Undoped GaAs/AlGaAs heterostructures have been used to fabricate quantum wires in which the average impurity separation is greater than the device size. We compare the behavior of the Zero-Bias Anomaly against predictions from Kondo and spin polarization models. Both theories display shortcomings, the most dramatic of which are the linear electron-density dependence of the Zero-Bias Anomaly spin-splitting at fixed magnetic field B and the suppression of the Zeeman effect at pinch-off.PACS numbers: 72.10. Fk, 72.25.Dc, 73.21.Hb, 73.23.Ad Split gates [1] can be used to restrict transport from a two-dimensional electron gas (2DEG) to a ballistic onedimensional (1D) channel. This results in the quantization of the differential conductance G = dI/dV sd in units of G 0 = 2e 2 /h at zero magnetic field [2,3]. A shoulder on the riser of the first quantized plateau, the "0.7 anomaly" or "0.7 structure" [4], is not completely understood but generally acknowledged to result from electronelectron interactions. Although spin polarization models [5,6,7,8,9,10] and 1D Kondo physics models [11,12,13] can describe many experiments, neither can explain all phenomena associated with the 0.7 structure. One example is the so-called zero-bias anomaly (ZBA): a peak in G centered at V sd = 0 for G < 2e 2 /h when sweeping sourcedrain bias V sd at a fixed gate voltage V gate at low temperature T . Spin polarization models cannot alone predict its occurrence in quantum wires, although an embedded impurity near or in the 1D channel could produce a ZBA via the 0D Kondo effect [14,15,16,17]. On the other hand, in 1D Kondo physics models, a bound state forms when G < G 0 . In this context, a resonance observed by a non-invasive detector capacitively coupled to a quantum wire at threshold [18] as well as a triple-peaked structure in G at fixed V gate below the 0.7 structure [19] are consistent with the presence of a localized state in 1D channels.Systematically studying the ZBA in modulation-doped 2DEGs has proven difficult because of the large variability of its characteristics from device to device [20,21], probably due to the randomly fluctuating background potential caused by the ionized dopants, significant even with the use of large (≥75 nm) spacer layers. This disorder is so pervasive that one can be led to wonder whether the ZBA always results from interactions between conduction electrons and a random localized state near the 1D channel. However, disorder can be dramatically reduced in undoped GaAs/AlGaAs heterostructures where an external electric field (via a voltage V top on a metal top gate) electrostatically induces the 2DEG [22,23]. Figure 1(a) shows the advantages of this technique, particularly at low carrier densities (see also Fig. 3 in Ref. [22]), a regime most relevant for the ZBA. In this Letter, we report on the study of the ZBA in ten quantum wires fabricated in undoped GaAs/AlGaAs heterostructures. We demonstrate that an unsplit ZBA does not result from interactions between conduction electrons and a random localiz...
