1: \begin{abstract}

2: Using numerical results from density matrix renormalization group

3: (DMRG) calculations for the $t$-$J$ model, on systems as large as

4: $10\times7$, we examine the structure of the one and two hole ground

5: states in ladder systems and in two dimensional clusters.  A simple

6: theoretical framework is used to explain why holes bind in pairs in

7: two-dimensional antiferromagnets.

8: For the case $J/t=0.5$, which we have studied, the hole pairs reside

9: predominantly on a $2\times2$ core plaquette with the probability

10: that the holes are on diagonal sites greater than nearest-neighbor

11: sites.  There is a strong singlet bond connecting the spins on the two

12: remaining sites of the plaquette.

13: We find that a general characteristic of dynamic holes in an

14: antiferromagnet is the presence of frustrating antiferromagnetic bonds

15: connecting next-nearest-neighbor sites across the holes. Pairs of

16: holes bind in order to share the frustrating bonds.

17: 

18: At low doping, in addition to hole pairs, there are two additional

19: low-energy structures which 

20: spontaneously form on certain finite systems. The

21: first is an undoped $L\times2$ spin-liquid region, or ladder. The

22: second is a hole moving along a one dimensional chain of sites. 

23: At higher doping we expect that hole pairing is always favored.

24: \end{abstract}

25: