---
title: Generalized Farey Fractions
url: https://www.emergentmind.com/topics/generalized-farey-fractions
type: topic
---

# Generalized Farey Fractions

Searching arXiv for recent and relevant papers on generalized Farey fractions and related extensions.
Searching for Farey sequence generalizations via quotient sets, congruence restrictions, thin groups, and multidimensional maps.
Generalized Farey fractions are extensions of the classical Farey sequence \(F_n\) and its associated maps, graphs, and counting problems in which one changes the admissible rational set, the adjacency relation, the ambient geometry, or the dynamical rule. Recent work studies several distinct extensions: quotient sets of subsets of \(F_n\), denominator-restricted subsequences, Farey fractions with \(k\)-free denominators, Farey subgraphs and continued fractions, generalized Farey sequences for thin groups, multidimensional Farey fractions on simplices and horospheres, and Ford-circle extractions by continuous curves [2005.04429] [2507.00228] [2106.14856] [1907.01854] [1003.4613] [2312.12539]. In all of these settings, the classical invariants—coprimality, mediants, adjacency, equidistribution, and continued-fraction structure—remain central, but they are reorganized by new arithmetic, geometric, or dynamical constraints.

## 1. Classical structure and the main directions of extension

The classical Farey sequence of order \(n\) is
\[
F_n=\left\{\frac{a}{b}:0\le a\le b\le n,\ \gcd(a,b)=1\right\},
\]
ordered increasingly [2005.04429]. Its basic local structure is governed by the mediant
\[
\frac{a}{b}\oplus\frac{c}{d}=\frac{a+c}{b+d},
\]
which lies strictly between \(\frac{a}{b}\) and \(\frac{c}{d}\), and by the adjacency condition for consecutive Farey fractions,
\[
bc-ad=1
\]
[2011.06318]. A broader “fraction space” viewpoint treats reduced fractions \(p/q\) with \(p,q>0\), together with \(0/1\) and \(1/0\), as a primary object and extends adjacency to all of this space by
\[
\left|\frac{p}{q},\frac{r}{s}\right|=|qs-pr|,\qquad \frac{p}{q}\,L\,\frac{r}{s}\iff |qs-pr|=1
\]
[1507.03952].

Within this broader framework, mediants and adjacency are no longer features of one bounded sequence alone. Every positive reduced fraction arises uniquely as the mediant of exactly one pair of adjacent nonnegative fractions, and iterative mediant subdivision of an adjacent-bounded interval generates all fractions in that interval without repetition [1507.03952]. This gives a unifying baseline against which later generalizations modify either the admissible set of fractions, the rule for adjacency, or the ambient space.

| Direction | Defining object | Representative source |
|---|---|---|
| Subset quotients | \(\mathcal Q(S)=\{x/y:x,y\in S,\ x\le y,\ y\neq 0\}\) inside \(F_n\) | [2005.04429] |
| Arithmetic restrictions | Denominator congruences, \(k\)-free denominators, truncated ranges | [2502.19881], [2507.00228], [2404.08283] |
| Graph and continued-fraction variants | Farey subgraphs \(\mathcal F_N\), well-directed paths, graph-based expansions | [2106.14856] |
| Thin-group generalization | Orbit-based generalized Farey sequences for Fuchsian groups | [1907.01854] |
| Multidimensional extensions | \(\mathbf p/q\in[0,1)^{d-1}\), projective simplex maps, Farey polyhedra | [1003.4613], [2506.02984], [2410.13091] |
| Geometric extraction | Fractions extracted from Ford circles by continuous curves | [2312.12539] |

## 2. Quotient-set generalizations and rigidity inside \(F_n\)

One precise generalization replaces the full Farey sequence by a subset \(S\subseteq F_n\) and studies its quotient set
\[
\mathcal{Q}(S)=\left\{\frac{x}{y}:x,y\in S,\ x\le y,\ y\neq 0\right\},
\]
with all fractions taken in reduced form [2005.04429]. The basic problem is to determine when \(\mathcal Q(S)\) remains inside \(F_n\), and when it can generate the whole sequence.

The central boundedness theorem states that if \(S\subseteq F_n\) and \(\mathcal Q(S)\subseteq F_n\), then \(|S|\le n+1\) [2005.04429]. The same paper proves that this statement is equivalent to Graham’s GCD Conjecture 1. This equivalence is structurally significant: it shows that a quotient-closure condition on rational subsets of \(F_n\) is tuned exactly to a maximal pairwise-\(\gcd\) phenomenon for finite sets of integers.

The extremal cases are rigid. If \(\mathcal Q(S)=F_n\), or if \(\mathcal Q(S)\subseteq F_n\) and \(|S|=n+1\), then one must have
\[
S=\left\{0,1,\frac12,\ldots,\frac1n\right\}
\]
or
\[
S=\left\{0,1,\frac1n,\ldots,\frac{n-1}{n}\right\},
\]
except for \(n=4\), where there is an additional set
\[
\left\{0,1,\frac12,\frac13,\frac23\right\}
\]
in the second case [2005.04429]. A common misconception is that many large subsets of \(F_n\) should behave well under pairwise quotients. The theorem shows the opposite: maximal quotient-stable or quotient-generating subsets are essentially forced into two explicit shapes, with one small exceptional case.

## 3. Arithmetic restrictions, local statistics, and analytic distribution

A major family of generalized Farey fractions is obtained by imposing arithmetic restrictions on denominators. In one direction, fractions in the classical Farey sequence \(\Phi_Q\) are “colored” by a congruence condition \(q\equiv c\pmod D\), and one studies gaps whose endpoints are colored and whose interior contains no other colored fractions [2502.19881]. The limit proportion of such gaps with exactly \(r\) intermediate fractions is denoted \(\nu(r;D,c)\). The BCZ-transform \(T\) on the Farey triangle \(\mathcal T=\{(x,y):0<x,y\le 1,\ x+y>1\}\) organizes the admissible denominator patterns, and the limiting proportions are expressed through areas of polygons \(\mathcal T(k_1,\ldots,k_r)\) [2502.19881].

For \(D=2, c=0\), explicit formulas include
\[
\nu(1;2,0)=6-8\ln 2,\qquad
\nu(2;2,0)=16\ln 2-\frac{1132}{105},
\]
and for \(r\ge 5\),
\[
\nu(r;2,0)=\frac{8}{(2r-3)(2r-1)(2r+1)}
\]
[2502.19881]. For \(D=3\), the sequel computes complete explicit formulas for \(c=1,2\); for \(r\ge 6\),
\[
\nu(r;3,c)=\frac{8}{3(2r-5)(2r-3)(2r-1)},
\]
and for fixed \(r=o\!\left((Q/\ln Q)^{1/3}\right)\),
\[
\nu(Q;r,3,c)=\nu(r;3,c)+O\!\left(\frac{\ln Q}{Q}\right)
\]
uniformly in \(r\) [2503.08176]. These results generalize gap statistics from the full Farey sequence to denominator residue classes.

A second arithmetic restriction selects Farey fractions with \(k\)-free denominators in an arithmetic progression:
\[
\mathscr{F}_{Q,k}^{(m)}:=\left\{\frac{a}{q}\ \middle|\ 1\le a\le q\le Q,\ \gcd(a,q)=1,\ q\text{ is }k\text{-free},\ q\equiv b\pmod m\right\}
\]
[2507.00228]. This sequence is equidistributed modulo one, has discrepancy
\[
D_{\mathcal N(Q,k,m)}\big(\mathscr F_{Q,k}^{(m)}\big)\asymp \frac1Q,
\]
admits explicit \(\nu\)-level correlation measures, and has a limiting pair correlation function \(\mathfrak g_{m,k}(\lambda)\) given by an explicit series [2507.00228]. Its local statistics are not Poissonian. The same paper proves an analogue of Franel–Landau: GRH holds if and only if, for any \(\epsilon>0\),
\[
\sum_{j=1}^{\mathcal N(Q,k,m)} R_{\mathcal N(Q,k,m)}(\gamma_j)=O\!\left(Q^{1/2+\epsilon}\right)
\]
[2507.00228].

Other arithmetic refinements focus on localized structure within \(F_n\). For fractions with fixed numerator \(h\),
\[
N_n(h)=n\frac{\varphi(h)}{h}-\varphi(h)-\sum_{d\mid h}\mu(d)\left\{\frac nd\right\}+\delta_{1h},
\]
and for the truncated Farey sequence \(F_n^{1/k}\), the rank of \(1/k\) is
\[
I_n(1/k)=n\sum_{j=1}^{[n/k]}\frac{\varphi(j)}{j}-k\sum_{j=1}^{[n/k]}\varphi(j)-\sum_{j=1}^{[n/k]}\sum_{d\mid j}\mu(d)\left\{\frac nd\right\}
\]
[2404.08283]. Partial Franel sums further localize discrepancy: for \(N=\mathrm{lcm}(1,\dots,i)\),
\[
P_{[0,i/N]}=O(\log(N)\delta_B(\log N))
\]
[1802.07792]. Taken together, these works shift Farey theory from a single ordered list to a family of arithmetic subensembles with their own rank formulas, gap laws, and correlation structures.

## 4. Farey subgraphs, paths, and continued-fraction variants

Another line of generalization is graph-theoretic. For \(N\in\mathbb N\), the Farey subgraph \(\mathcal F_N\) has vertex set
\[
\mathcal X_N=\left\{\frac pq: q>0,\ \gcd(p,q)=1,\ N\mid q\right\}\cup\{\infty\},
\]
with an edge between \(\frac pq\) and \(\frac rs\) when \(|ps-qr|=N\) [2106.14856]. The decisive connectivity theorem states that \(\mathcal F_N\) is connected if and only if \(N=1\) or \(N\) is a prime power. When \(N=2^\ell\), \(\mathcal F_N\) is a tree; for odd prime powers it is connected but not a tree [2106.14856].

These graphs support a corresponding continued-fraction theory. An \(\mathcal F_N\)-continued fraction is linked to a well-directed path from \(\infty\) in \(\mathcal F_N\), and finite expansions are in one-to-one correspondence with such paths [2106.14856]. For every real number \(x\) and prime power \(N=p^\ell\), there exists an \(\mathcal F_{p^\ell}\)-continued fraction expansion of \(x\). Uniqueness holds for \(N=2^\ell\), whereas for other prime powers rational points may have finitely many distinct expansions [2106.14856]. This makes explicit that the classical uniqueness paradigm for continued fractions is not stable under all Farey-type restrictions.

The Farey graph also supports broader combinatorial generalizations. Paths from \((1,0)\) to \((0,1)\), equivalently finite Stern–Brocot subtrees, form an operad; the paper on the Farey fractal isolates “coronas,” especially spiked such paths, and proves that
\[
S_R=\{(x,y)\in\mathbb N^+\times\mathbb N^+:\gcd(x,y)=1,\ x+y\le R\}
\]
is a corona [2105.14537]. More generally, sets defined by fundamentally monotone constraints such as \(\alpha x+\beta y\le R\), \(\max\{x,y\}\le R\), or \(x^p+y^p\le R^p\) are accommodated in the same framework [2105.14537]. A related strand identifies continued fractions, polygon triangulations, and walks in the Farey tessellation, extending Conway–Coxeter and Series to \(PSL(2,\mathbb Z)\) via polygon dissections and canonical presentations of modular-group elements [1811.01229].

## 5. Multidimensional, projective, and group-theoretic extensions

In higher dimension, Farey fractions become primitive lattice directions. For \(d\ge 2\),
\[
\mathcal F_Q=\left\{\frac{\mathbf p}{q}\in[0,1)^{d-1}:(\mathbf p,q)\in \widehat{\mathbb Z}^d,\ 0<q\le Q\right\},
\]
and these points may be embedded as \(n_-(\mathbf r)\Phi^t\) in large horospheres of \(\Gamma\backslash SL(d,\mathbb R)\) [1003.4613]. The resulting equidistribution statement depends on arithmetic commensurability: multidimensional Farey fractions become uniformly distributed in \(\Gamma\backslash G\) if and only if \(\Gamma\) is not commensurable with \(SL(d,\mathbb Z)\) [1003.4613]. This places generalized Farey fractions inside homogeneous dynamics rather than solely inside one-dimensional Diophantine orderings.

A distinct multidimensional extension concerns projective continued-fraction algorithms on the standard simplex
\[
\Delta=\{x\in\mathbb R^{n+1}:\sum_{i=0}^n x_i=1,\ x_i\ge 0\}.
\]
If \(A_0,A_1\in GL(n+1,\mathbb Z)\) are nonnegative and \(A_0[\Delta],A_1[\Delta]\) split \(\Delta\) into two sub-simplices, then, up to the natural action of the symmetric group, there are precisely three simplex-splitting, topologically contractive two-map projective IFSs in every dimension [2506.02984]. Only one yields a continuous Gauss-type map: the Farey–Mönkemeyer map [2506.02984]. The classification gives a precise sense in which the classical Farey map has a unique proper multidimensional analogue under these hypotheses.

Generalized Farey sequences also arise from thin Fuchsian groups. If \(\Gamma<PSL(2,\mathbb R)\) has a cusp at \(\infty\) and \(\mathcal Z=(0,1)\Gamma\), then
\[
\mathcal F_Q=\left\{\frac pq\in[0,1):(p,q)\in\mathcal Z,\ 0<q<Q\right\}
\]
defines a generalized Farey sequence [1907.01854]. These sequences equidistribute, their gap distribution converges, and in one sparse Ford-configuration example the limiting gap law is given explicitly [1907.01854]. The count satisfies
\[
|\mathcal F_Q|\sim c_\Gamma Q^{2\delta_\Gamma},
\]
and the associated analogue of Gauss measure is ergodic for the generalized Gauss map, yielding Gauss–Kuzmin statistics in this thin-group setting [1907.01854].

A further geometric generalization is the multidimensional Farey summation algorithm. It produces Farey polyhedra and sails, classifies them by prismatic diagrams, and leads to multidimensional Conway–Coxeter frieze patterns satisfying generalized Ptolemy relations [2410.13091]. Here the classical Farey tessellation is replaced by a tessellation of the positive orthant by Farey simplices, and continued fractions are encoded by higher-dimensional geometric invariants rather than by scalar partial quotients alone [2410.13091].

## 6. Ford circles, determinant matrices, and other reformulations

Ford circles provide another general mechanism for producing Farey-type sets. A continuous curve can extract the fractions whose Ford circles it meets, and every ordered sequence that satisfies the Farey sum and has two adjacent fractions can be extracted from Ford circles through continuous curves [2312.12539]. For inclined lines of slope \(1/m\), the extracted sequence is
\[
F_{1/m}=\left\{\frac pq:(p,q)=1,\ q\ge p,\ pq\le m\right\},
\]
which replaces the usual denominator bound by the multiplicative constraint \(pq\le m\) [2312.12539].

The growth of these sets is governed by prime-factor data. If
\[
S_m:=|F_{1/m}|-|F_{1/(m-1)}|,
\]
then
\[
S_m=2^{\omega(m)-1},
\]
where \(\omega(m)\) is the number of distinct prime divisors of \(m\), and hence
\[
|F_{1/m}|=\sum_{j=2}^m 2^{\omega(j)-1}+2
\]
[2312.12539]. The same paper gives three approximations, including
\[
|F_{1/m}|\approx \frac{m}{2\zeta(2)}(\ln m+1),
\]
showing \(m\ln m\)-type growth [2312.12539]. This is a genuine generalization of Farey fractions in which the extraction rule is geometric and the counting law is controlled by the Möbius and prime-omega functions rather than by a sharp denominator cutoff.

Matrix reformulations encode Farey arithmetic globally. For Farey fractions \(h_i/k_i\) and \(h_j/k_j\), the Farey determinant
\[
d_{ij}(N)=h_jk_i-h_ik_j
\]
is the numerator of their difference, and the matrix \(d(N)\) is skew-symmetric, symmetric about the secondary diagonal, and has rank \(2\) for \(N>2\) [1809.08642]. Its diagonal bands recover higher Farey indices,
\[
\nu(x_i)=d_{i-1,i+1},\qquad V_k(x_i)=d_{i-1,i+k-1},
\]
and it realizes the Hall–Shiu gcd symmetry
\[
\gcd(d_{k,j},d_{k,i})=\gcd(d_{k,j},d_{j,i})=\gcd(d_{k,i},d_{j,i})
\]
for \(i>j>k\) [1809.08642]. Some higher-order determinant matrices contain lower-order ones as block submatrices, giving a self-similar representation of Farey structure [1809.08642].

The additive-slow-Farey map supplies yet another reinterpretation: it can be used to generate integer partitions of \(n\) into two parts with multiplicity, and its multidimensional analogue, the additive-slow-Triangle map, generates partitions into three parts with multiplicity [2109.08962]. A plausible implication is that generalized Farey fractions are not only new sets of rationals; they also function as templates for transferring mediant and subtractive dynamics into combinatorial settings far from the original ordered list of reduced fractions.

Across these developments, generalized Farey fractions form a family of theories rather than a single definition. What persists is the Farey core: primitive lattice data, local unimodularity, mediant-type recursion, and strong rigidity under arithmetic or dynamical constraints. What changes is the ambient category—subsets, congruence classes, graphs, group orbits, simplices, polyhedra, Ford-circle configurations, or partition dynamics—and with it the appropriate notions of spacing, coding, and asymptotic distribution.

Source: https://www.emergentmind.com/topics/generalized-farey-fractions