# The invertible local factor of an Eichler order of level p^r

10 November 2026, America/Los_Angeles.

[project, proved]  An elementary orbit and volume count, not new deep mathematics, and not an RH claim.  Prior art was checked in the sources listed below and the explicit Dirichlet polynomial was not found there; if it appears in work that was missed, the priority is theirs.

## Statement

Let p be a prime and r a positive integer.  Let O be the standard Eichler order of level p^r in M_2(Q_p),

    O = { [[a, b], [p^r c, d]] : a, b, c, d in Z_p }.

Write a_e for the number of invertible right O-ideals of reduced norm p^e (equivalently, of index p^{2e}).  Then

    a_e = (e+1) p^e                         for 0 <= e < r,

    a_e = [ ((r+1)p - (r-1)) p^e - (p+1) p^{r-1} ] / (p-1)
                                            for e >= r.

The Dirichlet generating function, relative to the maximal-order factor 1/((1-u)(1-p u)) with u = p^{-2s}, is therefore the degree-r polynomial

    Q_r^{inv}(u) = 1 + p^r u^r + (p-1) sum_{j=1}^{r-1} p^{j-1} u^j

(for r=1 this is 1+p u).  It is self-dual: p^r u^r Q(1/(p^2 u))=Q(u).  Its value at the pole u=p^{-2} is 1+1/p, matching Voight's local mass factor lambda(O,p) for a residually split Eichler order, independently of r.

Together with the unit-circle argument in `order-eichler-unitcircle-proof.md`, this proves part (1) of the Block 9 project conjecture.

## Proof

Every invertible right ideal of a quaternion order is locally principal (Voight, 16.2-16.3).  Thus every invertible right O-ideal of reduced norm p^e is of the form alpha O for some alpha in O with v_p(det alpha)=e.  Two such generators give the same right ideal if and only if they differ by right multiplication by a unit of O.  So a_e is the number of right O^x-orbits on {alpha in O : v_p(det alpha)=e}.

The group O^x consists of those matrices in O with a and d in Z_p^x (the determinant is then automatically a unit, because v_p(p^r c b) >= r >= 1).  Parametrise O by the coordinates (a,b,c,d) in Z_p^4, and take the additive Haar measure with vol(Z_p)=1, so vol(O)=1 and vol(O^x)=(1-1/p)^2.

The orbit map phi: O^x -> O, u |-> alpha u, extends to the Z_p-linear map m_alpha: O -> O, u |-> alpha u.  This map is injective (alpha is invertible in M_2(Q_p)), and its image is alpha O, of index [O : alpha O] = nrd(alpha)^2 = p^{2e}.  Hence it has constant Jacobian p^{-2e}, and each orbit has volume p^{-2e} vol(O^x).  Therefore

    a_e = p^{2e} / (1-1/p)^2 * V_e,
    V_e = vol{ (a,b,c,d) in Z_p^4 : v_p(a d - p^r c b) = e }.

The volume V_e is computed by splitting on k = v_p(a).

If 0 <= k < r and k <= e, write a = p^k a_0 with a_0 a unit.  Then a d - p^r c b = p^k (a_0 d - p^{r-k} c b).  As d runs through Z_p, the term in parentheses runs through all of Z_p, so the condition that it have valuation e-k has volume p^{-(e-k)}(1-1/p), independently of b and c.  The slice of a with valuation k has volume p^{-k}(1-1/p).  The contribution of this k is therefore (1-1/p)^2 p^{-e}.  There are min(r, e+1) such values of k (namely k=0,1,...,min(r-1,e)), each contributing the same amount.

In particular, if e < r, only these slices appear, there are e+1 of them, and

    V_e = (e+1)(1-1/p)^2 p^{-e},
    a_e = (e+1) p^e,

which is the first formula.

If e >= r, the same low slices k=0,...,r-1 contribute r(1-1/p)^2 p^{-e}, hence r p^e ideals, and it remains to compute the contribution of k >= r.  Write k = r+s with s >= 0, and n = e-r, so the condition becomes v_p(p^s a_0 d - c b) = n.  The unit a_0 may be absorbed into d (the substitution d |-> a_0 d preserves Haar measure on Z_p).  Let Y(s,n) be the volume of the set of (b,c,d) with v_p(p^s d - b c)=n.

To compute Y, split on i = v_p(b).  (Organising on v_p(b c) instead leads to an awkward unit-cancellation at p=2; the split on a single coordinate does not.)  Write U=(1-1/p)^2 p^{-n}.

If i < s, the factor p^i can be pulled out and the unit part of b absorbs into c, leaving a shell of exact valuation n-i in the c-line, provided n>=i; the contribution of each such i is U, and is 0 if n<i.  If i=s and n>=s, the same volume U appears, now as a shell in the d-line.  If i>s and n>=s, the condition is a shell for d around a multiple of p^{i-s}, of d-volume p^{-(n-s)}(1-1/p), and summing over all i>s (a set of b-volume p^{-(s+1)}) contributes (1-1/p) p^{-n-1}.  If n<s, the cases i>=s contribute nothing.

Thus, if s>n, only the indices i=0,...,n (all <s) contribute, and Y(s,n)=(n+1)U=(n+1)(p-1)^2 p^{-n-2}.  If s<=n, the indices i=0,...,s contribute (s+1)U and the tail i>s contributes (1-1/p)p^{-n-1}, which together simplify to Y(s,n)=(s(p-1)+p)(p-1)p^{-n-2}.  The two formulae disagree at s=n, and the second (the s<=n line) is the correct value there.  Both formulae hold for p=2.

The contribution of all s>=0 is V_high = sum_{s>=0} p^{-s-r}(1-1/p) Y(s,n).  The inner sums are elementary.  For the tail s>=n+1, Y is independent of s, and sum_{s>=n+1} p^{-s} = p^{-(n+1)}/(1-1/p), giving (n+1)(p-1) p^{-2n-2} after the remaining prefactors.  For the finite range s=0,...,n one uses the standard evaluations sum p^{-s} = p/(p-1)(1-p^{-n-1}) and sum s p^{-s} = p/(p-1)^2 (1-(n+1)p^{-n}+n p^{-n-1}).  Collecting terms and restoring n=e-r yields

    V_high = (1 - p^{-2}) (p^{-e} - p^{r-1-2e}),

and therefore the high slices contribute

    a_e^{high} = p^{2e}/(1-1/p)^2 * V_high = (p+1)/(p-1) (p^e - p^{r-1}).

Adding the r p^e ideals from the low slices gives the second formula for a_e.  (The locus a=0 has measure zero and does not contribute.)

It remains only to pass from a_e to Q.  Let A(u) = sum_{e>=0} a_e u^e, and write b_e for the coefficients of (1-u)(1-p u) A(u), so b_e = a_e - (p+1) a_{e-1} + p a_{e-2} with a_negative=0.  Direct substitution of the two formulae for a_e gives:

- b_0 = 1;
- b_1 = p-1 for r > 1 (and b_1 = p for r=1);
- b_e = (p-1) p^{e-1} for 2 <= e <= r-1;
- b_r = p^r;
- b_e = 0 for every e > r.

Thus (1-u)(1-p u) A(u) is exactly Q_r^{inv}(u), as claimed.  Self-duality of Q is immediate from the palindrome of its coefficients, and Q(p^{-2}) = 1 + p^r p^{-2r} + (p-1) sum_{j=1}^{r-1} p^{j-1} p^{-2j} unfolds to 1+1/p by summing the geometric series.

The case r=1 recovers the known hereditary count a_e = (2 p^{e+1} - p - 1)/(p-1) and the factor 1+p u (Denert, Bushnell-Reiner).

## Prior art

Voight, 26.4-26.6, treats the zeta function of a maximal or hereditary order, and the mass of a general definite order, including the local factor lambda(O,p)=1+1/p used above; those sections do not give an explicit local Dirichlet polynomial for the invertible right ideals of a non-hereditary Eichler order of level p^r.  Denert 1990, and the type-(1,1) calculations of Bushnell-Reiner 1981, treat hereditary (squarefree-level) orders.  Brzezinski's classification of Bass and Gorenstein quaternion orders identifies the type of O, but not this generating function.  Bushnell-Reiner 1980 guarantees that a local zeta of this kind is a rational function, not the coefficients.  Hijikata-Pizer-Shemanske representatives are used for traces and class numbers; the explicit series A(u) was not found in the sources checked.

The formula is consistent with the double-computed counts already obtained at r=1 (known), r=2 (p=2,3,5), r=3 (p=2,3, and p=5 through degree 2), and r=4 (p=2,3).  It is an elementary count, not a new structural theorem about quaternion zeta functions.
