Light-pulse atom interferometry has been powerful in testing fundamental physics. Because
of its precision and sensitivity to inertial forces, the study of gravitational phenomena using an
atom interferometer with various configurations has become a new trend in this community.
This thesis presents my effort in characterising and improving the acceleration sensitivity of
an 87Rb atom interferometer that is sensitive to the horizontal acceleration of atomic motion,
which can be used to constrain chameleon dark energy force from an in-vacuum source mass.
We achieved ≈ 10^7 atoms for interferometry in 800 ms including loading 87Rb atoms into
a magneto-optical trap, vertically launching them upwards, preparing their internal state in
|F = 1,mF = 0⟩ state and velocity selection. After minimising AC Stark shifts with a novel
scheme using microwave spectroscopy, this centimetre-scale fountain configuration allowed us to
interrogate our atom interferometer with three Raman pulses separated by T = 33 ms, which
achieved an acceleration sensitivity of 30-40 μm/s^2 per shot. This acceleration sensitivity is
more than a factor of two better than 86-156 μm/s^2 per shot reported for the last generation.
We quantified the noise contribution to our acceleration sensitivity, which was found to be
dominated by phase noise. With this improved interferometer, we measured an acceleration
of −2.597 ± 0.186 μm/s^2 in 92 hours. This negative result is not compatible with a chameleon
dark energy induced fifth-force, which leads to a discussion of systematics.