Grothendieck rings of ordered subgroups of \mathbb{Q}

We first prove \mathrm{(i)}. For a, b\in G with a<b we have that

Since b-a\in G, the map x\to x+b-a gives a definable isomorphism from (a,\infty) to (b,\infty). It follows that [(a,b)]= -1 as desired, since the value of a singleton is 1.

For \mathrm{(ii)}, recall that S_G does not contain all primes and take p to be the smallest such prime. Suppose k\in \mathbb{N} such that 1/p^k\in G and 1/p^{k+1}\notin G. We get that

Here we use that x\to \frac{1}{p^{k}}-x and x\to x/(p-1) are definable isomorphisms. Note by the choice of p, all prime factors of p-1 are in S_G. By \mathrm{(i)}, we know that [(0,\frac{1}{p^{k}})]=-1 and hence [(0,\frac{1}{p^{k+1}})]=-\frac{1}{2}.

The assertions in \mathrm{(i), (ii)} are immediate, and then \mathrm{(iii)} follows by a direct application of Lemma 2.1\mathrm{(ii)}.

For I:=(0,\infty)\cap G, we set

where f and g are definable functions in (G;+,<), or f,g\in \{-\infty, +\infty\}, with f(x)<g(x) for all x\in I. We split I^2 into three disjoint parts and obtain

where \text{Id}_{I} is the identity map on I and \Gamma(\text{Id}_{I}) denotes its graph. We work with three definable isomorphisms given as follows.

Hence we may rewrite (Proof) as

and it follows that X^2+X=0 as claimed.

Let n\geqslant 2 such that n_{S_G}=1 and y=\frac{a}{b}\in G, with a and b coprime, be given.

By the definition of n_{K_G}, there are u, v \in \mathbb{Z} such that b=u\cdot v, u divides n_{K_G}, and \text{gcd}(v,n_{K_G})=1.

The fact that \text{gcd}(v,n_{K_G})=1 implies that \text{gcd}(v,n)=1 and we have that

Since \text{gcd}(v,n)=1, there is unique i\in \{0,\ldots n-1\} such that

and the proof is complete.

The idea is to partition (-\infty,+\infty)\cap G into equivalence classes modulo p. For i=0, \ldots, p-1, set N_i\coloneq\{x\in G:\ p\mid (x+\frac{i}{p^k})\}. Then by Lemma 2.4 we have

For i=0,\ldots,p-1 we have a definable bijection N_i \to G given by x\to\frac{x+\frac{i}{p^k}}{p}. This implies that

and hence (p-1)(2X+1)=0. Multiplying both sides by 2X+1 and employing Lemma 2.3 we get that p-1=0. Since 2 is invertible in the Grothendieck ring of (G;+,<) by Lemma 2.1\mathrm{(ii)}, the desired conclusion follows.

We begin with the proof of \mathrm{(i)}. It suffices to consider the case of a\in G and b\in \mathbb{Q}\setminus G.

By translating suitably we may assume that a=0 and b=m/n\notin G such that b<1 and m,n\in \mathbb{N} with \text{gcd}(m,n)=1. Since multiplying and dividing by elements of S_G are definable isomorphisms in (G;+,<), we may assume without loss of generality that all prime factors of m and n are not in S_G.

For b<1/2, we have that

We will show that [(\frac{m}{n},\frac{n-m}{n})]=0. It then follows that [(0,\frac{m}{n})]=-1/2 as desired for \frac{m}{n}<1/2. From this, we also obtain the case b>1/2 as follows

Note that our assumptions imply that 2\in S_G and hence \frac{m}{n}\neq \frac{n-m}{n}. Suppose \frac{m}{n}<1/2. Since G is dense in \mathbb{Q}, we have c\in G such that \frac{m}{n}<c<\frac{n-m}{n}. Using definable isomorphisms we get that

Next we aim to find Q\in S_G such that n divides Q+1 in the integers. Since we can then translate the interval in the first summand of equation (?) by the constant T=c+Q(1-c) -\frac{(Q+1)m}{n}\in G and multiply the second summand of equation (?) by Q respectively to get our desired claim as follows.

Here we used Lemma 2.1(\mathrm{i}). By our assumption, we have an odd prime q which divides p-1 for all p\notin S_G, and we solve for prime Q satisfying

Note that q\in S_G and hence we have that q is coprime to n as well. Then by the Chinese remainder theorem and Dirichlet’s theorem on primes in arithmetic progressions we have that there is a prime Q satisfying the displayed divisibility conditions. Note such a Q necessarily belongs in S_G since q does not divide Q-1. This finishes the proof for (\mathrm{i}).

For (\mathrm{ii}), we start with c,d\in\mathbb{Q}\backslash G, and take a,b\in G with a<c<d<b. Then we have that

It follows that [(c,d)]=0.

Assertion (\mathrm{iii}) follows by a direct application of (\mathrm{i}) and Lemma 2.2(\mathrm{i}).

It suffices to show that the value of all quasi-cells can be expressed as a linear combination of values of unary quasi-cells. Given a quasi-cell C\cap G^n, we proceed by induction on n. The base-cases 1 is immediate.

Take a quasi-cell C'=C\cap G^{n+1}, where C is a cell in \mathbb{Q}^{n+1}. We first consider the case where C is given as \Gamma(f) where f is a definable continuous function on a cell A\subseteq \mathbb{Q}^{n}. Let A'=A\cap G^n and we have that

for some a_i,b, m\in\mathbb{Z}. The function y=\frac{\sum_{i=1}^n a_ix_i+b}{m} is only definable on the elements of A divisible by m. Let m_T:=m/m_{S_G} and observe that an element of G is divisible by m if and only if it is divisible by m_T. The projection map onto the first n co-ordinates \pi gives a definable isomorphism from C to \pi(C)=A'\cap P_{m_T}(\sum_{i=1}^n a_ix_i+b) and we are done by induction.

The other case is that

where A is a cell in \mathbb{Q}^{n}, and

with f(x)<g(x) for all x\in A. Set A'=A\cap G^n as before.

If f=-\infty and g=+\infty, then C'=A'\times G and [C']=[A']\cdot[G]. We now proceed to the non-trivial case of f\neq -\infty and g= +\infty. We have that

for some a_i,b, m\in\mathbb{Z}. Let m_T:=m/m_{S_G} as before and observe that

For each j\in \{0,\ldots, m_T-1\}, we set A'_j:=A'\cap P_{m_T}(\sum_{i=1}^n a_ix_i+b+\frac{j}{m_K}) and C'_j:=\{(x,y)\in A'_j\times G:\ \sum_{i=1}^n a_ix_i+b<my\}, and observe that

is a definable bijection in (G;+,<). Hence we have that

and we are done by induction. The case of f=-\infty and g\neq \infty is similar.

Finally, suppose f\neq -\infty and g\neq \infty. There is a cell A\subseteq \mathbb{Q}^n such that

and we employ the the previous cases and induction to conclude the proof.