1: %\documentclass{aastex} 

2: \documentclass[12pt,preprint]{aastex} 

3: %\documentclass[preprint2]{aastex} 

4: \usepackage{psfig}

5: 

6: \newcommand{\et}{et al.}  

7: \newcommand{\ms}{\mbox{m s$^{-1}~$}}

8: \newcommand{\mse}{\mbox{m s$^{-1}$}}

9: \newcommand{\kms}{\mbox{km s$^{-1}~$}}

10: \newcommand{\kmse}{\mbox{km s$^{-1}$}}

11: \newcommand{\msun}{M$_{\odot}~$}

12: \newcommand{\msune}{M$_{\odot}$}

13: \newcommand{\rsun}{R$_{\odot}$}

14: \newcommand{\mjup}{M$_{\rm JUP}~$}

15: \newcommand{\mjupe}{M$_{\rm JUP}$}

16: \newcommand{\msat}{M$_{\rm SAT}$}

17: \newcommand{\rjup}{R$_{\rm JUP}~$}

18: \newcommand{\msini}{$M \sin i~$}

19: \newcommand{\msinie}{$M \sin i$}

20: \newcommand{\sini}{$sin i~$}

21: \newcommand{\Mv}{M$_{\rm v}~$}

22:   

23: \begin{document}  

24:   

25: \title{Optical SETI: A Spectroscopic Search for Laser Emission \\  

26:     from Nearby Stars$~^{1}$}  

27:   

28: \author{Amy E. Reines\altaffilmark{2} and  

29:     Geoffrey W. Marcy\altaffilmark{2,3}}  

30:   

31: \email{gmarcy@etoile.berkeley.edu}  

32:   

33: \altaffiltext{1}{Based on observations obtained at the  

34: W.M. Keck Observatory, which is operated jointly by the  

35: University of California and the California Institute of Technology.}  

36:   

37: \altaffiltext{2}{Department of Physics and Astronomy, San Francisco   

38:     State University, San Francisco, CA USA 94132}  

39:   

40: \altaffiltext{3}{Department of Astronomy, University of California,  

41: Berkeley, CA USA 94720}  

42:   

43:   

44: \begin{abstract}  

45: We have searched for nonastrophysical emission lines in the optical

46: spectra of 577 nearby F, G, K, and M main--sequence stars.  Emission

47: lines of astrophysical origin would also have been detected, 

48: such as from a time--variable chromosphere or infalling comets.  We

49: examined $\sim$20 spectra per star obtained during four years with the

50: Keck/HIRES spectrometer at a resolution of 5 \kms, with a detection

51: threshold 3\% of the continuum flux level.  We searched each

52: spectrum from 4000~\AA~-- 5000~\AA~  

53: for emission lines having widths too narrow to be natural

54: from the host star, as well as for lines broadened by

55: astrophysical mechanisms.  We would have detected lasers that emit a

56: power, $P>$50 kW, for a typical beam width of $\sim$0.01 arcsec

57: (diffraction--limit from a 10-m aperture) if directed toward Earth

58: from the star.  No lines consistent with laser emission were found.

59: \end{abstract}  

60:   

61: \keywords{astrobiology, extraterrestrial intelligence --- techniques: spectroscopic}   

62: \received{} 

63: \accepted{} 

64: \slugcomment{Submitted to Pub.Ast.Soc.Pac.}  

65:   

66: \section{Introduction}   

67:    Most searches for signals from extraterrestrial intelligent

68: civilizations (SETI) have concentrated on radio wavelengths

69: (e.g. Tarter 2001, Cullers 2000, Werthimer et al. 2000, Leigh and

70: Horowitz 2000).  However, a preferred wavelength for SETI work has

71: proved elusive.  The motivation for this project stemmed from the

72: suggestion of Schwartz and Townes (1961) that advanced civilizations

73: might use optical or near--infrared lasers as an efficient means of

74: interstellar communication (Townes 1998, Lampton 2000).  The

75: advantages of optical wavelengths over radio include the higher

76: bandwidth and the narrower diffraction--limited beam size to promote

77: private communication.  In addition, the technology needed to produce

78: lasers with detectable powers is not far from current Earth technology

79: (\S 2.3).

80: 

81: Optical SETI (OSETI) projects are currently designed to detect highly

82: energetic, pulsed lasers and continuous--wave lasers. Such efforts

83: include dedicated searches, piggyback searches, and data mining.  Many

84: OSETI projects are designed to detect pulsed lasers using fast

85: photometry (Howard \et~2000, Bhathal 2000, Werthimer \et~2001).  Two

86: or three photodetectors are employed to search for laser pulses having

87: a duration of $\sim$1 ns in broadband visible light.  A high--powered

88: nanosecond pulsed laser will outshine its host star during its brief

89: pulse because at most one photon will be received (for a typical 1-m

90: class telescope) from the star during the 1 ns duration.

91: 

92: Alternately, interstellar optical communication may occur via

93: nearly continuous laser emission.  In that case, lasers of even

94: moderate power, tens of kilowatts, can outshine the host star 

95: when focussed by large apertures during transmission (see \S 2.3).  

96: Here we report a search for ultra narrow emission

97: lines such as predicted by the Schwartz--Townes suggestion.

98: 

99: 

100: \section{Optical SETI with High Resolution Spectroscopy}  

101:    We examine high resolution spectra of 577 F, G, K, and M

102: stars,   most of which reside within 50 pc.  We have taken

103: $\sim$12,500   spectra using the HIRES echelle spectrometer on the

104: 10--m Keck 1 telescope.  All target stars are listed in

105: Nidever et al. (2002).  The stellar spectra were obtained during the

106: period 1997--2001 in a search for extrasolar planets using

107: relative radial   velocities.  We are indebted to Drs.~R.~P.~Butler

108: and S.~S.~Vogt for their contributions in obtaining these Keck

109: spectra.  The multiple spectra from 4000~\AA~-- 5000~\AA~ 

110: are compared to detect sharp differences such as those caused by

111: nonnatural emission lines, as described in the following subsections.  

112: 

113: \subsection{High Resolution Echelle Spectra}  

114:   

115: The acquired echelle spectra contain 34 spectral orders, covering  

116: wavelengths 3800 \AA -- 6100~\AA~with resolution $\lambda / \Delta  

117: \lambda$ = 70,000 (see Vogt et al. 1994, Vogt et al. 2000, Butler et  

118: al. 2000).  Not all orders are used here because the 

119: starlight is sent through a glass cell  

120: filled with iodine vapor before entering the spectrometer for use

121: in setting the wavelength scale in the planet search.

122: The iodine  lines contaminate the stellar spectrum at wavelengths

123: greater than 4980 \AA.  The remaining  

124: usable orders, 72 through 89, contain 1000~\AA~ranging from 3980~\AA~--  

125: 4980~\AA.  This region includes Balmer $\beta$ and $\gamma$, but not

126: the CaII H\&K lines.   The pixel width is given by  

127: $\lambda/ \Delta \lambda$ = 145,000 , 

128: corresponding to 2.1 \kms per pixel or $\sim$0.032~\AA~per pixel.

129: The actual spectral resolution is only 70,000 because the

130: instrumental profile has a width of 2.3 pixels.

131:   

132: The recorded two--dimensional raw spectra are reduced in the usual way  

133: to one--dimensional spectra.  The bias is subtracted from all images  

134: including both the stellar spectra and the flat--field exposures.  Each  

135: spectral order recorded by the CCD has a width proportional to the  

136: diameter of the stellar seeing disk along the length of the slit. The  

137: typical seeing at Keck/HIRES of 0.7--1.2 arcsec yields orders of width  

138: $\sim$4 CCD rows (each ``row'' being the on--chip sum of neighboring  

139: original CCD rows).  Cosmic rays are removed by searching for  

140: 7--$\sigma$ sharp intensity spikes at one row compared to the  

141: intensity in a neighboring row (at the same wavelength).  We are  

142: indebted to Jason Wright for this cosmic--ray removal algorithm.  The  

143: stellar spectra are divided by the flat--field exposures.    

144: The two--dimensional spectral orders are compressed to one dimension by  

145: summing the recorded counts within the $\sim$10 rows  

146: centered on each order.  We thank Eric Nielsen for the careful reduction of

147: CCD images. 

148: Figure \ref{reduction} shows the results  

149: of this process. The signal--to--noise ratio per pixel in orders 72  

150: through 89 of a typical 1-D spectrum is $\sim$200.  

151:   

152: \subsection{Overview of the Detection Scheme}  

153:    

154: We search for laser emission lines in the reduced spectra by comparing

155: each spectrum to a reference spectrum of the same star, searching for

156: statistically significant differences.  Over time, a star's spectrum

157: should remain nearly constant except for minor fluctuations due to

158: photon noise and photospheric changes.  A large difference at one

159: wavelength could signify the presence of a laser signal that had been

160: detected during one observation but not during the other.  

161: 

162: We would fail to detect an emission line if it occured at the

163: same wavelength in both the observed and reference spectra.

164: A laser fired continuously from the rest frame 

165: of the stellar photosphere would go undetected.  However,

166: such a continuous laser would be detected in both spectra at different

167: wavelengths if the laser's velocity vector were to change between

168: observations, e.g. due to orbital motion around its host star.  The resulting

169: Doppler shift permit detections in the difference spectrum.

170:   

171: For each star, a series of algorithms is applied to all of its spectra.

172: First a ``reference'' spectrum is chosen.  Then, one by one, the other

173: ``test'' spectra are aligned in wavelength with the reference for

174: comparison (see \S 3.2).  Candidate laser lines are found in the test spectra by

175: subtracting the reference spectrum and searching for positive

176: differences, or ``spikes.''  One test spectrum is then used as a

177: secondary reference to check for spikes in the primary reference

178: spectrum.  The spikes are subjected to a selection criterion based on

179: their width, to ensure that their shapes are consistent with the

180: instrumental profile of HIRES (see \S 3.3).  Spikes that are narrower than the

181: instrumental profile are likely to be caused by cosmic rays hitting

182: the CCD that eluded detection in the CCD reduction process described

183: in \S 2.1.  

184: 

185: A monochromatic laser line would have an intrinsic width narrower than

186: the resolution of the spectrometer.  However, as the laser light

187: passes through the HIRES optical system, it would be smeared out by

188: the HIRES point spread function (PSF), giving the detected emission

189: line an apparent distribution in wavelength characteristic of the

190: HIRES PSF.  Thus candidate laser lines can be distinguished from most

191: cosmic rays, most CCD flaws, and stellar emission lines by their

192: predictable spread in wavelength that would be merely the HIRES PSF,

193: as described in detail by Valenti et al. (1995).  Notably, emission

194: lines having widths that are narrower than possible from a star (with

195: its thermal, collisional, turbulent, and rotational broadening

196: mechanisms) would stand out as possibly ``nonastrophysical'' in

197: origin.  All of the resulting candidate emission lines of any width,

198: excluding the apparent cosmic rays hits, from the spectra of each star

199: are stored for further analysis as described in \S 4.  In sections \S

200: 3.1, \S 3.2, and \S 3.3, we describe in detail the processes by which

201: we searched for emission lines.

202:   

203: \subsection{Detection Limits for Laser Emission}  

204: 

205: To be detected, a laser signal must compete against the light of its host star.

206: A laser beam carries three   advantages over starlight.  First, the

207: laser light is concentrated in a narrow cone of tiny solid

208: angle relative to  starlight, which is omnidirectional.  Second, a

209: monochromatic laser   line falls into one resolution element,

210: with a width of $\sim$2.3 pixels, while starlight, composed of

211: a wide range ($\sim$4000 \AA) of wavelengths, is spread out

212: into roughly 50,000 resolution elements.  Lastly, the laser 

213: needs only to outshine the photon noise of the stellar flux, not the

214: flux itself, within that resolution element since we are looking

215: at differences in the spectra.    To estimate these advantages

216: of the laser signal, we consider the   solid angles of the starlight

217: and a diffraction--limited laser beam.     

218: \begin{eqnarray}  

219:   \Omega_{\rm star}&=& 4\pi \\  

220:   \Omega_{\rm laser}&\approx& \Bigg({\lambda \over a}\Bigg)^2  

221: \label{omegas}  

222: \end{eqnarray}   

223:   

224: \noindent  

225: Here $\lambda$ is the wavelength of the laser light and $a$ is the  

226: aperture size of the laser transmitter.  The fraction of starlight contained  

227: within the laser beam is then given by  

228:   

229: \begin{equation}  

230: f_{\rm star}={1 \over 4\pi} \Bigg({\lambda \over a}\Bigg)^2    

231: \label{frac1}  

232: \end{equation}  

233:   

234: We note that the intrinsic width of a laser line will be much   

235: narrower than the HIRES resolution of $\sim$0.075~\AA,

236: thereby placing most all of the laser light into a single  

237: resolution element.  

238: The stellar spectral energy distribution   

239: may be approximated by a blackbody   

240: curve.   For a Solar--type star, the SED has a width of   

241: $\sim$4000~\AA, yielding the number of resolution elements, $N_{\rm star}$, into  

242: which the stellar luminosity falls:  

243:   

244: \begin{equation}  

245: N_{\rm star}\approx{4000\mbox{ \AA} \over 0.075\mbox{ \AA}}=5.3 \times 10^4  

246: \label{nstar}  

247: \end{equation}  

248:   

249: \noindent  

250: The fraction of starlight contained within one   

251: resolution element   

252: is therefore given by  

253:   

254: \begin{equation}  

255: f_{\rm star, \mbox{ }\Delta\lambda}={1 \over 4\pi}   

256: \Bigg({\lambda \over a}\Bigg)^2 {1 \over 5.3 \times 10^4}    

257: \label{frac2}  

258: \end{equation}  

259:   

260: The laser light needs only to outshine the stellar photon noise since we  

261: are detecting differences in a star's spectrum over time.  

262: The photon noise relative to the flux itself, $\sigma$, for one spectrum is given by  

263: 1/(S/N) where S/N is the signal to noise ratio per resolution element.   

264: Since the laser must compete with the photon noise of two spectra in one  

265: resolution element, we have  

266:   

267: \begin{equation}  

268: \sigma \approx {\sqrt{2} \over \mbox{S/N}}  

269: \label{noise1}  

270: \end{equation}  

271:   

272: Finally we find that the laser power necessary for an $n\sigma$   

273: detection is  

274:   

275: \begin{equation}  

276: P_{\rm laser}={1 \over 4\pi} \Bigg({\lambda \over a}\Bigg)^2   

277: {1 \over 5.3 \times 10^4} {n \sqrt{2} \over \mbox{S/N}} L_{\rm star}  

278: \label{power1}  

279: \end{equation}  

280:   

281: \noindent  

282: where $L_{\rm star}$ is the luminosity of the star.  As a representative case,   

283: we take $\lambda$=4500~\AA, S/N=200, n=6, $L_{\rm star}=L_{\rm sun}=

284: 3.8~\times~10^{26}$ W, and assume Keck to Keck communication 

285: so $a$=10 m, which gives     

286:   

287: \begin{equation}  

288: P_{\rm laser} = 50 \mbox{ kW}  

289: \label{power2}  

290: \end{equation}  

291:   

292: \noindent  

293: Thus for this representative case, one can detect a 50 kW laser.  

294:   

295: The previous calculation assumes that   

296: the laser is on continuously.  If, however,  

297: a nanosecond pulsed laser signal were transmitted, Equations \ref{power1}  

298: and \ref{power2} would need to be multiplied by the factor  

299: ${\Delta t_{\rm exp} \over {N_{\rm pulses} \times 10^{-9}\mbox{ s}}}$  

300: where $\Delta t_{\rm exp}$ is the exposure time of the observation and $N_{\rm pulses}$  

301: is the number of pulses received in that time.  Using a typical exposure time  

302: of 10 minutes and assuming 1 pulse is received, the minimum necessary power of a  

303: detectable nanosecond pulsed laser is $P_{\rm pulsed}=3 \times 10^{16}$ W.    

304: Thus, the minimum necessary energy of

305: a detectable nanosecond pulsed laser is 30 MJ.

306: 

307: For comparison, the most powerful continuous laser on Earth is the Free Electron

308: Laser (FEL) developed at Jefferson Laboratory in Virginia.  The FEL has 

309: reached an average power of 1.7 kW and the laser is expected to achieve 

310: average powers up to $\sim$10 kW with an upgrade (Jefferson Lab 2001).  

311: Lawrence Livermore National Laboratory produced the world's most powerful 

312: pulsed laser, the Petawatt laser.  During its 440 femtosecond pulse, it 

313: delivered a power greater than 10$^{15}$ W (LLNL 2001). 

314:   

315: We can also determine a minimum detectable flux, $F_{\rm laser}$,

316:   

317: \begin{equation}  

318: F_{\rm laser} = {P_{\rm laser} \over A_{\rm laser}}  

319: \end{equation}  

320:   

321: \noindent  

322: where $A_{\rm laser}$ is the area of the laser beam at the Earth   

323: and given by  

324:   

325: \begin{equation}  

326: A_{\rm laser}={\pi \over 4} \Bigg({\lambda \over a}\Bigg)^2 d^2  ,

327: \end{equation}  

328:   

329: \noindent   

330: where $d$ is the distance from the laser to Earth.  The flux limit  

331: can now be expressed as  

332:   

333: \begin{equation}  

334: F_{\rm laser} = {4 \over \pi}{1 \over 5.3 \times 10^4}{n\sqrt{2} \over \mbox{S/N}}  

335: {L_{\rm star} \over {4 \pi d^2}} ,

336: \end{equation}  

337:   

338: \noindent  

339: where $L_{\rm star}/ 4 \pi d^2$ is the flux of the star.  

340: If we take $d=100$ light  

341: years with the previously adopted values, the minimum flux of a detectable  

342: laser is  

343:   

344: \begin{equation}  

345: F_{\rm laser} = 3.4 \times 10^{-17} \mbox{ W/m$^2$}  

346: \end{equation}  

347:   

348: \noindent  

349:   

350: Thus, we can detect monochromatic laser light from nearby stars  

351: if the flux is at least $3.4~\times~10^{-17}$~W/m$^2$ at Earth.  

352:   

353: \section{Spectral Analysis}  

354:   

355: A detailed description of the algorithms developed to analyze  

356: the spectra is given in this section.  

357:   

358: \subsection{Choosing Reference Spectra}  

359:    Two reference spectra are required for each star.  The primary  

360: reference spectrum is compared to all the other test spectra.    

361: The second is compared to the primary reference spectrum.  Since

362: only   positive differences are considered, any candidate laser

363: lines   in the first reference would not be flagged during its

364: comparison   with the test spectra.  A second reference, one of the

365: previous   test spectra, is used to account for this

366: possibility.

367: 

368: The two reference spectra are chosen as follows.  Almost every  

369: star has at least one ``template'' spectrum.  In the planet--search  

370: project, the stellar radial velocities are measured with respect  

371: to the template spectra.  These observations are of good quality with

372: high signal--to--noise ratios.  If there are multiple--template spectra for

373: a    star, the two with the highest signal to noise are chosen as

374: the    reference spectra.  If there is only one template, it is used

375: as the    primary reference spectrum and another observation with

376: high signal    to noise is chosen as the second reference.  Some

377: stars which are new    to the extrasolar planet search catalog do

378: not yet have templates.  For    these stars, the two observations

379: with the highest signal to noise are    chosen as the references.

380:      

381: \subsection{Aligning Spectra}  

382:   

383: Aligning a pair of spectra precisely in both wavelength and flux is   

384: important.  If the spectra are exactly the same, except for minor   

385: fluctuations due to photon noise, the fractional   

386: standard deviation, $\sigma$, of the difference spectrum is expected to be

387:   

388: \begin{equation}  

389: \sigma=\sqrt{{1 \over { \mbox{(S/N)}}_1^2}+{1 \over { \mbox{(S/N)}}_2^2}}  

390: \end{equation}  

391:   

392: \noindent  

393: where (S/N)$_1$ and (S/N)$_2$ are the signal--to--noise ratios of   

394: the two spectra.  For two spectra with signal--to--noise ratios of   

395: 300, $\sigma$ can, in principle, be as low as 0.005.  It is important  

396: to align the two spectra well enough such that the observed  

397: error is as close to the theoretical error as possible in order to   

398: detect the smallest differences.  

399: 

400: 

401: We line up a pair of spectra order by order.  Various normalization  

402: and shifting algorithms are applied to both the reference and test  

403: spectra, as follows.  

404: The first step is to put the orders on roughly the same flux  

405: scale by dividing by the continuum.  The continuum is found by first  

406: partitioning an order into 10 bins.  Each bin is assigned a temporary  

407: continuum at the 85th percentile in flux.  A third--order polynomial is  

408: fit to these individual continua to normalize the entire spectral  

409: order.  

410: 

411: Each spectrum is Doppler--shifted to correct for the Earth's orbital

412: motion, which causes a maximum wavelength shift of $\sim$1 \AA~

413: corresponding to $\sim$30 pixels.  We determine the Doppler shift of

414: the test spectrum relative to the reference spectrum by minimizing

415: $\chi^2$, and then shift the test spectrum so that the two are

416: aligned.  The final shifts are precise to $\sim$0.01 pixel, and

417: fractional pixel displacements of the spectra are accomplished with a

418: spline interpolation.  This shifting is first carried out on segments

419: of length 280 pixels, and then it is done on a 40--pixel scale

420: (``chunks'').  If any spectrum contains cosmic rays (or laser emission

421: lines), it is difficult to align.  Such features are removed

422: temporarily (replaced with a spline interpolation) to carry out the

423: alignment.  Operating on chunks of 40 pixels allows us to account for

424: the wavelength dependence of the Doppler shift and to correct any

425: slopes in the flux with wavelength.

426: 

427: 

428: The resulting (temporarily spike--free) reference and test spectra  

429: are subtracted to yield a difference spectrum, from which  

430: the standard deviation is determined and recorded as the  

431: ``noise'' for each spectral chunk.    

432: Finally the reference and test spectra are subtracted with all spikes  

433: included, leaving a difference chunk which can  

434: be searched for laser--line candidates.  Figure 2 shows a chunk of  

435: difference spectrum both with (top) and without (bottom)  

436: flux spikes of unidentified origin.  

437: 

438: 

439: \subsection{ Criteria for Laser Line Candidates}

440: 

441: 

442: A laser beam would travel through the telescope and the HIRES  

443: spectrometer optics just as starlight would.  Moreover, the laser

444: beam   travels through the same column of Earth's atmosphere,

445: acquiring the same wavefront distortions and seeing profile

446: as does the star.   Therefore, an emission line from a laser would

447: span the whole width of   a spectral order (perpendicular to

448: dispersion) as would the stellar spectrum on the CCD.  While the

449: intrinsic width of the laser line in wavelength would be smaller than

450: the resolution of HIRES, it would be broadened by the point--spread

451: function of HIRES.  The B1 slit used in HIRES projects to a width of

452: 2.0 pixels on the CCD.  The instrumental profile of HIRES from both

453: its optics and the slit width is known from careful measurements

454: (Valenti et al. 1995) and from simple Gaussian fitting of emission

455: lines from a thorium-argon lamp.  The instrumental profile with our

456: HIRES setup has a width of 2.3 pixels (FWHM).  Thus, we expect any

457: laser line to have an observed width of 2.3 pixels.

458: 

459: We search for candidate emission lines in the stellar spectra as

460: follows.   Each 40--pixel chunk of difference spectrum has a

461: corresponding   fractional photon noise, $\sigma$, associated with

462: it as described in   \S 3.2.  We pass through all the chunks and

463: note spikes having   heights above $6\sigma$.  This threshold is

464: chosen to exclude   essentially all noise but detect a signal from a

465: laser with the lowest   detectable power.    

466: 

467: We also require that candidate laser lines exhibit a width that

468: is   at least as broad as the HIRES instrumental profile, with

469: FWHM=2.3 pixels, allowing us to discriminate 

470: against   (i.e., reject) cosmic--ray

471: hits.  Most cosmic rays produce electrons in   a single pixel or

472: occasionally in two neighboring pixels, clearly   narrower than the

473: HIRES PSF.  To implement this ``width'' criterion,   we require that

474: the shorter of the two adjacent pixels on either side   of each

475: 6--$\sigma$ spike contain at least 30\% of the photons in that  

476: central spike.  Examination of the known HIRES PSF and thorium  

477: comparison lines (see Figure \ref{thor}) shows that this width  

478: criterion would be met by almost any emission line that had been

479: broadened   by the HIRES PSF.  Any spike in the difference spectrum

480: that does not   meet both the 6--$\sigma$ height criterion and this

481: width criterion is   rejected from future consideration as they are

482: probably cosmic rays.  Spikes   that do meet both criteria are

483: stored in a database that includes the   location and

484: characteristics of that spike, for later statistical analysis (\S

485: 4).  We note that emission lines having widths {\em   greater} than

486: the PSF will be detected and stored as candidate   signals.  Thus

487: our survey would detect intermittent emission lines of   any origin,

488: laser or astrophysical, if they are sufficiently strong.

489: 

490: \section{Statistical Analysis of Candidates}

491: 

492: The spectral analysis described in \S 3 yields a   database of

493: surviving candidate emission lines in each   of the $\sim$12,500

494: Keck spectra of our 577 target stars.  A typical star has 

495: $\sim$13 surviving candidate emission lines in its $\sim$20

496: spectra.  For each candidate, the

497: database contains the   star, date of observation, the wavelength,

498: the peak number of photons,    and the width of the candidate

499: line.   Detailed examination shows that most (if not all)    of

500: the surviving candidate   emission lines are simply cosmic ray hits

501: that we had failed to reject.  

502: 

503: To discriminate a true emission line from these cosmic rays,  

504: we apply a statistical analysis on the collection of emission--line   

505: candidates from all spectra (typically 20) for a given star.    

506: We search for an overabundance of  

507: candidates (``hits'') in the vicinity of any particular wavelength.  

508: In contrast, cosmic rays would hit at random locations on the CCD,  

509: and would not favor any particular wavelength.  

510: A clustering of hits near one wavelength from all the spectra of  

511: one star could indicate laser emission.  

512:   

513: 

514: We note that the Earth's orbital motion will not cause a Doppler   

515: shift of the laser line in our analysis, as we have already forceably

516: shifted the reference and test stellar spectra onto the same wavelength scale (\S 3).  

517: However, acceleration of the emitter relative to the star   

518: (such as due to orbital motion) during the  

519: typical 4 years of observations would cause the  

520: laser line to be Doppler--shifted to neighboring wavelengths.  

521:   

522: 

523: We make a grid representing the CCD with the spectral orders partitioned

524: into approximately 30 segments of roughly 100 pixels each.  The  

525: segments overlap one another by half of their size to account for the  

526: possibility of an abundance of hits split into two neighboring  

527: segments.  We record the number of emission--line candidates in each segment.  

528:   

529: 

530: For each star, we plot the number of 100--pixel segments against the

531: number of hits in the segments.  Most segments have no hits at all,

532: and only a few have one hit.  These infrequent hits (none or one hit

533: out of $\sim$20 observations in a wavelength segment) are consistent

534: with random cosmic ray hits.  To this histogram, we fit a Poisson

535: distribution (see Figure \ref{hist}), and determine the probabilities

536: of finding an excess number of hits, $n$, in any wavelength segment.

537: A star exhibiting an excess number of hits in a segment, i.e. a low

538: probability of the events occurring randomly, would stimulate further

539: investigation as a potential emission line.

540:   

541: 

542: The fit of a Poisson distribution to the histogram of hits is done as  

543: follows.  We fit the low--$n$ (0, 1, 2) portion of the histogram and  

544: neglect the high--$n$ tail.  This approach avoids contaminating the  

545: Poisson mean, $\mu$, with actual emission lines in the high--$n$ tail.  

546: The first empty bin in the histogram, $n_0$, is determined and an  

547: estimate of the expected mean number of hits per segment, $\mu$, is  

548: calculated neglecting those segments with $n \ge n_0$.  

549: 

550: The Poisson probability of finding $n$ cosmic ray hits   

551: in any given wavelength segment is given by (Taylor 1982):  

552:   

553: \begin{equation}  

554: P_{\mu}(n)=e^{-\mu} {\mu^n \over n!}  

555: \label{poisson}  

556: \end{equation}  

557:   

558: \noindent  

559: Using the estimated $\mu$, we calculate $P_{\mu}(n)$ for $n=0, 1, ...,  

560: n_{0}-1$, and check if any of the probabilities are less than 1\%.  If  

561: so, we exclude those bins from the fit as well if $n_0>2$.  The fit   

562: is then applied to bins from $n=0$ to some $n_{\rm max}$ hits.  

563:   

564: The estimated mean number of hits per segment  

565: may not necessarily yield a Poisson distribution that gives 

566: the best fit to the observed histogram.    

567: The best Poisson fit may be better determined by finding 

568: the lowest $\chi^2$ given by   

569:   

570: \begin{equation}  

571: {\chi^2}=\sum_{n=0}^{n_{\rm max}} {(O_n - E_n)^2 \over E_n}  

572: \label{chi2}  

573: \end{equation}  

574:   

575: \noindent

576: where $O_n$ is the observed number of segments containing $n$  

577: candidate hits and $E_n$ is the expected number with $n$ hits, based  

578: on a Poisson distribution.  The only free parameter is $\mu$, the mean  

579: number of hits per segment.  We vary $\mu$ to find the value  

580: corresponding to the minimum of $\chi^2$.  The resulting Poisson  

581: probabilities are multiplied by the total number of segments and  

582: adopted as the best fit to the observed histogram.  

583:   

584: We now determine the probabilities of finding $n$ hits in any given  

585: segment, extrapolating beyond $n_{\rm max}$.  Segments having an  

586: excessive number of hits with probabilities less than 0.01\% are  

587: noted as serious candidate emission lines, as they occur in excess  

588: numbers above the background cosmic ray hits.  Since there are about  

589: 500 wavelength segments on our grid of the CCD for each star, we  

590: expect to flag one in 20 stars spuriously, i.e. due to a chance  

591: coincidence of cosmic rays.  

592:   

593: We find that indeed 45 stars were flagged as having a wavelength  

594: segment with an excess number of hits at the 0.01\% false-alarm level.  

595: Each of the surviving flagged candidate laser lines, of which there  

596: were 187, were examined by eye in the raw CCD images.  

597:   

598: \section{An Apparent Emission--Line Detection}  

599:   

600: Our search for intermittent emission lines revealed an apparent  

601: detection for the star HD29528 (K0V).  Three separate spectra  

602: obtained on different dates contained strong candidates at the same  

603: wavelength.  The hits were noted in both of two overlapping spectral  

604: segments for this star. The probability that these three apparent  

605: emission lines were merely spurious cosmic rays occurring within a  

606: segment of length $\sim$200 km s$^{-1}$ is easily calculated from the  

607: Poisson occurrence of cosmic rays (\S 4).  

608: 

609: 

610: Figure \ref{hist} shows the observed histogram of the number of  

611: segments versus number of hits along with the best fitting Poisson  

612: distribution.  The probability of randomly obtaining 3 hits in a  

613: segment for this star is 1.23 $\times$ 10$^{-7}$ or 1 in 8 million.  

614: 

615: 

616: Figure \ref{las1} shows these candidate laser lines in the reduced  

617: spectra of the three observations.  The aligned reference and test  

618: spectra are shown and twice the difference spectra are below them.  

619: The candidate laser lines are in spectral order 83 and the central  

620: peaks are in pixel 1436 with respect to the reference spectrum.  

621: 

622: 

623: We examined the raw spectra to see if the candidate laser lines span  

624: the whole width of the spectral order, indicating whether the  

625: light passed through the telescope and spectrometer.  Figure  

626: \ref{fakelas} shows the raw CCD image of the spectrum of HD29528  

627: obtained on JD 2451793.1, centered on the first candidate laser line.  

628: The reference spectrum is shown below it.  Beneath the reference is  

629: the raw difference spectrum making it easier to see the apparent laser  

630: line.  The other two candidate laser lines found in observations  

631: obtained on JD 2451882.98 and on JD 2451884.08 are not shown, but  

632: those apparent emission lines also span the width of the spectral  

633: order.  Because all three hits occur at the same pixel (and hence  

634: wavelength) with respect to the star's reference spectrum, and because 

635: they   span the entire width of the spectral order, the hits are unlikely to  

636: be cosmic rays.  These are surviving laser line candidates thus far in  

637: the analysis.

638: 

639: To further investigate these candidate emission lines in HD29528, we  

640: compared the spectrum of HD29528 to the solar spectrum using ``{\it  

641: The Solar Spectrum 2935~\AA~to 8770~\AA}'' (Moore \et $~$1966).  The  

642: rest wavelengths of the lines are 4307.31~\AA~and there are no listed  

643: absorption lines due to the Earth's atmosphere nor anomalous Solar  

644: emissions lines at this wavelength.  

645: 

646: We further examined the spectrum of a rapidly rotating B--type star  

647: taken on the same night as the reference spectrum for HD29528.  The B  

648: star spectrum should have no prominent spectral features.  A strange  

649: drop in flux of 20\% is seen in the spectrum, exactly at pixel 1436 in  

650: order 83, precisely where the hits were found with respect to the  

651: reference spectrum of HD29528 (see Figure \ref{bstar}).  This clearly  

652: spurious feature in the B star strongly suggests that our candidate  

653: laser lines are merely artifacts caused by a flaw in the CCD.  

654: 

655: Indeed, a close examination of the raw CCD image of the reference  

656: spectrum of HD29528 shows the flaw in the CCD (see Figure \ref{flaw}).  

657: This flaw is certainly the source of the false detection.  

658:   

659: As a side note to this false detection of an emission line, one may  

660: calculate the necessary power of a laser that could have produced a  

661: ``signal'' similar to that detected.  In Eqn \ref{power1} we assume a  

662: transmitter aperture, $a$=10 meters.  The other variables are known.  

663: The wavelength of the 13$\sigma$ ``signal'' is 4307~\AA, and the  

664: signal--to--noise ratio of the spectra is $\sim$120.  The luminosity of a K0  

665: star is 0.42 $L_{\rm sun}$ (Carroll and Ostlie 1996).  Using these values  

666: in equations \ref{power1}, we find $P_{\rm laser}=68$ kW.  This would be the  

667: required power of a laser to produce the spurious signal in Figure  

668: \ref{fakelas}.  This test supports our expectation that lasers having  

669: power of 50 kW would have been detected.  

670:   

671: \section{Results}  

672: 

673: For each of the 45 stars that exhibited an excess of candidate  

674: emission lines (above the cosmic ray background noise) we examined the  

675: original raw CCD images from HIRES.  Approximately 200 raw spectra were  

676: examined to follow up on the 45 flagged stars.  

677: Stellar emission lines were detected in the spectra of two flare stars,

678: HIP5643 (M4.5) and HIP92403 (M3.5).  Both stars exhibited

679: varying emission of the Balmer line H$\beta$. 

680: 

681: Interestingly, many  

682: candidate emission lines were located along the steep slopes of  

683: absorption lines in the reduced spectra.  However, the corresponding  

684: emission spikes were not apparent in the raw CCD image.  Such emission  

685: spikes are apparently neither cosmic rays nor emission lines.  

686: Instead, it is likely that these spikes located in absorption lines  

687: are caused simply by tiny changes in the instrumental profile of HIRES  

688: from one observation to the next.  The changes in the instrumental  

689: profile could result from different atmospheric seeing of the star  

690: image at the slit (making the slit effectively a different size) or  

691: from small focus changes in HIRES itself.  The steep sides of  

692: absorption lines have local intensities that are very sensitive to the  

693: instrumental profile.  

694: It is also possible that the spikes located

695: in absorption lines are residual numerical artifacts from the

696: spectral alignment process since interpolation routines normally

697: give the largest errors on steep slopes.  

698: Thus these candidate emission spikes located  

699: within absorption lines were rejected as spurious.

700: 

701: All other candidate emission lines were visually inspected and found  

702: to be inconsistent with actual emission lines.  Some were multiple,  

703: surviving cosmic rays that occurred, by chance, near the same  

704: wavelength.  Some emission spikes were caused by bad pixels or flaws  

705: in the CCD.  {\em No lines consistent with laser emission

706: were found in any of our stars.}

707: 

708: \section{Discussion}

709: 

710: We would have detected time--varying emission lines from any of

711: the 577 main sequence stars above a threshold of a few percent of

712: the continuum flux.  Emission lines of constant 

713: intensity would also have been detected 

714: had they Doppler--shifted relative to the star 

715: during the $\sim$4 yr of observations.  Lines having

716: intrinsic widths from arbitrarily narrow to tens of \kms would have

717: been detected.  No narrow lines consistent with laser emission were 

718: detected.  However, naturally occurring emission lines were found

719: in the spectra of two stars.  

720: 

721: Emission lines in the spectra of Solar--type stars are rare between

722: 4000~\AA~-- 5000~\AA.  Such lines can arise occasionally in the

723: chromospheres or coronae of the most magnetically active stars.  In

724: some ($\sim$20\%) F, G, K--type main sequence stars, the cores of the

725: Balmer absorption lines are ``filled in'' by a few percent due to

726: chromospheric emission (Herbig 1985, Robinson, Cram, \& Giampapa

727: 1990).  We would have detected temporal variation in that

728: chromospheric filling, if it exceeded a few percent of the continuum

729: level.  No such variations were detected in any of our F, G, K--type

730: main sequence stars.  This nondetection indicates that such chromospheric

731: variation in the cores of the strong Balmer lines, notably H$\beta$

732: and H$\gamma$, did not exceed a few percent of the continuum level

733: in these stars.  Flare stars, most of spectral type M0 or later,

734: more commonly exhibit emission of H$\beta$.  Indeed, we detected

735: time--varying emission of H$\beta$ in two M--type flare stars.

736: 

737: Time--varying emission lines from stars may also arise from infalling

738: volatile material and comets, as is observed for $\beta$ Pic (see for

739: example Beust et al. 1998).  Our search would have detected such

740: astrophysical emission lines if they appeared or varied by more than a

741: few percent of the continuum flux of the star.  All such astrophysical

742: origins for emissions lines would yield line widths greater than our

743: resolution of 5 \kms due to the usual thermal and turbulent

744: broadening, as well as possible rotational and collisional broadening.

745: Hence such lines would be resolved.  Nonetheless, no emission lines due

746: to these effects were found in the 577 stars at our threshold of a

747: few percent of continuum flux.

748: 

749: The entrance slit of the HIRES spectrometer is rectangular with

750: dimensions 0.57 $\times$ 3.5 arcsec.  At the typical distance of our

751: stars of $\sim$50 pc, this slit corresponds to 30--175 AU at the star.

752: Thus, sources of emission lines residing within 30 AU of our target

753: stars would have been included in our spectra.  Of course, the

754: emission beam must be pointed toward the Keck telescope for detection.

755: Extinction from interstellar dust is negligible toward these target

756: stars due to their proximity ($d<$100 pc).

757: 

758: The motivation for this search stemmed from the suggestion of Schwartz

759: and Townes (1961) that advanced civilizations might communicate from

760: one colony to another within a planetary system, or from one planetary

761: system to another, by using optical lasers as the carrier.  The

762: advantages of optical communication over radio include its higher

763: bandwidth and its smaller diffraction beam to promote private

764: communication.

765: 

766: A nondetection of extraterrestrial intelligence carries no value

767: unless it excludes a plausible model of life in the universe.  One

768: might imagine the following model.  The Milky Way Galaxy has an age of

769: $\sim$10 Gyr, while only 4.6 Gyr were required to spawn a

770: technological species on Earth.  Approximately half of the stars in

771: the Galactic disk are older than the Sun, and they formed with

772: comparable amounts of heavy elements.  Approximately 50\% of these old

773: disk stars do not have a stellar companion within 100

774: AU (Duquennoy and Mayor 1991), thereby permitting stable planetary

775: orbits.  Therefore, $\sim$25\% of the stars in the Galactic disk have

776: requisite characteristics of adequate age, chemical composition, and

777: dynamical quiescence to serve as sites for the development of

778: technological species.  The anthropocentric nature of these stellar

779: criteria are apparent, and thus serve merely as a useful guide for a

780: model.

781: 

782: Coincidentally, the target stars used here are systematically

783: middle--aged or older and void of close companions.  They were

784: selected for the Doppler planet search at Keck from among stars in the

785: Solar neighborhood.  They were chosen to have ages greater than 2 Gyr,

786: as judged from the CaII H\&K chromospheric emission line, because of

787: the resulting photospheric stability of older stars.  Moreover the

788: target stars have no known stellar companions within 2 arcsec (typically

789: $\sim$50 AU) to prevent any companion's light from entering the

790: spectrometer slit (but see Vogt et al. 2002 and Nidever et al. 2002

791: for a few recently detected, but negligibly bright, close companions).

792: Moreover, spectroscopic analyses of the stellar spectra show that 90\%

793: of the target stars contain amounts of heavy elements that are 

794: within a factor of 2 of the Solar value ([Fe/H] = -0.3 to +0.3;

795: D.A.Fischer, private communication).

796: 

797: Thus, the majority of the 577 target stars for this SETI search are

798: older than the Sun, have roughly Solar chemical composition, and have

799: no perturbing stellar companions within 50 AU.  This stellar sample is

800: clearly biased.  Thus our observations test a limited model of the

801: Galaxy in which older, single stars harbor technological

802: civilizations.

803: 

804: The speculative model posits that some fraction of these target stars

805: contain civilizations that sent probes to, or established colonies on,

806: planets, moons or other platforms either in their host 

807: planetary system or in other systems.  

808: This model of behavior, albeit anthropocentric,

809: stems from the tendencies of Homo Sapiens toward

810: exploration.  To communicate with such probes or colonies over

811: distances of Astronomical Units to parsecs, some fraction of these

812: civilizations might be modeled as using electromagnetic waves, 

813: including optical lasers.

814: 

815: We model the solid angle subtended by laser beams as follows.  Some

816: fraction, $f_{\rm tech}$, of the total number of target stars, $N_{\rm

817: stars}$=577, harbor technological civilizations.  On average each such

818: civilization emits some number of laser beams, N$_{\rm beams}$,

819: perhaps in arbitrary directions.  We consider that each beam subtends

820: an average solid angle $\Omega_{\rm beam}$.

821: 

822: The total solid angle, $\Omega_{\rm Total}$, 

823: subtended by all arbitrarily oriented beams is

824: 

825: \begin{equation}

826: \Omega_{\rm Total} = N_{\rm stars} f_{\rm tech} N_{\rm beams} \Omega_{\rm beam}

827: \end{equation}

828: 

829: \noindent

830: We ignore here details of overlap of the beams, their actual distribution

831: of power and beam size, and any purposeful aiming toward or away from us.

832: Any properties of the lasers that are constructed with

833: the Earth as a consideration could drastically change the

834: probability of interception of such beams.

835: 

836: In our survey, we have monitored $N_{\rm stars}$= 577 stars. 

837: The nondetection obtained in the survey here suggests that the

838: total solid angle subtended by lasers emitting 

839: 50 kW or more is constrained as 

840: $\Omega_{\rm Total} < 4\pi$ sr .

841: 

842: Thus for arbitrarily directed beams, we  

843: constrain the model as

844:   

845: \begin{equation}

846: f_{\rm tech} N_{\rm beams} \Omega_{\rm beam} < 4\pi / 577 

847: \end{equation}

848:   

849: The nondetection reported here suggests lasers emitting

850: detectable power must satisfy the above constraint for the

851: combination of $f_{\rm tech}$,

852: $N_{\rm lasers}$, $\Omega_{\rm beam}$.  

853: Apparently, advanced civilizations

854: communicating by kilowatt optical lasers are not so common as to fill 4$\pi$ sr.

855: 

856: As a touchstone, we consider the case of the 

857: diffraction--limited laser with an aperture of 10-m (Keck--to--Keck),

858: giving a solid angle for one beam, $\Omega_{\rm beam} \approx 10^{-15}$ sr.  

859: For transmitters of such aperture, a laser emitting 50 kW would be detectable.

860: For this case, each star would have to emit $\sim10^{13}$ arbitrarily

861: oriented beams on average, each with power $P \ge$ 50 kW, in order 

862: that one would likely be oriented, serendipitously, toward us.

863: 

864: Clearly, the observed nondetection implies that such numerous,

865: narrow beams, however implausible, do not exist in sufficient numbers

866: to permit likely detection.  These observations

867: offer a poor constraint on lasers of arbitrary orientation, as their

868: solid angles are simply too small in this model.  One might consider

869: a Galactic model consisting of wider laser beams, enhancing the

870: chances of interception.  In that case,

871: the present nondetection would impose a constraint

872: of fewer lasers emanating from the target stars, but would

873: require proportionally higher power per laser for detection

874: at 6--$\sigma$ above the stellar continuum.  

875: 

876: More meaningfully, the nondetection reported here suggests that none of

877: the 577 target stars harbors a civilization that has purposefully

878: directed a laser with $P \ge $50 kW toward Earth (for the case of 

879: a 10--meter diffraction--limited transmitting aperture).  This modest detection

880: threshold of 50 kW highlights the ease with which an advanced

881: civilization could signal us, if desired.  But such is apparently not

882: the case.

883: 

884: %As an example, we might consider \Omega_{\rm beam} to be that given by  

885: %optical light from a diffraction--limited 10-m aperture transmitter.  

886:   

887:   

888: \section{Acknowledgements}  

889:   

890: 

891: We thank the SETI Institute for generous funding of this work.  We

892: wish to thank R.Paul Butler and Steve Vogt for their   endless hours

893: in obtaining the Keck/HIRES spectra.   S.Vogt's design of HIRES made

894: this work possible.   We thank Jason Wright for the new CCD

895: reduction algorithm of Keck spectra that efficiently removed cosmic

896: ray hits.  We thank Eric Nielsen for reducing the 12,500 CCD images

897: to one--dimensional spectra.  We thank Debra Fischer for chemical--abundance 

898: measurements of the target stars.  We also thank D.Nidever,

899: D.Werthimer, J.Tarter, F.Drake, M.Lampton, B.Walp and referee, William D. Cochran 

900: for many useful comments.   This work 

901: was carried out by A.R. as partial fulfillment of the of the

902: requirements toward a Master's Degree in physics at San Francisco

903:  State University.    

904: 

905: \begin{thebibliography}{}  

906:   

907: \bibitem[]{}  

908: Beust H., Lagrange A.-M., Crawford I.A., Goudard C., Spyromilio J., Vidal-madjar A. 

909: 1998, AA, 338, 1015

910: 

911: The {beta} Pictoris circumstellar disk. XXV. The CaII absorption lines and the Falling 

912: Evaporating Bodies model revisited using

913: UHRF observations. 

914: 

915: 

916: \bibitem[]{}  

917: Bhathal, R. 2000, in {\it Bioastronomy 99: A New Era in Bioastronomy},  

918: eds.~G. Lemarchand \& K. Meech, ASP Conf.~Proc., vol.~213, 553.  

919:   

920: \bibitem[Butler { et~al.} 2000]{BuMa00}  

921: Butler, R.~P., Vogt, S.~S., Marcy, G.~W., Fischer, D.~A.,  

922: Henry, G.~W. \& Apps, K. 2000,  

923: \newblock { ApJ, } {545}, 504.  

924:   

925: \bibitem[]{}  

926: Carroll, B. and Ostlie, D. 1996, in An Introduction to  

927: Modern Astrophysics, Addison-Wesley Publishing.     

928:   

929: \bibitem[]{}  

930: Cullers, K. 2000, in {\it Bioastronomy 99: A New Era in Bioastronomy},  

931: eds.~G. Lemarchand \& K. Meech, ASP Conf.~Proc., vol.~213, 451.  

932:   

933: \bibitem[]{}  

934: Duquennoy, A. and Mayor, M. 1991, A\&A, 248, 485

935: 

936: \bibitem[]{}  

937: Herbig, G. H. 1985, ApJ, 289, 269

938: 

939: \bibitem[]{}  

940: Howard, A., Horowitz, P., Coldwell, C., Klein, S., Sung, A.,  

941: Wolff, J., Caruso, J., Latham, D., Papaliolios, C., Stefanik, R., \& Zajac, J.  

942: 2000, in {\it Bioastronomy 99: A New Era in Bioastronomy},   

943: eds.~G. Lemarchand \& K. Meech, ASP Conf.~Proc., vol.~213, 545.  

944: 

945: \bibitem[]{}

946: Jefferson Lab. 2001. ``Free-Electron Laser Program.''

947: Obtained 6 January 2002 \\

948: $<$http://www.jlab.org/FEL/feldescrip.html$>$.  

949: 

950: \bibitem[]{}  

951: Lampton, M. 2000, in {\it Bioastronomy 99: A New Era in Bioastronomy},  

952: eds.~G. Lemarchand \& K. Meech, ASP Conf.~Proc., vol.~213, 565.  

953: 

954: \bibitem[]{}

955: Lawrence Livermore National Laboratory (LLNL). 2001.

956: ``The Amazing Power of the Petawatt.''  Obtained 6 January 2002 

957: $<$http://www.llnl.gov/str/MPerry.html$>$. 

958:     

959: \bibitem[]{}  

960: Leigh, D. and Horowitz, P.  2000, in {\it Bioastronomy 99: A New Era in Bioastronomy},  

961: eds.~G. Lemarchand \& K. Meech, ASP Conf.~Proc., vol.~213, 459.  

962:   

963: %\bibitem[]{}  

964: %Marcy, G. W. 1997, in {\it CYCLOPS II, Optical SETI Workshop},   

965: %Space Sciences Lab, University of California, Berkeley.   

966:   

967: \bibitem[]{}  

968: Moore, E., Minnaert, M. G. J., Houtgast, J. 1966, in   

969: The Solar Spectrum 2935~\AA~to 8770~\AA.  

970: National Bureau of Standards Monograph 61.   

971:   

972: \bibitem[]{}  

973: Nidever, D., Marcy, G.~W., Butler, R.~P., Fischer, D.~A., Vogt, S.~S. 2002,   

974: submitted to ApJ Supp.  

975:   

976: \bibitem[]{}  

977: Reines, A. E. 2001, Master's thesis, San Francisco State University  

978:   

979: \bibitem[]{}  

980: Robinson, R. D., Cram, L. E., Giampapa,M. S. 1990, ApJS, 74,891

981: 

982: \bibitem[]{}  

983: Schwartz, R. M. and Townes, C. H. 1961, Nature, 190, 205.   

984:   

985: \bibitem[]{}  

986: Tarter, J. 2001, ARAA, 39, 511-548.

987: 

988: \bibitem[]{}  

989: Taylor J. R. 1982, in An Introduction to Error Analysis:  

990: The Study of Uncertainties in Physical Measurements, University  

991: Science Books, ch. 11 \& 12.   

992:   

993: \bibitem[]{}  

994: Townes, C. 1997, in {\it Proc. 5th Int. Conf. on Bioastronomy},  

995: eds.~C. Cosmovici, S. Bowyer, \& D. Werthimer., IAU Colloq. No.~161, 585.  

996:   

997: 

998: \bibitem[Valenti {  et~al.} 1995]{val:95}  

999: Valenti, J. A., Butler, R. P., \& Marcy,G.W. 1995, PASP, 107, 966.

1000: 

1001: \bibitem[Vogt {  et~al.} 1994]{vogt:94}  

1002: Vogt, S.~S. et al. 1994, Proc. Soc. Photo-Opt. Instr. Eng., 2198, 362  

1003:   

1004: 

1005: \bibitem[Vogt {  et~al.} 2000]{vogt:00}  

1006: Vogt, S.~S., Marcy, G.~W., Butler, R.~P., Apps, K. 2000,  

1007: \newblock { ApJ, } {536}, 902  

1008: 

1009: \bibitem[Vogt {  et~al.} 2002]{vogt:02}  

1010: Vogt, S.~S.,  Butler, R.~P., Marcy, G.~W., Fischer,D.A., 

1011: Pourbaix,D., Apps, K., Laughlin,G. 2002,  

1012: \newblock {Accepted for publication, ApJ}

1013: 

1014:   

1015: \bibitem[]{}  

1016: Werthimer, D., Bowyer, S., Cobb, J., Lebofsky, M., Lampton, M. 2000, 

1017: in {\it Bioastronomy 99: A New Era in Bioastronomy},  

1018: eds.~G. Lemarchand \& K. Meech, ASP Conf.~Proc., vol.~213, 479.  

1019: \bibitem[]{}  

1020: Werthimer, D., Anderson, D., Bowyer, S., Cobb, J., Heien, E.,   

1021: Korpela, E., Lampton, M., Lebofsky, M., Marcy, G., McGarry, M., \&    

1022: Treffers, D. 2001, in {\it The Search for Extraterrestrial Intelligence  

1023: in the Optical Spectrum III}, ed.~S. Kingsley, Proc. of SPIE, vol.~4273   

1024:   

1025: \end{thebibliography}  

1026:   

1027: \clearpage  

1028:   

1029: \begin{figure}  

1030: %%\centerline{\scalebox{.75}{\rotatebox{90}{\includegraphics{fig1.ps}}}}  

1031: %\psfig{figure=fig1.ps,angle=90,width=\textwidth}

1032: %\psfig{figure=fig1.ps,angle=90}

1033: \plotone{fig1.ps}

1034: \caption{{\it Top:} A portion of a spectral order 

1035: from a raw CCD image of the spectrum.  Any emission line should

1036: span the full width of the stellar spectrum, as it is caused by

1037: seeing at the slit.

1038: {\it Bottom:} The reduction process converts a two dimensional spectrum   

1039: into a one dimensional spectrum.}  

1040: \label{reduction}  

1041: \end{figure}   

1042:   

1043: \begin{figure}  

1044: \plotone{fig2.ps}  

1045: \caption{{\it Top:} A full chunk of difference spectrum, including spikes, which  

1046: will be searched for laser line candidates.  The peak height is 15.80\% above

1047: the continuum.  {\it Bottom:} The same chunk  

1048: of difference spectrum with spikes removed and offset for clarity.  The standard deviation of  

1049: this spike-free chunk is the reported sigma.}  

1050: \label{diff}  

1051: \end{figure}  

1052:   

1053: \begin{figure}  

1054: \plotone{fig3.ps}  

1055: \caption{Emission lines in a thorium-argon spectrum showing the   

1056: spectrometer PSF.}  

1057: \label{thor}  

1058: \end{figure}  

1059:   

1060: \begin{figure}  

1061: \plotone{fig4.ps}  

1062: \caption{The observed histogram of the number of segments  

1063: versus number of hits for the star HD29528.  The dots indicate the best   

1064: fitting Poisson distribution with the mean number of hits per  

1065: segment equal to 0.00907.}  

1066: \label{hist}  

1067: \end{figure}   

1068:   

1069: \begin{figure}  

1070: %\psfig{figure=fig5.ps,width=\textwidth}

1071: %%\centerline{\scalebox{.8}{\includegraphics{fig5.ps}}}  

1072: \plotone{fig5.ps}

1073: \caption{Three hits in the reduced spectra of HD29528.  The test spectra  

1074: (dotted lines) have been aligned with the reference spectra (solid lines).  

1075: Twice the difference spectra are shown below.    

1076: The candidate laser lines are in order 83 and the  

1077: central peak is in pixel 1436 with respect to the reference spectrum.}  

1078: \label{las1}  

1079: \end{figure}  

1080:   

1081: \begin{figure}  

1082: %\centerline{\scalebox{.35}{\rotatebox{90}{\includegraphics{fig6a.ps}}}}  

1083: %\centerline{\scalebox{.35}{\rotatebox{90}{\includegraphics{fig6b.ps}}}}  

1084: %\centerline{\scalebox{.35}{\rotatebox{90}{\includegraphics{fig6c.ps}}}}  

1085: %\plotone{fig6a.ps}

1086: %\plotone{fig6b.ps}

1087: %\plotone{fig6c.ps}

1088: \psfig{figure=fig6a.ps,angle=90,width=\textwidth}

1089: \psfig{figure=fig6b.ps,angle=90,width=\textwidth}

1090: \psfig{figure=fig6c.ps,angle=90,width=\textwidth}

1091: %\centerline{\scalebox{.8}{\rotatebox{90}{\includegraphics{fig6.ps}}}}  

1092: \caption{An apparent narrow emission line. {\it Top:}  The raw 

1093: CCD image of an observation of HD29528 taken

1094: on JD 2451793.1, centered  

1095: on the first candidate laser line. {\it Middle:} The raw CCD image  

1096: of the reference spectrum.  {\it Bottom:} The difference of the two  

1097: raw spectra.  A bright line spanning the width of the spectral order   

1098: can easily be seen.}  

1099: \label{fakelas}  

1100: \end{figure}  

1101:   

1102: \begin{figure}  

1103: \plotone{fig7.ps}  

1104: \caption{{\it Top:} Order 83 of a B star spectrum.  A strong feature can  

1105: be seen at pixel 1436. {\it Bottom:} Zooming in on the bad pixels.}  

1106: \label{bstar}  

1107: \end{figure}   

1108:   

1109: \begin{figure}  

1110: %\centerline{\scalebox{.35}{\rotatebox{90}{\includegraphics{fig8a.ps}}}}  

1111: %\centerline{\scalebox{.35}{\rotatebox{90}{\includegraphics{fig8b.ps}}}}  

1112: %\plotone{fig8a.ps}

1113: %\plotone{fig8b.ps}

1114: \psfig{figure=fig8a.ps,angle=90,width=\textwidth}

1115: \psfig{figure=fig8b.ps,angle=90,width=\textwidth}

1116: %\centerline{\scalebox{.8}{\rotatebox{90}{\includegraphics{fig8.ps}}}}  

1117: \caption{A flaw in the CCD masquerades as an emission line. {\it Top:} A chunk of 

1118: raw spectrum from order 83 of the reference  

1119: spectrum for HD29528 centered on the location of the candidate laser lines.   

1120: {\it Bottom:} The same chunk of spectrum with the intensity levels adjusted.   

1121: A dark CCD flaw can be seen under the bright spectral order.}  

1122: \label{flaw}  

1123: \end{figure}  

1124:  

1125: \end{document}           

1126: