1: \begin{abstract} 

2: 

3:     We propose counting peaks in weak lensing (WL) maps, as a function

4:     of their height, to probe models of dark energy and to constrain

5:     cosmological parameters. Because peaks can be identified in

6:     two--dimensional WL maps directly, they can provide constraints

7:     that are free from potential selection effects and biases involved

8:     in identifying and determining the masses of galaxy clusters. We

9:     have run cosmological N--body simulations to produce WL

10:     convergence maps in three models with different constant values of

11:     the dark energy equation of state parameter, $w=-0.8, -1$, and

12:     $-1.2$, with a fixed normalization of the primordial power

13:     spectrum (corresponding to present--day normalizations of

14:     $\sigma_8=0.742,0.798$, and $0.839$, respectively).  By comparing

15:     the number of WL peaks in 8 convergence bins in the range of $-0.1

16:     < \kappa < 0.4$, in multiple realizations of a single simulated

17:     $3\times3$ degree field, we show that the first (last) pair of

18:     models can be distinguished at the 95\% (85\%) confidence level.

19:     A survey with depth and area (20,000 degree$^2$), comparable to

20:     those expected from LSST, should have a factor of $\approx 50$

21:     better parameter sensitivity.  We find that relatively

22:     low--amplitude peaks ($\kappa\sim 0.03$), which typically do not

23:     correspond to a single collapsed halo along the line of sight,

24:     account for most of this sensitivity.  We study a range of

25:     smoothing scales and source galaxy redshifts ($z_s$). With a fixed

26:     source galaxy density of 15 arcmin$^{-2}$, the best results are

27:     provided by the smallest scale we can reliably simulate, 1

28:     arcminute, and $z_s=2$ provides substantially better sensitivity

29:     than $z_s\leq 1.5$.

30: 

31: \end{abstract}}