As a continuation of the work reported by Thejll et al. (1995), we have measured the JH and K band fluxes from 72 more stars - mainly hot sdB and sdO stars, along with a few White Dwarfs. In 28 new sd cases (i.e., from June and October, 1994 - see Table 7) we have found a significant excess that we interpret to be mainly due to stellar companions. Of these, 14 are new binary discoveries: BD+25 4655, Feige 108, HD4539, HD149382, HD216135, KPD2109+440, LSI+63 198, LSIV+10 9, LSV+22 38, PG0011+221, PG0116+242, PG0314+103, PG2151+100 and TON139. BD+25 4655, HD4539 and HD216135 are however the weakest cases as their evidence for an IR excess becomes marginal when compared to other clearer cases. Two more discoveries are BD+37 1977 and BD+48 1777 which, from Paper I, are now found to display excesses with our present analysis. It should be kept in mind, however, that both are sdO cases now fitted with H-rich Kurucz models which, together with the previous marginal evidence for an excess pointed out in Paper I, render this conclusion weak as well. BD-13 842 and BD+33 2642 display also IR flux escesses but, being classified as CSPN, have not been analysed. The suspected binaries PB8555 (Kilkenny et al. 1988) and SB7 (Heber 1986) are confirmed. Our two analysis methods (extraction of 2 $\sigma$ simultaneously significant JHK flux excesses and fitting of two Kurucz atmospheric models) can jointly find agreeing spectral type companions for 21 out of 28 "good data'' (see Table 8) cases. 13 companions could be of G type (see Table 7).

The 28 analysed new cases, added to the 11 sds found by Thejll et al. (1995), represent 44% out of the 88 hot sds observed in total during 1994 (i.e., =24 from February+ 64 from June and October). Of the 41 (i.e., =28+ 11+ BD+37 1977+ BD+48 1777), fulfilling a 2 $\sigma$ excess condition, 15 are sdO, 15 are sdB, three are sdOBs, PG0110+262 is an sdB-O and the rest (i.e., 7) remain unclassified (i.e., sd).

For 28 stars (see Table 8) we were furthermore able to satisfactorily fit the sum of two spectral Kurucz models so that temperature estimates and an estimate of the relative radii could be obtained. Assuming that all the companions are ZAMS stars we can then calculate upper limits on log(g) for the hot subdwarfs. For 18 of the hot subdwarfs previous estimates of log(g) were available from spectroscopic analysis and our upper limits are in agreement with most cases except for a few discrepant ones. We discuss HD149382 as an exaple which we find inexplicable, as the excess flux is small and thus probably did not influence the several spectroscopic analysis that have been performed - we speculate that the source of the excess radiation is unrelated to the star itself, but cannot evaluate this suggestion without better imaging of possible faint sources in the sky around the bright main star. Other discrepant cases may be due to the presence of Giant or Subgiant, instead of ZAMS, companions.

Given the recent discovery of pulsators among hot subdwarfs with binary companions (O'Donoghue et al. 1997 and references therein) our sample provides suitable candidates to pursue that line of investigation. We are currently investigating, both theoretically and observationally, the presence of eventual pulsations in Feige 108 among others.

$\begin{figure} \centering \includegraphics[width=8cm,clip]{newfig4.ps}\end{figure}$

Figure 4: Derived upper limits for log(g $\theta^{4}$ ) of the sample sdB and sdO stars. Stars with known giant companions have been excluded

6.1 Gravities and evolutionary implications

We had hoped to find several telling cases of sdOs with companions so that we could "test'' NLTE modelling techniques, but we have found no strongly contradictory cases within the present analysis.

As our method only provides upper limits on log(g) for the hot star, the narrowness of the log(g) distributions can be interpreted to imply that the companion-types are all well described by our assumption of ZAMS status - real giant or sub-giant companions force our method to exaggerate the upper limits on log(g) but the absence of a substantial number of cases like that shows the predominance of un-evolved companions. This is what one would expect on the basis of probability considerations as most stars spend most of their time in the hydrogen burning phase.

The coincidence of the main peaks in the log(g) distributions - to the extent that we are able to "resolve'' it with the limited number of stars we have - may say something about the underlying mass-distributions. Saffer et al. $\,$ (1994) found that the distribution of log( $g\theta^{4}$ ) (where $\theta$ = 5040/ $T_{\rm eff}$ ) for the sdB stars they studied was a basically unresolved peak near 2.6 and a FWHM of 0.4 (i.e. standard deviation near 0.2). This corresponded to a single Gaussian mass distribution with a narrow width of 1 $\sigma\sim$ 0.04 $M_{\hbox{$\odot$}}$ centered on 0.5 $M_{\hbox{$\odot$}}$ . As $g\theta^{4}$ scales with luminosity and is constant for a given core mass (Greenstein & Sargent 1974) it is consistent to interpret the distribution found by Saffer et al. $\,$ as evidence for core He burning. In our sample of stars we find (after excluding the known cases of giant companions) that the distributions of log( $g\theta^{4}$ ) have mean and standard deviations of (2.9; 0.8) and (2.3; 0.7) for the sdBs and sdOs, respectively. The distributions are shown in Fig. 4.

These values are not significantly different from each other, nor from the more narrowly distributed value that Saffer et al. $\,$ found for their sample of sdBs. Analysis of the mass-luminosity ratio for helium-core burning objects (Thejll & MacDonald, unpublished models), shows that near 0.5 $M_{\hbox{$\odot$}}$ a doubling or halving of the mass corresponds to an order of magnitude decrease or increase in the M/L ratio, respectively. As the present distribution of log( $g\theta^{4}$ )for the sdOs contains most of the cases within one order of magnitude on either side of the mean we see that the distribution corresponds to a spread in masses for the sdOs between 0.25 and 1 $M_{\hbox{$\odot$}}$ , if they are assumed to be helium core burners. These are not unrealistic limits given what we know about the mass distribution of sdOs from suitable binary systems (Ritter 1990). The present data on the sdOs is therefore consistent with an interpretation of the sdOs as being He-core burners (i.e. Helium Main Sequence stars - HEMS) with a natural spread of masses.

Future improvements in the amount and quality of data available for sdOs will be needed to make a statement about whether a narrower distribution of masses - such as in a scenario where sdBs evolve into sdOs, and therefore keep their narrowly distributed mass-spectrum - is the case.

Peter Thejll gratefully acknowledges support from Nordita and the Danish Natural Research Council. Ana Ulla acknowledges support from the Spanish DGICYT for a contract under the MEC's programme "Incorporación de Doctores y Tecnólogos a grupos de Investigación Científica y Enseñanza Superior" associated to CICYT's research project ISOPHOT-S Fase Final (contract ESP94-0034), the EU (contract ERBCHBGCT930407) and Nordita for visiting grants in May and October, 1996, to work on this paper. Staff at the Carlos Sánchez Telescope are thanked for their skilled assistance during observations. C.S.Hansen is also thanked for obtaining for us some of the October, 1994, observations. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France and of the NASA's Astrophysics Data System (ADS). Use has also been made of the RDAF facility at the GSFC at Greenbelt Maryland and of the ULDA at the LAEFF in Madrid (Spain). Henrik Svensmark is thanked and admired for his insight into IDL font magics. Ana Ulla is an external scientific collaborator of the Laboratorio de Astrofísica Espacial y Física Fundamental in Madrid (Spain).

6 Summary and discussion

6.1 Gravities and evolutionary implications