The Dirichlet Condition

Boundary conditions are not suggestions. In string theory, they are the law.

An open string is parameterised by coordinates $X^\mu(\sigma, \tau)$ where $\sigma \in [0, \pi]$ and $\tau$ is world-sheet time. The endpoints sit at $\sigma = 0$ and $\sigma = \pi$. What happens there determines everything.

The Dirichlet boundary condition is the simplest and most unforgiving:

$$\partial_\sigma X^i(\tau, 0) \neq 0 \quad\text{but}\quad X^i(\tau, 0) = y^i$$

The endpoint does not slide. It is fixed. The position $y^i$ is a constant — a point in the transverse space, a coordinate on the brane, a decision that the string's end will not move in direction $i$.

This is not a small thing. Fixing an endpoint in a direction is a way of saying that the string cannot carry momentum in that direction away from the brane. Momentum perpendicular to the brane is not just suppressed — it is zero at the boundary. The string feels a wall. It is the field-theoretic analogue of a particle in a box, but the box is built from the string's own boundary conditions, and the walls are physical D-branes.

The Neumann condition, by contrast, says $\partial_\tau X^\mu = 0$ at the endpoint — the momentum flux vanishes, and the endpoint is free to move. A D$p$-brane imposes Neumann along its $p$ spatial directions and Dirichlet along the $9-p$ transverse ones (in ten-dimensional superstring theory). The split is the brane's fingerprint.

Here is what many introductions do not make sufficiently clear: imposing Dirichlet conditions does not break reparameterisation invariance on the world-sheet. It breaks target-space Poincaré invariance — the string can still reparameterise itself freely, but the spacetime symmetry is explicitly broken by the presence of a preferred hyper-surface. This is not a defect; it is a feature. The brane is a physical object, and physical objects break symmetries.

The Dirichlet condition also has a deep connection to T-duality. If you T-dualise a direction with Neumann boundary conditions along a circle of radius $R$, you get Dirichlet along a circle of radius $\alpha'/R$. The free endpoint becomes a stuck endpoint. The D-brane appears as a consequence of duality — it was always there, hiding in the boundary conditions.

At the quantum level, the Dirichlet condition constrains the mode expansion of $X^i$. The transverse coordinates expand in cosines rather than sines (or vice versa, depending on convention), and the zero-mode — the centre-of-mass position — becomes a classical parameter $y^i$ rather than a dynamical operator. The brane's position is a background field, a modulus that labels the vacuum.

Move the brane and you move the vacuum. The endpoints follow. They are stuck to it, literally — their position is the brane's position. This is the deepest truth of the Dirichlet condition: it is not a mathematical constraint imposed on a string. It is a relationship between two objects, a commitment written into the boundary of the world-sheet. The string end and the brane surface are bound together by a condition that is at once simple and irreversible.