We present a new mass estimate of a well-studied gravitational lensing cluster , Abell 1689 , from deep Chandra observations with a total exposure of 200 ks .
Within r = 200 h ^ { -1 } kpc , the X-ray mass estimate is systematically lower than that of lensing by 30-50 % .
At r > 200 h ^ { -1 } kpc , the mass density profiles from X-ray and weak lensing methods give consistent results .
The most recent weak lensing work suggest a steeper profile than what is found from the X-ray analysis , while still in agreement with the mass at large radii .
Fitting the total mass profile to a Navarro-Frenk-White model , we find M _ { 200 } = ( 1.16 ^ { +0.45 } _ { -0.27 } ) \times 10 ^ { 15 } h ^ { -1 } M _ { \odot } with a concentration , c _ { 200 } = 5.3 ^ { +1.3 } _ { -1.2 } , using non-parametric mass modeling .
With parametric profile modeling we find M _ { 200 } = ( 0.94 ^ { +0.11 } _ { -0.06 } ) \times 10 ^ { 15 } h ^ { -1 } M _ { \odot } and c _ { 200 } = 6.6 ^ { +0.4 } _ { -0.4 } .
This is much lower compared to masses deduced from the combined strong and weak lensing analysis .
Previous studies have suggested that cooler small-scale structures can bias X-ray temperature measurements or that the northern part of the cluster is disturbed .
We find these scenarios unlikely to resolve the central mass discrepancy since the former requires 70-90 % of the space to be occupied by these cool structures and excluding the northern substructure does not significantly affect the total mass profiles .
A more plausible explanation is a projection effect .
Assuming that the gas temperature and density profiles have a prolate symmetry , we can bring the X-ray mass estimate into a closer agreement with that of lensing .
We also find that the previously reported high hard-band to broad-band temperature ratio in A1689 , and many other clusters observed with Chandra , may be resulting from the instrumental absorption that decreases 10-15 % of the effective area at \sim 1.75 keV .
Caution must be taken when analyzing multiple spectral components under this calibration uncertainty .