The printed assertion is false in the usual pointwise meaning of nowhere continuous. The original PDF explicitly has this wording. Part (c) proves that has nondecreasing càdlàg paths; such paths can have only countably many discontinuities, so they are continuous at uncountably many levels. We prove the precise valid statement:For countability, fix a path and a bounded level interval . A discontinuity is a jump , since right continuity is already known. The sum of any finite family of these jumps is at most . Thus there are only finitely many jumps of size at least , for each positive integer . Taking the countable union over and over integer proves that all discontinuities form a countable set.
For density, work on the common probability-one event of Brownian continuity, unboundedness above, and the conclusion of part (a). The inverse is strictly increasing: for every level , path continuity gives , so two distinct levels cannot have the same passage time. Suppose there were a nonempty open level interval on which had no discontinuity. Choose inside it. Its restriction to would be continuous and strictly increasing, hence onto . For every time in this positive-length interval there would be a unique with , and then . Increasing increases , so the Brownian path would be nondecreasing throughout . This contradicts part (a).
Therefore every open level interval contains a jump. This establishes Countable dense jumps of the Brownian first-passage subordinator, and in particular rules out any interval on which is continuous throughout. It does not rule out continuity at an individual level. In fact, for every fixed deterministic , part (b) gives almost surely, so is almost surely continuous at that specified level. Its dense discontinuities occur at random levels.
Articles by others on the same topic
There are currently no matching articles.