Discovery and time series analysis of
chemically peculiar (CP2 and CP4) stars
Klaus Bernhard
My observing site on a balcony in Linz, Austria
GSC 1419-0091 = Brh V132
11.20
11.40
11.60
11.80
12.00
12.20
-0.25 0.00 0.25 0.50 0.75 1.00 1.25
Phase
V
combination of own observations
and data mining in publicly
available surveys (ASAS)
- slowly rotating hot main-sequence stars (spectral classes early B to
early F) with peculiar surface compositions (10-15% of upper mainsequence
stars)
- peculiar composition is due to processes that took place after the star
formed, in particular the competition between radiative levitation and
gravitational settling in surface layers (Michaud 1970; Richer et al. 2000)
- most chemical elements tend to sink under the force of gravity; those
with numerous absorption lines near the local flux maximum are driven
outwards by radiative pressure
Introduction
What are chemically peculiar stars?
CP2 and CP4 stars: organized magnetic fields with non-uniform
distribution of chemical elements; surface spots of enhanced or depleted
element abundance; elements involved: silicon, strontium, chromium,
europium, and many other elements.
Most suitable for variable star astronomy!
Photometrically variable CP2/4 stars are traditionally
referred to as 𝛼2 Canum Venaticorum (ACV)
Variables, after their bright prototype.
Elements like europium: not very rare on earth;
used for example in computer screens…
Introduction
What do ACV light curves look like?
• non-uniform distribution of chemical elements: brighter and fainter
spots/areas; brightness varies in different spectral bands
• "Oblique rotator model" (Stibbs 1950):
photometric period = rotation period
• Surface looks quite different from
that of our sun; Figure: ESA
Surface temperatures and overabundant elements
From: A plethora of new, magnetic chemically peculiar stars
from LAMOST DR4
(S. Hümmerich, E. Paunzen, K. Bernhard, 2020)
Absorption lines of the respective element near the emission
maximum of the star
What do ACV light curves look like?
Addition of Gaussian scatter ('artificial error')
of 1, 5 and 10 mmag
Characteristics of ACV light curves:
- single wave, double wave or complex; peak-to-peak
amplitude in most cases <0.2 mag
- usually one single frequency, in rare cases additionally
pulsations
- chemical star spots much more stable than Sun-like
spots
- periods usually between 1 and 10 d, longer or shorter
periods possible: 0.5 d - 5 yr (and more!)
- multiband photometry exhibits interesting features:
varying light curve shapes in different passbands, even
antiphase variation is possible
- confirmation with spectroscopic methods is required;
the groups of CP2 and CP4 stars are by definition
classified via their spectroscopic characteristics (mostly
enhanced lines of certain elements in the blue-violet
spectral region; see for example Gray & Corbally, 2009)
Where to get data?
- own observations or data from various sky surveys;
a few examples:
ASAS-SN: Kepler:
20 telescopes in different places Satellite mission for exoplanets
Nikon telephoto f400/2.8 lenses, diameter 14 cm 0.95 m Schmidt telescope
ProLine PL230 CCD cameras 21 modules with two 2200x2048 CCDs
https://asas-sn.osu.edu/ https://archive.stsci.edu/kepler/
ASAS telescopes, Las Campanas,
Chile; source: ASAS homepage
The All Sky Automated Survey (ASAS, Poland)
- "low-cost project" for photometric monitoring of the southern sky and parts of the
northern sky (δ < +28°)
- data of the third phase (ASAS-3; 2000 – 2009) publicly available
- good photometry for 10^7 stars in the range 7 < V < 14 mag
- suitable also for the search for microvariables (variability amplitude: 4-5 mmag;
especially in the range 8 < V < 10 mag, for example Pigulski 2014)
- even known exoplanets have been confirmed
in ASAS data (Hümmerich & Bernhard, 2015)
SuperWASP (Instituto de Astrofísica de
Canarias (IAC) and six universities from
the United Kingdom)
- Observatorio del Roque de los Muchachos (La
Palma)
- South African Astronomical Observatory
(SAAO)
- stations equipped with eight f/1.8 200mm
Canon lenses and 2048 x 2048 Andor CCDs,
field of view 7.8°x7.8°
- ca. 18 million suitable light curves in the
range of 8 – 14 mag
Some deep surveys
ATLAS (Asteroid Terrestrial-impact
Last Alert System)
- 50 cm diameter f/2 Wright-Schmidt
telescope with 110 megapixel
camera
- field of view: 7.4°x7.4°; complete
sky visible from Hawaii two times per
night; objects brighter than 20 mag
http://mastweb.stsci.edu/mcasjobs/
ZTF (The Zwicky Transient Facility new timedomain
survey)
- first light at Palomar Observatory in 2017.
- 47 square degree, 600 megapixel cryogenic
CCD mosaic science camera; objects brighter
than 20.5 mag
Could also identify faint ACVs in the galactic halo
Results
• analysis of 142 million light curves with at least 100 data
points from the first 2 years of observation
• 4.7 million variable star candidates with light curves
• among them 427.000 clearly variable stars, identified and
classified with "machine learning"
• among them 214.000 specific variables (eclipsers etc.)
• 141.000 new discoveries
Most surveys provide an automatic analysis and classification of variable
stars. Why does it still make sense to search for ACVs in these data bases?
example: ATLAS
The mysterious 'upside-down CBH' variables
- close eclipsing binaries?
Location of proposed new type of variables
in the period vs. amplitude diagram
These "unknown variables" look exactly like
(and in all likelihood are!) ACV variables.
 Many surveys ignore the rather rare class of
ACV variables. They frequently end up mixed
with other groups (like RS CVn stars) or are
listed as MISC-type or 'unsolved' variables.
Example of an ACV
project: download of
light curves of
spectroscopically
confirmed ACVs
(Renson & Manfroid
2009; 5000 objects)
- removal of obvious outliers
- removal of data points with quality
flag of 'D' (='worst data, probably
useless’)
- removal of instrumental trends,
zero point issues between data sets
- workflow may be applied to
many surveys
How to download time series data:
ZTF data query in the NASA/IPAC Infrared Science Archive:
Obtain the available g-band lightcurves within 5 arcsec of a source position:
https://irsa.ipac.caltech.edu/cgi-bin/ZTF/nph_light_curves?POS=CIRCLE
298.0025 29.87147 0.0014&BANDNAME=g
In this way the light curves of your seminar were downloaded
OR use the private homepage of Dr. Chen:
http://variables.cn:88/lcz.php?SourceID=109576
Survey data sets are tricky, take care of HJD/MJD issues!
Survey data sets are tricky!
MJD Modified Julian Date:
MJD=HJD-2400000.5
not: MJD=HJD-2400000
The modified Julian Date (MJD) was
introduced by the Smithsonian
Astrophysical Observatory in 1957, when
storage space was still very expensive.
On the other hand: don't give up, if you come across weird
data.
On the contrary: you might possibly be on the brink of an
interesting discovery! Although these are rare, they do
happen.
As in the case of AR Sco:
A striking example of the productivity of collaborations
between amateur and professional astronomers.
Other troubles with survey data sets
- blending
- saturation
What are time series analysis?
• A time-series is a series of observations (or measurements, data) taken at
different times. We thus obtain a set of data pairs (t, x), where t is the time
and x is the observation (data value).
• We assume that t is error free, and that x is a combination of the true
signal plus some error.
• Time-series analysis is the application of mathematics to find periodic
signals in data. In doing so, we ultimately want to learn something about the
physics involved in the phenomenon.
• 1. Fourier methods: these methods attempt to represent a set of
observations with a series of trigonometric functions (sines and cosines, with
different periods, amplitudes and phases). Examples: Lomb-Scargle,
Bloomfield, Discrete Fourier Transform (DFT; Deeming) ...
• 2. Statistical methods: instead of fitting the observational data with
trigonometric functions, statistical methods compare points in the
observational data to other points at fixed time intervals or "lags" to see how
different they are from one another. These methods are very suitable for the
analysis of observational data that include non-sinusoidal periodic
components. Examples: ANOVA (observations fit with periodic orthogonal
polynomials), PDM, Lafler-Kinman ...
Time series analysis of an improved data set
Many programs available: Period04, (University of Vienna, Austria), Peranso
(Vanmunster & Paunzen), vartools (University of Princeton, USA)
Period04 (Lenz & Breger 2005)
Sinusfit with fourier analysis
Period04: example workflow
Attention needs to be paid to the
presence of strong "daily alias" peaks
in the Fourier diagrams.
- aliases: dependant on sample
frequency
- daily aliases: 1-f; f+1, (1-f)+1 etc.
- sometimes alias frequencies can even
dominate over the true frequencies
Result of period analysis with Period04 (Lenz & Breger 2005)
Peranso (Vanmunster & Paunzen)
Peranso
Use Mouse to Zoom in/out
and to activate/deactivate
observations.
Peranso
Which method for which data set?
• Delta Cepheids and RR Lyrae variables can generally be well analyzed with
the Lafler-Kinman method.
• If you expect a variable to be multi-periodic, use CLEANest.
• If the light curve is highly non-sinusoidal, use ANOVA.
• PDM is also well suited for analyzing highly non-sinusoidal light curves
consisting of only a few observations procured over a limited period of time.
- Consult the manual for special purposes!
Peranso (Vanmunster&Paunzen)
finally you hopefully get something like this:
ZTF Survey:
Basis of the new ACVs of your seminar
The Zwicky Transient Facility (ZTF) new
time-domain survey
- first light at Palomar Observatory in 2017
- 47-square-degree, 600 megapixel
cryogenic CCD mosaic science camera
- objects brighter than 20.5 mag
- 3 band photometry in g, r and (partly) i
Filter transmission and mosaic CCDs
The ZTF Catalog of Periodic Variable Stars by Chen et al. (2020)
ZTF time series analysis
Fourth order Fourier function
amplitudes and phases for each order
represents how well LCs are fit by the
Fourier function
Selection criteria for variability types in Chen et al. (2020)
There is no type "ACV"!
Results
ACV variables may be included in the category 'RS CVn variables' (objects with starspots).
A few examples of interesting light curves
P = 0.4362979 d
g band amplitude larger than r band amplitude: pulsator!
Folded light curve with large scatter
P = 0.080869 d
rotational variable with large amplitude?
P = 0.364225 d
It is the central star of the planetary nebula Ou 5:
We come back to the ~81.000 RS CVn stars
and want to search ACVs within this group:
Assuming that 30 seconds are needed for the
visual inspection of a light curve, an inspection of
all stars within this group would take
81.000*30/60/60 = 675 hours!
Possible but impractical...
Therefore we need further constraints:
Gaia temperature >6000 K (ACVs are hot!)
variability amplitude <0.3 mag (typical for ACVs)
...
~1400 objects
time for visual inspection:
1400*30/60/60 ~ 12 hours: this is fine!
Homepage of Xiaodian Chen:
Visualization of the light curves
Every second counts…
Examples of objects that were sorted out:
unspecific light curve
Examples of objects that were sorted out:
unspecific light curve-similar object
Examples of objects that were sorted out:
likely Gamma Doradus pulsating variable
Only 2 types of objects were kept, which are almost certainly ACVs:
r band amplitude > g band amplitude
Only 2 types of objects were kept, which are almost certainly ACVs:
antiphase variations are present
Result: 87 objects, light curves made with MS EXCEL
examples with amplitude r > g
Result: 87 objects, light curves made with MS EXCEL
Examples with antiphase variation
Depending on the investigated passbands, ACV stars may show considerably
different light curve shapes:
Have fun with the new ACV
objects!
Thank you very much and all
the best for your seminar.
Thanks also to my colleague Stefan Hümmerich for
his help in preparing this talk!