Con empo a y Ma hema ics
Fixed Subg oups In F ee G oups: A Su ey
E. Ven u a
Abs ac . This no e is a su ey o he main esul s known abou ixed sub-
g oups o endomo phisms o ini ely gene a ed ee g oups. A his o ic poin
o iew is aken, emphasizing he e olu ion o his line o esea ch, om i s
beginning o he p esen ime. The a icle concludes wi h a sec ion con aining
he main open p oblems and conjec u es, wi h some commen s and discussions
on hem.
1. In oduc ion and no a ion
The pu pose o his no e is o su ey wha is known abou ixed subg oups o
ini ely gene a ed ee g oups. Desc ip ions, commen s and ema ks on he p oo s o
known esul s will be p esen ed ins ead o he p oo s hemsel es, o which we e e
he eade o he co esponding o iginal pape s. We will also ake he oppo uni y
o e iew he his o y o his pa icula opic in G oup Theo y, which goes back o
he 1970’s. Howe e , i is no ou in en ion o be exhaus i e in his su ey. We
apologize o hose au ho s who con ibu ed o his line o esea ch and a e no ci ed
he e. In sec ions 2-7 we will adop a his o ic poin o iew, desc ibing he esul s
ch onologically as hey appea ed in he li e a u e, e en when some o hem imp o e
olde ones. Finally, in sec ion 8, a lis o some conjec u es and open p oblems in
his a ea o esea ch will be p o ided and discussed.
The whole pape is abou ini ely gene a ed ee g oups. F om now on, Fnwill
deno e a ee g oup o ank n≥0, while Fwill be used o deno e an a bi a y ee
g oup (possibly wi h in ini e ank). I is well known ha e e y subg oup o a ee
g oup is also ee (see, o example, Theo em I.8.4 in [11]). Bu , in gene al, i s ank
can be la ge han he ank o he ambien g oup. In ac , i is easy o see ha he
subg oup H=hb− ab , ∈Zio F2=ha, biis ee o coun ably in ini e ank. So,
Fℵ0is a subg oup o F2. The e o e, e e y ee g oup o ini e o coun ably in ini e
ank can be iewed as a subg oup o e e y o he , wi h he ob ious excep ion o he
i ial g oup, and o he ee g oup wi h ank 1, which is he g oup o in ege s. Also,
i is well known ha he in e sec ion o ini ely gene a ed subg oups o a ee g oup
is again ini ely gene a ed (see [30] o Theo em I.8.8 in [11]), while i s ank can be
o he o de o he p oduc o he wo anks. This beha io , which is comple ely
di e en om wha happens in o he mo e classical algeb aic con ex s, mo i a ed
a lo o esea ch dedica ed o compu e o bound he ank o se e al subg oups
c
0000 (copy igh holde )
1
2 E. VENTURA
in di e en si ua ions. We a e in e es ed in he case in ol ing ixed subg oups o
endomo phisms o Fn.
The educed ank o a ee g oup F, deno ed ˜ (F), is max{0, (F)−1}, ha is,
one less han he ank, excep o he i ial g oup whe e he educed ank coincides
wi h he ank, which is ze o. So, a ee g oup has educed ank ze o i and only i
i is cyclic.
Le End(Fn) deno e he monoid o endomo phisms o Fn, and Au (Fn) he
g oup o au omo phisms o Fn(so, Au (Fn) is he g oup o uni s o End(Fn)). Le
Inj(Fn) deno e he se o injec i e endomo phisms o Fn, a submonoid o End(Fn)
con aining Au (Fn). Finally, we deno e by Ou (Fn) he g oup o ou e au omo -
phisms o Fn, ha is, Au (Fn) modulo he (no mal) subg oup o conjuga ions, also
called inne au omo phisms. Fo each u∈Fn, we le uagain deno e he co e-
sponding igh conjuga ion, u:Fn→Fn,x7→ xu=u−1xu.
We le elemen s o End(Fn) ac on he igh o Fnand, i he e is no isk o
con usion, we will omi he pa en hesis o he a gumen . Thus, xφ deno es he
image o xunde φ, and xφ1φ2deno es (xφ1)φ2.
Fo any S⊆End(Fn), le Fix Sdeno e he se consis ing o he elemen s o
Fnwhich a e ixed by e e y elemen o S(wi h he con en ion ha Fix S=Fn
when Sis emp y). Then, Fix Sis a subg oup o Fn, called he ixed subg oup o
So he subg oup ixed by S. When Sis a single on, S={φ}, we simply w i e
Fix φins ead o Fix {φ} o he ixed subg oup o φ. So, Fix S=∩φ∈SFix φ.
Clea ly, i S⊆End(Fn) and Mis he submonoid o End(Fn) gene a ed by S hen
Fix S= Fix M.
Following he no a ion in oduced in [38], a subg oup H≤Fnis called endo-
ixed i H= Fix S o some S⊆End(Fn). I Scan be chosen o lie in Inj(Fn)
( esp. Au (Fn)) we u he say ha His a mono- ixed ( esp. au o- ixed) subg oup
o Fn. And i Scan be chosen o be a single on, S={φ} o some φ∈End(Fn)
( esp. φ∈Inj(Fn), φ∈Au (Fn)) we say mo e explici ly ha His a 1-endo- ixed
( esp. 1-mono- ixed,1-au o- ixed) subg oup o Fn.
Le φ∈End(Fn). We will concen a e on hose esul s conce ning Fix φas
a subg oup o Fnand, o ins ance, only a ew commen s will be made abou he
nice ecen esul s on in ini e wo ds ixed by he ex ension o φ o he bounda y o
Fn. Also, we will no be conce ned in gene al abou isola ed ixed wo ds, o abou
simila esul s in non- ee g oups.
The s udy o ixed subg oups in ee g oups began in 1975 wi h he pape [15]
by J. Dye and P. Sco . Since hen, new esul s ha e been appea ing cons an ly.
I is ag eed ha he main and deepes esul in his line o esea ch is he Bes ina-
Handel Theo em, published in [2]. Fo his eason, he p esen pape is o ganized
as ollows. In sec ion 2 we e iew he o iginal mo i a ions, as well as he i s pa ial
esul s. In sec ion 3 we e iew he main esul s abou ixed subg oups ob ained in
he pe iod 1982-1992, be o e he Bes ina-Handel Theo em. Sec ion 4 is en i ely
dedica ed o he Bes ina-Handel heo y. Sec ion 5 is abou he esul s ob ained by
using he Bes ina-Handel Theo em and by gene alizing i . Sec ion 6 conside s he
maximal ank case, while sec ion 7 is dedica ed o he concep o ine ia, p obably
he s onges p ope y cu en ly known o ixed subg oups o ini ely gene a ed
ee g oups. Finally, sec ion 8 con ains a lis o conjec u es and open p oblems in
his line o esea ch, as well as se e al commen s and discussions.
FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 3
2. The Sco conjec u e
In 1975, J. Dye and P. Sco published he pape [15]. Using a heo em o
Ka ass–Pie owski–Soli a abou he s uc u e o ee-by- ini e g oups, he au ho s
ob ained he ollowing esul :
Theo em 2.1 (Dye -Sco , [15]).I Gis a ini e g oup o au omo phisms o a
ee g oup F, hen Fix Gis a ee ac o o F. In pa icula , o e e y ini e o de
au omo phism φ∈Au (Fn), we ha e (Fix φ)≤n.
This pape can be conside ed as he s a ing poin o he line o esea ch abou
ixed subg oups in ini ely gene a ed ee g oups. In [15], he au ho s men ioned
ha i was no known i Fix Gis ini ely gene a ed o a bi a y subg oups G≤
Au (Fn). In iew o his esul , P. Sco conjec u ed ha his is he case a leas
o single au omo phisms (i.e. when Gis cyclic). His main mo i a ion was he
ollowing classical esul due o J. Nielsen:
Theo em 2.2 (Nielsen, [41]).Le Gpbe he undamen al g oup o a closed
o ien able su ace wi h genus p≥1, and le Hbe he subg oup ixed by some
au omo phism o Gp. Then, ei he H=Gpo His ee wi h ank a mos 2p−1.
Fu he mo e, i His no cyclic, hen Gphas a se o 2pgene a o s such ha some
subse o i gene a es H.
In 1975, no example was known o a 1-au o- ixed subg oup o Fnwi h ank
bigge han n. So, one could also conjec u e ha (Fix φ)≤n o e e y φ∈
Au (Fn). The ini eness o his ank o he p e ious inequali y can also be s udied
o endomo phisms ins ead o au omo phisms, and mo e gene ally, o a bi a y se s
o endomo phisms. Soon, all hese s a emen s became gene ically e e ed o as he
Sco conjec u e. And ime jus i ied i , since i is now known ha (Fix S)≤n o
e e y subse S⊆End(Fn).
In 1977, W. Jaco and P. B. Shalen published [34], whe e hey s udied he
ixed and pe iodic subg oups o hose au omo phisms o he undamen al g oup
o a compac connec ed su ace Tinduced by homeomo phisms o he su ace T.
Such an au omo phism is called geome ic o T. In he case whe e Tis closed and
o ien able, his p o ides an al e na i e and en i ely di e en p oo o Nielsen’s
Theo em (using he ac ha , o hese pa icula su aces, e e y au omo phism o
π1(T) is geome ic o T). Fo he cases whe e Thas bounda y, he Jaco-Shalen
Theo em gi es he ollowing co olla y:
Theo em 2.3 (Jaco-Shalen, [34]).Le φbe an au omo phism o Fn. I φis
geome ic ( o some compac connec ed su ace), hen Fix φis ei he cyclic o a
ee ac o o Fn. In pa icula , (Fix φ)≤n.
Howe e , his esul did no comple ely sol e he Sco conjec u e because,
as J. S allings showed in [44], he e exis au omo phisms o ini ely gene a ed ee
g oups which a e no geome ic ( o any compac connec ed su ace). Conc e ely, J.
S allings in oduced he concep o PV-au omo phism o Fn:φ∈Au (Fn) is a PV-
au omo phism when he absolu e alue o all he eigen alues o i s abelianiza ion
a e less han 1, excep exac ly one which is la ge han 1. Wi h a simple homological
a gumen , he showed ha no powe o any PV-au omo phism is geome ic. The
simples such au omo phism is he one in F3=ha, b, cigi en by a7→ b,b7→ c,
c7→ ab. Fu he mo e, J. S allings also conjec u ed ha any PV-au omo phism o
4 E. VENTURA
Fnwi h n≥3 has i ial ixed subg oup. This was known as he S allings PV-
conjec u e, and i p o ided new e idence in suppo o he Sco conjec u e. As a
mo i a ion o his conjec u e, J. S allings p o ed (bu did no w i e) ha he ixed
subg oup o a PV-au omo phism o Fnis con ained in he hi d e m in he lowe
cen al se ies o Fn(see §1 in [46]).
The ollowing s ep, in 1982, was he wo k [20] due o S. Ge s en (and published
in 1984). He e, he au ho conside ed ano he amily o au omo phisms o Fn, wi h
mo i a ions om one-dimensional geome y, i.e. om g aphs. Le Fnbe iewed
as he undamen al g oup o a bouque Ywi h npe als. Fo e e y g aph Xand
e e y maximal sub ee Tin X, one can collapse T o a poin and ob ain X/T,
which is a bouque wi h as many pe als as he ank o ( he undamen al g oup o )
X. The e is also he na u al p ojec ion map X→X/T inducing an isomo phism
a he undamen al g oup le el. An isomo phism φo Fn=π(Y, ∗) is called a
change o maximal ee au omo phism (CMT-au omo phism o sho ) when he e
exis s a ( ank n) g aph X, wo maximal sub ees T, T 0in X, iden i ica ions o X/T
and X/T0wi h Y, and a e ex in Xsuch ha he co esponding isomo phisms
ηT:π(X, )→π(X/T, ∗)≃Fand ηT0:π(X, )→π(X/T 0,∗)≃Fsa is y η−1
TηT0=
φ. In [20], S. Ge s en sol ed he Sco conjec u e o CMT-au omo phisms p o ing
he ollowing esul , using only combina o ial a gumen s:
Theo em 2.4 (Ge s en, [20]).Le φbe an au omo phism o Fn. I φis a
CMT-au omo phism, hen (Fix φ)≤n.
Despi e being ini e o e e y n≥1, he amily o CMT-au omo phisms o Fn
con ains in e es ing elemen s. Fo example, all Whi ehead au omo phisms and all
Squie ’s skew-Nielsen au omo phisms o Fna e CMT-au omo phisms.
In addi ion, S. Ge s en ga e a me hod such ha , when i e mina es, i com-
pu es he ixed subg oup o a CMT-au omo phism. Howe e , his me hod is no
eally an algo i hm because i can go in o an in ini e loop. The au ho analyzed
some pa icula cases and p o ided he example gi en by he au omo phism o
Fn=hx1, . . . , xni,n≥2, gi en by xi7→ xi+1x1 o i= 1, . . . , n −1 and xn7→ x1,
which is simul aneously a CMT-au omo phism and a PV-au omo phism, and has
i ial ixed subg oup. This was he i s in ini e amily o examples o which he
PV-conjec u e was known o be ue.
3. Fixed subg oups be o e Bes ina-Handel
In he same yea o i s publica ion, an addendum o [20] appea ed wi h mo e
in o ma ion. Fi s , he au ho ga e a simple e ised e sion o his me hod o
compu e ixed subg oups o CMT-au omo phisms. Then, he announced a p oo o
he Sco conjec u e in gene al. This p oo was ou lined in [21] bu he de ails did
no appea un il 1987 in he pape [22]. The idea in ol es s udying composi ions
o Whi ehead au omo phisms. These au omo phisms gene a e Au (Fn), and hey
a e CMT-au omo phisms so Ge s en’s me hod applies o hem. A e a ca e ul
combina o ial analysis, he au ho deduces ha Fix φis ini ely gene a ed o e e y
φ∈Au (Fn). Then, using Howson’s Theo em, he same can be said o ini ely
gene a ed g oups o au omo phisms o Fn:
Theo em 3.1 (Ge s en, [22]).Le Gbe a ini ely gene a ed g oup o au omo -
phism o Fn. Then, Fix Gis ini ely gene a ed.
FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 5
This sol ed he ques ion o iginally asked by P. Sco o he ini ely gene a ed
case. Howe e , when one s udies he composi ion o CMT-au omo phisms, one loses
con ol on he uppe bound o he ank o he ixed subg oup, and he a gumen
gi en by S. Ge s en says only ha his subg oup is ini ely gene a ed. In [22], an
uppe bound o (Fix φ) is gi en, bu i depends on he au omo phism φi sel and
can be a bi a ily la ge, in con as wi h he uni o m bound (Fix φ)≤nknown
a he ime o pe iodic, geome ic and CMT-au omo phisms o Fn.
Du ing he pe iod 1982-1987, using comple ely di e en echniques, o he au-
ho s ob ained independen p oo s o Ge s en’s Theo em, as well as se e al ex en-
sions o i . And due o he nuances o ma hema ical publica ion, many o Ge -
s en’s successo s saw hei wo k published be o e his pape appea ed. Fo example,
see [24], [25], [26], [10] and [46], e iewed below.
In 1983, J. S allings published [45], whe e some powe ul g aph- heo e ical
echniques, e.g. he pull-back o g aphs, we e i s in oduced and used o gi e
al e na i e p oo s o classical esul s on ee g oups. A he end o [45], he au ho
en u ed a gene aliza ion o he Sco conjec u e, which has become known as he
S allings conjec u e o he equalize conjec u e. Gi en wo g oups G, G0, and wo
homomo phisms ϕ, φ:G→G0, he equalize o ϕand φ, deno ed Eq(ϕ, φ), is he
maximal subg oup o Gwhe e bo h homomo phisms coincide,
Eq(ϕ, φ) = {x∈G|xϕ =xφ}.
The S allings conjec u e says ha i ϕ, φ:F→F0a e wo homomo phisms o
ee g oups, Fis ini ely gene a ed, and ϕis injec i e, hen Eq(ϕ, φ) is also ini ely
gene a ed. No e ha his conjec u e is s onge han he Sco conjec u e since, i
ϕis an isomo phism, hen Eq(ϕ, φ) = Fix φϕ−1. No e also ha his s a emen is
easily seen o be alse when bo h mo phisms ha e non- i ial ke nel. Conside he
ee g oup F2=ha, biand he wo endomo phisms ϕ, φ gi en by aϕ = 1, bϕ =b
and aφ = 1, bφ =b−1; clea ly, he equalize Eq(ϕ, φ) is he no mal closu e o ain
F2, which has in ini e ank. In [22], ano he example o his ac is p oduced bu
wi h he common image o he equalize being no ini ely gene a ed.
Assuming ha ϕmus be injec i e, one can es ic he a en ion o he case
whe e Fis a subg oup o F0and ϕis he inclusion. Then, Eq(ϕ, φ) = Fix φ,
and φ:F→F0is an a bi a y homomo phism. This se ing is mo e gene al han
wo king wi h endomo phisms o F0, because he exis ence o an ex ension o φ o
an endomo phism o he whole o F0is no equi ed.
Simila ly o he case o ixed subg oups, one can also ask (o conjec u e)
whe he he ank o he equalize o ϕ, φ:F→F0is bounded abo e by ha o
F, when ϕis injec i e. O e en mo e, one can ask i (Eq(S)) ≤ (F) o e e y
se So mo phisms om F o F0con aining a leas one injec i e mo phism. As
we will see below, he S allings conjec u e was p o ed some yea s la e , while hese
s onge ques ions a e s ill open oday. We will gene ically e e o all o hem as
he equalize conjec u e.
A he 1983 AMS Summe mee ing, J. S allings poin ed ou ha he g aph
cons uc ed and used by S. Ge s en o p o e he ini eness o he ank o 1-au o-
ixed subg oups o Fn, could also be cons uc ed using he amilies o su aces
desc ibed by he 3-dimensional model cons uc ed by Whi ehead in he 1930’s (see
[54] and [55]). R. Golds ein and E. Tu ne de eloped his idea in a se ies o wo
pape s, [24] and [25]. As is said in he in oduc ion o he i s pape , i so happens

6 E. VENTURA
ha he use o h ee dimensions p o ides a eedom ha makes i easie o desc ibe
and cons uc Ge s en’s g aph, and o p o e i s main p ope ies. In he i s o hese
wo pape s, published in 1984, hey ga e an al e na i e p oo o Ge s en’s Theo em
and hey imp o ed Ge s en’s uppe bound o he ank o he ixed subg oup o a
gi en au omo phism o Fn. In he second one, which appea ed in 1985, hey wen
u he using he same echniques and p o ed he S allings conjec u e o he case
o wo injec i e homomo phisms.
Fu he mo e, in 1986, ano he pape by R. Golds ein and E. Tu ne appea ed,
[26]. He e, he au ho s imp o ed hei p e ious esul , sol ing comple ely he
S allings conjec u e:
Theo em 3.2 (Golds ein-Tu ne , [26]).Le Fand F0be wo ee g oups wi h
F ini ely gene a ed. I ϕ, φ:F→F0a e wo mo phisms and ϕis injec i e hen
Eq(ϕ, φ)is ini ely gene a ed.
In his case, he p oo is qui e simple, sho and independen o he p e ious
ones, and i p o es a s onge esul . In he si ua ion whe e Fis a subg oup o F0
and ϕis he inclusion, he au ho s p o ide a pa icula desc ip ion o he co e ing
space Xo a bouque co esponding o he inclusion Fix φ≤F. Then, hey choose
an app op ia e o ien a ion on he edges o Xand use i o show ha his g aph
has ini e ank, and hen so does Fix φ.
In 1987, D. Coope published [10] wi h ano he p oo o Ge s en’s Theo em.
This ime, he me hods used a e o a opological and dynamical na u e. They a e
inspi ed by wo k o W. Thu s on abou su ace g oups. The idea was o ex end
a gi en au omo phism o Fn o a homeomo phism o i s end comple ion ˆ
Fn, a
compac me ic space whe e Fnis dense. The au ho hen showed ha he ixed
poin se o his ex ension is ini ely gene a ed in a ce ain opological sense, which
implies he ini e gene a ion o he subg oup ixed by he o iginal au omo phism.
In pa inspi ed by D. Coope ’s wo k, h ee mo e pape s appea ed, de eloping
new concep s and esul s in ol ing in ini e wo ds. These wo ks made i clea ha
a close connec ion be ween he appa en ly di e en p oo s o [10] and [26] exis s.
Two o hem a e [31] and [32], bo h published in 1990. In [32], W. Im ich and E.
Tu ne conside ed an a bi a y homomo phism φ:F→F0 om a subg oup Fo
an a bi a y ee g oup F0, o F0. By ex ending Golds ein-Tu ne echniques, hey
ob ained new uppe bounds o he ank o Fix φ, and o he ank o he ixed
poin se (`a la Coope ) o an injec i e φ. These bounds a e exp essed in e ms o a
Nielsen educed basis o F.
The hi d pape men ioned abo e is [6], by M.M. Cohen and M. Lus ig, and
published in 1989. In his in e es ing wo k, he au ho s ex ended he Golds ein-
Tu ne me hod by imposing a ec o ield (i.e. a p e e ed o ien a ion o he edges o
he i s ba ycen ic subdi ision) on he g aph cons uc ed in [26]. They in oduced
he concep s o a ac i e edges, epulsi e edges, a ac ing ixed in ini e wo ds and
a ac ing ixed poin s a in ini y, and in es iga ed he esul ing dynamics.
In [6], he au ho s ga e wo explici o mulas o he exac alue o he ank
o he ixed subg oup o a gi en au omo phism o Fn, one in geome ic e ms and
he o he in a comple ely algeb aic con ex . In gene al, i is no easy o make
use o hese o mulas o conc e e examples and, also, hey do no seem o gi e
enough in o ma ion o de i e a uni o m uppe bound o he ank o he 1-au o-
ixed subg oups o Fndepending only on n. Howe e , hey applied hei me hod
FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 7
o ob ain an explici algo i hm o compu ing he ixed subg oup o a posi i e
au omo phism o Fn. Recall ha a posi i e au omo phism is one o which he e
exis s a basis o Fnwhose images a e posi i e wo ds, see sec ion 6 in [6]. They also
ob ained some use ul esul s o which he ollowing a e a sample:
Theo em 3.3 (Cohen-Lus ig, [6]).Le {x1, . . . , xn}be a basis o he ee g oup
Fn, and le φ∈Au (Fn). I ideno es he numbe o occu ences o xiin ( he
no mal o m o ) xiφo he o m xiφ= ·xi·w, wi h , w ∈Fnand he abelianiza ion
o wbelonging o he image o Id −φab, hen (Fix φ)≤Pn
i=1 i. In pa icula , i
o e e y i= 1, . . . , n,xidoes no occu in xiφ, hen Fix φ= 1.
Theo em 3.4 (Cohen-Lus ig, [6]).Le {x1, . . . , xn}be a basis o he ee g oup
Fn, and le φ∈Au (Fn)be an au omo phism such ha x1φ, . . . , xnφa e posi i e
wo ds. I no wo d xiφbegins o ends wi h xi, hen Fix φ= 1.
Theo em 3.5 (Cohen-Lus ig, [6]).Le Φ∈Ou (Fn)be an ou e au omo phism
o Fn. Then, (Fix φ)≤1 o e e y φ∈Φexcep o hose belonging o a ini e
numbe o conjugacy classes o elemen s in Φ.
The e is also ano he ex ension o Ge s en’s Theo em in a di e en di ec ion,
due o J. S allings in 1987. The pape [46], de eloped simul aneously o [25]
o Golds ein-Tu ne , p o ided a p oo o he ollowing a ia ion o he S allings
conjec u e:
Theo em 3.6 (S allings, [46]).Le Fand F0be wo ee g oups wi h F ini ely
gene a ed. Le ϕ, φ:F→F0be wo homomo phisms wi h ke ϕ= ke φ. Then he
(common) image o Eq(ϕ, φ)unde ϕo φis ini ely gene a ed. In pa icula , i ϕ
and φa e injec i e hen Eq(ϕ, φ)is ini ely gene a ed.
He e, J. S allings de eloped his dyads and he g aphic echniques o olds and
ladde a achmen s. He used hese echniques o build an ex ension o Ge s en’s
p oo sui able o s udy he mo e gene al si ua ion ea ed in his pape . So, [46]
ocuses on he g aphic and combina o ial poin o iew, wi hou in ol ing ei he 3-
dimensional a gumen s o in ini e wo ds (al hough i con ains an appendix ela ing
he dyads o Heegaa d diag ams in 3-mani olds).
Addi ionally, [46] con ains a lis o he main ques ions and conjec u es abou
his opic a ha ime. This lis con ains he abo e men ioned ques ions abou n
being a uni o m uppe bound o he ank o 1-au o- ixed, 1-endo- ixed, au o- ixed
o endo- ixed subg oups o Fn, o he co esponding ones o he equalize o a
amily o wo o mo e homomo phisms (one o hem being injec i e). The e is also
ano he in e es ing ques ion ela ing hese p oblems wi h subg oups o Fn×Fn.
Fu he mo e, one inds he pape s [27] and [28], published by A.H.M. Hoa e
in 1988 and 1990, espec i ely. He e, he Whi ehead 3-dimensional model ([54]
and [55]), Ge s en’s g aph ([22]), he Golds ein-Tu ne cons uc ion, ([24], [25])
and he dyads o S allings ([46]) a e conside ed and ela ed o each o he . A he
same ime, he au ho o e ed some a ia ions and ex ensions o all hese cons uc-
ions, as well as a combina o ial p oo o he classical Whi ehead cu e ex lemma
using hem. Howe e , no new esul s abou ixed subg oups we e gi en.
I is also wo h men ioning he ex emely sho pape [48] published by S.
Thomas in 1988. Using p e iously known esul s om A.G. Howson, J. S allings
8 E. VENTURA
and J. McCool, as well as Ge s en’s Theo em, he au ho de ini ely closed he o igi-
nal ques ion o Dye -Sco in [15] by d opping he hypo hesis on Gin Theo em 3.1:
Theo em 3.7 (Thomas, [48]).Le Gbe an a bi a y g oup o au omo phisms
o Fn. Then, Fix Gis ini ely gene a ed.
4. The wo k o Bes ina-Handel
The nex s ep in his his o y is he deepes and mos ele an one in he whole
line o esea ch abou ixed subg oups in ee g oups. In 1988 M. Bes ina and M.
Handel announced a p oo o he s onge e sion o he Sco conjec u e, namely
he uppe bound n o he ank o 1-au o- ixed subg oups o Fn:
Theo em 4.1 (Bes ina-Handel, [2]).Le φbe an au omo phism o Fn. Then,
(Fix φ)≤n.
This esul p o ides he i s (and he bes possible) uni o m uppe bound o
(Fix φ), in con as wi h he p e iously known ones, all o hem being ei he no
uni o m o alid only o special amilies o au omo phisms. This esul appea ed
in published o m, wi h a delay o almos ou yea s, in he i y-page pape [2] in
he Annals o Ma hema ics in 1992. Again, some applica ions and imp o emen s
o he Bes ina-Handel Theo em appea ed in p in be o e [2] was published.
This excellen and dense wo k de eloped a new and powe ul g aphic heo y o
ep esen a bi a y au omo phisms o Fn, and ended up wi h a p oo o Sco con-
jec u e as a consequence. I has a highly opological la o and i is no o iously ha d
o ead. I was inspi ed in he classical wo k o W. Thu s on on homeomo phism
o su aces.
W. Thu s on, ollowing a p og am s a ed many yea s be o e by J. Nielsen,
ga e a classi ica ion o homeomo phisms o a compac su ace S, up o iso opy.
Fu he mo e, he cons uc ed a pa icula ly e icien ep esen a i e o e e y iso opy
class, ha is, o e e y elemen o he mapping class g oup o he su ace, MCG(S).
This ep esen a i e can always be chosen o be ei he o ini e o de , o pseudo-
Anoso , o educible (i.e. cons uc ible in a simple way, om homeomo phisms o
simple su aces, which a e ei he o ini e o de o pseudo-Anoso ). Thu s on’s
wo k was i s desc ibed in he p ep in [49] in 1976, h ee yea s la e in [16] ( he
p oceedings o a semina held a O say in 1978), and inally published wel e yea s
la e in [50]. See also [5] o an in oduc ion, and [3] o an algo i hmic p oo o
Thu s on’s Theo em, based on [2].
Mapping class g oups o compac su aces, MCG(S), and ou e au omo phism
g oups o ee g oups, Ou (Fn), a e simila in many ways. In [2], M. Bes ina
and M. Handel de eloped a p ojec analogous o Thu s on’s, eplacing su aces by
g aphs, and homeomo phisms by homo opy equi alences, a so o one-dimensional
e sion o Thu s on’s wo k. Thei main esul was he cons uc ion o a g aph
Zand an e icien homo opy equi alence β:Z→Z, inducing a p e iously gi en
ou e au omo phism Φ ∈Ou (Fn) a he undamen al g oup le el. He e, e iciency
consis s o some good p ope ies con olling he amoun o cancella ion appea ing
in he images o single edges o Zunde i e a es o β.
One has o hink o Fnas he undamen al g oup o he bouque Rn, a g aph
wi h a single e ex and nedges. A ma ked g aph is a connec ed g aph Z oge he
wi h a homo opy equi alence ρ:Rn→Z. Then, e e y homo opy equi alence
FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 9
β:Z→Zo a ma ked g aph Zinduces an au omo phism o Fnup o conjuga ion,
ha is, an ou e au omo phism Φ ∈Ou (Fn). I is hen said ha βis a opological
ep esen a i e o Φ, o ha βinduces Φ.
In he i s pa o [2], he au ho s analyzed he case o i educible ou e au o-
mo phisms. An ou e au omo phism Φ ∈Ou (Fn) is said o be educible i he e
exis ee ac o s G1, . . . , Gko Fnwhose conjugacy classes a e pe mu ed by Φ, and
such ha G1∗ · · · ∗ Gkis s ill a ee ac o o Fn; o he wise, i is called i educible.
See [13] o a classi ica ion o he simples i educible au omo phisms. M. Bes ina
and M. Handel p o ed ha e e y i educible au omo phism can be opologically
ep esen ed by a ain ack ha is, a homo opy equi alence β:Z→Zsuch ha
βkis locally injec i e in he in e io o e e y edge (i.e., no cancella ion appea s in
he i e a e images o e e y edge eβk, o e∈EZ, and k≥1). Using hese ain
acks, hey p o ed he ollowing wo esul s:
Theo em 4.2 (Bes ina-Handel, [2]).Le Φ∈Ou (Fn)be an i educible ou e
au omo phism o Fn. Then, (Fix φ)≤1 o e e y φ∈Φ.
Theo em 4.3 (Bes ina-Handel, [2]).Le Φ∈Ou (Fn)be such ha Φkis
i educible o e e y k≥1. I he e exis a cyclic wo d ssuch ha sΦ = so
sΦ = s−1, hen Φis geome ically ealized by a pseudo-Anoso homeomo phism o
a compac su ace wi h one bounda y componen .
The o me is a s onge e sion o he Sco conjec u e o i educible au o-
mo phisms. The la e was quickly used by S. Ge s en and J. S allings o sol e he
PV-conjec u e (see [23] and he nex sec ion).
In sec ion 5 o [2], M. Bes ina and M. Handel ex ended hei heo y om i -
educible o a bi a y au omo phisms o Fn. Any homo opy equi alence β:Z→Z
de ines a na u al il a ion o Zby (no necessa ily connec ed) β-in a ian sub-
g aphs, ∅=Z0< Z1<· · · < Z =Z, such ha Zi−1is a maximal p ope
β-in a ian subg aph o Zi,i= 1, . . . , . The subg aph cl(Zi Zi−1) is called
he i- h s a um. I is p o en ha an ou e au omo phism is i educible p ecisely
when all i s opological ep esen a i es o e co e g aphs (i.e. ini e g aphs wi h-
ou e ices o alence one) con aining no non- i ial β-in a ian o es s, con ain
no non- i ial β-in a ian subg aphs. In his case, he co esponding il a ion has
only one s a um.
Now, hey in oduced he no ion o ela i e ain ack which, up o echnical
de ails, is a na u al ela i iza ion o he no ion o ain ack, wi h espec o he
p e ious il a ion. Wi h a conside able amoun o echnical wo k, he au ho s
ended up wi h he esul ha e e y au omo phism o Fnadmi s a opological
ep esen a i e which is a ela i e ain ack. Then, analyzing Nielsen pa hs, i.e.
pa hs in he g aph ixed up o homo opy, hey inally ob ained he desi ed p oo
o he Sco conjec u e.
The p ocess o building such a good opological ep esen a i e is con olled by
a ce ain ma ix. He e, he Pe on-F obenius Theo em (see Theo em 8.4.4 in [29]
o II.1 in [14]) plays a cen al ole. A squa e non-nega i e in ege ma ix Mis
called i educible when o e e y en y (i, j) he e exis s k≥0 such ha he (i, j)-
h en y o Mkis posi i e; o he wise, Mis educible. Now, he Pe on-F obenius
Theo em s a es ha e e y i educible ma ix Mhas he ollowing p ope ies: (i) i
has a unique eigen alue wi h maximum modulus, (ii) his eigen alue is eal, i.e. he
16 E. VENTURA
maximum- ank 1-au o- ixed subg oups o Fn. He e, Fab
ndeno es he abelianiza ion
o Fn, and o a, b ∈Fnwe w i e [a, b] = a−1b−1ab.
Theo em 6.2 (Collins-Tu ne , [9]).Le Hbe a subg oup o Fn, and le m
deno e he ank o he ( ee abelian) image o Hin Fab
n(called he abelian ank o
H). The ollowing a e equi alen :
a) His a maximum- ank 1-au o- ixed subg oup o Fn,
b) he e exis s a basis {x1, . . . , xn}o Fnsuch ha , se ing Fl=hx1, . . . , xli
o 0≤l≤n, he e exis s a basis {y1, . . . , yn}o Hsuch ha o 1≤j≤
m,yj=xj, and o m+1 ≤k≤n,yk= [wk, xk] o some wk∈H∩Fk−1
no being a p ope powe o any elemen o Fn(so, in pa icula , wk6= 1).
In his e en , {y1, . . . , yl}is a basis o H∩Fl, o 0≤l≤n. Fu he mo e,
e e y au omo phism φ∈Au (Fn)wi h H≤Fix φis o he o m xj7→ xjand
xk7→ w k
kxk o some non-ze o in ege s k,1≤j≤m,m+ 1 ≤k≤n.
The ypical example o a maximum- ank 1-au o- ixed subg oup is gi en by he
ollowing au omo phism o F2=ha, bi:a7→ a,b7→ a b, whe e is an in ege .
I s ixed subg oup is H=ha, [a, b]i=ha, b−1abi, and m= 1, excep when = 0
(in which case i is he whole F2). Howe e , o n≥3, he ixed subg oup o an
au omo phism o he ype desc ibed in he las pa ag aph o he p e ious heo em
can be bigge han H. Fo example, le be a non-ze o in ege and conside
he au omo phism φ o F3=ha, b, cigi en by a7→ a,b7→ ab,c7→ a c. Fo
6= 1 we ha e Fix φ =ha, b−1ab, c−1aci, which has abelian ank m= 1. Bu
Fix φ1is bigge since i also con ains b−1c, and has abelian ank m= 2 (in ac ,
Fix φ1=ha, b−1c, b−1abi=ha, b−1c, c−1aci).
Addi ionally, in [9] he e is also an analogous esul desc ibing he ixed sub-
g oups o au omo phisms o ee p oduc g oups, wi h maximal Ku oˇs ank.
Finally, he pape [53], published in 1997 by E. Ven u a, also conside ed he
case o maximal ank. The main esul in [53] was he ollowing:
Theo em 6.3 (Ven u a, [53]).Among he s ic ly ascending chains o maxi-
mum- ank 1-au o- ixed subg oups o Fn, he maximum leng h is exac ly n.
The p oo uses he Collins-Tu ne desc ip ion o maximum- ank 1-au o- ixed
subg oups o Fn. Essen ially, i is a g aphic p oo and i in ol es imme sions and
co e ings o g aphs, as well as S allings oldings. Also, a simple g aphic p oo o an
old esul om M. Takahasi was p o ided. Fo he pa icula case n= 2, and being
Ha maximum- ank 1-au o- ixed subg oup o F2, a desc ip ion was gi en o all hose
subg oups K≤F2sa is ying (K) = (H∩K) = 2. The main consequence de i ed
om his esul was he ollowing heo em, which is clea ly only alid in he ank
2 case:
Theo em 6.4 (Ven u a, [53]).Le Sbe a non-emp y se o non-iden i y en-
domo phisms o F2such ha (Fix S) = 2. Then, S⊆Au (F2), and Fix S=
ha, b−1abi o some basis {a, b}o F2, and Fix S= Fix φ o each φ∈S.
In pa icula , his implies ha e e y g oup o au omo phisms G≤Au (F2) wi h
(Fix G) = 2 is ei he i ial o in ini e cyclic. The pape [53] ends by showing
ha , in he ank 2 case, he amilies o 1-au o- ixed, 1-endo- ixed, au o- ixed and
endo- ixed subg oups do coincide. As men ioned abo e, in [40] an example is gi en
o a 1-endo- ixed no 1-au o- ixed subg oup o Fn, o n≥3.

FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 17
I is no known in gene al whe he he amilies o 1-au o- ixed and au o- ixed
(o 1-endo- ixed and endo- ixed) subg oups o Fncoincide (see 8.9 and 8.10 below).
Howe e , he pape [38] ( e iewed la e ) ga e a posi i e solu ion o his p oblem
in he maximal ank case, as a consequence o he main esul he e. The esul is
he ollowing:
Theo em 6.5 (Ma ino-Ven u a, [38]).Le H≤Fnbe a subg oup o Fnwi h
(H) = n. The ollowing a e equi alen :
a) His a 1-au o- ixed subg oup o Fn,
b) His a 1-mono- ixed subg oup o Fn,
c) His a 1-endo- ixed subg oup o Fn,
d) His an au o- ixed subg oup o Fn,
e) His a mono- ixed subg oup o Fn,
) His an endo- ixed subg oup o Fn.
7. Ine ia and ixed subg oups
In he monog aph [14], published in 1996 by W. Dicks and E. Ven u a, a new
concep was in oduced: ine ia o subg oups o a ee g oup. The main esul in
his wo k was he ine ia p ope y o mono- ixed subg oups o Fn, a s onge esul
han he Sco conjec u e o hose subg oups. This esul is a pa ial solu ion o
he unsol ed equalize conjec u e. A na u al pa icula case o ha conjec u e was
he mo i a ion o look a he no ion o ine ia.
The gene al equalize conjec u e says ha he equalize o wo homomo phisms
ϕ, φ:Fm→Fn, one o hem being injec i e, has ank bounded by m. I , o ex-
ample, ϕis injec i e hen he e is no lose o gene ali y in assuming ha Fmis
a subg oup o Fnand ϕis he inclusion. In his case, we ha e only one homo-
mo phism o conside , φ:Fm→Fn,Fm≤Fn, and Eq(ϕ, φ) = Fix φ. So, he
equalize conjec u e is he Sco conjec u e gene alized o pa ial endomo phisms,
i.e. homomo phisms om a ce ain subg oup o Fn, o Fn.
The case Fm=Fnis he Sco conjec u e al eady sol ed by Bes ina-Handel.
The simples s ep o conside nex is he case when φ:Fm→Fnex ends o an
endomo phism ˜
φo Fn; in his case, Fix φ=Fm∩Fix ˜
φ. We can ese his si ua ion
by aking φ∈End(Fn), and an a bi a y subg oup K≤Fn, and asking whe he
(K∩Fix φ)≤ (K). This was he o iginal mo i a ion o he ollowing de ini ion.
A subg oup H≤Fnis called ine when (K∩H)≤ (K) o e e y subg oup
K≤Fn. O cou se, aking K=Fn, i H≤Fnis ine hen (H)≤n.
Easy examples o ine subg oups o Fna e he cyclic ones, and he ee ac o s
o Fn. No so ob ious is he ac ha e e y ank 2 subg oup o Fnis ine . This
is because o wo k by G. Ta dos on he H. Neumann conjec u e. In [47] his
conjec u e, i.e. ˜ (H∩K)≤˜ (H)˜ (K) o e e y H, K ≤Fn, was p o ed when one
o he wo subg oups in ol ed has ank 2. This is p ecisely he same as saying ha
ank wo subg oups o Fna e all ine .
The main esul in [14] p o ided new examples o ine subg oups o Fn:
Theo em 7.1 (Dicks-Ven u a, [14]).E e y mono- ixed subg oup o Fnis ine .
In pa icula , (Fix S)≤n o e e y S⊆Inj(Fn).
This esul can be hough o bo h as he nex s ep owa ds he solu ion o
he gene al Sco conjec u e (be o e, i was only known o single endomo phisms,
18 E. VENTURA
and his was he i s esul o amilies), o as a pa ial solu ion o he gene al
equalize conjec u e.
Wi h a simple obse a ion one can see ha he amily o ine subg oups o
Fnis closed unde a bi a y in e sec ions ( his is clea o ini e in e sec ions, and
a s anda d a gumen on a descending chain o subg oups p o es i in he in ini e
case, see Co olla y I.4.13 in [14]). Then, i is enough o p o e Theo em 7.1 o
1-mono- ixed subg oups. This is wha i is done in [14].
The wo k [14] con ains a comple e e o mula ion and ex ension o he Bes i-
na-Handel heo y, imp o ing i in he ollowing ou di ec ions.
Though he pape [2], he e is a opological la o in many a gumen s (con-
inui y, limi s, densi y, Can o se s, e c). Howe e , one obse es ha his is no
essen ial because he inal esul , as well as all he in e media e ones, a e o a pu ely
combina o ial and algeb aic na u e. One o he goals o [14] was o e o mula e
he en i e heo y in [2] in a comple e algeb aic se ing, making anspa en i s al-
geb aic na u e. The only essen ial in o ma ion ha is ca ied by a pa h in a g aph,
is he educed sequence o edges c ossed by he pa h. So, i u ns ou o be mo e
con enien o hink o a g aph jus as a combina o ial objec , and con inuous maps
be ween g aphs as o mal maps sending edges (and educed pa hs) o educed pa hs.
A good way o modelling his is he ca ego ical language o g oupoids. Associa ed
o any g aph Z, one has he undamen al g oupoid πZ, which mus be hough
o as he se o pa hs in Zmodulo educ ion, and wi h conca ena ion, which is a
pa ially de ined ope a ion. Then, con inuous maps simply ansla e in o g oupoid
mo phisms. One o he ad an ages o his se ing is ha he no ion o igh ening
in [2] disappea s comple ely in he new language. This is because, when igh ening
a gi en pa h pin a g aph, one ob ains ano he pa h p0, bu hey bo h gi e exac ly
he same elemen o πZ. Ano he ad an age is ha he undamen al g oup o Z
a a e ex is now iewed as a subg oup o he g oupoid πZ. So, passing o he
undamen al g oup does no in ol e a mo e o a di e en con ex , i is jus he
es ic ion o a ce ain subg oup. This way, con inuous maps and wha hey induce
a he undamen al g oup le el a e uni ied in a common con ex . Also, he no ion o
opological ep esen abili y can jus be exp essed as a simple equi alence ela ion
in he con ex o g oupoid mo phisms.
Ano he impo an poin o his e o mula ion o [2] is ha i a oids alking
abou he in a ian il a ion ∅=Z0< Z1<· · · < Z =Z. Following an idea o
Gabo iau-Le i -Lus ig, he en i e a gumen o [2] is ew i en jus looking a he
op s a um, and hen a guing by induc ion on he educed ank o he unde lying
(no necessa ily connec ed) g aph. This simpli ies conside ably he echnical de ails,
since one only has o wo k wi h a single in a ian subg aph (ins ead o he whole
il a ion) and a single ma ix and Pe on-F obenius eigen alue and eigen ec o
(ins ead o one o each s a um in he il a ion).
While ansla ing in o he algeb aic se ing, he echnical de ails we e a anged
in such a way ha he su jec i i y p ope y o he au omo phism is inally used
nowhe e. Hence, he en i e a gumen , and so he inal conclusion, wo ks o in-
jec i e endomo phisms and no jus o au omo phisms o Fn. This is he i s
gene aliza ion p o ided. Howe e , he cons uc ion is a om being alid in he
p esence o a non- i ial ke nel.
FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 19
The ou h and main imp o emen is he gene aliza ion o he heo y o p o ide
a p oo o he ine ia o ixed subg oups. This was achie ed by mixing he Bes ina-
Handel heo y wi h he pull-back echnique de eloped by J. S allings in [45]. Gi en
an injec i e endomo phism φ∈Inj(Fn), he au ho s o [14] conside ed he g aph
Xand he map X→Zcons uc ed in sec ion 6 o [2], and inducing he inclusion
Fix φ≤Fn(see sec ion 4). This is whe e M. Bes ina and M. Handel used he
good g aphical p ope ies o his g aph and g aph map o show ha ˜ (X)≤˜ (Z)
(and consequen ly, (Fix φ)≤n). Ins ead, in [14] he au ho s conside ano he
a bi a y ini e connec ed co e-g aph Yand map Y→Z( his ep esen s an a bi-
a y di e en subg oup K≤Fn). Then, hey look a he pull-back o hese wo
g aphical maps (p e iously a anged o be imme sions), say W→Z. By [45], he
la e imme sion ep esen s he in e sec ion K∩Fix φ. And he good echnical
p ope ies we e enough no only o show ha ˜ (X)≤˜ (Z), bu also o p o e ha
he inequali y passes h ough he pull-back diag am. In his way, i was p o ed
ha ˜ (W)≤˜ (Y), which means (K∩Fix φ)≤ (K) o an a bi a y K≤Fn.
This is p ecisely he ine ia o he ixed subg oup.
In ac , in [14] he non-connec ed e sion o he p e ious esul was p o ed.
The analysis o he algeb aic meaning o he di e en componen s o Wyields
a inal inequali y in ol ing he sum o educed anks o subg oups o he o m
Ky∩Fix(φx−1), whe e x anges o e a se o ep esen a i es o Reidemeis e
classes o φ, and, o e e y x,y anges o e a se o double cose ep esen a i es o
K Fn/Fix(φx−1) (see Theo em IV.5.5 in [14] o he de ails). This non-connec ed
e sion o ine ia was codi ied using se s ac ed on by Fn.
Ano he in e es ing ac abou [14] is ha all he de ails le o he eade in [2]
we e me iculously e i ied in he new con ex .
Finally, ano he concep is in oduced in [14] by elaxing he ine ia condi ion.
A subg oup H≤Fnis called comp essed when (H)≤ (K) o e e y o he
subg oup H≤K≤Fn. Clea ly, ine subg oups a e comp essed, and i is easy
o see ha e ac s a e also comp essed. I is no known whe he comp essed
subg oups a e ine . I is e en unknown whe he e ac s a e ine . These p oblems
a e ela ed wi h some conjec u es and open ques ions abou ixed subg oups (see
sec ion 8).
A e [14], wo mo e pape s appea ed in he li e a u e using, o building upon,
he concep o ine ia. The i s is [1], due o G. Be gman and published in 1999.
Among o he esul s, [1] con ains he ollowing heo em, which de ini ely sol ed he
emaining piece o he gene al Sco conjec u e:
Theo em 7.2 (Be gman, [1]).Le S⊆End(Fn)be an a bi a y amily o
endomo phism o Fn. Then, (Fix S)≤n.
So, he Sco conjec u e was comple ely sol ed by 1999. Howe e , a new p ob-
lem came up na u ally om [14]. In despi e o Be gman’s esul , one can ask i
Fix φis ine when φhas non- i ial ke nel. As a as we know, his is s ill an open
p oblem oday (see sec ion 8 o addi ional commen s).
Be gman’s p oo o Theo em 7.2 is based on a educ ion o he injec i e case
by using his main esul :
Theo em 7.3 (Be gman, [1]).Le Gbe a ini ely gene a ed o sion- ee g oup,
Lan a bi a y g oup, and conside he ee p oduc G∗Land he p ojec ion π:G∗
20 E. VENTURA
L→G. I σ1, σ2:G→G∗La e wo g oup- heo e ic sec ions o π, hen Eq(σ1, σ2)
is a ee ac o o G.
The p oo o his esul used esul s o S allings-Swan and a de ailed analysis
o suppo s o de i a ions. Also, a he end o [1], he e is an in e es ing lis o
ques ions. Some o hem will be discussed in he nex sec ion.
I is also in e es ing o ema k ha W. Dicks and M. Dunwoody ga e a di e en
and mo e gene al p oo o Theo em 7.3. In [12], hey p o ed exac ly he same
esul , bu wi hou using he o sion- ee hypo hesis. This new p oo makes use o
he Almos S abili y Theo em, o p o ees and also o some new olding sequence
echniques.
Ano he wo k elying on he ine ia p ope y is [38], published in 2000 by A.
Ma ino and E. Ven u a. He e i was conjec u ed ha e e y au o- ixed subg oup
is 1-au o- ixed. The au ho s we e no able o p o e his ac (which is s ill open
oday), bu hei main esul p o ides suppo o his conjec u e:
Theo em 7.4 (Ma ino-Ven u a, [38]).Le S⊆End(Fn)and le Mbe he
submonoid o End(Fn)gene a ed by S. Then, he e exis s φ∈Msuch ha Fix S
is a ee ac o o Fix φ.
The p oo was comple ely algeb aic, and conside ed i s Scon aining wo ele-
men s, one o hem injec i e, hen S⊆Inj(Fn) and, inally, he gene al case using
Be gman’s esul . Theo em 7.4 na u ally aises he ques ion o whe he a ee ac-
o o a 1-endo- ixed subg oup is again 1-endo- ixed. This is ob iously ue when
n= 2 bu i is no ue o n≥3, and he ollowing coun e example is p o ided
in [38]. By Theo em 6.2, he subg oup
ha, [a, b],[a, c]i=hai∗h[a, b],[a, c]i
o F3=ha, b, ciis 1-au o- ixed. Bu he au ho s p o ed ha any endomo phism
o F3 ixing [a, b] and [a, c] is o ced o ix a. So, h[a, b],[a, c]iis a ee ac o o
a 1-au o- ixed subg oup o F3, which is no e en endo- ixed. The example can be
easily gene alized o an a bi a y n≥3.
As men ioned in sec ion 6, Theo em 7.4 also has consequences o he maximal
ank case (see Theo em 6.5).
The pape [38] also in oduced he concep o au o- ixed closu e. The au o-
ixed closu e o a subg oup H≤Fnis Hc= Fix(Au H(Fn)), whe e Au H(Fn) is he
subg oup o Au (Fn) o hose au omo phisms φsuch ha H≤Fix φ. As in classical
Galois heo y, H≤Hcbu he inequali y can be s ic . Fo example, he p e ious
example can be used o say ha he au o- ixed closu e o H=h[a, b],[a, c]iin F3
is Hc=ha, [a, b],[a, c]i. In gene al, he ela ionship be ween any gi en subg oup H
o Fn, and i s au o- ixed closu e is qui e obscu e.
8. Conjec u es and open p oblems
A lis o in e es ing p oblems and conjec u es ha s ill emain open is gi en in
his inal sec ion. We u he add some commen s and discussions on hem. Some
o he ques ions men ioned he e a e also asked and discussed in he las sec ion
o [14] and [1].
Fi s o all, i is comple ely sa is ac o y ha he gene al Sco conjec u e, he
o iginal mo i a ion o his line o esea ch, is al eady comple ely sol ed. The
wo k o Bes ina-Handel, Im ich-Tu ne , Dicks-Ven u a and Be gman, shows ha
FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 21
(Fix S)≤n o any a bi a y se So endomo phisms o Fn. Howe e , as usually
happens in ma hema ics, e en when he o iginal p oblem is comple ely sol ed,
he line o esea ch is no exhaus ed because o he in e es ing ques ions came up
na u ally om he de eloped wo k. Fo example, i is no known i ixed subg oups
o endomo phisms o Fna e necessa ily ine .
Conjec u e 8.1 (Ine ia conjec u e).E e y endo- ixed subg oup o Fnis ine .
Equi alen ly, e ac s o Fna e ine .
Discussion. Le S⊆End(Fn). In [14] i is p o ed ha Fix Sis ine when
S⊆Inj(Fn). So, he open p oblem is abou endomo phisms wi h non- i ial
ke nel. In his case, Be gman showed in [1] ha (Fix S)≤n o a bi a y S,
gi ing suppo ing e idence o he conjec u e.
The se o ine subg oups o Fnis closed unde a bi a y in e sec ions. So, i
is enough o conside 1-endo- ixed subg oups. Le φ∈End(Fn). By a esul o
E. Tu ne [51], φ es ic s o an isomo phism o he s able image Fnφ∞, which is
a e ac o Fncon aining Fix φ. So, Fix φis ine as a subg oup o a e ac o
Fn. Hence, he conjec u e is equi alen o saying ha e ac s a e ine . The e is
a ai ly explici desc ip ion o e ac s o Fn(see exe cise 15 in sec ion 3.2 o [36]),
bu his does no seem o help in sol ing his p oblem. 2
Ques ion 8.2 (Be gman, [1]).Le R, H ≤Fnbe wo ini ely gene a ed sub-
g oups o Fn. Is i always ue ha i Ris a e ac o Fn hen R∩His a e ac
o H?
Discussion. The mo i a ion o his ques ion is ha an a i ma i e answe
would immedia ely gi e a p oo o he ine ia conjec u e. Obse e also ha he
analogous ques ion o ee ac o s is ue, namely he in e sec ion o a ee ac o
Ro Fnwi h any subg oup H≤Fnis a ee ac o o H.
No e ha one o he esul s o Be gman in [1] is ha an a bi a y in e sec ion
o e ac s o Fnis again a e ac o Fn.2
Conjec u e 8.3 (Equalize conjec u e).I Sis a se o homomo phisms om
Fm o Fn, one o hem being injec i e, hen (Eq(S)) ≤m. Equi alen ly, Eq(S)is
ine in Fm.
Discussion. By es ic ing he homomo phisms in S o a gi en subg oup K≤
Fm, we see ha he wo s a emen s o he conjec u e a e in ac equi alen .
By a esul o Golds ein-Tu ne [26], Eq(S) is ini ely gene a ed when Sis
ini e. And by a esul o Be gman [1], i he e is a homomo phism ω:Fn→Fm
such ha , o e e y φ∈S,φω is he iden i y o Fm hen Eq(S) is a ee ac o o
Fm. In pa icula , (Eq(S)) ≤min his case.
Using an injec i e elemen in S, we can es a e he conjec u e in he ollowing
way: “Le Fm≤Fnand le Sbe an a bi a y se o mo phisms om Fm o Fn.
Then, (Fix S)≤m(equi alen ly, Fix Sis ine in Fm)”. Wi h his new o mula-
ion, he esul in [14] says ha he conjec u e is ue when he homomo phisms
in Sall ex end o injec i e endomo phisms o Fn.2
Conjec u e 8.4 (Comp essed-ine conjec u e).E e y comp essed subg oup
o Fnis ine .
Discussion. Fi s , le us no e ha , using he Sch eie o mula, he conjec u e
is clea ly ue o subg oups o Fno ini e index.

22 E. VENTURA
Since e e y e ac o Fnis comp essed, bu he e a e comp essed subg oups
which a e no e ac s, his conjec u e is s ic ly s onge han he ine ia con-
jec u e. So a posi i e solu ion would imply ha endo- ixed subg oups o Fna e
ine .
Le Fmbe a comp essed subg oup o Fn, and le Sbe an a bi a y amily o
homomo phisms om Fm o Fn. In [52], he ollowing is p o ed: i i is ue ha
comp essed subg oups a e ine , hen Eq(S) is ine in Fm. So, his conjec u e
seems o be o a simila le el o di icul y as he equalize conjec u e. 2
Ques ion 8.5.Is he e an algo i hm o decide i a gi en subg oup H≤Fnis
ine ?
Discussion. In [53] a simple g aphic algo i hm is gi en o decide i a gi en
subg oup H≤Fnis comp essed. Howe e , he algo i hm seems o be unable o
con ol subg oups no con aining H.2
Ques ion 8.6.Is he e an algo i hm o decide whe he a gi en subg oup H≤
Fnis 1-au o- ixed and, in his case, ind an au omo phism ixing H? Analogous
ques ions can be asked o 1-mono- ixed, 1-endo- ixed, au o- ixed, mono- ixed and
endo- ixed subg oups.
Discussion. This ques ion asks o an algo i hm dual o he one p o ided by M.
Lus ig in [35]. The desc ip ion o 1-au o- ixed subg oups p o ided by Theo em 5.8,
al hough being qui e explici , does no seem o be good enough om he algo i hmic
poin o iew. 2
Ques ion 8.7.Is he e any algeb aic cha ac e iza ion o 1-au o- ixed subg oups
o Fn? Simila ly o he o he i e ypes o ixed subg oups.
Discussion. Collins-Tu ne ga e in [9] such a cha ac e iza ion o he maximal
ank case. The only esul ha is known in his di ec ion o he gene al case is
Theo em 5.8 due o A. Ma ino and E. Ven u a.
I is known ha e e y endo- ixed subg oup His pu e (i.e. x ∈Himplies
x∈H) and ine . Howe e , hese condi ions, al hough being qui e es ic i e,
especially he second one, a e no enough o cha ac e ize ixed subg oups. 2
Ques ion 8.8.Is he e any algo i hm o compu e he au o- ixed closu e Hco
a gi en subg oup H≤Fn?
Discussion. Gi en an au omo phism φo Fn, his ques ion amoun s o unde -
s anding which elemen s o Fna e o ced o be ixed by φ, i some o he s a e. This
can happen e en wi h independen elemen s (see he example in [38] men ioned
in he p e ious sec ion). I is no clea how hese ela ionships wo k, no i is
unde s ood how a subg oup Ho Fnde e mines i s au o- ixed closu e Hc.2
Conjec u e 8.9 (Ma ino-Ven u a, [38]).E e y au o- ixed subg oup o Fnis
1-au o- ixed.
Discussion. This was conjec u ed in [38], which gi es much o he in o ma ion
known in suppo o his conjec u e. The e, i was p o ed ha e e y au o- ixed
( esp. mono- ixed, endo- ixed) subg oup o Fnis a ee ac o o some 1-au o-
ixed ( esp. 1-mono- ixed, 1-endo- ixed) subg oup o Fn. Howe e , an example was
shown o a ee ac o o a 1-au o- ixed subg oup o Fnwi h n≥3, which is no
e en endo- ixed.
FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 23
In pa icula , he conjec u e is ue o F2, and o maximum- ank au o- ixed
subg oups.
The same can conjec u ed wi h he p e ix mono- o endo- ins ead o au o-. 2
Ques ion 8.10.Wha a e he ela ionships (wi h espec o inclusion) be ween
he amilies o 1-au o- ixed, 1-mono- ixed, 1-endo- ixed, au o- ixed, mono- ixed and
endo- ixed subg oups o Fn?
Discussion. I is easy o see ha hese six amilies o subg oups o Fncoincide
when n= 2. By Theo em 6.5, hey also coincide in he maximal ank case.
In gene al, apa om he ob ious inclusions be ween hem, he only in o ma-
ion known is ha , o n≥3, he amily o 1-endo- ixed subg oups s ic ly con ains
ha o 1-au o- ixed ones (see [40]). In be ween, he e is he amily o 1-mono- ixed
subg oups, which is no known o coincide wi h any o he p e ious amilies.
Conjec u e 8.9 and he discussion he e a e closely ela ed wi h his ques ion.
2
Ques ion 8.11.Le Fbe a ee g oup and φ∈End(F). Is Fix φine in F?
Discussion. I Fis ini ely gene a ed, his ques ion has al eady been conside ed,
and coincides wi h he ine ia conjec u e abo e (see 8.1). The only new ques ion
asked he e is abou he in ini ely gene a ed case. The gene al Sco conjec u e
is acuous in his case, since he inequali y (H)≤ (F) is immedia e o e e y
subg oup H≤F. Howe e , he ine ia p ope y changes he si ua ion he e. I is
no a i ial ques ion o ask i he ixed subg oup o an endomo phism o Fis ine
in F. A simila ques ion abou equalize s can also be asked. No hing is known in
his di ec ion. 2
Acknowledgmen s
The au ho is indeb ed o he Ma hema ics Depa men o he Ci y College o
New Yo k o he hospi ali y ecei ed du ing he academic yea 2000-2001, when
his pape has been w i en. He is also indeb ed o J. Bu illo o se e al sugges-
ions imp o ing he pape . The au ho g a e ully acknowledges pa ial suppo by
he DGI (Spain) h ough g an BFM2000-0354 and by he DGR (Gene ali a de
Ca alunya) h ough g an 2001BEA1400176.
Re e ences
[1] G.M. Be gman, Suppo s o de i a ions, ee ac o iza ions and anks o ixed subg oups in
ee g oups, T ans. Ame . Ma h. Soc.,351 (1999), 1531-1550.
[2] M. Bes ina, M. Handel, T ain acks and au omo phisms o ee g oups, Ann. o Ma h.,135
(1992), 1-51.
[3] M. Bes ina, M. Handel, T ain acks o su ace homeomo phisms, Topology,34(1) (1995),
109-140.
[4] O. Bogopolski, Classi ica ion o au omo phisms o he ee g oup o ank 2 by anks o ixed-
poin subg oups, J. G oup Theo y,3(2000), 339–351.
[5] A.J. Casson, S.A. Bleile , Au omo phisms o su aces a e Nielsen and Thu s on, Camb idge
Uni e si y P ess, (1988).
[6] M.M. Cohen, M. Lus ig, On he dynamics and he ixed subg oup o a ee g oup au omo -
phism, In en . Ma h.,96 (1989), 613-638.
[7] D.J. Collins, E.C. Tu ne , An au omo phism o a ee g oup o ini e ank wi h maximal ank
ixed poin subg oup ixes a p imi i e elemen , J. Pu e Appl. Algeb a,88 (1993), 43-49.
24 E. VENTURA
[8] D.J. Collins, E.C. Tu ne , E icien ep esen a i es o au omo phisms o ee p oduc s, Michi-
gan Ma h. J.,41 (1994), 443-464.
[9] D.J. Collins, E.C. Tu ne , All au omo phisms o ee g oups wi h maximal ank ixed sub-
g oups, Ma h. P oc. Camb idge Philos. Soc.,119 (1996), 615-630.
[10] D. Coope , Au omo phisms o ee g oups ha e ini ely gene a ed ixed poin se s, J. Algeb a,
111 (1987), 453-456.
[11] W. Dicks, M.J. Dunwoody, G oups ac ing on g aphs, Ad anced S udies in Ma hema ics 17,
Camb idge Uni e si y P ess, (1989).
[12] W. Dicks, M.J. Dunwoody, On equalize s o sec ions, J. Algeb a,216 (1999), 20-39.
[13] W. Dicks, E. Ven u a, I educible ou e au omo phisms o g ow h a e one, J. Pu e Appl.
Algeb a,88 (1993), 51-62.
[14] W. Dicks, E. Ven u a, The g oup ixed by a amily o injec i e endomo phism o a ee g oup,
Con emp. Ma h.,195 (1996), 1-81.
[15] J.L. Dye , G.P. Sco , Pe iodic au omo phisms o ee g oups, Comm. Alg.,3(1975), 195-201.
[16] A. Fa hi, F. Laudenbach, V. Poena u e al., T a aux de Thu s on su les su aces, As ´e isque,
66-67, Soc. Ma h. F ance, Pa is, 1979.
[17] D. Gabo iau, A. Jaege , G. Le i , M. Lus ig, An index o coun ing ixed poin s o au omo -
phisms o ee g oups, Duke Ma h. J.,93(3) (1998), 425-452.
[18] D. Gabo iau, G. Le i , The ank o ac ions on R- ees, Ann. Sci. ´
Ecole No m. Sup.,28(5)
(1995), 549-570.
[19] D. Gabo iau, G. Le i , M. Lus ig, A dend ological p oo o he Sco conjec u e o au o-
mo phisms o ee g oups, P oc. Edinbu gh Ma h. Soc.,41(2) (1998), 325-332.
[20] S.M. Ge s en, On ixed poin s o ce ain au omo phisms o ee g oups, P oc. London Ma h.
Soc.,48 (1984), 72-90. Addendum: 49 (1984), 340-342.
[21] S.M. Ge s en, On ixed poin s o au omo phisms o ini ely gene a ed ee g oups, Bull. Ame .
Ma h. Soc.,8(1983), 451-454.
[22] S.M. Ge s en, Fixed poin s o au omo phisms o ee g oups, Ad . Ma h.,64 (1987), 51-85.
[23] S.M. Ge s en, J.R. S allings, I educible ou e au omo phisms o a ee g oup, P oc. Ame .
Ma h. Soc.,111 (1991), 309-314.
[24] R.Z. Golds ein, E.C. Tu ne , Au omo phisms o ee g oups and hei ixed poin s, In en .
Ma h.,78 (1984), 1-12.
[25] R.Z. Golds ein, E.C. Tu ne , Monomo phisms o ini ely gene a ed ee g oups ha e ini ely
gene a ed equalize s, In en . Ma h.,82 (1985), 283-289.
[26] R.Z. Golds ein, E.C. Tu ne , Fixed subg oups o homomo phisms o ee g oups, Bull. London
Ma h. Soc.,18 (1986), 468-470.
[27] A.H.M. Hoa e, On au omo phisms o ee g oups I, J. London Ma h. Soc.,38 (1988), 277-285.
[28] A.H.M. Hoa e, On au omo phisms o ee g oups II, J. London Ma h. Soc.,42 (1990), 226-
236.
[29] R.A. Ho n, C.R. Johnson, Ma ix analysis, Camb idge Uni e si y P ess, Camb idge (1990).
[30] A.G. Howson, On he in e sec ion o ini ely gene a ed ee g oups, J. London Ma h. Soc.,
29 (1954), 428-434.
[31] W. Im ich, S. K s i´c, E.C. Tu ne , On he ank o ixed poin se s o au omo phisms o ee
g oups, NATO Ad . Sci. Ins . Se . C Ma h. Phys. Sci.,301 (1990), 113-122.
[32] W. Im ich, E.C. Tu ne , Fixed subse s o homomo phisms o ee g oups, Topology and Com-
bina o ial G oup Theo y, Lec u e No es in Ma h., 1440 (1990), 130-147.
[33] W. Im ich, E.C. Tu ne , Endomo phisms o ee g oups and hei ixed poin s, Ma h. P oc.
Camb idge Philos. Soc.,105 (1989), 421-422.
[34] W. Jaco, P.B. Shalen, Su ace homeomo phisms and pe iodici y, Topology,16 (1977), 347-
367.
[35] M. Lus ig, S uc u e and conjugacy o au omo phisms o ee g oups I, II, p ep in s.
[36] W. Magnus, A. Ka ass, D. Soli a , Combina o ial G oup Theo y, In e science Publishe s,
New Yo k, (1966).
[37] A. Ma ino, An index heo em o au omo phisms o ee p oduc s, J. G oup Theo y,2
(1999), 199-211.
[38] A. Ma ino, E. Ven u a, On au omo phism- ixed subg oups o a ee g oup, J. Algeb a,230
(2000), 596-607.
[39] A. Ma ino, E. Ven u a, A desc ip ion o au o- ixed subg oups in he ee g oup, p ep in .
FIXED SUBGROUPS IN FREE GROUPS: A SURVEY 25
[40] A. Ma ino, E. Ven u a, Examples o e ac s in Fn ha a e no he ixed subg oup o any
au omo phism, p ep in .
[41] J. Nielsen, Un e suchungen zu opologie de geschlossenen l¨achen II, Ac a Ma h.,53 (1929),
1-76.
[42] J. Nielsen, Die au omo phismen de allgemeinen unendlichen g uppe mi zwei e zeugenden,
Ma h. Ann.,78 (1918), 385-397.
[43] Z. Sela, The Nielsen-Thu s on classi ica ion and au omo phisms o a ee g oup I, Duke Ma h.
J.,84 (1996), 379-397.
[44] J.R. S allings, Topologically un ealizable au omo phisms o ee g oups, P oc. Ame . Ma h.
Soc.,84 (1982), 21-24.
[45] J.R. S allings, Topology o ini e g aphs, In en . Ma h.,71 (1983), 551-565.
[46] J.R. S allings, G aphical heo y o au omo phisms o ee g oups, Combina o ial g oup heo y
and opology, Ann. o Ma h. S ud. 111, P ince on Uni e si y P ess, 1987.
[47] G. Ta dos, On he in e sec ion o subg oups o a ee g oup, In en . Ma h.,108 (1992), 29-36.
[48] S. Thomas, Fixed poin s o au omo phisms o ini ely gene a ed ee g oups, P oc. Ame .
Ma h. Soc.,103 (1988), 333.
[49] W. Thu s on, On he geome y and dynamics o di eomo phisms o su aces I, P ep in ,
1976.
[50] W. Thu s on, On he geome y and dynamics o di eomo phisms o su aces, Bull. Ame .
Ma h. Soc.,19(2) (1988), 417-431.
[51] E.C. Tu ne , Tes wo ds o au omo phisms o ee g oups, Bull. London Ma h. Soc.,28
(1996), 255-263.
[52] E. Ven u a, Endomo ismes de g ups lliu es ini amen gene a s (ca alan), PhD Thesis, Uni-
e si a Au `onoma de Ba celona, (1995).
[53] E. Ven u a, On ixed subg oups o maximal ank, Comm. Algeb a,25 (1997), 3361-3375.
[54] J.H.C. Whi ehead, On ce ain se s o elemen s in a ee g oup, P oc. London Ma h. Soc.,41
(1936), 48-56.
[55] J.H.C. Whi ehead, On equi alen se s o elemen s in a ee g oup, Ann. o Ma h.,37 (1936),
782-800.
[56] H. Zieschang, ¨
Ube au omo phismen ebene diskon inuie liche g uppen, Ma h. Ann. 166
(1966), 148-167.
Dep . Ma . Apl. III, Uni i e si a Poli `
ecnica de Ca alunya, Ba celona, SPAIN,
and, Dep . o Ma h., Ci y College o New Yo k (Cuny), USA.
E-mail add ess:[email p o ec ed]