Nimbers is a 2-person game, winnable only if you understand the arithmetic of the finite fields $\mathbb{F}_{2^{2^n}} $ associated to Fermat 2-powers.

It is played on a rectangular array (say a portion of a Go-board, for practical purposes) having a finite number of stones at distinct intersections. Here’s a typical position

The players alternate making a move, which is either

• removing one stone, or
• moving a stone to a spot on the same row (resp. the same column) strictly to the left (resp. strictly lower), and if there’s already a stone on this spot, both stones are removed, or
• adding stones to the empty corners of a rectangle having as its top-right hand corner a chosen stone and removing stones at the occupied corners

Here we illustrate two possible moves from the above position, in the first we add two new stones and remove 2 existing stones, in the second we add three new stones and remove only the top right-hand stone.

As always, the last player able to move wins the game!

Note that Nimbers is equivalent to Lenstra’s ‘turning corners’-game (as introduced in his paper Nim-multiplication or mentioned in Winning Ways Chapter 14, page 473).

If all stones are placed on the left-most column (or on the bottom row) one quickly realizes that this game reduces to classical Nim with Nim-heap sizes corresponding to the stones (for example, the left-most stone corresponds to a heap of size 3).

Nim-addition $n \oplus m $ is defined inductively by

$n \oplus m = mex(n’ \oplus m,n \oplus m’) $

where $n’ $ is any element of ${ 0,1,\ldots,n-1 } $ and $m’ $ any element of ${ 0,1,\ldots,m-1 } $ and where ‘mex’ stands for Minimal EXcluded number, that is the smallest natural number which isn’t included in the set. Alternatively, one can compute $n \oplus m $ buy writing $n $ and $m $ in binary and add these binary numbers without carrying-over. It is well known that a winning strategy for Nim tries to shorten one Nim-heap such that the Nim-addition of the heap-sizes equals zero.

This allows us to play Nimber-endgames, that is, when all the stones have been moved to the left-column or the bottom row.

To evaluate general Nimber-positions it is best to add another row and column, the coordinate axes of the array

and so our stones lie at positions (1,3), (4,7), (6,4), (10,3) and (14,8). In this way all legal moves follow the rectangle-rule when we allow rectangles to contain corners on the added coordinate axes. For example, removing a stone is achieved by taking a rectangle with two sides on the added axes, and, moving a stone to the left (or the bottom) is done by taking a rectangle with one side at the x-axes (resp. the y-axes)

However, the added stones on the coordinate axes are considered dead and may be removed from the game. This observation allows us to compute the Grundy number of a stone at position (m,n) to be

$G(m,n)=mex(G(m’,n’) \oplus G(m’,n) \oplus G(m,n’)~:~0 \leq m’ < m, 0 \leq n’ < n) $

and so by induction these Grundy numbers are equal to the Nim-multiplication $G(m,n) = m \otimes n $ where

$m \otimes n = mex(m’ \otimes n’ \oplus m’ \otimes n \oplus m \otimes n’~:~0 \leq m’ < m, 0 \leq n’ < n) $

Thus, we can evaluate any Nimbers-position with stone-coordinates smaller than $2^{2^n} $ by calculating in a finite field using the identification (as for example in the odd Knights of the round table-post) $\mathbb{F}_{2^{2^n}} = \{ 0,1,2,\ldots,2^{2^n}-1 \} $

For example, when all stones lie in a 15×15 grid (as in the example above), all calculations can be performed using

Here, we’ve identified the non-zero elements of $\mathbb{F}_{16} $ with 15-th roots of unity, allowing us to multiply, and we’ve paired up couples $(n,n \oplus 1) $ allowing u to reduce nim-addition to nim-multiplication via

$n \oplus m = (n \otimes \frac{1}{m}) \otimes (m \oplus 1) $

In particular, the stone at position (14,8) is equivalent to a Nim-heap of size $14 \otimes 8=10 $. The nim-value of the original position is equal to 8

Suppose your opponent lets you add one extra stone along the diagonal if you allow her to start the game, where would you have to place it and be certain you will win the game?

In 1956, Alexander Grothendieck (middle) introduced $\lambda $-rings in an algebraic-geometric context to be commutative rings A equipped with a bunch of operations $\lambda^i $ (for all numbers $i \in \mathbb{N}_+ $) satisfying a list of rather obscure identities. From the easier ones, such as

$\lambda^0(x)=1, \lambda^1(x)=x, \lambda^n(x+y) = \sum_i \lambda^i(x) \lambda^{n-i}(y) $

to those expressing $\lambda^n(x.y) $ and $\lambda^m(\lambda^n(x)) $ via specific universal polynomials. An attempt to capture the essence of $\lambda $-rings without formulas?

Lenstra’s elegant construction of the 1-power series rings $~(\Lambda(A),\oplus,\otimes) $ requires only one identity to remember

$~(1-at)^{-1} \otimes (1-bt)^{-1} = (1-abt)^{-1} $.

Still, one can use it to show the existence of ringmorphisms $\gamma_n~:~\Lambda(A) \rightarrow A $, for all numbers $n \in \mathbb{N}_+ $. Consider the formal ‘logarithmic derivative’

$\gamma = \frac{t u(t)’}{u(t)} = \sum_{i=1}^\infty \gamma_i(u(t))t^i~:~\Lambda(A) \rightarrow A[[t]] $

where $u(t)’ $ is the usual formal derivative of a power series. As this derivative satisfies the chain rule, we have

$\gamma(u(t) \oplus v(t)) = \frac{t (u(t)v(t))’}{u(t)v(t)} = \frac{t(u(t)’v(t)+u(t)v(t)’}{u(t)v(t))} = \frac{tu(t)’}{u(t)} + \frac{tv(t)’}{v(t)} = \gamma(u(t)) + \gamma(v(t)) $

and so all the maps $\gamma_n~:~\Lambda(A) \rightarrow A $ are additive. To show that they are also multiplicative, it suffices by functoriality to verify this on the special 1-series $~(1-at)^{-1} $ for all $a \in A $. But,

$\gamma((1-at)^{-1}) = \frac{t \frac{a}{(1-at)^2}}{(1-at)} = \frac{at}{(1-at)} = at + a^2t^2 + a^3t^3+\ldots $

That is, $\gamma_n((1-at)^{-1}) = a^n $ and Lenstra’s identity implies that $\gamma_n $ is indeed multiplicative! A first attempt :

hassle-free definition 1 : a commutative ring $A $ is a $\lambda $-ring if and only if there is a ringmorphism $s_A~:~A \rightarrow \Lambda(A) $ splitting $\gamma_1 $, that is, such that $\gamma_1 \circ s_A = id_A $.

In particular, a $\lambda $-ring comes equipped with a multiplicative set of ring-endomorphisms $s_n = \gamma_n \circ s_A~:~A \rightarrow A $ satisfying $s_m \circ s_m = s_{mn} $. One can then define a $\lambda $-ringmorphism to be a ringmorphism commuting with these endo-morphisms.

The motivation being that $\lambda $-rings are known to form a subcategory of commutative rings for which the 1-power series functor is the right adjoint to the functor forgetting the $\lambda $-structure. In particular, if $A $ is a $\lambda $-ring, we have a ringmorphism $A \rightarrow \Lambda(A) $ corresponding to the identity morphism.

But then, what is the connection to the usual one involving all the operations $\lambda^i $? Well, one ought to recover those from $s_A(a) = (1-\lambda^1(a)t+\lambda^2(a)t^2-\lambda^3(a)t^3+…)^{-1} $.

For $s_A $ to be a ringmorphism will require identities among the $\lambda^i $. I hope an expert will correct me on this one, but I’d guess we won’t yet obtain all identities required. By the very definition of an adjoint we must have that $s_A $ is a morphism of $\lambda $-rings, and, this would require defining a $\lambda $-ring structure on $\Lambda(A) $, that is a ringmorphism $s_{AH}~:~\Lambda(A) \rightarrow \Lambda(\Lambda(A)) $, the so called Artin-Hasse exponential, to which I’d like to return later.

For now, we can define a multiplicative set of ring-endomorphisms $f_n~:~\Lambda(A) \rightarrow \Lambda(A) $ from requiring that $f_n((1-at)^{-1}) = (1-a^nt)^{-1} $ for all $a \in A $. Another try?

hassle-free definition 2 : $A $ is a $\lambda $-ring if and only if there is splitting $s_A $ to $\gamma_1 $ satisfying the compatibility relations $f_n \circ s_A = s_A \circ s_n $.

But even then, checking that a map $s_A~:~A \rightarrow \Lambda(A) $ is a ringmorphism is as hard as verifying the lists of identities among the $\lambda^i $. Fortunately, we get such a ringmorphism for free in the important case when A is of ‘characteristic zero’, that is, has no additive torsion. Then, a ringmorphism $A \rightarrow \Lambda(A) $ exists whenever we have a multiplicative set of ring endomorphisms $F_n~:~A \rightarrow A $ for all $n \in \mathbb{N}_+ $ such that for every prime number $p $ the morphism $F_p $ is a lift of the Frobenius, that is, $F_p(a) \in a^p + pA $.

Perhaps this captures the essence of $\lambda $-rings best (without the risk of getting an headache) : in characteristic zero, they are the (commutative) rings having a multiplicative set of endomorphisms, generated by lifts of the Frobenius maps.

Next time you visit your math-library, please have a look whether these books are still on the shelves : Michiel Hazewinkel‘s Formal groups and applications, William Fulton’s and Serge Lange’s Riemann-Roch algebra and Donald Knutson’s lambda-rings and the representation theory of the symmetric group.

I wouldn’t be surprised if one or more of these books are borrowed out, probably all of them to the same person. I’m afraid I’m that person in Antwerp…

Lately, there’s been a renewed interest in $\lambda $-rings and the endo-functor W assigning to a commutative algebra its ring of big Witt vectors, following Borger’s new proposal for a geometry over the absolute point.

However, as Hendrik Lenstra writes in his 2002 course-notes on the subject Construction of the ring of Witt vectors : “The literature on the functor W is in a somewhat unsatisfactory state: nobody seems to have any interest in Witt vectors beyond applying them for a purpose, and they are often treated in appendices to papers devoting to something else; also, the construction usually depends on a set of implicit or unintelligible formulae. Apparently, anybody who wishes to understand Witt vectors needs to construct them personally. That is what is now happening to myself.”

Before doing a series on Borger’s paper, we’d better run through Lenstra’s elegant construction in a couple of posts. Let A be a commutative ring and consider the multiplicative group of all ‘one-power series’ over it $\Lambda(A)=1+t A[[t]] $. Our aim is to define a commutative ring structure on $\Lambda(A) $ taking as its ADDITION the MULTIPLICATION of power series.

That is, if $u(t),v(t) \in \Lambda(A) $, then we define our addition $u(t) \oplus v(t) = u(t) \times v(t) $. This may be slightly confusing as the ZERO-element in $\Lambda(A),\oplus $ will then turn be the constant power series 1…

We are now going to define a multiplication $\otimes $ on $\Lambda(A) $ which is distributively with respect to $\oplus $ and turns $\Lambda(A) $ into a commutative ring with ONE-element the series $~(1-t)^{-1}=1+t+t^2+t^3+\ldots $.

We will do this inductively, so consider $\Lambda_n(A) $ the (classes of) one-power series truncated at term n, that is, the kernel of the natural augmentation map between the multiplicative group-units $~A[t]/(t^{n+1})^* \rightarrow A^* $.
Again, taking multiplication in $A[t]/(t^{n+1}) $ as a new addition rule $\oplus $, we see that $~(\Lambda_n(A),\oplus) $ is an Abelian group, whence a $\mathbb{Z} $-module.

For all elements $a \in A $ we have a scaling operator $\phi_a $ (sending $t \rightarrow at $) which is an A-ring endomorphism of $A[t]/(t^{n+1}) $, in particular multiplicative wrt. $\times $. But then, $\phi_a $ is an additive endomorphism of $~(\Lambda_n(A),\oplus) $, so is an element of the endomorphism-RING $End_{\mathbb{Z}}(\Lambda_n(A)) $. Because composition (being the multiplication in this endomorphism ring) of scaling operators is clearly commutative ($\phi_a \circ \phi_b = \phi_{ab} $) we can define a commutative RING $E $ being the subring of $End_{\mathbb{Z}}(\Lambda_n(A)) $ generated by the operators $\phi_a $.

The action turns $~(\Lambda_n(A),\oplus) $ into an E-module and we define an E-module morphism $E \rightarrow \Lambda_n(A) $ by $\phi_a \mapsto \phi_a((1-t)^{-1}) = (1-at)^{-a} $.

All of this looks pretty harmless, but the upshot is that we have now equipped the image of this E-module morphism, say $L_n(A) $ (which is the additive subgroup of $~(\Lambda_n(A),\oplus) $ generated by the elements $~(1-at)^{-1} $) with a commutative multiplication $\otimes $ induced by the rule $~(1-at)^{-1} \otimes (1-bt)^{-1} = (1-abt)^{-1} $.

Explicitly, $L_n(A) $ is the set of one-truncated polynomials $u(t) $ with coefficients in $A $ such that one can find elements $a_1,\ldots,a_k \in A $ such that $u(t) \equiv (1-a_1t)^{-1} \times \ldots \times (1-a_k)^{-1}~mod~t^{n+1} $. We multiply $u(t) $ with another such truncated one-polynomial $v(t) $ (taking elements $b_1,b_2,\ldots,b_l \in A $) via

$u(t) \otimes v(t) = ((1-a_1t)^{-1} \oplus \ldots \oplus (1-a_k)^{-1}) \otimes ((1-b_1t)^{-1} \oplus \ldots \oplus (1-b_l)^{-1}) $

and using distributivity and the multiplication rule this gives the element $\prod_{i,j} (1-a_ib_jt)^{-1}~mod~t^{n+1} \in L_n(A) $.
Being a ring-qutient of $E $ we have that $~(L_n(A),\oplus,\otimes) $ is a commutative ring, and, from the construction it is clear that $L_n $ behaves functorially.

For rings $A $ such that $L_n(A)=\Lambda_n(A) $ we are done, but in general $L_n(A) $ may be strictly smaller. The idea is to use functoriality and do the relevant calculations in a larger ring $A \subset B $ where we can multiply the two truncated one-polynomials and observe that the resulting truncated polynomial still has all its coefficients in $A $.

Here’s how we would do this over $\mathbb{Z} $ : take two irreducible one-polynomials u(t) and v(t) of degrees r resp. s smaller or equal to n. Then over the complex numbers we have
$u(t)=(1-\alpha_1t) \ldots (1-\alpha_rt) $ and $v(t)=(1-\beta_1) \ldots (1-\beta_st) $. Then, over the field $K=\mathbb{Q}(\alpha_1,\ldots,\alpha_r,\beta_1,\ldots,\beta_s) $ we have that $u(t),v(t) \in L_n(K) $ and hence we can compute their product $u(t) \otimes v(t) $ as before to be $\prod_{i,j}(1-\alpha_i\beta_jt)^{-1}~mod~t^{n+1} $. But then, all coefficients of this truncated K-polynomial are invariant under all permutations of the roots $\alpha_i $ and the roots $\beta_j $ and so is invariant under all elements of the Galois group. But then, these coefficients are algebraic numbers in $\mathbb{Q} $ whence integers. That is, $u(t) \otimes v(t) \in \Lambda_n(\mathbb{Z}) $. It should already be clear from this that the rings $\Lambda_n(\mathbb{Z}) $ contain a lot of arithmetic information!

For a general commutative ring $A $ we will copy this argument by considering a free overring $A^{(\infty)} $ (with 1 as one of the base elements) by formally adjoining roots. At level 1, consider $M_0 $ to be the set of all non-constant one-polynomials over $A $ and consider the ring

$A^{(1)} = \bigotimes_{f \in M_0} A[X]/(f) = A[X_f, f \in M_0]/(f(X_f) , f \in M_0) $

The idea being that every one-polynomial $f \in M_0 $ now has one root, namely $\alpha_f = \overline{X_f} $ in $A^{(1)} $. Further, $A^{(1)} $ is a free A-module with basis elements all $\alpha_f^i $ with $0 \leq i < deg(f) $.

Good! We now have at least one root, but we can continue this process. At level 2, $M_1 $ will be the set of all non-constant one-polynomials over $A^{(1)} $ and we use them to construct the free overring $A^{(2)} $ (which now has the property that every $f \in M_0 $ has at least two roots in $A^{(2)} $). And, again, we repeat this process and obtain in succession the rings $A^{(3)},A^{(4)},\ldots $. Finally, we define $A^{(\infty)} = \underset{\rightarrow}{lim}~A^{(i)} $ having the property that every one-polynomial over A splits entirely in linear factors over $A^{(\infty)} $.

But then, for all $u(t),v(t) \in \Lambda_n(A) $ we can compute $u(t) \otimes v(t) \in \Lambda_n(A^{(\infty)}) $. Remains to show that the resulting truncated one-polynomial has all its entries in A. The ring $A^{(\infty)} \otimes_A A^{(\infty)} $ contains two copies of $A^{(\infty)} $ namely $A^{(\infty)} \otimes 1 $ and $1 \otimes A^{(\infty)} $ and the intersection of these two rings in exactly $A $ (here we use the freeness property and the additional fact that 1 is one of the base elements). But then, by functoriality of $L_n $, the element
$u(t) \otimes v(t) \in L_n(A^{(\infty)} \otimes_A A^{(\infty)}) $ lies in the intersection $\Lambda_n(A^{(\infty)} \otimes 1) \cap \Lambda_n(1 \otimes A^{(\infty)})=\Lambda_n(A) $. Done!

Hence, we have endo-functors $\Lambda_n $ in the category of all commutative rings, for every number n. Reviewing the construction of $L_n $ one observes that there are natural transformations $L_{n+1} \rightarrow L_n $ and therefore also natural transformations $\Lambda_{n+1} \rightarrow \Lambda_n $. Taking the inverse limits $\Lambda(A) = \underset{\leftarrow}{lim} \Lambda_n(A) $ we therefore have the ‘one-power series’ endo-functor
$\Lambda~:~\mathbf{comm} \rightarrow \mathbf{comm} $
which is ‘almost’ the functor W of big Witt vectors. Next time we’ll take you through the identification using ‘ghost variables’ and how the functor $\Lambda $ can be used to define the category of $\lambda $-rings.