Measure the tangential shear around a set of lenses. This technique is also known as cross-correlation shear, hence the name xshear
Runs of the code can easily be parallelized across sources as well as lenses.
Parallelizing across sources makes sense for cross-correlation shear, because the source catalog is generally much larger than the lens catalog. The source catalog can be split into as many small chunks as needed, and each can be processed on a different cpu across many machines.
Lenses can also be split up and the result simply concatenated.
# process the sources and lenses
cat source_file | xshear config_file lens_file > output_file
# you can parallelize by splitting up the sources.
cat sources1 | xshear config_file lens_file > output_file1
cat sources2 | xshear config_file lens_file > output_file2
# combine the lens sums from the different source catalogs using
# the redshear program (red for reduce: this is a map-reduce!).
cat output_file1 output_file2 | redshear config_file > output_file
# first apply a filter to a set of source files.
cat source_file | src_filter | xshear config_file lens_file > output_file
# The src_filter could be an awk command, etc.
cat source_file | awk '($6 > 0.2)' | xshear config_file lens_file > output_file
# xshear will output a list of source-lens pairs if you pass an additional
# filename to write this list to as a command line argument; use
cat source_cat | xshear config_file lens_file pair_logfile > output_file
# output will contain lens ID, source ID, radial bin ID, weight in mean shear,
# sigma_crit^-1 (sample), shear sensitivity in tangential direction, source redshift (sample)
# cosmology parameters
H0 = 100.0
omega_m = 0.30
# optional, nside for healpix, default 64
healpix_nside = 64
# masking style, for quadrant cuts. "none", "sdss", "equatorial"
mask_style = "equatorial"
# shear style
# "reduced": normal reduced shear shapes, source catalog rows are like
# ra dec g1 g2 weight ...
# "lensfit": for lensfit with sensitivities
# ra dec g1 g2 g1sens g2sens weight ...
# "metacal": for metacal with response tensor g1sens=R11, g2sens=R22, g12sens=(R12+R21)/2
# ra dec g1 g2 g1sens g2sens g12sens weight ...
shear_style = "reduced"
# source id style
# "none": no source id in first column (default)
# "index": integer source id in first column
# "somcell": integer source id in first column, integer SOMCELL id in second column
sourceid_style = "index"
# sigma crit style
# "point": point z for sources. Implies the last column in source cat is z
# "interp": Interpolate 1/sigma_crit calculated from full P(z)).
# Implies last N columns in source cat are 1/sigma_crit(zlens)_i
# "sample": random sample of p(z). Implies the second-to-last column in
# source cat is a mean-z estimate to be used for weighting and
# last column in source cat contains a random sample from the p(z).
sigmacrit_style = "point"
# units of shear, "shear", "deltasig" or "both". Defaults to "deltasig". For pipeline
# mode with variable shear styles, "both" is recommended (in which case output format is
# fixed and allows calculating all necessary quantities, see below).
shear_units = "deltasig"
# weight style
# only used in shear_units = "both"
#
# "uniform": calculate mean shear and mean sigma_crit^-1, weighting sources by the
# weight column in the input catalog only
# "optimal": weight sources by the product of weight colum in the input catalog and
# expectation value of sigma_crit^-1 for a minimum variance estimate (default)
#
# The output allows you to calculate (weighted) mean shear and DeltaSigma in both styles.
weight_style = "optimal"
# number of logarithmically spaced radial bins to use
nbin = 21
# index of innermost and outermost radius bin for which xshear should print pairs to logfile
# (optional, default is 0 for both which means no printing)
pairlog_rmin = 0
pairlog_rmax = 0
# maximum number of pairs logged in each radial bin
# first pairs in that bin are printed - make sure to shuffle your shape catalog
# if you need a representative sample; 0 (default) to print all pairs
pairlog_nmax = 0
# min and max radius (units default to Mpc, see below)
rmin = 0.02
rmax = 35.15
# units of radius (Mpc or arcmin). If not set defaults to Mpc; alternative is "arcmin"
r_units = "Mpc"
# demand zs > zl + zdiff_min
# optional, used for sigmacrit_style "point" and "sample" (on z_mean for the latter)
zdiff_min = 0.2
# note that if using weight_style "uniform" and sigma_crit_style "sample" the
#zdiff_min is not enforced in selecting source galaxies
sigmacrit_style = "interp"
# zlens values for the 1/sigma_crit(zlens) values tabulated for each source
# note the elements of arrays can be separated by either spaces or commas
zlvals = [0.02 0.035 0.05 0.065 0.08 0.095 0.11 0.125 0.14 0.155 0.17 0.185 0.2 0.215 0.23 0.245 0.26 0.275 0.29 0.305 0.32 0.335 0.35 0.365 0.38 0.395 0.41]
By default the code works in units of \Delta\Sigma (Msolar/pc^2) vs radius in Mpc. You can set the units of radius with "r_units" and the units for shear with "shear_units"
To measure tangential shear vs radius in arcminutes
r_units = "arcmin"
shear_units = "shear"
You can even measure \Delta\Sigma vs radius in arcminutes
r_units = "arcmin"
shear_units = "deltasig"
The most flexible output is generated in mode
shear_units = "both"
in which you can calculate mean shears, mean \Delta\Sigma, mean \Sigma_{crit}^{-1}, and mean shear responses, always from the same output column format (see below).
The format is white-space delimited ascii. The columns are
index ra dec z maskflags
For example:
0 239.5554049954774 27.24812220183897 0.09577668458223343 31
1 250.0985461117883 46.70283749181137 0.2325130850076675 31
2 197.8753117014942 -1.345204982754328 0.1821855753660202 7
...
The meaning of each column is
index: a user-defined index
ra: RA in degrees
dec: DEC in degrees
z: redshift
maskflags: flags for quadrant checking
The maskflags are only used if you set a mask style that is not "none" (no mask flags)
The format is white-space delimited ascii. The columns contained depend on the configuration.
When using point photozs (sigmacrit_style="point") the format is the following
For shear_style="reduced" (using simple reduced shear style)
ra dec g1 g2 source_weight z
For shear_style="lensfit" (lensfit style)
ra dec g1 g2 g1sens g2sens source_weight z
For shear_style="metacal" (metacal style)
ra dec g1 g2 g1sens=R11 g2sens=R22 g12sens=(R12+R21)/2 source_weight z
The format for sigmacrit_style="interp" includes the mean 1/sigma_crit in bins of lens redshift.
For shear_style="reduced" (using simple reduced shear style)
ra dec g1 g2 source_weight sc_1 sc_2 sc_3 sc_4 ...
For shear_style="lensfit" (lensfit style)
ra dec g1 g2 g1sens g2sens source_weight sc_1 sc_2 sc_3 sc_4 ...
For shear_style="metacal" (metacal style)
ra dec g1 g2 g1sens=R11 g2sens=R22 g12sens=(R12+R21)/2 source_weight sc_1 sc_2 sc_3 sc_4 ...
When using sigmacrit_style="sample", the source catalog needs to contain a mean-z estimate to be used for weighting and a random sample from the p(z).
For shear_style="reduced" (using simple reduced shear style)
ra dec g1 g2 source_weight z_mean z_sample
For shear_style="lensfit" (lensfit style)
ra dec g1 g2 g1sens g2sens source_weight z_mean z_sample
For shear_style="metacal" (metacal style)
ra dec g1 g2 g1sens=R11 g2sens=R22 g12sens=(R12+R21)/2 source_weight z_mean z_sample
You can pass catalogs with an integer source ID in the first column (with sourceid_style="index") or no ID (with sourceid_style="none", in which case the first column is expected to contain ra).
There is also the option to include a SOMPZ cell id, (with sourceid_style="somcell")
In sourceid_style = "index", add a first column such that source catalogs now look like
id ra dec g1 g2 ...
In sourceid_style = "somcell", add a first and second column such that source catalogs now look like
id somcell_id ra dec g1 g2 ...
Meaning of columns:
id: integer source id (optional)
ra: RA in degrees
dec: DEC in degrees
g1: shape component 1
g2: shape component 2
source_weight: a weight for the source
z: a point estimator (when sigmacrit_style="point")
sc_i: 1/sigma_crit in bins of lens redshift. The redshift bins
are defined in "zlvals" config parameter
z_mean: expectation value of source redshift, used for selecting/weighting the source
z_sample: random sample from source p(z), used for calculating sigma_crit^-1
The output catalog has a row for each lens. For shear_style="reduced", ordinary reduced shear style, and shear_units="deltasig" or "shear", each row looks like
index weight_tot totpairs npair_i rsum_i wsum_i dsum_i osum_i
For shear style="lensfit" or "metacal", and shear_units="deltasig" or "shear",
index weight_tot totpairs npair_i rsum_i wsum_i dsum_i osum_i dsensum_i osensum_i
where
index: index from lens catalog
weight_tot: sum of all weights for all source pairs in all radial bins
totpairs: total pairs used
npair_i: number of pairs in radial bin i. N columns.
rsum_i: sum of radius in radial bin i
wsum_i: sum of weights in radial bin i
dsum_i: sum of \Delta\Sigma_+ * weights in radial bin i.
osum_i: sum of \Delta\Sigma_x * weights in radial bin i.
dsensum_i: sum of weights times sensitivities
osensum_i: sum of weights times sensitivities
In shear_units="both", the output contains the same columns regardless of input format:
index weight_tot totpairs npair_i rsum_i wsum_i ssum_i dsum_i osum_i dsensum_w_i osensum_w_i dsensum_s_i osensum_s_i e1sum_s_i e2sum_s_i
where, in addition to the above,
rsum_i =Sum(w_i*(sigma_crit^-1)_i*r_i)
with w_i the weight defined by weight_style
in weight_style="optimal", this column and wsum, dsum, osum, dsensum, osensum are the same as above
wsum_i =Sum(w_i*(sigma_crit^-1)_i)
ssum_i =Sum(w_i)
to normalize to mean shear rather than mean \Delta\Sigma
dsum_i =Sum(w_i*(g_t)_i)
osum_i =Sum(w_i*(g_x)_i)
dsensum_w_i =Sum(w_i*(sigma_crit^-1)_i*(gsens_t)_i), where gsens_t is the sensitivity of the tangential component
osensum_w_i =Sum(w_i*(sigma_crit^-1)_i*(gsens_x)_i), where gsens_x is the sensitivity of the cross component
dsensum_s_i=Sum(w_i*(gsens_t)_i)
osensum_s_i=Sum(w_i*(gsens_x)_i)
In each case, _i means radial bin i, so there will be Nbins columns for each, ordered by radial bin.
Just divide the columns. For shear_units="deltasig" and shear_style="reduced",
r = rsum_i/wsum_i
\Delta\Sigma_+^i = dsum_i/wsum_i
\Delta\Sigma_x^i = osum_i/wsum_i
For shear_units="deltasig" and shear_style="lensfit" or "metacal"
\Delta\Sigma_+^i = dsum_i/dsensum_i
\Delta\Sigma_x^i = osum_i/osensum_i
In shear_units="both", you can calculate the following w-weighted quantities:
<r>_i = rsum_i/wsum_i
<\Sigma_{crit}^-1>_i = wsum_i/ssum_i
<e_t>_i = dsum_i/ssum_i
<e_x>_i = osum_i/ssum_i
<g_t>_i = dsum_i/dsensum_s_i
<g_x>_i = osum_i/osensum_s_i
<\Delta\Sigma>_i = dsum_i/dsensum_w_i
<\Delta\Sigma_x>_i = osum_i/osensum_w_i
Note that in shear_style = "reduced", these sens columns use gsens_i=1, so you can always calculate mean shears <g_t> and DeltaSigma as defined above.
# default build uses gcc
make
# use intel C compiler.
make CC=icc
make install prefix=/some/path
make uninstall prefix=/some/path
A C99 compliant compiler.