In this work we have investigated the relevant trend of the bolometric correction ( BC ) at the cool-temperature regime of red giant stars and its possible dependence on stellar metallicity .
Our analysis relies on a wide sample of optical-infrared spectroscopic observations , along the 3500 Å \Rightarrow 2.5 \mu m wavelength range , for a grid of 92 red giant stars in five ( 3 globular + 2 open ) Galactic clusters , along the full metallicity range covered by the bulk of the stars , -2.2 \leq [ Fe / H ] \leq + 0.4 .

Synthetic BVR _ { c } I _ { c } JHK photometry from the derived spectral energy distributions allowed us to obtain robust temperature ( T _ { eff } ) estimates for each star , within \pm 100 K or less .
According to the appropriate temperature estimate , black-body extrapolation of the observed spectral energy distribution ( SED ) allowed us to assess the unsampled flux beyond the wavelength limits of our survey .
For the bulk of our red giants , this fraction amounted to 15 % of the total bolometric luminosity , a figure that raises up to 30 % for the coolest targets ( T _ { eff } \lesssim 3500 K ) .
Allover , we trust to infer stellar M _ { bol } values with an internal accuracy of a few percent .
Even neglecting any correction for lost luminosity etc .
we would be overestimating M _ { bol } by \la 0.3 mag , in the worst cases .
Making use of our new database , we provide a set of fitting functions for the V and K BC vs. T _ { eff } and vs . ( B - V ) and ( V - K ) broad-band colors , valid over the interval 3300 ~ { } { K } \leq { T } _ { eff } \leq 5000 K , especially suited for Red Giants .

The analysis of the BC _ { V } and BC _ { K } estimates along the wide range of metallicity spanned by our stellar sample show no evident drift with [ Fe/H ] .
Things may be different for the B-band correction , where the blanketing effects are more and more severe .
A drift of \Delta ( B - V ) vs. [ Fe/H ] is in fact clearly evident from our data , with metal-poor stars displaying a “ bluer ” ( B - V ) with respect to the metal-rich sample , for fixed T _ { eff } .

Our empirical bolometric corrections are in good overall agreement with most of the existing theoretical and observational determinations , supporting the conclusion that ( a ) BC _ { K } from the most recent studies are reliable within \la \pm 0.1 over the whole color/temperature range considered in this paper , and ( b ) the same conclusion apply to BC _ { V } only for stars warmer than \simeq 3800 K. At cooler temperatures the agreement is less general , and MARCS models are the only ones providing a satisfactory match to observations , in particular in the BC _ { V } vs . ( B - V ) plane .