We present an analysis of a Chandra ACIS-S observation of the elliptical galaxy NGC 720 to verify the existence of a dark matter halo and to measure its ellipticity .
The ACIS-S3 image reveals over 60 point sources distributed throughout the field , most of which were undetected and therefore unaccounted for in previous X-ray studies .
For semi-major axes a \lesssim 150 \arcsec ( 18.2 h _ { 70 } ^ { -1 } kpc ) the ellipticity of the diffuse X-ray emission is consistent with a constant value , \hbox { { $ \epsilon _ { x } $ } } \approx 0.15 , which is systematically less than the values 0.2-0.3 obtained from previous ROSAT PSPC and HRI observations because of the unresolved point sources contaminating the ROSAT values .
The Chandra data confirm the magnitude of the \sim 20 \arcdeg position angle ( PA ) twist discovered by ROSAT over this region .
However , the twist in the Chandra data is more gradual and occurs at smaller a also because of the point sources contaminating the ROSAT values .
For a \gtrsim 150 \arcsec out to a = 185 \arcsec ( 22.4 h _ { 70 } ^ { -1 } kpc ) , which is near the edge of the S3 CCD , \epsilon _ { x } and PA diverge from their values at smaller a .
Possible origins of this behavior at the largest a are discussed .

Overall the ellipticities and PA twist for a \lesssim 150 \arcsec can be explained by the triaxial mass model of NGC 720 published by Romanowsky & Kochanek ( which could not produce the abrupt PA twist in the ROSAT HRI data ) .
Since the optical image displays no substantial isophote twisting , the X-ray PA twist requires a massive dark matter halo if the hot gas is in hydrostatic equilibrium .
Furthermore , the values of \epsilon _ { x } obtained by Chandra are too large to be explained if the gravitating mass follows the optical light ( M \propto \hbox { { $L _ { \star } $ } } ) irrespective of the PA twist : The M \propto \hbox { { $L _ { \star } $ } } hypothesis is inconsistent with the Chandra ellipticities at the 96 % confidence level assuming oblate symmetry and at the 98 % confidence level for prolate symmetry .
Thus , both the PA twist and the ellipticities of the Chandra image imply the existence of dark matter independent of the temperature profile of the gas .
This geometric evidence for dark matter can not be explained by alternative gravity theories such as the Modification of Newtonian Dynamics ( MOND ) .

To constrain the ellipticity of the dark matter halo we considered both oblate and prolate spheroidal mass models to bracket the full range of ( projected ) ellipticities of a triaxial ellipsoid .
The dark matter density model , \rho \propto ( a _ { s } ^ { 2 } + a ^ { 2 } ) ^ { -1 } , provides the best fit to the data and gives ellipticities and 1 \sigma errors of \epsilon = 0.37 \pm 0.03 for oblate and \epsilon = 0.36 \pm 0.02 for prolate models .
Navarro-Frenk-White ( NFW ) and Hernquist models give similar ellipticities for the dark matter .
These moderate ellipticities for the dark halo are inconsistent with both the nearly spherical halos predicted if the dark matter is self-interacting and with the highly flattened halos predicted if the dark matter is cold molecular gas .
These ellipticities may also be too large to be explained by warm dark matter , but are consistent with galaxy-sized halos formed in the currently popular \Lambda CDM paradigm .