Open XXZ chain and boundary modes at zero temperature

We thank the referees for their careful reading of the manuscript and their interesting comments. We have indeed implemented the two requested changes suggested in report 2. Concerning the many insightful comments of report 1, they deserve some more explanations.

• The sentence «  the boundary magnetization becomes a discontinuous function of the other field at h− = h+, point at which the boundary root jumps from one edge of the chain to the other » was supposed to be an image describing the significant change in the mathematical description of the ground state (defined as the lowest energy state in finite volume) when crossing this point: the boundary root changes localization in its mathematical description as a solution of the Bethe equation (from being exponentially close to a zero associated to the boundary parameter at one end of the chain, it becomes exponentially close to the zero associated to the boundary parameter at the other end of the chain, as it becomes clear from our study of the ground state in section 4, and this also corresponds to a change of localization of the corresponding contribution to the wave function), and this change of localization is in fact responsible for the discontinuity that we observe here. This is not really due to a crossing of level since the degeneration is not exact at h_-=h_+ for L even (so there is no crossing of level for L finite in this case, see Figure 3 that we have slightly modified to make this point more clear). This change of localization of the boundary root with respect to the two zeroes of the boundary factor corresponds of course to a rearrangement of all the Bethe roots, but the latter is continuous when crossing this point (since the two zeroes coincide at this point). Nevertheless, since it seems that this sentence can be misunderstood, we have slightly modified it, as well as a few sentences in the paragraph before, so as to make it more precise.

• "It is difficult to imagine that any measurement would detect the discontinuity investigated by the authors ». We are not expert in experimental measurements and, of course, since the discontinuity that we computed for the boundary magnetization may be smoothened by the temperature, it may not be easy to measure it experimentally (and we never pretended it should be). However, it seems to us that, in order to cool down the system to one of its ground state, it is sufficient to choose one of the magnetic fields at the edges to be larger (or smaller) than the other and then cool down the chain. This way the system is gapped, there is no degeneracy and the ground state 1 (or 2) has a finite energy difference with the other. We agree that this may be difficult to do in practice when h_- ~ h_+ (the temperature should be kept smaller than |h_- - h_+|) but in principle the protocol is clear.

• Concerning the discussion about finite temperature, we hope it is clear that we do not consider the temperature case in the present paper. This is indeed an interesting problem that we mention as a possible further development in the conclusion. In particular, our study of the boundary magnetization in the ground state is performed strictly at zero temperature. Therefore, we do not pretend to correct anything about references [50,51] (which are now [53,54] in the new version): indeed, these references should in fact not appear in the footnote on page 28, and we thank the referee for pointing out this misprint, we have corrected this. We mentioned these references only by completeness since the same quantity was computed there in the temperature case. In particular, in [51] (now [53]), the zero-temperature limit of the boundary magnetization has been recovered (in the case of a zero magnetic field at the other end).

• Concerning the question of whether the thermodynamic and the zero-temperature limits commute or not, it is difficult to answer this question without having investigated the temperature case. Nevertheless, the authors of [51] (now [53]) have investigated the zero-temperature limit of the expression they have obtained for the boundary magnetization at finite-temperature by taking the thermodynamic limit first, and they have recovered the result (including the discontinuity) that had been previously obtained at zero-temperature for the case of a zero-boundary magnetic field at infinity. This seems to indicate that, at least in the case considered in [51], these limits do commute.

• « What happens here when we change only one of the fields, and how does it relate to the size of the system, and the order of limits ? » We thank the referee for this very good question. What we have found is that the form factor of the sigma_1^z operator between the two ground states decays exponentially in system size whenever the two fields are chosen to be different. Therefore in order to have a finite t->infty limit for the auto-correlation one should take first h_- -> h+ and then the thermodynamic limit. We have commented this in the manuscript, and added the computation of the form factor for different fields.

• We focused on the even length case because it is the only case where there is an interesting quasi (and not exact) degeneracy at h_-=h_+. In the odd length case, there is no degeneracy at h_-=h_+ unless both are zero. There is however an exact degeneracy for h_-= - h_+, which is also present for finite L and can be seen as a consequence of the spin flip symmetry of the chain, and the two degenerate ground Bethe states are in this case in different magnetization sectors. It is obvious that in this case the discontinuity of the boundary magnetization is due to a crossing of levels of these two ground Bethe states with different magnetization, and therefore this case seemed to us a priori not so interesting from the physical point of view. It is nevertheless interesting to compare what happens in the even and odd length cases, since the appearance of the discontinuity is due to very different microscopic mechanisms. Hence, as suggested by the referee, we have added a section devoted to the odd L case, in which we redo the computation in this case and explain the differences with respect to the even L case. We have also commented this case in introduction and in conclusion.

• Finally, we thank the referee for pointing out the misprint in formula (1.8). Indeed in the bulk case, the spontaneous staggered magnetisation can be observed either by applying a small staggered field or a small field on one of the sites (or at the boundary). We have nevertheless preferred to remove the mention to this quantity, since finally we do not find it so significant for our study. We have preferred to add instead a small comment about the odd L case that we now also study in our paper.