Minimal pseudo-Anosov stretch factors on nonoriented surfaces (Q1986002)

A surface homeomorphism is called pseudo-Anosov if there are two transverse singular measured foliations which are stretched by \(\lambda\) and \(1/\lambda\), respectively, for some \(\lambda>1\). A construction of \textit{R. C. Penner} [Trans. Am. Math. Soc. 310, No. 1, 179--197 (1988; Zbl 0706.57008)] gives families of products of Dehn twists, whose stretch factor can be computed. The minimal stretch factor for orientation-preserving maps is known for several (small) genera and in all cases one obtains numbers with no Galois conjugate on the unit circle. \textit{H. Shin} and \textit{B. Strenner} [Geom. Topol. 19, No. 6, 3645--3656 (2015; Zbl 1358.37080)], however, proved that a map with a power arising from Penner's construction cannot have a stretch factor Galois conjugate into the unit circle. Thus the known examples of minimal stretch factor cannot arise from Penner's construction. In fact, they are constructed by a variety of methods. For orientation-reserving maps of orientable surfaces (with the invariant foliations orientable) or for arbitrary pseudo-Anosov maps of non-orientable surfaces (with one invariant foliation orientable and the other not) this obstruction does not arise: the authors show that their stretch factor never has Galois conjugates on the unit circle. Thus it might be possible that in these cases the maps of minimal stretch factor have a power arising from Penner's construction. In fact, this is the case for the examples considered in this paper. This paper studies certain pseudo-Anosov maps that arise as a generalization of the construction for the Arnoux-Yoccoz maps which the present authors gave in [``The Arnoux-Yoccoz mapping classes via Penner's construction'', Preprint, \url{arXiv:1805.01248}, to appear in Bull. Soc. Math. Fr.]. For a number of genera (\(g=4,5,6,7,8,10,12,14,16,18,20\) in the non-orientable case, \(g=1,3,5,7,9,11\) in the orientation-reversing orientable case) the authors show by extensive computations that these examples realise the minimal stretch factor. The idea, following \textit{E. Lanneau} and \textit{J.-L. Thiffeault} [Ann. Inst. Fourier 61, No. 1, 105--144 (2011; Zbl 1237.37027)], is to run a brute-force search for integral polynomials and to show there is none whose largest root is smaller than in the constructed example. The authors state the following conjectures that are compatible with the computed cases. For nonorientable surfaces of even genus \(g=2k, k\ge 2\) the minimal stretch factor of all pseudo-Anosov maps with one orientable invariant foliation should be the largest root of \(x^{2k-1}-x^k-x^{k-1}-1\). For orientable surfaces of odd genus \(g=2k-1, k\ge 2\) the minimal stretch factor for orientation-reversing pseudo-Anosov maps with orientable foliations should be the largest root of \(x^{4k}-x^{2k+1}-x^{2k-1}-1\).