Arnold Scholz: Between Mathematics and Politics Franz Lemmermeyer Abstract This is an English translation (with a few modifications) of the introduction to Scholz’s life and work taken from the edition [ 52 ] of the correspondence between Helmut Hasse, Arnold Scholz and Olga Taussky. I hope to find the time one day to include more information on Scholz’s mathematical results. In this article we want to present, apart from a brief biography, 1 1 1 N. Korrodi has kindly helped me to fill some gaps. an introduction to Arnold Scholz’s number-theoretic work and at the same time to convey an impression of the difficulties he faced from the National Socialists’ seizure of power until his death. Arnold Scholz was born on 24 December 1904 in the Charlottenburg district of Berlin. His father, Reinhold Scholz (1857–1933), was employed as a physicist and mathematician at the Military Research Institute in Berlin; his mother was Johanna Scholz, née Diesfeld (1874–1957). In 1923 Scholz passed the Abitur at the Kaiserin-Augusta-Gymnasium in Charlottenburg and then attended lectures in philosophy, mathematics and musicology at the University of Berlin. Scholz spent the summer of 1927 abroad in Vienna studying with Philipp Furtwängler; he received his Ph.D. in 1928 magna cum laude under Issai Schur and moved to Freiburg in April 1929. Arnold Scholz quickly became one of the foremost experts on Takagi’s class field theory; what others had to prove, he simply “saw”. This is one reason why his publications are sometimes very difficult to read. According to Taussky’s obituary of Scholz, Schur is said to have remarked about one of Scholz’s early papers: “More Landau, less Goethe!” A quite different difficulty is described by Scholz himself in his letter from 19 December 1935: The final formulation is so difficult for me with a new subject because I usually do not think in words, but in spatial images, furnished with colours and all sorts of other associations. Hasse, in an expert opinion on Scholz, criticised him for using too little modern algebra: On the other hand, this strongly developed sense for numbers is also a certain weakness, for it often prevents him from fully employing the meanwhile developed more formal methods of modern algebra, which in recent years have so decisively influenced number theory. In some sense this may well be true, even though the readability of Scholz’s articles suffers primarily from the fact that he does not condense recurring patterns of arguments into lemmas and propositions, but rather hides them in the text at their first appearance and then employs them a second time without any comment: too much Goethe, too little Landau. Concerning modern algebra, the following remark by Jehne from [ 41 , S. 221] seems to me more to the point: Without any cohomology theory at hand Scholz recognized that the number knot \nu_{K|k} is an epimorphic image of the fundamental group \pi{\mathfrak{g}} of the Galois group in a natural way. It is quite simply the case that the language of exact sequences and cohomology theory – so helpful in questions of embedding problems and norm residues – had not yet been developed, and Scholz was therefore unable to express his results with the desired clarity using the language available to him at the time. This also explains why most of his results had to be rediscovered first before one could find them in Scholz’s papers. One of the few who studied Scholz’s works thoroughly was Shafarevich: by extending Scholz’s methods for constructing number fields with given Galois group, he was able to show that every solvable finite group occurs as a Galois group over {\mathbb{Q}} (Scholz had proved this for p -groups and for certain two-step groups, i.e. those whose commutator subgroup is abelian). Other results of Scholz that were rediscovered much later include: • The theory of genus and central class fields, initiated by Chebotarev and extended by Scholz, was developed a second time by Fröhlich [ 20 ] . • The “Scholz reciprocity law”, already proved by Schönemann, was rediscovered by Emma Lehmer (see [ 49 ] ) and only then located in Scholz’s work; it triggered a flood of investigations into power residue characters of units. By contrast, Scholz’s generalization of this law to primes \ell>2 has so far played no role in the literature. • Scholz’s theory of knots was modernized by Jehne [ 41 ] (after Tate had rediscovered the main theorem in cohomological form) and subsequently further developed by many authors. • Parts of the work of Scholz & Taussky [ S17 ] were presented in modern form by Heider & Schmithals [ 34 ] . • The theory of abelian crossings, developed by Scholz in connection with his knots, was further elaborated by Heider [ 35 ] (cf. also Steinke [ 97 , § 2] ) using modern tools. Similar investigations are still pending for other works of Scholz; even as regards the theory of knots, Jehne’s paper represents only a first step towards a proper “translation” of Scholz’s work: since Jehne we know what Scholz proved, but his proofs have remained obscure to this day. In the following we wish to go through the most important stations in the life and work of Arnold Scholz in chronological order. 1927 Together with his former teacher Friedrich Neiß and his fellow student Klaus Müller, Scholz studied Hasse’s class field report and, in his first letter to Hasse dated 22 April 1927, made suggestions for simplifying several proofs, which ultimately led to their joint paper [ S1 ] . Scholz’s essential contributions are the following: He realised that, starting from Takagi’s definition of class fields, one can show directly and quickly • that class fields are necessarily Galois and possess a Galois group all of whose subgroups are normal; • that class fields possess simple functorial properties: if L/K/k is a tower of field extensions and L/k is a class field, then so are L/K and K/k . The proofs are of an analytic nature and make essential use of the pole of the Dedekind zeta function at s=1 . With a little group theory, one easily obtains from these statements a proof that class fields are abelian extensions. Some of Scholz’s ideas from 1927 were only published later in the paper [ S10 ] . On 23 April Scholz travelled to Vienna and remained there until September in order to study with Furtwängler the topic of his dissertation, namely the construction of field extensions with prescribed solvable Galois group using class field theory. Later others also began to take an interest in this topic, in particular Richter [ 89 , 90 ] , Reichardt [ 85 ] , and Tannaka [ 100 ] . Definitive results were obtained much later by Shafarevich [ 96 ] , who built upon Scholz’s ideas. In order to explain the basic idea of Scholz, we shall carry out his construction for a few very simple examples. Octic dihedral extensions Let G=D_{4}=\langle S,T:S^{4}=T^{2}=1,\ TST=S^{-1}\rangle be the dihedral group of order 8 . The construction of extensions K/{\mathbb{Q}} with {\operatorname{Gal}}(K/{\mathbb{Q}})\simeq D_{4} is of course no problem that would require class field theory for its solution; our aim here is to illustrate what the Scholz construction looks like in such a simple example. Scholz constructs such two-step extensions in two abelian steps; in the present case there are three possibilities: a) construct K as a suitable quadratic extension K/k of the biquadratic extension k={\mathbb{Q}}(\sqrt{m},\sqrt{n}\,) ; b) construct K as a suitable biquadratic extension of a quadratic subfield; c) construct K as a cyclic extension K/k of degree four over a quadratic subfield k={\mathbb{Q}}(\sqrt{m}\,) . We shall first choose the third variant here; in order to be able to describe the Galois group of the extension K/{\mathbb{Q}} precisely at each step, we wish to realise the generating automorphisms in each step as inertia automorphisms of suitable prime ideals, and for this purpose it is helpful if exactly one prime ideal ramifies in each step. In the present case we choose a quadratic extension k/{\mathbb{Q}} in which exactly one prime p ramifies. By the Kronecker–Weber theorem these fields are the quadratic subfields of the field of p -th roots of unity; in order that the quadratic subfield does not ramify at the infinite places (which would introduce an additional difficulty for continuing the construction), we must choose p\equiv 1\pmod{4} (alternatively: the unit -1 in the base field should be a quadratic residue modulo p ; we shall encounter such conditions quite frequently). Thus k={\mathbb{Q}}(\sqrt{p}\,) is a real quadratic subfield whose Galois group is generated by the Frobenius automorphism T of the prime ideal (\sqrt{p}\,) . Now we turn to the construction of the cyclic extension K/k . In order that T retains order 2 also in K/{\mathbb{Q}} , we shall choose K in such a way that the prime p does not ramify further in K/k ; then the inertia group of p in K/{\mathbb{Q}} will still have order 2 and be generated by an element \widetilde{T} of order 2 whose restriction to k is T . In what follows we shall, like Scholz, not distinguish between \widetilde{T} and T . The construction of a cyclic extension K/k with the aid of class field theory could certainly also be carried out directly; in any case, it is not sufficient to lift cyclic extensions from {\mathbb{Q}} to k , since the Galois group of such an extension would necessarily be abelian. In order to ensure that the cyclic extension K/k does not come from a cyclic extension F/{\mathbb{Q}} , Scholz proceeds via the construction of an extension L/{\mathbb{Q}} whose Galois group is the wreath product (at Scholz this construction is called, as we shall see below, a “disposition group”) \displaystyle\Gamma \displaystyle=C_{4}\,\wr\,C_{2}=\langle S_{1},S_{2},T:S_{1}^{4}=S_{2}^{4}=T^{2}=1, \displaystyle\qquad S_{1}S_{2}=S_{2}S_{1},\ TS_{1}T=S_{2},\ TS_{2}T=S_{1}\rangle (1) here and below we denote by C_{m} the cyclic group of order m . In this case, the construction is also carried out in such a way that exactly one prime ideal ramifies in each of the cyclic base extensions. If one introduces in \Gamma the additional relation S_{1}=S_{2} (or S_{1}=S_{2}^{-1} ), i.e. forms the quotient group \Gamma/\langle S_{1}S_{2}^{-1}\rangle (or \Gamma/\langle S_{1}S_{2}\rangle ), then the fixed field K of the normal subgroup \langle S_{1}S_{2}^{-1}\rangle (or \langle S_{1}S_{2}\rangle ) is a cyclic extension of k in which two prime ideals ramify; in the first case we have {\operatorname{Gal}}(K/{\mathbb{Q}})\simeq C_{2}\times C_{4} , in the second {\operatorname{Gal}}(K/{\mathbb{Q}})\simeq D_{4} (the Hasse diagram in Fig. 1 does not contain all subfields or subgroups). {\mathbb{Q}} k K_{1} K \widetilde{K} K_{2} L \Gamma \langle S_{1},S_{2}\rangle S_{1} S_{1}S_{2} S_{1}S_{2}^{-1} S_{2} 1 Figure 1: Extension with wreath product C_{4}\,\wr\,C_{2} Scholz therefore wishes to construct a cyclic extension K_{1}/k that is conjugate over k to a cyclic extension K_{2}/k with K_{1}\cap K_{2}=k . This situation arises automatically if K_{1}/k is a cyclic extension whose conductor {\mathfrak{f}}_{1}={\mathfrak{q}} is a prime ideal of degree 1; by class field theory, the corresponding class field K_{2} to the conductor {\mathfrak{f}}_{2}={\mathfrak{q}}^{T} will then also be cyclic of degree 4 and disjoint from K_{1} over k . To construct the class field K_{1}/k , we therefore choose a prime number q that splits in k/{\mathbb{Q}} : q\mathcal{O}_{k}={\mathfrak{q}}_{1}{\mathfrak{q}}_{2} . The ray class number for an extension with conductor {\mathfrak{f}}_{1}={\mathfrak{q}}_{1} is h_{k}({\mathfrak{q}}_{1})=h_{k}\cdot\frac{\Phi({\mathfrak{q}}_{1})}{(E:E^{(1)})}, where h_{k} is the class number of k , \Phi({\mathfrak{q}}_{1})=q-1 is the number of residue classes in k coprime to {\mathfrak{q}}_{1} , E=\langle-1,\varepsilon_{p}\rangle is the unit group of k , and E^{(1)} denotes the subgroup of all units \varepsilon\equiv 1\pmod{{\mathfrak{q}}_{1}} . The minimum requirement on q is therefore that this ray class number is divisible by 4 . For this to hold, q-1 must be a multiple of 4 , and in order for E^{(1)} to have index divisible by 4 in E , both -1 and the fundamental unit \varepsilon_{p} must be quadratic residues modulo {\mathfrak{q}}_{1} . If we choose q\equiv 1\pmod{4} , then q-1 is a multiple of 4 and -1 is a quadratic residue modulo {\mathfrak{q}}_{1} ; the question of when \varepsilon_{p} is a quadratic residue modulo {\mathfrak{q}}_{1} was later studied extensively by Scholz in [ S15 ] . The existence of such primes q can easily be proved: we need only choose q so that this prime splits completely in the extension {\mathbb{Q}}(\sqrt{p},\sqrt{-1},\sqrt{\varepsilon_{p}}\,) , which is always possible by classical density theorems (Kronecker, Frobenius, later Chebotarev). By this choice of q we have now obtained a cyclic extension K_{1}/k of degree 4 in which at most the prime ideal above {\mathfrak{q}}_{1} ramifies. In the present case this prime ideal must ramify, because k possesses no quadratic unramified extensions. Let K_{1} therefore be the cyclic subextension of degree 4 in the ray class field of k modulo {\mathfrak{q}}_{1} , and let S_{1} be the inertia automorphism of {\mathfrak{q}}_{1} ; then S_{1}^{4}=1 , and by Artin the extension K_{2}/k conjugate over k to K_{1}/k is also cyclic, its Galois group is generated by the inertia automorphism S_{2} of {\mathfrak{q}}_{2} , and the compositum L=K_{1}K_{2} has Galois group as in ( 1 ). The desired dihedral extension is obtained by setting S_{1}=S_{2}^{-1} in ( 1 ), i.e. by considering the quotient group of \Gamma by \langle S_{1}S_{2}\rangle : this normal subgroup then fixes a dihedral extension K/{\mathbb{Q}} that is cyclic over the quadratic subfield k={\mathbb{Q}}(\sqrt{p}\,) ; in K/k precisely the two prime ideals above q will ramify. We now also see that constructing a dihedral extension by the second method would have been much simpler, since the dihedral group D_{4} is isomorphic to the wreath product C_{2}\,\wr\,C_{2} , and we would therefore only need to construct an extension of a quadratic number field as a subfield of the ray class field modulo a primary prime ideal. The ramification problem In our above construction of a dihedral extension we were, at one point, somewhat fortunate in that the base field k={\mathbb{Q}}(\sqrt{p}\,) has odd class number. In general we cannot assume that our base field satisfies such arithmetic conditions. This fact leads to problems that we wish to illustrate again with a very simple example. Let k be an imaginary quadratic number field different from {\mathbb{Q}}(\sqrt{-1}\,) and {\mathbb{Q}}(\sqrt{-3}\,) , and let p>2 be a prime that splits in k : p\mathcal{O}_{k}={\mathfrak{p}}_{1}{\mathfrak{p}}_{2} . The ray class number of k modulo {\mathfrak{p}} is then h_{k}\{{\mathfrak{p}}\}=h_{k}\cdot\frac{p-1}{2}. The class number h_{k} is even if and only if there exist unramified quadratic extensions of k . For the existence of a quadratic extension ramified exactly at {\mathfrak{p}} , however, it is necessary that p\equiv 1\pmod{4} . This condition is usually also sufficient: for the prime ideal {\mathfrak{p}}=(3+2\sqrt{-5}\,) of norm 29 in the quadratic number field k={\mathbb{Q}}(\sqrt{-5}\,) with class number h_{k}=2 , we have h_{k}\{{\mathfrak{p}}\}=28\equiv 0\pmod{4} , but the degree- 4 subfield of the ray class field modulo {\mathfrak{p}} consists of the unramified extension K=k(i) , the Hilbert class field, and a quadratic extension L=K\big(\sqrt{(1+2i)(3+2\sqrt{-5}\,)}\big) adjoined to it, in which {\mathfrak{p}} ramifies. There is, however, no quadratic extension of k that is ramified exactly at {\mathfrak{p}} ; the “usual suspects” such as k(\sqrt{\pm 3+2\sqrt{-5}}\,) are in fact also ramified over the prime ideal above 2 . If, therefore, in a given base field k one has found a prime ideal {\mathfrak{p}} for which the ray class number h_{k}({\mathfrak{p}}) is divisible by \ell , then there exists a cyclic extension K/k of degree \ell in which at most {\mathfrak{p}} ramifies; however, this extension may also be unramified, which leads to insurmountable problems when continuing the construction, because one can say little about the corresponding automorphisms, as they are tied to the behaviour of the associated ideal classes in extensions. If one chooses {\mathfrak{p}} in such a way that h_{k}({\mathfrak{p}})/h_{k} is divisible by \ell , then there does indeed exist an extension F/H of the Hilbert class field H of k that is ramified exactly over {\mathfrak{p}} , but in general this extension cannot be pulled back down, except when the ray class group is the direct product of the ideal class group and a group of order divisible by \ell . Together with Furtwängler, Scholz discovers how one can achieve this decomposition of the ray class group by a clever choice of the prime ideal {\mathfrak{p}} . To this end, let K be a number field containing a primitive \ell -th root of unity \zeta (what follows also holds more generally for roots of unity of prime power degree); a prime ideal {\mathfrak{p}} is called \ell -primary if it is coprime to \ell and if (\omega/{\mathfrak{p}})_{\ell}=1 for all singular \omega\in K\setminus(1-\zeta) , i.e., all elements \omega coprime to \ell that are \ell -th ideal powers: (\omega)=\mathfrak{a}^{\ell} . In fields with class number coprime to \ell , {\mathfrak{p}} is \ell -primary if and only if every unit is an \ell -th power residue modulo {\mathfrak{p}} . The Frobenius density theorem guarantees the existence of such prime ideals. Scholz observes that he can also use the notion of primary prime ideal in number fields that do not contain the corresponding roots of unity: he calls a prime ideal {\mathfrak{p}} in K primary modulo \ell^{h} if N{\mathfrak{p}}\equiv 1\pmod{\ell^{h}} and all elements of K that are \ell^{h} -th ideal powers are also \ell^{h} -th power residues modulo {\mathfrak{p}} . In the case k={\mathbb{Q}}(\sqrt{-5}\,) , the prime ideal {\mathfrak{p}} above 29 is not 2 -primary, because the singular number \omega=2+\sqrt{-5} is a quadratic non-residue modulo {\mathfrak{p}} . By contrast, the prime ideal {\mathfrak{q}}=(-3+4\sqrt{-5}\,) of norm 89 is 2 -primary; since N{\mathfrak{q}}\equiv 1\pmod{4} , we have (\frac{-1}{{\mathfrak{q}}})=+1 , while 4\sqrt{-5}\equiv 3\pmod{{\mathfrak{q}}} implies 2+\sqrt{-5}\equiv\frac{1}{4}(8+4\sqrt{-5})\equiv\frac{11}{4}\equiv 25\pmod{{\mathfrak{q}}} and thus certainly (\frac{2+\sqrt{-5}}{{\mathfrak{q}}})=+1 . Indeed, k(\sqrt{-3+4\sqrt{-5}}\,) is a quadratic extension in which precisely the prime ideal {\mathfrak{q}} ramifies. For a prime ideal {\mathfrak{p}} that is primary modulo \ell^{h} , consider the ideal group \mathcal{H} consisting of all ideals \mathfrak{a} in k coprime to {\mathfrak{p}} such that some m -th power with \ell\nmid m equals a principal ideal generated by an \ell^{h} -th power residue \alpha : \mathfrak{a}^{m}=(\alpha) with (\alpha/{\mathfrak{p}})_{\ell^{h}}=1 . This last condition does not depend on the choice of \alpha , because every unit is an \ell^{h} -th power residue modulo {\mathfrak{p}} . If D_{k} denotes the group of all ideals coprime to {\mathfrak{p}} , then {\operatorname{Cl}}_{\ell}\{{\mathfrak{p}}\}=D_{k}/\mathcal{H} is a finite \ell -group that has the \ell -class group as a quotient group. More precisely, Scholz shows that when {\mathfrak{p}} is chosen to be primary, we always have {\operatorname{Cl}}_{\ell}\{{\mathfrak{p}}\}\simeq{\operatorname{Cl}}_{\ell}(k)\times A ; the class field corresponding to the quotient group D_{k}/{\operatorname{Cl}}_{\ell}(k) is what Scholz calls a residue field , i.e., a purely ramified cyclic extension of degree \ell^{h} in which exactly the prime ideal {\mathfrak{p}} (and only this one) is purely ramified. Figure 2: Arnold Scholz Disposition groups If G is a solvable group, then extensions with Galois group G can be constructed step by step in abelian steps. Scholz recognised that it suffices, for the construction of these abelian extensions, to consider two types of groups: abelian p -groups and “disposition groups”. In [ S2 ] Scholz treats “two-step disposition groups” G ; these are groups possessing the following two properties: 1. G contains an abelian subgroup A such that G/A is abelian; 2. A contains an element a such that the elements a^{\sigma} form a basis of A as \sigma runs through a system of representatives of G/A . In modern terminology, a two-step disposition group is therefore simply the wreath product C_{m}\,\wr\,N , where C_{m}\simeq\langle a\rangle and N=G/A . The simplest examples of wreath products are the dihedral group D_{4}\simeq C_{2}\,\wr\,C_{2} and the alternating group A_{4}=C_{2}\,\wr\,C_{3} . For two-step disposition groups Scholz can realise the step-by-step construction with exactly one ramified prime ideal in each step. Non-abelian extensions of degree pq Let p and q\equiv 1\pmod{p} be two primes. Then there exists a non-abelian group of order pq , namely \Gamma=\langle S,T:S^{p}=T^{q}=1,\ S^{-1}TS=T^{r}\rangle, where r^{p}\equiv 1\pmod{q} and r\not\equiv 1\pmod{q} . Following our preliminary considerations in the case of the dihedral group, the procedure should now be clear: we construct a cyclic extension k/{\mathbb{Q}} of prime degree p as a subextension of the field of \ell -th roots of unity, where \ell\equiv 1\pmod{p} is prime. Then we choose a q -primary prime ideal {\mathfrak{q}} in k and construct a cyclic extension of prime degree q with conductor {\mathfrak{q}} ; the normal closure of this extension then has Galois group C_{q}\,\wr\,C_{p} . The desired extension with Galois group \Gamma is finally obtained by a clever formation of quotient groups. Scholz shows in [ S2 ] how one can also construct the extension with Galois group \Gamma without the detour via the wreath product. To this end, he introduces (in [ S2 , p. 353, bottom] ) the notion of (S-1) -th power residues, which plays a quite central role in a large number of his papers but has received hardly any attention to this day. To do so, Scholz considers a number field K containing a primitive \ell -th root of unity \zeta and calls \alpha\in\mathcal{O}_{K} an (\ell,S-a) -th power residue modulo a prime ideal {\mathfrak{p}} (here we tacitly assume that {\mathfrak{p}} does not divide \ell and that \alpha is coprime to {\mathfrak{p}}\ell ) if \alpha\equiv\beta^{S-a}\gamma^{\ell}\pmod{{\mathfrak{p}}} holds. More generally, for any element \Lambda\in{\mathbb{Z}}[G] , where {\mathbb{Z}}[G] denotes the group ring of the Galois group of K/{\mathbb{Q}} , one can call an \alpha\in\mathcal{O}_{K} an (\ell,\Lambda) -th power residue modulo {\mathfrak{p}} if \alpha\equiv\beta^{\Lambda}\gamma^{\ell}\pmod{{\mathfrak{p}}} holds. In his papers Scholz applies this notion mainly to \Lambda=(S-1)^{2} . In the present case, the aim is to adjoin to a cyclic extension of degree p an extension L/K with Galois group \Gamma . To this end, Scholz observes that the symbolic order {\mathfrak{M}}={\operatorname{Ann}}(A) , i.e. the annihilator of A in the group ring {\mathbb{Z}}[G] , is in the present case equal to {\mathfrak{M}}=(q,S-r) and that this ideal in turn divides (q,S^{p}-1) . As before, Scholz chooses a q -primary prime ideal {\mathfrak{q}} , but he does not construct an extension whose residue field is the full disposition group C_{q}\,\wr\,C_{p} corresponding to an ideal group whose principal ideals consist of the q -th power residues modulo {\mathfrak{q}} ; instead he constructs the extension directly by including all (q,S-a) -th powers in the ideal group. Over the years a group-theoretic language has emerged that allows us to state Scholz’s main result from the paper [ S2 ] briefly and concisely as follows: given a field extension K/k with Galois group G , and given the embedding problem in which one seeks an extension L/K such that {\operatorname{Gal}}(L/k)=\Gamma and {\operatorname{Gal}}(L/K)=N , then this embedding problem is always solvable whenever the exact sequence \begin{CD}1@>{}>{}>N@>{}>{}>\Gamma @>{}>{}>G@>{}>{}>1\end{CD} splits. Of the many smaller theorems scattered throughout the text of [ S2 ] , we mention only one: if K/k is normal with Galois group G and L/K is a class field corresponding to the ideal group H , then K/k is normal if and only if H remains fixed under the action of all \sigma\in G . After his return to Berlin, Scholz turns his attention to cubic number fields with given discriminant and the related search for quadratic number fields whose 3 -class group has rank \geq 2 . These investigations were later subsumed into the joint paper with Olga Taussky on the capitulation problem. 1928 In February Scholz completes his dissertation on the construction of extensions K/K_{0} of a given number field K_{0} whose Galois group is a “disposition group”. He now turns to the group-theoretic question for which other groups one must solve the inverse Galois problem in order to bring all solvable finite groups under control. On 1 April Scholz takes up his assistant position with Erhard Schmidt in Berlin. In August Scholz begins investigating the unit group and ideal class group of number fields {\mathbb{Q}}(\sqrt{p},\sqrt{q}\,) for primes p\equiv q\equiv 1\pmod{4} , which will ultimately lead him to questions concerning the solvability of the non-Pell equation t^{2}-pqu^{2}=-4 , to the “Scholz reciprocity law”, and also to his counterexample to the conjecture that the local-global principle for norms holds in non-cyclic extensions of number fields. The corresponding investigations of composita of cyclic fields of prime degree \ell lead him to the discovery that there exist class field towers of arbitrarily great length. 1929 In 1929, Scholz receives his PhD in Berlin under Schur; as his topic, he had chosen the construction of extensions of {\mathbb{Q}} with prescribed Galois group (see [ S2 ] ). The “inverse Galois problem” remains a central theme for Scholz, and his articles move back and forth between class field theory on the one hand and group theory on the other e.g. in [ S3 , S4 ] . In [ S3 ] Scholz shows that the construction of number fields with arbitrary two-step Galois group can be reduced to the construction of extensions whose Galois group is a disposition group or a branch group. Branch groups are here a family of, in a certain sense, maximal p -groups that contain all p -groups as quotient groups. Branch groups A two-step branch group of type (\ell^{h_{1}},\ldots,\ell^{h_{m}};\ell^{k}) is defined by Scholz as a metabelian \ell -group G with commutator subgroup A and the following properties: 1. G/A has a basis S_{1}A , …, S_{m}A . 2. \operatorname{ord}\,(S_{\mu})=\operatorname{ord}\,(S_{\mu}A)=\ell^{h_{\mu}} for \mu=1,\ldots,m . In particular, the cyclic groups generated by the S_{\mu} intersect A only at the identity element. 3. Between the commutators a_{\mu\nu}=[S_{\nu},S_{\mu}] with \mu<\nu there exist only the relations that follow from their definition: a_{\lambda\mu}^{S_{\kappa}-1}a_{\mu\kappa}^{S_{\lambda}-1}a_{\kappa\lambda}^{S_{\mu}-1}=1 and a_{\kappa\lambda}^{f_{\kappa}}=a_{\kappa\lambda}^{f_{\lambda}}=1, where f_{\nu}=\sum_{\lambda=0}^{\ell^{h_{\nu}}-1}S_{\nu}^{\lambda} . In particular, a_{\mu\nu} has symbolic order {\mathfrak{N}}_{\mu\nu}^{(k)}=\Big(\ell^{k},\frac{S_{\mu}^{\ell^{h_{\mu}}}-1}{S_{\mu}-1},\frac{S_{\nu}^{\ell^{h_{\nu}}}-1}{S_{\nu}-1},\ldots,S_{\rho}^{\ell^{h_{\rho}}}-1,\ldots(\rho\neq\mu,\nu)\Big). In [ S3 ] Scholz proved that every two-step metabelian \ell -group with m generators and commutators a_{\mu\nu} whose orders divide \ell^{k} is a quotient group of such a branch group. The simplest branch group has type (2,2;2^{k}) . Here G is generated by two elements S_{1} and S_{2} of order 2 and is therefore a dihedral group; the commutator a=[S_{1},S_{2}] has order 2^{k} and generates the cyclic group A=G^{\prime} , and we have G/A\simeq C_{2}\times C_{2} . Scholz in Freiburg After receiving his doctorate, Scholz moved to Freiburg in April 1929 (and not only in 1930, as Taussky writes in her obituary; see Remmert [ 86 ] ); there he succeeded Baer (who had gone to Hasse in Halle) as Loewy’s assistant and met Zermelo, who was interested in the foundations of mathematics and set theory. Alfred Loewy was born on 20 June 1873 in Rawitsch near Posen and came from a strictly orthodox Jewish family. He studied in Breslau, Munich, Berlin and Göttingen and received his doctorate in 1894 under Lindemann in Munich. In 1897 he received his habilitation in Freiburg, became extraordinary professor in 1902 and full professor in 1919, succeeding Stickelberger. In 1916 Loewy went blind in one eye; a failed operation in 1928 cost him the second eye as well. His doctoral student Krull (PhD 1921) introduced modern algebra in the sense of Emmy Noether in Freiburg. Another of Loewy’s doctoral students, namely F.K. Schmidt, received, after his doctorate, an assistant position created at the request of Heffter and Loewy; as his successor Loewy brought Reinhold Baer to Freiburg in 1926, and after the latter’s habilitation the position went to Arnold Scholz. In 1933, Loewy is forced to retire by the National Socialists. Because the decree of April 1933 is temporarily suspended, Loewy is still able to give a lecture in the summer semester of 1933. Efforts by Heffter to obtain an exceptional permit for Loewy fail, just as similar initiatives elsewhere fail for, among others, Emmy Noether, Issai Schur, and Edmund Landau. Loewy’s succession is arranged by Doetsch; since algebra and number theory are said to have “taken a completely abstract direction”, and indeed “under strong influence from the Jewish side”, a geometer is to be appointed. The choice falls on Wilhelm Süss, who, with his diplomatic skill, knows how to exploit the National Socialist moment for his own purposes like few others. With Scholz’s departure, algebra is dead in Freiburg. Loewy writes his book Blinde große Männer (‘‘Blind Great Men’’), for which he finds no publisher in National Socialist Germany: the book appears posthumously in Zurich. Loewy dies on 25 January 1935 following an operation in Freiburg. As Heffter writes, at his funeral, apart from a few of his mathematical colleagues, only he himself and Hans Spemann 2 2 2 Hans Spemann (1869–1941) was a biologist, active in Freiburg from 1919 until his retirement in 1937, and rector from 1923 to 1924. He received the Nobel Prize in Physiology or Medicine in 1935. attend. His wife followed him a few years later by suicide. Ernst Zermelo was born on 27 July 1871 in Berlin. After studies in Berlin, Halle, and Freiburg, he received his doctorate in Berlin in 1894 with a thesis on the calculus of variations. In 1897 he habilitated in Göttingen and achieved fame in 1904 through his clarification of the axiom of choice, which earned him a professorship in Göttingen in 1905. In 1910 he accepted a professorship in Zurich, which he gave up in 1916 for health reasons and moved to the Black Forest. In 1926, on the application of Heffter and Loewy, he received an honorary professorship in Freiburg, which he held until his denunciation in 1935 by Doetsch’s assistant Eugen Schlotter (the full text of this denunciation can be found in [ 18 , pp. 296–297] ): Zermelo refused to begin his lectures with the Hitler salute. During their common time in Freiburg, Arnold Scholz became Zermelo’s closest friend until Scholz’s death in 1942, which deeply affected Zermelo. Zermelo died on 21 May 1953 in Freiburg. For Zermelo’s life and work, see [ 18 ] . In [ S5 ] , Scholz proves that there exist class field towers of arbitrarily great length. At the end of the 19th century, Hilbert had still conjectured in [ 37 , 38 , 39 ] that the class number of the Hilbert class field of number fields with class number 4 is always odd. This was disproved by Furtwängler in 1916 in [ 23 ] using counterexamples; since then, however, no further progress had been made on this question. Scholz now succeeded in proving that a cyclic number field of prime degree \ell possesses an \ell -class field tower of any prescribed minimal length, provided its \ell -class group has sufficiently large rank. The definitive answer to the question of infinitely long class field towers was provided only by Golod and Shafarevich; refinements of the question remain the subject of numerous investigations even today (Benjamin, Lemmermeyer & Snyder [ 6 , 7 , 8 , 9 ] ; Azizi et al. (see e.g. [ 3 ] , [ 66 ] ); Bartholdi [ 4 ] ; Bush [ 14 , 15 , 16 ] ; Hajir [ 31 , 32 ] ; Kuhnt [ 46 ] ; Maire [ 56 , 57 ] ; D. Mayer [ 58 , 59 , 60 , 61 ] ; McLeman [ 63 ] ; Nover [ 73 ] ; Steurer [ 98 ] ). The group theoretic side of this topic has also been thoroughly investigated (see Magnus [ 55 ] , Serre [ 95 ] , Nebelung [ 70 ] , and more recent articles e.g. on Schur \sigma -groups). In [ S5 ] , Scholz returns to the topic of the (S-1) -th power residues of units. Let S be a generator of the Galois group of the cyclic extension K/{\mathbb{Q}} of prime degree \ell , and set X=S-1 . We recall that an element \alpha\in K^{\times} coprime to \ell is an (\ell,X^{\lambda}) -th power residue (for 1\leq\lambda\leq\ell ) modulo a prime ideal {\mathfrak{p}} if \alpha\equiv\xi^{X^{\lambda}}\eta^{\ell}\pmod{{\mathfrak{p}}} for some \xi,\eta\in K^{\times} ; we agree to write [\alpha,{\mathfrak{p}}]_{\lambda}=1 in this case, and [\alpha,{\mathfrak{p}}]_{\lambda}\neq 1 otherwise. For \lambda=\ell , the element X^{\lambda}=(S-1)^{\ell}=\ell F_{1}(S) in the group ring {\mathbb{Z}}[G] is a multiple of \ell . Therefore, X^{\ell} -th power residues are always also \ell -th power residues, and conversely. Hence, Scholz’s notion of X^{\lambda} -th power residues provides a refinement of the division into \ell -th power residues and non-residues. In the case \ell=3 , the following possibilities arise: 1. [\alpha,p]_{1}\neq 1 ; 2. [\alpha,p]_{1}=1 , [\alpha,p]_{2}\neq 1 ; 3. [\alpha,p]_{2}=1 , [\alpha,p]_{3}\neq 1 ; 4. [\alpha/p]_{3}=1 . Units with norm +1 are X -th powers by Hilbert’s Theorem 90; consequently, [\varepsilon,{\mathfrak{p}}]_{1}=1 for all units \varepsilon with norm 1 . Furthermore, Scholz’s local-global principle holds: Proposition 1 . If [\varepsilon,{\mathfrak{p}}]_{2}=1 for a unit \varepsilon and for all prime ideals {\mathfrak{p}} , then \pm\varepsilon=\eta^{X-1} for some unit \eta . Indeed: if [\varepsilon,{\mathfrak{p}}]_{2}=1 , then \varepsilon^{S-1} is a cube modulo every prime ideal, hence a cube of a unit: \varepsilon^{S-1}=\eta_{1}^{3} . Since 3\sim(S-1)^{2} , it follows that \varepsilon^{S-1}=\eta_{1}^{(S-1)^{2}} , and thus \varepsilon=\eta^{S-1}\beta for some \beta with \beta^{S-1}=1 . This implies \beta\in{\mathbb{Q}} , and since \varepsilon and \eta are units, we must have \beta=\pm 1 . The question of whether [\varepsilon,{\mathfrak{p}}]_{2}=1 or not can be decided by computing the cubic residue symbols modulo the three prime ideals {\mathfrak{p}} lying above p : • If [\varepsilon,p]_{3}=1 , then (\varepsilon,{\mathfrak{p}})_{3}=1 ; • If [\varepsilon,p]_{3}\neq 1 but [\varepsilon,p]_{2}=1 , then (\varepsilon/{\mathfrak{p}})_{3}=(\varepsilon/{\mathfrak{p}}^{\prime})_{3}=(\varepsilon/{\mathfrak{p}}^{\prime\prime})_{3} have the same cubic residue character with respect to all three prime ideals above p , and these are pairwise distinct precisely when [\varepsilon,p]_{2}\neq 1 . Furthermore, Scholz proves the following reciprocity law: Theorem 1 . Let \ell and p\equiv q\equiv 1\pmod{\ell} be odd primes, and let \varepsilon_{p} and \varepsilon_{q} be the fundamental units of the subfields K_{p}^{\ell}\subset{\mathbb{Q}}(\zeta_{p}) and K_{q}^{\ell}\subset{\mathbb{Q}}(\zeta_{q}) of degree \ell , respectively. Then [\varepsilon_{p},q]_{2}=1 if and only if [\varepsilon_{q},p]_{2}=1 as well. In the case \ell=2 , the relation [\varepsilon_{p},q]_{2}=[\varepsilon_{q},p]_{2} also holds provided that p\equiv q\equiv 1\pmod{4} ; in this case the quadratic subfields of the cyclotomic fields are real and possess a fundamental unit. The part of the reciprocity law that concerns the case \ell=2 had already been discovered by Schönemann [ 92 ] . The power-residue characters [\varepsilon,p]_{\ell} are closely related to the divisibility of the class number of K_{p}^{\ell}K_{q}^{\ell} by powers of \ell . On these questions, the last word has not yet been spoken (see e.g. Naito [ 69 ] ). 1930 Olga Taussky, a student of Furtwängler who had studied the capitulation of ideal classes in cyclic unramified extensions, had written a letter to Scholz in November 1929 asking him for advice concerning group theoretic problems. At the meeting of the DMV in Königsberg (Prussia, now Kaliningrad, Russia) in September 1930, Arnold Scholz met Olga Taussky, whom he had met before when he was in Vienna. Scholz and Taussky decide to investigate the capitulation of ideal classes in cubic unramified extensions of quadratic number fields. In [ S6 ] , Scholz corrects an oversight in Furtwängler’s work [ 24 ] on the approximation of real numbers by rationals. Furtwängler started from a classical result of Dirichlet, according to which, for every real number \alpha , there exist infinitely many integers x , y such that \Big|\frac{x}{y}-\alpha\Big|<\frac{1}{y^{2}}. Hurwitz showed that the bound on the right-hand side can be improved to \frac{1}{\sqrt{5}\,y^{2}} , and that this bound is optimal. Already here one can recognize the influence of minimal discriminants: the constant \frac{1}{\sqrt{5}} in Hurwitz’s theorem is related to the fact that D=5 is the minimal discriminant of a real quadratic number field. If one restricts to \alpha\in{\mathbb{R}}\setminus{\mathbb{Q}}(\sqrt{5}\,) , this bound can even be improved to \frac{1}{\sqrt{8}} , and here too D=8 is the smallest discriminant of a real quadratic number field except for {\mathbb{Q}}(\sqrt{5}\,) . Kronecker [ 43 ] generalized Dirichlet’s theorem to higher dimensions: given real numbers \alpha_{1},\ldots,\alpha_{n} , there exist rational numbers with denominator q such that \Big|\alpha_{\nu}-\frac{p_{\nu}}{q}\Big|<\frac{\delta}{q\sqrt[n]{q}} holds for \nu=1,2,\ldots,n and \delta=1 . Minkowski [ 65 , p. 112] succeeded in reducing this bound to \delta=\frac{n}{n+1} . In the other direction, Borel [ 10 ] was able to show, by means of a set-theoretic existence theorem, that Kronecker’s approximation theorem cannot be improved to \Big|\alpha_{\nu}-\frac{p_{\nu}}{q}\Big|<\frac{C}{q^{1+s}} with s>\frac{1}{n} and any constant C>0 . Perron [ 83 ] in turn was able to generalize Borel’s result: let \alpha_{1},\ldots,\alpha_{n} be algebraic integers in a real number field K of degree n+1 ; we assume that there is no relation of the form \sum_{\nu=1}^{n}k_{\nu}\alpha_{\nu}-l=0 with rational k_{1},\ldots,k_{n},l except for the trivial equation with k_{\nu}=l=0 . Furthermore, let \alpha_{\nu}^{(\mu)} ( \mu=1,\ldots,n+1 ) denote the conjugates of \alpha_{\nu} ; we set \rho_{\mu}=\sum_{\nu=1}^{n}|\alpha_{\nu}^{(\mu)}-\alpha_{\nu}|\qquad(\mu=1,2,\ldots,n), as well as \sigma=\prod_{\mu=1}^{n}\rho_{\mu}. Then the system of n inequalities \bigg|\alpha_{\nu}-\frac{p_{\nu}}{q}\bigg|<\frac{\delta}{q\sqrt[n]{q}}\qquad(\nu=1,2,\ldots,n), where p_{\nu} and q>0 denote rational numbers, cannot hold for infinitely many values of q as soon as \delta<\frac{1}{n\sigma} . Perron handed over the topic of the analogous question for complex quadratic number fields to his student Hasan Aral, who wrote his doctoral thesis [ 2 ] on it; incidentally, the example of the unramified cubic extension of the quadratic number field {\mathbb{Q}}(\sqrt{-182}\,) defined by the equation 3 3 3 Scholz probably constructed the polynomial f(x)=x^{3}+ax+b with the solution a=17 and b=2\sqrt{-182} of the equation \operatorname{disc}f=-4a^{3}-27b^{2}=4 . The standard method of constructing unramified cubic extensions would be using ideas from Scholz’s reflection theorem, namely solving the dipohantine equation x^{2}-27\cdot 182y^{2}=a^{3} . Such a solution comes from the fundamental unit \varepsilon=701+30\sqrt{3\cdot 182} and provides us with the polynomial g(x)=x^{3}-3x-1402 , whose roots generate a cyclic cubic unramified extension of k={\mathbb{Q}}(\sqrt{-182}\,) . x^{3}+17x+2\sqrt{-182}=0 in this dissertation was contributed by Scholz (Neiß [ 71 , p. 47] also received from Scholz the example of the quadratic number field {\mathbb{Q}}(\sqrt{-19\,677}\,) with 3 -class group of rank 2 ). Hasan Aral was born on 25 March 1913 in Konya, Turkey; after his Abitur, he received a scholarship from the Turkish government which enabled him to learn the German language from 1932 to 1933 at the Gymnasium Lüben (formerly Silesia, today Lubin; a website Türkische Gastschüler am Lübener Gymnasium has been set up about this gymnasium, from which this information is taken) near Liegnitz in Silesia. In the winter semester 1934/35 he enrolled at Göttingen to study mathematics, and after two semesters he transferred to Munich, where he received his doctorate under Perron on 17 July 1939. Aral’s dissertation was cited by Hofreiter and Koksma, but was later improved using Minkowski’s methods. In [ 24 ] , Furtwängler was able to improve Perron’s bound: Theorem 2 . Let k<|D|^{-1/2(n-1)} be a positive real number, where D denotes the minimal discriminant of a real number field of degree n . Then there exist n-1 real irrational numbers \alpha_{1},\ldots,\alpha_{n-1} , linearly independent over the rationals, for which the n-1 inequalities \Big|\frac{x_{i}}{x_{n}}-\alpha_{i}\Big|0 the equation t^{2}-du^{2}=-4 is solvable. Since this is a central example for Scholz’s theory of knots, let us give a few more details in the special case where k is a real quadratic number field with discriminant d=pq for primes p\equiv q\equiv 1\bmod 4 . Let \varepsilon denote the fundamental unit of k . Here there are the following cases: \begin{array}[]{l|ccc}\raisebox{0.0pt}[11.62494pt][5.59721pt]{\rule{0.0pt}{8.61108pt} }\text{condition}&N\varepsilon&{\operatorname{Cl}}_{2}(k)&{\operatorname{Cl}}_{2}^{+}(k)\\ \hline\cr\raisebox{0.0pt}[11.62494pt][5.59721pt]{\rule{0.0pt}{8.61108pt} }(p/q)=-1&-1&[2]&[2]\\ \raisebox{0.0pt}[11.62494pt][5.59721pt]{\rule{0.0pt}{8.61108pt} }(p/q)=+1,\ (p/q)_{4}=-(q/p)_{4}&+1&[2]&[4]\\ \raisebox{0.0pt}[11.62494pt][5.59721pt]{\rule{0.0pt}{8.61108pt} }(p/q)=+1,\ (p/q)_{4}=(q/p)_{4}=-1&-1&[4]&[4]\end{array} If (p/q)=+1 and (p/q)_{4}=(q/p)_{4}=+1 , then the 2 -class group in the strict sense is cyclic of order divisible by 8 . Scholz proves the reciprocity law (\varepsilon_{p}/q)=(p/q)_{4}(q/p)_{4} for the fundamental unit \varepsilon_{p} in {\mathbb{Q}}(\sqrt{p}\,) already discovered by Schönemann; later Scholz will finde that this quadratic residue symbol is connected norm residues and number knots. The second major work, which was published in 1934 after long efforts, is the paper [ S17 ] , on which Scholz had been working for years together with Olga Taussky. Taussky certainly belonged to the “Jews and Communists” with whom Scholz, according to Doetsch and Schlotter, surrounded himself. The paper deals with 3 -class field towers of imaginary quadratic number fields; more precisely, Scholz and Taussky attempt to restrict the possibilities for the Galois group of the 3 -class field tower by computing the capitulation in such a way that its determination becomes possible or at least the termination of the tower can be proved. This work has left many traces in the mathematical literature; here we content ourselves with references to Brink [ 12 ] , Heider & Schmithals [ 34 ] , and the many recent articles by D. Mayer. The capitulation problem is also, like many other problems that Scholz investigated, of a cohomological nature. In normal unramified field extensions K/k with Galois group G , the capitulation kernel is isomorphic to H^{1}(G,E_{K}) (see the nice survey by Jaulent [ 40 ] ); like all problems in algebraic number theory that have to do with units, this one is also very difficult (or, as Artin once expressed to Taussky, “hopeless”), especially of course in non-cyclic extensions. Even such a simple theorem as the one stating that in abelian unramified extensions K/k a subgroup of order divisible by (K:k) capitulates was proved as late as 1991 by Suzuki [ 99 ] . In July Scholz explores the possibility of transferring his habilitation to Kiel; Hammerstein had only recently been appointed there. In August he visits Kiel for the first time, and at the end of October he moves there permanently. 1935 With the end of the Notgemeinschaft fellowship, Scholz once again begins an arduous struggle for financial independence. His attempts to obtain a paid teaching assignment in Kiel prove more difficult than expected, presumably also due to the continuing disruptive fire from Freiburg. In his report from May 1935, Schlotter writes (it is a sign of the times that even useful idiots like Schlotter are consulted for the assessment of a mathematician such as Arnold Scholz): Politically, Scholz is not reliable. He still does not seem to have come to terms with National Socialism; up to his departure from Freiburg he was constantly associating with Jews and former Communists. Scholz does not want to understand National Socialism. Schlotter, on the other hand, understood National Socialism very well: in 1935 he became Untersturmführer and in 1938 Hauptsturmführer of the Waffen-SS. As late as 1992 he writes: The Waffen-SS was an elite formation of the German Wehrmacht, to which the Occident owes the fact that Stalin’s world conquest did not take place. We are proud of it. Figure 4: Schlotter’s Totenkopfring As for the elite formation: From 6 June 1944 to 17 April 1945 Schlotter commanded the Division ‘‘Nordland’’ of the SS-Panzer-Nachrichten-Abteilung 11. Because he had recognized the hopeless situation in Berlin and, according to his own statements, 8 8 8 Sh. http://www.wehrmacht-awards.com did not want to die a hero’s death, he had himself transferred by a friend to Flensburg (his successor, Hauptsturmführer Schnick, actually fell a few days later). After the war Schlotter taught mathematics at a secondary school until he retired in 1974—two years after the so-called “Radikalenerlass” came into force in Germany, whose goal was keeping left-leaning persons from becoming teachers, postmen or conductors; this included in particular members of the Association of Persecutees of the Nazi Regime and other organizations. From 1935 onward Scholz supervised the dissertation of G. Hannink on the displacement and non-simplicity of groups. In the summer of 1935 it became clear that Scholz was to go to Jena as assistant to F.K. Schmidt. Already on 24 June F.K. Schmidt writes to Hasse, who had suggested Reichardt and Wecken 9 9 9 Franz Wecken passed the state examination in Göttingen in 1935 and received his doctorate in Marburg in 1938. as assistants: In addition, Blaschke has informed me in recent days of the difficulties with which Arnold Scholz unfortunately still has to deal in Kiel. According to this, Scholz has still not received the teaching assignment promised to him and is therefore without means after the expiry of his Notgemeinschaft fellowship. I would therefore like above all to endeavour to achieve something here for Scholz if possible. Melanie Schönbeck 10 10 10 In Kontinuität oder Wandel? writes about Scholz: [His] lack of party membership as well as participation in National Socialist movements were also criticized by the Dean of the Kiel Philosophical Faculty, Ferdinand Weinhandl: “He is neither in the SA nor in the SS, as far as I know, not even in the NSV.” In contrast, according to Theodor Menzel, who had been Dean of the Kiel Philosophical Faculty before Weinhandl, Scholz was “a man of genius, after […] Bieberbach and Tornier one of the best and most promising of the mathematical young generation.” From Kiel, Scholz regularly attended lectures in Hamburg; in the summer of 1935 Blaschke had paid a return visit to Kiel. In his letter to Hasse dated 15 July, F.K. Schmidt once again mentions Scholz: I also thank you very much for mentioning the names Schröder and Wieland[t]. I had already heard about both of them last week from Schur, whom I visited in Berlin. With us, however, things now stand such that I have already definitively offered our assistant position to Scholz. Scholz, who as you know has been in Kiel since last October, has been held up there in the strangest way. Although he has already substituted for Hammerstein for several months, his transfer of habilitation was delayed until the spring of this year and the granting of the teaching assignment to him has not yet been pushed through to this day. Scholz is therefore, after the expiry of his Notgemeinschaft fellowship, i.e., since 1 April of this year, without any payment. I cannot suppress the feeling that Doetsch may have had his hand in all these difficulties. In any case, I know that Doetsch, in the autumn of last year, immediately contacted Kaluza in order to erase the extremely favourable impression that Scholz had left during his visit in Kiel. I therefore believe that a change of location is best for Scholz, and that he should go to a university where he is already known. I have not yet received an answer from Scholz, but assume that it will arrive in the next few days. On 27 August 1935 Scholz writes to Taussky: Now it is as good as decided that I will become assistant to F.K. Schmidt in Jena. Not all formalities have been completed yet; but F.K.S. has paved the way so far. F.K. Schmidt had moved from Göttingen in 1934, where he had been attacked by radical students because of his dealings with Courant, to Jena; there he also encountered problems, also because he had published books by emigrants and Jewish authors for Springer-Verlag, and so in 1941 he moved to the Deutsche Forschungsanstalt für Segelflug in Ainring near Bad Reichenhall. In October 1935 the negotiations between Scholz and Jena end; the rector of Jena writes to Scholz on 28 October 1935 that during his visit in Berlin he was told that it was desired for Scholz to remain in Kiel. F.K. Schmidt also writes to Hasse on 30 October: Allow me to report briefly to you on the further progress of the Scholz matter. After rather protracted negotiations and after it had been possible to interest the Ministry of Culture in the matter, our rector has been informed that Berlin wishes Scholz to remain in Kiel, since several positions are open there. I therefore assume that Scholz will receive a paid position in Kiel in some form for the next semester. In any case, it seems that the objections raised against him have finally been clarified as unfounded by the responsible authorities. I am of course very pleased for Scholz’s sake about this arrangement, even if I regret having to do without him now. As we already discussed in Stuttgart, I have now offered the position to Herr Reichardt for the usual period. On 9 December F.K. Schmidt writes to Hasse again: To my greatest surprise, I heard from Scholz a few days ago that he still has not received the teaching assignment in Kiel. I would like to hope that this is merely one of those delays within the Reich Ministry of Culture that are so frequent today. I really do not want to believe that the Reich Ministry of Culture’s statement to our rector—that it is desired for Scholz to remain in Kiel—was merely a diplomatic maneuver by which one wanted to shake off further pressure in the matter from Jena’s side. As long as Scholz still has the possibility of obtaining something in Kiel, I would consider it unwise if he were to leave his post and accept a foreign offer. Thereupon Scholz inquires with Hasse whether Ore could obtain money for him for a stay in Cambridge. Hasse promises to inquire with Ore, but also writes that it would be unfavourable for Scholz’s prospects in Germany if he were to go abroad now. In December Scholz writes that he would also go to New Haven. In his letter to Hasse dated 19 December 1935 it becomes clear that Scholz does not like to leave his home country: But in the meantime I do have to concern myself with possibilities abroad, and I already reckon that if I go abroad, I will have to give up everything further here, at least for this era […]. Besides this possibility, Scholz also mentions others such as Amsterdam, Cambridge and Yugoslavia. Hasse then writes to Ore on 21 December 1935: We are at the moment somewhat worried about the shaping of the future of the number theorist A. Scholz, who is also known to you. He is currently in Kiel and has long been promised a teaching assignment. But since he evidently has an enemy somewhere, the granting of such a teaching assignment has not yet come about. Ore’s reply from January 1936 is discouraging: I met Scholtz [sic!] in Hamburg last year and he made the best impression upon me. I am afraid I shall have to be rather discouraging about the possibilities for obtaining a position in America at the present time, but there might still be some chances. […] I should, however, not be too hopeful on the outcome of an application. The Institute of Advanced Study in Princeton announces a set of fellowships in the last number of the Bulletin of the American Mathematical Society. Scholtz may also apply for one of these […]. The paper [ S18 ] , in which Scholz proves that the embeddability of the cyclotomic field K_{p}^{\ell^{h}}K_{q}^{\ell^{m}} into a purely ramified tower field is equivalent to the validity of the principal irreality criterion, is dedicated by Scholz to his “teacher I. Schur on his 60th birthday on 10 January 1935”. For this dedication, Scholz was denounced by Fritz Lettenmeyer (1891–1953). Lettenmeyer almost had been dismissed in 1934 because of “political unreliability”. He joined the NSLB (Nationalsozialistischer Lehrerbund) in 1933, the SA and the NSKK (Nationalsozialistisches Kraftfahrer-Korps) in 1934, the NSV (Nationalsozialistische Volkswohlfahrt) in 1936, the NSdAP in 1937 and the NSDDB (Nationalsozialistischer Deutscher Dozentenbund) in 1942 – quite an impressive list, which apparently had no consequences for him after the war: he remained professor in Kiel until his retirement in 1948. Scholz begins with the investigation of norm residues; the corresponding examples, some of which he searched for jointly with Artin’s doctoral student Harald Nehrkorn 11 11 11 Harald Nehrkorn (1910–2006) was a student of Emil Artin and received his Ph.D. in 1933 with his thesis On absolute ideal class groups and units in algebraic number fields . After the war he worked as a teacher. In his lecture on class field theory from the winter semester 1931/32, of which a transcript written after the Second World War has been preserved, Artin computes two examples, namely the field {\mathbb{Q}}(\sqrt{-3},\sqrt{13}\,) of Hasse and the field {\mathbb{Q}}(\sqrt{13},\sqrt{17}\,) “of Nehrkorn and Scholz”. , from Hamburg, he later publishes in [ S19 ] . Let L/K be an abelian extension of number fields, and let {\mathcal{N}} denote the group of all \alpha\in K^{\times} that are norm residues in L/K , i.e., local norms in all completions of L . Scholz now defines \begin{array}[]{lcl}\text{the number knot}&&{\mathcal{K}}_{\alpha}={\mathcal{N}}/N_{L/K}L^{\times},\\ \text{the unit knot}&&{\mathcal{K}}_{\varepsilon}=E_{K}\cap{\mathcal{N}}\Big/E_{K}\cap N_{L/K}L^{\times},\qquad\text{and}\\ \text{the ideal knot}&&{\mathcal{K}}_{(\alpha)}=\{(\alpha):\alpha\in{\mathcal{N}}\}\Big/\{(\alpha):\alpha\in N_{L/K}L^{\times}\}.\end{array} It follows from the definition that these knots are torsion groups, and that their exponents divide the degree (L:K) . Among Scholz’s results is the statement that the sequence \begin{CD}1@>{}>{}>{\mathcal{K}}_{\varepsilon}@>{}>{}>{\mathcal{K}}_{\alpha}@>{}>{}>{\mathcal{K}}_{(\alpha)}@>{}>{}>1\end{CD} is exact; in his own words it reads as follows: The unit knot is isomorphic to a subgroup of the total knot whose quotient group is the ideal knot. Thus for the individual group orders, the knot orders k_{(\alpha)} , k_{\varepsilon} , k_{\alpha} , the relation k_{(\alpha)}\cdot k_{\varepsilon}=k_{\alpha}. holds. Scholz also remarks that the knots for extensions with Galois group (\ell,\ell) are finite \ell -groups, and that in this case the order of {\mathcal{K}}_{\alpha} divides \ell . From earlier results it follows that Rédei fields have a nontrivial number knot. With the term Rédei fields, Scholz denotes composita of cyclic extensions K_{1}/K and K_{2}/K of prime degree, when their discriminants are coprime and when every prime ideal that ramifies in K_{1}/K splits in K_{2}/K and vice versa. Furthermore, Scholz proves by an example that the number field {\mathbb{Q}}(\sqrt{p},\sqrt{q}) , where p and q are primes of the form 4n+1 with (p/q)_{4}(q/p)_{4}=-1 , possesses a unit knot of order \#{\mathcal{K}}_{\varepsilon}=2 . In contrast, {\mathcal{K}}_{\varepsilon}=1 when (p/q)_{4}(q/p)_{4}=+1 . Even André Weil pointed out in [ 103 ] the significance of the discrepancy between norms and norm residues. It was not until the late 1960s that the theory of norm residues again came into focus for various mathematicians. Tate gave a cohomological approach to this theory at the Brighton conference, thereby opening the door to modern methods. Garbanati, in [ 25 ] , was able to prove little that had not already been established by Scholz, but in [ 26 ] he drew attention to Scholz’s work [ S25 ] . Scholz’s works on norm residues were subsequently taken up again by Jehne [ 41 ] , Heider [ 35 ] , Lorenz [ 54 ] , Opolka [ 75 , 76 , 77 , 78 ] , and others. Independently from this development, the arithmetic of tori was developed, initially primarily by Takashi Ono, in which the number knot \nu_{K/k} is occasionally identified with the Tate-Shafarevich group of the norm torus R_{K/k}^{(1)}({\operatorname{GL}}_{1}) ; see Pollio & Rapinchuk [ 84 ] . Since no one has yet emerged to present the elementary parts of this theory in an elementary way, this branch of number theory has not received the dissemination it deserves. 1936 After Scholz had been appointed a member of the scientific examination office, he once again believes that his academic career can take an upward turn and that he will receive the promised teaching assignment by Easter. In [ S19 ] , Scholz recognizes that the existence of knots is connected with the question of the embeddability of certain field extensions, an insight that ultimately leads to the major role played by Schur multipliers in this problem. In his paradigmatic example K=K_{p}^{\ell}K_{q}^{\ell} , a nontrivial knot exists precisely when K can be embedded into a nonabelian extension of degree \ell^{3} whose nontrivial elements, in the case \ell>2 , all have order \ell , while in the case \ell=2 the Galois group is a dihedral group of order 8 . The Galois group of such an extension is generated by the Frobenius substitutions S_{1} of a prime ideal above p and S_{2} of a prime ideal above q ; the commutator A=[S_{1},S_{2}] corresponds to an ideal class of order \ell in K , the so-called commutator class , which Scholz had already introduced in [ S18 ] . The absolute norms of the ideals in this commutator class are principal ideals generated by norm residues in K/{\mathbb{Q}} , without however being norms themselves. Another example from the work [ S19 ] is the following: let k={\mathbb{Q}}(\sqrt{-3}\,) , \eta=18-7\sqrt[3]{17} be the fundamental unit of F={\mathbb{Q}}(\sqrt[3]{17}\,) ; then \eta=\varepsilon^{S-1} in K_{1}=kF , where S is the generating substitution of K_{1}/k , and \varepsilon and \varepsilon^{S} are the two fundamental units of k . The extension K_{2}/k is cyclic of degree 3 with prime ideal conductor {\mathfrak{q}} , where {\mathfrak{q}} is to split completely in K_{1}/k . For a cyclic extension F/k of prime degree \ell , a prime ideal {\mathfrak{q}} in k , and an \alpha\in F^{\times} , denote by r(\alpha,{\mathfrak{q}}) the maximal exponent r\leq\ell such that \alpha\equiv\mu^{(S-1)^{r}}\nu^{\ell}\pmod{{\mathfrak{q}}}. Since units are (S-1) -power residues, we have r(\varepsilon,{\mathfrak{q}})\geq 1 . Scholz shows: the compositum K=K_{1}K_{2} possesses a unit knot if and only if r(\varepsilon,{\mathfrak{q}})=1 , and an ideal knot otherwise. Results such as this one are abundant in Scholz’s works and deserve far more attention than they have received in the past. A very convincing presentation of the theory of number knots was later given by Wolfram Jehne (1926–2018) in [ 41 ] . Here we content ourselves with presenting the essential definition of the knots as given by Jehne. Let k be a number field; the embedding of k^{\times} into the idele group J_{k} gives an exact sequence \begin{CD}1@>{}>{}>k^{\times}@>{}>{}>J_{k}@>{}>{}>C_{k}@>{}>{}>1,\end{CD} where C_{k} denotes the idele class group of k . The kernel of the map J_{k}\twoheadrightarrow I_{k} from the idele group onto the group of fractional ideals I_{k} is the group U_{k} of idele units, which gives us a second exact sequence \begin{CD}1@>{}>{}>U_{k}@>{}>{}>J_{k}@>{}>{}>I_{k}@>{}>{}>1.\end{CD} Finally, there is the classical sequence \begin{CD}1@>{}>{}>H_{k}@>{}>{}>I_{k}@>{}>{}>{\operatorname{Cl}}_{k}@>{}>{}>1,\end{CD} which defines the ideal class group {\operatorname{Cl}}_{k} as the quotient of the ideal group I_{k} by the group H_{k} of principal ideals. These exact sequences form the fundamental square \begin{CD}111\\ @V{}V{}V@V{}V{}V@V{}V{}V\\ 1@>{}>{}>E_{k}@>{}>{}>k^{\times}@>{}>{}>H_{k}@>{}>{}>1\\ @V{}V{}V@V{}V{}V@V{}V{}V\\ 1@>{}>{}>U_{k}@>{}>{}>J_{k}@>{}>{}>I_{k}@>{}>{}>1\\ @V{}V{}V@V{}V{}V@V{}V{}V\\ 1@>{}>{}>{\mathcal{E}}_{k}@>{}>{}>C_{k}@>{}>{}>{\operatorname{Cl}}_{k}@>{}>{}>1\\ @V{}V{}V@V{}V{}V@V{}V{}V\\ 111\\ \end{CD} of the algebraic number field k . Now Jehne considers exact sequences \begin{CD}1@>{}>{}>A_{k}@>{}>{}>B_{k}@>{}>{}>C_{k}@>{}>{}>1\end{CD} (2) of abelian groups like those from the fundamental square. If K/k is normal with Galois group G , and if N=N_{K/k} denotes the relative norm, then an application of the snake lemma to the diagram \begin{CD}1@>{}>{}>A_{K}@>{}>{}>B_{K}@>{}>{}>C_{K}@>{}>{}>1\\ @V{}V{N}V@V{}V{N}V@V{}V{N}V\\ 1@>{}>{}>A_{k}@>{}>{}>B_{k}@>{}>{}>C_{k}@>{}>{}>1\\ \end{CD} yields the exact sequence \begin{CD}1@>{}>{}>A_{K}[N]@>{}>{}>B_{K}[N]@>{}>{}>C_{K}[N]@>{\delta}>{}>\\ {}{}A_{k}/NA_{K}@>{}>{}>B_{k}/NB_{K}@>{}>{}>C_{k}/NC_{K}@>{}>{}>1,\end{CD} where A[N] denotes the kernel of the norm map on A . The connecting homomorphism \delta maps an element a=\alpha A_{K}\in{}_{N}C_{K} to \delta(a)=N_{K/k}(\alpha)N_{K/k}A_{K} ; since \alpha is an element of B_{K} whose relative norm lands in A_{k} , this is well-defined. Thus it follows that {\operatorname{im}\,}\delta=A_{k}\cap N_{K/k}B_{K}/N_{K/k}A_{K}=:[A,B]. Jehne then calls [A,B] the knot associated to the sequence ( 2 ). If one splits this sequence at the place \delta , one obtains two exact sequences in which this knot appears: 1\longrightarrow{}_{N}A_{K}\longrightarrow{}_{N}B_{K}\longrightarrow{}_{N}C_{K}\longrightarrow[A,B]\longrightarrow 1 (3) 1\longrightarrow[A,B]\longrightarrow A_{k}/NA_{K}\longrightarrow B_{k}/NB_{K}\longrightarrow C_{k}/NC_{K}\longrightarrow 1. (4) The exact sequences of the fundamental square thus yield six knots, of which one is trivial: [U,J]=1 , as can easily be shown. Thus five knots remain, namely 1. the number knot \nu=\nu_{K/k}=[K^{\times},J_{K}]=k^{\times}\cap NJ_{K}/NK^{\times} ; 2. the first unit knot \omega=\omega_{K/k}=[E_{K},K^{\times}]=E_{k}\cap NK^{\times}/NE_{K} ; 3. the second unit knot \omega^{\prime}=\omega^{\prime}_{K/k}=[E_{K},U_{K}]=E_{k}\cap NU_{K}/NE_{K} ; 4. the ideal knot \delta=\delta_{K/k}=[H_{K},I_{K}]=H_{k}\cap NI_{K}/NH_{K} ; 5. the idele knot \gamma=\gamma_{K/k}=[{\mathcal{E}}_{K},C_{K}]={\mathcal{E}}_{k}\cap NC_{K}/N{\mathcal{E}}_{K} . With very little effort one now obtains from the main theorems of class field theory Theorem 3 ( The Fundamental Knot Sequence). 1\longrightarrow\omega_{K/k}\longrightarrow\omega^{\prime}_{K/k}\longrightarrow\nu_{K/k}\longrightarrow\delta_{K/k}\longrightarrow\gamma_{K/k}\longrightarrow 1. Scholz’s unit knot \omega^{0} is the quotient \omega^{\prime}/\omega\simeq E_{k}\cap NU_{K}/E_{k}\cap NK^{\times}; using the knot \delta^{0}:={\operatorname{im}\,}\,(\nu\longrightarrow\delta) we obtain from the fundamental knot sequence Theorem 4 ( Scholz’s Knot Sequence). \begin{CD}1@>{}>{}>\omega^{0}_{K/k}@>{}>{}>\nu_{K/k}@>{}>{}>\delta^{0}_{K/k}@>{}>{}>1.\end{CD} These knots allow an interpretation as Galois groups of certain subfields of the Hilbert class field: Theorem 5 . We have \delta^{0}\simeq{\operatorname{Gal}}(K_{\operatorname{cen}}/K_{\operatorname{gen}}),\quad\delta\simeq{\operatorname{Gal}}(K_{\operatorname{cen}}/k^{1}K)\quad\text{and}\quad\gamma\simeq{\operatorname{Gal}}(K_{\operatorname{gen}}/k^{1}K). Rédei fields such as {\mathbb{Q}}(\sqrt{p},\sqrt{q}\,) with p\equiv q\equiv 1\bmod 4 and (p/q)=+1 admit an unramified cyclic quartic extension L/k , hence K_{\operatorname{cen}} is strictly bigger than K_{\operatorname{gen}}={\mathbb{Q}}(\sqrt{p},\sqrt{q}\,) . Thus it must have a nontrivial knot. The middle term of Scholz’s knot sequence is connected with the Schur multiplier {\mathfrak{M}}(G) of the Galois group G={\operatorname{Gal}}(K/k) . The Schur multiplier of a finite group G is defined as follows: a group extension E:\begin{CD}1@>{}>{}>A@>{}>{}>\Gamma @>{}>{}>G@>{}>{}>1\end{CD} is called central if A\subseteq Z(\Gamma) , and a splitting (or split extension) if A\subseteq\Gamma^{\prime} . Schur [ 93 ] has shown that the orders of the groups A in central splittings are bounded, and that all groups A in maximal central splittings are isomorphic to one another. The isomorphism class of these groups A is called the Schur multiplier of G , and the corresponding groups \Gamma are called representation groups (or covering groups). It must have deeply satisfied Scholz to have uncovered the major role played by the Schur multiplier—which his doctoral supervisor Schur had introduced into representation theory—in class field theory. In his works on norm residues published in Crelle’s Journal [ S19 , S25 ] , he did not cite Schur, or perhaps could not cite him; only in the group-theoretic part [ S23 ] , which appeared in the Monatshefte für Mathematik und Physik in Austria, do we find the relevant citations of Schur’s papers. After the development of cohomology, it turned out that {\mathfrak{M}}(G)\simeq H^{2}(G,{\mathbb{Q}}/{\mathbb{Z}}) . This implies Theorem 6 . Let K/k be a normal extension of number fields with Galois group G . Then there exists an exact sequence \begin{CD}\coprod_{\mathfrak{P}}{{\mathfrak{M}}(G_{\mathfrak{P}})}\,\widehat{\phantom{n}\!\!}@>{}>{}>{{\mathfrak{M}}(G)}\,\widehat{\phantom{n}\!\!}@>{}>{}>\nu_{K/k}@>{}>{}>1,\end{CD} where G_{\mathfrak{P}} denotes the Galois group of the local extension K_{\mathfrak{P}}/k_{\mathfrak{p}} , and where K_{\mathfrak{P}} and k_{\mathfrak{p}} are the completions of K and k at the prime ideal {\mathfrak{P}}\mid{\mathfrak{p}} . Furthermore, \widehat{A}={\operatorname{Hom}}(A,{\mathbb{Q}}/{\mathbb{Z}}) denotes the Pontryagin dual group of A . Since the Schur multiplier of cyclic groups is trivial, this theorem has the Hasse norm theorem as a corollary. In addition to his first investigations of knots and their connection with the construction of field extensions with prescribed Galois groups, Scholz studies Furtwängler’s work on rational approximability and the related question of the size of minimal discriminants of number fields with prescribed degree. He publishes his results in his article on minimal discriminants [ S21 ] . In doing so, Scholz takes up again a question from [ S6 ] : how can one find towers of field extensions whose discriminants remain relatively modest? In this direction, apart from Perron’s example of the fields {\mathbb{Q}}(\sqrt[n]{2}\,) , nothing was known. Among all number fields of degree n , let K_{n} be a number field with minimal discriminant (meaning the absolute value). Scholz considers the real number E_{n}=\sqrt[n]{|\operatorname{disc}K_{n}|} , which in the modern literature is called the root discriminant \operatorname{rd}\,(K_{n}) . Results of Minkowski show \liminf_{n\to\infty}E_{n}\geq e^{2}, and Blichfeldt improved this bound to \geq e\pi . Perron’s fields {\mathbb{Q}}(\sqrt[n]{2}\,) yield E_{n}<2n ; from the existence of infinite class field towers, as Scholz writes, it would follow that \liminf_{n\to\infty}E_{n}/n=0. By means of cyclotomic fields K/{\mathbb{Q}} one can only slightly improve Furtwängler’s upper bound, because for such fields E_{n}\geq C\cdot\frac{n\log\log n}{\log n}. for K={\mathbb{Q}}(\zeta_{p}) we have \operatorname{disc}K=p^{p-2} , so \operatorname{rd}\,(K)=p^{\frac{p-2}{p-1}} . By considering ray class fields over cyclotomic fields, however, Scholz succeeds in proving that E_{n}=O\Big(\frac{\log n}{\log\log n}\Big)^{2}. Scholz’s construction in fact yields unramified abelian extensions of suitable cyclotomic fields; see Lemmermeyer [ 50 ] . Scholz’s conjecture that \lim_{n\to\infty}E_{n}/n=0 remains open even despite the techniques of Golod and Shafarevich. The difficulty lies in proving the existence of prime-degree extensions with small discriminant. Similarly, Scholz’s conjecture that the minimal discriminant for fields of degree n=400 is smaller than that for n=397 is still open. There are a whole series of further questions from this area that seem interesting. For example, Ankeny [ 1 ] has shown that there exists a constant c>0 such that \log\operatorname{rd}\,(K)>c\cdot\log_{m}[K:{\mathbb{Q}}] for all normal extensions K/{\mathbb{Q}} whose Galois group G is metabelian of step m\geq 1 , for which the m -th commutator subgroup G^{(m)}=1 vanishes, but the (m-1) -th does not. Here, \log_{m}(x)=\log\log\cdots\log(x) denotes the m -fold iterated logarithm. It is, to my knowledge, not known whether this bound is optimal. For m=1 , i.e., abelian groups, the fields K={\mathbb{Q}}(\zeta_{p}) have root discriminant p^{\frac{p-2}{p-1}} , and thus \log\operatorname{rd}\,(K)=\frac{p-2}{p-1}\log p<\log(p-1)=\log[K:{\mathbb{Q}}]. Scholz’s result shows that such families of fields also exist for m=2 , since for the metabelian fields he constructed the inequality \log\operatorname{rd}\,(K)<2\log_{2}[K:{\mathbb{Q}}] holds. From de Gruyter, Scholz receives the offer to write an introduction to elementary number theory for the Göschen Collection. 1937 In search of a permanent position, Scholz writes to Hasse on 3 November 1937 that he might be well placed in Tübingen, where a successor to Kamke was currently being sought. Kamke was married to a woman of Jewish descent, although she had converted to Protestantism. Already in April 1937, Kamke was given the choice of divorcing his wife or losing his position; Kamke (see Segal [ 94 , pp. 105] ) chose the latter and was dismissed on 1 November. That Kamke survived the war years unharmed is due, among other things, to the fact that Süss kept his protective hand over him. In 1937, Scholz publishes his paper [ S20 ] on the construction of algebraic number fields whose Galois group is a prescribed p -group. The fact that he dedicates this article “to the memory of Emmy Noether” will hardly have helped his efforts to secure a position. In this paper, Schur’s representation groups appear in connection with embedding problems, once again demonstrating how closely such questions are linked to the theory of knots that Scholz would later develop. In 1937, Scholz’s Problem 253 on “addition chains” appears in the Jahresberichte, which, due to its simple formulation and its complexity, quickly became famous and notorious. Alfred Brauer succeeds in [ 11 ] in answering some of the problems, but Scholz’s actual conjecture has remained open to this day and is discussed in Guy’s “Unsolved Problems in Number Theory” [ 30 , C6] . It was only later discovered that the problem of addition chains had already been formulated in 1894 by H. Dellac in [ 17 ] . Scholz revises his second paper on norm residues for the sixth time; this work [ S25 ] will only appear in 1940 as one of his last publications. Since the mathematical language suitable for describing Scholz’s insights (above all cohomology theory) has not yet been developed, this paper is among the most difficult to read. Here too, Scholz has hidden many gems so well within the text that his priority was only recognized much later: on p. 227, for example, he writes: Moreover, it is shown that every finite group occurs as the Galois group of an unramified relative extension. This immediately implies that there exist arbitrarily long class field towers. Artin had already proved this theorem in 1934, but did not publish it: cf. [ 19 , 15.1] and [ 51 , 7.27] . Later, Fröhlich [ 21 ] and Uchida [ 102 , Cor. 1] rediscovered this result; a strengthening for p -groups can be found in Nomura [ 72 ] , and Ozaki [ 82 ] was finally able to show that for every p -group G there exists not only an unramified extension K/k with Galois group G , but moreover one may additionally require that K is the p -class field tower of k . 1938 In October 1938, Hammerstein asks Süss – who had been proposed as chairman of the DMV in August 1937 and was elected shortly afterwards at the annual meeting in Bad Kreuznach – for an expert opinion on Scholz (Hasse was also contacted; he replies (see [ 86 , p. 62–63] ): I must honestly confess that, in this sense, I would certainly not yet propose Mr. Scholz to the government of the Third Reich at the present moment. His attitude has left much to be desired for years. In 1938, Süss also actively participated in removing Issai Schur from the editorial board of the Mathematische Zeitschrift (see Remmert [ 87 ] ). The highly ambiguous role that Süss played during the Third Reich did not harm him in any way after 1945: Süss was certainly a master of walking the tightrope between diplomatic skill and opportunism. Süss faced almost no problems after the war, although Ferdinand Springer, backed by F. K. Schmidt, accused Süss of having tried to have Schmidt arrested in 1939 when Schmidt was visiting the US (see Remmert [ 86 ] ). Springer had sent Schmidt to the USA in connection with the Zentralblatt; Süss accused Schmidt of having strong ties to Jewish emigrants and suggested he himself should go. After the war, Süss accused Schmidt of having given an erroneous report. Scholz completes his work on his Introduction to Number Theory [ S22 ] in the Göschen Collection. This book begins with the derivation of the most important fundamental laws of the natural numbers from the axioms 12 12 12 Here Zermelo’s influence can be seen. , develops elementary number theory within the framework that has been essentially standard since Gauss, and then provides an introduction to the theory of binary quadratic forms. In the final chapter, Scholz explains various tricks and methods for computations in number theory. Scholz’s hopes for a position in Vienna are dashed; depressions make it hard for him to do mathematics. In the following years, Scholz limits himself to completing works already begun. In particular, the continuation of the work on minimal discriminants, which had already been announced, never appeared: On 27 July 1938, Scholz responds to an inquiry from Hasse concerning minimal discriminants of number fields with prescribed degree, and in doing so introduces the Dirichlet series D_{n,G}(s)=\sum|D_{K}|^{-s} , where the summation extends over all fields of degree n whose normal closure has Galois group G . He also communicates various results on the abscissa of convergence a(G) of such series; in particular, for degree- 4 extensions, • a(D_{4})=1 for the dihedral group D_{4} of order 8 , • a(V_{4})=a(C_{4})=\frac{1}{2} for the Klein four group V_{4} and the cyclic group C_{4} of order 4 . Moreover, Scholz correctly conjectures that a(A_{4})=\frac{1}{2} and a(S_{4})=1 . These results are connected with asymptotic questions about the number of number fields with given Galois groups, which have been intensively studied in recent years. One of the central conjectures in the theory of counting Galois extensions is the following: If N(G,x) denotes the number of all extensions of degree n with discriminant \leq x whose normal closure has Galois group G , then N(G,x)=c\cdot x^{a(G)}(\log x)^{b(G)} holds for certain constants a , b , c that depend only on the base field and G . Writing Z_{G}(s)=\sum D_{K}^{-s}=\sum a_{n}n^{-s} , we have \sum_{k=1}^{n}a_{k}=N(G,n) . From known properties of Dirichlet series, it then follows that a(G) equals the abscissa of convergence of Z_{G}(s) . In particular, the values claimed by Scholz, namely a(D_{4})=1 and a(C_{4})=a(V_{4})=\frac{1}{2} , are all correct. Moreover, Scholz’s conjectures a(S_{4})=1 and a(A_{4})=\frac{1}{2} turned out to be true, as has the claim that \lim_{s\to 1/2}\frac{Z(s)}{V(s)}=0 on account of b(C_{4})=0 and b(V_{4})=1 . Nothing is known about Scholz’s proofs. 1939 In 1939, Scholz publishes an original introduction to number theory [ S22 ] in the Göschen Collection. From 26 to 30 June, Hasse organizes a group theory conference in Göttingen, to which he has invited Philip Hall. The first four days each begin with a two-hour lecture by Hall, after which Magnus, Zassenhaus, Grün, Speiser, Wielandt, Witt, van der Waerden, Krull, and Specht give talks; Scholz delivers a lecture on 29 June from 18:00 to 19:00 on Extension and splitting of groups , thus on the content of his papers [ S23 , S24 ] . Many of these lectures were published in volume 182 of Crelle’s Journal. In [ S23 ] , Scholz investigates the group-theoretic side of field constructions as they are needed in his theory of norm residues. In order to convey at least a rough idea of Scholz’s terminology, we here provide a brief introduction to the presentation of the theory as it was developed by Heider [ 35 ] and later by Steinke [ 97 , § 2] . To this end, let K/k be a normal extension of number fields. If an extension L/K is normal over k , then we have an exact sequence \begin{CD}1@>{}>{}>{\operatorname{Gal}}(L/K)@>{}>{}>{\operatorname{Gal}}(L/k)@>{}>{}>{\operatorname{Gal}}(K/k)@>{}>{}>1\end{CD} A normal extension L/K/k is called • central over K/k if {\operatorname{Gal}}(L/K)\subseteq Z({\operatorname{Gal}}(L/k)) , that is, the Galois group of L/K lies in the center of the Galois group of L/k . • abelian over K/k if {\operatorname{Gal}}(L/K)\cap{\operatorname{Gal}}(L/k)^{\prime}=1 , that is, the Galois group of L/K and the commutator subgroup of the Galois group of L/k are disjoint. Observe that if L/K is abelian over K/k , then L/k is not necessarily abelian. For example, L/K is abelian over K/k whenever L=KF for some abelian extension F/k . Now let L_{1}/k and L_{2}/k be central extensions over K/k . If there exists a central extension L/k over K/k that contains both L_{1} and L_{2} and which is abelian over L_{1}/k and over L_{2}/k , then L_{1} and L_{2} belong to the same genus . Opolka has shown in [ 74 ] that the genera of central extensions of a Galois extension K/k of global or local fields correspond bijectively to the quotient groups of H^{-1}(G,C_{K}) , where C_{K} denotes the idele class group of K in the global case and the multiplicative group K^{\times} in the local case. Here once again it becomes apparent how Scholz, with his own language, created a substitute for the cohomological structures that were not yet available at the time. If, in addition, there exists an abelian extension M/k that contains the maximal abelian subextension of K/k and for which L_{1}M=L_{2}M and L_{1}\cap M=L_{2}\cap M hold, then one says that L_{1} and L_{2} belong to the same class, or that they arise from each other by abelian crossing . Scholz calls the group F the relative product of F_{1} and F_{2} if F is chosen minimally in Fig. 5 . Hilbert introduced the technique of abelian crossing in his proof of the Kronecker–Weber theorem: there he showed that one can make an abelian extension K/{\mathbb{Q}} disappear by abelian crossing with cyclotomic fields, from which it immediately follows that K is contained in a compositum of cyclotomic fields. Chebotarev was able to prove his density theorem using the same technique, and Artin employed it in the proof of his reciprocity law. Despite its outstanding importance, Scholz’s penetration of this concept with the help of group theory found little resonance. From the group-theoretic point of view, the matter comes down to the following: let G={\operatorname{Gal}}(K/k) be a finite group and set F_{j}={\operatorname{Gal}}(L_{j}/k) ; then the sequences \begin{CD}1@>{}>{}>A_{j}@>{}>{}>F_{j}@>{}>{}>G@>{}>{}>1\end{CD} (5) are exact, and if we identify A_{j} with its image in F_{j} , then A_{j}\subseteq Z(F_{j}) . If there exist abelian group extensions \begin{CD}1@>{}>{}>N_{j}@>{}>{}>F@>{}>{}>F_{j}@>{}>{}>1,\end{CD} for which the diagram in Fig. 5 is commutative, then the exact sequences ( 5 ) belong to the same genus. {\lx@inpgf@ignorespaces N_{1}} {\lx@inpgf@ignorespaces N_{2}} {\lx@inpgf@ignorespaces A_{2}} {\lx@inpgf@ignorespaces F} {\lx@inpgf@ignorespaces A_{1}} {\lx@inpgf@ignorespaces F_{2}} {\lx@inpgf@ignorespaces F_{1}} {\lx@inpgf@ignorespaces G} \scriptstyle{\lx@inpgf@ignorespaces g_{2}} \scriptstyle{\lx@inpgf@ignorespaces g_{1}} \scriptstyle{\lx@inpgf@ignorespaces p_{2}} \scriptstyle{\lx@inpgf@ignorespaces p_{1}} Figure 5: Scholz’s group genera Furthermore, if there exists a normal subgroup N of F such that F/N is abelian and with NN_{1}=NN_{2} and N_{1}\cap N=N_{2}\cap N=1 , then the exact sequences ( 5 ) belong to the same class, or arise from each other by abelian crossing. 1940 In the year 1940, G. Hannink receives his Ph.D. under Scholz in Kiel with the thesis 13 13 13 Hannink cites Schur in a footnote. “Transfer and non-simplicity of groups”. In May, Scholz is drafted, trained as a radio operator, and sent to the East. 1941 In the year 1941, Scholz is assigned as a mathematics teacher at the Naval Academy Flensburg-Mürwik, together with Ott-Heinrich Keller (1906–1990). On 28 February, Scholz is in Kiel to attend the funeral of Hammerstein, who had died on 25 February “due to malnutrition affecting the lungs”. Scholz organizes a colloquium on 19 July in Göttingen on the occasion of Zermelo’s 70th birthday; Zermelo himself delivers three lectures, and the other speakers are Konrad Knopp and Bartel van der Waerden. In December, Scholz learns that he cannot count on remaining in Kiel after the end of the war. Mathematically, he occupies himself with infinite algebraic extensions of number fields, which had already been studied before him by Krull and Herbrand; Moriya [ 67 , 68 ] had also concerned himself in 1936 with the development of a class field theory over infinite algebraic number fields. The corresponding paper, which Scholz had originally intended to deliver to Hasse in Berlin at Christmas, was only published posthumously. 1942 Scholz dies on 1 February 1942 in Flensburg. In [ 5 , p. 141] it is stated about this: Occasionally there were also musical contacts with the number theorist Arnold Scholz, who taught as a lecturer at the University of Kiel. Heesch later reported of him that he had hardly known any other person who could listen to music so well and understand its spirit. Scholz considered Franz Schubert the greatest musician in the entire history of mankind and his String Quartet in C major, op. 163 as the greatest musical work. While Heesch was active during the war as a “civilian teacher” at the Naval Gunnery School in Kiel, Scholz taught as a civilian teacher at the similarly named school in Flensburg. They had mutual friends at whose homes they were often invited. At one of these festive evenings in January 1942, where “despite the darkness of the times, there was also dancing”, Scholz struck the hostess by his unusual thirst. Eight days later Arnold Scholz was dead. Cause: “galloping diabetes”. “Galloping” was an archaic adjective used in medicine to describe diseases that advance quickly and aggressively. Heesch had already met Olga Taussky in Göttingen in 1931, who occasionally joined him, Franz Rellich, Ernst Witt, Stefan Cohn-Vossen, and Fritz John for communal lunches. Scholz’s mother sends his estate to relatives in Poland; during the expulsion of the Germans in 1945, all documents are lost. Regarding the history of Scholz’s estate, Hasse wrote to Olga Taussky on 29 October 1949: Mrs. Scholz 14 14 14 Mars. Scholz is his mother, who was living in Zurich at the time. initially wrote to me about the estate as follows: It was stored in a hard case and several suitcases with his relatives Josef Stenzel in Günthersdorf near Brieg/Oder 15 15 15 Today: Brzeg . When the Poles arrived, these relatives had to vacate their property, and Arnold Scholz’s entire estate along with some of his furniture was “taken into custody” by the Poles. Given the situation, it is hardly possible for us today to gain access to these items. This indeed proved to be the case, so that the entire estate of Scholz (as well as that of Emmy Noether) must be considered completely lost; Hasse himself calls this “a very painful loss”. 1943 Scholz’s last paper [ S26 ] , which he dedicated to his recently deceased colleague Hammerstein, is published posthumously. It concerns a new foundation for the number theory of algebraic field extensions of infinite degree, in which he recovers earlier results of Krull [ 44 , 45 ] and Herbrand [ 36 ] by different means. At that time, it was in no way foreseeable that these ideas would one day bear rich fruit: following the groundbreaking work of Iwasawa, infinite extensions of number fields became an indispensable tool for studying arithmetic questions in finite extensions. 1949 Ott-Heinrich Keller (1906–1990) dedicates his paper [ 42 ] “to the memory of Arnold Scholz”. 1952 Olga Taussky publishes an obituary [ 101 ] for Arnold Scholz. 1969 Emma Lehmer discovers the reciprocity law (\varepsilon_{p}/q)=(p/q)_{4}(q/p)_{4} for primes p\equiv q\equiv 1\bmod 4 with (p/q)=1 and asks Hasse (letter from October 24) if he knows this result. Hasse replies that he does not. Lehmer finds the result in Scholz’s article and tells Hasse about in a letter from January 24, 1970. 1979 Wolfram Jehne publishes his major work [ 41 ] on number-theoretic knots “In memoriam Arnold Scholz” in Crelle’s Journal. 1982 On the occasion of the 40th anniversary of his death and the publication of Heider and Schmithals [ 34 ] , Crelle’s Journal publishes a photograph of Arnold Scholz. 2016 Lemmermeyer and Roquette publish the correspondence between Hasse, Scholz and Taussky. Publications of Arnold Scholz Problems in the Jahresberichte der DMV Problems posed: 194, vol. 45 ; 208, vol. 45 ; 211, vol. 46 ; 212, vol. 46 ; 222, vol. 46 ; 232, vol. 46 ; 249, vol. 47 ; 253, vol. 47 ; Problems solved: 153, vol. 43 ; 169, vol. 44 ; 171, vol. 44 ; 175, vol. 44 ;