HIGH-PRECISION TIMING OF MILLISECOND PULSARS .3. LONG-TERM MONITORINGOF PSRS B1885+09 AND B1937+21

Biweekly timing observations of PSRs B1855 + 09 and B1937 + 21 have be en made at the Arecibo Observatory for more than 7 and 8 yr, respectiv ely, with uniform procedures and only a few modest gaps. On each obser ving date we measure an equivalent pulse arrival time for PSR B1855 09 at 1.4 GHz, with typical accuracies of about 0.8 mus, and for PSR B 1937 + 21 at both 1.4 and 2.4 GHz, with accuracies around 0.2 mus. The pulse arrival times are fitted to a simple model for each pulsar, yie lding high-precision astrometric, rotational, and orbital parameters, and a diverse range of conclusions. The celestial coordinates and prop er motions of the two pulsars are determined with uncertainties less-t han-or-equal-to 0.12 mas and less-than-or-equal-to 0.06 mas yr-1 in th e reference frame of the DE200 planetary ephemeris. The annual paralla xes are found to be pi = 1.1 +/- 0.3 mas and pi < 0.28 mas for PSRs B1 855 + 09 and B1937 + 21, respectively. The general relativistic Shapir o delay is measured in the PSR B1855 + 09 system and used to obtain ma sses m1 = 1.50(+0.26/-0.14) M. and m2 = 0.258(+0.028/-0.016) M. for th e pulsar and its orbiting companion. The extremely stable orbital peri od of this system provides a phenomenological limit on the secular cha nge of Newton's gravitational constant, G/G = (-9 +/- 18) x 10(-12) yr -1. Variations in the dispersion measure of PSR B1937 + 21 indicate th at the spectrum of electron-density fluctuations in the interstellar m edium has a power-law index beta = 3.874 +/- 0.01 1, slightly steeper than the Kolmogorov value of 11/3, and we find no strong evidence for an ''inner scale'' greater than about 2 x 10(9) cm. 2 x 10(9) cm. In t he residual pulse arrival times for PSR B1937 + 21 we have observed sm all systematic trends not explained by our deterministic timing model. We discuss a number of possible causes; although the results are not yet conclusive, the most straightforward interpretation is that the un modeled noise (a few microseconds over 8 yr, or DELTAt/T almost-equal- to 10(-14)) is inherent to the pulsar itself. In the present data set, PSR B1855 + 09 exhibits no discernible timing noise. With conventiona l assumptions we derive a limit OMEGA(g)h2 < 6 x 10(-8) (95% confidenc e) for the energy density, per logarithmic frequency interval, in a co smic background of stochastic gravitational waves. We discuss the feas ibility of establishing a pulsar-based timescale that might be used to test the stabilities of the best available atomic clocks. In an Appen dix, we propose guidelines for the archiving of pulsar timing observat ions. Instructions are provided for obtaining copies of our own archiv al data, via Internet.