"Measurement of the Proton Structure Function F₂ from the 1994 HERA Data using a Neutral Network for the Identification of the Scattered Lepton"

The proton structure function F₂ is measured from deep inelastic positron-proton scattering in the kinematic range 5 ≤ Q² ≤ 40 GeV² and 1 · 10^-4 ≤ x ≤ 6 · 10^-3 using the '94 ZEUS data. This measurement extends the kinematic region so far studied to lower values of x and Q². For the event selection a neural network is used which efficiently identifies the scattered positron in the final state and reduces the contribution from background events significantly. For the training of the neural network a new approach to error function minimization is developed. When compared to the F₂ measurement using the '93 ZEUS data the errors on the F₂ are typically reduced by a factor 2. It is further shown that the value of the predicted power behavior of F₂ as a function of the x at fixed Q² from the BFKL leading-order approximation does not match the data for 5 ≤ Q² ≤ 40 GeV² when fitting a function F₂ ∼ x^-λ. The double-leading-logarithmic approximation of the DGLAP evolution equation reproduces the Q² behavior of λ within errors. The predicted Q² behavior of F₂ at fixed values of y from the BFKL and the DGLAP evolution equations are compared and shown to differ significantly in their curvature. It is demonstrated that the current errors on F₂ need to be reduced by at least an additional factor 2 at lower Q² and by a factor ∼ 4 at larger Q² to uniquely distinguish between the predicted behavior of F₂ from both evolution equations.