THE MONOPERIODIC SCUTI STAR UY CAMELOPARDALIS: AN ANALOG TO SX PHOENICIS AND RR LYRAE VARIABLES A.-Y. Zhou and Z.-L. Liu

The Astronomical Journal, 126:2462 2472, 2003 November # 2003. The American Astronomical Society. All rights reserved. Printed in U.S.A. E THE MONOPERIODIC SCUTI STAR UY CAMELOPARDALIS: AN ANALOG TO SX PHOENICIS AND RR LYRAE VARIABLES A.-Y. Zhou and Z.-L. Liu National Astronomical Observatories, Chinese Academy of Sciences, 20A Datun Road, 100012 Beijing, China; aiying@bao.ac.cn Received 2003 April 18; accepted 2003 July 30 ABSTRACT We present the results of a 4 year photometric study of the high-amplitude Scuti (HADS) star UY Camelopardalis. Analysis of the available data from 1985 to 2003 shows that UY Cam is monoperiodic. Fourier solutions for individual data sets do not reveal period changes in the star. Although forced parabolic fits to the O C residuals indicate a measurable period change, the distribution of the data points in the O C diagram and the deviations between fits and observations suggest that the existence of the period change still has not been established. We demonstrate the presence of amplitude variations from cycle to cycle and on longer timescales. The pulsation amplitude seemed to change from 1985 to the 2000s, but it remained constant in 2000 2003. UY Cam is located in the upper portion of the Scuti instability region. Its photometric properties and estimated physical parameters reveal that UY Cam is an interesting object in terms of its poor metallicity, long period, high luminosity, low surface gravity, and large radius among the HADS stars. UY Cam could be a younger (0.7 0.1 Gyr) Population I HADS star with low metal abundance (Z = 0.004) evolving on its post main-sequence shell hydrogen burning evolutionary phase. UY Cam intervenes among the Population I/II HADS and type c RR Lyrae variables. These characteristics suggest the star to be an analog of HADS, SX Phoenicis, and RRc variables. Key words: RR Lyrae variable Scuti stars: individual (UY Camelopardalis) stars: oscillations techniques: photometric On-line material: color figures 1. INTRODUCTION The Scuti stars are regularly pulsating variables situated in the lower classical Cepheid instability strip, on or near the main sequence. In general, the period range of Scuti stars lies between 0.02 and 0.25 days and the spectral types range from A2 to F2. The majority of Scuti stars pulsate with a number of nonradial p-modes simultaneously excited to low amplitudes, but some are (pure) radial pulsators with larger amplitudes, and others pulsate in a mixture of radial and nonradial modes. We launched a program dedicated to the investigation of poorly studied Scuti stars in 1996. UY Camelopardalis (=HIP 39009 = GSC 04369-01129, J2000 =07 h 58 m 5990, = +72 47 0 23>7, V=11.44 mag, A3 A6 III; Rodríguez, López-Gonález, & López de Coca 2000) belongs to the subclass of high-amplitude Scuti (HADS) stars, and there is less knowledge about its nature. Therefore it was selected as one of the targets for the program. UY Cam was discovered to be a variable by Baker (1937), who classified the star as a Cepheid. Observations on five nights in 1962 led Williams (1964) to resolve a period of 0.267 days. Williams suggested that UY Cam was a type c RR Lyrae star. He pointed out that further observations with a larger telescope (than his 20 inch [0.5 m] one) were necessary to establish the existence of possible cycle-to-cycle changes in the light curves and to study their details. During 1946 and 1965, Beyer (1966) collected 21 maxima and derived a constant period. According to Beyer, the light curve was unstable and its amplitude changed from 0.17 to 0.50 mag. However, the potential instability of the light curve of UY Cam, suspected by both Williams and Beyer, was not observed late in 1985 by Broglia & Conconi (1992), who obtained a total of five nights of Johnson BV data covering five maxima. 2462 Considering the insufficiency of previous data in revealing the possibile amplitude variations, we observed UY Cam from 1999 to 2003 with an emphasis on time resolution and a longer time baseline. We collected a number of CCD and photoelectric photometric data in the Johnson V band. In this paper, we present the results of a comprehensive analysis aimed at the short-term and/or secular behavior of the light variations using all available data. Section 2 contains an outline of the observations and data reduction. Section 3 is devoted to analysis of the data. The nature of the variable is briefly discussed in x 4, and our main results are summarized in x 5. 2. DATA ACQUISITION New observations of UY Cam were secured between 1999 November 17 and 2003 March 3. The data consist of 8337 measurements (merged into 60 s bins) in the Johnson V band collected over 19 observing nights (129.5 hr). A journal of the observations is given in Table 1. 2.1. CCD Photometry From 1999 November 19 to 2000 January 24, Johnson V photometry of UY Cam was performed with the light-curve survey CCD photometer (Wei, Chen, & Jiang 1990; Zhou et al. 2001a) mounted on the 85 cm Cassegrain telescope at Xinglong Station of the Beijing Astronomical Observatory. The photometer employs a red-sensitive Thomson TH7882 576 384 CCD with an overall imaging size of 13.25 8.83 mm 2, corresponding to a field of view of 11<5 7<7, which allows sufficient stars to be selected in a frame as reference. Depending on the nightly seeing, the integration times varied from 20 to 60 s. All the monitored

TABLE 1 Journal of Johnson V Photoelectric and CCD Photometry of UY Camelopardalis UY CAMELOPARDALIS 2463 Run Night (UT) JD (2,450,000+) Time Interval (days) Points 2000... 2002... 2003... 2000 Jan 7 1,551 0.126 273 2000 Jan 13 1,557 0.291 498 2000 Jan 14 1,558 0.157 350 2000 Jan 15 1,559 0.261 469 2000 Jan 16 1,560 0.333 567 2000 Jan 17 1,561 0.317 601 2000 Jan 19 1,563 0.268 386 2000 Jan 23 1,567 0.089 161 2000 Jan 24 1,568 0.310 430 2002 Jan 23 2,298 0.309 438 2002 Jan 26 2,301 0.352 491 2002 Jan 27 2,302 0.362 515 2002 Jan 28 2,303 0.379 524 2002 Jan 29 2,304 0.364 507 2002 Jan 31 2,306 0.203 289 2002 Dec 31 2,640 0.337 486 2003 Feb 28 2,699 0.356 514 2003 Mar 2 2,701 0.369 532 2003 Mar 3 2,702 0.212 306 reference stars in the field of UY Cam were detected as nonvariables within the observational error (0.006 mag), which is the typical standard deviation of the magnitude differences between the reference stars. Among them, GSC 4369-1457 ( =07 h 58 m 11912, = +72 45 0 5198, J2000; V = 11.6 mag) was confirmed to be the best comparison against the variable. The differential magnitudes of UY Cam were thus measured with respect to GSC 4369-1475. Atmospheric extinction was ignored because of the proximity of the two stars. The standard data reduction procedure, including on-line bias subtraction, dark reduction, and flat-field correction, is outlined in Zhou et al. (2001a). 2.2. Photoelectric Photometry From 2002 January 23 to 2003 March 3, UY Cam was reobserved with the three-channel high-speed photoelectric photometer (Jiang & Hu 1998) attached to the same telescope as before. This detector is used in the Whole Earth Telescope campaigns (Nather et al. 1990). GSC 4380-1705 ( =08 h 00 m 13929, =+72 43 0 47>9, J2000; V = 11.2 mag) was chosen as the comparison star. The variable, comparison star, and sky background were simultaneously monitored in continuous 10 s intervals through a standard Johnson V filter throughout the two observing runs in 2002 and 2003. The typical observational accuracy with the three-channel photometer is about 0.006 mag. 3. DATA ANALYSIS 3.1. Frequency Detection Previous works on UY Cam have not suggested any pulsation frequencies aside from the known primary frequency. We decided to make a further investigation to find a complete set of pulsation frequencies. To do so, we merged all the data collected from 1999 to 2003. The frequency analysis was carried out using the programs Period98 (Breger 1990; Sperl 1996) and MFA (Hao 1991; Liu 1995), with which Fig. 1. Spectral window and amplitude spectra of UY Cam (2000 2003). Note the high noise at 0 5 day 1 that is visible in the two bottom panels, where significance curves with S/N = 4.0 are drawn. single-frequency Fourier transforms and multifrequency least-squares fits were processed. The two programs use the discrete Fourier transform method (Deeming 1975) and basically lead to identical results. We first computed the noise level at each frequency using the residuals from the original measurements with all trial frequencies prewhitened. Then the confidence levels of the trial frequencies were estimated following Scargle (1982). Last, to judge whether a peak was significant in the amplitude spectra, we followed the empirical criterion of Breger et al. (1993), that an amplitude signal-to-noise ratio (S/N) larger than 4.0 usually corresponds to an intrinsic peak of the variable. Note that this S/N criterion assumes a good spectral window, typical of multisite campaigns. However, in the case of single-site observations, the noise level can be enhanced by the spectral window patterns of the noise peaks and possible additional frequencies. Therefore, a significant peak s S/N value might be a little less than 4.0 in the single-site case. The amplitude spectra and spectral window are given in Figure 1, in which each spectrum panel corresponds to the residuals with all the previous frequencies prewhitened. The bottom panel ( Data 2f ) shows the residuals after subtracting the fit of the two outstanding frequencies, together with the significance curve spline-connected points at 4 times the noise values obtained at a spacing of 0.2 day 1. The observational noise is frequency dependent and was defined as the average amplitude of the residuals in a range of 1 day 1 (11.57 lhz), close to the frequency under consideration. We see high noise in the low-frequency region (0 5 day 1 ), which might be caused by instrumental shifts and nonidentical zero points during different runs or even nightto-night zero-point shifts. Apart from the outstanding primary frequency f 0 and its first harmonic 2f 0, a peak at 0.7394 day 1 ( f n ) with S/N = 5.0 has an effect on the spectrum (see the two bottom panels of Fig. 1). Frequencies f 0 and 2f 0 fit the light curves with the standard deviation of the residuals = 0.0312 mag and a zero point of 0.0019 mag. If f n is considered, these values become 0.0242 and 0.00164 mag, respectively. The term f n evidently improves the quality of the fit by about 22%. However, there is no reason that allows us to attribute this frequency ( f n ) to the variable, and we regard it as a noise component involved in the data. So,

2464 ZHOU & LIU deviation of the fit with f 0 and 2f 0 after removing the noise componenent ( f n ), 5 times the practical observational accuracy. We further used the real nights with data for the time baseline rather than the span of observations. Even with this approach, we found that the frequency errors might have been underestimated, because the single-frequency Fourier transform showed the second frequency term at 7.48685 day 1 differing from 2f 0 by 0.0026 day 1, which is larger than the theoretically calculated errors above (0.00007 and 0.00054 day 1 for f 0 and 2f 0, respectively). Therefore, we finally adopted the values at 5 times the calculated errors, so that the error for 2f 0 becomes 0.0027 mag, conforming to the difference of 0.0026 mag. As mentioned by Montgomery & O Donoghue (1999), this is a perfectly valid thing to do, since hidden correlations in the errors in the data lead to an underestimate of the true errors, because correlations in the noise can modify the formulae used to calculate the errors. Fig. 2. Phase diagram of the observed differential light curves of UY Cam in 2000 2003. Phases were calculated relative to HJD 2,451,551.38219, at which the phase is zero. prewhitening f n played a role in denoising the data. As a final result, the pulsation frequency f 0 = 3.7447 day 1 and its harmonic 2f 0, with amplitudes of 0.1638 and 0.020 mag, respectively, are found to be intrinsic to UY Cam. No additional peak is significant in the residual spectrum. Therefore, the variable is monoperiodic. Figure 2 displays the light curves folded with the main frequency f 0. Differential light curves along with the fits using f 0 and 2f 0 after denoising are presented in Figure 3. Note that there exists a zero-point shift on one night in the phase diagram. This figure appears to provide evidence of amplitude variability. We will explore this in more detail in next subsection. The Fourier parameters of the best-fitting sinusoids, m(t) =A 0 + P N i¼1 A i cos (2f i t + i ), are listed in Table 2, where the errors in frequency, amplitude, and phase were estimated using the formulae of Montgomery & O Donoghue (1999). We assumed the rms deviation of the observational noise to be 0.00242 mag, the standard 3.2. Frequency and Amplitude Variability It is possible to search the data for amplitude or frequency variability in the star. We dealt with the data in individual subsets corresponding to different observing runs. The CCD data in 1999 were ignored because of their bad quality. In addition, a 1985 data set (five nights) was adopted from Broglia & Conconi (1992). Table 3 summarizes these data sets. Individual Fourier analyses were then carried out for each set. However, we are aware that the Fourier results depend highly on the structure of the data, that is, the number of data points, sampling rate, time span, etc. Fourier analysis assumes the frequency and amplitude of a signal to be constant in the time domain. Therefore, our results for the time dependence of the frequency and amplitude should be regarded as averaged values for the period under investigation. Given the trial frequency values of f 0 = 3.7447 day 1 and its two harmonics 2f 0 and 3f 0, nonlinear least-squares sinusoidal fits to each data set were performed. Our results are listed in Table 4, where the errors were estimated following the method in the previous subsection. The standard deviations of the fits for the four subsets are = 0.0117, 0.0222, 0.0382, and 0.0216 mag, respectively. The fitting error for the 1985 data set agrees with its observational accuracy. However, the fit quality is generally low for the current data. For the 2000 and 2003 sets, the errors are about 4 times the observational accuracy, and more than 6 times for the 2002 set. In view of these inconsistent large fitting errors, the pulsation frequency and amplitude seemed to change. However, as can be seen from Table 4, within the error bars the primary frequency was quite stable over the past years. On the other hand, except for the 1985 data set, the amplitudes are basically consistent with each other. According to Figures 2 and 4, as well as the light curve on HJD 2,452,304 (2002 January 29, the panel with abscissa TABLE 2 Results of the Frequency Analysis of UY Camelopardalis Based on the 2000 2003 Data Frequency Value (day 1 ) Amplitude Phase Epoch a (days) S/N Conf. (%) f 0... 3.74475 0.00035 163.8 1.9 0.266 0.010 7.010 44.3 100 2f 0... 7.48685 0.00270 20.0 1.9 0.188 0.095 7.108 11.9 100 a HJD 2,451,550.0.

Fig. 3. Observed differential light curves of UY Cam in 2000 2003 together with the sinusoidal fits (solid lines; with f 0 = 3.7447 day 1 and 2f 0 ). The abscissae are HJD 2,451,500 days, and the ordinates are magnitudes.

2466 ZHOU & LIU Vol. 126 TABLE 3 Subsets of Data Used in the Time-dependent Analysis Run Dates Nights Measurements 1985... 1985 Apr 14 28 5 364 a 2000... 2000 Jan 7 24 9 3735 2002... 2002 Jan 23 31 6 2786 2003... 2002 Dec 31 2003 Mar 3 4 1838 a From Broglia & Conconi 1992. 804 in Fig. 3), amplitude variations, at least at the cycleto-cycle level, might be present. A glance at Table 4 shows that the pulsation amplitudes in 2000 2003 were the same within the errors. However, the amplitude in 1985 is a bit higher than the others. Because the best fits optimize all three Fourier parameters, we attempted to check the case in which the frequencies are fixed. By fixing f 0 = 3.7447 day 1 and its harmonics 2f 0 = 7.4894 day 1 and 3f 0 = 11.2341 day 1 and allowing their amplitudes and phases to change, we obtained the amplitudes of the three frequencies for the different years data. Deviations between the fits and light curves are relatively smaller in the 1985 data set than in the other three sets: 0.0149, 0.0226, 0.0383, and 0.0219 mag for the four runs, respectively. The results in Table 5 are consistent with those in Table 4. The amplitudes for 2000, 2002, and 2003 are consistent with each other. However, in both Tables 4 and 5 the amplitude of f 0 in 1985 is obviously higher than in the other three years. Given the estimated error range, this difference probably is real. Consequently, we think that the amplitude of f 0 changed in 1985 but remained constant during 2000 and 2003. In addition, the contribution of the third-order harmonic 3f 0 to the light variations can be neglected. This picture of amplitude variability is similar to that mentioned by Breger (2000), in that the HADS stars typically have less amplitude variation than the low-amplitude Scuti stars. In order to further examine the stability of the primary frequency, we determined the times of maximum light and made use of those existing maxima. To determine a maximum, we applied a polynomial fit to a portion of the light curve around each maximum and took the extrema of the polynomials as the observed times of maximum light. We usually tried several times of fitting by selecting polynomials TABLE 4 Fourier Parameters of the Best-fitting Sinusoids for the Four Subsets of Data from 1985 to 2003 Run Frequency (day 1 ) Amplitude Phase Fig. 4. Light curve of UY Cam on 2002 January 28. The third-order polynomial fits to the different regions around the maxima show the determination of times of maximum. This figure also displays the cycle-to-cycle amplitude variation. of different orders (e.g., third, sometimes fifth) and different regions until a satisfactory fit was reached. Generally, we chose a range with an amplitude of about one-third the full amplitude. (See Fig. 4 for an illustration.) The fitting errors are typically 0.008 mag, consistent with the observational precision. Thus 16 maxima were determined from the present data. The errors in this determination are about 0.00045 days. In order to avoid the difficulty of fitting a polynomial to determine the precise time of maximum of each peak, one could perform a sine fit including the harmonics for the given season of observations, that is, the observations in 1985 and in 2000 2003. However, given that the relative amplitude of the harmonics may change with respect to the amplitude of the fundamental frequency f 0 (and that this would change when the time of maximum occurred even if the frequency f 0 were absolutely constant), it may be safer to look at the time of maximum of only the f 0 component, and not 2f 0 or 3f 0. In this way, we selected four times of maximum light for the 1985, 2000, 2002, and 2003 seasons: 2,446,173.4361, 2,451,560.21625, 2,452,301.2504, and 2,452,699.13721. Their corresponding O C values are 0.04249, 0.00678, 0.00493, and 0.00214 days. We note that these times of maximum have a lag ranging from 0.00375 to 0.01197 days with respect to their observed peaks determined from the polynomial fits. Moreover, the fits with only f 0 are poor, so these times act as mean values. We adopted five maxima from Broglia & Conconi (1992). In addition, there are 21 maxima in the B band by Beyer (1966) in the literature. In terms of the BV light curves of 1985... 2000... 2002... 2003... f 0 = 3.745 0.004 182.3 4.5 0.620 0.015 2f 0 = 7.492 0.008 20.8 4.5 0.521 0.126 3f 0 = 11.245 0.015 5.8 4.5 0.460 0.134 f 0 = 3.748 0.001 166.8 2.4 0.077 0.015 2f 0 = 7.477 0.009 16.1 2.4 0.825 0.126 3f 0 = 11.233 0.019 10.3 2.4 0.834 0.106 f 0 = 3.745 0.002 165.2 3.0 0.328 0.018 2f 0 = 7.492 0.012 21.9 3.0 0.896 0.150 3f 0 = 11.284 0.066 6.5 3.0 0.501 0.242 f 0 = 3.745 0.012 162.4 8.7 0.137 0.090 2f 0 = 7.489 0.039 19.2 8.7 0.907 0.280 3f 0 = 11.233 0.150 7.4 8.7 0.570 0.358 Frequency TABLE 5 Pulsation Amplitudes of the Primary Frequency ( f 0 = 3.7447 day 1 ) of UY Camelopardalis and Its Two Harmonics in 1985 2003 1985 2000 2002 2003 f 0... 181.0 4.5 166.2 2.4 165.5 3.0 162.1 8.7 2f 0... 21.2 4.5 14.9 2.4 21.5 3.0 19.0 8.7 3f 0... 6.0 4.5 10.2 2.4 6.3 3.0 7.2 8.7

No. 5, 2003 UY CAMELOPARDALIS 2467 is, we defined an initial ephemeris Fig. 5. BV light curves of UY Cam on 1985 April 15. The phase difference between B and V can be ignored with respect to the errors in the determination of the maxima. (Data adopted from Broglia & Conconi 1992.) Broglia & Conconi (IAU Archives of Unpublished Observations, file 246E), we found that the phase difference between the V and B filters can be ignored. In fact, when fixing f 0 and 2f 0, fits to the BV data produced an equivalent phase difference of 0.0003 days, which is less than the error in the determination of the maxima. We refer the reader to Figure 5, where the BV light curves on HJD 2,446,171 are shown. We therefore combined the 21 maxima in B with the others in V. In total, we have 42 observed times of maximum light (see Table 6). We took our first maximum as the initial epoch and P 0 =1/f 0 = 0.26704 days as the trial period, that HJD max ¼ 2;451;557:27203 þ 0:26704E : Then the observed minus the calculated times of maxima (O C residuals) and the cycles elapsed from this epoch were calculated. We note that the counting of cycles produces integers, so whenever a maximum occurred at a time near the next cycle, that is, more than half a cycle away from the previous cycle, the next cycle number was chosen for this maximum. To determine whether the times of maximum light are consistent with a constant period (O C residuals lying on a truly straight line) or a changing period (on a parabola), a parabolic fit was applied to the data. We fitted the O C residuals to the equation O C = DT 0 + DPE +0.5E 2, following, for example, Kepler et al. (2000) and Zhou (2001). Here DT 0 = T new 0 T ini 0 (improved new epoch minus assumed initial epoch), DP refers to the improved-minus-adopted trial period, and refers to dp/dt, the rate of period change. Using all available times of maximum light, we obtain ð1þ O C ¼ 0:00853 0:00234 þð9:11 2:21Þ10 7 E ð4:43 3:38Þ10 12 E 2 ð2þ with a fitting error of = 0.0104 days. If the quadratic term is ignored disregarding period changes the fit becomes linear, as O C ¼ 0:00819 0:00234 þð1:19199 0:05309Þ10 6 E with = 0.01049 days (see Fig. 6, dot-dashed line). If the 21 maxima in 1985 2003 are replaced with the four corresponding mean times (Fig. 6, horizontal bars), a parabolic fit ð3þ TABLE 6 Times of Maximum Light of UY Camelopardalis HJD (2,400,000+) E O C Ref. HJD (2,400,000+) E O C Ref. 32,144.42300... 72696 0.10919 1 46,170.49480... 20172 0.04635 2 32,240.56500... 72336 0.10159 1 46,171.56400... 20168 0.04531 2 32,643.78700... 70826 0.10999 1 46,172.35900... 20165 0.05143 2 32,985.61000... 69546 0.09819 1 46,173.43020... 20161 0.04839 2 33,116.46600... 69056 0.09179 1 46,184.38150... 20120 0.04573 2 33,228.61800... 68636 0.09659 1 51,557.27203... 0 0.00000 3 33,399.51900... 67996 0.10119 1 51,560.20428... 11 0.00519 3 33,565.09000... 67376 0.09499 1 51,561.27899... 15 0.00136 3 33,805.43700... 66476 0.08399 1 51,568.21634... 41 0.00433 3 34,019.06700... 65676 0.08599 1 52,298.30404... 2775 0.00399 3 34,328.83900... 64516 0.08039 1 52,301.24094... 2786 0.00453 3 35,565.23500... 59886 0.07959 1 52,302.31347... 2790 0.00016 3 36,208.81400... 57476 0.06699 1 52,303.12009... 2793 0.00534 3 36,550.62700... 56196 0.06519 1 52,303.37114... 2794 0.01065 3 36,927.15600... 54786 0.06259 1 52,304.17633... 2797 0.00658 3 37,690.88500... 51926 0.06799 1 52,306.30818... 2805 0.01105 3 38,056.75300... 50556 0.04479 1 52,640.11870... 4055 0.00053 3 38,289.07200... 49686 0.05059 1 52,640.38168... 4056 0.00459 3 38,459.97900... 49046 0.04919 1 52,699.13346... 4276 0.00161 3 38,681.61400... 48216 0.05739 1 52,701.26168... 4284 0.00971 3 38,831.16800... 47656 0.04579 1 52,702.06761... 4287 0.00490 3 Note. Elapsed cycles (E) and O C are based on the initial ephemeris HJD max = 2,451,557.27203 + 0.26704 E. References. (1) Beyer 1966; (2) Broglia & Conconi 1992; (3) this work.

2468 ZHOU & LIU Vol. 126 This linear fit appears to be perfect the quality of the fit ( = 0.00618 days) has improved evidently compared with that given by equation (4) ( = 0.01049 days). This fit is also shown in Figure 6 (dashed line). It has almost the same slope as that for the early 1946 1965 data. For these shifted data, a parabolic fit results in O C ¼ 0:06378 0:00133 þð2:02 0:03Þ10 6 E ð4:82 1:92Þ10 12 E 2 ð6þ Fig. 6. The O C diagram for UY Cam. Horizontal bars: season s mean value; thick solid line: linear fit, 1946 1965 data; dotted line, linear fit, 1985 2003 data; dot-dashed line, linear fit, all data; thin solid line, quadratic fit, all data; dashed line, linear fit, all data with the 1985 2003 data shifted by 0.07243 days. The validity of this offset is unresolved. to the 25 data points (21 points from 1946 to 1965) leads us to O C ¼ 0:0000180187 0:00311 þð4:58639 2:54803Þ10 7 E ð1:3416 0:3528Þ10 11 E 2 ( = 0.00829 days). Compared with the fits in equations (2) and (4), the quality of the fit is improved in equation (4). Equation (2) implies a rate of period change (decrease) of dp/dt = ( 8.86 6.76) 10 12 days cycle 1 = ( 3.32 2.53) 10 11 days day 1, and P 1 dp/ dt =( 4.54 3.46) 10 8 yr 1. In the case of equation (4), dp/dt =( 10.0 2.6) 10 11 days day 1 and P 1 dp/ dt =( 13.68 3.56) 10 8 yr 1, about 3 times the previous values. The values of the period change are comparable to those listed by Breger & Pamyatnyhk (1998) and Rodríguez et al. (1995). We note here that the error bars for the maxima from 1985 to 2003 are around 0.0005 days, which is about a quarter of the size of the symbols in Figure 6, so they are invisible in the figure. At first glance, however, if one looks only at the data from 1946 to 1965, they appear very linear. Furthermore, if one looks at the 1985 plus the 2000 2003 data, one can also draw a straight line through them. Therefore, we made individual linear fits to the two parts of the data. For the 1946 1965 data, we obtain O C = (0.06357 + 2.32624) 10 6 E with = 0.00556 days (Fig. 6, thick solid line), while for the 1985 2003 data we have O C = (0.00886 + 1.8807) 10 6 E with = 0.00557 days (Fig. 6, dotted line). There is a small difference between the slopes of these two lines, but the latter intercept (corresponding to the improved initial epoch of maximum light) differs from the former by 0.07243 days. If we were to shift the 1985 2003 data upward by this offset, then we would find that all the measurements lie on a straight line described by O C ¼ 0:06383 0:00140 þð2:32083 0:03132Þ10 6 E : ð5þ ð4þ ( = 0.0059 days). However, there is not adequate evidence to justify our inserting such an offset. We recall that the maxima in 1946 1965 are in the B band, but the phase difference between V and B could not possibly cause an offset as large as can be seen in Figure 6. This offset, if it existed, would mean that there is some sort of inconsistency in the zero-point time calibrations between the old and the new data. That is, each observed time of maximum from 1985 to 2003 had a time lag equivalent to this offset, so that the resulting O C values should be shifted relative to the 1946 1965 data by this value. Because the 1985 2003 data comprise two different groups observations collected over 4 years, this kind of consistent error in the zero-point times (i.e., the starting observation times) in each observing year is implausible. Therefore, this possibility can be ruled out. Another possible cause of such an offset is a period change: the pulsation period of the star was changing from 1965 to 1985 when we were not looking at it, but in 1985, the star went back to exactly the same period that it had in 1946 1965. Because there were no observations during this time span, we are unable to affirm whether or not any changes occurred during this period. Furthermore, there is also a time span between 1985 and 2000 without observations. These two time spans without data bring about great uncertainties in deriving a period change through a parabolic fit to the O C residuals. Therefore, we think that an O C analysis is not a suitable method for determining this star s period variability with the current data. To summarize, it is clear that the overall profile of the O C residuals indicates a measurable period change, but the Fourier solutions for the different data sets did not resolve any such changes. We note that the small number of data (42 maxima), with two gaps from 1965 to 1986 and 1985 to 2000, over the long period under investigattion (1946 2003) do not support a reliable O C analysis. 4. DISCUSSION 4.1. Light Curves and Types of Variability An inspection of all the light curves show that they are slightly asymmetric or nearly sinusoidal with rounded maxima. It was estimated that about 55% of time (3.5 hr) the variable is in the descending branch. Amplitude variations at a cycle-to-cycle timescale did not always occur from one cycle to the next. The deviation between the fit and the observations on HJD 2,452,304 (2002 January 29) is unique in our 19 nights data, as well as in the five nights data from Broglia & Conconi (1992). The difference in amplitude between 1985 and the 2000s does not seem to be strong evidence for longer timescale amplitude variability. We could not infer any periodicity of the amplitude variation from these light curves. On the contrary, we might interpret the deviation on 2002 January 29 to be a peculiar cycle. In

No. 5, 2003 UY CAMELOPARDALIS 2469 view of the current Fourier solution, there are other types of variability that can explain the light variations. We know the Scuti stars are pulsating variables with periods less than 0.3 days and visual light amplitudes in the range from a few thousandths of a magnitude to about 0.8 mag. They occupy a position on the Hertzsprung-Russell (H-R) diagram either on or somewhat above or below the main sequence (MS). Most Scuti stars belong to Population I, but a few variables show low metal abundances, low masses, and high space velocities, typical of Population II (i.e., SX Phoenicis stars). The majority of the known Scuti stars have evolved to post-ms stage. The effective temperature range corresponds well to the extension of the Cepheid and RR Lyrae instability strip to the MS. Population I HADS stars with normal mass and chemical composition, evolving away from the MS, together with SX Phoenicis stars with low mass and metal content, have periods of 1 5 hr and amplitudes of 0.3 0.8 mag. For a long time, this small group (dwarf Cepheids) was included in the large RR Lyrae family. They were later distinguished from the RR Lyrae stars by their shorter periods and weaker absolute luminosities (H. A. Smith 1955; H. J. Smith 1995). RR Lyrae stars usually reside in Galactic globular clusters (about 130 in our Galaxy, e.g., IC 4499; Clement, Dickens, & Bingham 1979) of age about 10 Gyr or more, as well as in the bulge region of our Galaxy and other dwarf galaxies. The RRc variables have lower light amplitudes (0.5 mag) and nearly sinusoidal light curves with a rounded maximum. In all known RRc stars, the dominant mode is the radial first overtone, and the periods are mostly in a range from about 0.2 to 0.5 days. From the above, UY Cam can be seen to be similar to the SX Phoenicis stars in amplitude, while it is similar to the RRc stars in period and in the shape and amplitude of its light curves. A HADS star might be misclassified and could be reclassified as either SX Phoenicis or RR Lyrae type. To go into details of the variability types, we give further information on the nature of the star below. 4.2. Physical Parameters By adopting the values b y =0.149, m 1 =0.110, c 1 =1.140,and = 2.754 mag for UY Cam from Rodríguez et al. (2000), we can deredden these indices by making use of the dereddening formulae and calibrations for A-type stars given by Crawford (1979). In this way we derive a color excess of E(b y) = 0.003 mag and the following intrinsic indices: (b y) 0 =0.146, m 0 =0.111, and c 0 = 1.139 mag. Furthermore, deviations from the zero-age MS values of m 0 =0.077andc 0 = 0.452 mag are also found. A mean metal abundance of [M/H] = 0.732 dex (or Z = 0.0037) was derived from m 0 using the calibrations for the metallicity of A-type stars by Smalley (1993). A more metal-poor value of [M/H] = 1.51 was derived by Fernley & Barnes (1997). The metallicity is quite a bit lower than those of other HADS stars and is comparable to those of SX Phoenicis stars (see, e.g., Table 2 of McNamara 2000). Usually, the shorter period variables are metal-poor (SX Phoenicis stars), while the longer period variables are metalrich, as can be seen from Table 2 and Figure 1 of McNamara (2000). However, the period of UY Cam (log P = 0.573) is the second longest among the HADS stars (almost equal to that of V1719 Cyg and just shorter than the period of SS Psc), but its metal abundance as derived above is so poor that the star does not fit the general profile or [Fe/H] log P relation for the HADS stars (see Fig. 1 of McNamara 2000). Stellar physical parameters of UY Cam, including effective temperature, absolute magnitude, surface gravity, and other quantities, can be derived by applying suitable calibrations for the uvby indices. Using the same method as in Zhou et al. (2001b) and Zhou, Liu, & Rodríguez (2002), we derived T eff = 7300 150 K. We used, which is free of interstellar extinction effects, as the independent parameter for measuring temperature (Crawford 1979). The effective temperature was determined from the model atmosphere calibrations of uvby photometry by Moon & Dworetsky (1985). Because of the large value of c 0 (>0.28 mag, too luminous relative to normal Scuti stars), we were misled to an unusual value for the absolute visual magnitude, M v = 1.16 mag, as well as a small surface gravity of log g = 2.79 0.06, by Moon s (1985) program UVBY- BETA. However, this value of M v is generally unreasonable because it would place the star outside the Scuti instability strip according to several H-R diagrams of the pulsating variables (e.g., Fig. 2 of Breger 2000, Fig. 2 of McNamara 2000, and Fig. 8 of Rodríguez & Breger 2001), even though we know there are a few Scuti stars outside the instability strip. The gravity is also too low for a normal Scuti star. Therefore, we tried to determine M v and log g in other ways. Based on = 2.754 mag, assuming v sin i =25kms 1 (a statistically average value for HADS stars, according to Jiang, Siek, & Min 2000), we obtained M v = 0.228 mag using the relation from Domingo & Figueras (1999). This value is comparable to the results derived from Crawford (1979) and from Mathew & Rajamohan (1992): 0.454 and 0.218 mag, respectively. Furthermore, this value has a small difference from that predicted (0.246 0.15 or 0.235 0.18 mag) by the period-luminosity relation for fundamental-mode HADS, SX Phoenicis, and RR Lyrae stars (Høg & Petersen 1997; Petersen & Høg 1998; McNamara 1997, 2002; Laney, Joner, & Schwendiman 2002). On the other hand, we can check M v from the Hipparcos parallax. Unfortunately, the parallax has a large uncertainty ( 1.64 2.08 mas). An examination of the catalog of Scuti stars by Rodríguez et al. (2000) shows that there are only eight stars with negative parallax values that have determination errors larger than the parallaxes themselves (the stars are TV Lyn, CW Ser, V974 Oph, V567 Oph, BQ Ind, BP Peg, DE Lac, and UY Cam). Taking the positive value 1.64 + 2.08 = 0.44 mas for UY Cam, we have M v = 0.38 mag if hv i = 11.4 mag, from the formula M v = m + 5 + 5 log p, where p is the parallax in parsecs. Therefore, we are led to take M v = 0.2 0.2 for UY Cam. As a consequence, these parameters suggest that the variable is a Scuti star with a poor abundance in metals and is situated in the upper Scuti region in the H-R diagram, as can be seen from Figure 8 of Rodríguez & Breger (2001) and Figure 2 of McNamara (2000). In addition, we derived log g = 3.46 0.06 from the biparametric calibrations by Ribas et al (1997) and the grids of uvby colors for [M/H] = 0.5 by Smalley & Kupka (1997). The interstellar reddening can be neglected because of the color excess of E(b y) = 0.003 mag and the Galactic latitude (b =30=8) of the star. In fact, following Crawford & Mandwewala (1976), the reddening value is only about 0.0006 mag. Consequently, a distance modulus m M v = 11.2 0.2 mag is obtained, that is, a distance of

2470 ZHOU & LIU Vol. 126 TABLE 7 Dereddened Indices and Physical Parameters for UY Camelopardalis Parameter Value Parameter Value E(b y) (mag)... 0.003 0.01 Age (Gyr)... 0.7 0.1 (b y) 0 (mag)... 0.146 0.01 M (M )... 2.0 0.3 m 0 (mag)... 0.111 0.01 R (R )... 5.50 0.5 c 0 (mag)... 1.139 0.01 log (L/L )... 1.90 0.08 m 0 (mag)... 0.077 0.01 T eff (K)... 7300 150 c 0 (mag)... 0.452 0.01 log g (dex)... 3.46 0.06 M v (mag)... 0.2 0.2 [M/H] (dex)... 0.732 0.1 M bol (mag)... 0.0 0.2 ( )... 0.012 0.005 1.74 kpc according to the formula 5 log r = m M v +5. This distance in turn results in a parallax of 0.57 mas. Assuming a bolometric correction BC = 0.02 mag, derived from Malagnini et al. (1986) for T eff = 7300 K, the bolometric magnitude is M bol =0.0 0.2 [log (L/L )= 1.90 0.08]. Moreover, it is possible to gain some insight into the mass and age of this star using the evolutionary tracks of Claret & Giménez (1998) for Z = 0.004. In this case, an evolutionary mass M =2.0 0.3 M and an age of 0.7 0.1 Gyr are found with T eff = 7300 K and log g = 3.46. So, a radius R =5.5 0.5 R from the radiation law or period-radius relation (McNamara & Feltz 1978; Fernie 1992; Laney et al. 2002) and a mean density of M/R 3 = 0.012 0.005 are derived. Table 7 lists the parameters derived for UY Cam. Regarding its younger age and lower surface gravity, UY Cam is similar to a Population I star. But it could be a Population II star (SX Phoenicis type) by virtue of its poor metal abundance and advanced evolution stage, or an RR Lyrae star given its distance, period, luminosity, and metallicity and the morphology of the light curves. Furthermore, with respect to the long pulsation periods, Rodríguez & Breger (2002) discussed the implications of the Scuti stars with periods longer than 0.25 days. There are only 14 stars in this subgroup from the total list of 636 Scuti stars (Rodríguez et al. 2000). On the basis of an examination of the pulsational behaviors, luminosities, metallicities, and light curves, they reclassified three HADS stars, DH Peg, UY Cam, and YZ Cap, as RR Lyrae variables, while suspecting three other HADS stars, including SS Psc, to not be of Scuti type. We found that UY Cam resembles V1719 Cyg in several aspects. However, V1719 Cyg remains a Scuti pulsator. A comparison between them is given in Table 8. It is quite interesting to conduct a further comparison of the two stars. V1719 Cyg has a rotational velocity of v sin i =31kms 1 (Solano & Fernley 1997). No observed v sin i value was found for UY Cam in the literature, for example, in the catalog of stellar projected rotational velocities by Goębocki & Stawikowski (2000). According to the statistics compiled by Jiang et al. (2000) and Rodríguez et al. (2000), as our previous estimate, UY Cam s rotational velocity should be around 20 km s 1. Observations have shown that the HADS stars appear to be much more slowly rotating than the other Scuti stars, as pointed out by, for example, Breger (2000). Statistical results on pulsation amplitudes versus rotational velocities for the 191 Scuti stars with known rotational velocity indicate that amplitude decreases with increasing v sin i value, that is, high-amplitude pulsators have low v sin i values and fewer modes, while low-amplitude pulsators are faster rotators with multiple simultaneously excited modes (Breger 2000; Jiang et al. 2000; Rodríguez et al. 2000). This correlation between slow rotation and high amplitude may be an important clue to the excitation and damping mechanisms for the pulsations in these stars. However, limited by our observational conditions, we could not obtain a spectrum for this fainter star (V = 11.4 mag). If a high-resolution spectrum of this star could be taken in the future so that some sort of model atmosphere fit could be done, we may then expect to derive its rotational velocity, effective temperature, surface gravity, metallicity, and other atmospheric parameters. RR Lyrae variables have no detectable rotation Peterson, Carney, & Latham (1996) estimated an upper limit of v sin i < 10 km s 1 for the RR Lyrae stars. So, a spectroscopic study of UY Cam would be very useful for determining the star s nature. SX Phoenicis variables are not yet fully explained by stellar evolution theory. Both SX Phoenicis and RR Lyrae stars are indicators of the distance scale and are very important objects for the study of stellar evolution, as well as the study of clusters which are inhabited by quite a number of SX TABLE 8 A Comparison between Two Similar High-Amplitude Scuti Stars Parameter UY Cam V1719 Cyg Similarities Primary period (day 1 )... 0.26704 0.2673 Stable period?... Yes Yes Radius (R )... 4.9 5.5 Gravity, log g... 3.46 3.1 3.4 DV (mag)... 0.34 0.35 M v (mag)... 0.2 0.2 0.37 Evolutionary status... Post-MS Post-MS Population... I or II I Differences Mass (M )... 2.0 0.3 2.0 Periodicity... Single Double Spectral type... A3 A6 III F5 III ht eff i (K)... 7300 150 6750 7300 Metallicity [Fe/H]... Poor: 0.732 Rich: 0.25 Distance (pc)... 1740 100 324 53 Reddening, E(b y) (mag)... 0.02 0.006 Age (Gyr)... 0.7 0.1... Note. Data from Rodríguez & Breger 2002 and Peña et al. 2002 and references therein.

No. 5, 2003 UY CAMELOPARDALIS 2471 Phoenicis and RR Lyrae stars and the structure of galaxies. Its sharing some of the photometric properties of Scuti, SX Phoenicis, and RR Lyrae variables makes UY Cam an important star. 4.3. Pulsation Mode From a multicolor study of phase shift and amplitude ratio, UY Cam was identified as a radial pulsator (Rodríguez et al. 1996). By means of the empirical formula of Petersen & Jørgensen (1972), we obtained a pulsation constant of Q = 0.037 0.007 days for f 0. The error might be underestimated because of various calibrations used in deriving physical parameters, but in any event, this Q-value means that the primary frequency is a radial p ffiffiffi mode. From the basic pulsation equation, Q 0 = P 0 (here P0 is the fundamental period), the mean density = 0.02 is about double the value estimated above (M/R 3 0.01 ). If = 0.01, then Q = 0.0267 days, a value largely corresponding to first-overtone pulsation. So, f 0 is probably not the fundamental mode. Furthermore, owing to the uncertainty in the value of M bol, f 0 has the possibility of being a fundamental or first-overtone mode according to the empirical periodluminosity-color relations (Stellingwerf 1979; López de Coca et al. 1990; Tsvetkov 1985). If f 0 is the firstovertone mode, we could expect the fundamentalized mean period to be 2.913 day 1 by assuming a normal frequency ratio f 0 /f 1 = 0.778, the ratio predicted from models. We note that there is a term at 2.98 day 1 that appears in the residual spectrum after removing f 0 and 2f 0, but it is not significant in the bottom panel of Figure 1. Succeeding Fourier transforms output two frequencies at 1.364 and 2.987 day 1 with amplitudes of 0.018 and 0.014 mag (below the significance level), respectively, but they are not found to be real at all. It is very hard to safely detect an intrinsic oscillation frequency in this lowfrequency region (0 4 day 1 ) at these amplitude levels. If f 0 were the fundamental mode, then the expected second frequency would be at 4.8 day 1 as a potential firstovertone mode. Similarly, this term is difficult to resolve with the current data. Finally, obtaining an accurate value of M bol is very important to identify the mode of the primary frequency. If M bol is positive, for example, 0.2 mag, the mode will be the fundamental. In conclusion, the oscillatory nature of f 0 is radial, with the possibility of being in either fundamental or first-overtone mode. 5. CONCLUSIONS Based on the available data from 1985 to 2003, we confirm that UY Camelopardalis is a monoperiodic radial pulsator. The light curves of UY Cam are slightly asymmetric. About 55% of the time (3.5 hr), the variable is seen in the descending branch. We searched data for the 4 years for possible frequency and amplitude variability. Fourier analysis cannot resolve any period change (see Table 4). Forced parabolic fits to the observed-minus-calculated times of maximum light (eqs. [2] and [4]) suggest a period-change rate dp/dt of ( 3.32 2.53) 10 11 or ( 10.0 2.6) 10 11 days per day. We discussed the distribution of the O C data points in 1985 2003 and finally dismissed the approach of inserting an offset for this part of the data. Because of gaps in the data in the two periods 1965 1985 and 1985 2000 and fewer data points in the O C plot, fits to the O C residuals suffered from great uncertainties. The O C diagram is shown in Figure 6. At present, we do not think that period changes in UY Cam have been established, given the current status of the data and the fitted results. Within the errors, the amplitude of f 0 was constant in 2000 2003, but it appeared to change from 1985 to the 2000s. Table 5 shows the amplitudes in 1985, 2000, 2002, and 2003. According to the star s location in the color-magnitude diagram (see Fig. 8 of Rodríguez & Breger 2001), UY Cam is located in the upper portion of the Scuti instability region. Using photometric indices, we derived the main physical parameters of UY Cam. The photometric properties and stellar parameters are given in Table 7. UY Cam could be a Population I high-amplitude Scuti star with poor metal abundances (Z = 0.0037, about 18.5% of solar) evolving on its post main-sequence stage after the turnoff point (age 0.7 0.1 Gyr) in a shell hydrogen burning phase. UY Cam probably pulsates in the radial first-overtone mode. UY Cam is an interesting object because of its longer period, poorer metallicity, higher luminosity, lower gravity, and larger radius among the HADS stars. The star overlaps the Population I HADS stars, Population II HADS or SX Phoenicis stars, and RR Lyrae stars. In light of these properties and features, the variable may be considered an analog of both the dwarf Cepheid (HADS plus SX Phoenicis) and RR Lyrae stars. This research was funded by the National Natural Science Foundation of China (grant 10273014). Baker, E. A. 1937, MNRAS, 98, 65 Beyer, M. 1966, Astron. Nachr., 289, 95 Breger, M. 1990, Commun. Asteroseismol., No. 20, 1. 2000, in ASP Conf. Ser. 210, Delta Scuti and Related Stars, ed. M. Breger & M. H. Montgomery (San Francisco: ASP), 3 Breger, M., et al. 1993, A&A, 271, 482 Breger, M., & Pamyatnyhk, A. A. 1998, A&A, 332, 958 Broglia, P., & Conconi, P. 1992, Inf. Bull. Variable Stars, No. 3748 Claret, A., & Giménez, A. 1998, A&AS, 133, 123 Clement, C. C., Dickens, R. J., & Bingham, E. E. 1979, AJ, 84, 217 Crawford, D. L. 1979, AJ, 84, 1858 (erratum 85, 621 [1980]) Crawford, D. L., & Mandwewala, N. 1976, PASP, 88, 917 Deeming, T. J. 1975, Ap&SS, 36, 137 Domingo, A., & Figueras, F. 1999, A&A, 343, 446 Fernie, J. D. 1992, AJ, 103, 1647 Fernley, J. A., & Barnes, T. G. 1997, A&AS, 125, 313 Goębocki, R., & Stawikowski, A. 2000, Acta Astron., 50, 509 REFERENCES Hao, J.-X. 1991, Publ. Beijing Astron. Obs., 18, 35 Høg, E., & Petersen, J. O. 1997, A&A, 323, 827 Jiang, S.-Y., Siek, H., & Min, K. K. 2000, in ASP Conf. Ser. 210, Delta Scuti and Related Stars, ed. M. Breger & M. H. Montgomery (San Francisco: ASP), 572 Jiang, X.-J., & Hu, J.-Y. 1998, Acta Astron. Sinica, 39, 438 Kepler, S. O., et al. 2000, ApJ, 534, L185 Laney, C. D., Joner, M., & Schwendiman, L. 2002, in ASP Conf. Ser. 259, Radial and Nonradial Pulsations as Probes of Stellar Physics, ed. C. Aerts, T. R. Bedding, & J. Christensen-Dalsgaard (San Francisco: ASP), 112 Liu, Z.-L. 1995, A&AS, 113, 477 López de Coca, P., Rolland, A., Rodríguez, E., & Garrido, R. 1990, A&AS, 83, 51 Malagnini, M. L., Morossi, C., Rossi, L., & Kurucz, R. L. 1986, A&A, 162, 140 Mathew, A., & Rajamohan, R. 1992, J. Astrophys. Astron., 13, 61

2472 ZHOU & LIU McNamara, D. H. 1997, PASP, 109, 1221. 2000, in ASP Conf. Ser. 210, Delta Scuti and Related Stars, ed. M. Breger & M. H. Montgomery (San Francisco: ASP), 373. 2002, in ASP Conf. Ser. 259, Radial and Nonradial Pulsations as Probes of Stellar Physics, ed. C. Aerts, T. R. Bedding, & J. Christensen-Dalsgaard (San Francisco: ASP), 116 McNamara, D. H., & Feltz, K. A., Jr. 1978, PASP, 90, 275 Montgomery, M. H., & O Donoghue, D. 1999, Commun. Asteroseismol., No. 116, 28 Moon, T. T. 1985, Commun. Univ. London Obs., No. 78 Moon, T. T., & Dworetsky, M. M. 1985, MNRAS, 217, 305 Nather, R. E., Winget, D. E., Clemens, J. C., Hansen, J. C., & Hine, B. P. 1990, ApJ, 361, 309 Peña, J. H., Paparó, M., Peniche, R., Rodríguez, M., Hobart, M. A., de la Cruz, C., & García Cole, A. 2002, PASP, 114, 214 Petersen, J. O., & Høg, E. 1998, A&A, 331, 989 Petersen, J. O., & Jørgensen, H. E. 1972, A&A, 17, 367 Peterson, R. C., Carney, B. W., & Latham, D. W. 1996, ApJ, 465, L47 Ribas, I., Jordi, C., Torra, J., & Giménez, A. 1997, A&A, 327, 207 Rodríguez, E., & Breger, M. 2001, A&A, 366, 178. 2002, in ASP Conf. Ser. 259, Radial and Nonradial Pulsations as Probes of Stellar Physics, ed. C. Aerts, T. R. Bedding, & J. Christensen-Dalsgaard (San Francisco: ASP), 328 Rodríguez, E., López de Coca, P., Costa, V., & Martín, S. 1995, A&A, 299, 108 Rodríguez, E., López-González, M. J., & López de Coca, P. 2000, A&AS, 144, 469 Rodríguez, E., Rolland, A., López de Coca, P., & Martín, S. 1996, A&A, 307, 539 Scargle, J. 1982, ApJ, 263, 835 Smalley, B. 1993, A&A, 274, 391 Smalley, B., & Kupka, F. 1997, A&A, 328, 349 Smith, H. A. 1995, RR Lyrae Stars (Cambridge: Cambridge Univ. Press) Smith, H. J. 1955, Ph.D. thesis, Harvard Univ. Solano, E., & Fernley, J. A. 1997, A&AS, 122, 131 Sperl, M. 1996, Commun. Asteroseismol., No. 111, 5 Stellingwerf, R. F. 1979, ApJ, 227, 935 Tsvetkov, Ts. G. 1985, Ap&SS, 117, 137 Wei, M.-Z., Chen, J.-S., & Jiang, Z.-J. 1990, PASP, 102, 698 Williams, J. A. 1964, PASP, 76, 50 Zhou, A.-Y. 2001, A&A, 374, 235 Zhou, A.-Y., Liu, Z.-L., & Rodríguez, E. 2002, MNRAS, 336, 73 Zhou, A.-Y., Rodríguez, E., Liu, Z.-L., & Du, B.-T. 2001a, MNRAS, 326, 317 Zhou, A.-Y., Rodríguez, E., Rolland, A., & Costa, V. 2001b, MNRAS, 323, 923