On the class number of real multiquadratic fields

\bullet Let us firstly note that if h_2(K)=2 and d is odd, then the rank of the 2-class group of K equals 1; and so, according to [ChemseddinGreenbergConjectureI, Lemma 3.1], d is in one of the following forms:

1. d=r is a prime number such that r\equiv1\pmod 8 and \left(\frac{r}{q}\right)=1,

2. d=rs where r and s are two distinct prime numbers.

Assume that d takes the first form (i.e. as in a)). So the extension K/\mathbb{Q}(\sqrt{qr}) is an unramified quadratic extension. Since the 2-class group of \mathbb{Q}(\sqrt{qr}) is cyclic (cf. [conner88, Corollary 18.6]), it follows that h_2(K)=2 if and only if h_2(qr)=4 (this justifies the form A)).

Now assume that d takes the second form (i.e. as in b)). Let us characterize the integers d such that h(K)=2. Put k=\mathbb{Q}(\sqrt{qrs}). We distinguish the following two cases:

229 Assume that K/k is ramified, i.e., rs\equiv 3\pmod4. So h_2(k) divides h_2(K), and thus h_2(k)=1 or 2. The case h_2(k)=1 is rejected as h_2(k) is divisible by 2 (cf. [conner88, Corollary 18.6]). Therefore, h_2(k)=2;

and so, by Lemma 1.2, we have q\equiv s \equiv3 \pmod4 and r\equiv1\pmod4 with \left(\dfrac{q}{r}\right)=-1 or \left(\dfrac{s}{r}\right)=-1. We distinguish two sub-cases:

1. Let r\equiv5\pmod 8. Using Tables ? and ?, and Wada’s algorithm (cf. Page ?), we check that E_K= \langle{-1,\varepsilon_q,\varepsilon_{rs},\eta(r,s,q)\rangle}, where \eta(r,s,q) is as in Table Proof:

   The paper has a table here, which this page does not show.

   Thus, q(K)=2. Since h_2(rs)=h_2(qrs)=2 (cf. Lemma 1.2), it follows by Kuroda’s class number formula that we have h_2(K)=2 (this justifies the form B)-1)).

2. Now let r\equiv1\pmod 8. According to Lemma 1.2, [h_2(qrs)=2 and h_2(rs)=2] if and only if \left(\dfrac{s}{r}\right)=-1. Furthermore, from Tables ? and ?, we deduce that E_K= \langle{-1,\varepsilon_q,\sqrt{\varepsilon_{q}\varepsilon_{rs}},\eta\rangle} where \eta=\varepsilon_{qrs} or \sqrt{\varepsilon_{qrs}} according to whether \left(\dfrac{s}{r}\right)=\left(\dfrac{s}{r}\right)=-1 or not. Therefore, q(K)=2 exactly when \left(\dfrac{s}{r}\right)=\left(\dfrac{q}{r}\right)=-1. It follows, by Kuroda’s class number formula (cf. (?)), that h_2(K)=2 if and only if \left(\dfrac{q}{r}\right)= \left(\dfrac{s}{r}\right)=-1 (this justifies the form B)-2)).

229 Now assume that K/k is unramified, i.e., rs\equiv1\pmod 4. We distinguish two sub-cases:

1. Let r\equiv s\equiv 3\pmod4, then [BS95, Theorem 1] implies that k^{(1)}=k^{(2)}. Therefore,

   h_2(K)=\frac{h_2(k)}{2};

   and thus h_2(K)=2 if and only if h_2(k)=4. Under the conditions of this sub-case, the quadratic fields k=\mathbb{Q}(\sqrt{qrs}) such that h_2(k)=4 can be deduced from [BS25I, Form d_i, for i\in \{1,...,8\}, page 11]. In fact, we have h_2(k)=4 if and only if, after a suitable permutation of the prime numbers q,r and s, we have \left(\dfrac{r}{s}\right)=\left(\dfrac{s}{q}\right)=1 and one of the conditions in Table Proof :

   1.31.6

   The paper has a table here, which this page does not show.

   This justifies the forms listed from B)-3) to B)-6).

2. Let r\equiv s \equiv 1\pmod4. Notice that if h_2(K)=2, then the class field tower of k is abelian, i.e., k^{(1)}= k^{(2)}. Thus, using the classification in [BS95, Form c_i, for i\in \{15,...,20\}, page 177-178], by interpreting the graphs corresponding to the abelian case to those of type 7 (for a definition of these graphs reader can see page 168 in the same reference), we deduce that r, s and q satisfy one of the conditions in Table Proof:

So let us assume that q, r and s satisfy one of the conditions presented in Table Proof. Notice that h_2(qrs)=4 and since in all cases we have \left(\dfrac{r}{s}\right)=-1, it follows that h_2(rs)=2 (cf. Lemma 1.2). Thus, by Kuroda’s class number formula, h_2(K)=2 if and only if q(K)=1.

On the other hand, using Tables ? and ?, we check that E_K=\langle-1, \varepsilon,\varepsilon,\varepsilon\text{ or } \sqrt{\varepsilon} \rangle according to whether [r\equiv s\equiv5\pmod8, q\equiv7\pmod8 \text{ and } \left(\dfrac{q}{r}\right)=\left(\dfrac{r}{s}\right)=\left(\dfrac{s}{q}\right)=-1] or not. Hence, h_2(K)=4 or 2, more precisely,

Which justifies the forms listed from B)-7) to B)-9).

\bullet Finally let d be an even positive square-free integer co-prime with q, i.e., d=2d', where d' is an odd positive square-free integer (co-prime with q). Notice that if h_2(K)=2, then the fact that the extension K/\mathbb{Q}(2qd') is ramified, implies that h_2(2qd')\in\{1,2\}. It follows, by Lemma 1.2, that the integer d' is in one of the following sub-cases: [d'=1], [d'=p \text{ is a prime number with } p\equiv5\pmod8 \text{ or } \left(\dfrac{p}{q}\right)=-1] or [d'=q' \;\text{ is a prime numer with } q'\equiv3\pmod8].

We note that the sub-case d'=1 is eliminated, as it implies, by Lemma 1.3, that h_2(K)=1. So let us assume that d'=p is a prime number with p\equiv5\pmod8 or \left(\dfrac{p}{q}\right)=-1. If p\equiv5\pmod8, then h_2(2p)=h_2(2pq)=2 (cf. Lemma 1.2); and on the other hand, we have E_K=\langle{-1,\varepsilon_q,\varepsilon_{2p}, \sqrt{\varepsilon_{q}\varepsilon_{2pq}}\rangle} (cf. Table ?), i.e., q(K)=2. Thus, by Kuroda’s class number formula, we have h_2(K)=2. This justifies the form C).

Now let p\equiv1\pmod8 and \left(\dfrac{p}{q}\right)=-1. So we have q(K)\geq 2 (In fact \sqrt{\varepsilon_{2p}\varepsilon_{q}} or \sqrt{\varepsilon_{q}\varepsilon_{2pq}} \in K). As Lemma 1.2 implies that h_2(2pq)=2 and h_2(2p)\geq 4, we deduce that h_2(K)\geq 4.

To complete our proof let us consider that last sub-case, i.e., d'=q' is a prime number such that q'\equiv3\pmod8. In this sub-case we have E_K=\langle{-1,\varepsilon_q,\sqrt{\varepsilon_{q}\varepsilon_{2q'}}, \sqrt{\varepsilon_{q}\varepsilon_{2qq'}}\rangle} (cf. Table ?) and h_2(2qq')=2 (cf. Lemma 1.2), which gives h_2(K)=2. This justifies the form D).

Hence, putting all of the above together, we obtain the complete classification stated in the theorem. This completes the proof.

Notice that if h_2(K)=2, then by Kuroda’s class number formula and Lemma 1.1, we deduce that h_2(2qd)\leq 8; and so by genus theory we conclude that d has at most three distinct prime divisors. So we have the following cases to investigate:

1. Let d=r be a prime number that is co-prime with 2q. Assume that h_2(K)=2. If r\equiv3\pmod4, the quadratic extension K/\mathbb{Q}(2qr) is ramified, and so h_2(2qr)=1 or 2. As there is no real quadratic field of this form that has an odd class number (cf. Lemma 1.1), we have h_2(2qr)=2, which is equivalent to q or r \equiv3\pmod8 (cf. Lemma 1.2). Conversely, if r\equiv3\pmod 4 and q or r \equiv3\pmod8, then by using Wada’s algorithm (cf. Page ?), we check that E_K=\langle{-1,\varepsilon_r, \sqrt{\varepsilon_{r}\varepsilon_{2q}}, \sqrt{\varepsilon_{2q}\varepsilon_{2qr}}\rangle} (cf. Table ? and Lemma 5.2), i.e., q(K)=4. It follows, by Kuroda’s class number formula, that h_2(K)=2.

   Now if r\equiv1\pmod4, then K/\mathbb{Q}(2qr) is an unramified quadratic extension. Therefore h_2(K)=2 if and only if h_2(2qr)=4. This justifies the form A).

2. Let d=rs for some distinct prime numbers r and s coprime with 2q. Put k=\mathbb{Q}(\sqrt{2qrs}) and assume that h_2(K)=2. Notice that if K/k is ramified, then h_2(k)\in \{1,2\}, which is impossible according to Lemmas 1.1 and 1.2. Therefore, K/k is unramified, and so necessarily rs\equiv1\pmod4. Yields that we have two cases to investigate:

   229 Assume that r\equiv s\equiv 3\pmod4. According to [BS95, Theorem 1] we have k^{(1)}=k^{(2)}, and so h_2(K)=\frac{h_2(k)}{2}. Therefore, h_2(K)=2 if and only if h_2(k)=4. On the other hand, the quadratic fields k=\mathbb{Q}(\sqrt{2qrs}) such that h_2(k)=4 can be deduced from [BS25I, Form c_i, for i\in \{1,...,3\}, page 11]. In fact, we have h_2(k)=4 if and only if, after a suitable permutation of q, r and s, we have r\equiv3\pmod8, q\equiv7\pmod8, \left(\dfrac{r}{s}\right)=1 and one of the conditions in Table ?.

   1.31.6

   The paper has a table here, which this page does not show.

   229 Now assume that r\equiv s \equiv 1\pmod4. Under the hypothesis h_2(K)=2, the 2-class field tower of k is abelian, i.e., k^{(1)}= k^{(2)}. Hence, using the classification in [BS95, Form c_i, for i\in \{19,...,22,25,26\}, page 176], by interpreting the graphs corresponding to the abelian case to those of type 6, we have :

   The paper has a table here, which this page does not show.

   Conversely, every row of Table ? guarantees that h_2(k)=4. Hence, to obtain h_2(K)=2, it remains to verify that k^{(1)}=k^{(2)}. Therefore, by [BLS98, Theorem 2], apart from the case c_{19}, the 2-class tower of k is abelian in all the remaining cases.

   This justifies the form B).

3. Let d=rst, where r, s and t are distinct prime numbers co-prime with 2q. By genus theory, we obtain h_2(rst)\geq 2 and h_2(2qrst)\geq 8. It follows that h_2(K)\geq 4, which forces the rejection of this case.

4. Let d=2d', for an odd positive integer d' coprime with q. Assume that h_2(K)=2. As K/\mathbb{Q}(\sqrt{qd'}) and K/\mathbb{Q}(\sqrt{2d'}) ramifies, we have h_2(\mathbb{Q}(\sqrt{qd'})), h_2(\mathbb{Q}(\sqrt{2d'})) \in \{1,2\}\}. It follows, by Lemmas 1.1 and 1.2, that we have:

   d' = \begin{cases} 1, \;\text{ or } \\ q_1, \text{ with } q_1\equiv3\pmod 4, \text{ or } \\ p, \text{ with } p\equiv5\pmod 8 \text{ or } [ p\equiv1\pmod 8 \text{ and } \left(\frac{2}{p}\right)_4\left(\frac{p}{2}\right)_4=\left(\frac{p}{q}\right)=-1 ], \text{ or } \\ q_1p, \text{ with } \left(\frac{p}{q_1}\right)=-1 \text{ or } [ p\equiv5\pmod8 \text{ and } \left(\frac{p}{q}\right)=-1]. \end{cases}

   Where p\equiv-q_1\equiv1\pmod4 distinct prime numbers. For the converse, if d'=1 then h_2(K)=1 (cf. Lemma 1.3) and so this case is rejected. Moreover, if d'=q_1 for a prime number q_1\equiv3\pmod4 distinct to q, then according to Lemma 1.1 and Kuroda’s class number formula, we have h_2(K)=q(K)/4 and making use of Table ? and Wada’s algorithm (cf. Page ?), we check that E_K=\langle{-1,\varepsilon_{2q},\sqrt{\varepsilon_{2q}\varepsilon_{2q_1}}, \sqrt{\varepsilon_{2q}\varepsilon_{qq_1}}\rangle}, this gives h_2(K)=1. So the second case is also rejected.

   Now let d' takes the third form, i.e., d'=p, such that p\equiv5\pmod 8 or [ p\equiv1\pmod 8 \text{ and } \left(\frac{2}{p}\right)_4\left(\frac{p}{2}\right)_4=\left(\frac{p}{q}\right)=-1 ]. By Kuroda’s class number formula and Lemma 1.2, we have : h_2(K)=\frac{q(K)h_2(2p)h_2(pq)}{4}=q(K). Moreover, by making use of Table ? and Lemma 5.2, we check that if p\equiv5\pmod8 then E_K=\langle{-1,\varepsilon_{2q},\varepsilon_{2p},\sqrt{\varepsilon_{2q}\varepsilon_{pq}} \;\text{or}\; \sqrt{\varepsilon_{pq}}\rangle} according to whether \left(\frac{p}{q}\right)=1 or not, thus h_2(K)=2. But if p\equiv1\pmod 8 \text{ and } \left(\frac{2}{p}\right)_4\left(\frac{p}{2}\right)_4=\left(\frac{p}{q}\right)=-1, we check that E_K=\langle{-1,\varepsilon_{2q},\varepsilon_{2p}, \sqrt{\varepsilon_{2q}\varepsilon_{2p}} \rangle} (notice that accordint to [kuvcera1995parity, Theorem 2] we have N(\varepsilon_{2p})=1). Therefore h_2(K)=2. This justifies the form C).

   Finally let d'=q_1p, with \left(\frac{p}{q_1}\right)=-1 \text{ or } [ p\equiv5\pmod8 \text{ and } \left(\frac{p}{q}\right)=-1]. Then h_2(K)=\frac{q(K)h_2(2pq_1)h_2(pqq_1)}{4}=q(K). We begin by the first case where \left(\frac{p}{q_1}\right)=-1, then if p\equiv1\pmod8 we have E_K=\langle{-1,\varepsilon_{2q },\sqrt{\varepsilon_{2q}\varepsilon_{2pq_1}}, \varepsilon_{pqq_1}\; \text{or}\; \sqrt{\varepsilon_{2q}\varepsilon_{pqq_1}} \rangle} according to whether \left(\frac{p}{q}\right)=-1 or not. But if p\equiv5\pmod8 we have E_K=\langle{-1,\varepsilon_{2q },\varepsilon_{2pq_1},\sqrt{\varepsilon_{2q}\varepsilon_{pqq_1}}\; \text{or}\; \sqrt{\varepsilon_{2pq_1}\varepsilon_{pqq_1}} \rangle} according to whether \left(\frac{p}{q}\right)=1 or not (cf. Tables ? and ?). Therefore, if \left(\frac{p}{q_1}\right)=-1, we have h_2(K)=2 if and only if p\equiv5\pmod8 or [p\equiv1\pmod8 \;\text{and}\; \left(\frac{p}{q}\right)=-1]. At the end if \left(\frac{p}{q_1}\right)=1, p\equiv5\pmod8 \text{ and } \left(\frac{p}{q}\right)=-1, we check that E_K=\langle{-1,\varepsilon_{2q },\varepsilon_{2pq_1}, \sqrt{\varepsilon_{2pq_1}\varepsilon_{pqq_1}} \rangle} (cf. Tables ? and ?), then h_2(K)=2. This justifies the form D).

5. Now let d=qd' for an odd positive integer d' co-prime with q. Then K=\mathbb{Q}(\sqrt{2q},\sqrt{qd'})=\mathbb{Q}(\sqrt{2q},\sqrt{2d'}), and this case has already been treated in the preceding case.

Plugging all of the above investigations together, we complete the proof of our classification.

Let d be a positive square-free integer. We have two main cases to investigate.

229 Assume that \mathrm{gcd}(d,q_1q_2)=1. Put k=\mathbb{Q}(\sqrt{q_1q_2d}). We note that K/k is an unramified quadratic extension. So by Remark 1.6, we have h_2(K)=2 if and only if h_2(k)=4 and k^{(1)}=k^{(2)}. In particular, if h_2(K)=2, then the 2-class group of k is cyclic or it is the Klein 4-group; and so by genus theory, we have one of the following forms for d:

where p\equiv p_i\equiv -q_i\equiv1\pmod4 are distinct prime numbers. Now let us explore each form.

1. Let d=2 (resp. d=p). Since A(k) is cyclic, we deduce from class field theory and Lemma 1.2 that h_2(K)=2 if and only if q_1\equiv q_2\equiv 7\pmod 8 and h(2q_1q_2)=4 (resp. \left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=1 and h_2(pq_1q_2)=4). This justifies the forms A) and B) respectively.

2. Let d=2p. Since K/\mathbb{Q}(\sqrt{d}) ramify, this implies that h_2(2p) = 2, and thus p\equiv5\pmod8 (cf. Lemmas 1.2 and 1.1). According to [BS95, page 175, cases c_{12}-c_{14}], we have k^{(1)}=k^{(2)} and h_2(k)=4 if and only if one of the following conditions holds:

   1. p\equiv5\pmod8,q_1\equiv q_2\equiv3\pmod8 and \left(\frac{p}{q_1}\right)=-\left(\frac{p}{q_2}\right)=-1, or

   2. p\equiv5\pmod8,q_1\equiv7\pmod8, q_2\equiv3\pmod8 and \left(\frac{p}{q_1}\right)=-\left(\frac{p}{q_2}\right)=\left(\frac{q_2}{q_1}\right)=-1.

   This justifies the form C).

3. Let d=q_3. By Theorem 2.1, h_2(K)=2 if and only if we have one of the conditions listed in D).

4. Let d=2q_3. By Lemma 1.2, h_2(K)=2 if and only if we have one of the conditions listed in E).

5. Let d=p_1p_2. According to [BS95, page 175, cases c_{12}-c_{14}] we have h_2(p_1p_2q_1q_2)=4 and k^{(1)}=k^{(2)} if and only if we have one of the following conditions:

   1. \left(\frac{p_1}{p_2}\right)=\left(\frac{p_1}{q_1}\right)=\left(\frac{p_1}{q_2}\right)=\left(\frac{p_1}{q_1}\right)=-\left(\frac{p_1}{q_2}\right)=-1,

   2. \left(\frac{p_1}{p_2}\right)=-\left(\frac{p_1}{q_1}\right)=\left(\frac{p_1}{q_2}\right)=\left(\frac{p_1}{q_1}\right)=-\left(\frac{p_1}{q_2}\right)=\left(\frac{q_1}{q_2}\right)=-1.

   This justifies the form F).

6. Let d=q_3q_4. We note that according [BS95, Theorem 1] the 2-class field of k is abelian. Moreover, using [BS25I, page 11, c_1-c_3] we deduce that h_2(q_1q_2q_3q_4)=4 if and only if one of the following conditions holds:

   1. \left(\frac{q_1}{q_2}\right)=\left(\frac{q_1}{q_4}\right)=\left(\frac{q_4}{q_3}\right)=-\left(\frac{q_1}{q_3}\right)=-\left(\frac{q_2}{q_4}\right)=1, or

   2. \left(\frac{q_1}{q_2}\right)=\left(\frac{q_1}{q_4}\right)=\left(\frac{q_4}{q_3}\right)=\left(\frac{q_1}{q_3}\right)=-\left(\frac{q_2}{q_3}\right)=-\left(\frac{q_2}{q_4}\right)=1.

   This justifies the form G).

229 Now assume that \mathrm{gcd}(d,q_1q_2)=q_1. Put k=\mathbb{Q}(\sqrt{q_2d'}), where d'=\frac {d}{q_1}. As K/k ramify, the equality h_2(K)=2 implies that h_2(k)\in\{1,2\}; and thus Lemmas 1.1 and 1.2 imply that:

where q\equiv q_3\equiv -p\equiv3\pmod4 are prime numbers satisfying some extra conditions as below. For the converse, let us explore these forms of d'. Firstly, notice that the forms d'=1, 2 and q are eliminated. In fact, each of these cases gives h_2(K)=1 (cf. Lemma 1.3). Now let us investigate each of the remaining forms.

1. d'=2q_3, for some prime number q_3\equiv3\pmod8 distinct from q_1 and q_2. Using Table ? and Wada’s algorithm (cf. Page ?) we conclude that E_K= \langle{-1,\varepsilon_{q_1q_2},\varepsilon_{2q_1q_3},\eta\rangle}, where

   \eta \in \{\sqrt{\varepsilon_{2q_1q_3}\varepsilon_{2q_2q_3}},\sqrt{\varepsilon_{q_1q_2}\varepsilon_{2q_2q_3}},\sqrt{\varepsilon_{q_1q_2}\varepsilon_{2q_2q_3}\varepsilon_{2q_1q_3}}\}.

   Therefore h_2(K)=2. This justifies the form H).

2. Let d'=p, for some prime number p\equiv1\pmod4 such that p\equiv5\pmod8 or \left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=-1. From Table ? we get E_K= \langle{-1,\varepsilon_{q_1q_2},\varepsilon_{pq_1}, \sqrt{\varepsilon_{pq_1}\varepsilon_{pq_2}}\;or\;\sqrt{\varepsilon_{pq_1}\varepsilon_{pq_2}\varepsilon_{q_1q_2}}\rangle}. So by Kuroda’s class number formula and Lemma 1.2, we have h_2(K)=2. This justifies the form I).

3. Now let d'=2p, for some prime number p\equiv1\pmod4 such that p\equiv5\pmod8 or \left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=-1. Notice that q(K)=4 or 2 according to whether p\equiv5\pmod8 and \left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=1 or not. In fact using Table ? and Wada’s algorithm (cf. Page ?) we check that:

   The paper has a table here, which this page does not show.

   Therefore by applying Kuroda’s class number formula and Lemma 1.2, we deduce that h_2(K)=2 if and only if

   1. \left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=-1, or

   2. \left(\frac{p}{q_1}\right)=-\left(\frac{p}{q_2}\right)=1 and p\equiv5\pmod8.

   This justifies the form J). This completes the proof of our theorem.

Let K=\mathbb{Q}(\sqrt{2},\sqrt{d}), with d an odd positive square-free integer. Our proof consists of two parts, according to the equivalence class of d \pmod4.

229 Assume that d\equiv3\pmod4. So K/\mathbb{Q}(\sqrt{2d}) ramify and thus the conditions h_2(K)=2, implies that we have h_2(2d)=1 or 2. Conversely, notice that if h_2(2d) =1, then d=q\equiv3\pmod4 is a prime number (cf. Lemma 1.1); and this implies that the class number of K is odd (cf. Lemma 1.3). So this case eliminated.

Now let h_2(2d)=2. So Lemma 1.2 implies that d=pq, where p\equiv-q\equiv1\pmod4 are prime numbers such that p\equiv5\pmod8 or \left(\frac{p}{q}\right)=-1. Thus, we have h(pq)=h(2pq)=2 (cf. Lemma 1.2) and q(K)=2 (In fact E_K=\langle{-1,\varepsilon_{2},\varepsilon_{pq}, \sqrt{\varepsilon_{pq}\varepsilon_{2pq}}\rangle} by Wada’s algorithm and Table ?). Consequently, Kuroda’s class number formula gives h_2(K)=2. This justifies the form A).

229 Now assume that d\equiv1\pmod4. Thus, K/\mathbb{Q}(\sqrt{2d}) is an unramified quadratic extension. Notice that if h_2(K)=2, then h_2(2d)= 4 and so h_2(d)q(K)=2. This implies that h_2(d)=1 or 2. So we have two sub-cases to investigate.

1. Assume that h_2(d)=1. So, by Lemma 1.1, we have d=p, where p \equiv1\pmod4 is a prime number, or d=q_1q_2, where q_1\equiv q_2\equiv3\pmod4 are distinct prime numbers. Moreover, we have:

   1. If d=p, then [kuvcera1995parity, Theorem 2] implies that, p\equiv1\pmod 8 and \left(\frac{2}{p}\right)_4=(-1)^{\frac{p-1}{8}}. Therefore, by class field theory, h_2(K)=2 if and only if h_2(2p)=4. This justifies the form B).

   2. Now if d=q_1q_2, then, by class field theory and Lemma 1.2, h_2(K)=2 if and only if q_1\equiv q_2\equiv7\pmod8 and h_2(2q_1q_2)=4. This justifies the form C).

2. Assume that h_2(d)=2. By Lemma 1.2, we have one of the following cases:

   1. If d takes the first form, i.e., d=p_1p_2 \text{ with } p_1\equiv p_2\equiv1\pmod 4 \text{ and } \left(\frac{p_1}{p_2}\right)=-1, then according to [kuvcera1995parity, Theorem 1] we have N_{\mathbb{Q}(\sqrt{p_1p_2})/\mathbb{Q}}(\varepsilon_{p_1p_2})=-1 and using [AzZektaous, Theorem 2.2] we deduce that q(K)=2 or 1, according to whether one of the units \{ {\varepsilon_{2}\varepsilon_{p_1p_2}\varepsilon_{2p_1p_2}} , {\varepsilon_{2p_1p_2}}\} is a square in K or not.

      Since by Kuroda’s class number formula, we have h_2(K)=\frac12 q(K)h_2(2p_1p_2), we deduce that h_2(K)=2 if and only if q(K)=1 and h_2(2p_1p_2)=4. This justifies the form D)-1).

   2. Now if d takes the second form, i.e., d=p_1p_2 \text{ with } p_1\equiv p_2\equiv1\pmod 4 \text{ and } [\left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\not=\left(\frac{p_2}{p_1}\right)_4], then according to [kuvcera1995parity, Theorem 1] we have N_{\mathbb{Q}(\sqrt{p_1p_2})/\mathbb{Q}}(\varepsilon_{p_1p_2})=-1 and using [AzZektaous, Theorem 2.2] we deduce that q(K)=2 or 1, according to whether one of the units \{ {\varepsilon_{p_1p_2}\varepsilon_{2p_1p_2}} , {\varepsilon_{2p_1p_2}}\} is a square in K or not. Thus, Kuroda’s class number formula gives h_2(K)=\frac12 q(K)h_2(2p_1p_2). Therefore, h_2(K)=2 if and only if q(K)=1 and h_2(2p_1p_2)=4. This justifies the form D)-2).

   3. Now if d takes the third form, i.e., d=pq_1q_2 with q_1\equiv q_2\equiv-p\equiv3\pmod 4 and (\left(\frac{p}{q_1}\right)=-1 or \left(\frac{p}{q_2}\right)=-1), then according to [BS95, Page 175 : c_{12}-c_{14}], we have Table ? that characterizes when do we have exactly h(2pq_1q_2)=4.

      1.31.6

      The paper has a table here, which this page does not show.

      On the other hand, assuming that d=2pq_1q_2, where p, q_1 and q_2 are prime numbers satisfying one of the conditions in Table ?, we make use of Table ? and Wada’s method to check that we have q(K)=1. Thus under each of these conditions we have h_2(K)=2. This justifies the form E).

   Combining all of the above investigations, we complete the proof of our classification in the above theorem.

As N(\varepsilon_{pp_1q_1q_2}) = N(\varepsilon_{p_1q_1q_2}) = 1 and N(\varepsilon_p) = -1 we need to determine the the forms of:

Here we give the proof of the first item in details, and the prove of the second item is analogous. Let us assume that \left(\frac{p}{p_1}\right)=\left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=\left(\frac{p_1}{q_1}\right)=-\left(\frac{p_1}{q_2}\right)=-1. So we have three main cases to investigate:

229 Assume that x,y, a and b are integers. Then x^2-p_1q_1q_2y^2=a^2-pp_1q_1q_2b^2=1. Then by taking into account the unique factorization in \mathbb{Z} and using Lemma 5.1 we have one of the following systems:

where y_1 and y_2 are two integers such that y=2y_1 y_2.

1. The first system can not occur, since it implies that :

   \left(\frac{2}{p_1}\right)=\left(\frac{\mp 2}{p_1}\right)=\left(\frac{x\pm1\mp 2}{p_1}\right)=\left(\frac{x\mp1}{p_1}\right)=\left(\frac{2q_1q_2}{p_1}\right)=-\left(\frac{2}{p_1}\right).

2. Similarly the second system implies that:

   \left(\frac{2}{p_1}\right)=\left(\frac{\pm 2}{p_1}\right)=\left(\frac{x\mp1\pm 2}{p_1}\right)=\left(\frac{x\pm1}{p_1}\right)=\left(\frac{2q_1}{p_1}\right)=-\left(\frac{2}{p_1}\right).

From the third system we get 2q_2(x+1) or 2q_2(x-1) is a square in \mathbb{N} according to whether \left(\frac{q_1}{q_2}\right)=-1 or not. Then \sqrt{\varepsilon_{p_1q_1q_2}}=y_1\sqrt{q_2}+y_2\sqrt{p_1q_1}. Similarly we have the following possible systems :

where b_1 and b_2 are two integers such that b=2b_1 b_2.

1. The first system implies that:

   \left(\frac{2}{q_2}\right)=\left(\frac{a+1 }{q_2}\right)=\left(\frac{a-1+ 2}{q_2}\right)=\left(\frac{2pp_1}{q_2}\right)=-\left(\frac{2}{q_2}\right), \;\;\text{or}
   \left(\frac{-2}{q_1}\right)=\left(\frac{a-1 }{q_1}\right)=\left(\frac{a+1- 2}{q_1}\right)=\left(\frac{2pp_1}{q_1}\right)=\left(\frac{2}{q_1}\right)

   which is a contradiction.

2. By the same way we prove that the third system can bot occur. Finally it remains the second system and from which we get 2pq_1(x+1) or 2pq_1(x-1) is square according to whether \left(\frac{q_1}{q_2}\right)=-1 or not. Therefore \sqrt{\varepsilon_{pp_1q_1q_2}}=b_1\sqrt{pq_1}+b_2\sqrt{p_1q_2}. Then by using Wada’s algorithm we conclude that q(\mathbb{Q}(\sqrt{p},\sqrt{p_1q_1q_2}))=1.

229 Now we assume that x and y are integers, a and b are semi integers such that a=\alpha/2 and b=\beta/2 for some integers \alpha and \beta. According to the previous case we have \sqrt{\varepsilon_{p_1q_1q_2}}=y_1\sqrt{q_2}+y_2\sqrt{p_1q_1}. For the second square we proceed as before. Then from \alpha^2-4=pp_1q_1q_2\beta^2 and according to Lemma 5.1 we have the following possible systems:

where \beta_1 and \beta_2 are integers such that \beta=\beta_1\beta_2.

1. The first system can not occur since it implies:

   1=\left(\frac{\pm4}{p_1}\right)=\left(\frac{\alpha\mp2\pm4}{p_1}\right)=\left(\frac{\alpha\pm2}{p_1}\right)=\left(\frac{p}{p_1}\right)=-1.

2. Similarly for the second system:

   1=\left(\frac{\pm4}{p}\right)=\left(\frac{\alpha\mp2\pm4}{p}\right)=\left(\frac{\alpha\pm2}{p}\right)=\left(\frac{p_1}{p}\right)=-1.

3. Similarly for the third system:

   1=\left(\frac{\pm4}{p}\right)=\left(\frac{\alpha\mp2\pm4}{p}\right)=\left(\frac{\alpha\pm2}{p}\right)=\left(\frac{q_1}{p}\right)=-1.

4. Similarly for the fourth system:

   1=\left(\frac{\pm4}{p}\right)=\left(\frac{\alpha\mp2\pm4}{p}\right)=\left(\frac{\alpha\pm2}{p}\right)=\left(\frac{q_2}{p}\right)=-1.

5. Similarly for the fifth system:

   1=\left(\frac{4}{q_2}\right)=\left(\frac{\alpha-2+4}{q_2}\right)=\left(\frac{\alpha+2}{q_2}\right)=\left(\frac{pp_1}{q_2}\right)=-1,\;\text{or}
   -1=\left(\frac{-4}{q_1}\right)=\left(\frac{\alpha+2-4}{q_1}\right)=\left(\frac{\alpha-2}{q_1}\right)=\left(\frac{pp_1}{q_1}\right)=1.

6. The same thing for the seventh system:

   1=\left(\frac{\pm4}{p_1}\right)=\left(\frac{\alpha\mp2\pm4}{p_1}\right)=\left(\frac{\alpha\pm2}{p_1}\right)=\left(\frac{pq_2}{p_1}\right)=-1.

Then it remains the sixth system which implies that pq_1(x+2) or pq_1(x-2) is a square in \mathbb{N} according to whether \left(\frac{q_1}{q_2}\right)=-1 or not. Therefore \sqrt{\varepsilon_{pp_1q_1q_2}}=\frac{1}{2}(\beta_1\sqrt{pq_1}+\beta_2\sqrt{p_1q_2}) and so q(\mathbb{Q}(\sqrt{p},\sqrt{p_1q_1q_2}))=1.

229 Now if that x, y, a and b are semi-integers. Then \sqrt{\varepsilon_{pp_1q_1q_2}}=\frac{1}{4}(\beta_1\sqrt{pq_1}+\beta_2\sqrt{p_1q_2}). Moreover we have z^2-4=t^2p_1q_1q_2, where z and t are integers such that x=z/2 and y=t/2, then Lemma 5.1 implies the following possible systems:

where t_1 and t_2 are two integers such that t=t_1 t_2. Using the same argument as in the previous two cases we check that q_2(x+2) or q_2(x-2) is square in \mathbb{N} according to whether \left(\frac{q_1}{q_2}\right)=-1 or not. Thus \sqrt{\varepsilon_{p_1q_1q_2}}=\frac{1}{2}(t_1\sqrt{q_2}+t_2\sqrt{p_1q_1}) and q(\mathbb{Q}(\sqrt{p},\sqrt{p_1q_1q_2}))=1.

229 Finally if x and y are semi integers, a and b are integers. Then from the previous sub-cases we have \sqrt{\varepsilon_{p_1q_1q_2}}=\frac{1}{2}(t_1\sqrt{q_2}+t_2\sqrt{p_1q_1}) and \sqrt{\varepsilon_{pp_1q_1q_2}}=b_1\sqrt{pq_1}+b_2\sqrt{p_1q_2}. Then q(\mathbb{Q}(\sqrt{p},\sqrt{p_1q_1q_2}))=1.

This completes the proof.

Let d be a positive square-free integer co-prime with p. Notice that the extension K/k, with k=\mathbb{Q}(pd), is an unramified quadratic extension. Assume that h_2(K)=2. Then h_2(pd)=4 and by Kuroda’s class number formula, we deduce that h_2(d)=1 (in this case q(K)=2) or h_2(d)=2 (in this case q(K)=1). So we have two main cases to investigate.

229 Assume that h_2(d)=1. By Lemma 1.1, d\in \{2,p_1,q,2q, q_1q_2\}, where p_1\equiv-q\equiv-q_1\equiv-q_2\equiv1\pmod4 are distinct prime numbers. In particular, we have:

1. Letting d=2, then as in the proof of Theorem 2.4, we deduce that h_2(K)=2 if and only if p\equiv1\pmod8, \left(\frac{2}{p}\right)_4=(-1)^{\frac{p-1}{8}} and h_2(2p)=4. This justifies the form A).

2. Letting d=p_1, with p_1\equiv1\pmod4 a prime number. Thus, by class field theory, h_2(K)=2 exactly when h_2(pp_1)=4. This justifies the form B).

3. Letting d=\eta q with q\equiv3\pmod4 prime number, we deduce from the condition h_2(\eta pq)=4 that we have p\equiv1\pmod8 and \left(\frac{p}{q}\right)=1 (cf. Lemma 1.2). Thus, in this case, by class field theory h_2(K)=2 if and only if p\equiv1\pmod8, \left(\frac{p}{q}\right)=1 and h_2(\eta pq)=4. This justifies the form C).

4. d=q_1q_2 where q_1\equiv q_2\equiv3\pmod4 are distinct prime numbers. Notice that the condition h_2(pq_1q_2)=4 implies that \left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=1 (cf. Lemma 1.2). So by class field theory we have h_2(K)=2 if and only of h_2(pq_1q_2)=4. This justifies the form D).

229 Now assume that h_2(d)=2. By Lemma 1.2, d\in \{p_1q,2p_1q,2q_1q_2,p_1q_1q_2, 2p_1, p_1p_2\}, where p_1\equiv p_2\equiv-q_1\equiv q_2 \equiv 1\pmod4 are prime numbers that satisfy extra condition as in each case below.

1. Let d=p_1q, where p_1\equiv-q\equiv1\pmod4 are prime numbers such that p_1\equiv5\pmod 8 or \left(\frac{p_1}{q}\right)=-1. Thus, h_2(pp_1q)=4, then (by letting p=r and p_1=s) the primes p, p_1 and q satisfy the conditions c_{15}-c_{20} in Table Proof (cf. [BS95, page 177-178]).

   Conversely using Tables ?, ? and Wada’s algorithm (cf. Page ?) we check that E_K=\langle{-1,\varepsilon_{p},\varepsilon_{p_1q} , \varepsilon_{pp_1q}\rangle} for all such cases of Table Proof except the case c_{17} which gives q(K)=2 (In fact, in this case we have E_K=\langle{-1,\varepsilon_{p},\varepsilon_{p_1q} , \sqrt{\varepsilon_{p_1q}\varepsilon_{pp_1q}}\rangle}). Hence, h_2(K)=2 if and only if we have one of the conditions listed from E)-1) to E)-3).

2. Let d=2p_1q, where p_1\equiv-q\equiv1\pmod4 are prime numbers such that p_1\equiv5\pmod 8 or \left(\frac{p_1}{q}\right)=-1. Notice that if h_2(2pp_1q)=4, then the primes p, p_1 and q satisfy one of the conditions c_{19}-c_{22}, c_{25}- c_{26} in Table ? (cf. [BS95, Page 176]). Conversely by invoking Table ? in all previous cases except case c_{19} we check that E_K=\langle{-1,\varepsilon_{p},\varepsilon_{2p_1q} , \varepsilon_{2pp_1q}\rangle}, i.e., q(K)=1, which gives h_2(K)=2. For the case c_{19} we get E_K=\langle{-1,\varepsilon_{p},\varepsilon_{2p_1q} , \sqrt{\varepsilon_{2pp_1q}\varepsilon_{2p_1q}}\rangle}, then h_2(K)=4 and so this case is rejected. This justifies the forms listed from F)-1) to F)-3).

3. Let d=2q_1q_2, with q_1\equiv q_2\equiv3\pmod 4 are distinct prime numbers such that q_1 or q_2\equiv3\pmod8. We note that the primes satisfying h_2(2pq_1q_2)=4 are exactly those satisfying one of the conditions in Table ?. Conversely, by making use of Tables ? and ? and Wada’s algorithm we check that in all these cases we have q(K)=1, which means that h_2(K)=2. This justifies the forms G)-1) and G)-2).

4. Let d=p_1q_1q_2, where p_1\equiv-q_1\equiv-q_2\equiv1\pmod 4 are distinct prime numbers such that \left(\frac{p_1}{q_1}\right)=-1 or \left(\frac{p_1}{q_2}\right)=-1. Notice that K/\mathbb{Q}(\sqrt{pp_1q_1q_2}) is an unramified quadratic extension. So h_2(K)=2 implies that h_2(pp_1q_1q_2)=4. Moreover, if h_2(pp_1q_1q_2)=4, then according to [BS95, page 175, cases c_{12}-c_{14}], the primes p, p_1, q_1 and q_2 satisfy one of the following conditions:

   1. \left(\frac{p}{p_1}\right)=\left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=\left(\frac{p_1}{q_1}\right)=-\left(\frac{p_1}{q_2}\right)=-1,

   2. \left(\frac{p}{p_1}\right)=-\left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=\left(\frac{p_1}{q_1}\right)=-\left(\frac{p_1}{q_2}\right)=\left(\frac{q_1}{q_2}\right)=-1.

   Conversely, we check that if one of the conditions \mathrm{C}1) and \mathrm{C}2) holds, then making use of Lemma 2.5, we check that q(K)=1. Since 2h_2(p_1q_1q_2)=h_2(pp_1q_1q_2)=4h_2(p)=4, it follows that h_2(K)=2. This justifies the forms H)-1) and H)-2).

5. Let d=2p_1, where p_1\equiv5\pmod8 is a prime number. According to [kuvcera1995parity, Theorem 2] we have N(\varepsilon_{2p_1})=-1 and by [AzZektaous, Theorem 2.1] we check that q(K)=2 or 1 according to whether one of the units \{\varepsilon_{p}\varepsilon_{2p_1}\varepsilon_{2pp_1}, \varepsilon_{2pp_1}\} is a square in K or not. Furthermore h_2(K)=2 exactly when p_1\equiv5\pmod8, h_2(2pp_1)=4 and none of the units \{\varepsilon_{p}\varepsilon_{2p_1}\varepsilon_{2pp_1}, \varepsilon_{2pp_1}\} is a square in K. This justifies the the first item in form I).

6. Let d=2p_1, where p_1\equiv1\pmod8 is a prime number such that \left(\frac{2}{p_1}\right)_4\neq \left(\frac{p_1}{2}\right)_4. According to [kuvcera1995parity, Theorem 2] we have N(\varepsilon_{2p_1})=1 and by [AzZektaous, Theorem 2.1] we check that q(K)=2 or 1. according to whether N(\varepsilon_{2pp_1})=1 or not. Furthermore h_2(K)=2 exactly when p_1\equiv1\pmod8, h_2(2pp_1)=4, \left(\frac{2}{p_1}\right)_4\neq \left(\frac{p_1}{2}\right)_4 and N(\varepsilon_{2pp_1})=-1. This justifies the the second item in form I).

7. Let d=p_1p_2, where p_1\equiv p_2\equiv1\pmod4 distinct prime number such that \left(\frac{p_1}{p_2}\right)=-1 (cf. Lemma 1.2). Then by [kuvcera1995parity, Theorem 1] we have N(\varepsilon_{p_1p_2})=-1, thus using Wada’s method we conclude that q(K)=2 or 1 according to whether one of the units \{\varepsilon_{p}\varepsilon_{p_1p_2}\varepsilon_{pp_1p_2}, \varepsilon_{pp_1p_2}\} is a square in K or not. Hence, h_2(K)=2 exactly when \left(\frac{p_1}{p_2}\right)=-1, h_2(pp_1p_2)=4 and none of the above units is a square in K. This justifies the first item in forms J).

8. Let d=p_1p_2, where p_1\equiv p_2\equiv1\pmod4 distinct prime number such that \left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\neq\left(\frac{p_2}{p_1}\right)_4. Then by [kuvcera1995parity, Theorem 1] we have N(\varepsilon_{p_1p_2})=1, moreover using Wada’s method we conclude that q(K)=1 if and only if none of the units \{\varepsilon_{p_1p_2}\varepsilon_{pp_1p_2},\varepsilon_{p}\varepsilon_{p_1p_2}\varepsilon_{pp_1p_2}, \varepsilon_{p}\varepsilon_{pp_1p_2}\} is a square in K. Hence, h_2(K)=2 exactly when \left(\frac{p_1}{p_2}\right)=1, \left(\frac{p_1}{p_2}\right)_4\neq\left(\frac{p_2}{p_1}\right)_4, h_2(pp_1p_2)=4 and none of the above units is a square in K. This justifies the second item in forms J).

This completes the proof of our result.

Let K=\mathbb{Q}(\sqrt{d_1},\sqrt{d_2}), with d_1 and d_2 distinct positive square-free integers such that h_2(d_1)=h_2(d_2)=h_2(d_3)=2, where d_3:=\mathrm{sqf}(d_1d_2). Suppose also that h_2(K)=2. It follows that the integers d_1 and d_2 must appear among the possible forms listed in Table ? presented in the introduction. For the converse, we treat the cases independently as in what follows.

1. Let d_1=pq_1 and d_2=2q_1q_2 where p\equiv -q_1 \equiv -q_2 \equiv1\pmod4 are distinct prime numbers such that [\left(\frac{p}{q_1}\right)=\left(\frac{p}{q_2}\right)=-1 or p\equiv5\pmod8] and [q_1\equiv 3\pmod 8 or q_2\equiv 3\pmod 8]. Then according to Table ? and Wada’s method we have q(K)\geq 2. In fact we have :

   The paper has a table here, which this page does not show.

   Where the notation [m,n] is explained in the introduction (cf. Notation 1.5). Furthermore such field must me rejected.

2. Let d_1=pq_1q_2 and d_2=2q_1q_2 where p\equiv -q_1 \equiv -q_2 \equiv1\pmod4 are distinct prime numbers such that p\equiv5\pmod8, [\left(\frac{p}{q_1}\right)=-1 or \left(\frac{p}{q_2}\right)=-1] and [q_1\equiv 3\pmod 8 or q_2\equiv 3\pmod 8]. Using Tables ? and ? together and Wada’s method we check that q(K)=1 or 2. More precisely, we have Table ?.

   The paper has a table here, which this page does not show.

   It is worth noting that, by virtue of [kuvcera1995parity, Theorem 2], we have N_{\mathbb{Q}(\sqrt{2p})/\mathbb{Q}}(\varepsilon_{2p})=-1. Moreover, as is evident from the tables cited above, the first three cases appearing in Table ? are realised if and only if

   1. p\equiv5\pmod8, q_1\equiv 3\pmod 8 and q_2\equiv 7\pmod 8 and \left(\frac{p} {q_1}\right)=-\left(\frac{p}{q_2}\right)=-1,

   2. p\equiv5\pmod8, q_1\equiv 7\pmod 8 and q_2\equiv 3\pmod 8 and \left(\frac{p} {q_1}\right)=-\left(\frac{p}{q_2}\right)=1,

   3. p\equiv5\pmod8, q_1\equiv q_2\equiv 3\pmod 8 and \left(\frac{p} {q_1}\right)=\left(\frac{p}{q_2}\right)=-1.

   Which gives the exact cases when q(K)=2 using Wada’s method. Consequently, by negating these conditions and then intersecting the result with the hypotheses of this sub-case, we derive the precise conditions under which q(K)=1, thereby justifying case 1) of the theorem.

3. Let d_1=pq_1q_2 and d_2=2q_1q_2 where p\equiv1\pmod8, \left(\frac{2}{p}\right)_4\neq \left(\frac{p}{2}\right)_4, [\left(\frac{p}{q_1}\right)=-1 or \left(\frac{p}{q_2}\right)=-1] and [q_1\equiv 3\pmod 8 or q_2\equiv 3\pmod 8]. According to [kuvcera1995parity, Theorem 2], we have N_{\mathbb{Q}(\sqrt{2p})/\mathbb{Q}}(\varepsilon_{2p})=1, then \sqrt{\varepsilon_{2p}}=y_1\sqrt{2}+y_2\sqrt{p}, for some rational numbers y_1 and y_2, moreover using Tables ? and ? in the above conditions we check that a least one of the following square roots \sqrt{\varepsilon_{2p}\varepsilon_{2q_1q_2}}, \sqrt{\varepsilon_{2p}\varepsilon_{pq_1q_2}}, \sqrt{\varepsilon_{2p}\varepsilon_{2q_1q_2}\varepsilon_{pq_1q_2}} or \sqrt{\varepsilon_{2q_1q_2}\varepsilon_{pq_1q_2}} belongs to E_K, more precisely we have

   The paper has a table here, which this page does not show.

   Therefore q(K)\geq 2, and such case must be rejected.

4. 4)-5) Let (d_1,d_2)=(p_1q,p_2q) or (2p_1q,p_1p_2) where p_1\equiv p_2 \equiv -q \equiv1\pmod4 are distinct prime numbers such that [\left(\frac{p_1}{q}\right)=-1 or p_1\equiv 5\pmod 8], [\left(\frac{p_2}{q}\right)=-1 or p_2\equiv 5\pmod 8] and \left(\frac{p_1}{p_2}\right)=-1. Recalling [kuvcera1995parity, Theorem 1], we have N_{\mathbb{Q}(\sqrt{p_1p_2})/\mathbb{Q}}(\varepsilon_{p_1p_2})=-1. With the aid of Table ? and Wada’s method, we then conclude that q(K) is either 2 or 1, the distinction resting on the fulfilment or otherwise of one of the conditions set forth below:

   1. p_1\equiv p_2\equiv1\pmod8 and \left(\frac{p_1}{q}\right)=\left(\frac{p_2}{q}\right)=-1,

   2. p_1\equiv p_2\equiv5\pmod8 and \left(\frac{p_1}{q}\right)=\left(\frac{p_2}{q}\right).

   An entirely analogous procedure to that of the previous paragraph furnishes the conditions that establish cases 2)–i) and 2)–ii).

5. 6) Let d_1=p_1q and d_2=p_2q where [\left(\frac{p_1}{q}\right)=-1 or p_1\equiv 5\pmod 8], [\left(\frac{p_2}{q}\right)=-1 or p_2\equiv 5\pmod 8], \left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\neq \left(\frac{p_2}{p_1}\right)_4. According to [kuvcera1995parity, Theorem 1], we have N_{\mathbb{Q}(\sqrt{p_1p_2})/\mathbb{Q}}(\varepsilon_{p_1p_2})=1, then \sqrt{\varepsilon_{p_1p_2}}=y_1\sqrt{p_1}+y_2\sqrt{p_2}, for some rational numbers y_1 and y_2. Combining this with Table ? in this case, we check that q(K)\geq2, in fact at least one of the following square roots \sqrt{\varepsilon_{p_1p_2}\varepsilon_{p_1q}}, \sqrt{\varepsilon_{p_1p_2}\varepsilon_{p_2q}}, \sqrt{\varepsilon_{p_1q}\varepsilon_{p_2q}} or \sqrt{\varepsilon_{p_1q}\varepsilon_{p_2q}\varepsilon_{p_1p_2}} is in E_K. This such case must be rejected.

6. 7) Let d_1=2p_1q and d_2=2p_2q where [\left(\frac{p_1}{q}\right)=-1 or p_1\equiv 5\pmod 8], [\left(\frac{p_2}{q}\right)=-1 or p_2\equiv 5\pmod 8], \left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\neq \left(\frac{p_2}{p_1}\right)_4. Using the same argument as in the previous case one can conclude that q(K)=2. In fact E_K=\langle{-1,\varepsilon_{2p_1q},\varepsilon_{2p_2q},\eta \rangle} with \eta=\sqrt{\varepsilon_{p_1p_2}\varepsilon_{2p_1q}}, \sqrt{\varepsilon_{p_1p_2}\varepsilon_{2p_2q}}, \sqrt{\varepsilon_{2p_1q}\varepsilon_{2p_2q}} \;\text{or}\; \sqrt{\varepsilon_{p_1p_2}\varepsilon_{2p_1q}\varepsilon_{2p_2q}}. Thus such case must be rejected.

7. 8) Let d_1=p_1q_1q_2 and d_2=p_2q_1q_2 where p_1\equiv p_2\equiv -q_1 \equiv -q_2 \equiv1\pmod4 are distinct prime numbers such that [\left(\frac{p_1}{q_1}\right)=-1 or \left(\frac{p_1}{q_2}\right)=-1], [\left(\frac{p_2}{q_1}\right)=-1 or \left(\frac{p_2}{q_2}\right)=-1] and \left(\frac{p_1}{p_2}\right)=-1. According to [kuvcera1995parity, Theorem 1] we have N_{\mathbb{Q}(\sqrt{p_1p_2})/\mathbb{Q}}(\varepsilon_{p_1p_2})=-1, it follows that the unit \sqrt{\varepsilon_{p_1q_1q_2}\varepsilon_{p_2q_1q_2}} is the only square candidate belonging to E_K, because the other two units, \sqrt{\varepsilon_{p_1q_2q_2}} and \sqrt{\varepsilon_{p_2q_2q_2}}, take possible forms that do not belong to E_K (cf. Table ?). By using Table ? and Wada’s method we get q(K)=2 exactly in one of the following cases:

   1. \left(\frac{p_1}{q_1}\right)=\left(\frac{p_1}{q_2}\right)=\left(\frac{p_2}{q_1}\right)=\left(\frac{p_2}{q_2}\right)=-1,

   2. \left(\frac{p_1}{q_2}\right)=-\left(\frac{p_1}{q_1}\right)=-1 and \left(\frac{p_2}{q_2}\right)=-\left(\frac{p_2}{q_1}\right)=-1.

   Therefore in this sub-case we get q(K)=1 if and only if

   \left(\frac{p_1}{q_1}\right)=\left(\frac{p_1}{q_2}\right)=\left(\frac{p_2}{q_1}\right)\cdot\left(\frac{p_2}{q_2}\right)=\left(\frac{p_1}{p_2}\right)=-1.

   Which justifies the third case.

8. 9) Let d_1=p_1q_1q_2 and d_2=p_2q_1q_2 where [\left(\frac{p_1}{q_1}\right)=-1 or \left(\frac{p_1}{q_2}\right)=-1], [\left(\frac{p_2}{q_1}\right)=-1 or \left(\frac{p_2}{q_2}\right)=-1], \left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\neq \left(\frac{p_2}{p_1}\right)_4. As N_{\mathbb{Q}(\sqrt{p_1p_2})/\mathbb{Q}}(\varepsilon_{p_1p_2})=1 (cf. [kuvcera1995parity, Theorem 1]), Then by Table ? we conclude that q(K)=2, more precisely we have E_K=\langle{-1,\varepsilon_{p_1q_1q_2},\varepsilon_{p_2q_1q_2},\eta \rangle} with \eta=\sqrt{\varepsilon_{p_1p_2}\varepsilon_{p_1q_1q_2}}, \sqrt{\varepsilon_{p_1p_2}\varepsilon_{p_2q_1q_2}}, \sqrt{\varepsilon_{p_1q_1q_2}\varepsilon_{p_2q_1q_2}} \;\text{or}\; \sqrt{\varepsilon_{p_1p_2}\varepsilon_{p_1q_1q_2}\varepsilon_{p_2q_1q_2}}. Thus such case must be rejected.

9. 10) If d_1=2p_1 and d_2=2p_2 for distinct primes p_1\equiv p_2\equiv 5 \pmod 8 with \left(\frac{p_1}{p_2}\right)=-1, then [kuvcera1995parity, Theorems 1 and 2] gives that all three fundamental units of the quadratic subfields of K have norm -1. Wada’s method then implies that q(K)=2 or 1 according to whether \sqrt{\varepsilon_{2p_1}\varepsilon_{2p_2}\varepsilon_{p_1p_2}}\in K or not.

10. 11) If d_1=2p_1 and d_2=2p_2 for distinct primes with p_1\equiv p_2\equiv5\pmod8, \left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\neq \left(\frac{p_2}{p_1}\right)_4. By [kuvcera1995parity, Theorems 1 and 2] we have N(\varepsilon_{2p_1})=N(\varepsilon_{2p_1})=-N(\varepsilon_{p_1p_2})=-1, then q(K)=1, in fact the only square root candidate to be in E_K is \sqrt{\varepsilon_{p_1p_2}}, but it takes the form [\sqrt{p_1}, \sqrt{p_2}] which is not in K.

11. 12) If d_1=2p_1 and d_2=2p_2 for distinct primes with p_1\equiv5\pmod8, p_2\equiv1\pmod8, \left(\frac{2}{p_2}\right)_4\neq \left(\frac{p_2}{2}\right)_4 and \left(\frac{p_1}{p_2}\right)=-1. According to o [kuvcera1995parity, Theorems 1 and 2] we have N(\varepsilon_{2p_1})=N(\varepsilon_{p_1p_2})=-N(\varepsilon_{2p_2})=-1, thus q(K)=1 as \sqrt{\varepsilon_{2p_2}}\notin K.

12. 13) If d_1=2p_1 and d_2=2p_2 for distinct primes with p_1\equiv5\pmod8, p_2\equiv1\pmod8, \left(\frac{2}{p_2}\right)_4\neq \left(\frac{p_2}{2}\right)_4, \left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\neq \left(\frac{p_2}{p_1}\right)_4. Then According to o [kuvcera1995parity, Theorems 1 and 2] we have N(\varepsilon_{2p_2})=N(\varepsilon_{p_1p_2})=-N(\varepsilon_{2p_1})=1, moreover \sqrt{\varepsilon_{2p_2}}=[\sqrt{2},\sqrt{p_2}] and \sqrt{\varepsilon_{p_1p_2}}=[\sqrt{p_1},\sqrt{p_2}], then \sqrt{\varepsilon_{2p_2}\varepsilon_{p_1p_2}}\in K and q(K)=2.

13. 14) If d_1=2p_1 and d_2=2p_2 for distinct primes with p_1\equiv p_2\equiv1\pmod8, \left(\frac{2}{p_1}\right)_4\neq \left(\frac{p_1}{2}\right)_4, \left(\frac{2}{p_2}\right)_4\neq \left(\frac{p_2}{2}\right)_4 and \left(\frac{p_1}{p_2}\right)=-1. According to [kuvcera1995parity, Theorems 1 and 2] we have N(\varepsilon_{2p_1})=N(\varepsilon_{2p_2})=-N(\varepsilon_{2p_2})=1, and \sqrt{\varepsilon_{2p_1}}=[\sqrt{2},\sqrt{p_1}] and \sqrt{\varepsilon_{2p_2}}=[\sqrt{2},\sqrt{p_2}], thus \sqrt{\varepsilon_{2p_2}\varepsilon_{2p_2}}\in K and q(K)=2.

14. 15) If d_1=2p_1 and d_2=2p_2 for distinct primes with p_1\equiv p_2\equiv1\pmod8, \left(\frac{2}{p_1}\right)_4\neq \left(\frac{p_1}{2}\right)_4, \left(\frac{2}{p_2}\right)_4\neq \left(\frac{p_2}{2}\right)_4, \left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\neq \left(\frac{p_2}{p_1}\right)_4. Then according to [kuvcera1995parity, Theorems 1 and 2] we have N(\varepsilon_{2p_1})=N(\varepsilon_{2p_2})=N(\varepsilon_{2p_2})=1, and \sqrt{\varepsilon_{2p_1}}=[\sqrt{2},\sqrt{p_1}], \sqrt{\varepsilon_{2p_2}}=[\sqrt{2},\sqrt{p_2}] and \sqrt{\varepsilon_{p_1p_2}}=[\sqrt{p_1},\sqrt{p_2}], thus \sqrt{\varepsilon_{2p_2}\varepsilon_{2p_2}} and \sqrt{\varepsilon_{2p_2}\varepsilon_{2p_2}}\in K and q(K)=4.

15. 16) If d_1=p_1p_2 and d_2=p_2p_3 for some distinct prime numbers p_i\equiv1\pmod4 and \left(\frac{p_1}{p_2}\right)=\left(\frac{p_2}{p_3}\right)=\left(\frac{p_1}{p_3}\right)=-1, this case is similar to the previous, hence q(K)=2 or 1 according to whether \sqrt{\varepsilon_{p_1p_2}\varepsilon_{p_2p_3}\varepsilon_{p_1p_3}} is in K or not.

16. 17) If d_1=p_1p_2 and d_2=p_2p_3 for some distinct prime numbers p_i\equiv1\pmod4 such that \left(\frac{p_1}{p_2}\right)=\left(\frac{p_2}{p_3}\right)=-1, \left(\frac{p_1}{p_3}\right)=1 and \left(\frac{p_1}{p_3}\right)_4\neq \left(\frac{p_3}{p_1}\right)_4. Then according to [kuvcera1995parity, Theorems 1 ] we have N(\varepsilon_{p_1p_2})=N(\varepsilon_{p_2p_3})=-N(\varepsilon_{p_1p_3})=-1, then q(K)=1 since the only square root candidate to be in E_K is \sqrt{\varepsilon_{p_1p_3}}=[\sqrt{p_1},\sqrt{p_3}], which is not.

17. 18) If d_1=p_1p_2 and d_2=p_2p_3 for some distinct prime numbers p_i\equiv1\pmod4 such that \left(\frac{p_1}{p_2}\right)=-1, \left(\frac{p_2}{p_3}\right)=\left(\frac{p_1}{p_3}\right)=1, \left(\frac{p_1}{p_3}\right)_4\neq \left(\frac{p_3}{p_1}\right)_4 and \left(\frac{p_2}{p_3}\right)_4\neq \left(\frac{p_3}{p_2}\right)_4. Then according to [kuvcera1995parity, Theorems 1 ] we have N(\varepsilon_{p_2p_3})=N(\varepsilon_{p_1p_3})=-N(\varepsilon_{p_1p_2})=1, thus \sqrt{\varepsilon_{p_2p_3}}=[\sqrt{p_2},\sqrt{p_3}] and \sqrt{\varepsilon_{p_1p_3}}=[\sqrt{p_1},\sqrt{p_3}]. Therefore \sqrt{\varepsilon_{p_2p_3}\varepsilon_{p_1p_3}}\in K and q(K)=2.

18. 19) If d_1=p_1p_2 and d_2=p_2p_3 for some distinct prime numbers p_i\equiv1\pmod4 such that \left(\frac{p_1}{p_2}\right)=\left(\frac{p_2}{p_3}\right)=\left(\frac{p_1}{p_3}\right)=1, \left(\frac{p_1}{p_3}\right)_4\neq \left(\frac{p_3}{p_1}\right)_4, \left(\frac{p_2}{p_3}\right)_4\neq \left(\frac{p_3}{p_2}\right)_4 and \left(\frac{p_1}{p_2}\right)_4\neq \left(\frac{p_2}{p_1}\right)_4. Then according to [kuvcera1995parity, Theorems 1 ] we have N(\varepsilon_{p_2p_3})=N(\varepsilon_{p_1p_3})=N(\varepsilon_{p_1p_2})=1, thus \sqrt{\varepsilon_{p_2p_3}}=[\sqrt{p_2},\sqrt{p_3}], \sqrt{\varepsilon_{p_1p_3}}=[\sqrt{p_1},\sqrt{p_3}] and \sqrt{\varepsilon_{p_1p_2}}=[\sqrt{p_1},\sqrt{p_2}] . Therefore \sqrt{\varepsilon_{p_2p_3}\varepsilon_{p_1p_3}} and \sqrt{\varepsilon_{p_2p_3}\varepsilon_{p_1p_2}}\in K and q(K)=4.

Plugging all the above investigations together, we complete the proof of our theorem.

We have two main cases to investigate.

229 Assume that K is unramified over exactly one of its three quadratic subfields denote k. Notice that if the class number of K is odd, then h_2(k)=2. So Lemma 1.2 implies that k is in one of the following forms:

1. k=\mathbb{Q}(\sqrt{pq}) or \mathbb{Q}(\sqrt{2pq}) with \left(\frac{p}{q}\right)=-1 or p\equiv5\pmod8.

2. k=\mathbb{Q}(\sqrt{2q_1q_2}) with q_1\equiv3\pmod8 or q_2\equiv3\pmod8.

3. k=\mathbb{Q}(\sqrt{pq_1q_2}) with \left(\frac{p}{q_1}\right)=-1 or \left(\frac{p}{q_2}\right)=-1.

4. k=\mathbb{Q}(\sqrt{2p}) with \left(\frac{2}{p}\right)=-1 or [\left(\frac{2}{p}\right)=1 and \left(\frac{2}{p}\right)_4\not=\left(\frac{p}{2}\right)_4].

5. k=\mathbb{Q}(\sqrt{p_1p_2}) with \left(\frac{p_1}{p_2}\right)=-1 or [\left(\frac{p_1}{p_2}\right)=1 and \left(\frac{p_1}{p_2}\right)_4\not=\left(\frac{p_2}{p_1}\right)_4].

Hence, by class field theory, the class number of K is odd if and only if K takes one of the forms listed in items enumerated from 1) to 5) in the above theorem.

229 Now assume that K is ramified over each of its three quadratic subfields. Assume that the class number of K is odd and let k be any of of the quadratic subfields of K. Thus we have h_2(k)=1 or 2. It follows from Lemmas 1.1 and 1.2 that we have

Here q_1, q_2 and q_3 are prime numbers congruent to 3\pmod4 and satisfy some extra conditions as follows.

1. If K=\mathbb{Q}(\sqrt{2},\sqrt{q}), then q\equiv3\pmod4 is a prime number. Notice that in this case, we deduce from Lemma 5.2 that we have E_K= \langle{-1,\varepsilon_{2},\sqrt{\varepsilon_{q}},\sqrt{\varepsilon_{2q}}\rangle}. Therefore, Kuroda’s class number formula and Lemma 1.1 implies that the class number of K is odd.

2. If K=\mathbb{Q}(\sqrt{q_1},\sqrt{q_2}), then q_1\equiv q_2\equiv3\pmod4 are distinct prime numbers. Notice that in this case, by Lemma 5.2 we get E_K= \langle{-1,\varepsilon_{q_1},\sqrt{\varepsilon_{q_1}\varepsilon_{q_2}},\sqrt{\varepsilon_{q_1q_2}}\rangle} and so q(K)=2. Thus, by Kuroda’s class number formula, the class number of K is odd.

3. If K=\mathbb{Q}(\sqrt{2q_1},\sqrt{2q_2}), then q_1\equiv q_2\equiv3\pmod4 are distinct prime numbers. Notice that in this case, Lemma 5.2 implies that E_K= \langle{-1,\varepsilon_{2q_1},\sqrt{\varepsilon_{2q_1}\varepsilon_{2q_2}},\sqrt{\varepsilon_{2q_1}\varepsilon_{q_1q_2}}\rangle} and so q(K)=2. Then Kuroda’s class number formula implies that the class number of K is odd.

4. If K=\mathbb{Q}(\sqrt{q_1q_2},\sqrt{q_2q_3}), then q_1\equiv q_2\equiv q_3\equiv3\pmod4 are distinct prime numbers. Notice that in this case, Lemma 5.2 gives E_K= \langle{-1,\varepsilon_{q_1q_2},\sqrt{\varepsilon_{q_1q_2}\varepsilon_{q_2q_3}},\sqrt{\varepsilon_{q_1q_2}\varepsilon_{q_1q_3}}\rangle} and so q(K)=2. Then by Kuroda’s class number formula, the class number of K is odd.

Hence, putting all of the above together, we obtain the classification stated in this theorem.

Let K=\mathbb{Q}(\sqrt{\ell_1},\sqrt{\ell_2},\sqrt{\ell_3}), where \ell_1, \ell_2 and \ell_3 be distinct positive square-free integers. Notice that if h_2(K)=1, then by Theorem 4.1, we have:

Here \mathbb{P} denotes the set of prime numbers and \delta_K denotes the absolute discriminant of K. Notice that we have M_K=3 or 4. Thus, we have two main cases to investigate.

229 Let us start by assuming that M_K=3. This implies that all the integers \ell_i are prime numbers and so we have the following sub-cases:

1. Let \ell_1 \equiv\ell_2 \equiv\ell_3 \equiv 1\pmod4 be prime numbers. Then according to [Bulant], the class number of K is odd if and only if K takes the form in the first item of our theorem.

2. Let \ell_1=2 and \ell_2\equiv \ell_3 \equiv1\pmod4 be distinct prime numbers. Notice that K/\mathbb{Q}(\sqrt{2},\sqrt{\ell_1\ell_2}) is a quadratic unramified extension. Therefore, h(K) is odd if and only if the 2-class number of \mathbb{Q}(\sqrt{2},\sqrt{\ell_1\ell_2}) equals 2, which is, by Theorem 2.4 (cf. case D)), equivalent to the fact that K takes the forms in the second item of our theorem.

3. Let \ell_1=2 and \ell_2\equiv -\ell_3 \equiv1\pmod4 be distinct prime numbers. As K/k is an unramified quadratic extension with k=\mathbb{Q}(\sqrt{2\ell_2},\sqrt{\ell_3}), we have h(K) is odd if and only if h_2(k)=2. So by Theorem 2.1 (cf. case C)), we conclude that, h_2(k)=2 exactly when \ell_2\equiv 5 \pmod8.

4. Let \ell_1=2 and \ell_2\equiv \ell_3 \equiv3\pmod4 be distinct prime numbers. As K/k is an unramified quadratic extension with k=\mathbb{Q}(\sqrt{2\ell_2},\sqrt{\ell_3}), then by Theorem 2.2 (cf. case A)-1)) h(K) is odd if and only if \ell_2 or \ell_3 \equiv 3\pmod8.

5. Let \ell_1\equiv \ell_2\equiv\ell_3\equiv3\pmod4 be distinct prime numbers. Notice that K/k, with k=\mathbb{Q}(\sqrt{\ell_1\ell_2},\sqrt{\ell_3}), is an unramified quadratic extension. Therefore, by Theorem 2.3 (cf. Item D)), h(K) is odd if and only if d takes one of the forms in d takes one of the forms listed in Item 5) of our theorem.

6. Let \ell_1\equiv -\ell_2\equiv-\ell_3\equiv 1\pmod4 be distinct prime numbers. Consider the quadratic unramified extension K/k, k=\mathbb{Q}(\sqrt{\ell_1\ell_2},\sqrt{\ell_3}). So by Theorem 2.1 (cf. Items B)-1) and B)-2)), h_2(K)=1 if and only if d takes one of the conditions listed in 6) of our theorem.

7. Let \ell_1\equiv \ell_2\equiv-\ell_3\equiv 1\pmod4 be distinct prime numbers. Notice that K/k, with k=\mathbb{Q}(\sqrt{\ell_1\ell_3},\sqrt{\ell_2}), is an unramified quadratic extension. So by Theorem 2.6 (cf. Item E)), h_2(K)=1 if and only if d takes one of the conditions listed in 7) of our theorem.

229 Now assume that M_K=4. So K is in one of the following forms:

where \ell_i are distinct prime numbers. Thus, we have 3 main cases to investigate.

1. Let K=\mathbb{Q}(\sqrt{\ell_1},\sqrt{\ell_2},\sqrt{\ell_3\ell_4}). Notice that \mathbb{Q}(\sqrt{\ell_1},\sqrt{\ell_2},\sqrt{\ell_3},\sqrt{\ell_4})/K is ramified if and only if one of the following three congruence conditions holds:

   1. \ell_1\equiv \ell_2\equiv-\ell_3\equiv-\ell_4\equiv1\pmod4,

   2. \ell_1=2 and \ell_2\equiv-\ell_3\equiv-\ell_4\equiv1\pmod4,

   3. \ell_1\equiv\ell_2\equiv-\ell_3\equiv1\pmod4 and \ell_4=2.

   Otherwise we have h_2(K)\geq 2, and such fields must be rejected. Now, for each of the cases above, we choose k=\mathbb{Q}(\sqrt{\ell_1\ell_2},\sqrt{\ell_3\ell_4}). It follows that K/k is an unramified quadratic extension. For the case \mathrm{C}1) we have according to Theorem 2.3 (cf. F)-1) and F)-2)), h_2(K)=1 if and only of one of the following conditions holds:

   1. \left(\frac{\ell_1}{\ell_2}\right)=\left(\frac{\ell_1}{\ell_3}\right)=\left(\frac{\ell_1}{\ell_4}\right)=\left(\frac{\ell_1}{\ell_3}\right)=-\left(\frac{\ell_1}{\ell_4}\right)=-1,

   2. \left(\frac{\ell_1}{\ell_2}\right)=-\left(\frac{\ell_1}{\ell_3}\right)=\left(\frac{\ell_1}{\ell_4}\right)=\left(\frac{\ell_1}{\ell_3}\right)=-\left(\frac{\ell_1}{\ell_4}\right)=\left(\frac{\ell_3}{\ell_4}\right)=-1.

   For the case \mathrm{C}2) we have according to Theorem 2.3 (cf. C)-1) and C)-2)), h_2(K)=1 if and only of one of the following conditions holds:

   1. \ell_2\equiv5\pmod8,\ell_3\equiv \ell_4\equiv3\pmod8 and \left(\frac{\ell_2}{\ell_3}\right)=-\left(\frac{\ell_2}{\ell_4}\right)=-1, or

   2. \ell_2\equiv5\pmod8,\ell_3\equiv7\pmod8, \ell_4\equiv3\pmod8 and \left(\frac{\ell_2}{\ell_3}\right)=-\left(\frac{\ell_2}{\ell_4}\right)=\left(\frac{\ell_4}{\ell_3}\right)=-1.

   For the remaining case \mathrm{C}3) in this sub-case we have according to Theorem 2.2 (cf. B)-3) to B)-5)), h_2(K)=1 if and only of one of the following conditions holds:

   1. \ell_1\equiv5\pmod8, \ell_2\equiv1\pmod{8}, \ell_3\equiv3\pmod{8} and \left(\dfrac{\ell_3}{\ell_1}\right)=-\left(\dfrac{\ell_3}{\ell_2}\right)=-\left(\dfrac{\ell_1}{\ell_2}\right)=1,

   2. \ell_1\equiv \ell_2\equiv5\pmod8 and \left(\dfrac{\ell_3}{\ell_1}\right)=-\left(\dfrac{\ell_3}{\ell_2}\right)=-\left(\dfrac{\ell_1}{\ell_2}\right)=1,

   3. \ell_1\equiv1\pmod8, \ell_2\equiv5\pmod8 and \left(\dfrac{\ell_3}{\ell_1}\right)=\left(\dfrac{\ell_3}{\ell_2}\right)=\left(\dfrac{\ell_1}{\ell_2}\right)=-1.

2. Let K=\mathbb{Q}(\sqrt{\ell_1\ell_2},\sqrt{\ell_1\ell_3},\sqrt{\ell_4}). Notice that \mathbb{Q}(\sqrt{\ell_1},\sqrt{\ell_2},\sqrt{\ell_3},\sqrt{\ell_4})/K ramify if and only if one of the following conditions holds :

   1. \ell_1\equiv \ell_2\equiv-\ell_4\equiv3\pmod4 and \ell_3=2,

   2. \ell_1=2 and \ell_2\equiv\ell_3\equiv-\ell_4\equiv3\pmod4,

   3. \ell_1\equiv\ell_3\equiv-\ell_4\equiv3\pmod4 and \ell_2=2.

   4. \ell_1\equiv\ell_2\equiv\ell_3\equiv3\pmod4 and \ell_4=2.

   Following the same idea as in the previous case, and using the unramified (under all above conditions C1)–C4)) quadratic extension K/k where k=\mathbb{Q}(\sqrt{\ell_1\ell_2},\sqrt{\ell_1\ell_3\ell_4}), we obtain the following investigation:
   In case C1), it follows from Theorem 2.3 (cf. J)-1) and J)-2)) we have h_2(K)=1 if and only if one of the following conditions holds:

   1. \left(\frac{\ell_4}{\ell_1}\right)=\left(\frac{\ell_4}{\ell_2}\right)=-1,

   2. \left(\frac{\ell_4}{\ell_1}\right)=-\left(\frac{\ell_4}{\ell_2}\right)=1 and \ell_4\equiv5\pmod8.

   For the case \mathrm{C}2), according to Theorem 2.2 (cf. D)-1) and D)-2)) we have h_2(K)=1 if and only if one of the following conditions holds:

   1. \ell_4\equiv5\pmod8 and \left(\dfrac{\ell_4}{\ell_3}\right)=-1 or [\left(\dfrac{\ell_4}{\ell_3}\right)=-\left(\dfrac{\ell_4}{\ell_2}\right)=1],

   2. \ell_4\equiv1\pmod8 and \left(\dfrac{\ell_4}{\ell_3}\right)=\left(\dfrac{\ell_4}{\ell_2}\right)=-1.

   Now for case C3). By applying Theorems 2.2 and 1.3 we conclude that h_2(K)\geq2. Then such field must be rejected.
   Finally if condition C4) holds, then from Theorem 2.3 (cf. H)) it follows that h_2(K)=1 if and only if \ell_3\equiv 3\pmod8.

3. Let K=\mathbb{Q}(\sqrt{\ell_1\ell_2},\sqrt{\ell_1\ell_3},\sqrt{\ell_1\ell_4}). Notice that \mathbb{Q}(\sqrt{\ell_1},\sqrt{\ell_2},\sqrt{\ell_3},\sqrt{\ell_4})/K ramify if and only if one of the following conditions holds :

   1. \ell_1\equiv \ell_2\equiv\ell_3\equiv\ell_4\equiv3\pmod4,

   2. \ell_3=2 and \ell_1\equiv\ell_2\equiv\ell_4\equiv3\pmod4,

   3. \ell_1=2 and \ell_2\equiv\ell_4\equiv-\ell_3\equiv3\pmod4.

   By the same investigation as in the two above forms of K, by using the suitable unramified (under conditions C1)–C3)) quadratic extension K/k, where k=\mathbb{Q}(\sqrt{\ell_2\ell_3},\sqrt{\ell_1\ell_4}) we caracterise eauch case separably:
   If conditions C1) holds, then by Theorem 2.3 (cf. G)-1) and G)-2)) we have h_2(K)=1 if and only if one of the following conditions holds:

   1. \left(\frac{\ell_2}{\ell_3}\right)=\left(\frac{\ell_2}{\ell_4}\right)=\left(\frac{\ell_4}{\ell_1}\right)=-\left(\frac{\ell_2}{\ell_1}\right)=-\left(\frac{\ell_3}{\ell_4}\right)=1, or

   2. \left(\frac{\ell_2}{\ell_3}\right)=\left(\frac{\ell_2}{\ell_4}\right)=\left(\frac{\ell_4}{\ell_1}\right)=\left(\frac{\ell_2}{\ell_1}\right)=-\left(\frac{\ell_3}{\ell_1}\right)=-\left(\frac{\ell_3}{\ell_4}\right)=1.

   If conditions C2) holds, then by Theorem 2.2 (cf. B)-1) and B)-2)) we have h_2(K)=1 if and only if one of the following conditions holds:

   1. \ell_2\equiv \ell_4\equiv7\pmod{8}, \ell_1\equiv3\pmod{8} and \left(\dfrac{\ell_2}{\ell_1}\right)=-\left(\dfrac{\ell_2}{\ell_4}\right)=\left(\dfrac{\ell_1}{\ell_4}\right)=1,

   2. \ell_1\equiv\ell_4\equiv3\pmod{8}, \ell_2\equiv7\pmod{8} and \left(\dfrac{\ell_2}{\ell_1}\right)=-\left(\dfrac{\ell_2}{\ell_4}\right)=-\left(\dfrac{\ell_1}{\ell_4}\right)=-1.

   Finally if the case C3) holds. We have by Theorems 2.2 and 1.3 we have h_2(K)\geq 2. Hence such field must be rejected.
   

   Hence, putting all of the above together, we obtain the complete classification stated in the theorem. This completes the proof.

Put K=\mathbb{Q}(\sqrt{\ell_1}, \sqrt{\ell_2}, \sqrt{\ell_3}, \sqrt{\ell_4}), where \ell_1, \ell_2, \ell_3 and \ell_4 are distinct positive square-free integers. Assume that one of the integers \ell_{i} is composite. Then, the fact that all \ell_i’s are pairwise distinct, implies that there are at least five finite primes that ramify in K/\mathbb{Q}; this Theorem 4.1, implies that the class number of K is even.

It follows by Mouhib’s Theorem (cf. [paritymohib, The Main Theorem, p. 1206]) that h(K) is odd if and only if \ell_1, \ell_2, \ell_3 and \ell_4 are distinct prime numbers satisfying one of the conditions in the above theorem.

As d\equiv1\pmod4, then \varepsilon_{d}=x+y\sqrt{d}, where x and y are integers or semi integers :
229 If x and y are integers then x+1 and x-1 are not squares in \mathbb{N} and x^2-1=dy^2 (cf. [AzAekhTouCor, First item in Lemma 2]). Assume that there is a divisor \delta of d such that: \delta(x+1) or \delta(x-1) is a square in \mathbb{N}. Then we have

where y_1, y_2 and d' are integers such that y=y_1y_2 and d=\delta d'. Then

then

As d\equiv 1\pmod4 and \delta divides d, is follows that \delta\equiv \pm1\pmod4. Reducing modulo 4 we get

which is an absurd.

229 If x and y are semi integers : x=a/2 and y=b/2, for some odd integers a and b such that a^2-4=db^2. First of all if a-2 or a+2 is a square in \mathbb{N}, then we have:

where b_1 and b_2 are two integers such that b=b_1b_2. Then \varepsilon_d=\frac{1}{4}(b_1^2+b_2^2 d+2b\sqrt{d}), thus \sqrt{\varepsilon_d}\in \mathbb{Q}(\sqrt{d}), which is absurd. Therefore, a+2 and a-2 are not squares in \mathbb{N}; hence, a fortiori, x-1 and x+1 are not squares in \mathbb{Q}.

Now if there is some divisor \delta of d such that \delta(x+1) or \delta(x-1) is a square in \mathbb{Q}. Then as a is odd we have \delta(x\pm1)=\frac{\delta(a\pm2)}{2} which can not be a square in \mathbb{Q} because \delta(a\pm2) and 2 are coprime and both square free (A rational number is a square in \mathbb{Q} if and only if both its numerator and its denominator are squares in \mathbb{N}.).

We use the same techniques as in the proof of Lemma 2.5.