Electrokinetic injection (EKI) is usually considered as one of the useful approaches to improve sensitivity of CZE analysis. In the present study, we explored the relationship between electrode position and sample amount injected during EKI process by using 2D computer simulation (CFD-ACE+) and real experiments, aiming to obtain higher detection sensitivity. Two different models of electrode configuration, a capillary inserted in a hollow electrode and a capillary surrounded by a cylindrical electrode on the reservoir wall, were simulated to evaluate the efficiency of EKI. It was found that analytes, occurring only in an effective potential field, could be introduced into the capillary while the other analytes remain outside of the field because of slow diffusion. Consequently, the longer distance between the electrode and the end of capillary, the higher efficiency of EKI was found by the simulation. This finding was verified by the real CZE analysis of dilute rare-earth metal ions in a chloride solution (pH almost neutral). In fact, when the distance of Pt electrode and the capillary end in a CE apparatus (an Otsuka CAPI-3100) was default (ca. 1 mm), LOD of Er was 0.27 mu g/L. When the distance was increased to 19.5 mm, the LOD was improved over ten times down to 0.02 mu g/L. The LOD achieved is 50-fold better than that of inductively coupled plasma atomic emission spectrometry (1-2 mu g/L for Er).