The ULLYSES team produces several types of High Level Science Products (HLSPs), described here. Products are made using both archival data and new HST observations obtained through the ULLYSES program. Data products are available from this website (HLSPs and contributing data), the MAST Data Discovery Portal (HLSPs and contributing data), or directly as a High-Level Science Product collection using the DOI (HLSPs only).

• Data Product Preparation
• Combining Spectra with a Common Grating
• Combining Spectra with Different Gratings and Instruments
• Creating Time-series Spectra
• LCOGT Data Processing
• FUSE Data Processing
• Latest Data Release
• Download Data
• Data Product Description
• Data Product Format
• Single-epoch Spectra
• Time-series Spectra
• WFC3 Drizzled Images
• YAML files
• Publications

Data Product Preparation

Spectra of ULLYSES targets were obtained with multiple instruments, multiple gratings, multiple settings of a grating central wavelength, and multiple telescopes. Some of the spectra are obtained with an echelle grating, some with single-order small or large aperture, and some with single-order long-slit configurations. The approach for combining data depends upon whether the input spectra share a common instrument and grating. Input COS spectra are obtained by running the instrument calibration pipeline, CalCOS. For COS, _x1d products are used, not _x1dsum products, as _x1dsum are created using a linear interpolation method that introduces noise correlation between neighboring output pixels. For all T Tauri stars, custom-calibrated _x1d products are created for STIS/G230L, G430L, and G750L observations. For STIS data of all other targets, default _x1d products are used. Virtual Observatory (VO) files are used as the input data for FUSE.

Combining Spectra with a Common Grating

This approach applies to 1) combining adjacent spectral orders within a single echelle exposure and 2) different exposures obtained with a common grating with the same or different central wavelength settings. Each input pixel measurement is treated as an estimate of the monochromatic flux at its assigned wavelength. The output flux is obtained by calculating a weighted average of all the flux measurements that fall within the output pixel's bounds. The throughput (net count rate divided by flux) times the exposure duration is used as the weighting factor for each input pixel, so that measurements derived from more counts have higher weights. Only input pixels with corresponding Data Quality (DQ) flags that are not considered serious contribute to the output flux. Figure 1 shows an example of how fluxes from two overlapping spectra are mapped to the imposed wavelength grid of the output spectrum.

The error array is calculated as the square root of the total counts that contribute to the output pixel, converting to flux units by multiplying by the flux/net counts at that wavelength. If the net counts and flux are zero, the conversion ratio is interpolated using neighboring values. The signal to noise ratio (SNR) is calculated for each wavelength bin as the ratio of the flux to the error.

This method of combination avoids correlating errors in neighboring pixels, at the cost of a very small loss in spectral sampling.

Combining Spectra with Different Gratings and Instruments

For all other cases spectra are spliced, meaning that:

• For non-overlapping spectra the arrays are abutted, and the spectral sampling will be discontinuous at the transition.
• For overlapping spectra a transition wavelength is selected. The input spectra are truncated and abutted at that wavelength, so the spectral sampling will be discontinuous at the transition.

Creating Time-series Spectra

Both HST and LCOGT data are used to create spectral time-series products. LCOGT data are used to create exposure-level time-series products only. LCOGT time-series products are created for both survey and monitoring stars. Images are taken approximately 90 and 10 days before, during, and 10 and 90 days after HST observations.

HST data are used to create both exposure-level and sub-exposure-level time-series products. HST time-series products are only created for monitoring targets. Monitoring T Tauri stars in the ULLYSES program are observed with HST COS/G160M/(1589,1623) and COS/G230L/(2635,2950). Targets are observed with each setting a total of 24 times. These observations are taken in two epochs, separated by approximately a year. There are four observations per rotation period, and three rotation periods in each epoch.

Input Data Calibration

For LCOGT time-series products, aperture photometry and flux calibration are performed on calibrated images. Full details on LCOGT data calibration are included below. Only exposure level products are created with LCOGT data.

For exposure-level time-series products using HST data, default x1d files from MAST are used as input to the HLSP creation code.

For sub-exposure time-series products using HST/COS data, default corrtag files from MAST are split into smaller time bins using the costools.splittag routine. This creates multiple “split corrtags”, each of which is then calibrated using CalCOS to produce individual “split x1d” products.

Time and Wavelength Sampling

LCOGT images are taken over four epochs, each consisting of daily observations over 10-day intervals. In addition, 15-minute cadence observations are also taken during the HST observations. For each daily observation, back-to-back images are taken in the applicable filters for the star type. Monitoring stars use SDSS u’ and i’ filters and the Bessel V filter, while survey stars only use the i’ and V filters. Unsuccessful observations will cause some gaps in either time or wavelength coverage. LCOGT “wavelength arrays” are limited to the central wavelengths of each filter.

Each HST visit is executed in the following order: a COS/G230L/2950 observation, COS/G230L/2635 observation, COS/G160M/1589 exposures at FP-POS 3 and 4, and COS/G160M/1623 exposures at FP-POS 1 and 2.

The time sampling for exposure-level time-series products is determined by the time of observation of each exposure. Native wavelength sampling is always used for exposure-level time-series spectra (i.e. data are not binned in wavelength).

The time and wavelength sampling for subexposure time-series spectra is optimized to probe the smallest time interval possible while maintaining a S/N ≥ 5 per resolution element at the peak of the most important emission lines: C IV (1548 Å), and Mg II (2800 Å). The final sampling is also chosen to ensure that products are ready for use out-of-the-box and require no additional binning for most science goals. However, if a higher S/N is required, data can be binned further in time or wavelength by the user.

For all NUV observations, the optimal time sampling and wavelength sampling were determined to be, respectively, 10s and the native wavelength grid (i.e. no wavelength binning, or ~0.415 Å). For FUV the following time and wavelength bins have been adopted for the four monitoring T Tauri stars:

• TW Hya: time = 30s, wavelength = 3 pixels
• RU Lup: time = 30s, wavelength = 6 pixels
• BP Tau: time = 30s, wavelength = 6 pixels
• GM Aur: time = 90s, wavelength = 6 pixels

For reference, the COS/FUV resolution element is 6 pixels in the dispersion direction.

Creation and Format of Time-series Spectra

The exact method of creating the time-series products differs when using LCOGT and HST data, but the data formats of the output products are identical.

For LCOGT data, an ASCII photometric file is used as input- this file includes the LCOGT filename, MJD start, MJD stop, central wavelength, flux, and error of each observation. For HST data, x1d files are used as input.

All input data are rebinned onto the same wavelength grid using the wavelength sampling appropriate for the product type. Data are then assembled into a 2-D array by inserting each rebinned spectrum into a row of a 2-D image, where the rows are ordered in time. The fluxes and errors are used to create 2-D arrays of flux vs. wavelength and vs. time. Separate 1-D arrays in the data extension give the wavelength as a function of the column number and the MJD start and stop times as a function of the row number.

Table 1: An illustration of the 2-D flux arrays computed for the spectral time-series products. The values represent the flux at each time and wavelength sample. Values of zero correspond to times or wavelengths where no data were obtained. In this exaggerated example, red/italic text could correspond to G230L observations while blue/normal text could correspond to G160M observations. Both gratings’ wavelength coverage overlap at wavelength index 1 (“Wavelength1”), which then has flux values at all time samples. An identically formatted table is made for the error measurements as well. The corresponding wavelength and time arrays are provided as separate 1-D arrays in the same data extension.

Time3	0	1.5	1.0	1.0
Time2	0	1.4	1.0	1.0
Time1	1.0	1.3	0	0
Time0	1.0	1.2	0	0
	Wavelength0	Wavelength1	Wavelength2	Wavelength3

LCOGT Data Processing

LCOGT images reduced with the BANZAI pipeline (McCully et al. 2018) are available in the LCOGT archive (see DDT2020B, DDT2021A). BANZAI performs bad-pixel masking, bias and dark removal, and flat-field correction. It also determines the astrometric solution and extracts a catalog of sources. Using the BANZAI-reduced images, an absolute flux calibration is determined based on magnitudes cataloged by the AAVSO Photometric All-Sky Survey (APASS, funded by the Robert Martin Ayers Sciences Fund and NSF AST-1412587).

The ULLYSES calibration pipeline is written in IDL. It first performs aperture photometry on sources in the BANZAI-generated catalog using aper.pro from the IDL Astronomy User's Library (Landsman 1993). It sums counts in a five-pixel radius and subtracts the modal signal in an annulus extending from 10 to 20 pixels. These sources are matched to sources in the APASS tables, using a matching radius of 2 arcsec. The relationship between APASS magnitudes and –2.5 times the logarithm of the LCOGT counts is fit with a line, ignoring three-sigma outliers. The slope of this line can vary to account for non-linearity in the instrument response; this term is typically very close to 1.

Next, the pipeline attempts to find a point source at the expected coordinates of the ULLYSES target. If a source is found, aperture photometry is performed using the same parameters described above. The measured counts are converted to a magnitude with the relationship determined above. Magnitude is then converted to a flux density using the zero-magnitude flux for the observed bandpass. Central wavelengths and zero-magnitude fluxes come from Bessell et al. 1998 for the Bessell V filter and from Fukugita et. al 1996 for the Sloan Digital Sky Survey (SDSS) i' filter. The uncertainty is assumed to be dominated by the uncertainty in the measured counts of the ULLYSES target. Measurements with uncertainties greater than 20% of the fluxes (S/N < 5) are discarded. Output photometry files are then used as input to the spectral time-series creation code.

For the u' photometry, there are some differences in the procedure due to the sparseness of the science fields at this wavelength. Separate calibration fields are observed close in time and airmass to the science fields. If BANZAI is unable to determine an astrometric solution for the calibration field, a 500 pixel box is searched for the brightest star around its expected location, and 200 pixel boxes are searched for additional calibration stars at their expected offsets from the brightest calibration star. Magnitudes for the calibration stars at u' are obtained from Version 2.4.2 of STScI's Guide Star Catalog. For the science field, each u' image is almost always followed by a V image with no intervening telescope motion, so the u' astrometry is copied from the subsequent V image. The flux vs. counts relationship found for the calibration field is applied to the science target after correction for the different exposure times and airmasses. The airmass correction, central wavelength, and zero-magnitude flux for the SDSS u' filter come from Fukugita et. al 1996.

FUSE Data Processing

All archival FUSE data used in the ULLYSES sample are examined and vetted by the ULLYSES team. Some targets exhibit various issues in their spectra, such as spectral channel drifting. In DR6, the ULLYSES team has begun to deliver improved spectra for such targets. Using the strategy outlined below, FUSE data for 23 targets previously excluded from the sample were able to be rectified and included in products. Even with extra processing, 4 targets were still unable to be rectified and will not be included in the ULLYSES sample:

1. PGMW3120: multiple stars in aperture
2. AV22: multiple stars in aperture
3. AV287: Target not in aperture
4. SK-69D220: multiple stars in aperture

Flux Differences Due to Guiding

Some ULLYSES FUSE data suffer from drifts among the spectral channels; FUSE was essentially four independent spectrographs, and thermal instabilities on orbit could cause each one to drift out of alignment. One of the four channels was used for guiding, and the flux in this channel was generally the most accurate. Usually, thermal drifts in the other channels led the target to drift out of the aperture, resulting in lower count rates and thus spectral fluxes. In crowded fields, these drifts could allow a neighboring star to drift into the aperture, resulting in higher fluxes in the final spectra. The following strategy was used to repair these data:

1. Begin by examining the NVO file, which was initially created by splicing together pieces of the extracted spectra from the eight FUSE detector segments.
• If the NVO file does not meet data quality needs (e.g., depressed flux or mismatching flux at channel transition points), a new NVO file is created by using the eight individual extracted spectra in the “ALL” files. These eight spectra are shifted to a common wavelength zero point and rescaled to create a new NVO file.
2. The guide channel is identified (LiF1A for the first half of the mission, and LiF2A for the second) and its spectrum adopted as a reference.
3. If the spectra from other channels are less than 50% brighter than the reference, then they are rescaled to match the reference in the region of overlap.
4. If they are more than 50% brighter than the reference, they are assumed to be contaminated by nearby stars and not included in the final spectrum.

Even with these corrections, some flux mismatches still remain at the transition point between FUSE and HST data- these are not corrected by the ULLYSES team. If smooth transitions are required, one of the contributing spectra may be manually scaled to the other.

Background Subtraction Corrections

Parts of the FUSE calibration pipeline (CalFUSE) were run only in cases where the background subtraction failed. FUSE did not have a shutter, so the detector received light from all three apertures (LWRS, MDRS, and HIRS) at all times. CalFUSE assumes that only the target aperture contains a star and fits a background model to the rest of the detector. In crowded fields, nearby stars occasionally fell in a non-target aperture, leading to an over-subtraction of the background. In these cases, the region of the detector used to model the background (stored as header keywords in the intermediate data file) was modified, and spectra were re-calibrated.

Latest Data Release

The latest data release notes can be found here.

Download Data

Data products for are available from this website (HLSPs and contributing data), the MAST Data Discovery Portal (HLSPs and contributing data), or directly as a High-Level Science Product collection using the DOI.

Data Product Description

The file names for ULLYSES science data products follow a naming scheme which encodes the target designation and the instruments and observing configuration(s) that contribute to the product. However, not all products will appear in the early releases. File names have the form:

hlsp_ullyses_<telescope>_<instrument>_<target>_<opt_elem>_<version>_<product-type>

where

• <target> is the target name
• <version> is the data release identifier (dr1, dr2, etc.)

The <telescope>, <instrument>, <opt_elem>, and <product-type> templates take names from the following table:

Description	Telescope	Instrument	Opt-Elem	Product-Type	HLSP Level
Custom calibrated STIS 1D spectra	hst	stis	g230l	spec.fits	0
			g430l
			g750l
STIS custom calibration parameter files	hst	stis	g230l	spec.yaml	0
			g430l
			g750l
STIS echelle single grating, where the orders have been extracted and merged. No level 1 products exist.	hst	stis	e140h	mspec.fits	1
			e230h
			e140m
			e230m
Combined spectra, with common instrument and grating, and in some cases with different cenwave settings.	hst	cos	g130m	cspec.fits	2
		cos	g160m
		cos	g185m
		cos	g230l
		stis	e140h
		stis	e140m
		stis	e230h
		stis	e230m
		stis	g230l
		stis	g430l
		stis	g750l
Combined spectra, with common instrument, different gratings and cenwave settings, and grouped by resolution^	fuse	fuv	lwrs or mdrs	aspec.fits^	3
	hst	cos	g130m-g160m-g185m
		stis	e140h-e230h
		stis	e140m-e230m
		stis	g230l-g430l-g750l
All instruments and settings abutted together*	hst	cos-stis	uv	preview-spec.fits*	4
		cos-stis	uv-opt
	hst-fuse	fuse-cos-stis	uv-opt
Exposure-level time-series spectra	hst	cos	g130m	tss.fits	5
			g160m
	lcogt	04m	v-ip
Subexposure-level time-series spectra	hst	cos	g130m	split-tss.fits	5
			g160m
WFC3 drizzled images	hst	wfc3	f225w	drc.fits	6
			f275w
			f336w
			f475w
			f814w

Data Product Format

Most High Level Science Products are in FITS format. The organization of each FITS file depends on the HLSP type. There are three broad categories of HLSPs: single-epoch spectra, time-series spectra, and drizzled images.

Single-epoch Spectra

FITS File Structure

Spectral data and information is stored in two BINTABLE extensions:

• a Science extension for the combined product, and
• a matching Provenance extension to record attributes of each spectrum that contributed to the combined product

Primary Header	Metadata common to all contributing spectra
Extension 1 Header	Metadata specific to science EXTNAME = 'SCIENCE'
Table 1 Data	Science data specific to single-epoch spectrum
Extension 2 Header	Metadata specific to provenance EXTNAME = 'PROVENANCE'
Table 2 Data	Metadata specific to contributing spectra

Single-epoch Science Table

Various elements of a single spectrum of M wavelength bins are stored in a single table row; each element is stored in a separate field (i.e., column). The table extension headers also contain informative metadata.

Column Name	Dimensions	Units	Data Type
WAVELENGTH	M	Angstrom	single-precision float
FLUX	M	erg/cm2/s/Angstrom	single-precision float
ERROR	M	erg/cm2/s/Angstrom	single-precision float
SNR	M	—	single-precision float
EFF_EXPTIME	M	s	single-precision float

Single-epoch Provenance Table

Select metadata for each spectrum that contributes to the combined spectrum in the SCIENCE extension will populate a row in the provenance table. The fields in the following table are metadata harvested from the headers of the contributing spectra.

Column Name	Units	Data Type
FILENAME	—	string
PROPOSID	—	string
TELESCOPE	—	string
INSTRUMENT	—	string
DETECTOR	—	string
DISPERSER	—	string
FILTER	—	string
CENWAVE	—	string
MINWAVE	Angstrom	double-precision float
MAXWAVE	Angstrom	double-precision float
APERTURE	—	string
SPECRES	—	double-precision float
CAL_VER	—	string
MJD_BEG	d	double-precision float
MJD_MID	d	double-precision float
MJD_END	d	double-precision float
XPOSURE	s	double-precision float

Time-series spectra

Time-series Spectra File Structure

Spectral data and information is stored in two BINTABLE extensions:

• a Science extension for the combined product, and
• a matching Provenance extension to record attributes of each spectrum that contributed to the combined product

Primary Header	Metadata common to all contributing spectra
Extension 1 Header	Metadata specific to science EXTNAME = 'SCIENCE'
Table 1 Data	Science data specific to multi-epoch spectra
Extension 2 Header	Metadata specific to provenance EXTNAME = 'PROVENANCE'
Table 2 Data	Metadata specific to contributing spectra

Time-series Spectra Science Table

The time-series spectral products have slightly different table columns compared to single-epoch spectra. The FLUX and ERROR arrays are 2D arrays with wavelength increasing X, and time increasing along Y. The wavelength values for each column of the 2D data are stored in the WAVELENGTH array, while the MJDSTART and MJDEND columns store the start and end times for each row of the FLUX and ERROR arrays.

Column Name	Dimensions	Units	Data Type
MJDSTART	M	d	double-precision float
MJDEND	M	d	double-precision float
WAVELENGTH	N	Angstrom	single-precision float
FLUX	M x N	erg/cm2/s/Angstrom	single-precision float
ERROR	M x N	erg/cm2/s/Angstrom	single-precision float

Time-series Spectra Provenance Table

Select metadata for each spectrum that contributes to the time-series product will populate a row in the provenance table. The fields in the following table are metadata harvested from the headers of the contributing spectra. Some columns are only present for HST data or LCOGT data.

Column Name	Units	Data Type
FILENAME	—	string
PROPOSID	—	string
TELESCOPE	—	string
INSTRUMENT	—	string
DETECTOR	—	string
DISPERSER (HST data only)	—	string
FILTER (LCOGT data only)	—	string
CENWAVE (HST data only)	—	string
MINWAVE (HST data only)	Angstrom	double-precision float
MAXWAVE (HST data only)	Angstrom	double-precision float
APERTURE	—	string
SPECRES (HST data only)	—	double-precision float
CAL_VER	—	string
MJD_BEG	d	double-precision float
MJD_MID	d	double-precision float
MJD_END	d	double-precision float
XPOSURE	s	double-precision float

WFC3 Drizzled Images

FITS File Structure

Drizzled and weight images are stored in two IMAGEHDU extensions, and PROVENANCE information stored in a BINTABLE extension:

• a Science extension for the drizzled science image
• a Weight extension for the weight map image
• a matching Provenance extension to record attributes of each image that contributed to the combined product

Primary Header	Metadata common to all contributing images
Extension 1 Header	Metadata specific to drizzled science images EXTNAME = 'SCI'
Image 1 Data	Drizzled science image
Extension 2 Header	Metadata specific to drizzled images EXTNAME = 'WHT'
Image 2 Data	Weight map image
Extension 3 Header	Metadata specific to provenance EXTNAME = 'PROVENANCE'
Table 3 Data	Metadata specific to contributing images

Drizzled Image Provenance Table

Select metadata for each image that contributes to the drizzled image in will populate a row in the provenance table. The fields in the following table are metadata harvested from the headers of the contributing images.

Column Name	Units	Data Type
FILENAME	—	string
PROPOSID	—	string
TELESCOPE	—	string
INSTRUMENT	—	string
DETECTOR	—	string
FILTER	—	string
APERTURE	—	string
CAL_VER	—	string
MJD_BEG	d	double-precision float
MJD_MID	d	double-precision float
MJD_END	d	double-precision float
XPOSURE	s	double-precision float

YAML Files

All STIS/G230L, G430L, and G750L data of T Tauri stars require tailored calibration. For these targets, all non-standard calibration parameters are recorded in YAML configuration files.

Publications

A description of the ULLYSES observations and data products is given in:

Ultraviolet Legacy Library of Young Stars as Essential Standards (ULLYSES): Data Release I, Roman-Duval, J. et al. 2020

For more information on how to cite ULLYSES data, see ULLYSES References.