We have developed a wide-field mosaic CCD camera, MOA-cam3, mounted at the prime focus of the Microlensing Observations in Astrophysics (MOA) 1.8-m telescope. The camera consists of ten E2V CCD4482 chips, each having 2kx4k pixels, and covers a 2.2 deg^2 field of view with a single exposure. The optical system is well optimized to realize uniform image quality over this wide field. The chips are constantly cooled by a cryocooler at -80C, at which temperature dark current noise is negligible for a typical 1-3 minute exposure. The CCD output charge is converted to a 16-bit digital signal by the GenIII system (Astronomical Research Cameras Inc.) and readout is within 25 seconds. Readout noise of 2--3 ADU (rms) is also negligible. We prepared a wide-band red filter for an effective microlensing survey and also Bessell V, I filters for standard astronomical studies. Microlensing studies have entered into a new era, which requires more statistics, and more rapid alerts to catch exotic light curves. Our new system is a powerful tool to realize both these requirements.

We report a measurement of the shape of the source star in microlensing event MOA 2002-BLG-33. The lens for this event was a close binary whose centre-of-mass passed almost directly in front of the source star. At this time, the source star was closely bounded on all sides by a caustic of the lens. This allowed the oblateness of the source star to be constrained. We found that a/b = 1.02^{+0.04}_{-0.02} where a and b are its semi-major and semi-minor axes respectively. The angular resolution of this measurement is approximately 0.04 microarcsec. We also report HST images of the event that confirm a previous identification of the source star as an F8-G2 turn-off main-sequence star.

We present a 4 year light curve of a cataclysmic variable in M22, based on an analysis of accumulated data from the MOA microlensing survey. The position of the star coincides with that of a transient event observed by HST in 1999, originally attributed to microlensing but later suspected to be a dwarf nova outburst. Two outburst episodes, one in 2002 and one in 2003, with $\Delta I\sim3$ are seen in the MOA data thus confirming that the HST event was a dwarf nova outburst. The MOA and HST data show that this dwarf nova underwent at least three outburst episodes during 1999--2004. Further close monitoring of this event is encouraged as future outburst episodes are expected.

Observations of the gravitational microlensing event MOA 2003-BLG-32/OGLE 2003-BLG-219 are presented for which the peak magnification was over 500, the highest yet reported. Continuous observations around the peak enabled a sensitive search for planets orbiting the lens star. No planets were detected. Planets 1.3 times heavier than Earth were excluded from more than 50 % of the projected annular region from approximately 2.3 to 3.6 astronomical units surrounding the lens star, Uranus-mass planets from 0.9 to 8.7 astronomical units, and planets 1.3 times heavier than Saturn from 0.2 to 60 astronomical units. These are the largest regions of sensitivity yet achieved in searches for extrasolar planets orbiting any star.

We present observations of the unusual microlensing event OGLE 2003-BLG-235/MOA 2003-BLG-53. In this event a short duration (~7 days) low amplitude deviation in the light curve due a single lens profile was observed in both the MOA and OGLE survey observations. We find that the observed features of the light curve can only be reproduced using a binary microlensing model with an extreme (planetary) mass ratio of 0.0039 +/- (11, 07) for the lensing system. If the lens system comprises a main sequence primary, we infer that the secondary is a planet of about 1.5 Jupiter masses with an orbital radius of ~3 AU.

We analyze the Galactic bulge microlensing event MOA-2003-BLG-37. Although the Einstein timescale is relatively short, t_e=43 days, the lightcurve displays deviations consistent with parallax effects due to the Earth's accelerated motion. We show that the chi^2 surface has four distinct local minima that are induced by the ``jerk-parallax'' degeneracy, with pairs of solutions having projected Einstein radii, \tilde r_e = 1.76 AU and 1.28 AU, respectively. This is the second event displaying such a degeneracy and the first toward the Galactic bulge. For both events, the jerk-parallax formalism accurately describes the offsets between the different solutions, giving hope that when extra solutions exist in future events, they can easily be found. However, the morphologies of the chi^2 surfaces for the two events are quite different, implying that much remains to be understood about this degeneracy.

We report a measurement of limb darkening of a solar-like star in the very high magnification microlensing event MOA 2002-BLG-33. A 15 hour deviation from the light curve profile expected for a single lens was monitored intensively in V and I passbands by five telescopes spanning the globe. Our modelling of the light curve showed the lens to be a close binary system whose centre-of-mass passed almost directly in front of the source star. The source star was identified as an F8-G2 main sequence turn-off star. The measured stellar profiles agree with current stellar atmosphere theory to within ~4% in two passbands. The effective angular resolution of the measurements is <1 micro-arcsec. These are the first limb darkening measurements obtained by microlensing for a Solar-like star.