Astrometry of OH/IR stars using 1612 MHz hydroxyl masers. I. Annual parallaxes of WX Psc and OH138.0+7.2

0.5 MB·

by  G. Orosz

Astrometry of OH/IR stars using 1612 MHz hydroxyl masers.
I. Annual parallaxes of WX Psc and OH138.0+7.2
7 G. Orosz1, H. Imai1, R. Dodson2, M. J. Rioja2,3,4, S. Frey5, R. A. Burns1,6, S. Etoka7,
1
A. Nakagawa1, H. Nakanishi1,8,9, Y. Asaki10,11, S. R. Goldman12, and D. Tafoya13
0
2 1Department of Physics and Astronomy, Graduate School of Science and Engineering, Kagoshima
n University, 1–21–35 Korimoto, Kagoshima 890–0065, Japan
a 2International Centre for Radio Astronomy Research, M468, The University of Western Australia, 35
J
Stirling Hwy, Crawley, Western Australia 6009, Australia

8 3CSIRO Astronomy and Space Science, 26 Dick Perry Ave, Kensington, Western Australia 6151, Australia
1
4Observatorio Astron´omico Nacional (IGN), Alfonso XII, 3 y 5, E-28014 Madrid, Spain

] 5Konkoly Observatory, MTA Research Centre for Astronomy and Earth Sciences, Konkoly Thege Mikl´os u´t
R
15–17, 1121 Budapest
S 6Joint Institute for VLBI ERIC, Postbus 2, 7990 AA Dwingeloo, The Netherlands
h. 7Hamburger Sternwarte, Gojenbergsweg 112, 21029 Hamburg, Germany
p 8Institute of Space and Astronautical Science, Japan Aerospace Exploring Agency, 3–1–1 Yoshinodai,
-
Sagamihara, Kanagawa 252–5210, Japan
o
r 9SKA Organization, Jodrell Bank Observatory, Lower Withington, Macclesfield, Cheshire SK11 9DL, UK
t
s 10Chile Observatory, National Astronomical Observatory of Japan, National Institute of Natural Science,
a
Joaquin Montero 3000 Oficina 702, Vitacura, Santiago, C.P.7630409, Chile
[
11Joint ALMA Observatory, Alonso de Cordova 3107, Vitacura 763 0355, Santiago, Chile

1 12Astrophysics Group, Lennard-Jones Laboratories, Keele University, Keele, Staffordshire ST5 5BG, UK
v
13Chalmers University of Technology, Onsala Space Observatory SE–439 92 Onsala, Sweden
1
0 [email protected]
1 Accepted to The Astronomical Journal (January 17, 2017)
5
0
.
1
0
7 ABSTRACT
1
: We reporton the measurementofthe trigonometric parallaxesof 1612MHz hydroxylmasers
v
around two asymptotic giant branch stars, WX Psc and OH 138.0+7.2, using the NRAO Very
i
X LongBaselineArraywithin-beamphasereferencingcalibration. Weobtaineda3σupperlimitof
r 5.3masontheparallaxofWXPsc,correspondingtoalowerlimitdistanceestimateof&190pc.
a T≤he obtained parallax of OH 138.0+7.2is 0.52 0.09 mas ( 18%), corresponding to a distance
of 1.9 +0.4 kpc, making this the first hydroxy±l maser para±llax below one milliarcsecond. We
−0.3
also introduce a new method of error analysis for detecting systematic errors in the astrometry.
Finally, we compare our trigonometric distances to published phase-lag distances toward these
stars and find a good agreement between the two methods.
Subject headings: masers — astrometry — techniques: interferometric — stars: AGB and post-AGB —
stars: individual(WX Psc, OH 138.0+7.2)
1. Introduction surements of maser (and also radio continuum)
sources. Masersareexcellenttracersofvariousen-
Relativeastrometryusingverylongbaselinein-
vironments related to the young and evolvedstel-
terferometry (VLBI) has proven to be very suc-
lar populations in the Milky Way Galaxy: high-
cessfulforconductingtrigonometricparallaxmea-
1
mass star forming regions (HMSFRs), asymptotic over decadal timescales (e.g. Etoka & Le Squeren
giant branch stars (AGBs) or their massive coun- 2000).
terparts, red supergiants (RSGs). CH3OH or However, VLBI astrometric observations at
H2O masers at the relatively high radio frequen- low frequencies (especially below approximately
cies of 12 and 22 GHz respectively are excellent 2 GHz, i.e. L-band) are challenging to calibrate
forhigh-precisionparallaxmeasurementsofHMS- accurately due to the dominant ionospheric error
FRs at a 10 µas-level. This astrometric pre- contributions with typical residual path length
∼
cision is achieved by using techniques that en- errors of hundreds of centimeters at 1.6 GHz,
able us to calibrate errors due to atmospheric compared to only a few centimeters from the
effects. The geodetic blocks in VLBI observa- troposphere. These dispersive terms are slowly
tions(Reid et al.2009)andthedual-beamsystem changing spatial irregularities of plasma density
(Honma et al. 2008) introduced in the Japanese in the atmosphere that cause serious direction
VLBI Exploration of Radio Astrometry (VERA) dependent errors in astrometry. In turn, they
are examples of such calibration techniques, both drastically degrade the accuracy of conventional
designed to address the dominant error sources phase-referencing techniques by introducing sys-
above 10 GHz: the static (temporally stable tematic astrometric offsets into the observations.
∼
and systematic) and dynamic (short-term turbu-
Despite the challenges, there have been a few
lent and random) terms of the non-dispersive ex-
results of annual parallax measurements using
cess path delays caused by the troposphere, re-
OH masers. van Langevelde et al. (2000) and
spectively.
Vlemmings et al. (2003) used the two main-line
Thanks to these calibration techniques, there OH masers at 1665 and 1667 MHz with mostly
have been a series of successful parallax mea- a conventional source-switching phase referencing
surements of AGB stars using mostly circumstel- strategy to measure parallaxes at a 1 mas-level
lar H2O masers (see Nakagawa et al. 2016, and precision. Vlemmings & van Lange∼velde (2007)
references therein), with also a few results from refined and continued these measurements, using
SiO masers at 43 GHz (e.g. Min et al. 2014). in-beam phase referencing – i.e. simultaneously
H2O masers have also been used to measure dis- observing the maser with a reference calibrator
tancestoacoupleofpost-AGBstars(Tafoya et al. whichlieswithinthesameantennabeam–topush
2011; Imai et al. 2011, 2013b) and RSGs (e.g. the astrometric precision firmly into the sub-mas
Asaki et al. 2010). However, H2O masers are nei- regime. As we will also discuss in this paper, the
therthestrongestnorthemoststableofthestellar problems encountered at low frequencies can be
masersformeasuringtheparallaxesofAGBstars. mitigated by substantially decreasing the target–
Instead, the strongest and most commonly found calibrator separations, from a few degrees down
ones are the 1.6 GHz ground state OH masers, to a few tens of arcminutes.
withthousandsofknownsourcesintheMilkyWay
Looking beyond spectral line VLBI, we see
Galaxy (Engels & Bunzel 2015a).
that L-band pulsar astrometry has flourished in
In the case of OH/IR stars – AGB stars heav- the past decade, as those observations are not
ily enshrouded in optically thick circumstellar en- hindered by resolved spatial and velocity struc-
velopes (CSEs) – OH masers are suitable for as- tures like OH masers. Since pulsars are contin-
trometry, especially the strongest line at 1612 uum sources, wider bands, higher recording rates
MHz(e.g.Herman & Habing1985). Thesemasers and pulsar gating – recording signals only dur-
arelocatedintheCSEatadistanceofseveralhun- ing on-pulse periods – can be employed to sig-
dred stellar radii from the central host star, ex- nificantly increase signal-to-noise ratios and, as
panding outwardat terminal velocity. OH masers a result, reduce (random noise-induced) astro-
arepumpedbyinfraredcontinuumbackgroundra- metric errors. For the systematic errors, so far,
diation to which stellar photons are converted by there have been two main ionospheric calibra-
a heavy dust shell. As a result, 1612 MHz masers tion strategies for L-band continuum astrometry:
are excellent tracers of OH/IR stars: they are measure and remove the dispersive component
saturated, radially amplified and located in rel- of the ionospheric delays by using wide-spread
atively calm regions, with strong features stable bands to detect its frequency-dependent curva-
2
ture (Brisken et al. 2000, 2002); or use in-beam 2. Observations
astrometry to almost completely remove dynamic
We observed two OH/IR stars, WX Psc (also
(random) and mitigate static (systematic) error
known as IRC +10011 or OH 128.6 50.1) and
terms (Chatterjee et al. 2001, 2009). Deller et al.
−
OH 138.0+7.2 (hereafter abbreviated to OH138)
(2013, 2016) demonstrated that by using in-beam
with the NRAO1 Very Long Baseline Array
calibrators with careful scheduling and data re-
(VLBA). Table 1 lists basic details on the tar-
duction, it is possible to measure parallaxes at a
get maser sources and calibrators used. The
10 µas-level precision despite the low frequency.
∼ target sources were selected from the “Nanc¸ay
An important side note is that ionospheric er-
1612 MHz monitoring of OH/IR stars” project2
rors are not only a problem for L-band astrom-
(Engels et al. 2015b), based on having calibrators
etry, but can dominate the error budget even at
inthe sameVLBAbeamwith preciseandreliable
5GHz(seee.g.Krishnan et al.2015;Kirsten et al.
positions3,i.e. within0◦.7oftheOHmasersources
2015). In addition, in-beam astrometry can be
in our present case. Also, only double-peakedOH
limited by the availability of suitable calibrators,
sources were considered, which already had their
and by residual systematic errors in the mea-
periods and phase-lag distances measured.
surements due to non-zero source separations.
Each source was observed at four epochs over
We are therefore working on developing an al-
a period of one year. The details of the observing
ternative multi-calibrator approach (Rioja et al.
sessions are given in Table 2, whereas the source
2009; Dodson et al. 2013) that can fully remove
scan pattern on the sky is shown in Fig. 1. The
even the effects of the static ionosphere and make
sessions were scheduled near the peaks of the si-
our astrometry completely free of systematic er-
nusoidal parallax signatures to maximize the sen-
rors. Thistechnique,knownasMultiView,willbe
sitivity for parallax detection and ensure that we
demonstratedinaforthcomingpaper(Rioja et al.
can separate the linear proper motions from the
2017); while our current focus centers on explor-
parallactic modulations. The desired sampling
ing the limits of in-beammaser astrometryat low
wasonlypartiallyachievedduetofailedobserving
frequencies.
epochsandothertechnicaldifficulties. Allourob-
In this paper, we report on the VLBI obser-
servations are publicly available from the NRAO
vations of 1612 MHz OH masers to measure the
VLBA archive under project codes BO047A and
trigonometric parallaxes and proper motions of
BO047B.
two long-period variable OH/IR stars, WX Psc
Each observing session was 4h long to ensure
and OH 138.0+7.2. Their respective pulsation
sufficient (u,v) coverage and sensitivity, spend-
periods are 650 and 1410 days (Engels et al.
ing 70% of the time on the in-beam calibrator
2015b), and both exhibit high mass loss rates of
∼
10−5 M yr−1 (calculated based on the method (Cib) and OH maser pairs with intermittent ob-
⊙
∼ servations of fringe finders (FF) in every 2 hours
in Goldman et al. 2017). We describe the VLBI
and bright secondary calibrators (C ) in every
observations in Sect. 2, then the flow of data re- 1
15 min. For the target scans, antennas were
duction and maser detections in Sect. 3. A com-
pointed halfway between the OH maser and C
prehensiveastrometricerroranalysisis conducted ib
positions, except for Epochs I–II of the WX Psc
inSect.4,describingthe differenterrortermsand
sessions, where the pointing center was the OH
theireffectsonourmeasurements. Wealsopresent
maser source. The 2-bit quantized signals were
a new method to try to identify systematic errors
recorded on Mark5C units in dual circular po-
and estimate the uncertainties in our astromet-
larization with 128 Mbps using 4 intermediate
ric measurements. The astrometric results and
frequency (IF) bands, each with a bandwidth of
parallaxes are presented in Sect. 5, along with a
comparison to distances derived by the phase-lag
technique (e.g. van Langevelde et al. 1990). Sec- 1TheNationalRadioAstronomyObservatoryisafacilityof
the National Science Foundation operated under coopera-
tion6givesasummaryandoutlooktoafollow-up
tiveagreementbyAssociatedUniversities,Inc.
paper that will introduce theoretical topics at- 2Projecthome: www.hs.uni-hamburg.de/nrt-monitoring
tainable with our present and future OH maser 3Calibrators with status “C” in the Astrogeo Catalog (as-
parallaxes. trogeo.org).
3
Table 1: List of observed sources.
Target Period V Calibrator ID Right ascension Declination Sep. S
LSR 1.6GHz
(days) (km s−1) (J2000) ( h m s ) ( ◦ ′ ′′) (◦) (mJy beam−1)
WX Psc 650 8.9 0.1 J0106+1300 C 01 06 33.35651 13 00 02.6039 0.40 70
ib
±
J0121+1149 C /FF 01 21 41.59504 11 49 50.4130 3.81 2100
1
OH138.0+7.2 1410 37.7 0.1 J0322+6610 C 03 22 27.22883 66 10 28.3005 0.70 750
ib
− ±
J0257+6556 C 02 57 01.34302 65 56 35.4270 2.92 270
1
J0102+5824 FF 01 02 45.76238 58 24 11.1366 18.74 1300
Notes. For the stellarcoordinatesreferto Sect.3, while forthe preciseOHmaseroffsets to Table4. Stellar
pulsation periods are from the “Nanc¸ay 1612 MHz monitoring of OH/IR stars” project. Calibrators were
selectedusing the AstrogeoCatalogandtheir coordinatesareaccurateto.0.3mas. Systemicvelocitiesand
S flux densities are measured using the VLBA observations. Calibrator IDs refer to Fig. 1 and go as
1.6GHz
follows: C = in-beam; C =secondary;FF=fringe finder.
ib 1
Table 2: Summary of observing sessions conducted with the VLBA.
Target Epoch Date (DOY) MJD UT range Project code Flagged data Remarks
(days) (day/hh:mm)
WX Psc I 2014 Aug 01 (213) 56870 0/09:32 0/13:33 BO047A6 LA no FD
−
II 2015 Feb 17 (048) 57070 0/20:22 1/00:22 BO047A3 MK, SC
−
III 2015 Jun 08 (159) 57181 0/12:06 0/16:06 BO047A7 MK, SC
−
IV 2015 Jul 07 (188) 57210 0/11:12 0/15:12 BO047A4 MK, SC
−
OH138.0+7.2 I 2014 Feb 16 (047) 56704 0/22:25 1/02:26 BO047B1 MK
−
II 2014 May 07 (127) 56784 0/18:40 0/22:40 BO047B2 KP, PT no NL
−
III 2014 Aug 07 (219) 56876 0/11:08 0/15:08 BO047B3 KP
−
IV 2015 Feb 22 (053) 57075 0/22:02 1/02:02 BO047B4 SC
−
Notes. Flags refer to data from antennas not used in the astrometry, which were identified using methods
describedinSect.4.2. IntheNRAOarchiveWXPschasseveralfailedepochsduetorecordingorcorrelation
problems: BO047A1, A5, & A7.
VLBA stations. BR=Brewster, WA; FD=Fort Davis, TX; HN=Hancock, NH; KP=Kitt Peak, AZ;
LA=Los Alamos, NM; MK=Mauna Kea, HI; NL=North Liberty, IA; OV=Owens Valley, CA; PT=Pie
Town, NM; SC=Saint Croix, VI
4
1
J0322+6610 (C )
ib
J0106+1300 (Cib) (∆OH138 ≈ 0.70°) J0257+6556 (C1)
(∆WX Psc ≈ 0.40°) (∆OH138 ≈ 2.92°)
)
g 0
e
d
(
δ
∆ WX Psc
OH138.0+7.2 (OH138)
-1
J0121+1149 (C )
1
(∆ ≈ 3.81°)
WX Psc
4 3 2 1 0 -1 -2 -3 -4
∆α cosδ (deg)
Fig. 1.—Observingsetupoftheastrometricmeasurementsforbothtargetsplottedtogether. Solidandopen
symbolsrepresentthesourcesassociatedwiththetwodifferentmonitoringcampaignsofWXPscandOH138,
respectively. The target masers are marked as circles and the calibrators as squares. Angles in parentheses
show the separationbetween the respectivetargets andcalibrators. The twoconcentric circlesrepresentthe
half-powerbeamwidth andfull beamwidth ofthe VLBA.Dashed arrowsdenote sourceswitching (with 5–15
min cycles). Coordinates are relative to the target pointing centers, different for each campaign.
5
4 MHz. The IFs were spread out over 300 MHz, before phase referencing.
centered around the four ground-state OH maser We flagged channels contaminated by radio-
lines of 1612, 1665, 1667, 1720 MHz and the HI frequencyinterference,thenperformedinstrumen-
line at1420MHz. Eachbandhada channelspac- tal phase calibration using single FF calibrator
ing of 1.95 kHz, corresponding to a velocity reso- scans with the phase rates zeroed. Next, scans
lution of 0.36 km s−1. Note that for the last two on C were used to determine the group delays,
1
epochs of WX Psc (BO047A7, & A4) we used a phase rates and bandpass characteristics. Fringe-
slightly modified observing setup for our parallel fitting was performed by averaging polarizations,
investigations of high-precision astrometry at L- as Stokes V values were always less than 10%
band (described in Rioja et al. 2017, but not re- of Stokes I, making the difference between ∼polar-
lated to this work). The on-source time on the izations negligible (see Fig. 2). After applying
target was decreased to 40%; the switching cycle this calibration to the Maser–C pair, the final
ib
to C1 (and two other calibrators) was decreased phase calibration solutions were obtained using
to 5minandthe IFbandwidths wereincreasedto C (with IFs and polarizations averaged) and
ib
8MHz(256Mbpsrecordingrate),butkeepingthe transferred to all channels in the maser scans.
same spectral resolution. The changed setup did Thus WX Psc and OH138 were effectively phase
not affect our in-beam parallax measurements. referenced to their respective in-beam calibra-
Since the masers were always observed si- tors. Finally, phases were rotated to shift the
multaneously with the C phase reference cal- phase-tracking center to the vicinity of the maser
ib
ibrators in the same primary beam, all bands emission before imaging. The shifted J2000.0
were correlated with VLBA-DiFX (Deller et al. phase-tracking centers used for astrometry are:
2007, 2011) in a single run using two phase cen- (α,δ)shifted =(01h06m26s.02574,+12◦35′52′.′8242)
WXPsc
ters set to the maser and C calibrator posi- and(α,δ)shifted=(03h25m08s.42975,+65◦32′07′.′0900).
ib OH138
tions. Phase tracking centers for the calibrators Thoughourin-beamphasereferencingobserva-
were set to the coordinates described in Table 1, tions were affected by the primary beam attenua-
whereas for the OH maser targets the following tion of the antennas, a correction scheme was not
a-priori J2000.0 positions were used during corre- applied in the present observations. As a result,
lation: (α,δ)apriori =(01h06m25s.98,+12◦35′53′.′0) thefluxdensityscalesmentionedinthispaperare
WXPsc
and(α,δ)apriori=(03h25m08s.80,+65◦32′07′.′0). For systematically lower than the true values. Using
OH138
details on the observations and correlatoroutput, anAirydisk modelpresentedinMiddelberg et al.
we refer to the VLBA File Server4. (2013) with an antenna diameter of D=25.47m,
we estimate that the amplitudes of WX Psc and
3. Data Reduction and Maser Detections J0106+1300 are lower by 47%, and those of
∼
OH138 and J0322+6610are lower by 90%. Un-
The data analysis was carried out using the ∼
calibrated amplitudes do not seriously affect our
NRAO Astronomical Image Processing System
astrometricmeasurements,becausethephasepat-
(AIPS)packagewithanin-beamphasereferencing
tern in the primary beam is expected not to have
strategy. Flux density calibration was performed
a significant systematic offset and also the re-
using system temperatures and gain information
sulting instrumental error is believed to be con-
recorded at each station. The Earth orientation
stant throughout our observing sessions. How-
parameters from the VLBA correlator were re-
ever, in order to minimize possible error sources
fined by the U.S. Naval Observatory final solu-
and achieve better signal-to-noise ratios, proper
tions. Initial ionospheric delay corrections were
flux density corrections should be applied in fu-
performed using the NASA Jet Propulsion Lab-
turein-beamastrometricobservationsbyadopting
oratory IONEX files, which contain zenith total
a suitable primary beam model and beam squint
electron content (TEC) maps derived from global
corrections as done in Middelberg et al. (2013).
navigation satellite system observations. Finally,
We observed all four ground-state OH maser
phases were corrected for parallactic angle effects
transitions, but could only detect the double-
peaked 1612 MHz satellite lines for both OH/IR
4VLBAFileServer: www.vlba.nrao.edu/astro/VOBS/astronomy
stars. The full resolution VLBA spectra of the
6
1612 MHz OH masers are shown in Fig. 2. In ligible; less than 1◦ per baseline.
both profiles, the blueshifted features relative to The biggest residual errors in L-band are re-
the stellar systemic velocities have peak flux den- lated to spatial static terms, direction dependent
sities of 0.2 Jy, compared to 20 Jy from sin- systematicerrorsfromtheinadequatemodelingof
∼ ∼
gle dish observations with, e.g, the Nan¸cay Radio the ionospheric sky-plane TEC distribution. This
Telescope (NRT), meaning we could only recover means that reducing the target–calibrator sepa-
1% of the total maser emission from the OH re- ration is of utmost importance in mitigating at-
∼
gions. The redshifted features have a similar flux mospheric errors in low-frequency VLBI astrom-
density recovery percentage for WX Psc, but are etry. Also, even with in-beam phase referencing
almostcompletelymissingforOH138,perhapsdue the dynamic terms from ionospheric phase fluc-
tomoreseriousforegroundscatteringfromthecir- tuations are not zero. This is because the trav-
cumstellar envelope. However, in all cases maser eling waves causing the temporal disturbances in
emissiononbaselineslongerthan 4000kmarere- the ionosphere have spatial scales of hundreds of
∼
solvedoutsignificantlyforboththeblue-andred- kilometers, which again reflect as residual errors
shifted regions, similar to that seen in Imai et al. due to the non-zero target–calibratorseparations.
(2013a). Assuming these components are independent and
adding them in quadrature, we estimate the to-
4. Astrometric Error Analysis tal atmospheric phase errors per baseline to be
16◦ / 29◦ for WX Psc / OH138. These φ
4.1. Theoretical Predictions atmo
phaseerrorsperbaselinecanberoughlyconverted
The dominant error sources after our in-beam to σatmo astrometric errors in the VLBA maps as
phase referencing are composed of various com- σatmo ≈ (cid:0)φ[a◦t]mo/360◦(cid:1) · (cid:0)θ/√N(cid:1), where θ is the
ponents of uncompensated atmospheric terms, size of the synthesized beam and N is the num-
and additional contributions from source/velocity berofstationsinthearray. UsingN=8andbeam
structures and thermal noise (Reid & Honma sizes described in Sect. 5.1, we estimate the total
2014). Since the in-beam calibrators are observed atmosphericerrorsinourastrometrytobeapprox-
simultaneously in the same beam with the maser imately 0.3 mas / 0.4 mas for WX Psc / OH138,
targets,thederivedphasesolutionscanbedirectly dominated by the effects of the static ionosphere.
applied to the masers and do not have to be in- In bad ionospheric conditions (i.e. having larger
terpolated in time between the calibrator scans. residual TEC values), deviations even as large as
This mitigates the effects of temporal phase fluc- 0.7 mas can be expected.
∼
tuations, a dynamic term causing random errors Contribution from the target source structure
in the astrometry. The small target–calibrator to systematic astrometric errors is hard to pre-
separations also reduce excess path errors. dict, due to the variable behavior in the spatial
Given our observing parameters (1.6 GHz, 0 andvelocitystructuresofmasers,andthepossibil-
min switching time due to simultaneous observa- ityofmultiplemaserspotsblendingtogether. Our
tions, 0◦.4 0◦.7 target–calibrator separations) we OHmaserspotsusedfortheastrometricmeasure-
−
can estimate the expected errors from the static ments seem to show complex spatial structures
and dynamic components of the troposphere and at lower resolutions when using a (u,v) taper of
ionosphere(Asaki et al.2007),byassumingatyp- 13 Mλ (corresponding to baselines of 2000 km).
∼
icalzenithpatherrorof3 cm(Reid et al.1999),a However,when using the full resolution of our ar-
6 TECU uncertainty5 in the adopted ionospheric ray,weonlydetectcompact,albeitnotcompletely
maps (Ho et al. 1997), and a typical zenith an- unresolved emission. We also try to minimize the
gle of 45◦. This predicts, per baseline, a dy- systematic effects caused by the maser velocity
namic ionospheric phase error of 3◦ / 5◦, and structure by fitting the parallax using individual
a static ionospheric phase error of 16◦ / 28◦ for spotswiththesamevelocitybetweenepochs. Fea-
WX Psc / OH138. Due to the low frequency, the ture fitting is not feasible due to the few detected
non-dispersive tropospheric phase errors are neg- spots in each feature6.
5TECUnit;1TECU=1016 electrons m−2 6Amaserspotreferstoanindividualmaserbrightnesspeak
7
Frequency (MHz) Frequency (MHz)
1612.3 1612.2 1612.1 1612.5 1612.45 1612.4
0.9 0.9
Stokes I Stokes I
((aa)) WWXX PPsscc ((bb)) OOHH113388..00++77..22
Stokes V Stokes V
0.8 0.8
0.7 0.7
) )
y y
J AAuugguusstt 11,, 22001144 J FFeebbrruuaarryy 1166,, 22001144
( (
y 0.6 y 0.6
t t
si si
n n
e 0.5 e 0.5
d d

x FFeebbrruuaarryy 1177,, 22001155 x MMaayy 77,, 22001144
u u
fl 0.4 fl 0.4

d d
e e
t t
a a
el 0.3 el 0.3
rr JJuunnee 88,, 22001155 rr AAuugguusstt 77,, 22001144
o o
C 0.2 C 0.2
0.1 0.1
JJuullyy 77,, 22001155 FFeebbrruuaarryy 2222,, 22001155
0 0
-10 0 10 20 30 -50 -45 -40 -35 -30 -25
-1 -1
LSR velocity (km s ) LSR velocity (km s )
Fig. 2.— Scalar averagedcross-powerspectra of the 1612 MHz OH maser emission for each epoch towards
(a) WX Psc and (b) OH138. Thick and thin lines denote the Stokes I (total intensity) and V (degree of
circular polarization) parameters, respectively. The spectra are unsmoothed and have a channel spacing of
1.95 kHz (0.36 km s−1). Maser spots used for the parallax fitting are located in the brighter blueshifted
peaks. Thick dashed lines mark the systemic velocities, calculated by averaging the strongest blue- and
redshifted maser channels from all epochs. Spectra of different epochs have been shifted along the vertical
axis for clarity.
8
Astrometric errors from the image thermal that despite the smaller target–calibrator separa-
noiseareapproximatedasσtherm 0.5 (cid:0)θ/SNR(cid:1), tion, the astrometric errors of WX Psc can be
≈ ·
whereθ isthesizeofthesynthesizedbeamincase worsethanthatofOH138mainlybecausethe ref-
thesourceisunresolved,andSNRisthesignal-to- erence source of WX Psc is weaker. Although
noise ratio in our VLBA maps. The trade-off of transferred errors from calibrator structure can
limiting ourselvesonly to the most compact parts be ruled out as we modeled their structure before
of the OH maser emission for our astrometry – as phase referencing, the thermal noise in the trans-
an effort to reduce systematic effects from source ferred visibility phase solutions from the weaker
structure – is a reduced SNR and thus a larger J0106+1300to WX Psc is higher than in the case
thermal noise component in the total astrometric of J0322+6610 and OH138. Our simulations also
error budget. Fortunately, the thermal noise is show that the astrometric errors are particularly
a random error source, so it has a more benign sensitive to the peak value of the target Gaussian
effect on the parallax measurements than leaving component: the astrometric error for a maximum
possible systematic errors in our datasets. How- peak of 0.4 Jy is a factor of 1.6 worse than for
∼
ever, this highlights one of the major difficulties a maximum peak of 0.6 Jy, while the error for a
in low-frequency astrometry. As σ ν−1, maximumpeak of0.8Jyisafactorof 1.3better
therm
∝ ∼
whereν isfrequency,the intrinsiclimitofastrom- than for a maximum peak of 0.6 Jy. These errors
etry is lower than for CH OH or H O masers at canbe differentfromepochto epochas the maser
3 2
higher frequencies. source strength varies.
We quantitatively investigated the astrometric We then repeated the same simulations with
errorsnotonlybytheanalyticalmethoddescribed the ionospheric model errors turned off, and ob-
above, but also by simulating our 1.6 GHz VLBA tained astrometric errors of 1.0 / 0.4 mas for
observations with ARIS (Asaki et al. 2007). We WX Psc / OH138 for a maximum peak of 0.6 Jy.
adopted input parameters based on our observa- Comparing the two sets of simulations the contri-
tion parameters and typical error values in VLBI butionoftheionosphericmodelerrortotheastro-
observations: the target source is a single circu- metric measurements was found to be 0.6 mas,
∼
lar Gaussian component with a full width at half which is consistent with the previous analytical
maximumof10mas anda maximumpeak of0.4– estimates. We can also see that the limited band-
0.8 Jy in 1.95 kHz bandwidth. The reference width on the calibratorsis a major contributor to
sources for WX Psc and OH138 are J0106+1300 thetotalerrorbudget,whichcanbeavoidedinfu-
andJ0322+6610withfluxdensitiesof0.07Jyand tureobservationsbyusinglargertotalbandwidths
0.75 Jy in 32 MHz bandwidths, respectively. Be- for an increased continuum sensitivity. The sizes
causetheabovesourcestrengthswereassignedby of datasets can be kept manageable by using sev-
referringtoourdatareductionresults,theprimary eral correlatorpasses and spectral “zooming”,i.e.
beam attenuation in our in-beam phase referenc- correlating all scans with a coarse resolution on
ing was not considered. Imaging was conducted the full bandwidth and the maser scans with a
withoutMK andSC asdescribedinSect. 5.1. We high resolution on a narrow bandwidth contain-
simulated 200 samples and estimated astrometric ing the spectralfeatures. This feature is routinely
errors as the position offsets in RA and Dec from available on both DiFX (Deller et al. 2011) and
the phase tracking centers that contained 67% of SFXC (Keimpema et al.2015)correlatorsusedat
the simulated positions. most VLBI arrays.
The obtained astrometric errors from the sim-
4.2. EmpiricalErrorsfromSubarrayImag-
ulationsare1.2/0.7masforWXPsc/OH138in
ing
the case of a 0.6 Jy target Gaussian component,
showinggoodconsistencywithourobservationre- As we discussed, astrometric errors are com-
sults (see Sect. 4.2). Our ARIS simulations show posed of systematic and random errors. Because
the former can yield systematic shifts in the mea-
imagedinonespectralchannel, andamaserfeaturerefers sured maser positions, their identification is cru-
to a group of spots which are considered to relate to the
cial for accurate astrometry. We do this by imag-
samephysicalmasercloud.
ing the strongest maser channel for each source
9
with all possible three-antenna subarrays of the on the mainland US are of the order of 1TECU.
VLBA using the automatedCLEANprocedure in After the phasereferencingcorrectionswithacal-
AIPS, then measure the position of the peak in ibrator .1◦away, the residual systematic differ-
each resulting map with the verb MAXFIT7 (see ences would be .1TECU sin(1◦) 0.02TECU.
× ≈
Fig. 3). By limiting ourselves to three antennas, These would cause systematic phase errors of
each with an independent static ionospheric er- .6◦(Asaki et al.2007),whichtranslateto.1mas
ror, we form coherent but shifted images of these astrometric errors for 600km baselines, i.e. the
subsets of data. Therefore, comparing the maser average separation between the core antennas of
positions determined from these subarraysclearly the VLBA. Although this seems to explain the
exposetheantennasthatarecontaminatedbysys- observed offsets, it is curious that some close-
tematic errors,because their subarrayimages will by core stations are more heavily affected than
also be affected and shifted systematically. other mainland antennas further away, as would
Looking at Fig. 3, it is clear that the sub- be expected from Dodson et al. (2016). Future
array imaging worked better for OH138, which observations should be conducted to investigate
has a strong and compact maser spot, and seems this issue in more detail. For the present paper,
less conclusive for WX Psc where the detected data related to the marked offset antennas were
emission is more resolved and much weaker (see flagged out before making the final image cubes
Sect. 5.1). In most cases, the distribution of the and getting the astrometric solutions used for the
derivedpositionsiselongatedindeclination,which parallax fitting (see Fig. 3e–h).
is due to the geometry of the VLBA. Among the WX Psc sessionEpoch I.Asystematicshift
120possiblesubarraysthemajorityaredominated of 6mascanbeseenforLA,whichismuchlarger
∼
by East–West baselines, which yield poorer angu- than expected from ionospheric errors. Instead,
lar resolution in the North–South direction. This wesuspectthatthis mightbeanartifactfromthe
is just a random effect which does not skew the imaging of a weak source with only three anten-
astrometric results (see Fig. 3a,c,e–h). Next, we nas. Although we flagged the data related to LA
discuss the measured patterns, with the specific for safetyandbecause flaggingprovideda slightly
sessions and epochs shown in bold. better parallax fit, it only changed the parallax
OH138 session Epochs I–IV. Systematic value by 4% which is well below our relative fit-
shifts could be found and linked to specific an- ting error of 30% (see Fig. 3a).
∼
tennas (Epoch I: MK in RA; Epoch II: PT and WX Psc session Epoch II.Themaseristoo
KP in positive and negative Dec, respectively; weaktobeusefulforerroranalysisasitsmeasured
Epoch III: KP in Dec; Epoch IV: SC in RA). positions show scatter over the whole mapped re-
The size of these shifts are approx. 0.5–1 mas, gionof100 100mas. Asaresult,onlyafewmea-
×
which agree with our expectations for the static sured positions are found around the mean, with
ionosphericerrorcontributions. Whilesubstantial the restnotdisplayeddue tothelargescatter(see
offsetsfoundinthecaseofMKandSCarepossible Fig. 3b).
duetothelongbaselinesanddifferenceinantenna WX Psc session Epoch III. Subarrayimag-
elevations, it comes as a surprise that some of the ing works, but no specific pattern can be seen
core southwestern antennas (FD, KP, LA, OV, in the measured maser positions for any antenna.
PT) would also be affected by ionospheric model Data points related to antennas MK and SC are
errorstosuchadegree. FromDodson et al.(2016) randomly scattered over the whole map area and
wefoundthatundernormalconditionsthetypical are not displayed beyond the central region (see
∆TECmodeldifferencesbetweenVLBAantennas Fig. 3c).
WX Psc session Epoch IV. Measured po-
7Formeasuringthepeakpositions,wecomparedtheGaus-
sitions cluster into two complex groups offset by
sian model fitting of IMFIT/JMFIT, the quadratic func-
tionfittingofMAXFITandsimplyselectingthebrightest 6 mas, similar in size to that seen in Epoch I.
∼
pixelwithIMSTAT.Aslongasthemappedareawasafew Also, no antenna could be linked to the pattern.
times largerthanthe fitted maserspot andthepixel sizes
We again suspect that this might be an artifact
were adequately small compared to the synthesized beam
–inourcase0.1×0.1mas–allthreeapproaches produced due to the low image quality of the three-antenna
nearlyidenticalresults. subarrays. Thecomplexgeometricstructureofthe
10