A property of two-parameter martingales with path-independent variation (Q1822134)

The paper is devoted to some delicate and specific properties of two- parameter martingales. The authors consider continuous, vanishing on the axes, two-parameter martingales with path-independent variation (p.i.v.), i.e. their quadratic variations along all increasing paths from the origin and with the same end point have the same values. The main result of the paper states that if \(M_ t\) is a p.i.v. martingale from \(L^ 4\) then it is orthogonal to the martingale \[ \tilde M_ t=\lim \sum_{u}M(\Delta^ 1_ u)M(\Delta^ 2_ u) \] (M\({}_ t\) can be formally regarded as the stochastic integral \(\int \partial_ 1M\partial_ 2M)\). Orthogonality means that \(<M, \tilde M>=0.\) The authors show that the converse is not true. They construct martingales without p.i.v. and satisfying \(<M,\tilde M>=0\). The construction of such martingales is based on the existence of a non- trivial solution for the stochastic differential equation \(XdX+YdY=0.\) If \(M_ t\neq 0\) on the axes the condition p.i.v. does not imply \(<M,\tilde M>=0\), as follows from some example.