Nature 418, 955–959 (2002).

This Letter reported numerical simulations of one-dimensional disordered binary systems, and found a threshold value for the exponent characterizing the long-range power-law correlations of the system. Below this threshold, the system behaves as an insulator and above it, in the thermodynamic limit, the system behaves as a conductor. Unfortunately, we have now found that this observation was a consequence of the algorithm used to generate long-range correlations in binary chains, because above the threshold value of the exponent only a finite number of segments of atoms of the same type (A or B) exists, even in the thermodynamic limit of an infinitely large system. Thus, the system studied was not truly disordered. As a result, what we observed at the critical threshold value for the correlation exponent was not a transition from insulator to metal behaviour in a disordered system (as reported), but a transition from a disordered to an ordered system. For this reason, the authors retract the claim of a metal–insulator transition in the infinite binary chain with correlated disorder. The results are still valid that relate to the behaviour of a binary chain below the critical threshold value of the correlation exponent, and to large but finite system sizes (as found in the DNA example discussed in the Letter).

We thank L. Hufnagel and T. Geisel for drawing this to our attention.