Felix' Math Place » baby steps http://math.fontein.de Focussed on, but not limited to Computational Number Theory Sat, 30 Jul 2011 12:35:49 +0000 en hourly 1 http://wordpress.org/?v=3.1 Infrastructures and Global Fields. http://math.fontein.de/infrastructures/ http://math.fontein.de/infrastructures/#comments Thu, 23 Jul 2009 05:59:40 +0000 Felix Fontein http://math.fontein.de/?page_id=259 The following posts give an introduction to infrastructures and how to obtain these from global fields:

  1. The Discrete Logarithm Problem and Generalizations.
  2. One-dimensional Infrastructures.
  3. Interpreting One-dimensional Infrastructures as Groups: f-Representations.
  4. n-dimensional Infrastructures.
  5. How to Obtain Reduction Maps for n-dimensional Infrastructures.
  6. Obtaining Infrastructures from Global Fields.

See also my article on infrastructures at Wikipedia.

]]> http://math.fontein.de/infrastructures/feed/ 1 Obtaining Infrastructures from Global Fields. http://math.fontein.de/2009/07/21/obtaining-infrastructures-from-global-fields/ http://math.fontein.de/2009/07/21/obtaining-infrastructures-from-global-fields/#comments Tue, 21 Jul 2009 09:39:48 +0000 Felix Fontein http://math.fontein.de/?p=196 Basics on Global Fields.

Let K be a global field, i.e. an algebraic number field or an algebraic function field with a finite constant field. In the first case, let k^* be the roots of unity and k = k^* \cup \{ 0 \}. In the latter case, let k be the exact field of constants.

Let S = \{ \frakp_1, \dots, \frakp_{n+1} \} be the set of infinite places of K. If K is a number field, the elements of S correspond to embeddings of K into \C up to complex conjugation. Define q := \exp(1), and for \frakp \in S let \sigma : K \to \C be a corresponding embedding. Then define \nu_\frakp(f) := -\log \abs{\sigma(f)} for f \in K^* and \deg \frakp := 1 if \sigma(K) \subseteq \R, or \deg \frakp := 2 otherwise, and define \G_\frakp := \R. If K is a function field, let q := \abs{k}, i.e. k = \F_q; in this case, there exists an element x \in K \setminus k whose poles are exactly the elements of S, i.e. are the places of K lying above the infinite place of k(x). In all cases, S is finite and non-empty.

For a non-archimedean place \frakp of K, let \calO_\frakp be the valuation ring and \frakm_\frakp its maximal idea, and denote the discrete valuation by \nu_\frakp. Then set \deg \frakp := \log_q \abs{\calO_\frakp / \frakm_\frakp} and \abs{f}_\frakp := q^{-\nu_\frakp(f) \deg \frakp}, f \in K. Define \G_\frakp := \Z. In the number field case, let \G := \R, and otherwise \G := \Z.

Denote the set of places of K by \calP_K. The divisor group of K is \Div(K) := \coprod_{\frakp \in \calP} \G_\frakp, and for D = \sum_{\frakp \in \calP_K} n_\frakp \frakp define \deg D := \sum_{\frakp \in \calP_K} n_\frakp \deg \frakp. This is a homomorphism \deg : \Div(K) \to \G; denote its kernel by \Div^0(K). For f \in K^*, (f) := \sum_{\frakp \in \calP_K} \nu_\frakp(f) \frakp \in \Div^0(K) is a principal divisor; let the group of all these be denoted by \Princ(K). Then \Pic(K) := \Div(K) / \Princ(K) is the divisor class group of K and \Pic^0(K) := \Div^0(K) / \Princ(K) its degree zero part.

The support of a divisor D = \sum_{\frakp \in \calP_K} n_\frakp \frakp is the set \support(D) = \{ \frakp \in \calP_K \mid n_\frakp \neq 0 \}. Consider the subgroups \Div_{fin}(K) :={} & \{ D \in \Div(K) \mid \support(D) \cap S = \emptyset \} \\ \text{and} \qquad \Div_\infty(K) :={} & \{ D \in \Div(K) \mid \support(D) \subseteq S \}; then \Div(K) = \Div_{fin}(K) \oplus \Div_\infty(K). Moreover, let \Div_\infty^0(K) := \Div^0(K) \cap \Div_\infty(K). The set \calO := \calO_S := \{ f \in K \mid \nu_\frakp(f) \ge 0 \text{ for all } \frakp \in S \} is a Dedekind domain, whose maixmal ideals correspond to the places in \calP_K \setminus S. Moreover, the fractional ideal group \Id(\calO_S) is isomorphic to \Div_{fin}(K) by \divisor(\fraka) = \sum_{\frakp \not\in S} n_\frakp \frakp, in case \fraka = \prod_{\frakp \not\in S} (\frakm_\frakp \cap \calO_S)^{-n_\frakp}; the inverse is given by the restriction of Formula does not parse: \ideal : \Div(K) \to \Id(\calO_S), \sum n_\frakp \frakp \mapsto \prod_{\frakp \not\in S} (\frakm_\frakp \cap \calO_S)^{-n_\frakp} to \Div_{fin}(K). The group of fractional principal ideals \PId(\calO_S) equals Formula does not parse: \ideal(\Princ(K)). The quotient \Id(\calO_S) / \PId(\calO_S) is the ideal class group \Pic(\calO_S) of \calO_S. Putting all these things together, we get the following diagram with exact rows and columns: \displaystyle \xymatrix{ & 0 \ar[d] & 0 \ar[d] & 0 \ar[d] & \\ 0 \ar[r] & \calO_S^* / k^* \ar[r] \ar[d] & \Div^0_\infty(K) \ar[r] \ar[d] & T \ar[r] \ar[d] & 0 \\ 0 \ar[r] & K^* / k^* \ar[r] \ar[d] & \Div^0(K) \ar[r] \ar[d] & \Pic^0(K) \ar[r] \ar[d] & 0 \\ 0 \ar[r] & K^* / \calO_S^* \ar[r] \ar[d] & \Id(\calO_S) \ar[r] \ar[d] & \Pic(\calO_S) \ar[r] \ar[d] & 0 \\ & 0 & H \ar@{=}[r] \ar[d] & H \ar[d] & \\ & & 0 & 0 & } Here, T and H are essentially defined by the diagram, i.e. are the kernels and cokernels of the respective maps. In the number field case, H = 0, and in the function field case, H \cong (\deg \frakp \mid \frakp \in \calP_K) / (\deg \frakp \mid \frakp \in S).

A Geometry of Numbers in Global Fields.

Let \fraka \in \Id(\calO_S) and t_1, \dots, t_{n+1} \in \G. Define \displaystyle B(\fraka, (t_1, \dots, t_{n+1})) := \{ f \in \fraka \mid \forall i : \abs{f}_{\frakp_i} \le q^{t_i \deg \frakp_i} \}. If D := \divisor(\fraka) + \sum_{i=1}^{n+1} t_i \frakp_i \in \Div(K) and L(D) is the Riemann-Roch space of D, then L(D) = B(\fraka, (t_1, \dots, t_{n+1})). In particular, the set is finite and invariant under multiplication by elements of k; in case K is a function field, L(D) is a finite-dimensional k-vector space, whose dimension is described by the Riemann-Roch theorem. In the number field case, we can make statements on L(D) with Minkowski’s Lattice Point Theorem.

Consider the map \displaystyle \Psi : K^* \to \G^n, \quad f \mapsto (-\nu_{\frakp_1}(f), \dots, -\nu_{\frakp_n}(f)). Then \Lambda := \Psi(\calO^*) \cong \Z^n is a lattice by Dirichlet’s Unit Theorem, and \ker \Psi|_{\calO^*} = k^*. We get \calO^* \cong k^* \times \Z^n, and n is called the unit rank of \calO_S. This \Lambda will be the lattice for our n-dimensional infrastructure.

Reduced Ideals.

The elements of X will be principal reduced fractional ideals, modulo an equivalence relation. We begin by defining minima, which are similar to the ones introduced in the previous post for lattices.

Definition.
Let \fraka \in \Id(\calO_S) and \mu \in \fraka \setminus \{ 0 \}. We say that \mu is a minimum of \fraka if every f \in \fraka \setminus \{ 0 \} with \abs{f}_{\frakp_i} \le \abs{\mu}_{\frakp_i} for all i satisfies \abs{f}_{\frakp_i} = \abs{\mu}_{\frakp_i} for all i. Denote the set of all minima of \fraka by \calC(\fraka).

Using them, we can define reduced ideals:

Definition.
An ideal \fraka \in \Id(\calO_S) is said to be reduced if 1 \in \fraka is a minimum. Write \Red_S(K) for the set of all reduced ideals of \calO_S. For \frakb \in \Id(\calO_S) let \Red_S(\frakb) := \{ \fraka \in \Red_S(K) \mid \exists f \in K^* : f \fraka = \frakb \}.

The equivalence relation we need is defined by \displaystyle \fraka \sim_S \fraka' :\Leftrightarrow \exists f \in K^* : \fraka = f \fraka' \wedge \forall \frakp \in S : \abs{f}_\frakp = 1 for \fraka, \fraka' \in \Id(\calO_S). We then get the following results:

Theorem.
  1. We have that \Red_S(K) is a finite set.
  2. In case \deg \frakp = 1 for some \frakp \in S, we get \fraka \sim_S \fraka' for \fraka, \fraka' \in \Red(K) if, and only if, \fraka = \fraka'.
  3. We have that \calO^* acts on \calC(\fraka) by multiplication.
  4. The map \displaystyle \calC(\fraka) / \calO^* \to \Red(\fraka), \quad \mu \calO^* \mapsto \frac{1}{\mu} \fraka is a bijection.
  5. If \fraka \in \Red(K) and \frakb \in \Id(\calO) satisfies \fraka \sim_S \frakb, then \frakb \in \Red(\fraka).
□

The proofs of these and the following results or hints to the proofs can be found here. We next construct the map d:

Theorem (Infrastructure, Part I).
Fix an ideal \fraka \in \Id(\calO). Define X_\fraka := \Red(\fraka)/_{\sim_S} and define \displaystyle d_\fraka : X \to \G^n / \Lambda, \quad [\tfrac{1}{\mu} \fraka]_{\sim_S} \mapsto \Psi(\mu) + \Lambda. Then d_\fraka is well-defined and injective.
□

For a, a' \in K^*, write \displaystyle a \sim_S a' :\Longleftrightarrow \forall \frakp \in S : \abs{a}_\frakp = \abs{a'}_\frakp. Define \hat{X} := \calC(\fraka)/_{\sim_S} and \displaystyle \hat{d} : \hat{X} \to \G^n, \quad [\mu]_\sim \mapsto \Psi(\mu). Then (\hat{X}, \hat{d}) is the unrolled version of (X, d): if \pi : \G^n \to \G^n / \Lambda, x \mapsto x + \Lambda is the projection, and \psi : \hat{X} \to X, [\mu]_\sim \mapsto [\frac{1}{\mu} \fraka]_\sim, then the following diagram commutes: \displaystyle \xymatrix{ \hat{X} \ar[d]_{\psi} \ar[r]^{\hat{d}} & \G^n \ar[d]^{\pi} \\ X \ar[r]_{d} & \G^n/\Lambda } In particular, \hat{d}(\hat{X}) is the set \hat{X} from the previous post.

The Reduction Map, f-Representations, and the Infrastructure.

We proceed by defining f-representations, as giving these is equivalent to give a reduction map. Fix an ideal \fraka \in \Id(\calO_S).

First, define for f, f' \in K^* \displaystyle f \le_S f' :\Longleftrightarrow (\abs{f}_{\frakp_{n+1}}, \abs{f}_{\frakp_1}, \dots, \abs{f}_{\frakp_n}) \le_{\ell ex} (\abs{f'}_{\frakp_{n+1}}, \abs{f'}_{\frakp_1}, \dots, \abs{f'}_{\frakp_n}), where \le_{\ell ex} is the lexicographic order on \R^{n+1}.

Definition.
A tuple ([\frakb]_{\sim_S}, (t_1, \dots, t_n)) \in \Red_S(\fraka)/_{\sim_S} \times \G^n is said to be an f-representation if 1 is a smallest element of \displaystyle B(\frakb, (t_1, \dots, t_n, 0)) \setminus \{ 0 \} with respect to \le. Denote the set of all f-representations by \fRep(\fraka).

One quickly sees that this is well-defined. We have two auxilliary results:

Lemma (Uniqueness).
Let A = ([\frakb]_{\sim_S}, (t_1, \dots, t_n)) \in \fRep(\fraka) and f \in K^* such that \displaystyle B = ([\tfrac{1}{f} \frakb]_{\sim_S}, (t_1 + \nu_{\frakp_1}(f), \dots, t_n + \nu_{\frakp_n}(f))) \in \fRep(\fraka). Then \abs{f}_\frakp = 1 for all \frakp \in S, i.e. A = B.
□
Lemma (Reduction).
Let v = (v_1, \dots, v_n) \in \G^n. Then there exists a smallest \ell \in \G such that B_\ell := B(\fraka, (v_1, \dots, v_n, \ell)) \setminus \{ 0 \} \neq \emptyset. If \mu is minimal with respect to \le in that B_\ell, then \displaystyle ([\tfrac{1}{\mu} \fraka]_{\sim_S}, (v_1 + \nu_{\frakp_1}(\mu), \dots, v_n + \nu_{\frakp_n}(\mu))) \in \fRep(\fraka) and \Phi(\mu) + (v_1 + \nu_{\frakp_1}(\mu), \dots, v_n + \nu_{\frakp_n}(\mu)) + \Lambda = v + \Lambda.
□

From that, we get the following result:

Theorem (Infrastructure, Part II).
Let \fraka \in \Id(\calO). Then the map \Phi :{} & \fRep(\fraka) \to \G^n / \Lambda \\ & ([\tfrac{1}{\mu} \fraka]_{\sim_S}, (t_1, \dots, t_n)) \mapsto \Psi(\mu) + (t_1, \dots, t_n) + \Lambda is a bijection.
□

This allows to equip \fRep(\fraka) with a group operation. We will see that the group operation of \fRep(\calO_S) can be described in a very explicit form. This extends to a broader interpretation of the infrastructure, whence we will do this in the next section.

Before ending this section, we want to state a result which shows that f-representations are small.

Theorem.
Let ([\frakb]_{\sim_S}, (t_1, \dots, t_n)) \in \fRep(\fraka), then \divisor(\frakb) \ge 0, t_i \ge 0 for all i and \displaystyle \deg \divisor(\fraka) + \sum_{i=1}^n t_i \deg \frakp_i \le \kappa, where \displaystyle \kappa := \begin{cases} g + \deg \frakp_{n+1} - 1 & \text{if } K \text{ is a function field} \\ s \log \tfrac{2}{\pi} + \tfrac{1}{2} \log \abs{\Delta} & \text{if } K \text{ is a number field;} \end{cases} here, g is the genus of K in case K is a function field, and in case K is a number field, s denotes the number of places of degree two and \Delta is the discriminant of \calO_S.
□

Therefore, f-representations are small.

The Infrastructure and the Divisor Class Group.

Assume for a moment that \deg \frakp_{n+1} = 1, or that K is a number field. Then we have a short exact sequence \displaystyle \xymatrix{ 0 \ar[r] & T \ar[r] & \Pic^0(K) \ar[r] & \Pic(\calO_S) \ar[r] & 0, } and T \cong \G^n / \Lambda \cong \fRep(\fraka). This means that the divisor class group \Pic^0(K) is covered by copies of \G^n/\Lambda, where the copies are indexed by the elements of the divisor class group. If \fraka and \fraka' are in the same ideal class, X_\fraka and X_{\fraka'} differ by a translation, i.e. they give essentially the same infrastructure; in fact, \fRep(\fraka) = \fRep(\fraka'). Hence, one could get the idea to cover \Pic^0(K) by \fRep(\fraka), where \fraka ranges over the distinct ideal classes, i.e. by \fRep(K) := \bigcup_{\fraka \in \Id(\calO_S)} \fRep(\fraka). It turns out that this is indeed the case, and the arithmetic on \fRep(\calO_S) and \Pic^0(K) turn out to be the same under the bijection we get.

In case K is a function field and \deg \frakp_{n+1} > 1, we have T \not\cong \G^n / \Lambda in general (this is the case if, and only if, \deg \frakp_{n+1} = \gcd(\deg \frakp_1, \dots, \frakp_n, \frakp_{n+1})), and \Pic^0(K) \to \Pic(\calO_S) does not needs to be surjective. It would be nice to change the above sequence to \displaystyle \xymatrix{ 0 \ar[r] & \G^n/\Lambda \ar[r] & \Pic^0(K) \ar[r] & \Pic(\calO_S) \ar[r] & 0 } in any case, but this is not possible with \Pic^0(K) as it is; we have to replace it by something bigger. It turns out that the right replacement is \Pic(K) / \ggen{[\frakp_{n+1}]}, which is canonically isomorphic to \Pic^0(K) in case \deg \frakp_{n+1} = \gcd(\deg \frakp \mid \frakp \in \calP_K). We then get the diagram \displaystyle \xymatrix{ 0 \ar[r] & T \ar[r] \ar@{^(->}[d] & \Pic^0(K) \ar@{^(->}[d] \ar[r] & \Pic(\calO_S) \ar@{=}[d] & \\ 0 \ar[r] & \G^n/\Lambda \ar[r] & \Pic(K) / \ggen{[\frakp_{n+1}]} \ar[r] & \Pic(\calO_K) \ar[r] & 0 } with exact rows.

The complete result is stated in the following theorem:

Theorem (Infrastructure, Part III).
  1. Let K be a number field. Then the map \Phi :{} & \fRep(K) \to \Pic^0(K), \\ & ([\frakb]_{\sim_S}, (t_1, \dots, t_n)) \mapsto \biggl[ \divisor(\frakb) + \sum_{i=1}^n t_i \frakp_i - \frac{\dots}{\deg \frakp_{n+1}} \frakp_{n+1} \biggr], where \dots equals \deg \divisor(\frakb) + \sum_{i=1}^n t_i \deg \frakp_i, is a bijection.
  2. Let K be a function field. Then the map \Phi :{} & \fRep(K) \to \Pic(K) / \ggen{[\frakp_{n+1}]}, \\ & ([\frakb]_{\sim_S}, (t_1, \dots, t_n)) \mapsto \biggl[ \divisor(\frakb) + \sum_{i=1}^n t_i \frakp_i \biggr] + \ggen{[\frakp_{n+1}]} is a bijection.
Moreover, \Phi|_{\fRep(\calO_S)} is a group homomorphism, where the group structure on \fRep(\calO_S) is the one induced by the bijection \fRep(\calO_S) \to \G^n/\Lambda.
□

Finally, we explicitly describe the group operation induced by this bijection on \fRep(K) without using the bijection itself.

Theorem.
Let \Phi be the bijection from the previous theorem, and let A = ([\fraka]_{\sim_S}, (t_1, \dots, t_n)), A' = ([\fraka']_{\sim_S}, (t'_1, \dots, t'_n)) \in \fRep(K).
  1. Set B_\ell := B(\fraka \fraka', (t_1 + t'_1, \dots, t_n + t'_n, \ell)) \setminus \{ 0 \} for \ell \in \G. There exists a minimal \ell with B_\ell \neq \emptyset, and if \mu is a smallest element of B_\ell with respect to \le, we get \displaystyle B := ([\tfrac{1}{\mu} \fraka \fraka']_{\sim_S}, (t_i + t'_i + \nu_{\frakp_i}(\mu))_{i=1,\dots,n}) \in \fRep(K) with \Phi(A) + \Phi(A') = \Phi(B).
  2. Set B_\ell := B(\fraka^{-1}, (-t_1, \dots, -t_n, \ell)) \setminus \{ 0 \} for \ell \in \G. There exists a minimal \ell with B_\ell \neq \emptyset, and if \mu is a smallest element of B_\ell with respect to \le, we get \displaystyle C := ([\tfrac{1}{\mu} \fraka^{-1}]_{\sim_S}, (-t_i + \nu_{\frakp_i}(\mu))_{i=1,\dots,n}) \in \fRep(K) with -\Phi(A) = \Phi(C).

This shows that the n-dimensional infrastructure we defined has a very close connection to the arithmetic of the divisor class group. This connection was first shown for real hyperelliptic curves by H.-G. Rück and S. Paulus, “Real and Imaginary Quadratic Representations of Hyperelliptic Function Fields”. The first relation between the infrastructure of number fields and the Arakelov divisor class group was described by R. Schoof in his paper Computing Arakelov class groups.

What about… Baby Steps?

As I mentioned, there is no known construction for baby steps in general n-dimensional infrastructures, but there exists a construction for infrastructures obtained from global fields. I want to describe this construction here.

For i \in \{ 1, \dots, n + 1 \} and \fraka \in \Red(K), consider \displaystyle B_\ell := \biggl\{ f \in \fraka \;\biggm|\begin{matrix} \abs{f}_{\frakp_j} \le 1 \text{ for all } j \neq i, \\ \exists j' : \abs{f}_{\frakp_{j'}} < 1, \; \abs{f}_{\frakp_i} \le \ell \end{matrix} \biggr\} \setminus \{ 0 \} for \ell > 0. There exists a minimal \ell such that B_\ell \neq \emptyset. In case \deg \frakp_i = 1, B_\ell contains exactly one k^*-orbit, which gives a unique element \mu \in B_\ell. Otherwise, one has to add an order (lexicographic order as \le above) to chose an element. In any case, define \bs_i([\fraka]_{\sim_S}) := [\frac{1}{\mu} \fraka]_{\sim_S}; then this gives a function \Red(K) \to \Red(K) resp. \Red(\frakb) \to \Red(\frakb) for any \frakb \in \Id(\calO). Opposed to the one-dimensional case, this function neither has to be injective nor surjective, as examples below will show.

We begin with a “small” example: the infrastructure (X_{\calO_S}, d_{\calO_S}) of the function field defined by y^3 = x^6 + x^5 + x^4 + 4 x^2 over \F_7. The red arrows show \bs_1, the blue arrows \bs_2 and the green arrows \bs_3. The small black circles denote usual minima, the larger black circles denote elements of \Lambda, and the shaded areas denote translates of an fundamental parallelepiped of \Lambda:

\setlength{\unitlength}{12cm}\psset{unit=12cm}\begin{picture}(1,1)(0,0)\put(0,0){\white\line(1,0){0.001}}\put(1,1){\white\line(-1,0){0.001}}\put(0,0){\begin{pspicture}(0,0)(1,1) \begin{psclip}{\psframe[linestyle=none](0,0)(1,1)} \pspolygon[linewidth=0pt,fillstyle=solid,fillcolor=lightgray,linecolor=white](0.712334, 1.783083)(0.342881, 1.536781)(0.589183, 0.921026)(0.958636, 1.167328) \pspolygon[linewidth=0pt,fillstyle=solid,fillcolor=lightgray,linecolor=white](-0.026571, 1.290479)(-0.396024, 1.044177)(-0.149722, 0.428423)(0.219731, 0.674724) \pspolygon[linewidth=0pt,fillstyle=solid,fillcolor=lightgray,linecolor=white](0.589183, 0.921026)(0.219731, 0.674724)(0.466032, 0.058970)(0.835485, 0.305272) \pspolygon[linewidth=0pt,fillstyle=solid,fillcolor=lightgray,linecolor=white](-0.149722, 0.428423)(-0.519175, 0.182121)(-0.272873, -0.433634)(0.096580, -0.187332) \pspolygon[linewidth=0pt,fillstyle=solid,fillcolor=lightgray,linecolor=white](1.204938, 0.551573)(0.835485, 0.305272)(1.081787, -0.310483)(1.451240, -0.064181) \pspolygon[linewidth=0pt,fillstyle=solid,fillcolor=lightgray,linecolor=white](0.466032, 0.058970)(0.096580, -0.187332)(0.342881, -0.803087)(0.712334, -0.556785) \psline[linewidth=0.333pt,linecolor=darkgray](0.557594, 1.000000)(0.589183, 0.921026) \psline[linewidth=0.333pt,linecolor=darkgray](0.089620, 1.000000)(0.219731, 0.674724) \psline[linewidth=0.333pt,linecolor=darkgray](0.707644, 1.000000)(0.589183, 0.921026) \psline[linewidth=0.333pt,linecolor=darkgray](0.589183, 0.921026)(0.835485, 0.305272) \psline[linewidth=0.333pt,linecolor=darkgray](0.589183, 0.921026)(0.219731, 0.674724) \psline[linewidth=0.333pt,linecolor=darkgray](0.219731, 0.674724)(0.466032, 0.058970) \psline[linewidth=0.333pt,linecolor=darkgray](0.219731, 0.674724)(0.000000, 0.528237) \psline[linewidth=0.333pt,linecolor=darkgray](0.000000, 0.054117)(0.021647, 0.000000) \psline[linewidth=0.333pt,linecolor=darkgray](1.000000, 0.414948)(0.835485, 0.305272) \psline[linewidth=0.333pt,linecolor=darkgray](0.835485, 0.305272)(0.957594, 0.000000) \psline[linewidth=0.333pt,linecolor=darkgray](0.835485, 0.305272)(0.466032, 0.058970) \psline[linewidth=0.333pt,linecolor=darkgray](0.466032, 0.058970)(0.489620, 0.000000) \psline[linewidth=0.333pt,linecolor=darkgray](0.466032, 0.058970)(0.377578, 0.000000) \pscircle(0.466032, 0.921026){0.01} \pscircle(0.096580, 0.674724){0.01} \pscircle(0.096580, 0.921026){0.01} \pscircle(0.712334, 0.674724){0.01} \pscircle(0.712334, 0.797875){0.01} \pscircle(0.958636, 0.674724){0.01} \pscircle(0.835485, 0.797875){0.01} \pscircle(0.342881, 0.428423){0.01} \pscircle(0.342881, 0.551573){0.01} \pscircle(0.589183, 0.428423){0.01} \pscircle(0.466032, 0.551573){0.01} \pscircle(0.712334, 0.305272){0.01} \pscircle(0.589183, 0.797875){0.01} \pscircle(0.589183, 0.921026){0.01} \pscircle(0.589183, 0.921026){0.01} \pscircle(0.589183, 0.921026){0.013} \pscircle(0.589183, 0.921026){0.016} \pscircle(0.712334, 0.551573){0.01} \pscircle(0.219731, 0.182121){0.01} \pscircle(0.096580, 0.305272){0.01} \pscircle(0.342881, 0.058970){0.01} \pscircle(0.219731, 0.551573){0.01} \pscircle(0.219731, 0.674724){0.01} \pscircle(0.219731, 0.674724){0.01} \pscircle(0.219731, 0.674724){0.013} \pscircle(0.219731, 0.674724){0.016} \pscircle(0.342881, 0.305272){0.01} \pscircle(0.958636, 0.058970){0.01} \pscircle(0.958636, 0.182121){0.01} \pscircle(0.835485, 0.182121){0.01} \pscircle(0.835485, 0.305272){0.01} \pscircle(0.835485, 0.305272){0.01} \pscircle(0.835485, 0.305272){0.013} \pscircle(0.835485, 0.305272){0.016} \pscircle(0.466032, 0.058970){0.01} \pscircle(0.466032, 0.058970){0.01} \pscircle(0.466032, 0.058970){0.013} \pscircle(0.466032, 0.058970){0.016} \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.466032, 1.290479)(0.546080, 1.111910)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.096580, 1.044177)(0.176628, 0.865608)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.342881, 1.044177)(0.472190, 0.994917)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.466032, 0.921026)(0.527608, 0.927184)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.466032, 1.167328)(0.539923, 1.050335)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(-0.026571, 0.797875)(0.102737, 0.748615)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.096580, 0.674724)(0.158155, 0.680882)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.096580, 0.921026)(0.170470, 0.804033)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.712334, 0.674724)(0.792382, 0.496156)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.712334, 0.797875)(0.773910, 0.804033)(0.835485, 0.797875) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.958636, 0.674724)(1.087944, 0.625464)(1.204938, 0.551573) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.835485, 0.797875)(0.903218, 0.742457)(0.958636, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.342881, 0.428423)(0.422930, 0.249854)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.342881, 0.551573)(0.404457, 0.557731)(0.466032, 0.551573) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.589183, 0.428423)(0.718492, 0.379162)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.466032, 0.551573)(0.533765, 0.496156)(0.589183, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.712334, 0.305272)(0.773910, 0.311429)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.589183, 0.797875)(0.650759, 0.804033)(0.712334, 0.797875) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.589183, 0.921026)(0.835485, 0.945656)(1.081787, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.712334, 0.551573)(0.786225, 0.434580)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(-0.026571, 0.182121)(0.053477, 0.003552)(0.096580, -0.187332) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(-0.026571, 0.305272)(0.035004, 0.311429)(0.096580, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.219731, 0.182121)(0.349039, 0.132860)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.096580, 0.305272)(0.164313, 0.249854)(0.219731, 0.182121) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.342881, 0.058970)(0.404457, 0.065127)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.219731, 0.551573)(0.281306, 0.557731)(0.342881, 0.551573) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.219731, 0.674724)(0.466032, 0.699355)(0.712334, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.342881, 0.305272)(0.416772, 0.188278)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(-0.149722, 0.428423)(0.096580, 0.453053)(0.342881, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(-0.026571, 0.058970)(0.047319, -0.058024)(0.096580, -0.187332) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.958636, 0.058970)(1.038684, -0.119599)(1.081787, -0.310483) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.958636, 0.182121)(1.020211, 0.188278)(1.081787, 0.182121) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.835485, 0.182121)(0.897061, 0.188278)(0.958636, 0.182121) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.835485, 0.305272)(1.081787, 0.329902)(1.328089, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkred]{->}(0.466032, 0.058970)(0.712334, 0.083600)(0.958636, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.466032, 0.921026)(0.398299, 0.976444)(0.342881, 1.044177) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.096580, 0.674724)(0.028847, 0.730142)(-0.026571, 0.797875) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.096580, 0.921026)(-0.155880, 0.957971)(-0.396024, 1.044177) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.712334, 0.674724)(0.644601, 0.730142)(0.589183, 0.797875) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.712334, 0.797875)(0.644601, 0.853293)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.958636, 0.674724)(0.761595, 0.779403)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.835485, 0.797875)(0.706177, 0.847136)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(1.081787, 0.551573)(1.014054, 0.606991)(0.958636, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(1.081787, 0.797875)(0.829328, 0.834821)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.342881, 0.428423)(0.275148, 0.483840)(0.219731, 0.551573) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.342881, 0.551573)(0.275148, 0.606991)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.589183, 0.428423)(0.392142, 0.533101)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.466032, 0.551573)(0.336724, 0.600834)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.712334, 0.305272)(0.644601, 0.360690)(0.589183, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.589183, 0.797875)(0.521450, 0.853293)(0.466032, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.589183, 0.921026)(0.502978, 1.161170)(0.466032, 1.413630) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.712334, 0.551573)(0.459875, 0.588519)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.219731, 0.182121)(0.022689, 0.286799)(-0.149722, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.096580, 0.305272)(-0.032729, 0.354532)(-0.149722, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.342881, 0.058970)(0.275148, 0.114388)(0.219731, 0.182121) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.219731, 0.551573)(0.151998, 0.606991)(0.096580, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.219731, 0.674724)(0.133525, 0.914869)(0.096580, 1.167328) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.342881, 0.305272)(0.090422, 0.342217)(-0.149722, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.958636, 0.058970)(0.890903, 0.114388)(0.835485, 0.182121) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.958636, 0.182121)(0.890903, 0.237539)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(1.204938, 0.058970)(1.007896, 0.163648)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(1.081787, 0.182121)(0.952478, 0.231381)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(1.328089, 0.182121)(1.075629, 0.219066)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.589183, -0.064181)(0.521450, -0.008763)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.835485, -0.187332)(0.638444, -0.082654)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.712334, -0.064181)(0.583026, -0.014921)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.835485, 0.182121)(0.767752, 0.237539)(0.712334, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.835485, 0.305272)(0.749279, 0.545416)(0.712334, 0.797875) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.958636, -0.064181)(0.706177, -0.027236)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.466032, -0.064181)(0.398299, -0.008763)(0.342881, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkgreen]{->}(0.466032, 0.058970)(0.379827, 0.299114)(0.342881, 0.551573) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.712334, 1.290479)(0.669231, 1.099595)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.589183, 1.413630)(0.613813, 1.167328)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.835485, 1.167328)(0.724649, 1.031862)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.096580, 1.044177)(0.102737, 0.982602)(0.096580, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.342881, 1.044177)(0.299779, 0.853293)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.219731, 1.167328)(0.244361, 0.921026)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.466032, 0.921026)(0.355197, 0.785560)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.096580, 0.674724)(-0.014256, 0.539258)(-0.149722, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.096580, 0.921026)(0.041162, 0.853293)(-0.026571, 0.797875) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.712334, 0.674724)(0.718492, 0.613149)(0.712334, 0.551573) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.712334, 0.797875)(0.718492, 0.736300)(0.712334, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.958636, 0.674724)(0.915533, 0.483840)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.835485, 0.797875)(0.860115, 0.551573)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(1.081787, 0.551573)(0.970951, 0.416107)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.958636, 1.044177)(0.780067, 0.964129)(0.589183, 0.921026) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(1.081787, 0.797875)(1.026369, 0.730142)(0.958636, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.342881, 0.428423)(0.349039, 0.366847)(0.342881, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.342881, 0.551573)(0.349039, 0.489998)(0.342881, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.589183, 0.428423)(0.546080, 0.237539)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.466032, 0.551573)(0.490663, 0.305272)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.712334, 0.305272)(0.601498, 0.169806)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.589183, 0.797875)(0.410614, 0.717827)(0.219731, 0.674724) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.589183, 0.921026)(0.595341, 0.859451)(0.589183, 0.797875) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.712334, 0.551573)(0.656916, 0.483840)(0.589183, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.219731, 0.182121)(0.176628, -0.008763)(0.096580, -0.187332) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.096580, 0.305272)(0.121210, 0.058970)(0.096580, -0.187332) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.342881, 0.058970)(0.232046, -0.076496)(0.096580, -0.187332) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.219731, 0.551573)(0.041162, 0.471525)(-0.149722, 0.428423) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.219731, 0.674724)(0.225888, 0.613149)(0.219731, 0.551573) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.342881, 0.305272)(0.287464, 0.237539)(0.219731, 0.182121) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.958636, 0.058970)(0.964794, -0.002606)(0.958636, -0.064181) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.958636, 0.182121)(0.964794, 0.120545)(0.958636, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(1.204938, 0.428423)(1.026369, 0.348374)(0.835485, 0.305272) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.835485, 0.182121)(0.656916, 0.102073)(0.466032, 0.058970) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.835485, 0.305272)(0.841643, 0.243696)(0.835485, 0.182121) \pscurve[arrowsize=1.5pt 3,linecolor=darkblue]{->}(0.466032, 0.058970)(0.472190, -0.002606)(0.466032, -0.064181) \end{psclip} \end{pspicture}}\end{picture}

Unfortunately, the second example is too large for WordPress.

]]> http://math.fontein.de/2009/07/21/obtaining-infrastructures-from-global-fields/feed/ 0 One-dimensional Infrastructures. http://math.fontein.de/2009/07/20/one-dimensional-infrastructures/ http://math.fontein.de/2009/07/20/one-dimensional-infrastructures/#comments Mon, 20 Jul 2009 03:45:16 +0000 Felix Fontein http://math.fontein.de/?p=100 One-dimensional infrastructures first appeared in the 1970′s in Daniel Shanks‘ work on real quadratic number fields \Q(\sqrt{D}), D > 1 a squarefree integer, when he tried to fasten regulator computations. The previous algorithms used continued fraction expansion to obtain the regulator in \calO(D^{1/2 + \varepsilon}) binary operation, \varepsilon > 0 arbitrary. Shanks found out that one can obtain a multiplication like operation, which he dubbed giant steps, as opposed to the baby steps taken by one step in the continued fraction expansion. He described a baby step-giant step method to compute the regulator in \calO(D^{1/4 + \varepsilon}) binary operations, requiring \calO(D^{1/4 + \varepsilon}) bytes of storage. His methods were analysed, written up more clearly and extended by various people, including Hendrik Lenstra, Hugh Williams, Johannes Buchmann, Rene Schoof, and many others. Extensions of the method to function fields exist as well, most notably due to the work of Andreas Stein and Renate Scheidler.

I begin with giving an abstract definition of a one-dimensional infrastructure.

Definition (One-dimensional infrastructure).
Let R > 0 be a real number. A one-dimensional infrastructure of circumference R is a pair (X, d), where X \neq \emptyset is a finite set and d : X \to \R/R\Z is an injective map.

If you interpret \R/R\Z as a circle of circumference R (think of it as folding up the real line, such that two numbers whose difference is an integer multiple of R are identified), a one-dimensional infrastructure can be seen as a circle with a finite number of dots on it. The map d gives the distance between 0 \in \R/R\Z and some element x \in X on the circle, whence d is called the distance map.

Now one can define two operations on a one-dimensional infrastructure. Due to Shanks’ nomenclature, these are called baby steps and giant steps. To define a baby step, let x \in X. Then consider the set F_x := \{ f \in \R \mid f > 0, \; d(x) + f \in d(X) \}. It is non-empty as R \in F_x and bounded from below. Moreover, it is discrete as X is finite; therefore, f := \min F_x exists and d(x) + f \in d(X), say d(x) + f = d(x') with x' \in X. In that case, we define \bs(x) := x'. This gives a bijective map \bs : X \to X which, in case \abs{X} > 1, has no fixed points. If \R/R\Z is interpreted as a circle and X identified with d(X), \bs will send each point to the “next one” in positive direction on the circle.

To define giant steps, let x, x' \in X. For that, note that \R/R\Z is naturally a group, whence we can add d(x) and d(x'). Now d(x) + d(x') \in \R/R\Z, but in general d(x) + d(x') \not\in d(X). But we can use a similar trick as in the baby step case: we jump back to the previous point of d(X). For that, define F_{x,x'} := \{ f \in \R \mid f \ge 0, \; d(x) + d(x') - f \in d(X) \}. It is bounded from above, non-empty and discrete, whence f := \max F_{x,x'} exists with d(x) + d(x') - f' \in d(X), say d(x) + d(x') - f = d(y) for y \in X; then we define \gs(x, x') := y. This gives a binary operation \gs : X \times X \to X which is in general not associative.

But even though, we have \displaystyle d(\gs(x, x')) \approx d(x) + d(x') in general, assuming that D := \max\{ d(\bs(x)) - d(x) \mid x \in X \} is small (here, we identify d(\bs(x)) - d(x) \in \R/R\Z with its smallest non-negative representant). More precisely, we have d(\gs(x, x')) + f = d(x) + d(x') with 0 \le f < D, whence the giant step operation is “almost” associative.

Finite Cyclic Groups as One-dimensional Infrastructures.

Let G = \ggen{g} be a finite cyclic group of order R. For h \in G, one can write h = g^n with n \in \Z; note that n = \log_g h \in \Z/R\Z is the discrete logarithm of h with respect to g. Hence, we get the isomorphism G \cong \Z/R\Z induced by \log_g : G \to \Z/R\Z. As \Z/R\Z \subseteq \R/R\Z, we get the injective map d : G \to \R/R\Z, h \mapsto \log_g h, turning (G, d) into a one-dimensional infrastructure.

Let h, h' \in G; then we get \bs(h) = g h and \gs(h, h') = h h', i.e. baby steps are multiplications by the generator g and the giant steps equals the group operation. In particular, this provides an example for giant steps being associative.

Therefore, one-dimensional infrastructures can be seen as generalizations of finite cyclic groups.

Remarks.

Finally, we want to sketch some ideas, which will allow generalizing infrastructures to higher dimensions. For that, let (X, d) be a one-dimensional infrastructure. First, define the map red : \R/R\Z \to X as follows. For r \in \R/R\Z, define F_r := \{ f \in \R \mid f \ge 0, \; r - f \in d(X) \}. Again, F_r is non-empty, bounded from below and discrete, whence f := \min F_r exists and r - f \in d(X), say r - f = d(x) for some x \in X. Define red(r) := x. Then red satisfies red \circ d = \id_X and \gs(x, x') = red(d(x) + d(x')) for all x, x' \in X.

If red' : \R/R\Z \to X would be any other map satisfying red' \circ d = \id_X, one would obtain another giant step function \gs' : X \times X \to X, (x, x') \mapsto red'(d(x) + d(x')). In case X comes from a finite cyclic group, as above, \gs' would again be the group operation. If this is not the case, \gs and \gs' could be two distinct binary operations on X. If red' satisfies d(red'(r)) \approx r for all r \in \R/R\Z, we would also have \displaystyle d(\gs'(x, x')) \approx d(x) + d(x') \text{ for all } x, x' \in X.

This shows that our choice of red is rather random; we could also define red(r) = d^{-1}(d(x) + f) with f = \min \{ f \in \R \mid f \ge 0, \; r + f \in d(X) \}, or chose f such that \abs{f} = \min\{ \abs{f} \mid r + f \in d(X) \}, with some additional condition to rule out ties. Any other arbitrary choice of red is also possible, as long as red \circ d = \id_X is satisfied. We will later see that our definition of red is exactly the one we obtain in a canonical way if we obtain infrastructures from global fields of unit rank one. We call such maps reduction maps.

]]> http://math.fontein.de/2009/07/20/one-dimensional-infrastructures/feed/ 0