 0.0 < ra < 0.01deg and 0.0 < dec < 0.01deg.''' 
    154  
    155 {{{ 
    156 SELECT 
    157     deepSourceId, 
    158     deepForcedSourceId, 
    159     exp.run, 
    160     fsrc.ra, 
    161     fsrc.decl, 
    162     fsrc.timeMid, 
    163     scisql_dnToAbMag(fsrc.psfFlux, exp.fluxMag0) as g, 
    164     scisql_dnToAbMagSigma(fsrc.psfFlux, fsrc.psfFluxSigma, exp.fluxMag0, exp.fluxMag0Sigma) as gErr 
    165 FROM 
    166     DeepForcedSource AS fsrc, 
    167     Science_Ccd_Exposure AS exp 
    168 WHERE 
    169         exp.scienceCcdExposureId = fsrc.scienceCcdExposureId 
    170     AND fsrc.filterId = 1 
    171     AND NOT (fsrc.flagPixEdge     | fsrc.flagPixSaturAny | 
    172              fsrc.flagPixSaturCen | fsrc.flagBadApFlux   | 
    173              fsrc.flagBadPsfFlux) 
    174     AND fsrc.ra BETWEEN 0.0 AND 0.01 
    175     AND fsrc.decl BETWEEN 0.0 AND 0.01 
    176 ORDER BY 
    177     fsrc.deepSourceId, fsrc.timeMid 
    178 }}} 
    179  
    180 Notes: 
    181  * Expect this query to take 1-2 minutes to complete. It will return 2,014 rows. 
    182  
    183 === Covered footprint === 
    184  
    185 A quick visualization of the footprint available in Winter 2013 Stripe 82 data, created by plotting all 15.9 million detected objects: 
    186  
    187 [[Image(DC/Winter2013:Winter2013-Stripe82-Footprint.png)]] 
    188 [[Image(DC/Winter2013:Winter2013-Stripe82-Footprint-Zoomed.png)]] 
    189  
    190 === RGB Color composites === 
    191  
    192 RGB color composite of an area in the vicinity of M2: 
    193  
    194 [[Image(DC/Winter2013:M2Composite.png)]] 
    195  
    196 The full-sized image can be viewed/panned/zoomed at http://moe.astro.washington.edu/sdss/. 
    197  
    198 === Quality assessment === 
    199  
    200 == Comparison to S2012 and Completeness == 
    201  
    202 We took a small subset of data from both the Summer 2012 DC and the Winter 2013 early production DC.  Using the DEEP2 catalogs [http://adsabs.harvard.edu/abs/2004ApJ...617..765C Coil et al. (2004)] as reference, we compare the completeness as a function of magnitude between the two reductions. 
    203  
    204 [[Image(DC/Winter2013:S12_W13_comp.png, 400px)]] 
    205  
    206 The Winter 2013 (blue) completeness tracks very well with the Summer 2012 (red).  This shows that we have not changed anything substantial between the two reduction runs. 
    207  
    208 We next look at a much large section of the survey covering the Deep2 Field 4 photometric catalogs.  We construct completeness and contamination profiles for the Winter 2013 DC. 
    209  
    210 [[Image(DC/Winter2013:completeness.png, 400px)]] 
    211  
    212 In addition to comparing to the DEEP2 catalogs, we compare the completeness of the [http://adsabs.harvard.edu/abs/2011arXiv1111.6619A Annis (2011)] catalogs to the Winter 2013 results.  The Winter 2013 catalog is significantly less complete at bright magnitudes.  We are looking more into this, but early evidence suggests this is due primarily to background subtraction around bright stars and to the fact that multiple peaks within a single detection footprint are not de-blended into individual sources for the Winter 2013 runs.  We have placed a 5-sigma S/N threshold on the Annis catalog and the Winter 2013 catalog does not go significantly below 5-sigma.  With these cuts the Winter 2013 catalog goes ~0.2 mag deeper than the Annis catalog. 
    213  
    214 The completeness plot is not the whole story.  We also look at the trends in S/N between the Annis (2011) catalog and the Winter 2013 catalog.   
    215  
    216 [[Image(DC/Winter2013:snr.png, 400px)]] 
    217  
    218 This shows that for constant S/N the Winter 2013 catalog goes about 0.75 mag deeper than the Annis (2011) catalog.  We also see that the Winter 2013 catalog is 10-sigma at our 50% limiting magnitude of 24.2.  This suggests that a 5-sigma threshold on the coadd to seed forced photometry is too conservative and that we should have pushed to 3-sigma (or fainter) in the coadd to reach completeness in the coadded catalog at 5-sigma. 
    219  
    220 We also looked for contamination in the Winter 2013 catalog.  We define contamination simply as any object in the Winter 2013 catalog that is not in the DEEP2 catalog.  The following figure shows that there is less than 5% relative contamination to our limiting magnitude. 
    221  
    222 [[Image(DC/Winter2013:contamination.png, 400px)]] 
    223  
    224  
    225 The production pipelines perform photometric calibration using the catalog of [http://adsabs.harvard.edu/abs/2007AJ....134..973I Ivezic (2007)].  In this analysis we look at the distribution of forced photometry principal colors of stellar sources, described in [http://adsabs.harvard.edu/abs/2004AN....325..583I Ivezic (2004)].  We use the star-galaxy separation provided by the [http://adsabs.harvard.edu/abs/2011arXiv1111.6619A Annis (2011)] Stripe82 catalog to select point sources for the analysis; we do not do any native star-galaxy separation.  The figure below illustrates the process of defining a principal color (adopted from [http://adsabs.harvard.edu/abs/2004AN....325..583I Ivezic (2004)]): 
    226  
    227 [[Image(zeljko_w.png, 240px, align=top)]] 
    228  
    229 The width of the stellar locus perpendicular to principal color P1 (top) is a function of underlying stellar astrophysics, and errors on the photometry.  As the bottom panel demonstrates, this width increases as a function of magnitude, as photometric uncertainties start to dominate.  In our analysis, we look at the principal colors '''''w''''', shown in the figure above, and '''''x''''', which is the width perpendicular to the vertical distribution in the (r-i) vs. (g-r) diagram above.  We examine below the width of the principal loci as a function of the number of epochs for forced photometry: using 1 epoch (i.e. all the data), the median (in flux) of two epochs (where the flux medians to a value > 0.0), and the median of 10 and then 40 epochs.  We first show the results for Summer2012 processing below: 
    230  
    231 [[Image(S12_2.png, 360px, align=left)]] 
    232 [[Image(S12.png, 360px, align=top)]] 
    233  
    234 The '''''left''''' image provides the distribution of points around the principal colors '''''w''''' and '''''x''''' (i.e. the principal locus is at x=0 in all plots).  Each panel shows the all-data distribution, and then the median across epochs for all objects with N>9 epochs.  When medianing across many measurements, the locus becomes tighter, and is less dominated by the photometric uncertainties to fainter magnitudes.  The '''''right''''' panel shows how the width of this locus improves as a function of the number of epochs, for N=1,2,10,40 epochs, along with a histogram of the number of objects vs. r-band magnitude. 
    235  
    236 We examine below the results of the Winter2013 processing for one of the 6 SDSS camcols.  This includes data from camcol=1 of both the N and S strips of the stripe.  We subdivide the data into areas 10 degrees wide in RA, and provide measurements of the median and standard deviation of the distributions (computed as 0.741 times the interquartile range) in tabular form for the first RA range. 
    237  
    238 == -40 < RA < -30 == 
    239 [[Image(ra_dec_full0.dat_1.png, 360px, align=left)]] 
    240 [[Image(ra_dec_full0.dat_2.png, 360px, align=top)]] 
    241  
    242 {{{ 
    243 Mag             w;N=1        w;N=2       w;N=10       w;N=40                             x;N=1        x;N=2       x;N=10       x;N=40 
    244 15.25      -0.004,0.015 -0.006,0.015 -0.002,0.010 -0.002,0.010                      -0.027,0.016      ...          ...          ...  
    245 15.75      -0.004,0.016 -0.003,0.015 -0.003,0.011 -0.003,0.010                      -0.009,0.032 -0.008,0.030 -0.010,0.022 -0.006,0.022 
    246 16.25      -0.003,0.016 -0.002,0.014 -0.002,0.010 -0.002,0.010                      -0.015,0.037 -0.006,0.021 -0.013,0.027 -0.009,0.022 
    247 16.75      -0.003,0.016 -0.004,0.014 -0.003,0.010 -0.002,0.009                      -0.016,0.037 -0.015,0.036 -0.019,0.026 -0.018,0.028 
    248 17.25      -0.003,0.016 -0.003,0.014 -0.002,0.010 -0.002,0.008                      -0.007,0.037 -0.007,0.040 -0.005,0.034 -0.004,0.029 
    249 17.75      -0.002,0.017 -0.001,0.015 -0.002,0.010 -0.002,0.008                       0.001,0.041 -0.008,0.044  0.002,0.034  0.004,0.035 
    250 18.25      -0.002,0.019 -0.002,0.016 -0.002,0.011 -0.002,0.009                      -0.003,0.044 -0.007,0.042  0.001,0.032  0.000,0.031 
    251 18.75      -0.002,0.021 -0.003,0.018 -0.001,0.012 -0.001,0.009                       0.000,0.050  0.001,0.047  0.002,0.036  0.002,0.035 
    252 19.25      -0.002,0.026 -0.002,0.021 -0.002,0.013 -0.002,0.010                       0.004,0.063  0.005,0.059  0.004,0.042  0.006,0.038 
    253 19.75      -0.002,0.035 -0.003,0.029 -0.001,0.015 -0.002,0.010                      -0.000,0.082 -0.000,0.076  0.001,0.050 -0.000,0.043 
    254 20.25      -0.002,0.050 -0.003,0.040 -0.002,0.021 -0.002,0.012                       0.003,0.112  0.011,0.094  0.005,0.060  0.003,0.050 
    255 20.75      -0.002,0.074 -0.003,0.060 -0.003,0.029 -0.003,0.017                       0.003,0.161  0.001,0.140  0.006,0.076  0.003,0.059 
    256 21.25      -0.001,0.113 -0.008,0.083 -0.005,0.042 -0.006,0.024                       0.005,0.234  0.003,0.197  0.009,0.101  0.007,0.070 
    257 21.75       0.017,0.173  0.007,0.141 -0.008,0.066 -0.007,0.037                      -0.008,0.337  0.004,0.291  0.015,0.159  0.017,0.093 
    258 22.25       0.056,0.274  0.043,0.225 -0.008,0.104 -0.012,0.060                      -0.044,0.466 -0.034,0.388  0.033,0.225  0.034,0.136 
    259 22.75       0.128,0.390  0.081,0.350 -0.000,0.174 -0.011,0.097                      -0.176,0.591 -0.096,0.525  0.064,0.364  0.064,0.215 
    260 23.25       0.249,0.467  0.178,0.422  0.035,0.254  0.005,0.155                      -0.403,0.672 -0.309,0.631 -0.019,0.490  0.095,0.352 
    261 23.75       0.470,0.495  0.361,0.478  0.138,0.338  0.043,0.240                      -0.722,0.681 -0.625,0.674 -0.275,0.610  0.026,0.499 
    262 24.25       0.780,0.493  0.632,0.549  0.372,0.372  0.145,0.257                      -1.058,0.654 -0.985,0.668 -0.641,0.621 -0.179,0.618 
    263 24.75       1.112,0.514  0.934,0.480  0.673,0.362  0.437,0.293                      -1.400,0.637 -1.169,0.675 -1.064,0.514 -0.597,0.743 
    264 25.25       1.448,0.525  1.299,0.411  0.963,0.503      ...                          -1.742,0.659 -1.615,0.879 -1.434,0.437 -1.155,0.668 
    265 }}} 
    266  
    267 == -30 < RA < -20 == 
    268 [[Image(ra_dec_full1.dat_1.png, 360px, align=left)]] 
    269 [[Image(ra_dec_full1.dat_2.png, 360px, align=top)]] 
    270  
    271 == -20 < RA < -10 == 
    272 [[Image(ra_dec_full2.dat_1.png, 360px, align=left)]] 
    273 [[Image(ra_dec_full2.dat_2.png, 360px, align=top)]] 
    274  
    275 == -10 < RA < 0 == 
    276 [[Image(ra_dec_full3.dat_1.png, 360px, align=left)]] 
    277 [[Image(ra_dec_full3.dat_2.png, 360px, align=top)]] 
    278  
    279 == 0 < RA < 10 == 
    280 [[Image(ra_dec_full4.dat_1.png, 360px, align=left)]] 
    281 [[Image(ra_dec_full4.dat_2.png, 360px, align=top)]] 
    282  
    283 == 10 < RA < 20 == 
    284 [[Image(ra_dec_full5.dat_1.png, 360px, align=left)]] 
    285 [[Image(ra_dec_full5.dat_2.png, 360px, align=top)]] 
    286  
    287 == 20 < RA < 30 == 
    288 [[Image(ra_dec_full6.dat_1.png, 360px, align=left)]] 
    289 [[Image(ra_dec_full6.dat_2.png, 360px, align=top)]] 
    290  
    291 == 30 < RA < 40 == 
    292 [[Image(ra_dec_full7.dat_1.png, 360px, align=left)]] 
    293 [[Image(ra_dec_full7.dat_2.png, 360px, align=top)]] 
    294  
    295 == 40 < RA < 50 == 
    296 [[Image(ra_dec_full8.dat_1.png, 360px, align=left)]] 
    297 [[Image(ra_dec_full8.dat_2.png, 360px, align=top)]] 
    298  
    299 == 50 < RA < 55 == 
    300 [[Image(ra_dec_full9.dat_1.png, 360px, align=left)]] 
    301 [[Image(ra_dec_full9.dat_2.png, 360px, align=top)]]