The Milky Way's Rotation Curve Out To 100 KPC and Its Constraint On The Galactic Mass Distribution

Mon. Not. R. Astron. Soc. 000, ??–?? (2015) Printed 19 August 2016 (MN LATEX style file v2.
2)
The Milky Way’s rotation curve out to 100 kpc and its constraint on
the Galactic mass distribution
arXiv:1604.01216v2 [astro-ph.GA] 18 Aug 2016
Y. Huang1⋆ †, X.-W. Liu1,2⋆, H.-B. Yuan3 , M.-S. Xiang4 †, H.-W. Zhang1,2, B.-Q. Chen1†
J.-J. Ren1 , C. Wang1 , Y. Zhang5, Y.-H. Hou5, Y.-F. Wang5 , Z.-H. Cao4
1 Department of Astronomy, Peking University, Beijing, 100871, People’s Republic of China
2 KavliInstitute for Astronomy and Astrophysics, Peking University, Beijing, 100871, People’s Republic of China
3 Department of Astronomy, Beijing Normal University, Beijing, 100875, People’s Republic of China
4 Key Laboratory of Optical Astronomy, National Astronomical Observatories, Chinese Academy of Sciences, Beijing 100012, People’s Republic of China
5 Nanjing Institute of Astronomical Optics & Technology, National Astronomical Observatories, Chinese Academy of Sciences, Nanjing 210042
ABSTRACT
The rotation curve (RC) of the Milky Way out to ∼ 100 kpc has been constructed using
∼ 16, 000 primary red clump giants (PRCGs) in the outer disk selected from the LSS-GAC
and the SDSS-III/APOGEE survey, combined with ∼ 5700 halo K giants (HKGs) selected
from the SDSS/SEGUE survey. To derive the RC, the PRCG sample of the warm disc pop-
ulation and the HKG sample of halo stellar population are respectively analyzed using a
kinematical model allowing for the asymmetric drift corrections and re-analyzed using the
spherical Jeans equation along with measurements of the anisotropic parameter β currently
available. The typical uncertainties of RC derived from the PRCG and HKG samples are
respectively 5-7 km s−1 and several tens km s−1 . We determine a circular velocity at the
solar position, Vc (R0 ) = 240 ± 6 km s−1 and an azimuthal peculiar speed of the Sun,
V⊙ = 12.1 ± 7.6 km s−1 , both in good agreement with the previous determinations. The
newly constructed RC has a generally flat value of 240 km s−1 within a Galactocentric dis-
tance r of 25 kpc and then decreases steadily to 150 km s−1 at r ∼ 100 kpc. On top of this
overall trend, the RC exhibits two prominent localized dips, one at r ∼ 11 kpc and another at
r ∼ 19 kpc. From the newly constructed RC, combined with other constraints, we have built a
parametrized mass model for the Galaxy, yielding a virial mass of the Milky Way’s dark matter
halo of 0.90+0.07 12 +0.02
−0.08 × 10 M⊙ and a local dark matter density, ρ⊙,dm = 0.32−0.02 GeV cm
−3
.
Key words: Galaxy: disc – Galaxy: halo – Galaxy: kinematics and dynamics – Galaxy:
fundamental parameters – Galaxy: structure.
1 INTRODUCTION Generally speaking, for the inner region (i.e. inside the so-
lar circle) of the Galactic disc, the RC can be accurately mea-
The rotation curve (hereafter RC) of the Milky Way gives the mea-
sured simply using the so-called tangent-point (TP) method with
sured circular velocity Vc as a function of the Galactocentric dis-
the H I 21 cm or the CO 2.6 mm gas emissions in the Galactic plane
tance r. The RC provides important constraints on the mass distri-
as tracer (Burton & Gordon 1978; Gunn, Knapp & Tremaine 1979;
bution of our Galaxy, including its dark matter (DM) content, as
Clemens 1985; Fich, Blitz & Stark 1989; Levine, Heiles, & Blitz.
well as the local DM density (e.g. Salucci et al. 2010; Weber & de
2008; Sofue et al. 2009). In principle, a well defined RC could be
Boer 2010). The latter is crucial for the interpretation of any sig-
established by this method for the entire Galactic inner region if
nals that DM search experiments, direct or indirect, are expected to
one assumes that the gas moves in perfect circular orbits around
detect. The RC can also be used to construct realistic Galactic mass
the Galactic centre. However, the distribution and kinematics of gas
model by fitting the RC with a parameterized multi-component
can be easily perturbed by non-axisymmetric structures, in partic-
Milky Way, consisting of, for instance, a bulge, a disc and a dark
ular by the bar near the centre. Given the presence of those pertur-
matter halo (e.g. Sofue, Honma, & Omodaka 2009; Xin & Zheng
bations, the TP method only works well in deriving the RC for the
2013).
projected Galactocentric distance R from ∼ 4.5 kpc to R0 (Galac-
tocentric distance of the Sun; Chemin, Renaud & Soubiran 2015).
For the outer disc beyond the solar circle, the TP method can not
⋆ E-mails: [email protected] (YH); [email protected] (XWL) be used to derive the RC. Instead, the RC is derived using a variety
† LAMOST Fellow
2 Y. Huang et al.
of tracers belonging to the cold disc populations from the measured In this paper, we report a newly constructed RC of our Galaxy,
line-of-sight velocities (Vlos ) and estimated distances, such as the the Milky Way, extending out to 100 kpc, derived from ∼ 16, 000
thickness of H I gas (Merrifield 1992; Honma & Sofue 1997), H II primary red clump giants (PRCGs) selected from the LAMOST
regions (Fich et al. 1989; Brand & Blitz 1993; Turbide & Moffat Spectroscopic Survey of the Galactic Anti-centre (LSS-GAC; Liu
1993), OB stars (Frink et al. 1996; Uemura et al. 2000; Bobylev & et al. 2014; Yuan et al. 2015) and the SDSS-III/APOGEE survey
Bajkova 2015), carbon stars (Demers & Battinelli 2007; Battinelli (Eisenstein et al. 2011; Majewski et al. 2015) in the (outer) disc,
et al. 2013) and classical cepheids (Pont et al. 1997). However, two as well as from ∼ 5700 halo K giants (HKGs) selected from the
important issues limit the accuracy of RC derived from those disc SDSS/SEGUE survey (Yanny et al. 2009) for the halo region. The
tracers. First, it is difficult to determine the distances of those disc usage of PRCGs in deriving the RC in the outer disc region help
tracers and the poorly determined distances could lead large uncer- solve the above described two issues neatly. Firstly, PRCGs are
tainties (generally of the order of tens km s−1 , see, e.g. the Fig. 1 considered as excellent standard candles given that their intrinsic
of Sofue et al. 2009) in the derived circular velocity Vc . Another luminosities are insensitive to the stellar populations (i.e. metallic-
issue is that, similar to the TP method, the underlying assumption ity and age; e.g. Cannon 1970; Paczyński & Stanek 1998). Thus
that the disc tracers used move in purely circular orbits can be eas- their distances can be determined to a much higher precision (typ-
ily broken. Disc tracers belonging to the cold populations, espe- ically 5–10 percent) than for most other tracers belonging to cold
cially those young objects, are generally associated with the spiral disc population. Secondly, PRCGs are of intermediate- to old-age
arms and thus their kinematics are often perturbed by the arms. At stellar populations. Thus they have enough time to dynamically
present, it is difficult to correct for the effects of those perturbations mix in the disc and are therefore less affected by non-axisymmetric
given the properties including dynamics of arms are still poorly un- structures than those cold gaseous or young stellar tracers. On the
derstood. Recently, accurate distances and values of Vlos have been other hand, given that our PRCG sample stars are of relative old-
measured for a number of masers1 by the Bar and Spiral Structure age (i.e. warm), they need to be corrected the so-called asymmet-
Legacy (BeSSeL) survey (Brunthaler et al. 2011), allowing, in prin- ric drifts (the offsets between the circular velocity and the mean
ciple, the determination of RC to a very high precision, say better rotational speed of the population concerned), which can be calcu-
than few km s−1 (e.g. Xin & Zheng 2013; Reid et al. 2014). How- lated from the velocity dispersions of our sample stars. The large
ever, in deriving the RC from those measurements, possible pertur- number of PRCGs employed in the current study dramatically re-
bations to the measured velocities caused by the spiral arms remain duces the random errors of the newly derived RC. To derive the RC
to be properly accounted for. for the halo region, we have chosen SEGUE HKGs as tracers con-
For regions beyond the Galactic disc, the RC needs to be mea- sidering that, 1) They are intrinsically bright and also span about
sured using halo stars, such as the blue horizontal branch (BHB) 4 mag in r-band absolute magnitude (Mr ∼ −1 to 3 mag), allow-
stars (Xue et al. 2008; Deason et al. 2012; Kafle et al. 2012; ing one to determine the RC out to a distance as far as 100 kpc;
Williams & Evans 2015) and the K giants (Bhattacharjee, Chaud- 2) They are abundantly observed in the SDSS/SEGUE survey. We
hury, & Kundu 2014, hereafter BCK14), globular clusters or dwarf note that BCK14 have analyzed the same SEGUE HKG sample
galaxies. For those tracers of halo populations (assuming isotropi- (and other two halo tracer samples) using the spherical Jeans equa-
cally distributed), the radial velocity dispersion σr , number density tion and derived the RC in the halo region. However, the analysis
ν and velocity anisotropy parameter β ≡ 1 − σt2 /σr2 , are linked either assumes a constant anisotropy parameter β or takes its value
to the circular velocity Vc through the Jeans equation (see, e.g. from numerical simulations, and thus could be liable to potential
Binney & Tremaine 2008, pp. 349) for spherical systems. For halo systematic uncertainties. To break the RC/mass–anisotropy degen-
tracers, profile of the radial velocity dispersion σr can be easily es- eracy, we have re-analyzed the SEGUE HKG sample to derive the
timated from the line-of-sight velocity dispersion σlos (Battaglia RC in the halo region using measurements of β now available in
et al. 2005; Dehnen et al. 2006), while their number density is the literature (see Section 4.1). Finally, we have constructed a new
found to follow a double power law with a break radius rb around parameterized mass model for the Milky Way by combining con-
20 kpc (Bell et al. 2008; Watkins et al. 2009; Deason et al. 2011; straints provided by the current, newly constructed RC and other
Sesar et al. 2011). However, the anisotropy parameter β has only available data.
been accurately measured in the solar neighborhood, with a radi- The paper is organized as follows. In Section 2, we describe
ally biased value between 0.5 – 0.7 (Kepley et al. 2007; Smith et the LSS-GAC and SDSS data sets. We derive the RC by model-
al. 2009; Bond et al. 2010; Brown et al. 2010). Due to the lack of ing the PRCG and HKG samples in Sections 3 and 4, respectively.
accurate proper motion measurements of distant halo tracers, the The combined, final RC out to 100 kpc is presented in Section 5. A
anisotropy parameter β is still poorly constrained beyond the solar Galactic mass model derived by fitting the newly constructed RC is
neighborhood, particularly for the outer halo (> 25 kpc). Hence, the presented in Section 6. Finally, we summarize in Section 7.
existing determinations of RC suffer from the so-called RC/mass–
anisotropy degeneracy. To solve this problem, various values of the
anisotropy parameter, either of arbitrary nature (e.g. BCK14) or
predicted by numerical simulations (e.g. Xue et al. 2008; BCK14) 2 DATA
have been adopted in the spherical Jeans equation to derive the
RC. Only more recently, some constraints on the anisotropy pa- 2.1 Coordinate systems
rameter, mainly for the inner halo (6 25 kpc), have become avail- In this study, we use three sets of coordinate systems: (1) A right-
able, based on some direct/indirect measurements (e.g. Deason et handed Cartesian system (X, Y, Z) positioned at the Galactic cen-
al. 2012; Kafle et al. 2012; Deason et al. 2013). tre with X pointing in the direction opposite to the Sun, Y in the
direction of Galactic rotation and Z towards the North Galactic
Pole; (2) A Galactocentric cylindrical system (R, φ, Z) with R, the
1 The masers are generally associated with young massive stars and com- projected Galactocentric distance, increasing radially outward, φ
pact H II regions in the spiral arms. in the direction of Galactic rotation and Z the same as that in the
Rotation curve & mass distribution of the Milky Way 3
Cartesian system; (3) A Galactocentric spherical coordinate system
(r, θ, φ) with r, the Galactocentric distance, increasing radially out-
ward, θ towards the Sourth Galactic Pole and φ in the direction of
Galactic counter-rotation. The Sun is assumed to be at the Galactic
mid-plane (i.e. Z = 0 pc) and has a value of R0 of 8.34 kpc (Reid
et al. 2014). The former two coordinate systems are mainly used for
disc stars and the spherical coordinate system is used for halo stars.
The three velocity components are represented by (U, V, W ) in the
Cartesian system centred on the Sun, (VR , Vφ , VZ ) in the Galac-
tocentric cylindrical system and (Vr , Vθ , Vφ ) in the Galactocentric
spherical system.
2.2 LSS-GAC, SDSS/SEGUE and SDSS-III/APOGEE data

In this work, we use the second release of value-added catalogues
of LSS-GAC (LSS-GAC DR2; Xiang et al. 2016, in preparation),
the ninth SDSS/SEGUE public data release (SDSS/SEGUE DR9;
Ahn et al. 2012) and the twelfth SDSS-III/APOGEE public data
release (SDSS-III/APOGEE DR12; Alam et al. 2015).
LSS-GAC is a major component of the on-going LAMOST Figure 1. Spatial distribution of PRCG sample stars in the X–Y plane.
Experiment for Galactic Understanding and Exploration (LEGUE; Black and red dots represent stars selected from LSS-GAC and APOGEE,
Deng et al. 2012). LSS-GAC aims to collect optical (λλ3800– respectively. Blue dash lines denote different Galactocentric radii.
9000), low resolution (R ∼ 1800) spectra under dark and grey lu-
nar conditions for a statistically complete sample of over three mil-
lion stars of all colours and of magnitudes 14.0 6 r < 17.8 mag
(18.5 mag for limited fields), in a continuous sky area of ∼ 3400
square degrees, centred on the GAC, covering Galactic longitudes
150 < l < 210◦ and latitudes |b| < 30◦ . Over 2.5 million spectra
of very bright stars (9 < r < 14.0 mag) in the equatorial Decli-
nation range −10 < δ < 60◦ will also be obtained under bright
lunar conditions. The survey, initiated in the fall of 2012, is ex-
pected to last for five years. Details about the survey, including the
scientific motivations, target selections and data reduction, can be
found in Liu et al. (2014) and Yuan et al. (2015). The stellar at-
mospheric parameters and line-of-sight velocity Vlos of LSS-GAC
targets are derived with the LAMOST Stellar Parameter Pipeline at
Peking University (LSP3; Xiang et al. 2015) using template match-
ing with empirical spectral libraries. LSP3 achieves an accuracy Figure 2. Spatial distribution of the SEGUE HKG sample stars in the r–Z
of 5.0 km s−1 , 150 K, 0.25 dex, 0.15 dex for Vlos , effective tem- plane. The two blue lines represent |Z| = 4 kpc.
perature, surface gravity and metallicity [Fe/H], respectively, for
spectra of FGK stars of signal-to-noise ratios (SNRs) per pixel at
4650 Å higher than 10. (Mészáros et al. 2013). Benefited from the high resolution and high
SDSS/SEGUE survey, a Galactic extension of the SDSS-II/III SNRs of APOGEE spectra, the random errors of Vlos delivered for
surveys, has obtained a total of about 360,000 optical (λλ3820– APOGEE stars are at the level of ∼ 0.1 km s−1 with a zeropoint
9100), low resolution (R ∼ 2000) spectra of Galactic stars at differ- uncertainty at the level of ∼ 0.5 km s−1 (Nidever et al. 2015).
ent distances, from 0.5 to 100 kpc (Yanny et al. 2009). The spectra
are processed with SEGUE Stellar Parameter Pipeline (SSPP; Lee
2.3 PRCG and HKG samples
et al. 2008a,b; Allende Prieto et al. 2008; Smolinski et al. 2011),
providing estimates of stellar parameters and Vlos . The typical ex- Specifically, as mentioned earlier, we use PRCGs selected from
ternal errors of the stellar atmospheric parameters yielded by SSPP LSS-GAC and SDSS-III/APOGEE to derive the RC in the (outer)
are ∼ 5 km s−1 in Vlos , 180 K in Teff , 0.24 dex in log g and 0.23 dex disc and HKGs selected from SDSS/SEGUE to derive that in the
in [Fe/H] (Smolinski et al. 2011). halo. The PRCG stars are selected based on their positions in
The SDSS-III/APOGEE survey collects high-resolution the metallicity dependent effective temperature-surface gravity and
(R ∼ 22, 500) and high SNRs (∼ 100 per pixel) spectra in the near- colour-metallicity stellar parameters spaces, as developed by Bovy
infrared (H-band; 1.51 to 1.70 µm) for over one hundred thousand et al. (2014) and applied to the APOGEE data. From SDSS DR12,
stars (mainly the red giant stars) in the Milky Way. The scientific a total of 19,937 PRCGs are identified. Huang et al. (2015a) apply
motivations and target selections are described in Majewiski et al. the same method to the LSS-GAC DR2 and identify over 0.11 mil-
(2015) and Zasowski et al. (2013), respectively. The data reduction lion PRCGs. The almost constant absolutes magnitude of PRCGs
and stellar parameter determinations are introduced by Nidever et allow us to assign distances to the individual PRCGs with an accu-
al. (2015) and Garcı́a Pérez et al. (2015), respectively. Calibrated racy of 5–10 per cent. For consistency of analysis, we have applied
with open clusters, the accuracy of APOGEE stellar parameters are a zeropoint correction to Vlos values of LSS-GAC PRCG sample
better than 150 K in Teff , 0.2 dex in log g and 0.1 dex in [Fe/H] stars by adding a constant 2.7 km s−1 to those values. This zero-
4 Y. Huang et al.
Table 1. Parameters of the kinematical model employed
Parameter Adopted value

Rd (kpc) 2.5 ± 0.5
2 /σ 2
σφ 0.5 ± 0.3
R
Rσ (kpc) 16.40 ± 1.25
σR0 (km s−1 ) 35.32 ± 0.52
R0 (kpc) 8.34 ± 0.16
Ω⊙ (km s−1 kpc−1 ) 30.24 ± 0.11
VR,⊙ (km s−1 ) −7.01 ± 0.20
Figure 3. Schematic diagram showing the observed heliocentric line-of-

sight velocity of a star in the Galactic plane at position (R, l), where R is
the Galactocentric radius and l the Galactic longitude. The velocities are
defined in cylindrical coordinates centred on the Galactic centre (GC).
point offset between LSS-GAC and APOGEE Vlos values is de-

rived from a comparison of the two sets of measurements for 1500
common PRCG sample stars. The correction is consistent with the Figure 4. Profile of line-of-sight velocity dispersion given by PRCG disc
finding of Xiang et al. (2015) who compare the values of Vlos for stars of [Fe/H] > −1.0 in the direction of Galactic centre (l ∼ 0◦ ) selected
the LSS-GAC DR1 and APOGEE full samples that have about from APOGEE and in the direction of Galactic anti-centre (l ∼ 180◦ ) se-
3800 common sources. To ignore the vertical motions in the fol- lected from LSS-GAC with the requirement of |b| 6 5◦ (blue dots) or
lowing kinematic analysis and to minimize the contamination of |Z| 6 0.5 kpc (red dots). The red line represents an exponential best fit to
halo stars, we have restricted the PRCG sample to stars of |b| 6 3◦ the data points of requirement |Z| 6 0.5 kpc as described by Eq. (4).
and [Fe/H] > −1.0. Finally, a total of 15,634 PRCGs are selected,
with 11,572 stars from LSS-GAC and 3792 stars from APOGEE. helio
centric line-of-sight velocities V los of stars at a given position
As Fig. 1 shows, our PRCG sample spans from R = 6 to 16 kpc in (R, l) in the Galactic plane in the Galactocentric cylindrical frame
the Galactic plane. is given by,
The HKGs used here are taken from the SEGUE K giant cat-
helio
alog compiled by Xue et al. (2014). The catalog provides unbiased V los = V φ (R) sin β − Vφ,⊙ sin l
(1)
distance estimates with a typical precision of 16 per cent, as well as + V R (R) cos β + VR,⊙ cos l,
values of Vlos and metallicities for a total of 6036 K giant stars. To
exclude possible contamination from the disc population, we have where V φ (R) = Vc (R) − Va (R). Vφ,⊙ and VR,⊙ are the Sun’s
selected only those HKGs of |Z| > 4 kpc from the catalog. In ad- azimuthal velocity and the radial component of its peculiar velocity,
dition, we cull of r 6 8.0 kpc considering that only a few stars are respectively. V R is the mean radial motion. β is the angle between
found inside that radius. Finally, a total of 5733 HKGs are selected. the Sun and the Galactic centre with respect to the given position
As Fig. 2 shows, the sample HKGs span a large range in Galacto- and the value of this angle is given by (see Fig. 3),
centric radius r, from 8 to about 100 kpc. R0
β = sin−1 ( sin l). (2)
R
Va (R) is the so-called asymmetric drift and is given by (e.g. Binney
3 RC FROM PRCGS & Tremaine 2008),
2 2
3.1 Kinematical model σR (R) σφ (R) 1 2
Va (R) = [ 2 − 1 + R( + )
2Vc (R) σR (R) Rd Rσ
For the PRCGs, our approach to determine the RC is relied on the (3)
R ∂VR VZ
imprint that the Galactic rotation leaves in the observed heliocen- − 2
],
helio σR (R) ∂Z
tric line-of-sight velocity Vlos , as illustrated in Fig. 3. Specifi-
cally, the Galactic rotation yields a significant sinusoidal depen- assuming that both the number density ν of tracers and their (pro-
dence on the Galactic longitude l of the observed Vlos helio
at a certain jected) radial velocity dispersion σR are exponentially declining as
value of R. Accordingly, one can derive the RC by fitting the ob- a function of R with scale lengths of Rd and Rσ , respectively. The
helio
served Vlos as a function of l of the PRCG sample stars at differ- covariance VR VZ does not show obvious variations with Z since
ent Galactocentric radii using a kinematical, axisymmetric model our data are very close to the Galactic plane (Büdenbender et al.
constructed as follows. Note that throughout the paper, the vertical 2015). Therefore, we can ignore the last term in the above Equa-
motions are ignored and only those in the Galactic plane are con- tion in the following analysis.
helio
sidered since the sample includes only PRCGs of |b| 6 3◦ (see From Eq. (1), V los of stars at a certain R are essentially de-
Section 2.3). As a result of the Galactic rotation, the average helio- scribed by two trends of variations: 1) A sinusoidal dependence
helio ) as a function of the Galactic longitude in the individual annuli of R for the PRCG
Figure 5. Distributions of the heliocentric line-of-sight velocities (Vlos
sample. Back and red dots represent LSS-GAC and APOGEE PRCGs, respectively. Blue dots in each annulus represent the mean heliocentric line-of-sight
velocities of the individual Galactic longitude bins. Red lines show the best fits to the data of the kinematical model described in the text.
on the Galactic longitude governed by the mean Galactic rotation anti-centre area (i.e. l ∼ 180◦ ), we can measure the profile of σR
V φ of the (warm) stellar populations at the given R and the az- directly from the line-of-sight velocity dispersion σlos , since for
imuthal velocity of the Sun Vφ,⊙ ; 2) A cosinusoidal dependence those two areas, σR is essentially identical to σlos . For this purpose,
on the Galactic longitude governed by the mean radial motion2 V R a total of ∼ 4900 PRCG disc stars of [Fe/H] > −1.0 are selected
of the (warm) stellar populations at the given R and the Sun’s pe- from LSS-GAC and APOGEE with |l − 180| 6 3.5◦ for the area
culiar velocity in the radial direction VR,⊙ . For the first trend of toward the Galactic anti-centre and with |l| 6 3.5◦ for that toward
variations, the mean Galactic rotation is a combination of the cir- the Galactic centre. In doing so, we have also widened the cut on
cular velocity (that we want to determine) and the asymmetric drift Galactic latitude by slightly, to |b| 6 5◦ in order to include more
(that we need to correct for). With Rd , σφ2 (R)/σR2
(R) and σR (R) stars. Then we divide those stars into different bins in the radial
known, the asymmetric drift Va (R) becomes dependent on Vc only direction and derive the line-of-sight velocity dispersion σlos for
[see Eq. (3)]. The scale length of the exponential disc, Rd , has been each bin. The binsize in the radial direction is allowed to vary to
studied extensively and is generally known as about 2.5 kpc (e.g. contain a sufficient number of stars in each bin. We require that the
Benjamin et al. 2005; Jurić et al. 2008). For σφ2 (R)/σR 2
(R), we binsizes are no smaller than 0.3 kpc and each bin contains at least
assume it is independent of R and has a fixed value of 0.5, ap- 80 stars. As the blue dots in Fig. 4 show, σlos shows a clear trend
proximately the mean value of existing measurements in the solar of declining with R. The profile is not well constrained, given the
neighborhood (e.g. Dehnen & Binney 1998a; Bovy et al. 2012). limited range of R covered, 10 6 R 6 14 kpc (no data points
The main unknown of Eq. (3) is σR (R), i.e. the value of the expo- available from APOGEE). To better constrain the profile, we have
nentially declining radial velocity dispersion as a function of R. replaced the requirement |b| 6 5◦ with |Z| 6 0.5 kpc in select-
Fortunately, from the existing data of APOGEE in the Galactic ing the stars. For distant stars, the effect of the new requirement
centre area (i.e. l ∼ 0◦ ) and those from LSS-GAC in the Galactic is similar to the original one, but it allows to include more nearby
stars of relative high Galactic latitudes that are still close enough to
the Galactic plane such that their vertical motions can be ignored.
2 helio
Actually, V los also varies with angle β in a cosinusoidal form, as a re- Again, we derive σlos by binning the stars in the radial direction
sult of the mean radial motion. Given that cos β = − cos(φ+l) ∝ − cos l, and the results are overplotted in Fig. 4 by red dots. As expected,
helio
there is also an approximately cosinusoidal dependence of V los on the the new profile is similar to the original one for R > 10 kpc, ex-
Galactic longitude l governed by the mean radial motion.
6 Y. Huang et al.
Figure 6. Circular velocities of the Milky Way derived from our PRCG
sample for the Galactocentric radius range 8 6 R 6 16 kpc (black
dots). Blue line is a linear fit to the RC for R 6 11 kpc. The red star
denotes the circular velocity at the solar position as predicted by the linear
Figure 7. χ2tot , sum of the reduced χ2 values of fit for the 11 annuli defined
fit. Also overplotted cyan triangles and magenta boxes represent, respec-
in Fig. 5, as a function of the assumed value of R0 . The blue line connecting
tively, measurements based H II regions (Fich et al. 1989) and carbon stars
the dots has been smoothed over three adjacent points. Red dashed line
of 60 6 l 6 150◦ (Demers & Battinelli 2007)
represents the adopted value 8.34 kpc of R0 and the 1 σ error of the adopted
value of R0 as estimated by Reid et al. (2014) is shown in grey shade.
cept that now it has data points in the inner disc (R ∼ 7–10 kpc).
To quantitively describe the profile of σR , we fit the measured data
points of σlos obtained with the requirement of |Z| 6 0.5 kpc with
an exponential function,
R − R0
σR (R) = σR0 exp(− ), (4)
Rσ
where σR0 is the radial velocity dispersion at the solar position and
Rσ the scale length. As shown by the red line in Fig. 4, the best fit
yields σR0 = 35.32 ± 0.52 km s−1 and Rσ = 16.40 ± 1.25 kpc.
The value of σR0 found here is consistent with the previous mea-
surements for stars in the solar neighborhood (e.g. Dehnen & Bin- Figure 8. Mean radial motion as a function of R deduced from the PRCG
ney 1998a; Bensby et al. 2003; Soubiran et al. 2003). The value sample.
of Rσ agrees well with the recent determination of Sharma et al.
(2014), who report Rσ ∼ 14 kpc based on the RAVE (Steinmetz et
al. 2006) data. 3.2 Fitting and results
In principle, the mean radial motion V R involved in the sec-
ond trend of variations should be zero under our axisymmetric as- With the asymmetric drift properly modeled and a series of parame-
sumption. However, based on the RAVE data, Siebert et al. (2011) ters fixed as described above, we are now left with two free param-
and Williams et al. (2013) recently show that the mean radial mo- eters, i.e. the circular velocity Vc (R) and the mean radial motion
helio
tion V R in the solar neighborhood is not zero and has a gradient V R (R), to be determined by fitting Vlos (l) measurements of our
in the radial direction. To accomodate the possibility of a non-zero PRCG sample. To do so, we divide the stars into annuli in the ra-
mean radial motion, we have left V R as a free parameter in our dial direction with width 0.5 kpc from R0 to 12.34 kpc, with width
kinematical modeling. We note that the effect of mean radial mo- 1 kpc from 12.34 to 14.34 kpc and with width 2 kpc for the last
helio
tion on V los can be easily disentangled from that of mean Galac- annulus, i.e. 14.34 < R < 16.34 kpc. The choice of the width
tic rotation considering that they have an opposite dependence on are consistent with the typical distance uncertainties of our PRCG
the Galactic longitude. Finally, we fix the values of the azimuthal sample (i.e. 5 per cent). For most annuli, the stars span from ∼ 100
velocity of the Sun Vφ,⊙ (involved in the first trend of variations), to 210◦ in Galactic longitude, wide enough to simultaneously ob-
as well as the radial peculiar velocity, VR,⊙ (involved in the second tain robust estimates of Vc (R) and V R (R). Stars of R 6 R0 in
trend of variations), using the measurements in the literature. For our sample are excluded given their narrow range of distribution in
Vφ,⊙ , it is identical to Ω⊙ R0 . As mentioned earlier, R0 has been set Galactic longitude. For each annulus, with the kinematical model
helio
to 8.34 kpc (Reid et al. 2014). The value of Ω⊙ is well constrained described above, we fit the V los as a function of Galactic longi-
by the proper motions of Sgr A∗ measured by Reid & Brunthaler tude. To calculate the average heliocentric line-of-sight velocities
helio
(2004). For VR,⊙ , we take the value of −7.01 km s−1 determined at different Galactic longitudes, V los (l) measurements, we di-
by Huang et al. (2015b). Table 1 summaries all the fixed parameters vide the stars of each annulus into bins of Galactic longitude. The
employed in our kinematical model. binsize is allowed to vary but set to be no less than 2.5◦ and each
bin contains no less than 20 stars. Finally, the best-fit values of
Vc (R) and V R (R) of each annulus are found by nonlinear fitting
that minimizes χ2 defined as,
N helio helio
X [V los,obs (li , R) − V los,model (li , R| p)]2
χ2 = , (5)
i=1
σ 2 helio
V los,obs (li )
where N is the total number of data points to be fitted, σV helio

los,obs (li )
helio
is the uncertainty of V los (li ), li the mean longitude of the ith
Galactic longitude bin, and p represents the parameters in the kine-
matical model [see Eq. (1)], including those of fixed values as listed
in Table 1 and, the circular velocity and mean radial motion to be
derived from the fitting.
The fits are presented in Fig. 5. The derived RC, i.e., circular
velocity Vc as a function of R, is presented in Fig. 6. To prop- Figure 9. Relative difference, (Vc∗ − Vc )/Vc , between the value of RC de-
duced by varying the assumed value of parameters fixed in the kinematical
erly evaluate the errors of the derived Vc , we consider not only
model , i.e. Vc∗ , and that derived assuming the canonical values of those
the fitting error σVfitc but also the error σVpara that propagates from
c parameters, i.e. Vc . Different colours of triangles represent the results for
the uncertainties of parameters fixed in the kinematical model (see different sets of parameters as labelled in the top-right corner of the dia-
Table 1) as measured by previous or current work. Values of the gram. Lines of different colours are used to connect triangles of the same
latter are calculated using Monte Carlo simulations. In practice, we colour to guide the eye.
obtain a distribution of the derived values of Vc by repeating the
fitting 5000 times, and infer the error σVpara c
from the distribution.
For each fit, values of those fixed parameters are randomly sam- eters in the kinematical model affect the results significantly. To
pled assuming Gaussian distributions of values with uncertainties check the possible effects of σφ2 /σR 2
on the derived RC, we have
qlisted in Table 1. The final error of the derived Vc is then given by
as tried two values, i.e. a lowest value 0.35 and a highest value 0.70 re-
(σVfitc )2 + (σVpara
c
)2 . The typical error of RC thus derived is only ported in the literature, and redo the fitting. The relative differences
5-7 km s−1 . The relatively large errors (few tens km s−1 ) in the few between the original RC, i.e. Vc , that derived assuming the canoni-
annuli of large R are due to the relatively poor sampling in Galactic cal values of those parameters, and those derived after changing the
longitude for those annuli. The newly derived RC shows a smooth canonical value of σφ2 /σR 2
to the two extreme values above, denoted
∗
trend of variations with R, except for a clear dip at R around by Vc , are presented in Fig. 9. The differences are all smaller than
11 kpc. To infer the circular velocity at the solar position, Vc (R0 ), 0.5 per cent (∼ 1.25 km s−1 ). Similarly, by choosing a large scale
we apply a linear fit to the data point inside the dip, i.e. R 6 length Rd = 3.7 kpc from Chang, Ko & Peng (2011) and an al-
11 kpc (Fig. 6). The fit yields Vc (R0 ) = 239.89 ± 5.92 km s−1 , most flat radial velocity dispersion profile (σR0 = 31.4 km s−1
along with a local estimate of the slope of RC, i.e. ∂Vc /∂R, of and Rσ = 270 kpc) from Bovy et al. (2012), we find that the re-
−6.85 ± 3.90 km s−1 kpc−1 . Combining this estimate of Vc (R0 ) sultant changes in our results (cf. Fig. 9) are again very small, less
and the known Vφ,⊙ , we find an azimuthal peculiar velocity of the than 2 per cent (∼ 5 km s−1 ). As discussed in the above Section,
Sun, V⊙ , of 12.09 ± 7.61 km s−1 . As defined in Section 2.1, the our adopted value of R0 is self-consistent with the current data
current analysis assumes R0 = 8.34 kpc as determined by Reid set, suggesting our derived RC should suffer from negligible sys-
et al. (2014). Actually, as evident from Eq. (1), the current data set tematics, if any, as a result of our chosen value of R0 . Finally, at
also provide some constraints on R0 and therefore can be used to present, the best constraint on the Vφ,⊙ comes from the proper mo-
check whether they are consistent with the adopted value of R0 . tion measurements of Sgr A∗ , yielding a value that is also in ac-
For this purpose, we have calculated χ2tot , the sum of the reduced cordance with other most recent independent determinations (e.g.
χ2 values of the fit for the 11 annuli defined in Fig. 5, for various Bovy et al. 2012; Reid et al. 2014; Sharma et al. 2014). Therefore,
assumed values of R0 ranging from 7.2 to 10.0 kpc with a con- we have assumed that there are no systematic errors arising from
stant step 0.1 kpc. The results, plotted in Fig. 7, show a clear mini- our adopted value of Vφ,⊙ in deriving the RC. To conclude, the
mum around 8.3 kpc, identical to the value assumed above. Finally, systematic errors of RC resultant as a consequence of our adopted
we present the mean radial motions V R (R) deduced from the fit- canonical values of parameters are likely to be smaller than 2 per
ting in Fig. 8. The mean radial motion increases from ∼ 1 km s−1 cent (∼ 5 km s−1 ).
at R = 8.5 kpc to ∼ 9 km s−1 at R = 12.5 kpc, and then decrease The kinematical model described above assumes a simplified
to ∼ 4 km s−1 at R = 15 kpc. This trend of variations of V R found Gaussian distribution of azimuthal velocity Vφ . In reality, for stars
here, together with a negative gradient of the radial motion found of warm population (such as those of our PRCG sample), the dis-
previously from the RAVE data for 7 . R . 8.5 kpc (Siebert tribution of Vφ is significantly skewed (e.g. Schönrich & Binney
et al. 2011; Williams et al. 2013), suggest that the value of mean 2012, hereafter SB12). However, the simplification is expected to
radial motion oscillates with R. This interesting result is worth of have only minor effects on our derived RC. As pointed out by Bovy
further investigations but is out of the scope of the current study. et al. (2012), only at the tangent points the non-Gaussianity of Vφ
distribution is fully visible as at those points the line-of-sight veloc-
ity is identical to Vφ . The PRCG sample employed in the current
3.3 Discussion study are mostly spread over the Galactic longitudes between 90
and 240◦ . There is no tangent points in within this Galactic lon-
3.3.1 Systematics
gitude range, hence the skewness of Vφ distribution is much less
To assess the systematic uncertainties of the newly derived RC, we significant for oursample. To check the validity of the above simpli-
first examine whether our choices of values of those fixed param- fication, we have further performed tests using mock-data, as pre-
8 Y. Huang et al.
sented in Appendix A. When generating the mock-data sets, the which is in excellent agreement with most of the recent indepen-
line-of-sight velocities are no longer sampled assuming a Gaussian dent determinations (e.g. McMillan 2011; Schönrich 2012; Reid et
distribution of Vφ but using a distribution calculated from the anal. 2014; Sharma et al. 2014). Their analysis also yields an estimate
alytic formula given by SB12 (that well describes the skewness of of V⊙ of 26 ± 3 km s−1 . A very similar value, about 24 km s−1 ,
Vφ distribution). A detailed description of the mock-data sets and is obtained by Bovy et al. (2015), using a sample of 8155 PRCG
the tests can be found in Appendix A. We fit the mock-data sets us- stars within 250 pc from the Galactic mid-plane selected from the
ing the above kinematical model (assuming a simplified Gaussian APOGEE, quite similar to the sample employed in the current anal-
distribution of Vφ ). The results show that the simplification does ysis. Both estimates are more than 10 km s−1 higher than the val-
not introduce any significant bias in the derived RC. ues estimated from stars in the solar neighborhood (e.g. Schönrich,
In constructing the kinematical model, we have assumed an Binney & Dehnen 2010; Huang et al. 2015b) as well as the value
axisymmetric Galactic disc. In reality, there are prominent non- deduced in the current study. We note however, the analyses of both
axisymmetric structures in the Milky Way, such as the central bar Bovy et al. (2012) and Bovy et al. (2015) assume a certain shape
and spiral arms, that may bias the derived RC. In principle, as dis- of the RC – a flat RC in fact. The analyses also adopt a flat radial
cussed earlier, the PRCGs belong to relatively old and warm popu- velocity dispersion profile. Both assumptions have a major impact
lations, and consequently should be relatively insensitive to pertur- on the estimated value of V⊙ yet are not supported by the current
bations by non-axisymmetric structures. It is thus notable that, for data.
the current PRCG sample, we have detected a non-zero mean radial
motions, increasing from ∼ 1 km s−1 at R = 8.5 kpc to ∼ 9 km s−1
at R = 12.5 kpc, and then decreasing to ∼ 4 km s−1 at R = 15 kpc 3.3.3 The dip at R ∼ 11 kpc
(see Fig. 8). Generally, non-axisymmetric structures could induce Our newly derived RC shows a prominent dip at R ∼ 11 kpc. Such
such streaming motions, both in the radial and azimuthal direc- a similar feature (dip at R ∼ 11 kpc) of RC has also been noted
tions. Therefore, the possibility that the derived RC is affected by and studied by many previous studies (e.g. Sikivie 2003; Duffy &
the non-axisymmetric structures cannot be ruled out, considering Sikivie 2008; Sofue et al. 2009; de Boer & Weber 2011). Sikivie
that the trend seen in V R may be caused by such perturbations. By (2003) and Duffy & Sikivie (2008) interpret the dip by the existence
numerical simulations, one can quantitatively examine the effects of hypothetical caustic rings of dark matter in the Galactic plane,
of non-axisymmetric perturbations on warm populations, such as with ring radii predicted at an ≃ 40kpc n
(n = 1, 2, 3, 4, ...) for the
the PRCGs analyzed here, and figure out how large are the effects Milky Way. In this model, the n = 3 hypothetical caustic ring is
on the derived RC. However, we leave this to future work. located at a3 ≃ 13 kpc, and is interpreted to be responsible for the
Finally, we note that the systematics caused by uncertainties in dip of RC at R ∼ 11 kpc. Alternatively, de Boer & Weber (2011)
our determinations of distance and Vlos of our PRCG sample stars fit the dip with a donut-like ring of dark matter in the Galactic plane
are likely to be negligible, given the high accuracies of the current and conclude that the ring is actually at R ∼ 12.4 kpc with a total
distance and Vlos estimates. mass of ∼ 1010 M⊙ . We will address this feature as revealed by our
more accurate RC in a quantitatively way in Section 6. Finally, we
note that the possibility that some unknown perturbations, such as
3.3.2 Comparisons with other work those induced by non-axisymmetric structures discussed above, are
As introduced, there have already many determinations of the RC actually responsible for this localized dip at 13 kpc, cannot be ruled
in the disc based on data of various disc tracers of cold populations. out, considering that the derived RC is based on survey data that
To compare those earlier results with ours, we have used the data encompass only limited volumes of the whole Galaxy.
(i.e. distances, radial velocities and errors) of H II regions published
by Fich et al. (1989) and those of carbon stars of 60 6 l 6 150◦
published by Demers & Battinelli (2007), and recalculate the circu- 4 RC FROM HKGS
lar velocities, adopting the same value of R0 assumed in the current
4.1 Spherical Jeans model and results
study, the value of Vc (R0 ) deduced from the current study, as well
as the solar peculiar velocities in the radial and vertical directions To derive the RC in the halo region, the spherical Jeans equation is
(U⊙ , W⊙ ) from Huang et al. (2015b) for consistency reason. The applied to the HKG stars to estimate the circular velocity Vc (R) in
results are presented in Fig. 6. In generally, they show a trend of equilibrium through the relation,
variations similar to our new measurements but with much large
d lnν d lnσr2
scatters. The large scatters are possibly due to the large distance Vc2 (R) = −σr2 ( + + 2β), (6)
errors in those data, or the perturbations of the non-axisymmetric d lnr d lnr
structures, or both of them. More recently, Bovy et al. (2012) de- where σr is the radial velocity dispersion of the HKGs and ν the
rive the RC for R from 4 to 14 kpc with a kinematical model3 quite number density of HKGs. The velocity anisotropy parameter β is
similar to ours, using data of 3365 stars of warm populations se- defined as,
lected from APOGEE. In their analysis, they have assumed a flat σθ2 + σφ2
and a power-law form of the RC when fitting the data. Both ap- β =1− , (7)
2σr2
proaches yield similar results and give Vc (R0 ) = 218 ± 6 km s−1 .
This estimate of circular velocity at the solar position is substan- where σθ and σφ are the polar and azimuthal velocity dispersions
tially smaller than our value of Vc (R0 ) = 239.89 ± 5.92 km s−1 , in the spherical coordinate system defined in Section 2.1.
Recent extensive studies show that the number density of halo
stellar population follows a broken power-law (ν ∝ r −α ) distri-
3 We note that there is actually some differences between the kinematical bution with a minor- to major-axis ratio q = 0.5–1, and a shallow
analyses of Bovy et al. (2012) and ours that their analysis has to assume a slope of α ∼ 2–3 out to a break radius rb ∼ 16–27 kpc followed
certain shape of RC (either flat or power-law) while ours does not. by a steeper slope of α ∼ 3.8–5 beyond rb (e.g. Bell et al. 2008;
Table 2. The velocity anisotropy parameter β at different radii
r β
(kpc)
[8.0, 12.0] +0.50+0.10
−0.10
13.2 +0.47+0.23
−0.29
14.0 +0.21+0.32
−0.23
15.1 −0.23+0.51
−0.60
16.1 −0.64+0.68
−0.87
16.9 −1.08+0.78
−1.01
17.9 −0.62+0.74
−0.98
[18.0, 30.0] +0.00+0.20
−0.40
> 50 +0.40+0.20
−0.20
Figure 10. Velocity dispersions derived from the HKG sample. Green tri-
angles and black circles represent the values of line-of-sight velocity dis-
Watkins et al. 2009; Sesar et al. 2011, 2013; Deason 2011; Faccioli persion in the GSR frame σGSR , and those of radial velocity dispersions
et al. 2014; Xue et al. 2015). Similar to Kafle et al. (2014), here we σr , respectively. The dotted line indicates the broken radius (r = 20 kpc)
adopt a spherical (i.e. q = 1), broken power-law distribution for of the σr profile. Red and blue lines show the best power-laws fits to the
the stellar halo, assuming α = 2.4 for the inner halo (r 6 rb ) and profile within and beyond the broken radius, respectively.
α = 4.5 for the outer halo (r > rb ) from Watkins et al. (2009).
In agreement with the recent measurements (e.g. Bell et al. 2008;
Sesar et al. 2013; Kafle et al. 2014; Xue et al. 2015), we set the
broken radius rb to 20 kpc.
As mentioned earlier, determinations of the RC for the halo re-
gion suffer from the so-called RC–anisotropy degeneracy because
of the poorly constrained velocity anisotropy parameter β, espe-
cially for the outer halo. There are however some recent progress
in the measurements of β to large distances, using direct or in-
direct methods that help break the degeneracy. For the inner halo
(r 6 12 kpc), β is well, both directly and indirectly, measured and
is found to have a radial biased value around 0.5 (e.g. Smith 2009;
Brown 2010; Kalfe et a. 2012). For the region 12 < r 6 18 kpc,
based on a sample of ∼ 4600 BHB stars, Kafle et al. (2012) find
that β declines steadily, reaching a tangential value of ∼ −1.1
at ∼ 17 kpc. From the proper motions of main-sequence stars Figure 11. Circular velocities of the Milky Way derived from the HKG sam-
measured by HST, Deason et al. (2013) find the halo is isotropic ple for the range 8 6 r 6 100 kpc (black circles). Magenta boxes are values
(β = 0.0+0.2
−0.4 ) at r = 24 ± 6 kpc. The halo beyond 50 kpc is again
derived from the PRCG sample. Also overplotted as green downward trian-
found to be radially biased with β = 0.4 ± 0.2, based on the most gles and cyan triangles are, respectively, determinations taken from Kafle et
recent study of Kafle et al. (2014). Table 2 summarizes these latest al. (2012) and BCK14.
measurements of β. The Table is used to infer β at any given r by
interpolation in our following analysis. where
Finally, we determine the last unknown term in the Jeans equa-
r 2 + R02 (r 2 − R02 )2 r + R0
tion – the profile of radial velocity dispersion. To do so, we first A(r) = − ln| |. (10)
helio 4r 2 8r 3 R0 r − R0
convert the observed Vlos to the Galactic standard of rest (GSR)
frame by, The profile of σr , estimated from the above equations, is also pre-
sented in Fig. 10. The uncertainty of σp GSR for each bin is estimated
helio
VGSR = Vlos +U⊙ cos b cos l+Vφ,⊙ cos b sin l+W⊙ sin b, (8) by the classical method, ∆σGSR = 1/[2(N − 1)]σGSR , where
N is the total number of stars in the bin. Then the uncertainty of
where U⊙ and W⊙ are taken from Huang et al. (2015b), and Vφ,⊙ σr for each bin is propagated from that of σGSR using Eq. (9). To
is again set to the value as adopted in Section 3.1. Next, we cal- account for possible uncertainties induced by the correction factor
culate the GSR line-of-sight velocity dispersion, σGSR , by divid- for σr , we run Monte Carlo simulations. This is realized using the
ing HKGs into different radial bins. The binsize in radial direc- uncertainties of the input quantities (β, r and σGSR ) in Eq. (9) for a
tion is allowed to vary such that each bin contains at least 40 given bin and randomly sampling those quantities assuming Gaus-
stars. We require that the radial binsizes are no smaller than 1.0 kpc sian error distributions. For each bin, we obtain a distribution of σr
for r 6 20 kpc, 2.5 kpc for 20 < r 6 50 kpc and 5.0 kpc for by repeating the sampling 1000 times, and infer the error of σr for
r > 50 kpc, respectively, to match with the typical distance uncer- that bin from the distribution. Similar to the density profile, the pro-
tainties of our HKG sample stars (i.e. 16 per cent). The results are file of σr also shows a broken radius around 20 kpc, with a steeper
presented in Fig. 10. Then values of the radial velocity dispersion
slope inside the broken radius than beyond. To better describe the
σr can be obtained by applying a correction factor (Dehnen et al. profile, we apply a double power-law fit (σr ∝ r −γ ) to the pro-
2006) to σGSR , file with a broken radius of 20 kpc. The fit yields an inner slope of
σGSR γ = 0.43 and an outer slope of γ = 0.24.
σr = p , (9)
1 − βA(r) We now can obtain the RC by solving the spherical Jeans
10 Y. Huang et al.
Eq. (6) with values of ν, β and σr for the HKG sample properly Fig. 11 shows, the measurements also show two prominent dips at
specified above. The derived RC is presented in Fig. 11. The un- r ∼ 11 and ∼ 19 kpc, consistent with our result. We notice that the
certainty of circular velocity Vc in each bin is again calculated us- two dips revealed by their data are much deeper than ours, espe-
ing a Monte Carlo approach similar to that described above. The cially for the first dip. The discrepancies are largely caused by the
typical errors of the RC are several tens km s−1 . The largest er- differences in the radial velocity profiles between their BHB and
rors found at radii around 20 kpc are due to the poorly constrained our HKG samples. As mentioned earlier, BCK14 have recently de-
velocity anisotropy parameter β around that region. In the region rived the RC in the outer halo of 25 6 r 6 200 kpc, also with the
8 6 r 6 25 kpc, the RC shows two localized dips at radii ∼ 11 and spherical Jeans equation, using three halo tracer samples, including
∼ 19 kpc, respectively. The inner one is exactly that already found the HKG sample used in the current work. We overplot in Fig. 11
above from the PRCG sample. their values that are the combined results from the three halo tracer
samples deduced by setting R0 = 8.3 kpc, Vc (R0 ) = 244 km s−1 ,
and assuming a β profile taken from the numerical simulation of
4.2 Discussion Rashkov et al. (2013). As one can see in Fig. 11, their circular ve-
4.2.1 Systematics locities are somewhat smaller than ours for r 6 30 kpc. Beyond
this, they are in good agreement. The discrepancies inside 30 kpc
As evident from Eq. (6), varying the power-law indexes of stellar are largely due to the differences of β adopted – they use β ∼ 0.6
density of the inner and/or the outer halo change the circular veloc- as given by the simulation while we set it to 0.0 based on the direct
ities derived systematically. To explore the sensitivity of the newly measurement of Deason et al. (2013). For r > 30 kpc, the values
derived RC to the density profile adopted, we have repeated the of β given by the simulation, between 0.50 – 0.75, are close to the
analysis using another two sets of indexes, corresponding, respec- value of 0.4 adopted by us. Thus it is no surprising that the sets of
tively, the lower and upper limits allowed by the current available two RC agree in general within the errors.
measurements. For the lower limit, we use a power-law index of 2.0
and 3.8 for the inner and outer halo, respectively. We find that the
circular velocities thus derived are on average 8 per cent smaller 4.2.3 The dip at r ∼ 19 kpc
than the original values in both the inner and outer halo regions.
In addition to the significant, localized dip at r ∼ 11 kpc discussed
For the upper limit, we adopt a power-law index of 3.0 and 5.0
in Section 3.3.3, another prominent dip at r ∼ 19 kpc is revealed in
for the inner and outer halo, respectively, and find that the circular
the RC derived by the current HKG sample. As mentioned above,
velocities deduced in the inner and outer halo regions are, respec-
the latter is also seen in the RC derived by Kafle et al. (2012) using
tively, 9 and 6 per cent larger than the original ones on average.
over 4000 BHB stars. Recall that the profiles of stellar number den-
Thus the errors of the newly derived RC resultant from the possible
sity, velocity anisotropy and radial velocity dispersion, described in
uncertainties in density profile are probably less than 10 per cent
Section 4.1, all have a break around 20 kpc. Thus it is tempting to
as a whole (i.e. ∼ 20 km s−1 ), which are comparable to the random
conjecture that the dip seen in RC at r ∼ 19 kpc is a direct conse-
errors of the newly derived RC.
quence of the breaks in those profiles that may all have a common
In the current study, we assume that all the HKGs in our sam-
cause. On the other hand, the possibility that the dip is artificial
ple are of a single halo population in the Jeans equation. Recently,
cannot be completely ruled out, given the current measurement un-
some studies (Kafle et al. 2013; Hattori et al. 2013) claim that there
certainties of the break radii as well as the slopes of those profiles.
is a correlation between the metallicity and kinematics of halo stars
Finally, note that the position of the dip at r ∼ 19 kpc coincides
such that metal-rich ([Fe/H]> −2) and metal-poor ([Fe/H]< −2)
roughly with an n = 2 ring radius, a2 ≃ 20 kpc, of a hypothet-
halo stars may actually belong to different populations. The corre-
ical caustic ring of dark matter in the Galactic plane proposed by
lation, if exists, may potentially affect our analysis. On the other
Sikivie (2003) and Duffy & Sikivie (2008). In Section 6, we will
hand, a revisit of the problem by Fermani and Schörich (2013) find
present further quantitive analysis of this dip.
no correlation at all. In the future, with even larger halo samples
than the current available, it is possible to examine this effect quan-
titatively by modeling metal-rich and metal-poor populations sepa-
rately using the Jeans equation. 5 FINAL COMBINED RC
Finally, we note that the current analysis assumes a spherical In this Section, we combine the two segments of RC derived above
stellar halo. It may well be that the stellar halo is not spherical. We from, respectively, a sample of PRCGs selected from LSS-GAC
will however leave the determination of the RC in a non-spherical and APOGEE (Fig. 6) and from a sample of HKGs selected from
halo to future studies. SEGUE (Fig. 11). For the overlap region (i.e. 8 6 r 6 15 kpc) of
the two segments, the circular velocities derived from the PRCGs
are adopted as the final values given their high accuracy almost an
4.2.2 Comparisons with other work
order of magnitude higher than those derived from the HKGs. In
We first compare the circular velocities derived from the HKG addition, to provide circular velocities for the inner disk region (in-
sample to those from the PRCG sample in the overlap region side the solar circle), we take the H I measurements of Fich et al.
8 6 r 6 16 kpc. As Fig. 11 shows, they are in good agreement (1989) based on the TP method described above. We only provide
within the errors and the dip at r ∼ 11 kpc discussed above is again H I data for the region between ∼ 4.5 kpc and R0 , believed to be
revealed by data from the HKG sample. The close agreement be- less affected by the non-axisymmetric structures (e.g. the central
tween the two sets of independent determinations suggests the ro- bar; Chemin et al. 2015). For consistency, we have recalculated the
bustness of our analysis based on two types of tracer of different circular velocities from those H I data adopting R0 = 8.34 kpc and
populations. Recently, Kafle et al. (2012) estimate the RC in the Vc (R0 ) = 239.89 km s−1 described above. The circular velocities
region 8 6 r 6 25 kpc with the spherical Jeans equation using for the inner disk region are provided by taking the mean of cir-
4664 BHB stars. Their results are overplotted in Fig. 11. Again, as cular velocities derived from the H I data in every 0.5 kpc radial
Table 3. Final combined RC of the Milky Way
r Vc σVc tracer r Vc σVc tracer

(kpc) (km s−1 ) (km s−1 ) (kpc) (km s−1 ) (km s−1 )
4.60 231.24 7.00 HI 17.56 240.66 49.91 HKG
5.08 230.46 7.00 HI 18.54 215.31 24.80 HKG
5.58 230.01 7.00 HI 19.50 214.99 24.42 HKG
6.10 239.61 7.00 HI 21.25 251.68 19.50 HKG
6.57 246.27 7.00 HI 23.78 259.65 19.62 HKG
7.07 243.49 7.00 HI 26.22 242.02 18.66 HKG
7.58 242.71 7.00 HI 28.71 224.11 16.97 HKG
8.04 243.23 7.00 HI 31.29 211.20 16.43 HKG
8.34 239.89 5.92 MRCG 33.73 217.93 17.66 HKG
8.65 237.26 6.29 MRCG 36.19 219.33 18.44 HKG
9.20 235.30 5.60 MRCG 38.73 213.31 17.29 HKG
9.62 230.99 5.49 MRCG 41.25 200.05 17.72 HKG
10.09 228.41 5.62 MRCG 43.93 190.15 18.65 HKG
10.58 224.26 5.87 MRCG 46.43 198.95 20.70 HKG
11.09 224.94 7.02 MRCG 48.71 192.91 19.24 HKG
11.58 233.57 7.65 MRCG 51.56 198.90 21.74 HKG
12.07 240.02 6.17 MRCG 57.03 185.88 21.56 HKG
12.73 242.21 8.64 MRCG 62.55 173.89 22.87 HKG
13.72 261.78 14.89 MRCG 69.47 196.36 25.89 HKG
14.95 259.26 30.84 MRCG 79.27 175.05 22.71 HKG
15.52 268.57 49.67 HKG 98.97 147.72 23.55 HKG
16.55 261.17 50.91 HKG – – – –
bin. The uncertainties of the mean circular velocities are assumed Except for the rings, we adopt density profiles of the other three ma-
to be 7.0 km s−1 . The final combined values of circular velocity jor components similar to those employed by DB98b and McMillan
at different radius r, their associated 1σ errors (σVc ) and the tracer (2011). They are briefly describe below.
used, are presented in Table 3. This final combined RC is plotted (1) The discs. Three sub-components are included in the disc
in Fig. 12. Generally, the combined RC has a flat value 240 km s−1 component, i.e. a gas, a thin and a thick stellar disc. Their surface
within r ∼ 25 kpc. Beyond this, it starts to decline steadily, reach- densities are all described by,
ing 150 km s−1 at r ∼ 100 kpc. In addition to the overall trend, two R Rhole
prominent localized dips, as described earlier, are clearly seen in Σ(R) = Σd,0 exp(− − ), (12)
Rd R
the RC, with one at r ∼ 11 kpc and another at r ∼ 19 kpc.
with a central surface density Σd,0 and a scale length Rd . The pa-
rameter Rhole is used specially for the gas disc to create a central
cavity in the surface density that match with the observations (e.g.
Dame 1987). As in DB98b, we adopt Rhole = 4 kpc for the gas disc
6 GALACTIC MASS MODELS BASED ON THE and Rhole = 0 for the stellar discs. We further fix the surface den-
COMBINED RC sity for the individual sub-discs to the local measurements, denoted
by ΣR0 . We adopt ΣR0 ,gas = 17.0 M⊙ pc−2 from Read (2014)
6.1 Galactic mass models and the fit results and ΣR0 ,thick = 7.0 M⊙ pc−2 from Flynn et al. (2006). The local
Modeling the mass distribution of the Milky Way is a fundamental surface density of the thin disc can be derived from the total local
stellar surface density ΣR0 ,stellar = 38.0 M⊙ pc−2 estimated by
task of Galactic astronomy (e.g. Dehnen & Binney 1998b, here-
after DB98b; Klypin, Zhao & Somerville 2002). It is also of vital Bovy et al. (2013) by subtracting contributions from the thick disc
importance for understanding the Galaxy formation and evolution, and the stellar halo (ΣR0 ,halo = 0.6 M⊙ pc−2 ; Flynn et al. 2006).
Then, for each sub-discs, the central surface density can be calcu-
bearing fundamental questions such as whether the Galactic disc
is maximal (e.g. Sackett 1997), whether there is a ‘missing baryon lated from Σd,0 = ΣR0 exp(R0 /Rd + Rhole /R0 ). Moreover, we
problem’ in our Galaxy (e.g. Klypin et al. 1999) and whether our fix the ratio of gas disc scale length to the thin disc scale length to
′
Galaxy is an archetypical spiral galaxy comparing to other local 2. For any axisymmetric component with surface density Σ(R ),
spiral galaxies (e.g. Hammer et al. 2007). As introduced in Sec- the circular velocity is given by (Binney & Tremaine 2008; Xin &
tion 1, the RC provides the most fundamental, direct probe of the Zheng 2013),
mass distribution of the Milky Way. In doing this, we have con- Z R
a2 da
Z ∞
dΣ(R )
′
structed a parametrized Galactic mass model by fitting the model Vc2 (R) = −4G √ √ ′ . (13)
0 R 2 − a2 a R 2 − a2
predicted RC to our newly derived combined one presented above.
This parametrized Galactic mass model consists of four major com- Specifically, for the exponential razor-thin stellar discs, Eq. (13)
ponents, i.e. three discs, a bulge, a dark matter halo and two rings. has analytic solution given by,
The model predicted circular velocities as a function of Galactic Vc2 (R) = 4πGΣd,0 Rd y 2 [I0 (y)K0 (y) − I1 (y)K1 (y)], (14)
radius are contributed by the four components as given by,
where y = R/(2Rd ). In and Kn (n = 0, 1) are the first and second
Vc2 = Vc,disc
2 2
+ Vc,bulge 2
+ Vc,halo 2
+ Vc,ring . (11) kind modified Bessel functions, respectively. Finally, the circular
12 Y. Huang et al.
Figure 12. Final combined RC of the Milky Way to ∼ 100 kpc derived from H I data (green dots), PRCGs (red dots) and HKGs (blue dots). Lines of different
colours as labeled in the bottom right corner of the diagram represent the best-fit RCs contributions to the components of the Milky Way, with the line in gold
representing the sum of contributions from all the mass components (see Section 6.1 for details).
Table 4. Best-fit mass model parameters and derived quantities
Galactic component Parameter Value Unit Notea

Bulge Mb 8.9 109 M⊙ fixed
+203.5
discs Σd,0,thin 726.9−123.6 M⊙ pc−2 fixed
Rd,thin 2.63+0.16
−0.21 kpc fitted
Md,thin 3.15+0.35
−0.19 1010 M⊙ derived
+36.2
Σd,0,thick 30.4−10.3 M⊙ pc−2 fixed
Rd,thick 5.68+2.22
−1.99 kpc fitted
Md,thick 0.62+0.16
−0.06 1010 M⊙ derived
+18.8
Σd,0,gas 134.3−12.1 M⊙ pc−2 fixed
Rd,gas 5.26+0.32
−0.42 kpc fixed
Md,gas 0.55+0.02
−0.02 1010 M⊙ derived
Md,total 4.32+0.39
−0.20 1010 M⊙ derived
Dark matter halo rs 14.39+1.30
−1.15 kpc fitted
ρs 0.0121+0.0021
−0.0016 M⊙ pc−3 fitted
ρ⊙ 0.0083+0.0005
−0.0005 M⊙ pc−3 derived
c 18.06+1.26
−0.90 – derived
rvir 255.69+7.67
−7.67 kpc derived
Mvir 0.90+0.07
−0.08 1012 M⊙ derived
+13.47
Rings Σ0,ring1 44.89−10.32 M⊙ pc−2 fitted
Rring1 12.32+0.49
−0.37 kpc fitted
σring1 1.51+0.54
−0.45 kpc fitted
Mring1 1.32+0.71
−0.50 1010 M⊙ derived
+19.16
Σ0,ring2 27.37−13.69 M⊙ pc−2 fitted
Rring2 20.64+1.03
−1.03 kpc fitted
σring2 1.76+0.97
−0.74 kpc fitted
Mring2 1.57+0.83
−0.75 1010 M⊙ derived
All Mtotal 0.97+0.07
−0.08 1012 M⊙ derived
a
Here “fixed”, “fitted” and “derived” denote the parameter/quantity of concern is
either fixed or fitted in our mass model, or derived from the resultant model.
Figure 13. Two-dimensional marginalized PDFs for the ten model parameters (described in detail in Section 6.1) obtained from the MCMC. Histograms on top
of each column show the one-dimensional marginalized PDFs of each parameter labeled at the bottom of the column. The red contour in each panel delineates
the 1-σ confidence level. The red solid and dotted lines in each histogram represent, respectively, the best-fit value and the 68 per cent probability intervals of
the parameter concerned. The best-fit values and uncertainties of the model parameters are also labeled near the top of the individual columns.
velocity of main disc component is given by quadratic sum of con- radius rt . The indexes γ and β describe respectively the inner (r ≪
tributions from the three sub-dics, Vc,2 disc = Vc,2 thin + Vc,2 thick + r0 ) and outer (r0 ≪ r ≪ rt ) slopes of the radial density profile.
Vc,2 gas . Following McMillan (2011), we set our bulge mass model
(2) The bulge and dark matter halo. The density distributions similar to that constructed by Bissantz & Gerhard (2002) given
of the bulge and the dark matter halo are each described by, that we have no data to constrain the bulge mass distribution. The
ρ0 model has parameters γb = 0, βb = 1.8, rb,0 = 0.075 kpc,
ρ(R, Z) = exp[−(mr0 /rt )2 ], (15) rb,t = 2.1 kpc, an axis ratio q = 0.5 and a scale density ρb,0 =
mγ (1 + m)β−γ
9.93×1010 M⊙ kpc−3 . The total bulge mass corresponding to these
where, parameters is 8.9 × 109 M⊙ . The contribution to circular velocity
p of this bulge component can be calculated numerically (Binney &
m(R, Z) = (R/r0 )2 + (Z/qr0 )2 , (16)
Tremaine 2008, pp. 92).
with scale radius r0 , scale density ρ0 , axis ratio q and truncated For the dark matter halo, a spherical NFW density profile is
14 Y. Huang et al.
adopted with qh = 1, γh = 1, βh = 3 and rt ≃ ∞. The free
parameter ρh,0 , also denoted as ρs , is given by,
ρcr Ωm δth c3
ρs = , (17)
3 ln(1 + c) − c/(1 + c)
where ρcr = 3H 2 /8πG is the critical density of the universe, Ωm
the contribution of (dark and baryonic) matter to the critical density,
δth the critical overdensity at virialization and c the concentration
parameter (the ratio of the virial radius rvir to the scale radius rh,0 ,
also denoted as rs ). In the following analysis, we adopt Ωm = 0.28
and H0 = 69.7 km s−1 Mpc−1 from Hinshaw et al. (2013), and
set δth = 340 (Bryan & Norman 1998). The enclosed virial mass
within the virial radius of the NFW dark matter halo is given by,
4π 3
Mvir = ρcr Ωm δth rvir . (18)
3
The contribution to circular velocity of the NFW dark matter halo
can be calculated as
ln(1 + x) − x/(1 + x) Figure 14. Two-dimensional PDF of virial mass Mvir and concentration
Vc2 (r) = 4πGρs rs3 , (19)
r parameter c. The red solid line delineates the Mvir –c relation predicted by
the ΛCDM simulations of Bullock et al. (2001). The dashed red lines border
where x = r/rs .
the scatters of the relation. The red contour represents the 1-σ confidence
(3) The mass rings. As discussed earlier, there are two promi-
level.
nent localized dips in the RC, one at r ∼ 11 kpc and another at
r ∼ 19 kpc. By chance or not, the two dips are almost at the ex-
act positions of the n = 3 and 2 hypothetical caustic rings of dark
some correlations amongst the parameters. For example, Rring1 is
matter, at a3 ≃ 13 kpc and a2 ≃ 20 kpc, respectively, as proposed
strongly correlated twith σring1 . In addition, anticorrelations are
by Sikivie (2003) and Duffy & Sikivie (2008). To quantify the mass
found between Rd,thin and Rd,thick , and between ρs and rs . Other
distributions that may be associated with the two dips, we consider
pairwise parameters are generally independent of each other. Fi-
two ring-like structures in the Galactic plane in our mass models.
nally, the best-fit values of model parameters are estimated by
Following de Boer & Weber (2011), the surface density profiles of
the median values of their marginalized PDFs. The uncertainties
the two rings are each described by,
are computed from the 68 per cent probability intervals of the
(R − Rring )2 marginalized PDF of each parameter. The final best-fit values of
Σ(R) = Σ0,ring exp[− 2
]. (20)
2σring parameters of our mass models and other derived quantities (e.g.
mass of each component), together with their corresponding uncer-
This Gaussian-like ring has a central surface density Σ0,ring , a ring
tainties, are presented in Table 4. As Fig. 12 shows, the best-fit RC
radius Rring and a Gaussian width σring . The contribution to cir-
is in excellent agreement with the observed one. The contributions
cular velocity of the ring component is given by the quadratic sum
2 2 2 to RC from each of the components corresponding to the best-fit
of contributions of the two rings, Vc,ring = Vc,ring1 + Vc,ring2 and
parameters are also overplotted in Fig. 12.
can be calculated numerically with Eq. (13) numerically.
In total, there are ten free parameters in our Galactic mass
model: two for the discs (Rd,thin and Rd,thick ), two for the dark
matter halo (rs and ρs ), and six for the rings (Σ0,ring1,2 , Rring1,2 6.2 Discussion
and σring1,2 ). To derive the ten free parameters, we fit the model 6.2.1 Dark matter halo
circular velocities given by Eq. (11), to match our newly derived
values (see Table 3 & Fig. 12). We note that the latter six parame- From the best-fit values of ρs and rs , we estimate a virial mass
ters are only sensitive to the two localized features (i.e. the dips) of the dark matter halo, Mvir = 0.90+0.07 12
−0.08 × 10 M⊙ within the
in the RC and thus do not affect the overall fit controlled by the virial radius rvir = 255.69+7.67
−7.67 kpc. The results are in excellent
former four parameters. To efficiently explore the parameter space agreement with the recent measurements of Kafle et al. (2014). The
in searching for the best mass model, we use a Markov Chain concentration parameter c is 18.06+1.26
−0.90 , which also agrees well
Monte Carlo (MCMC) technique to sample the likelihood of the with the recent determinations of Kafle et al. (2014) and Piffl et
data, which is defined as, al. (2014). As expected from Eq. (18), there is a strong anticorre-
N obs model
lation between c and Mvir as shown in Fig. 14. The Mvir –c joint
Y 1 −[Vc,R − Vc,R (p)]2 PDF is presented in Fig. 14. Also overplotted in the Figure is the
L= √ exp i
2
i
, (21)
i=1
2πσV obs 2σV obs relation predicted by ΛCDM simulations taken from Bullock et al.
c,Ri c,Ri
(2001). The values of c predicted by the simulations are systemati-
where N is the total data points, σV obs is the uncertainty of the cally smaller than that yielded by our newly derived RC. However,
c,Ri
observed circular velocity and p represents the ten free parame- as argued by Kafle et al. (2014) and Piffl et al. (2014), the the-
ters of the above Galactic mass model that we want to determine. oretical relation is constructed from simulations with dark matter
The parameters after post-burn period in the MCMC chain give the only. The presence of baryons, not considered in the simulations,
probability distribution functions (PDFs) of the ten free parame- is expected to increase the concentration. Finally, we note that the
ters. We present the marginalized one- and two-dimensional PDFs slightly lighter dark matter halo estimated here could lessen ten-
of the model parameters in Fig. 13. The joint PDFs clearly show sion in hierarchical structure formation in the ΛCDM cosmological
paradigm imposed by the so-called missing satellite and too big to 0.90+0.07
−0.08 × 10
12
M⊙ , a concentration parameter c = 18.06+1.26
−0.90
fail problems (e.g. Springel et al. 2008; Vera-Ciro et al. 2013). and a local dark matter density ρ⊙,dm = 0.32+0.02
−0.02 GeV cm
−3
. We
From ρs and rs , we also find a local dark matter density, also find the two RC dips can be well described by two Gaussian-
ρ⊙,dm = 0.0083+0.0005
−0.0005 M⊙ pc
−3
(0.32+0.02
−0.02 Ge V cm
−3
). The like ring structures with the ring radii almost identical to the ones
very small uncertainty (only 5 per cent) is due to the strong an- predicted by the hypothetical n = 2 and 3 caustic rings of dark
ticorrelation between ρs and rs . Our estimate of ρ⊙,dm is in good matter. The two rings have a mass of the order of 1010 M⊙ .
agreement with the previous global (e.g. Salucci et al. 2010, Catena
& Ullio 2010 and McMillan 2011) as well as local determinations
(e.g. Bovy & Tremaine 2012; Zhang et al. 2013; Bovy & Rix 2013),
ACKNOWLEDGEMENTS
pointing to a nearly spherical local Milky Way dark matter halo
(Read 2014). This work is supported by the National Key Basic Research Pro-
gram of China 2014CB845700 and the National Natural Science
Foundation of China 11473001. We thank Zuhui Fan and Xi-
6.2.2 The rings angkun Liu for valuable discussions. We also thank Prajwal Raj
As Fig. 12 shows, the two localized dips in our RC are fitted quite Kafle for kindly providing his circular velocity measurements. The
well by two Gaussian-like ring structures in the Galactic plane. LAMOST FELLOWSHIP is supported by Special Funding for Ad-
The best-fit values of the radii of the two rings are, respectively, vanced Users, budgeted and administrated by Center for Astronom-
12.32+0.49 +1.03 ical Mega-Science, Chinese Academy of Sciences (CAMS).
−0.37 kpc and 20.64−1.03 kpc, in good agreement with the
radii of the n = 3 and 2 hypothetical caustic rings of dark mat- The Guoshoujing Telescope (the Large Sky Area Multi-Object
ter, at a3 ≃ 13 kpc and a2 ≃ 20 kpc, respectively, as proposed by Fiber Spectroscopic Telescope, LAMOST) is a National Major Sci-
Sikivie (2003) and Duffy & Sikivie (2008). The radius of the inner entific Project built by the Chinese Academy of Sciences. Funding
ring is also in excellent agreement with the value of 12.4 kpc es- for the project has been provided by the National Development and
timated by de Boer & Weber (2011). The rings are quite massive, Reform Commission. LAMOST is operated and managed by the
of the order of 1010 M⊙ , again matching well with estimate of de National Astronomical Observatories, Chinese Academy of Sci-
Boer & Weber (2011). At present, observational evidence linking ences.
the dips seen in the RC to hypothetical caustic rings of dark matter This work has made use of data products from the Sloan Dig-
is still marginal . To better understand the origin of the rings (or the ital Sky Survey (SDSS).
dips in the RC), further observations and simulations are needed.
REFERENCES
7 SUMMARY
Ahn C. P., Alexandroff R., Allende Prieto C., et al., 2012, ApJS,
Based on 16, 000 PRCGs selected from LSS-GAC and SDSS- 203, 21
III/APOGEE, and 5700 HKGs selected from SDSS/SEGUE, we Alam S., Albareti F. D., Allende Prieto C., et al., 2015, ApJS, 219,
have derived the RC of the Milky Way out to ∼ 100 kpc. For the 12
warm disk tracers PRCGs, a kinematic model with asymmetric drift Allende Prieto C., Sivarani T., Beers T. C., et al., 2008, AJ, 136,
correction is used to drive the RC. Benefited from the high accu- 2070
racy of distances and line-of-sight velocities of the PRCG sample, Battaglia G., Helmi A., Morrison H., et al., 2005, MNRAS, 364,
the typical uncertainties of the newly derived circular velocities are 433
only 5-7 km s−1 . From the new accurate RC yielded by the PRCG Battinelli P., Demers S., Rossi C., & Gigoyan K. S,. 2013, Astro-
sample, we have also obtained an estimate of the circular velocity physics, 56, 68
of 240 ± 6 km s−1 at the solar position, of the solar peculiar veloc- Bell E. F., Zucker D. B., Belokurov V., et al., 2008, ApJ, 680, 295
ity in the rotation direction of 12.1±7.6 km s−1 . For the HGK halo Benjamin R. A., Churchwell E., Babler B. L., et al., 2005, ApJ,
tracers, we derive the RC by spherical Jeans equation. We use the 630, L149
current available measurements of the velocity anisotropy param- Bensby T., Feltzing S., & Lundström I., 2003, A&A, 410, 527
eter β to break the so-called RC/mass-anisotropy degeneracy. The Bhattacharjee P., Chaudhury S., & Kundu S., 2014, ApJ, 785, 63
typical uncertainties of the derived circular velocities are several Binney J., & Tremaine S., 2008, Galactic Dynamics, 2nd edn.
tens km s−1 . Princeton Univ. Press, Princeton, NJ
By combining circular velocities from H I measurements for Bissantz N., & Gerhard O., 2002, MNRAS, 330, 591
the inner disk inside the solar circle, we present a combined RC Bobylev V. V., & Bajkova A. T., 2015, Astronomy Letters, 41, 473
for Galactocentric distance r ranging from ∼ 4 to ∼ 100 kpc. The Bond N. A., Ivezić Ž., Sesar B., et al., 2010, ApJ, 716, 1
combined RC show an overall flat value of ∼ 240 km s−1 within Bovy J., Allende Prieto C., Beers T. C., et al., 2012, ApJ, 759, 131
r ∼ 25 kpc, beyond which it declines steadily to 150 km s−1 at Bovy J., & Tremaine S. 2012, ApJ, 756, 89
r ∼ 100 kpc. The newly derived RC also established the existence Bovy J., & Rix H.-W., 2013, ApJ, 779, 115
of significant localized dips at r ∼ 11 and ∼ 19 kpc, respectively. Bovy J., Nidever D. L., Rix H.-W., et al., 2014, ApJ, 790, 127
The dips are possibly related to the n = 3 and 2 hypothetical Bovy J., Bird J. C., Garcı́a Pérez A. E., Majewski S. R., Nidever
caustic rings of dark matter with ring radii at a2 ≃ 20 kpc and D. L., Zasowski G., 2015, ApJ, 800, 83
a3 ≃ 13 kpc, respectively, as proposed by Sikivie (2003) and Duffy Brand J., & Blitz, L., 1993, A&A, 275, 67
& Sikivie (2008). Brown W. R., Geller M. J., Kenyon S. J., & Diaferio A., 2010, AJ,
Finally, we construct a parametrized Galactic mass model con- 139, 59
strained by the newly derived RC and other available constraints. Brunthaler A., Reid M. J., Menten K. M., et al., 2011, Astronomis-
The best-fit yields a virial mass of the dark matter halo Mvir = che Nachrichten, 332, 461
16 Y. Huang et al.
Bryan G. L., & Norman M. L., 1998, ApJ, 495, 80 Lee Y. S., Beers T. C., Sivarani T., et al., 2008a, AJ, 136, 2022
Büdenbender A., van de Ven G., & Watkins L. L., 2015, MNRAS, Lee Y. S., Beers T. C., Sivarani T., et al., 2008b, AJ, 136, 2050
452, 956 Levine E. S., Heiles C., & Blitz L., 2008, ApJ, 679, 1288
Bullock J. S., Kolatt T. S., Sigad Y., Somerville R. S., Kravstsov Liu X. -W., et al., 2014, in Feltzing S., Zhao G., Walton N., White-
A. V., Klypin A. A., Primack J. R., Dekel A., 2001, MNRAS, lock P., eds, Proc. IAU Symp. 298, Setting the scene for Gaia and
321, 559 LAMOST, Cambridge University Press, pp. 310-321, preprint
Burton W. B., & Gordon M. A,. 1978, A&A, 63, 7 (arXiv: 1306.5376)
Cannon R. D., 1970, MNRAS, 150, 111 Majewski S. R., Schiavon R. P., Frinchaboy P. M., et al., 2015,
Chang C.-K., Ko C.-M., & Peng T.-H., 2011, ApJ, 740, 34 arXiv:1509.05420
Chemin L., Renaud F., & Soubiran C., 2015, A&A, 578, A14 Markwardt C. B., 2009, Astronomical Data Analysis Software
Clemens D. P., 1985, ApJ, 295, 422 and Systems XVIII, 411, 251
Dame T. M., Ungerechts H., Cohen R. S., et al., 1987, ApJ, 322, McMillan P. J., 2011, MNRAS, 414, 2446
706 Merrifield M. R., 1992, AJ, 103, 1552
Deason A. J., Belokurov V., & Evans N. W., 2011, MNRAS, 416, Mészáros S., Holtzman J., Garcı́a Pérez A. E., et al., 2013, AJ,
2903 146, 133
Deason A. J., Belokurov V., Evans N. W., & An J., 2012, MNRAS, Nidever D. L., Holtzman J. A., Allende Prieto C., et al., 2015, AJ,
424, L44 150, 173
Deason A. J., Van der Marel R. P., Guhathakurta P., Sohn S. T., & Paczyński B., & Stanek K. Z., 1998, ApJ, 494, L219
Brown T. M., 2013, ApJ, 766, 24 Piffl T., Binney J., McMillan P. J., et al., 2014, MNRAS, 445, 3133
de Boer W., & Weber M., 2011, JCAP, 4, 002 Pont F., Queloz D., Bratschi P., & Mayor M., 1997, A&A, 318,
Dehnen W., & Binney J. J., 1998a, MNRAS, 298, 387 41
Dehnen W., & Binney J. J., 1998b, MNRAS, 294, 429 Press W. H., Teukolsky S. A., Vetterling W. T., & Flannery B. P.,
Dehnen W., McLaughlin D. E., & Sachania J., 2006, MNRAS, (ed.) 2007, Numerical Recipes. The Art of Scientific Computing
369, 1688 (3rd ed.; Cambridge: Cambridge Univ. Press)
Demers S., & Battinelli P., 2007, A&A, 473, 143 Rashkov V., Pillepich A., Deason A. J., Madau P., Rockosi C. M.,
Deng L.-C., Newberg H. J., Liu C., et al., 2012, Research in As- Guedes J., Mayer L., 2013, ApJ, 773, L32
tronomy and Astrophysics, 12, 735 Read J. I., 2014, Journal of Physics G Nuclear Physics, 41, 063101
Duffy L. D., & Sikivie P., 2008, PhRvD, 78, 063508 Reid M. J., & Brunthaler A., 2004, ApJ, 616, 872
Eisenstein D. J., Weinberg D. H., Agol E., et al., 2011, AJ, 142, Reid M. J., Menten K. M., Brunthaler A., et al., 2014, ApJ, 783,
72 130
Faccioli L., Smith M. C., Yuan H.-B., Zhang H.-H., Liu X.-W., Sackett P. D., 1997, ApJ, 483, 103
Zhao H.-B., Yao J.-S., 2014, ApJ, 788, 105 Salucci P., Nesti F., Gentile G., & Frigerio Martins C., 2010,
Fermani F., & Schönrich R., 2013, MNRAS, 432, 2402 A&A, 523, A83
Fich M., Blitz L., & Stark A. A., 1989, ApJ, 342, 272 Schönrich R., Binney J., & Dehnen W., 2010, MNRAS, 403, 1829
Flynn C., Holmberg J., Portinari L., Fuchs B., & Jahreiß H., 2006, Schönrich R., 2012, MNRAS, 427, 274
MNRAS, 372, 1149 Sesar B., Jurić M., & Ivezić Ž., 2011, ApJ, 731, 4
Frink S., Fuchs B., Röser S., & Wielen R., 1996, A&A, 314, 430 Sesar B., Ivezić Ž., Stuart J. S., et al., 2013, AJ, 146, 21
Garcı́a Pérez A. E., Allende Prieto C., Holtzman J. A., et al., 2015, Sharma S., Bland-Hawthorn J., Binney J., et al., 2014, ApJ, 793,
arXiv:1510.07635 51
Gunn J. E., Knapp G. R., & Tremaine S. D., 1979, AJ, 84, 1181 Siebert A., Famaey B., Minchev I., et al., 2011, MNRAS, 412,
Hammer F., Puech M., Chemin L., Flores H., & Lehnert M. D., 2026
2007, ApJ, 662, 322 Sikivie P., 2003, PhLB, 567, 1
Hattori K., Yoshii Y., Beers T. C., Carollo D., & Lee Y. S., 2013, Smith M. C., Evans N. W., Belokurov V., et al., 2009, MNRAS,
ApJ, 763, L17 399, 1223
Hinshaw G., Larson D., Komatsu E., et al., 2013, ApJS, 208, 19 Smolinski J. P., Lee Y. S., Beers T. C., et al., 2011, AJ, 141, 89
Honma M., & Sofue Y., 1997, PASJ, 49, 453 Sofue Y., Honma M., & Omodaka T., 2009, PASJ, 61, 227
Huang Y., Liu X.-W., Zhang H.-W., et al. 2015a, Research in As- Soubiran C., Bienaymé O., & Siebert A., 2003, A&A, 398, 141
tronomy and Astrophysics, 15, 1240 Springel V., Wang J., Vogelsberger M., et al., 2008, MNRAS, 391,
Huang Y., Liu X.-W., Yuan H.-B., Xiang, M.-S., Huo Z.-Y., Chen 1685
B.-Q., Zhang Y., Hou Y.-Y., 2015b, MNRAS, 449, 162 Steinmetz M., Zwitter T., Siebert A., et al., 2006, AJ, 132, 1645
Jurić M., Ivezić, Ž. Brooks A., et al., 2008, ApJ, 673, 864 Turbide L., & Moffat A. F. J., 1993, AJ, 105, 1831
Kafle P. R., Sharma S., Lewis G. F., & Bland-Hawthorn J., 2012, Uemura M., Ohashi H., Hayakawa T., Ishida E., Kato T., Hirata
ApJ, 761, 98 R., 2000, PASJ, 52, 143
Kafle P. R., Sharma S., Lewis G. F., & Bland-Hawthorn J., 2013, Catena R., & Ullio P., 2010, JCAP, 8, 004
MNRAS, 430, 2973 Vera-Ciro C. A., Helmi A., Starkenburg E., & Breddels M. A.,
Kafle P. R., Sharma S., Lewis G. F., & Bland-Hawthorn J., 2014, 2013, MNRAS, 428, 169
ApJ, 794, 59 Watkins L. L., Evans N. W., Belokurov V., et al., 2009, MNRAS,
Kepley A. A., Morrison H. L., Helmi A., et al., 2007, AJ, 134, 398, 1757
1579 Weber M., & de Boer W., 2010, A&A, 509, A25
Klypin A., Kravtsov A. V., Valenzuela O., & Prada F., 1999, ApJ, Williams M. E. K., Steinmetz M., Binney J., et al. 2013, MNRAS,
522, 82 436, 101
Klypin A., Zhao H., & Somerville R. S., 2002, ApJ, 573, 597 Williams, A. A., & Evans, N. W. 2015, MNRAS, 454, 698
Xiang M.-S., et al., 2015, MNRAS, 448, 822 and 2) A parabolic curve,
Xin X.-S., & Zheng X.-W., 2013, Research in Astronomy and As-
Vc = 3 × (R − 12)2 + 220. (A4)
trophysics, 13, 849
Xue X. X., Rix H. W., Zhao, G., et al., 2008, ApJ, 684, 1143 helio
Using the PDF of Vlos described above, we create five
Xue X.-X., Ma Z., Rix H.-W., et al., 2014, ApJ, 784, 170 mock-data sets, two with the flat and three with the parabolic RC.
Xue X.-X., Rix H.-W., Ma Z., et al., 2015, ApJ, 809, 144 Then the same fitting technique used for the real data as described
Yanny B., Rockosi C., Newberg H. J., et al., 2009, AJ, 137, 4377 in Section 3.2 is applied to the mock-data sets. The RCs derived
Yuan H.-B., et al., 2015, MNRAS, 448, 855 from the five mock-data sets are plotted in the left panel of Fig. A1.
Zasowski G., Johnson J. A., Frinchaboy P. M., et al., 2013, AJ, The mean radial motions as a function of R derived from the mock-
146, 81 data sets are also shown in the right panel of Fig. A1. For both as-
Zhang L., Rix H.-W., van de Ven G., Bovy J., Liu C., Zhang G., sumed shapes of RC, the true values of circular velocity are all ex-
2013, ApJ, 772, 108 cellently recovered by our methodology described in Section 3. In
addition, as we expect, the true values of the mean radial motions
as a function of R are also recovered quite well.
APPENDIX A: MOCK-DATA TESTS

Here, we use mock-data to test the effects of the approxi-
mations used in our kinematical model and the performance
of the whole methodology described in Section 3 on deriv-
ing Vc (R) and V R (R). The mock-data sets are created by
helio
re-sampling the heliocentric line-of-sight velocity, Vlos , for
each data point from the probability distribution function (PDF)
p(Vlos |l, b, d, R0 , VR,⊙ , Vφ,⊙ , Vφ (R), VR (R), σR (R)). The posi-
tions (l, b, d) of the stars in the PDF are taken as exactly as those
of the real data, i.e. 15,634 PRCGs. The values of Galactic con-
stants (R0 , Vφ,⊙ , VR,⊙ ) are the same as those fixed in Table 1 and
the radial velocity dispersion profile, σR (R), is taken from the one
derived in Section 3.1 by ourselves. At each R, a Gaussian distri-
bution is used for VR with a velocity dispersion, σR (R), as just
mentioned. The centre of the Gaussian distribution, i.e. the mean
radial velocity V R , is a function of R given by,
V R = −10 × cos(2πR/10 + 2). (A1)

Rather than using a simplified Gaussian distribution of azimuthal-
velocities assumed in our kinematical modeling described in Sec-
tion 3.1, here, we use an azimuthal-velocity distribution generated
from an analytic formula given by SB12. As argued by SB12,
the analytic formula can nicely and naturally reproduce the non-
Gaussianity of the observed azimuthal velocity distribution of the
Geneva-Copenhagen Survey (GCS) local sample of stars with ac-
curate space velocities. In addition, the distribution of Vφ given by
this analytic formula also yields excellent fit to the distribution of
Vφ produced by rigorous torus-based dynamics modelling (Binney
& McMillan 2011). The distribution of Vφ generated from this for-
mula is given by,
Rg − R0 2πRg K
n(Vφ |R, z) = N exp (− )
Rd σR (Rg )
(A2)
∆Φad
× exp [− 2 ]f (z, Rg − R),
σR (Rg )
where N is a normalization factor. Rg is the guiding-centre radius
given by Rg = RVφ /Vc . K is a factor that can be numerically cal-
culated [cf. Eq. (12) of SB12]. ∆Φad and f (z, Rg − R) are the the
so-called adiabatic potential and z factor, respectively (cf. Section 3
of SB12 for detailed descriptions of the two terms). The values of
Rd and R0 are adopted from Table 1. The profile of radial veloc-
ity dispersion, σR (Rg ), is again the same as the one derived in
Section 3.1 by ourselves. For the rotation curve (RC), Vc (R), two
types of shape are assumed and tested: 1) A flat one,
Vc = 220; (A3)
18 Y. Huang et al.
Figure A1. The left and right panels show the rotation curves and mean radial motions, respectively, recovered from the five mock-data sets using the exactly
the same fitting method as applied to the real data. Red lines in the left and right panels plot the true rotation curves and mean radial motions as a function of
R, respectively, that are assumed in generating the mock-data sets. The top two rows show the case of mock-data sets assuming a flat rotation curve of value
of 220 km s−1 [see Eq. (A3)], while the bottom rows are parabolic rotation curve described by Eq. (A4). The true mean radial motions as a function of R are
described by Eq. (A1).

The Milky Way's Rotation Curve Out To 100 KPC and Its Constraint On The Galactic Mass Distribution

Uploaded by

Copyright:

Available Formats

The Milky Way's Rotation Curve Out To 100 KPC and Its Constraint On The Galactic Mass Distribution

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Milky Way's Rotation Curve Out To 100 KPC and Its Constraint On The Galactic Mass Distribution

Uploaded by

Copyright:

Available Formats

Mon. Not. R. Astron. Soc. 000, ??–?? (2015) Printed 19 August 2016 (MN LATEX style file v2.

2.2 LSS-GAC, SDSS/SEGUE and SDSS-III/APOGEE data

Table 1. Parameters of the kinematical model employed

Parameter Adopted value

Figure 3. Schematic diagram showing the observed heliocentric line-of-

point offset between LSS-GAC and APOGEE Vlos values is de-

where N is the total number of data points to be fitted, σV helio

Table 2. The velocity anisotropy parameter β at different radii

Table 3. Final combined RC of the Milky Way

r Vc σVc tracer r Vc σVc tracer

Table 4. Best-fit mass model parameters and derived quantities

Galactic component Parameter Value Unit Notea

APPENDIX A: MOCK-DATA TESTS

V R = −10 × cos(2πR/10 + 2). (A1)

You might also like