A Search for Fast Pulsars along the Galactic PlaneD. J. Nice1National Radio Astronomy Observatory2,Edgemont Road, Charlottesville, VA 22903A. S. Fruchter1Space Telescope Science Institute,3700 San Martin Drive, Baltimore, MD 21218J. H. TaylorJoseph Henry Laboratories and Physics Department,Princeton University, Princeton, NJ 08544Accepted for publication in The Astrophysical Journal.1Some of this work was undertaken while a�liated with the Physics Department, Princeton University2Operated by Associated Universities, Inc., under contract with the National Science Foundation.

{ 2 {ABSTRACTWe present results from a survey for fast pulsars at the Arecibo Observatory. Ap-proximately 260 square degrees of sky were searched, primarily at Galactic latitudejbj < 8�. Signals were sampled at intervals of 517�s, allowing for the possibility ofdetecting pulsars with periods as short as 1ms or less. Sensitivity varied as a functionof sky position. In a typical direction, long period pulsars with ux densities of order0.9mJy at 430 MHz could be detected, although sensitivity to the fastest pulsars wassomewhat curtailed. Sixty-one pulsars were detected, twenty-four of which were previ-ously unknown. Four of the pulsars, two newly discovered, have millisecond periods.Subject headings: Stars: Pulsars | Surveys | Galaxy: Stellar Content

{ 3 {1. IntroductionOver the past quarter century, large-scale surveys speci�cally designed to detect pulsed radiosignals have proven to be a highly productive means of �nding pulsars. While most of the sky hasbeen searched for long period pulsars (Manchester et al. 1978; Stokes et al. 1986), similar searchesfor pulsars with millisecond periods have been undertaken only recently. We report here on highsensitivity observations of 260 square degrees of sky along the Galactic plane, made with equipmentcapable of detecting pulsars with rotation periods as short as 1 ms. The survey was sensitive toslow pulsars (period of order 1 second) as well as millisecond pulsars, and provided sensitivity tosuch objects comparable to or greater than that attained by previous surveys.The observations were made at the Arecibo Observatory. Portions of the region of the skyobserved have been searched for pulsars before, using the same telescope at the same radio frequency(430 MHz) as the present work. The �rst pulsar survey in this region was undertaken by Hulse andTaylor (1974, 1975), who had ux limits comparable to the present observations for long{periodpulsars. Fifty pulsars were detected in their survey, including 40 previously unknown sources, in140 square degrees of sky. Particularly notable was the discovery of the relativistic binary systemcontaining PSR B1913+16. The data sample interval of 16.7 ms used in this survey made itincapable of detecting millisecond pulsars (which were not then known to exist).Segelstein (1986; see also Stokes et al. 1986) conducted a survey in the same direction but withsu�ciently fast data acquisition equipment to detect pulsars with periods as short as 0.6 ms. Be-cause a relatively narrow passband was used, this survey was somewhat less sensitive to slow pulsarsthan the Hulse{Taylor survey. This survey covered 289 square degrees. Nineteen known sourceswere re-detected, and �ve new pulsars discovered, including the millisecond pulsar B1855+09. Per-haps the most signi�cant result of the Segelstein survey was its demonstration that surveys with ux limits of a few milliJanskys directed towards the Galactic plane �nd far fewer millisecondpulsars than slow pulsars.Several advances have allowed us to perform a much deeper search for millisecond pulsars.The telescope receivers were upgraded, resulting in system noise temperatures as low as 50{75 K.A new spectrometer allowed us to utilize the full 10 MHz passband of the telescope feed whilesampling at an interval of only 516.625 �s. Newly available supercomputer centers provided uswith the means to process the data quickly and e�ciently. We detected 61 pulsars, 24 of whichwere previously unknown. Of the 61, four are millisecond pulsars. PSRs B1937+21 and B1855+09had been previously observed. PSRs B1957+20 and J2019+2425 were new discoveries. Detailedobservations of these millisecond pulsars have been reported elsewhere (Fruchter, Stinebring, &Taylor 1988; Fruchter et al. 1990; Ryba & Taylor 1991; Nice, Taylor, & Fruchter 1993). The slowpulsars, including the slowest one yet discovered (J1951+1123, with period 5.094s), form a typicalsubset of the overall pulsar population. Extensive observations of these sources have been describedelsewhere (Camilo & Nice 1995).

{ 4 {2. Sky CoverageThe observations were made using the 305 m radio telescope at Arecibo, Puerto Rico. Itslarge collecting area makes it the instrument of choice for pulsar searches. However, its pointingis limited to zenith angles less than 20�; given its latitude, this limits observations to declinations�1� < � < +39�, and within this region there are wide variations in the antenna's forward gain. Twoportions of the Galactic plane can be observed: the Galactic anti-center (l � 180�) and the range35� < l < 80�. Our survey covered a portion of the latter area, primarily within �8� < b < +8�.Survey observations were made by tracking individual points on the sky for 67.7 s. Pointswere selected from a rectangular grid with spacing 90, approximately the half-power beam width ofthe telescope. The grid was de�ned in current (1987.0) coordinates to allow for the possibility ofswitching conveniently to constant-declination, \drift mode" observations, although this mode wasnever implemented. Grid points had coordinates right ascension �(1987:0) = 18h21m00s+ (38s)n�and declination �(1987:0) = +0�04:05 + (9:00)n� where n� and n� are integers. The separationbetween points was �xed in right ascension, so the density of beam areas on the sky was slightlyhigher at higher declinations and lower at lower declinations.Figure 1 shows the sky coverage of the survey. A total of 10,300 beam areas were successfullyobserved, processed, and found free of terrestrial interference. Taking the area of the sky observedaround each point as a 90 � 90 square, the total sky coverage was 260 square degrees. Within asingle pointing, the gain varied by up to a factor of four due to variations in the radiation patternof the telescope (M. Robison, personal communication). Median gain within a single pointing boxwas approximately 70% of peak gain. The e�ect of this gain variation, along with other factorsin uencing sensitivity, is quanti�ed below. Non-uniformities in sky coverage re ect the irregularitiesof telescope scheduling as well as the unpredictable nature of radio frequency interference.3. Data CollectionData were collected across two 10 MHz passbands, one in each sense of circular polarization,centered at 430 MHz. The 430 MHz feed at Arecibo has a peak gain of 19.7 K Jy�1, but e�ciencytapers o� towards the edges of the passband, and average gain across the band was 15 K Jy�1.The system temperatures of the two receivers were 55 K and 70 K measured on cold sky. High skytemperatures along the Galactic plane and spillover at high zenith angles increased the true systemtemperatures, typically by a factor of two or more.Due to the geometry of the telescope, gain was degraded and system temperature increasedduring observations directed away from zenith. At large zenith angles part of the feed illuminationpattern falls on the ground rather than the primary re ector. The reduction in useful collecting areacauses a corresponding reduction in gain; an increase in system temperature results from detectingground noise (� 300 K) rather than sky noise (20� 200 K).

{ 5 {The variation in gain and system temperature from one beam area to another caused variationsin the search sensitivity of a factor of four. Using the known zenith response of the telescope(National Astronomy and Ionosphere Center 1989), and extrapolating sky temperature from the408 MHz map of Haslam et al. (1982), we calculated the relative sensitivity of each beam area.Results of this calculation, along with other details of the sensitivity of the search, are given in x4.A correlation spectrometer was used to synthesize 128 narrow channels within the receiverpassbands. The signal from each polarization was �rst divided into four 2.5 MHz passbands, eachof which was three-level sampled at 10MHz, twice the Nyquist frequency. (Double Nyquist samplingpartially compensates for the loss in signal-to-noise ratio due to coarse quantization. Estimatedattenuation due to 3-level sampling was 11%; see Hagen 1985.) Each signal was auto-correlated,and thirty-two non-redundant lags were accumulated. Thus each 2.5 MHz passband yielded 32spectral channels spaced at intervals of 78.125 kHz, and the overall 10 MHz band was split into128 channels in each polarization. Lags were accumulated for � = 516:625�s, values from oppositepolarizations were summed, and the results were dumped to a computer. The raw data were toovoluminous to write directly to magnetic tape, so samples were scaled and reduced to four bitsbefore being stored for o�-line processing.Because the optimal quantization of the measurement of the (highly correlated) zero lag of theautocorrelation is di�erent from the other (less correlated) lags, a separate measure of the totalsignal power was made in each 2.5 MHz passband. These total power values were appropriatelyscaled and substituted for the zero lags of the autocorrelation functions during data analysis.4. Search Sensitivity: An OverviewPulsars are weak radio sources, and searching for them without regard to their pulsed natureis not feasible with �lled aperture telescopes such as the Arecibo instrument. Pulsars can bee�ciently recognized only by detection of the periodicities in their emission. This introduces biasesin the sensitivity to pulsars of varying rotational periods and dispersion measures. The relativesensitivity of our survey to pulsars with various parameters is most easily understood by Fourierdomain analysis.Figure 2a shows the intrinsic emission pattern of PSR J2019+2425, averaged over many pe-riods. This particular pro�le is typical of short-period pulsars; longer-period pulsars have one ormore peaks, usually close together, and total duty cycles (encompassing all peaks) of at most a fewpercent. The pulse shape as a function of time, s(t), where 0 � t < P and P is the pulse period,can be approximated by a Fourier series,s(t) = 1nh 2412S0 + nhXj=1 Sj exp�2�jtiP �35 : (1)We have employed a �nite sum over nh harmonic components because this is what is normally

{ 6 {available in a pulsar search program. This decomposition is illustrated in Figure 2b, which empha-sizes two important points. First, as the Fourier harmonic number increases, the correspondingamplitude decreases, but only slowly. Thus it is important to include, where possible, many har-monics of the signal in the pulsar search algorithm. Second, the phases of the harmonics are closeto alignment: note that each harmonic has a crest beneath the main peak. Thus the phases as wellas the amplitudes of the harmonics can be used to distinguish pulsar signals from noise.The observed amplitudes Sj of a harmonic with frequency f = j=P is reduced from its intrinsicvalues by two e�ects: di�erential pulse arrival times due to dispersion in the interstellar medium,and the use of a �nite sample integration time. We discuss these e�ects in turn. Within a narrowradio frequency range, j���0j � �0, around center frequency �0, the propagation time through theinterstellar medium of radio signals from a source with dispersion measure DM varies as (t� t0) =���1(���0) where � = 1:205�10�4 �30 DM�1 pc cm�3MHz�2 s�1. Di�erences in propagation timebetween signals at slightly di�erent frequencies within a single spectral channel cause the signal tobe smeared, reducing the amplitudes of the highest modulation frequencies. The spectral channelsof an autocorrelator have response function�(� � �0) = 1�� sinc2�� � �0�� � (2)where sinc(x) � sin(�x)=�x; (3)and �� is the spacing between channels. The detected signal from an impulse emitted by the pulsarsimultaneously at all frequencies but dispersed by the interstellar medium therefore has temporalbehavior ��(��t) within each spectral channel. The Fourier transform of this function,Td(f) = �(��1��f); (4)where �(x) � 1� jxj for jxj < 1 and zero elsewhere, is the transmission function of the dispersionprocess.A typical Galactic disk pulsar might have DM= 100 pc cm�3, so with our observing frequency�0 = 430MHz and spectral channel spacing �� = 78:125 kHz, we have ��1�� = 0:0008 s. For aslow pulsar with f = 1 s�1, then, Td(f) = 0:9992, so signals are nearly fully transmitted. However,transmission drops to Td(f) = 0:5 at f = 610 s�1, close to the spin frequency of the fastest knownpulsar.In our search system, the detected signals were boxcar integrated and recorded at �xed intervalsof � = 516:625�s. Ideally the recorded signals would be �ltered such that pulsar signals up to theNyquist frequency, fN = (2�)�1 = 967:8Hz, would be transmitted and those above this frequencywould be blocked. The �ltering performed as a side-e�ect of the boxcar accumulation is far fromideal. The Fourier transform of the boxcar function,Ts(f) = sinc(f�); (5)

{ 7 {gives the transmission of the sampling process. At the Nyquist frequency, transmission is nearly40%, and it declines only slowly with increasing frequency. Thus, the fastest pulsar signals willbe attenuated signi�cantly. On the other hand, the possibility of detecting signals with frequencyhigher than fN is left open for a su�ciently strong source. Such signals will be aliased into thespectrum at, e.g., frequency 2fN � f . Note that, while noise power is attenuated at the same rateas signal power, high-frequency noise aliased into the noise spectrum has the e�ect of attening thespectrum. Thus, as f is increased, the signal level is reduced while the noise level remains constant.Figure 2d shows the e�ect of passing the Fourier components from Figure 2b through the�lters of dispersion smearing and �nite sampling used in this search. In Figure 2c, the observedcomponents are re-assembled into the observed time-domain pro�le, showing the smeared-out pulseshape. The widening of the pulse peak translates directly into a loss of sensitivity for the survey.The amplitudes of pulse pro�le harmonics for each of the four millisecond pulsars detectedin this survey are plotted in Figure 3, along with the same amplitudes reduced by the e�ects ofdi�erential dispersion delays and �nite sample intervals. There are dramatic losses in the high-harmonic signals from the fastest-spinning pulsars (B1937+21 and B1957+20), each of which hasa period of about 1.6ms (only 3 times the sample interval) and a fairly high DM. By contrast,while there are losses for the 5.3ms pulsar B1855+09 and the 3.9ms pulsar J2019+2425, the e�ecton search sensitivity is relatively modest, both because of the longer pulse periods and because ofrelatively large duty cycles (and hence intrinsically weaker high harmonics).5. The Search AlgorithmData were recorded on magnetic tape at the observatory and were processed o� line. Eachbeam area, consisting of 131; 072 = 217 time samples in each of 128 spectral channels, was processedindependently. Using the algorithm of Taylor (1974), data from di�erent channels were delayedrelative to one another to compensate for dispersion. Time series were generated using 256 disper-sion measures evenly spaced between 0 and 127pc cm�3, 128 dispersion measures between 127 and253pc cm�3, and 128 dispersion measures between 254 and 507pc cm�3. Since dispersion smearingwas larger than 2� in the second range (and larger than 4� in the third range), we summed adjacentpairs of time samples in the second range (and sets of four adjacent samples in the third range)before processing. Thus each beam area provided 256 de-dispersed sequences of length 217, 128 oflength 216, and 128 of length 215, each corresponding to a di�erent dispersion measure.Each time series was Fourier transformed to yield spectra extending to frequencies as high asfN = 968 s�1. Recognizable terrestrial signals (in particular the 60Hz commercial power frequencyand its harmonics) were removed from the spectrum.For the highest-frequency pulsars (i.e., those with f > fN=2), only one Fourier componenthas signi�cant power in the Fourier spectrum. To search for such pulsars, the strongest singleperiodicity in each Fourier series was found by searching the amplitudes of the Fourier spectrum

{ 8 {for peaks, either single Fourier bins or split between two adjacent bins (for which amplitudes werecomputed by interpolation of complex Fourier coe�cients). Although this step was primarily aimedat detecting pulsars in the range fN=2 < f < fN , the search for Fourier peaks was not restrictedto this range; thus strong (or high-duty-cycle) pulsars with lower rotation frequencies could also befound in this step, although in general they could be detected more readily in other phases of theanalysis.The data reduction algorithm for pulsars with f < fN=2 was based on the a priori assumptionthat pulsar signals have low duty cycles and therefore many strong harmonics. As discussed above,the greatest sensitivity to these signals is attained by using both amplitude and phase information.This can be done by searching for strong, narrow peaks in time-domain pro�les generated by foldingsearch data at candidate periods. However, it is not computationally practical to do this for allperiods, so some procedure must be used to cull the best candidates for time-domain analysis. Wesummed nh =2, 3, 4, 8, and 16 harmonically related Fourier amplitudes, searching for relativelystrong signals. For each candidate period found in this way, a time-domain pro�le, with 4nh phasebins, was produced by inverting a series of up to 2nh complex Fourier coe�cients. The signal-to-noise ratios for these pro�les were calculated, based on the assumption that the signal should bepresent in at most two bins of the pro�le.Most beam areas, of course, did not contain pulsars. Due to the large number of degrees offreedom in the search, noise uctuations cause spurious \detections" with signal to noise ratio of 6in a typical beam area. Over the 10,300 beam areas, random uctuations produce numerous falsedetections up to SNR= 8:5. Any signal with a SNR above this empirical (but reasonable, giventhe quantity of data) cuto� was considered a candidate source and was re-observed to determinewhether it was a true pulsar signal or simply a statistical uctuation or an artifact of terrestrialinterference.The signal-to-noise ratio of a time domain pro�le of a pulsar with mean ux density S andduty cycle (fractional pulse width) w � 1 is given bySNR = S�w1=2 ; (6)where � = TsysG(npBt)1=2 ; (7)and G, Tsys, B, t, and np are telescope gain, system temperature, bandwidth, integration time,and number of polarizations. As discussed above, the gain and system temperature varied widelyduring this survey, so � varied as well. The median value was � = 0:43mJy.For a discrete n-point pulsar pro�le (perhaps generated by inverting nh = n=2 Fourier coe�-cients), a Fourier domain version of equation 6 isSNR = � 1��� 2nh�1=2 nhXj=1Sj (8)

{ 9 {where we have assumed the pulsar signal to be entirely con�ned to a single bin; this is equivalent torequiring the Sj in equation 1 to be real, with the pulse peak occurring at t = 0. For a very narrowpulsar signal the intrinsic Sj is S0 for each j. However, the measured amplitudes are reduced bythe dispersion and sampling �lters discussed above. The observed SNR isSNR = � 1��� 2nh�1=20@ nhXj=1Td(jf)Ts(jf)1AS0; (9)where f is the fundamental frequency of the pulsar.Extensive tests of the search code on a variety of long period pulsars showed that equation 9typically overestimates the SNR by a factor of 2. This can largely be attributed to deviations ofreal pulsar pro�les away from the model of an in�nitesimally narrow pulse peak. In particular,few pulsars have su�ciently low duty cycles (w < 0:015) to give 32 equally strong harmonics. Anadditional source of reduced SNR at long periods was a moderate increase in background noise atlow frequencies (indeed, we found it expedient to �lter out all signals of f < 2Hz, removing loworder harmonics of the slowest pulsars). While these processes a�ect only long period sources, thewide variety of millisecond pulsar pulse shapes (particularly the prevalence of interpulses) suggestthey too will tend to have somewhat less than the theoretical SNR. Therefore we have elected toinclude a factor of 2 reduction in sensitivity for all pulsars in the calculations below.The sensitivity of the search to a pulsar with very small intrinsic duty cycle in a typicaldirection (with median system noise and gain) is shown in Figure 4. The curves are based on ourlimiting SNR of 8.5, and including the empirical factor-of-2 upward adjustment. The small stepsare at points where the number of (non-aliased) harmonics changes; had we incorporated aliasedsignals into our data reduction scheme, these steps would have been smoothed out. The sensitivityvaried as a function of sky position due to di�erences in gain across the telescope beam, di�erencesin gain and spillover at di�erent zenith angles, and di�erences in the background sky temperature.Figure 5 shows the amount of sky searched with di�erent sensitivities, quanti�ed as the minimumdetectable ux density for long period pulsars.6. ResultsWe detected 61 pulsars, of which 24 were not previously known. These pulsars are listedin Tables 1 and 2. Signal-to-noise ratios quoted in the tables are those of the original searchdata. For technical reasons, our SNR values tend to be underestimates for SNR> 20 (i.e., thepulses were \clipped"). We have indicated this by showing such SNRs as lower limits in thetables. Because the relation between ux and SNR depends on the detailed pulse shape, thereis no robust scheme by which uxes can be derived from the SNR values to better than a factorof two or so. Further, ux density values from a single observation are necessarily unreliable dueto interstellar scintillation. Therefore we will not attempt to estimate uxes here, but we presentcareful measurements elsewhere (Camilo & Nice 1995).

{ 10 {The known pulsars in Table 2 were re-discovered in the search in an unbiased manner. Severalcataloged pulsars were nominally within the main beam of the telescope but were not detected (seeTable 3). This is likely because reduction of their already low ux densities by scintillation pushedthem below the search threshold. Also, it should be noted that the positions of these sources havehigh uncertainties (about the same as our beam size), and the error boxes surrounding three ofthem (B1906+09, B1911+09, and B1929+15) were not fully covered in the present search.The positions of the detected pulsars are illustrated in Figure 1. Note that the previouslyknown sources tend to be near the Galactic plane (jbj < 4�), and the newly discovered ones tendto be farther away (jbj > 4�). This is largely because the Hulse & Taylor (1975) survey, in whichthe bulk of the previously known sources were discovered, was restricted to jbj < 4�. The locationsof the pulsars projected onto the Galactic plane are given in Figure 6. It is clear that the detectedmillisecond pulsars are much closer to the Sun than the slow pulsars (mean distance of 1.1 kpc versus3.3 kpc). This can be attributed in part to our reduced sensitivity at shorter periods (Figure 4);however, millisecond pulsars with periods of 3 to 5 ms, such as B1855+09 or J2019+2425, mustalso have mean luminosities somewhat less than their slow pulsar counterparts.For a ux limited survey of a locally homogeneous population, the volume of space observed(and hence the number of objects detected) is proportional to sensitivity raised to the 3=2 power, or��3=2. (In principle this should be modi�ed for fast pulsars, for which dispersion smearing reducessomewhat the gains achieved by decreasing the limiting ux density; we will ignore such detailshere.) For each pointing, then, the (relative) volume searched can be calculated. The absolutevolume depends on type of pulsar and intrinsic luminosity, but beam-to-beam ratios of volumesremain �xed. Figures 7a and 7b show the average (relative) volume searched per beam as a functionof Galactic latitude and longitude. The average volume increases with decreasing jbj because, forreasons related to telescope scheduling, observations at large jbj tended to have large zenith anglesand therefore low gain and large spillover. The small cusp at jbj < 2� is due to the dramatic increasein sky temperature along the Galactic plane.The total volumes searched as a function of Galactic latitude and longitude are given in Fig-ures 7c and 7d. The densities of pulsars detected (number of pulsars divided by volume) are givenin Figures 7e and 7f. The relatively uniform distribution of density with Galactic latitude can beattributed to the concentration of our survey towards low latitudes, below the height at which thepulsar population tapers o�. The distinct increase in slow pulsar density with decreasing Galacticlongitude re ects the concentration of these pulsars inwards of the solar circle, a common feature ofpulsar population models (e.g., Lyne, Manchester, & Taylor 1985). Because the millisecond pulsarsdetected in this survey span a much smaller region of the Galaxy, a similar large-scale patternwould be impossible to detect in our data.

{ 11 {7. SummaryWe have presented results of a survey sensitive to both slow and fast pulsars. Sensitivity variedsigni�cantly due to variations in sky temperature, telescope system temperature, and telescope gain.Long period pulsars with ux densities of 0.9 mJy or more at 430 MHz could be detected in a typicalpointing. Sensitivity to pulsars at short periods was reduced, with attenuation by a factor of � 10at the period of the fastest known pulsar (1.56 ms). Still, strong pulsars with sub-millisecondperiods could have been detected.Fifty-seven slow and four millisecond pulsars were detected. The slow pulsars show someevidence of concentration towards the Galactic center. The distances to the millisecond pulsars areon average only one third the distances to the slow pulsars due to a combination of lower sensitivityto fast pulsars and lower intrinsic luminosities of fast pulsars.The Arecibo Observatory is part of the National Astronomy and Ionosphere Center, operatedby Cornell University under contract with the National Science Foundation. The survey data wereanalyzed at the John von Neumann Supercomputer Center (R.I.P.) and the Pittsburgh Super-computer Center. This work was supported by the NSF and the U.S. Department of Education(through a fellowship to D.J.N.).

{ 12 {REFERENCESCamilo, F. & Nice, D. J. 1995, ApJ. In pressFruchter, A. S. et al. 1990, ApJ, 351, 642Fruchter, A. S., Stinebring, D. R., & Taylor, J. H. 1988, Nature, 333, 237Hagen, J. 1985, NAIC Manual No. 8319Haslam, C. G. T., Salter, C. J., Sto�el, H., & Wilson, W. E. 1982, A&AS, 47, 1Hulse, R. A. & Taylor, J. H. 1974, ApJ, 191, L59Hulse, R. A. & Taylor, J. H. 1975, ApJ, 201, L55Lyne, A. G., Manchester, R. N., & Taylor, J. H. 1985, MNRAS, 213, 613Manchester, R. N., Lyne, A. G., Taylor, J. H., Durdin, J. M., Large, M. I., & Little, A. G. 1978,MNRAS, 185, 409National Astronomy and Ionosphere Center 1989, Arecibo Observatory User's ManualNice, D. J., Taylor, J. H., & Fruchter, A. S. 1993, ApJ, 402, L49Ryba, M. F. 1991. PhD thesis, Princeton UniversityRyba, M. F. & Taylor, J. H. 1991, ApJ, 380, 557Segelstein, D. J. 1986. PhD thesis, Princeton UniversityStokes, G. H., Segelstein, D. J., Taylor, J. H., & Dewey, R. J. 1986, ApJ, 311, 694Taylor, J. H. 1974, A&AS, 15, 367Taylor, J. H. & Cordes, J. M. 1993, ApJ, 411, 674Taylor, J. H., Manchester, R. N., & Lyne, A. G. 1993, ApJS, 88, 529This preprint was prepared with the AAS LATEX macros v3.0.

{ 13 {Table 1. New pulsars discovered in the survey.Pulsar l b Period DM SNR(deg) (deg) (ms) (pc cm�3)J1848+0823 40.5 5.5 329. 90 10.3J1859+1526 47.6 5.1 934. 95 16.1J1901+1306 45.8 3.7 1831. 75 14.6J1902+0723 40.8 1.0 488. 105 8.7J1903+2225 54.4 7.5 651. 110 10.9J1906+1854 51.5 5.2 1019. 155 15.0J1908+0500 39.2 �1.4 291. 200 � � �J1908+0734 41.6 �0.2 212. 10 10.4J1909+1450 48.1 2.8 996. 120 12.2J1909+1859 51.9 4.8 542. 65 >20.0J1910+0714 41.6 �0.8 2712. 125 16.8J1911+1758 51.2 3.8 460. 50 24.7J1912+2525 57.8 7.1 622. 40 12.2J1915+0738 42.5 �1.7 1543. 40 13.7J1915+1541 49.9 1.4 371. 10 17.4J1941+1026 48.0 �6.2 905. 140 >23.9J1946+2611 62.3 0.6 435. 165 9.1J1951+1123 50.0 �7.7 5094. 30 12.3J1953+1149 50.7 �8.1 852. 140 17.5B1957+20 59.2 �4.7 1.61 29 14.0J2002+1637 55.9 �7.5 276. 95 15.9J2008+2513 64.1 �4.0 589. 60 >23.2J2015+2524 65.1 �5.2 2303. 10 14.2J2019+2425 64.7 �6.6 3.93 50 9.2

{ 14 {Table 2. Previously known pulsars detected in the survey.Pulsar l b Period DM SNR Fluxa(deg) (deg) (ms) (pc cm�3) (mJy)B1839+09 40.1 6.3 381. 49 >42.3 20B1842+14 45.6 8.1 375. 41 9.5 19B1848+13 45.0 6.3 346. 59 18.3 7B1852+10 42.9 4.3 573. 250 13.6 11B1855+09 42.3 3.1 5.36 13 9.3 31B1901+10 43.3 1.8 1857. 140 9.0 2B1907+10 44.8 1.0 284. 148 >34.3 55B1907+12 46.2 1.6 1442. 260 12.5 5B1910+20 54.1 5.0 2233. 88 >31.1 5B1911+13 47.9 1.6 521. 144 12.2 6B1913+10 44.7 �0.6 404. 240 11.9 6B1913+167 50.6 2.5 1616. 66 >23.5 5B1914+09 44.6 �1.0 270. 61 >29.9 25B1914+13 47.6 0.5 282. 237 >22.1 13B1915+13 48.3 0.6 194. 95 >86.9 45B1916+14 49.1 0.9 1180. 30 18.0 6B1918+19 53.9 2.7 821. 154 >22.4 30B1919+21 55.8 3.5 1337. 12 >38.9 200B1920+20 54.4 2.6 1172. 203 10.8 5B1920+21 55.3 2.9 1078. 217 >32.5 30B1921+17 51.7 0.9 547. 135 >20.3 2B1924+14 49.9 �1.0 1325. 205 10.7 9B1924+16 51.9 0.1 580. 177 10.6 9B1925+18 53.8 1.0 483. 250 8.9 3B1925+188 53.8 0.9 298. 90 8.2 3B1925+22 57.0 2.7 1431. 180 19.3 6B1929+10 47.4 �3.9 227. 3 >57.0 250B1929+20 55.6 0.6 286. 210 >27.3 17B1930+13 49.4 �3.1 928. 165 12.4 2B1930+22 57.4 1.6 144. 211 >22.3 7B1931+24 59.2 2.4 814. 89 >47.2 8B1933+16 52.4 �2.1 359. 159 >51.3 200B1935+25 60.8 2.3 201. 62 >49.3 6B1937+21 57.5 �0.3 1.56 71 >88.3 240B1937+24 60.3 1.4 645. 100 >63.6 3B1939+17 54.5 �2.7 696. 175 18.7 3B1944+17 55.3 �3.5 440. 16 >37.5 31aFrom the catalog of Taylor, Manchester & Lyne (1993).

{ 15 {Table 3. Previously known pulsars not detected.Pulsar l b Period DM Fluxa(deg) (deg) (ms) (pc cm�3) (mJy)B1848+12 44.5 5.9 1205. 71 4B1904+12 46.1 2.4 827. 260 2B1906+09 43.2 0.4 830. 250 5B1910+10 44.8 0.1 409. 140 2B1911+09 44.0 �0.6 1242. 155 3B1913+16 50.0 2.1 59. 167 4B1927+13 49.1 �2.3 760. 207 5B1929+15 51.4 �1.6 314. 120 3B1933+17 53.7 �1.3 654. 210 3aFrom the catalog of Taylor, Manchester & Lyne (1993).

{ 16 {

Fig. 1.| Observations. Each dot represents one pointing of the telescope. Circles represent newlydiscovered sources, triangles represent known sources that were detected.

{ 17 {

Fig. 2.| (a) Intrinsic pulse shape of PSR J2019+2425. (b) Fourier decomposition of intrinsic shape,showing the �rst ten terms. (c) Observed pulse shape. (d) Fourier decomposition of observed shape.

{ 18 {

Fig. 3.| Amplitude vs. harmonic number of Fourier decompositions of pro�les of the four mil-lisecond pulsars detected in this survey. The upper line in each plot shows the intrinsic Fourieramplitudes (observed with coherent de-dispersion circuitry); the lower line shows the amplitudesas �ltered by the search system. Intrinsic pulse pro�les for all but PSR J2019+2425 are from Ryba(1991).

{ 19 {

Fig. 4.| Minimum detectable pulsar ux density for a pulsar with very low duty cycle (< 1:5%for long period pulsars, with a somewhat relaxed criterion for shorter-period pulsars). Curves areshown for dispersion measures 0, 10, 20, 40, 80, 160, and 320pc cm�3. Curves were calculatedassuming the median background noise of the survey.

{ 20 {

Fig. 5.| Amount of sky surveyed as a function of relative sensitivity, quanti�ed as the minimumdetectable ux density for long period pulsars. Total area is 260 square degrees.

{ 21 {

Fig. 6.| Positions of detected pulsars projected onto the Galactic Plane. Circles are superimposedat locations of millisecond pulsars. Distances are from the model of Taylor and Cordes (1993).

{ 22 {

Fig. 7.| Survey statistics analyzed by Galactic latitude and Galactic longitude. These �guresdisplay the relative depth of this survey as a function of Galactic coordinates l and b. Absolutecalibration of these histograms depends on the luminosity, period, and duty cycle of the pulsar(s)under consideration. (a,b) Relative volume searched in each telescope pointing at a given latitudeor longitude (c,d) Total volume searched (on a relative scale) at a given latitude or longitude (e,f)Density of slow (solid line) and millisecond (broken line) pulsars detected at a given latitude orlongitude.