Publications of the Astronomical Society of the Pacific 104: 1173-1176, 1992 December

Variation of the Radial Velocity of Epsilon Cygni A

R. S. McMillan, P. H. Smith, T. L. Moore, and M. L. Perry Lunar & Planetary Laboratory, Space Sciences Building, University of Arizona, Tucson, Arizona 85721 Electronic mail: [email protected] Received 1992 August 4; accepted 1992 September 16

ABSTRACT. A series of 217 measurements of the radial velocity of the KO— III star Epsilon Cygni A were made between 1987 May 16 and 1992 May 17 with an uncertainty per observation of ±12 m s_1. The results are inconsistent with the listed status of Epsilon Cygni A as a spectroscopic binary with an amplitude of 5 km s_1 and a period of 20 days. Instead, they indicate a sustained drift of —60 m s_ 1 yr-1. If the companion inducing this deceleration is in a circular orbit, the period must be at least 15 yr. Lower limits on the companion's mass are presented as functions of period and primary mass. The companion is probably more massive than a planet.

1. INTRODUCTION files by Gray ( 1982), who did not report evidence for dou- bled spectral lines or temporal changes of line asymmetry We are conducting an observing program to detect large that would indicate a spectrum of a companion. However, planets orbiting stars by measuring the periodic variations Gray was not specifically looking for a secondary spec- such companions induce in the star's Doppler shifts. We trum, especially not one at nearly the same radial velocity started observing Epsilon Cygni A in 1985 to demonstrate as Epsilon Cyg A. lanna and Culver's (1985) trigonomet- our capability to improve the orbit of a spectroscopic bi- ric parallax and the visual magnitude in Blanco et al. nary using a schedule of frequent, accurate observations. (1970) indicate an intrinsic luminosity more than a mag- Epsilon Cygni A [=HR 7949A=ADS 14274A=HD nitude fainter than that normally expected for an early Κ 197989 = Gliese (1969) 806.1 A] is of spectral class K0— giant star (Allen 1973). The angular diameter was mea- III (Keenan and Pitts 1980). Accurate radial-velocity ob- sured by Koechlin and Rabbia (1985) and Mozurkewich servations of it are feasible because it is bright, has many et al. ( 1991 ), with results that were reasonable for a single sharp lines, and is conveniently placed for observing from Κ giant at that distance. Kyrolainen et al. (1986) list sur- our latitude for a large part of the year. The companion Β face gravity and effective temperature within the normal is "optical," that is, not gravitationally associated (Hoffleit range for a star of this spectral type. Therefore, a close and Jaschek 1982). The latter authors also list a compan- companion to A, if present, must have a comparatively ion "C" with a proper motion similar to that of "A" at a faint luminosity. separation of 78 arcsec and a visual magnitude of 13. Cur- tis and Burns (quoted by Campbell and Moore 1906) first 2. OBSERVATIONS announced that A's radial velocity was variable. Harper Descriptions of our instrument and technique are given (1920) analyzed 68 measurements of radial velocity, 50 of by McMillan et al. (1985, 1986, 1990, 1993), McMillan which were made by him, and derived a tentative period of -1 and Smith (1987), and Smith et al. (1987). A brief sum- 20 days and an amplitude of 5 kms from the best 41 mary of the instrument's precision and accuracy is given measurements that spanned 1.3 yr. Harper's data show a -1 below. probable error of ±1.5 kms , which according to him Epsilon Cyg A was observed 246 times between 1985 was unexpectedly high for "...the good quality of the October 22 and 1992 May 17 UT inclusive, with the 0.9-m lines." A phase diagram of his data folded modulo his telescope of the Steward Observatory on Kitt Peak. Expo- period (not reproduced here) admits a curve with a semi- 1 sure times ranged from 15 min to 1 h. The observation amplitude of 1 km s" drawn through considerable scatter. series was interrupted in 1986 to upgrade the instrument. Campbell and Moore (1928) listed Epsilon Cyg A in a This created a boundary in the time series across which compilation of stellar radial velocities, showing a full span -1 data cannot be compared easily. Therefore we present two of 8 kms among 12 additional measurements spread separate series of observations with different instrumental over 23 years. Woolley et al. (1960) and Jones and Wool- "zero points." ley (1961) list a few measurements of its radial velocity In Table 1, 29 observations are given for 18 nights, 15 of and acknowledge the above studies that had "well estab- which were consecutive. The data span 33 days. The ve- lished" the radial-velocity variation. However, our inspec- locities are referred to a "zero" arbitrarily assigned to one tion of Woolley et al.'s few measurements does not support of the observations. The standard deviation of the data Harpers' claim. during the first 15 consecutive nights is ±30 m s-1. (All Many studies have been made since then of the proper- error estimates in this contribution correspond to the stan- ties of the atmosphere of the star, some of them with very dard deviation per sample.) The observations in Table 1 high spectroscopic-resolving power and signal-to-noise ra- are illustrated in Fig. 1. This star could not have been tio. One of the most relevant to the question of whether varying in the manner described by Harper when we ob- star A is binary was the high-resolution study of line pro- served it. Therefore, Harper's variation, if real, could not 1173 © 1992. Astronomical Society of the Pacific

Table 1 Table 2 Differential Radial Velocities of Epsilon Cyg; 1985 Differential Radial Velocities of Epsilon Cygni; 1987-1992 V -1 V V V Time (ms ) Time (m s_1) MJD (ms-1) MJD (ms-1) 60.247 -11 69.082 + 86 46931.4556 61.093 + 9 69.165 +91 -28.76 47691.4173 -141.45 61.126 -23 70.149 + 54 46931.4688 -15.76 47691.4389 -146.93 61.158 +72 71.102 + 53 46931.4833 -16.86 47691.4600 -153.36 61.194 +46 72.098 + 56 46945.4306 -14.42 47695.4433 -171.97 62.109 + 14 72.142 +43 46945.4465 -0.12 47695.4644 -179.52 63.158 +47 73.100 +20 46947.4583 -14.05 47696.3864 -170.88 63.201 46957.4667 -5.95 47696.4205 -150.97 + 19 73.142 +49 46959.4681 0.00 47790.2349 64.131 + 12 74.195 + 11 -172.77 64.169 +23 91.069 -47 46974.4451 -25.53 47790.2565 -184.17 65.163 -2 92.109 -19 46977.3465 + 16.44 47790.2787 -160.76 66.184 +0 (ref.) 92.156 -22 46977.3764 -10.97 47794.1075 -149.19 67.192 -8 93.112 -146 46977.4178 + 1.97 47794.1285 -151.40 68.091 +40 93.158 -102 46978.4694 -35.32 47795.1238 -150.80 68.157 + 19 47073.0932 -57.83 47795.1894 -156.31 47073.1161 -64.21 47798.1043 -151.25 "Time"=Julian Date minus 2,446,300.5. 47073.1386 -64.95 47809.0820 -94.08 47073.1611 -76.31 47809.0960 -109.90 47073.1829 -75.91 47809.1100 -103.11 47073.2050 -84.77 47810.0782 -143.27 have been due to orbital motion. Figure 1 also suggests a 47088.1463 -60.48 47810.0922 -134.26 deceleration of the star along the line of sight. However, 47088.1688 -62.31 47810.1063 -139.74 considering the primitive state of our equipment at that 47088.1965 -64.79 47812.1358 -141.26 time we do not attach great significance to the difference of 47089.1118 -70.84 47812.1568 -141.81 47107.0794 -88.66 47814.0666 -152.15 velocities between the two observing runs in 1985. 47107.1089 -88.52 47814.0878 -160.29 In Table 2, 217 observations span 5.0 yr. The much- 47107.1466 -88.33 47838.0719 -140.32 improved instrument was not modified during this interval 47108.1062 -69.83 47838.0931 -138.39 and calibrations were rigorous. The stability of the velocity 47111.0913 -78.37 47843.0792 -124.56 metric over this interval is presented and analyzed by Mc- 47139.0668 -60.49 47843.1011 -110.62 Millan et al. (1993). No celestial sources are used for cal- 47277.4843 -67.89 47845.0601 -150.20 47278.4650 -56.62 47845.0812 -144.46 ibration or velocity correction. Calibrations of the interfer- 47280.4708 -62.94 47988.5253 -196.11 ometer are based on measurements of emission lines from 47281.4646 -80.44 47989.5291 -200.28 an Fe-Ar hollow cathode lamp. The properties of the in- 47284.4767 -82.77 47991.5131 -218.53 terferometer are determined in absolute physical units, 47285.4036 -78.61 47994.5000 -178.12 which provide absolute vacuum wavelengths of the inter- 47285.4348 -84.91 47996.5202 -202.66 47286.4436 -58.13 48017.5030 -197.98 ference orders constructively transmitted by the interfer- 47287.4064 -90.70 48018.4741 -194.83 ometer. The short-term consistency of the individual cali- 47287.4486 -89.34 48020.4649 -194.91 brations made immediately before and after each setting on 47288.4687 -78.17 48022.4707 -184.83 a star is ±3-4 ms_1 (McMillan et al. 1990). Our long 47289.4356 -87.04 48027.4968 -187.94 series of observations of sunlight reflected off the Moon's 47289.4682 -95.00 48027.5047 -184.08 47305.3992 -97.75 48028.4953 -196.00 surface shows that periodic errors of the instrument are 47307.4011 -71.07 48029.4982 -177.69 47308.4011 -93.41 48029.5063 -153.01 47322.4700 -111.63 48029.5143 -87.39 47326.3491 -99.81 48049.4069 -177.49 47328.3939 -125.24 48163.1058 -113.49 47328.4365 -130.81 48168.1617 -223.14 47329.3824 -161.05 48168.1828 -220.66 ARV 47332.4752 -117.19 48170.1399 -234.70 (ms"1) 47333.4958 -124.08 48170.1540 -223.36 47334.4740 -136.65 48171.0807 -247.07 47335.4877 -138.82 48171.0958 -254.40 ε Cyg Α Κ0-ΙΠ 47336.4265 -96.58 48172.1883 -240.62 1985 Get 22-Nov 24 47336.4697 -118.63 48193.0903 -225.97 -150 - 47338.4592 -111.51 48193.1187 -223.51 47338.4839 -132.91 48193.1481 -215.09 -200 L 46370 46380 46390 47340.4808 -134.82 48193.1803 -202.27 Modified Julian Date —> 47341.4899 -107.52 48195.0846 -232.01 47342.4852 -114.09 48195.1063 -230.73 Fig. 1—Radial velocities of Epsilon Cygni A in the autumn of 1985 vs 47343.4859 -93.99 48195.1283 -239.57 modified Julian date (MJD) = Julian Date minus 2400000.5. The velocity 47345.4084 -123.86 48195.1514 -239.73 scale is referred arbitrarily to one of the observations, indicated in Table 47345.4412 -111.99 48200.1124 -222.80 1. A typical error bar is indicated on one of the data points.

Table 2 ( Continued )

V V MJD (ms"1) MJD (ms"1) 47346.4819 -94.45 48200.1266 -230.16 (m/s) 47352.4189 -107.95 48223.0507 -244.94 47353.4828 -135.93 48223.0653 -214.66 47424.1053 -115.17 48223.0796 -223.49 47424.1477 -86.70 48225.0689 -201.01 47000 47200 47400 4760,0 47800 48000 48200 48400 48600 47424.1942 -71.02 48225.0831 -204.24 Modified Julian Date 47424.2408 -78.16 48225.0973 -220.23 47424.2878 -80.33 48229.0587 -214.46 47426.2352 -110.66 48347.4829 -273.41 Fig. 2—Radial velocities of Epsilon Cygni A from 1987-1992 vs MJD. The velocity scale is referred arbitrarily to one of the observations. A 47431.1105 -61.81 48347.5005 -261.72 _1 47431.1573 -74.15 48348.5154 -275.28 typical error bar is ± 12 m s . 47432.1043 -75.00 48350.5221 -265.43 47432.1547 -86.56 48371.4668 -227.04 47452.0940 -75.91 48371.4844 -213.06 tion series within single nights to estimate this internal 47452.1347 -81.27 48372.4486 -255.18 consistency, the quantity we call "precision." This is inde- 47452.1796 -93.66 48372.4697 -227.67 pendent of noise models and lamp calibrations, and in the 47453.0907 -78.00 48403.4804 -269.32 case of Epsilon Cyg A is about ±10 m s-1. Adding this 47453.1355 -82.11 48404.3632 -261.35 _1 47453.1809 -70.17 48404.3816 -277.04 quadratically to the calibration error ( ±6 m s ) we ob- 47455.0820 -64.21 48405.4254 -275.05 tain an estimate of the external errors of the reduced ob- 47455.1205 -73.23 48429.4632 -283.47 servations of Epsilon Cyg A: approximately ±12 ms-1. 47481.0682 -74.41 48429.4772 -295.27 This estimate is based on the assumption that the star does 47482.0683 -75.73 48430.4685 -295.01 47484.0581 -58.75 48430.4828 -289.80 not vary on short time scales, so it is an upper limit to our 47487.0521 -54.32 48432.4563 -310.18 measurement error. 47508.0649 -60.37 48432.4705 -295.54 Figure 2 illustrates the radial velocity as a function of 47509.0938 -78.97 48524.2091 -288.40 time between 1987 May 16 and 1992 May 17 UT, inclu- 47510.0730 -86.82 48525.1854 -283.09 sive. The long trend is — 60 m s_1 yr_1 and does not ap- 47511.0673 -87.47 48549.0615 -265.03 47515.0645 -117.96 48549.0798 -263.51 pear in our observations of other target stars or the lunar 47601.5272 -152.99 48550.1413 -250.97 crater Mösting A (McMillan et al. 1993). In addition, 47603.5158 -188.25 48550.1555 -257.01 faster variations are present in Fig. 2. These are qualita- 47604.5174 -168.28 48577.0486 -248.21 tively consistent with the instability of the radial velocities 47605.5130 -168.09 48577.0628 -241.10 of other Κ giant stars (Smith et al. 1987; Cochran 1988, 47635.4689 -157.46 48577.0769 -231.47 Walker et al. 1989) and also do not appear in our obser- 47635.4906 -158.56 48733.4960 -337.63 47664.3965 -156.49 48755.4406 -368.90 vations of the Moon or other stars. We postpone study of 47664.4180 -147.81 48755.4557 -365.15 the rapid variations to a separate paper on the intrinsic 47664.4392 -156.24 48755.4698 -368.74 properties of Κ giants. Radial velocities in the historical 47665.4755 -158.90 48759.4131 -360.77 literature are too noisy to contribute to the present inves- 47668.4582 -150.41 48759.4275 -352.71 47668.4793 -165.25 48759.4421 -350.39 tigation. 47669.4706 -153.50 48759.4690 -381.66 47689.4457 -175.37 48759.4833 -372.10 3. INTERPRETATION -175.02 47689.4670 Some constraints can be placed on the mass and orbit of the companion responsible for the sustained deceleration. An elementary calculation with reasonable assumptions rules out the common proper-motion companion Epsilon Cygni C, 78 arcsec away, as the cause. In addition, the less than ±4 m s_1 for periods between 100 and 1000 days. period must be at least 15 yr, otherwise more curvature in On longer time intervals the instrument shows no drift the long trend would be noticeable. The lower limit on the greater than 1.5 ms-1 yr_1. These short-term and long- companion's mass can be shown in a mass versus period term calibration errors add quadratically to an estimate of diagram if we assume circular orbits and that the primary ±6 m s_1 for our calibration accuracy. and secondary are presently observed in the geometry that Observations of stars are affected by additional sources maximizes deceleration along the line of sight. Figure 3 of noise that in the case of the emission lamp are insignif- displays the minimum mass of the companion versus the icant. Photon statistics and CCD readout noise imply the period of its orbit about Epsilon Cygni A. Companion velocity noise of our individual observations of Epsilon masses less than 8 Jupiters are already ruled out. No useful Cyg A should be ±5 m s_1. However, an empirical deter- upper limit can yet be placed on the mass of the compan- mination of the internal consistency of the observations is ion; it could be as massive as the Sun and still be consistent more relevant. We use the standard deviations of observa- with the available information. If it is a white dwarf it

Period (years) Blanco, V. M., Demers, S., Douglass, G. G., and Fitzgerald, M. 10 20 40 80 100 160 240 320 500 P. 1970, Publ. US Naval Obs., 21, Second Series log m Campbell, W. W., and Moore, J. H. 1906, Lick Obs. Bull., 4, 96 Campbell, W. W., and Moore, J. H. 1928, Publ. Lick Obs., 16, 305 Cochran, W. D. 1988, ApJ, 334, 339 Gliese, W. 1969, Veröif. Astron. Rechen-Inst. Heidelberg, No. 22 Gray, D. F. 1982, ApJ, 262, 682 Harper, W. E. 1920, Publ. Dom. Obs. Ottawa, 4, 323 Hoffleit, D., and Jaschek, C. 1982, The Bright Star Catalogue (New Haven, Yale University Observatory) lanna, P. Á., and Culver, R. B. 1985, AJ, 90, 1870 Jones, D. H. P., and Woolley, R. v. d. R. 1961, R. Obs. Bull. log Ρ (Herstmonceux), Ser. Ε, No. 33 Keenan, P. C, and Pitts, R. E. 1980, ApJS 42, 541 Koechlin, L., and Rabbia, Y. 1985, A&A, 153, 91 Fig. 3—Minimum mass of the companion vs period of a circular orbit, Kyrolainen, J., Tuominen, I., Vilhu, O., and Virtanen, H. 1986, for four different masses of the primary, Epsilon Cygni A. The left and A&AS, 65, 11 bottom scales are labeled logarithmically; at top and right are tic marks in McMillan, R. S., Moore, T. L., Perry, M. L., and Smith, P. H. linear units. 1993, ApJ, 403, No. 2, in press McMillan, R. S., and Smith, P. H. 1987, PASP, 99, 849 McMillan, R. S., Smith, P. H., Frecker, J. E., Merline, W. J., and could have escaped notice in conventional observations Perry, M. L. 1985, in Proc. IAU Colloq. No. 88, Stellar Radial done for other purposes. We will continue to observe the Velocities, ed. A. G. Davis Philip and D. W. Latham (Schenectady, Davis), p. 63 primary star in hopes of seeing the deceleration change. McMillan, R. S., Smith, P. H., Frecker, J. E., Merline, W. J., and We are grateful to the late Krzysztof M. Serkowski Perry, M. L. 1986, Proc. SPIE, Vol. 627 (Instrumentation in (1930-1981), as well as R. L. James, W. J. Merline, and Astronomy-VI ), ed. D. L. Crawford, (Bellingham, WA, SPIE), p. 2 M. S. Williams. This investigation was supported by McMillan, R. S., Smith, P. H., Perry, M. L., Moore, T. L., and NASA grants NAG2-52 and NAGW-1283, NSF grants Merline, W. J. 1990, Proc. SPIE, Vol. 1235 (Instrumentation AST-8403285, AST-8714817, and AST-9016572, private in Astronomy-VII), ed. D. L. Crawford (Bellingham, WA, donations to T. Gehrels' Spacewatch Project, and the allo- SPIE), p. 601 cation of telescope time by the Director of Steward Obser- Mozurkewich, D., Johnston, K. J., Simon, R. S., Bowers, P. F., vatory. Guame, R, Hutter, D. J., Colavita, Μ. M., Shao, M., and Pan, X. P. 1991, AJ, 101, 2207 Smith, P. H., McMillan, R. S., and Merline, W. J. 1987, ApJ, 317, L79 REFERENCES Walker, G. A. H., Yang, S., Campbell, B., and Irwin, A. W. 1989, ApJ, 343, L21 Allen, C. W. 1973, Astrophysical Quantities (London, Athlone), Woolley, R. v. d. R., Jones, D. H. P., and Mather, L. M. 1960, R. 3rd ed. Obs. Bull. (Herstmonceux), Ser. Ε, No. 23