Analysis of H O Masers in Sharpless 269 2 using VERA Archival data — Effect of maser structures on astrometric accuracy 2 1 Makoto Miyoshi 0 2 Divisionof Radio Astronomy, National Astronomical Observatory of Japan,2-21-1 Osawa, n Mitaka, Tokyo181-8588, Japan a J Yoshiharu Asaki 0 2 Institute of Space and Astronautical Science, 3-1-1 Yoshinodai, Chuou, Sagamihara, Kanagawa 229-8510, Japan ] A Keiichi Wada, Hiroshi Imai G Graduate School of Science and Engineering, Kagoshima University,1-21-35, Korimoto, . Kagoshima, Kagoshima 890-0065, Japan h p - o r t s a Abstract [ Astrometry using H O maser sources in star forming regions is expected to be 1 2 v a powerful tool to study the structures and dynamics of our Galaxy. Honma et 8 al. (2007) (hereafter H2007) claimed that the annual parallax of Sharpless 269 3 is determined within an error of 0.008 milliarcsec (mas), concluding that S269 2 is located at 5.3 kpc ± 0.2 kpc from the sun, and its galactrocetnricdistance is 4 R = 13.1 kpc. From the proper motion, they claimed that the galacto-centric . 1 rotational velocity of S269 is equal to that of the sun within a 3% error. This 0 smallerror,however,is hardlyunderstoodwhentakinginto accountthe results 2 of other observations and theoretical studies of galactic dynamics. We here 1 : reanalyzed the VERA archival data using the self calibration method (hybrid v mapping), and found that clusters of maser features of S269 are distributed in i X muchwiderareathanthatinvestigatedinH2007. Weconfirmedthat,ifwemake r anarrowregionimagewithoutconsideringthepresenceofmultiplemaserspots, a and only the phase calibration is applied, we can reproduce the same maser structures in a maser feature investigated in H2007. The distribution extent of maser spots in the feature differs 0.2 mas from eastto west between our results and H2007. Moreover, we found that change of relative positions of maser spots in the cluster reaches 0.1 mas or larger between observational epochs. Email addresses: [email protected] (MakotoMiyoshi), [email protected] (YoshiharuAsaki),[email protected] (KeiichiWada), [email protected] (HiroshiImai) Preprint submitted toNew Astronomy January 23, 2012 Thissuggeststhatifonesimplyassumesthetime-dependent,widelydistributed maser sources as a stable single point source, it could cause errors of up to 0.1masintheannualparallaxofS269. Takingintoaccounttheinternalmotions ofmaserspotclusters,thepropermotionofS269cannotbedeterminedprecisely. WeestimatedthatthepeculiarmotionofS269withrespecttoaGalacticcircular rotation is ∼ 20 km s−1. These results imply that the observed kinematics of maser emissions in S269 cannot give a strong constraint on dynamics of the outer part of the Galaxy, in contrast to the claim by H2007. Keywords: ISM:star forming regions, ISM:individual (Sharpless 269), masers (H O), INSTRUMENTS:VERA 2 1. Introduction Very long baseline interferometric (VLBI) astrometry of the Galactic maser sources is expected to be a powerful probe for investigation of the structure and kinematics of the Milky Way. The Very Long Baseline Array (VLBA), the European VLBI Network (EVN), and the Japanese VERA (VLBI Exploration of Radio Astrometry) have been used to measure annual parallaxesand proper motionsofstar-formingregionsandredsuper-giantsinspiralarmsoftheMilky Way (Reid et al., 2009; Sato et al., 2010, and references therein). Reid et al. (2009) reported that these young sources have large peculiar motions (i.e., de- viations from circular rotation) as large as 30 km s−1. Such large peculiar motions are incompatible with the prediction from the conventional theory of quasi-stationaryspiralarms(Lin & Shu ,1964;Bertin & Lin,1996),butingood agreementwithrecenttheoreticalhigh-resolutionN-body/hydrodynamicalsim- ulations (Baba et al., 2009; Wada, Baba, & Saitoh, 2011). Baba et al. (2009) suggestedthatspiralarmsinthe MilkyWayarenotstationary;intheirsimula- tionsthearmsrecurrentlyformandvanish. Owingtogravitationalinteractions betweenthe time-dependentspiralpotentialandtheISM,theyshowedthatthe dense gas and star forming regions have large peculiar velocities. Among the star forming regions whose distances have been measured using VLBI,Sharpless269(S269)isofspecialinterest. Honma et al.(2007)(hereafter H2007)measuredtheannualparallaxandthesecularpropermotionofS269us- ingtheVERA,andreportedthatS269islocatedatthegalactocentricdistance, R,of13.1kpc, andits galactocentricrotationalvelocityisequalto(within 3%) that of the Sun with assumptions of R = 8 kpc, and Θ = 200 km s−1. From 0 0 these results, they concluded that the flat rotation of the disk of the Milky Way extends to 13 kpc from the Galactic center. On the contrary, Oh et al. (2010) observed star forming regions AFGL 2789 and IRAS 06058+2138using theVERA,andconcludedthattheirrotationvelocitiesaresignificantlysmaller thanthevaluederivedfromtheassumptionoftheflatrotation. Sincethegalac- tocentric distances of these two objects are 8.8 and 9.7 kpc, respectively, they suggested that there is a dip in the rotational velocity at around R ∼ 9 kpc. If those objects have large peculiar velocities as suggested by the theoretical 2 studies, their motions may not place strong constraints on the rotation curve. If S269 has almost the same rotational speed of the sun at R = 13 kpc with a small peculiar motion as suggested by H2007, then we should consider how we can reconcile this to other observations and theories. H2007 reported that the annual parallax, π, of S269 is 0.189±0.008 mas, which corresponds to 5.28+0.24 kpc. Given the source distance of 5.28 kpc, the −0.22 proper motion vector was estimated to be (v, v ) = (−4.60±0.81, −3.72± l b 0.72)kms−1. Theerrorsinthe annualparallaxandpropermotionsofS269are a factor of 3 to 5 smaller than those in recent other VLBI astrometric observa- tionsforstarformingregions(e.g.,Sato et al.,2010). H2007showedthatmaser sourcesof S269 have a simple disk-like structure alignedin the east–westdirec- tiononascaleof0.4masandaradialvelocitytotheLSRintherangeof19.0and 20.1kms−1. Previousobservations,howeverhavesuggestedmore complexand time-varyingstructurein awiderfield(Lo & Burke ,1973;Genzel et al.,1977; White et al.,1979;Cesaroni ,1990;Migenes et al.,1999;Lekht et al.,2001a,b). From the H O maser spectrum shape of the double or triple peaks around 2 V =14 to 22 km s−1, two possible structures were proposed by Lekht et al. LSR (2001a): one is anexpanding envelope,and the other is an edge-on(Keplerian) disk around a protostar. Lekht et al. (2001b) reported on a sinusoidal velocity drift at V ∼ 20 km s−1, and concluded that it is due to turbulent motions LSR of masing clouds because the estimated central mass is too small to be a pro- tostar. A wide field spatial distribution for the H O in S269 was shown with 2 the VLBI fringe rate mapping technique by Migenes et al. (1999). They found four velocity components at V = 16.5,17.3,19.4,and 20.7 km s−1 spread LSR over 1.3 arcseconds on the sky. They also reported that the spatial component V = 19.4 km s−1 was the strongest in their VLBI observation. However, LSR the velocity structure of these sources was not studied. Single-dish observation in July 1996 (Lekht et al., 2001b) obtained a single peak around 20.3 km s−1, which probably coincides with the 19.4 km s−1 peak in Migenes et al. (1999). Three-dimensional velocity estimate of star forming regions may cause big uncertainty in the derived three-dimensional motion in the Milky Way because wecanoftenfindoutflow-likestructureinmaserswhichmaynotreflectthemo- tions of the mass center. We have to search the velocity components carefully to estimate the motion of the mass centers. In addition, as demonstrated later spatial distributions of maser sources also affect the accuracy of VLBI astrom- etry even for the individual maser spots. Therefore we have to investigate the distribution in a wide field for star forming regions. In this paper, we focus on the structures of water maser source in S269 whether it is simple and stable enough to achieve the high accuracy in astrom- etry using the VERA archivaldata1. In section2 we describe the observational specifications. Maser emissions in a wide sky area (1.6 arcseconds square) as 1Rygletal.(2008)andRygletal.(2010)reportedthattheyfailedtomeasuretheannual parallaxforS269’smethanolmaserswiththeEVNwhiletheyobtainedtheannualparallaxes offiveother starformingregionswithaccuracyasgoodas∼0.02mas. 3 well as their time variations, or relative proper motions, of the maser spots are shown in section 4. We discuss comparison our results with H2007 and ef- fects of the internal motion of S269 on the Galactic dynamics in section 5 and summarize this study in section 6. 2. Observations As described in H2007, observations have been conducted over one year at 6 epochs on Nov 18 in 2004 (Day of Year, or DOY, 2004/323), and Jan 26, Mar 14, May 14, Sep 23, and Nov 21 in 2005 (DOY2005/026, DOY2005/073, DOY2005/134,DOY2005/266andDOY2005/326). TheH Omasersat22GHz 2 fromtheS269regionhavebeenobservedusinga2300kmscalearrayconsisting of four antennas of the VERA (Mizusawa, Iriki, Ogasawara,and Ishigaki-jima; see Kobayashiet al. 2008 in more detail) with the left hand circular polariza- tion for almost 8 hours. The recording bandwidth for the maser emission was 8 MHz at epochs 1, 4, 5, and 6, covering the velocity range of 112 km s−1. At epochs 2 and 3, the recording bandwidth was 4 MHz to cover the veloc- ity range of 57 km s−1. The recorded data was processed with the Mitaka FX correlatorto produce cross correlateddata with the 256 and 512 frequency channels for epochs 2 and 3, and the others, respectively, so that the frequency spacing is 15.625 kHz for all the epochs, corresponding to the velocity spacing of 0.21 km s−1. All the VERA antennas have a dual beam receiving system for phase-referencing (Kobayashi et al., 2008), and a closely located reference source,J0613+1306,wasobservedsimultaneouslyintheobservations. However, we did not carry out data analysis of the reference source because our purpose here is concentratedoninvestigationof the spatial andvelocity distributions of S269’s H O masers. 2 3. Reduction Methods In our data reduction, we could not obtain uniform signal-to-noise ratio through all of the epochs because the atmospheric attenuation were unexpect- edly highly variable dependent on observing season in Japan. In Table 1 and Figure1,we showthe variationsofsystemnoisetemperaturesforallthe anten- nas. Basically we followed a standard manner of spectral line VLBI data reduc- tions with AIPS (NRAO) package. However, because of insufficient (u, v) cov- erage of the VERA observations, we found many confusing emission peaks due to the side-lobes coupled with still-not-perfect amplitude calibration. In such a situation, mapping accuracy can depend on the details of data reduction. Here we note the details of our data reduction in order to assure the reproducibility of our mapping results. 4 Table 1 Time average of the system noise temperature of each stations for all the epochs. Epoch Mizusawa [K] Iriki [K] Ogasawara[K] Ishigaki [K] epoch 1 692 596 280 289 epoch 2 242 230 181 799 epoch 3 240 161 428 328 epoch 4 205 479 311 834 epoch 5 276 234 283 443 epoch 6 137 117 1050 417 3.1. Visibility Calibrations In order to get reliable images, we must perform calibrations of phase, am- plitude, andbandpass characteristicsof visibility data. For visibility-amplitude calibrations, we first performed the task ACCOR with SOLINT=0.1. Using auto-correlation spectra, the task ACCOR corrects amplitude errors in cross- correlation spectra suffered from sampling thresholds. Then we used the task APCAL to generate an amplitude calibration SN table, which includes the in- formationofantennagaincurve(fromGCtable)andsystemnoisetemperatures (from TY table) of each station. Furthermore, we applied the amplitude solu- tions obtained from the self calibration method using the task CALIB.2 For visibility-delay, rate, and phase calibrations, we used the task FRING with SOLINT=1.0 (SOLSUB=0.1) in order to obtain the clock offset and rate from the strong continuum source, J0530+13 inserted between S269 observing scans. As for fine phase calibrations, we relied on the self calibration solutions from the CALIB in AIPS at the last stage of calibrations (SOLINT=0.1, SOL- SUB=0.05). The selfcalibrationswereatfirstperformedatthe peak frequency channelscorrespondingto the V =19.5kms−1(248ch, 88ch,90ch, 266ch, LSR 239 ch, and 240 ch in frequency at respective epochs). The frequency and ve- locityresolutionswerecommonthroughallthe observationalepochs. Although these channelscontainedstrongmaseremissions,the maserstructureswerenot asinglespotbutatleasttwospotswithcomparableintensities. Thesolutionsof calibrations from these methods were applied to not only the reference channel but all of the velocity channels. As for bandpass calibrations we used the task BPASS in AIPS with total 2Ingeneral,measurementsofthesystemnoisetemperaturesandantennagainparameters areinsufficienttocalibratetheVLBIdata,becausetheerrorsinVLBIdataaresolargethat we cannot rely on conventional calibration and mapping methods often done in connected interferometers. Selfcalibrationinhybridmappingmethodprovidesuspowerfulsolutionsfor calibrating VLBI data. Today most VLBI maps are obtained after the calibration through hybridmappingmethod. FortheVERA,duetotheinsufficient(u,v)coverage,itissometimes difficulttogettheoptimizedsolutionwithhybridmapping. 5 power spectra of calibratorcontinuum sources and got the amplitude bandpass characteristics. After these calibrations, we performed corrections of velocity- shift due to diurnal rotation of the earth using the task CVEL in AIPS. 3.2. How to Make Maps 3.2.1. Search for Masing Regions To find the whole region of H O maser emissions, instead of fringe rate 2 mapping, we used wide area synthesis imaging with low spatial resolutions. We performed synthesis imaging of the 2 × 2 arcseconds area in all velocity channels of four epoch’s data (the 1st, the 3rd, the 5th, and the 6th epochs). The 2×2 arcseconds area was covered with 4096×4096 grids. Namely, each cell size is about 0.5 mas. From these coarse maps, we selected positions for synthesis imaging with fine spatial resolution. We selected positions where the firstCLEANcomponentsofthedifferentepochs’datacoincidedwitheachother within 0.1 arcsecond. In addition to the positions, we added 0.3 arcseconds squareareasaroundthethreepositionswherestrongmaseremissionswerefound (around positions C, D, and E shown in Figure 9). We thus selected 42 of 100×100 mas areas to be mapped with higher spatial resolution. 3.2.2. Fine Synthesis Imaging of the Selected 42 Areas Wemappedthe42squareswith4096×4096grids. Eachcellsizeis24.4µarc- seconds. For the synthesis imaging, we used the task IMAGR in AIPS with parametersNITER=3000,GAIN=0.01,andFLUX=15mJy. The selectedmin- imumflux densitylevelofa CLEANcomponent, FLUX=15mJyis presumably lower than the array sensitivity. Because the absolute flux density of the data has anuncertaintydue to insufficient amplitude calibrations,we used the lower level in order to achieve an adequate subtraction. 3.3. Measurements of Maser Positions In order to avoid subjective selections of maser spots, we ran the task SAD in AIPS automatically with its default parameters. We divided the respective 4096×4096gridareasinto25sub-areas(4×4massquare),andranthetaskSAD in each sub-area and measured maser spot positions. This selection method partially failed to select some maser positions around strong masers because suchregionsincludenotafewnumbersofhighlevelpeaksduetotheside-lobes. However, we adopted the automatic SAD selection to prioritize objectivity in selecting maser spots. By the SAD selection we found a lot of peak positions. Thenumbersofpeakswithfluxdensity ≥6σ aregiveninTable2. Presumably, side-lobesor notrealmaserspots mingle amongthe selectedpeaks by the SAD method. It was quite difficult to select only real maser spots from these maps. To completely avoid selecting peaks that are not real, criteria other than their signal-to-noise ratios (SNR) are required. 6 Obs Epoch Peak Number 1σ Noise Level (>6σ) (Jy/Beam) 1 238 1.37 2 354 4.82×10−1 3 379 4.88×10−1 4 660 1.22 5 177 1.09 6 789 5.78×10−1 Table 2 Numbers of the maser emission peaks selected by the SAD method. 1σ noise levels were measured from the V = 18.5 km s−1channel at the LSR 100×100 mas field centered at (980 mas, 370 mas) in the map of Figure 9. 4. Results Section 4.1 shows the cross power spectra of the H O maser in the present 2 dataanalysis. Sections4.2and4.3showthespatialdistributionsoftheH Omaser 2 emissions and its time variation in the individual maser spot clusters and the wholeareaofS269. Relativepropermotionsofthe maserclustersarepresented in section 4.4. Here we define a “maser spot” as the origin of a maser emission in a single velocity channel map, and a “maser feature” as a group of maser spots with different velocities gathered at a common place. We also use the term “maser cluster” as a group of maser features with a common motion. 4.1. Cross power spectra of H O masers in S269 2 Figure 2 shows the cross power spectra of the H O maser emissions in 2 S269 obtained with the Mizusawa–Iriki baseline (1300 km length) for all the six epochs. It is important to note that there are complex maser emission peaks inthe radialvelocityrangefrom 8 to 20 kms−1, andthat the line profile is changed on a time scale shorter than one year, as reported by Lekht et al. (2001a). Themostprominentemissioncanbe seenatV of19.5kms−1,and LSR the line profile of H O maser spectrum changes significantly in one year. Note 2 alsothatthereareseveralemissionpeaksinthe spectraatV of17kms−1at LSR epoch2,V of18kms−1atepoch3,andV from8to11kms−1atepochs5 LSR LSR and 6. The signal-to-noise ratio (SNR) of the cross power spectrum at epoch 1 seems worsethan those at the other epochs. This is mainly because the system noise temperatures of these two stations at epoch 1 were unusually a factor of 2 to 5 higherthanthose atother epochs andalsobecause the observingtime of S269 was about half of those in other epochs. In the cross power spectrum at epoch 1, there are several peaks between V =0 to 5 km s−1, but they are not maser emissions. This is due to strong LSR artificial signals at Iriki station. We found the artificial signals at 21.233 GHz (corresponding to V = 35 km s−1), 22.235 GHz (corresponding to V = LSR LSR 0 km s−1), and 22.237 GHz (corresponding to V = −22 km s−1) in sky LSR frequency, one of which caused the peaks at V from 0 to 5km s−1. LSR 7 4.2. Arcsecond scale distribution of H O masers in S269 2 Figures 3, 4, and 5 show the spatial distribution of the maser spots and position-velocity diagrams for all the six epochs. We identified seven maser groups (A through G). All of them show strong time variation. At epoch 1, only group C was strong (SNR >40). C and E were strong at epoch 2, D and E at epoch 3, C and D at epochs 4 to 6. Groups C, E, F and G all lay in V ranging from 18 to 20 km s−1. The maser emissions from V = 8 to LSR LSR 14 km s−1 came from a single group denoted as D. The distribution of masers was qualitatively consistent with the map obtained by Migenes et al. (1999). They identified four sources at 16.5, 17.3,19.4 and 20.7 km s−1, which roughly coincide with groups A, C, E and G. They used a fringe-rate mapping method to obtain the positions ofthe four peaks in their spectrum. Therefore,they did not obtain the velocity structures of the individual sources. The fact that they didnotreportthe velocitydistributionwithin eachmaser groupdoes notmean the observed groups were single points. 4.3. Structure of individual maser groups Figure6showstheinternalmaserstructuresofgroupsAtoG.Thestructure of A is taken from epoch 6, that of B is from epoch 4, those of C, D, and E are from epoch 3, and those of F and G are from epoch 6. Each maser group consists of one or two features. One feature typically has a spatial size over 1 mas and a velocity range of 2 km s−1. Groups D and E consist of two features, while other groups consist of one feature. As shown in the panel for group A, the typical beam size is 1.4×0.9 mas. For maser spots in group C, SNR is higher than 100. From the high SNR, we can expect that the error of the relative position is small down to 0.01 mas. For other maser spots, SNR is in a rangeof10–20. This implies that apossible positionerroris 0.1mas orlarger. Group C is the brightest among the clusters A-F, and it should correspond to the maser cluster studied in H2007. H2007 detected the maser spots with V from 19.0 to 20.1 km s−1, which were distributed over an angular range LSR of0.4masinthe east–westdirection. While wedetectedfromgroupC inwider velocity range of 18.8 to 21.4 km s−1 as shown in Figure 6. As shown in Fig. 7 (a), the maser spots in the same velocity range as reported by H2007 show a more compact distribution within an angular range of 0.2 mas. We discuss the difference between the two maps in section 5.1. The left two panels of Figure 8 show the relative positions of the maser spots in group C with respect to the position of the maser spot at V = LSR 19.5 km s−1. Positional deviations at epochs 1 and 4 are larger than those at other epochs: positions at epochs 2, 3,5, and6 are consistentwith one another within the positional accuracy of 0.05 mas. Judging from the system noise temperatures, the large positional deviations at epochs 1 and 4 were caused by the bad atmospheric conditions as noted in section 2. The plot shows that changeofrelativepositionsofmaserspotsintheclusterreaches0.5masbetween observationalepochs (see section 4.1 on possible errorsin the annual parallax). 8 4.4. Relative Proper Motions of Maser Clusters Figure10showsthe propermotionsofmaserclustersrelativeto the V = LSR 19.5 km s−1 spot in group C. First, we searchedproper motions of maser spots from groups A to G (Figure 9). Among the selected peaks, we searched for relative proper motions of maser spots with the following three criteria. 1. Peaks which are higher than 20σ noise level at each epoch. 2. Peaks whose displacements are less than a corresponding proper motion of 3.5 mas yr−1. 3. Peaks whose velocity shifts are less than 0.5 km s−1. For peaks in the two fields of groups F and G, where the maser spots were well isolated, we selected peaks which are higher than 7σ noise level at each epoch. Because these two maser groups show relatively weak emissions but the existence of the groups is certain, these masers cannot be neglected in order to investigatetheinternalmotionsofthemasersinS269regions. Wefound22sets of proper motions using the first criterion. Using the looser criterion on SNR, wefoundmore4setsofpropermotionsand1setofpropermotionfromthetwo groups F and G respectively. In Table 3 we show the detected proper motions of maser spots with these criteria. After detecting propermotions ofmaserspots,we combinedallmaserspots data showing common proper motions in order to obtain independent motions. Firstofall,fromthepropermotionsdataset,weomitteddatausingpositionsof epoch1and4asthesetwoepochswereunderfairlybadatmosphericconditions. We combined and averaged proper motions and velocities of maser spot data whose velocity difference was within 1 km s−1 in order to derive independent motions of maser clusters. Table 4 gives the parameters of derived proper motions of maser clusters. Apparently, the maser clusters in groups C, D, E and G form an arc-like struc- turewith the propermotionsimplying anexpanding shell. Since the numberof clusters is limited, it is possible to judge that this structure and motion could be just a coincidence. 5. Discussion We have re-analyzed the VERA archival data of S269. We found that (a) maser emissions in S269 are distributed over around 1.6 arcseconds, (b) the maser emissions are found from a wide radial velocity range between 8 and 20 km s−1, (c) there are multiple maser groups in the 1.6 × 1.6 arcsecond area at aroundthe radialvelocityof19.5kms−1,and(d) the structureofthe brightest source(groupC)isdifferentfromthesinglesourcereportedinH2007. Insection 5.1, we discuss the origin of the discrepancy in the obtained maser structure betweenourresultandthatofH2007. Insection5.2,wediscusstheimplication of the internal maser motions on the constraint on the galactic rotation curve of the Milky Way. 9 5.1. Comparison with the map of H2007 In Figure 7, we show the distribution of maser spots of groupC and that of H2007,forthe sameepoch. Bothshowvelocitygradientsintheeast-westdirec- tion, but H2007’s spots in the velocity range of V =19.0to 20.1 km s−1 are LSR distributed over 0.4 mas, while those in our result have a more compact distri- bution within 0.2 mas. We consider the possibility that the difference between our methods of analysis and those of H2007 caused the difference in the im- ages. The primary difference between the two methods is that we have made theimagefromhybridmapping,while H2007usedthe phase-referenceddatato determinetheabsolutepositionsofthemaserspots. Thisdifferenceofthemap- ping techniques themselves does not seem to cause significantdifferences in the determination of the internal structure. Rather, the reasons for the difference lie in differences in the treatment of the data. We noted the following two major differences. First, H2007 did not find groupsEandG,whichareinthesamevelocityrangeasgroupC.Second,when they performed the analysis of the data, they applied corrections for the atmo- spheric delays of stations to the visibility data in order to obtain a reasonable result. They noted: “To calibrate them, residual zenith delays were estimated as a constant offset that maximizes the coherence of the phase-referencedmap. Typical residuals of zenith delay are 1 to 5 cm, but in the worst case (during the summer at Ishigaki-jima station) it was as large as 20 cm”. Theymaximizedthe coherencebyadjustingthe residualatmosphericexcess path-length3, L of individual stations, under the following assumptions: s 1. Theexcesspath-lengthδL tobeadjustedatstationSisδL =L ×(sec Z − S S S S269 sec Z ), where Z is the zenith angle of source X at station S and L ref X S is the residual atmospheric zenith delay at station S. 2. The residualatmosphericzenith delay L is constantduring one observa- S tional session. 3. Thecorrectestimateoftheresidualexcesspath-lengthL givesthecorrect S map of S269. We investigated how these assumptions can change the map. For this pur- pose we created a map using the method which is effectively equivalent to that usedinH2007. Thedifferencebetweenthestandardimagingmethodandthisis summarizedas follows. In the standardimagingmethod one uses fine solutions ofboth amplitude and phase fromself-calibrationwith an optimizedmodel im- age,howeverhereweusedself-calibrationonlyforphasesolutionsusingapoint asthe imagemodel. Calibrationofphase φis equivalentto thatofexcesspath- length L due to the equation φ = 2πL/λ (where λ is observing wavelength). We assumedthat the source in one velocity channel has a single point when we performed the phase self-calibration. We limited the imaging area to a narrow region(100×100mas span). If they performedwide-areaimaging,they should have found multiple sources as we found. (Hereafter we call the map obtained 3Inotherwords,”residualatmosphericdelay”. 10