Contents lists available atScienceDirect
Finite
Fields
and
Their
Applications
www.elsevier.com/locate/ffa
A
characterization
of
linearized
polynomials
with
maximum
kernel
✩Bence Csajbóka, Giuseppe Marinob,c, Olga Polverinob,
Ferdinando Zullob,∗
aMTA–ELTEGeometricandAlgebraicCombinatoricsResearchGroup,ELTE
EötvösLorándUniversity,Budapest,Hungary,DepartmentofGeometry,1117
Budapest,Pázmány P.stny.1/C,Hungary
b
DipartimentodiMatematicaeFisica,UniversitàdegliStudidellaCampania “LuigiVanvitelli”,VialeLincoln5,I-81100Caserta,Italy
cDipartimentodiMatematicaeApplicazioni“RenatoCaccioppoli”,Universitàdegli
StudidiNapoli“FedericoII”,ViaCintia,MonteS.Angelo,I-80126Napoli,Italy
a r t i c l e i n f o a b s t r a c t
Articlehistory:
Received15August2018 Receivedinrevisedform26 November2018
Accepted29November2018 Availableonline6December2018 CommunicatedbyDieterJungnickel
MSC:
11T06 15A04
Keywords:
Linearizedpolynomials
Weprovidesufficient andnecessaryconditions forthe coef-ficientsofa q-polynomial f overFqn whichensure that the
numberofdistinctrootsof f inFqn equals thedegreeof f .
Wesaythatthesepolynomialshavemaximumkernel.Asan applicationwestudyindetail q-polynomialsofdegree qn−2 overFqn whichhave maximumkernelandforn≤ 6 welist
all q-polynomialswithmaximum kernel.Wealso obtain in-formationonthesplittingfieldofanarbitraryq-polynomial. Analogousresultsareprovedforqs-polynomialsaswell,where
gcd(s,n)= 1.
©2018ElsevierInc.Allrightsreserved.
✩ TheresearchwassupportedbyMinistryforEducation,UniversityandResearchofItalyMIUR(Project PRIN2012“GeometriediGaloisestrutturediincidenza”)andbytheItalianNationalGroupforAlgebraic andGeometricStructuresandtheirApplications(GNSAGA- INdAM).Thefirstauthorwassupportedby theJánosBolyai ResearchScholarshipoftheHungarian AcademyofSciencesandbyOTKAGrantNo. K 124950.
* Correspondingauthor.
E-mailaddresses:csajbokb@cs.elte.hu(B. Csajbók),giuseppe.marino@unicampania.it, giuseppe.marino@unina.it(G. Marino),olga.polverino@unicampania.it(O. Polverino), ferdinando.zullo@unicampania.it(F. Zullo).
https://doi.org/10.1016/j.ffa.2018.11.009
Lineartransformations Semilineartransformations
1. Introduction
Aq-polynomialoverFqnisapolynomialoftheformf (x)=
iaixq i
,whereai∈ Fqn.
Wewill denotethesetofthesepolynomials byLn,q.LetK denotethealgebraicclosure
ofFqn.ThenforeveryFqn≤ L≤ K,f definesanFq-lineartransformationofL,whenL
isviewedasanFq-vectorspace.IfL isafinitefieldofsizeqmthenthepolynomialsofLn,q
consideredmodulo(xqm−x) formanF
q-subalgebraoftheFq-lineartransformationsofL.
OncethisfieldL isfixed,wecandefinethekernel off asthekernelofthecorresponding Fq-linear transformation ofL, which isthe sameas theset of roots of f inL; and the
rank off astherankofthecorrespondingFq-lineartransformation ofL.Notethatthe
kernelandtherankoff dependonthisfieldL andfromnowonwewillconsiderthecase L=Fqn.Inthis caseLn,q consideredmodulo(xq
n
− x) isisomorphictotheFq-algebra
of Fq-lineartransformationsof then-dimensionalFq-vectorspace Fqn.The elementsof
this factor algebraare representedby L˜n,q :={
n−1 i=0 aixq
i
: ai ∈ Fqn}.For f ∈ ˜Ln,q if
deg f = qk then wecall k theq-degree of f .It is clearthatinthis casethekernel off hasdimensionat mostk and therankoff isat leastn− k.
Let U = u1,u2,. . . ,ukFq be ak-dimensional Fq-subspace of Fqn. It is well known
that, up to a scalar factor, there is a unique q-polynomial of q-degree k, which has kernel U . Wecangetsuchapolynomialasthedeterminantofthematrix
⎛ ⎜ ⎜ ⎜ ⎜ ⎝ x xq · · · xqk u1 uq1 · · · u qk 1 .. . uk uqk · · · uq k k ⎞ ⎟ ⎟ ⎟ ⎟ ⎠.
Theaimofthispaperistostudytheotherdirection,i.e.whenagivenf ∈ ˜Ln,q with
q-degreek haskernelofdimensionk.Ifthishappensthenwesaythatf isaq-polynomial
with maximumkernel.
If f (x) ≡ a0x+ a1xσ+· · · + akxσ k
(mod xqn
− x), with σ = qs for some s with
gcd(s,n)= 1,thenwesaythatf (x) isaσ-polynomial(orqs-polynomial)withσ-degree
(or qs-degree)k.Regardingσ-polynomialsthefollowingisknown.
Result 1.1.[7, Theorem 5] Let L be a cyclic extension of a field F of degree n, and suppose thatσ generates the Galois group of L over F. Let k bean integer satisfying
1 ≤ k ≤ n, and let a0,a1,. . . ,ak be elements of L, not all them are zero. Then the
f (x) = a0x + a1xσ+· · · + akxσ k
haskernelwithdimensionatmostk inL.
Similarlyto thes= 1 case wewill saythataσ-polynomialis ofmaximum kernel if thedimensionofitskernelequalsitsσ-degree.
Linearized polynomials have been used to describe families of Fq-linear maximum
rank distancecodes (MRD-codes), i.e. Fq-subspacesof L˜n,q of order qnk inwhich each
element has kernel of dimension at most k. The first examples of MRD-codes found were the generalized Gabidulin codes [3,5], thatis Gk,s = x,xq
s
,. . . ,xqs(k−1)Fqn with
gcd(s,n)= 1;thefactthatGk,sisanMRD-codecanbeshownsimplybyusingResult1.1.
Itisimportant tohaveexplicitconditionsonthecoefficientsofalinearizedpolynomial characterizingthenumberofitsroots.Furtherconnectionswithprojective polynomials canbe foundin[8].
Ourmain resultprovides sufficient and necessary conditions onthe coefficients ofa
σ-polynomialwith maximumkernel.
Theorem1.2. Consider
f (x) = a0x + a1xσ+· · · + ak−1xσ
k−1 − xσk,
with σ = qs,gcd(s,n)= 1 and a
0,. . . ,ak−1 ∈ Fqn. Thenf (x) is of maximumkernel if
andonly ifthematrix
A = ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ 0 0 · · · 0 a0 1 0 · · · 0 a1 0 1 · · · 0 a2 .. . ... ... ... 0 0 · · · 1 ak−1 ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ (1) satisfies AAσ· · · Aσn−1 = Ik,
whereAσi isthematrixobtainedfrom A byapplying toeachof itsentriesthe
automor-phism x→ xσi andI
k istheidentity matrixoforder k.
Animmediateconsequenceofthisresultgivesinformationonthesplittingfieldofan arbitraryσ-polynomial,cf. Theorem4.1.
InSection3.1westudyindetailstheσ-polynomialsofσ-degreen− 2 for eachn.For
n≤ 6 wealsoprovidealistofallσ-polynomialswithmaximumkernelcf.Sections3.2,3.3
and3.4.TheseresultsmightyieldfurtherclassificationresultsandexamplesofFq-linear
2. Preliminaryresults
In this section we recall some results of Dempwolff, Fisher and Herman from [4], adaptingthemto ourneedsinordertomake thispaperself-contained.
LetV beak-dimensionalvectorspaceoverthefieldF andletT beasemilinear
trans-formationofV .AT -cyclicsubspaceofV isanF-subspaceofV spannedby{v,T (v),. . .} overF forsomev∈ V ,whichwillbedenotedby[v].Wefirstrecallthefollowinglemma. Lemma2.1.[4,Theorem1] LetV beann-dimensionalvectorspaceoverthefieldF,σ an
automorphism ofF and T an invertible σ-semilineartransformation onV .Then
V = [u1]⊕ . . . ⊕ [ur]
forT -cyclicsubspaces satisfying dim[u1]≥ dim[u2]≥ . . . ≥ dim[ur]≥ 1.
Theorem 2.2.Let T be an invertible semilinear transformation of V = V (k,qn) of
or-der n,withcompanionautomorphismσ∈ Aut(Fqn) suchthatFix(σ)=Fq.ThenFix(T )
is ak-dimensionalFq-subspaceof V andFix(T )Fqn = V .
Proof. First assumethatthe companionautomorphism of T is x→ xq and thatthere
exists v∈ V suchthat
V =v, T (v), . . . , Tk−1(v)Fqn.
Followingtheproofof[4,MainTheorem],considertheorderedbasisBT = (v,T (v),. . . , Tk−1(v)) andletA bethematrixassociatedwithT withrespect tothebasisBT,i.e.
A = ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ 0 0 · · · 0 α0 1 0 · · · 0 α1 0 1 · · · 0 α2 .. . ... ... ... 0 0 · · · 1 αk−1 ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ ∈ Fk×k qn , (2) where Tk(v) = k
i=1αi−1Ti(v) with α0,. . . ,αk−1 ∈ Fqn and, since T is invertible,
we have α0 = 0. Denote by T the semilinear transformation of Fkqn having A as the
associated matrix withrespect to thecanonical orderedbasis BC = (e1,. . . ,ek) of Fkqn
and companion automorphism x→ xq.Note thatc
BT(Fix(T ))= Fix(T ),where cBT is
thecoordinatization withrespectto thebasisBT. Also,sinceT hasordern,we have
AAq· · · Aqn−1 = Ik, (3)
where Aqi,fori∈ {1,. . . ,n− 1},is thematrixobtainedfrom A byapplying toeachof itsentriestheautomorphismx→ xqi.Avectorz= (z0,. . . ,zk−1)∈ Fkqn isfixedbyT if
⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ α0zk−1q = z0 zq0+ α1zqk−1= z1 .. . zqk−2+ αk−1zk−1q = zk−1
Eliminatingz0,. . . ,zk−2, weobtain theequation
αq0k−1zqk−1k + αq1k−2zqk−1k−1+ . . . + αk−1zk−1q − zk−1= 0,
whichhasqk distinct solutionsinthealgebraic closureK ofFqn bythederivativetest.
Each solution determines a unique vector of Fix(T ) in Kk. Also, the set Fix(T ) is an
Fq-subspace of Kk and hence dimFqFix(T ) = k. Let {w1,. . . ,wk} be an Fq-basis of
Fix(T ) andnote thatsince|Fix(T )|= qk, avector k
i=1
aiwi isfixed byT ifand onlyif ai∈ Fq.Thisimpliesthatw1,. . . ,wkarealsoK-independent.ThusFix(T )K=Kk and {w1,. . . ,wk} isalsoaK-basisofKk.Denotebyφ theK-lineartransformationsuchthat
φ(wi)= ei and by P theassociated matrix withrespect to the canonicalbasis BC, so
P ∈ GL(k,K).Thesemilineartransformation φ◦ T ◦ φ−1 hascompanionautomorphism
x→ xq, ordern andassociated matrixwithrespect to thecanonicalbasisP · A· P−q,
whereP−q istheinverseofP in whichtheautomorphismx→ xq isappliedentrywise.
Notethatφ◦ T ◦ φ−1(ei)= φ(T (wi))= φ(wi)= ei,hence
P· A · P−q= Ik, (4)
i.e.
Pq= P · A. (5)
ByEquations(3) and(5) andusinginductionweget Pqn = P· A · Aq· . . . · Aqn−1 = P,
i.e. P ∈ Fk×kqn . This implies thatFix(T ) is an Fq-subspace of Fkqn of dimension k and
henceFix(T )= c−1BT(Fix(T )) isak-dimensionalsubspaceofV (k,qn) withtheproperty
thatFix(T )Fqn = V .
Considernowthegeneralcase,i.e.supposeT asinthestatement,thatisT isan invert-iblesemilinearmapofordern withcompanionautomorphismx→ xqsandgcd(s,n)= 1.
Sincegcd(s,n)= 1 thereexistl,m∈ N suchthat1= sl + mn, andhencegcd(l,n)= 1. Then thesemilinear transformation Tl hasorder n, companion automorphism x→ xq
andFix(T )= Fix(Tl).ByLemma2.1,we maywrite V = [u1]⊕ . . . ⊕ [ur],
where [ui] is a Tl-cyclic subspace of V of dimension mi ≥ 1, for each i ∈ {1,. . . ,r},
and ri=1mi = k.Then we can restrictTl to eachsubspace [ui] and byapplying the
previousargumentswe getthatUi= Fix(Tl|[ui]) isanFq-subspaceof[ui] ofdimension mi withthepropertythatUiFqn = [ui].Thus
Fix(T ) = Fix(Tl) = U1⊕ . . . ⊕ Ur
is anFq-subspaceofdimensionk ofV withtheproperty thatFix(T )Fqn = V . 2
TheexistenceofamatrixP ∈ GL(k,K),withK thealgebraicclosureofafinitefield of order q,satisfying(4) isalso aconsequenceofthecelebratedLang’s Theorem[9] on connectedlinearalgebraic groups.Moreprecisely,byLang’sTheorem,sinceGL(k,K) is a connected linear algebraic group, the map M ∈ GL(k,K) → M−1· Mq ∈ GL(k,K)
is onto. InTheorem 2.2 it is proved that, if the semilinear transformation of V (k,qn)
having A asassociated matrixhasorder n,thenP ∈ GL(k,Fqn).
Remark 2.3.Let T be an invertible semilinear transformation of V = V (k,qn) with companion automorphism x → xq and let K be the algebraic closure of Fqn. Denote
by T the semilinear transformation of Kk associated with T as in the proof of
The-orem 2.2. If λ ∈ K, then the set E(λ) := {v ∈ Kk: T (v) = λv} is an F
q-subspace
of Kk. By [4, page 293], it follows that E(λ) = λq−11 Fix(T ) and by [4, Main
The-orem] E(λ) is a k-dimensional Fq-subspace of Kk. Also, when T has order n and λq−11 ∈ Fqn, by Theorem 2.2, E(λ) is a k-dimensional Fq-subspace contained in Fkqn
suchthatE(λ)Fqn =F
k qn.
3. Mainresults
Now weareableto proveourmain result:
Proof of Theorem1.2. First suppose dimFqker f = k. Then there exist u0,u1, . . . , uk−1∈ Fqn whichform anFq-basisofker f .
Put u := (u0,u1,. . . ,uk−1)∈ Fkqn.Sinceu0,u1,. . . ,uk−1 areFq-linearlyindependent,
by [10, Lemma 3.51],we get thatB := (u,uqs,. . . ,uqs(k−1)) is anordered F
qn-basisof Fk qn.Also, uq sk = a0u + a1uq s
+· · · + ak−1uqs(k−1). It canbe seenthatthe matrix(1)
representstheFqn-linearpartoftheFqn-semilinearmap σ : v∈ Fkqn→ vq s
∈ Fk qn w.r.t.
thebasisB. Sincegcd(s,n)= 1,σ hasordern and hencetheassertionfollows. Viceversa, letτ bedefinedasfollows
τ : ⎛ ⎜ ⎜ ⎜ ⎜ ⎝ x0 x1 .. . xk−1 ⎞ ⎟ ⎟ ⎟ ⎟ ⎠∈ F k qn → A ⎛ ⎜ ⎜ ⎜ ⎜ ⎝ x0 x1 .. . xk−1 ⎞ ⎟ ⎟ ⎟ ⎟ ⎠ qs ∈ Fk qn, (6)
whereA isas in(1) with thepropertyAAqs· · · Aqs(n−1) = Ik. Thenτ has ordern and,
byTheorem 2.2, it fixesa k-dimensionalFq-subspace S ofFkqn with the property that SFqn =F
k qn.
LetBS = (s0,. . . ,sk−1) beanFq-basisofS and notethat,sinceSFqn =F k
qn,BS is
alsoanFqn-basisofFkqn, thendenotingbyBC thecanonicalordered basisofFkqn,there
existsauniqueisomorphismφ ofFk
qn suchthatφ(si)= ei foreachi∈ {1,. . . ,k}.Then σ = φ◦ τ ◦ φ−1,whereσ : v∈ Fk qn→ vq s ∈ Fk qn.Also, σi= φ◦ τi◦ φ−1, (7)
foreachi∈ {1,. . . ,n− 1}.Also,by(6) τ (e0) = e1, τ (e1) = τ2(e0) = e2, .. . τ (ek−1) = τk(e 0) = (a0, . . . , ak−1) = a0e0+· · · + ak−1ek−1.
So,wegetthat
τk(e0) = a0e0+ a1τ (e0) +· · · + ak−1τk−1(e0),
andapplyingφ itfollowsthat
φ(τk(e0)) = a0φ(e0) + a1φ(τ (e0)) +· · · + ak−1φ(τk−1(e0)).
By(7) thepreviousequationbecomes
σk(φ(e0)) = a0φ(e0) + a1σ(φ(e0)) +· · · + ak−1σk−1(φ(e0)).
Putu = φ(e0),then
uqsk= a0u + a1uq
s
+· · · + ak−1uqs(k−1).
Thisimpliesthatu0,u1,. . . ,uk−1 areelementsofker f ,whereu = (u0,. . . ,uk−1).Also,
theyareFq-independentsinceB = (u,. . . ,uq s(k−1)
)= (φ(e0),. . . ,φ(ek−1)) isanordered
Fqn-basisofFk
qn.Thiscompletes theproof. 2
As a corollary we get the second part of [6, Theorem 10], see also [12, Lemma 3] for the case s = 1 and [11] for the case when q is a prime. Indeed, by evaluating the determinantsinAAqs· · · Aqs(n−1) = Ik weobtainthefollowing corollary.1
1 Forx∈ Fqn andforasubfieldF
qm ofFqn wewilldenotebyNqn/qm(x) thenormofx overFqm andby
Trqn/qm(x) wewilldenotethetraceofx overFqm.Ifn isclearfromthecontextandm= 1 thenwewill
Corollary3.1. Ifthekernelofaqs-polynomialf (x)= a0x+a1xq
s
+· · ·+ak−1xqs(k−1)−xqsk
has dimensionk, thenN(a0)= (−1)n(k+1).
Corollary 3.2.LetA beamatrix asinTheorem1.2.The condition AAqs· · · Aqs(n−1) = Ik
is satisfiedifandonly if AAqs
· · · Aqs(n−1) fixes e
0= (1,0,. . . ,0).
Proof. The only if part is trivial, we prove the if part by induction on 0 ≤ i ≤ k − 1. Suppose AAqs
· · · Aqs(n−1)eT
i = eTi for some 0 ≤ i ≤ k − 1. Then by
tak-ing qs-th powers of each entry we get AqsAq2s· · · AeTi = eTi . Since AeTi = eTi+1 this yields AqsAq2s· · · Aqs(n−1)eT
i+1 = eTi . Then multiplying both sides by A yields AAqsAq2s· · · Aqs(n−1)eT i+1 = eTi+1. 2 Consideraqs-polynomialf (x)= a 0x+ a1xq s +· · · + ak−1xqs(k−1)− xqsk, thematrix A∈ Fkqn×k asinTheorem1.2andthesemilinearmapτ definedin(6).
Note that eτ0 = (0, 1, 0, . . . , 0) = e1 eτ02 = (0, 0, 1, . . . , 0) = e2 .. . eτ0k−1 = (0, 0, 0, . . . , 1) = ek−1 eτ0k= (a0, a1, a2, . . . , ak−1) eτ0k+1= (a0aq s k−1, aq s 0 + a1aq s k−1, aq s 1 + a2aq s k−1, . . . , aq s k−2+ aq s+1 k−1 ). (8) Hence,if
eτ0i= (Q0,i, Q1,i, . . . , Qk−1,i)
where Qj,icanbeseenaspolynomials ina0,a1,. . . ,ak−1,fori≥ 0,then eτ0i+1 = (a0Qq s k−1,i, Qq s 0,i+ a1Qq s k−1,i, . . . , Qq s k−2,i+ ak−1Qq s k−1,i),
i.e. the polynomials Qj,i for 0 ≤ j ≤ k − 1 can be defined by the following recursive
relationsfor0≤ i≤ k − 1:
Qj,i=
1 if j = i, 0 otherwise, and bythefollowing relationsfori≥ k:
Q0,i+1= a0Qq s k−1,i Qj,i+1= Qq s j−1,i+ ajQq s k−1,i. (9) Now,weareabletoprovethefollowing.
Theorem3.3. The kernel of a qs-polynomial f (x)= a0x+ a1xq
s
+· · · + ak−1xqs(k−1)− xqsk ∈ Fqn[x],wheregcd(s,n)= 1, has dimensionk ifandonly if
Qj,n(a0, a1, . . . , ak−1) =
1 if j = 0,
0 otherwise. (10)
Proof. Relations(9) canbewrittenasfollows ⎛ ⎜ ⎜ ⎜ ⎜ ⎝ Q0,i+1 Q1,i+1 .. . Qk−1,i+1 ⎞ ⎟ ⎟ ⎟ ⎟ ⎠= ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ 0 0 · · · 0 a0 1 0 · · · 0 a1 0 1 · · · 0 a2 .. . ... ... ... 0 0 · · · 1 ak−1 ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ ⎛ ⎜ ⎜ ⎜ ⎜ ⎝ Qq0,is Qq1,is .. . Qqk−1,is ⎞ ⎟ ⎟ ⎟ ⎟ ⎠,
with i ∈ {0,. . . ,n − 1}. Also, (Q0,0,Q1,0,. . . ,Qk−1,0) = (1,0,. . . ,0) and eτ
t
0 =
(Q0,t,. . . ,Qk−1,t) for t ∈ {0,. . . ,n}. By Theorem 1.2 and by Corollary 3.2, the
ker-nel of f (x) has dimension k if and only if e0 = (Q0,0,Q1,0,. . . ,Qk−1,0) is fixed by
AAqs· · · Aqs(n−1),sothis happensifandonlyif
eτ0n= (Q0,n, Q1,n, . . . , Qk−1,n) = (1, 0, . . . , 0). 2
Theorem 3.3 with k = n− 1 and s = 1 givesthe following well-known result as a corollary.
Corollary3.4.[10,Theorem2.24] Thedimensionof thekernel ofaq-polynomialf (x)∈ Fqn[x] isn− 1 ifandonly ifthere existα,β ∈ F∗qn suchthat
f (x) = α Tr(βx). AgainfromTheorem3.3 wecandeducethefollowing. Corollary3.5. [10,Ex. 2.14] Theqs-polynomiala0x− xq
sk
∈ Fqn[x], withgcd(s,n)= 1
and1≤ k ≤ n− 1, admitsqk roots ifand onlyif k| n and N
qn/qk(a0)= 1.
3.1. Whentheqs-degreeequalsn− 2
Inthissectionweinvestigateqs-polynomials f (x) = a0x + a1xq
s
+· · · + an−3xq s(n−3)
with gcd(s,n)= 1.ByTheorem 3.3,dim ker f (x) = n− 2 if andonlyifa0,a1,. . . ,an−3
satisfythefollowingsystemofequations ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ Q0,n= a0(aq 2s n−4+ aq 2s+qs n−3 ) = 1, Q1,n= aq s 0 a q2s n−3+ a1(aq 2s n−4+ aq 2s +qs n−3 ) = 0, Q2,n= aq 2s 0 + a q2s n−3aq s 1 + a2(aq 2s n−4+ aq 2s+qs n−3 ) = 0, Q3,n= aq 2s 1 + a q2s n−3aq s 2 + a3(aq 2s n−4+ aq 2s +qs n−3 ) = 0, .. . Qn−3,n = aqn−52s + aqn−32s an−4qs + an−3(aqn−42s + aqn−32s+qs) = 0, (11) whichisequivalent to ⎧ ⎪ ⎨ ⎪ ⎩ a0(aq 2s n−4+ aq 2s +qs n−3 ) = 1, a1=−aq s+1 0 a q2s n−3=: g1(a0, an−3), aj=−aq 2s j−2a0− aq 2s n−3aq s j−1a0=: gj(a0, an−3), for 2≤ j ≤ n − 3. (12)
So,dimFqker f (x) = n− 2 if andonlyifa0 andan−3 satisfytheequations
a0(gn−4(a0, an−3)q 2s + aqn−32s+qs) = 1, an−3 = gn−3(a0, an−3), and aj= gj(a0,an−3) forj∈ {1,. . . ,n− 4}.
Theorem 3.6.Suppose thatf (x)= a0x+ a1xq+· · · + an−3xqn−3− xqn−2 has maximum
kernel.Thenfort≥ 2 withgcd(t− 1,n)= 1 thecoefficients at−2 andan−t arenon-zero
and, with s= n− t+ 1,
an−2t+1aqt−22s+qs=−an−tqs+1aq2t2s−3. (13)
Also, itholdsthat
−an−t(−aq s t−2aq 2s 3t−4+ aq 2s+qs 2t−3 ) = aq 2s+qs+1 t−2 . (14)
In particular,fort≥ 2 withgcd(t− 1,n)= 1 weget
N(an−t) = (−1)nN(at−2) (15)
and
N(an−2t+1) = (−1)nN(a2t−3), (16)
Proof. Lett≥ 2 withgcd(t− 1,n)= 1 andconsiderthepolynomialF (x)= f (xqt),that is, F (x) = a0xq t + a1xq t+1 +· · · + an−3xqn+t−3− xqn+t−2.
Clearly dimFqker F = dimFqker f = n− 2. Byrenaming the coefficients, F (x) can be
writtenas F (x) = α0x + α1xq n−t+1 + α2xq 2(n−t+1) +· · · + αn−3xq (n−t+1)(n−3) + αn−2xq (n−t+1)(n−2) = α0x + α1xq n−t+1 +· · · + αn−3xq3t−3+ αn−2xq2t−2.
SinceF (x) hasmaximumkernel,bythesecondequationof(12) wegetα0= 0,αn−2= 0
andthefollowingrelation
− α1 αn−2 =− − α0 αn−2 qs+1 −αn−3 αn−2 q2s . (17)
The coefficient αj of F (x) equals the coefficient ai of f (x) with i ≡ n− t+ j(1− t)
(mod n),inparticular ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ α0= an−t, α1= an−2t+1, αn−3= a2t−3, αn−2= at−2, αn−4= a3t−4, (18)
andby(17),wegetthatat−2 andan−t arenonzero,and an−2t+1aq 2s+qs t−2 =−aq s+1 n−t aq 2s 2t−3,
whichgives(13).Thefirstequationof(12) gives
− α0 αn−2 −αn−4 αn−2 q2s + −αn−3 αn−2 q2s+qs = 1, thatis, −α0(−αq s n−2αq 2s n−4+ αq 2s+qs n−3 ) = αq 2s+qs+1 n−2 .
Then(18) andαn−4= a3t−4 imply
−an−t(−aq s t−2aq 2s 3t−4+ a q2s+qs 2t−3 ) = a q2s+qs+1 t−2 ,
whichgives(14).ByCorollary3.1withs= n− t+ 1 weobtain N − α0 αn−2 = 1, and taking(18) intoaccountwe get
N(an−t) = (−1)nN(at−2).
Then (13) andthepreviousrelationyield
N(an−2t+1) = (−1)nN(a2t−3). 2
Proposition 3.7. Let f (x) be a qs-polynomial with qs-degree n− 2 and with maximum kernel.Ifthecoefficientofxqs iszero,thenn isevenandf (x)= α Trqn/q2(βx) forsome
α,β∈ F∗qn.
Proof. We mayassumef (x)= a0x+ a1xq
s
+· · · + an−3xqs(n−3)− xqs(n−2) with a
1= 0.
Bythesecondequation of(12),itfollowsthatan−3= 0.Bythethirdequationof (12), we getthataj= 0 for everyoddintegerj∈ {3,. . . ,n− 3}.Ifj iseventhenwehave
aj = (−1) j
2aq
sj+qs(j−2)+···+q2s+1
0 . (19)
Ifn−3 iseven,thenthisgivesusacontradictionwithj = n−3.Itfollowsthatn−3 isodd andhencen iseven.ByN(a0)= (−1)n,thereexistsλ∈ F∗qn suchthata0=−λ1−q
s(n−2) . So,by(19) wegetaj = λq js−qs(n−2) ,and hence f (x) = Trqn/q2(λx) λqs(n−2) . 2
In the next sections we list all the qs-polynomials of Fqn with maximum kernel for n≤ 6. ByCorollaries 3.4and 3.5 then≤ 3 case canbe easily describedhence wewill consider onlythen∈ {4,5,6} cases.
For f (x) = n−1i=0 aixq i ∈ ˜Ln,q we denote by f (x)ˆ := n−1i=0 aq n−i i xq n−i the adjoint (w.r.t.thesymmetricnon-degeneratebilinear formdefinedbyx,y= Tr(xy))off .
By [1, Lemma 2.6],see also [2, pages 407–408],the kernel of f and f hasˆ the same dimensionand hencethefollowingresultholds.
Proposition 3.8. Iff (x)∈ ˜Ln,q is aqs-polynomialwith maximumkernel, thenf (x) isˆ a
qn−s-polynomialwithmaximum kernel.
3.2. The n= 4 case
InthissectionwedeterminethelinearizedpolynomialsoverFq4withmaximumkernel.
Withoutlossofgenerality,wecansupposethattheleadingcoefficientofthepolynomial is−1.
Because of Proposition 3.8, we canassume s = 1. Corollaries 3.4 and 3.5 cover the caseswhentheq-degreeoff is1 or3 sofromnowonwesupposef (x)= a0x+a1xq−xq
2
. Ifa1= 0 thenwecanuseagainCorollary3.5andwegeta0x− xq
2
,withNq4/q2(a0)= 1.
Supposea1= 0.ByEquation(12),wegettheconditions
a0(aq 2 0 + a q2+q 1 ) = 1, a1=−aq+10 a q2 1 , whichisequivalentto Nq4/q(a0) = 1, aq+11 = aq02+q+1− aq0, see(A1) ofSection5.
InTable1we listtheq-polynomialsof L4,q withmaximum kernel,upto anon-zero
scalarin F∗q4. Applying theadjoint operation we can obtain thelist of q3-polynomials
overFq4 withmaximumkernel.Inthetabletheq-degreewillbe denotedbyk.
Table 1
LinearizedpolynomialsofFq4 withmaximumkernelwiths= 1.
k polynomial form conditions
3 Tr(λx) λ∈ F∗q4 2 a0x− xq 2 Nq4/q2(a0) = 1 2 a0x + a1xq− xq 2 Nq4/q(a0) = 1 aq+11 = a q2+q+1 0 − a q 0 1 a0x− xq Nq4/q(a0) = 1 3.3. The n= 5 case
InthissectionwedeterminethelinearizedpolynomialsoverFq5withmaximumkernel.
Withoutlossofgenerality,wecansupposethattheleadingcoefficientofthepolynomial is−1.BecauseofProposition3.8,wecanassumes∈ {1,2}.Corollaries3.4and3.5cover thecaseswhentheqs-degreeoff is1 or4.Firstwesupposethatf hasqs-degree3,i.e.
f (x) = a0x + a1xq
s
+ a2xq 2s
From (12), f (x) hasmaximum kernel ifand only ifa0, a1 and a2 satisfy thefollowing system: ⎧ ⎪ ⎨ ⎪ ⎩ a1=−aq s+1 0 a q2s 2 , −aq3s+q2s+1 0 a q4s 2 + a q2s+qs 2 a0= 1, a2=−aq 2s+1 0 + a q3s+q2s 2 a q2s+qs+1 0 , whichis equivalentto ⎧ ⎪ ⎨ ⎪ ⎩ N(a0) = 1, a1=−aq s+1 0 a q2s 2 , −aq3s+q2s+1 0 a q4s 2 + a0aq 2s+qs 2 = 1,
see(A2)ofSection5.
Supposenow thattheqs-degreeis2,i.e. f (x) = a0x + a1xq
s − xq2s.
ByTheorem3.3thepolynomialf (x) hasmaximumkernelifandonlyifitscoefficients satisfy a0(aq 2s 0 a q3s 1 + a qs 1 (a q3s 0 + a q3s+q2s 1 )) = 1, aq0s+1(aq03s+ aq13s+q2s) + a1= 0, whichis equivalentto N(a0) =−1, a0qs+ aq1s+1= a0q2s+qs+1aq13s, see(A3)ofSection5.
In Table 2we list the qs-polynomials,s ∈ {1,2} of L
5,q with maximum kernel, up
to a non-zero scalar in F∗q5. Applying the adjoint operation we can obtain the list of
qt-polynomials, t ∈ {3,4}, over F
q5 with maximum kernel. As before, the qs-degree is
denotedbyk.
Table 2
LinearizedpolynomialsofFq5withmaximumkernelwiths∈ {1,2}.
k polynomial form conditions
4 Tr(λx) λ∈ F∗ q5 3 a0x + a1xq s + a2xq 2s − xq3s ⎧ ⎪ ⎨ ⎪ ⎩ N(a0) = 1 a1=−aq s+1 0 a q2s 2 −aq3s+q2s+1 0 a q4s 2 + a0aq 2s+qs 2 = 1 2 a0x + a1xq s − xq2s N(a0) =−1 aq1s+1+ a qs 0 = a q3s 1 a q2s+qs+1 0 1 a0x− xq s N(a0) = 1
3.4. The n= 6 case
InthissectionwedeterminethelinearizedpolynomialsoverFq6withmaximumkernel.
Without loss of generality, we can suppose that the leading coefficient of the polyno-mial is −1. Because of Proposition 3.8, we can assume s = 1. Corollaries 3.4 and 3.5
coverthe cases whenthe q-degree of f is1 or 5.As before, denote by k theqs-degree
of f .
Wefirstconsiderthecasek = 2,i.e.f (x)= a0x+ a1xq
s − xq2s
.ByTheorem3.3,f (x) hasmaximumkernelifandonlyifthecoefficientssatisfy
⎧ ⎪ ⎨ ⎪ ⎩ N(a0) = 1, (aq0+ aq+11 )q3 = aq05+q4+q3(aq0+ aq+11 ), a1q4aq03+ aq12(a0q4+ aq14+q3) =− a1 aq+10 ,
see(A4) ofSection5.
Ifk = 3,thenf (x)= a0x+a1xq
s
+a2xq 2s
−xq3s
,andbyTheorem3.3ithasmaximum kernelifandonlyifthe coefficientsfulfill
⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, aq03+q+1+ aq23a1q2aq+10 − aq2a1= aq0, a2q+1=−a0q3+q2+q+1aq14− aq1, aq+11 = a2aq0+ a q2+q+1 0 a q3 2 ,
see(A5) ofSection5.Notethata1= 0 ifandonlyifa2= 0 andinthiscasewegetthe
traceover Fq3.
Finally,letk = 4.Thenthepolynomialf (x)= a0x+ a1xq
s + a2xq 2s + a3xq 3s − xq4s
hasmaximumkernelifandonlyifthecoefficientssatisfy ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, a0(−aq 4+q2 0 + a q5+q4 3 a q4+q3+q2 0 + a q2+q 3 ) = 1, a1=−aq+10 a q2 3 , a2=−aq 2+1 0 + a q3+q2 3 a q2+q+1 0 , a3= aq 4 3 a q3+q2+1 0 + a q2 3 a q3+q+1 0 − a q3+q2+q+1 0 a q4+q3+q2 3 ,
see(A6) ofSection5.
InTable3we listtheq-polynomialsof L6,q withmaximum kernel,upto anon-zero
scalarin F∗q6. Applying theadjoint operation we can obtain thelist of q5-polynomials
Table 3
LinearizedpolynomialsofFq6withmaximumkernelwiths= 1.
k polynomial form conditions 5 Trq6/q(λx) λ∈ F∗q6 4 a0x + a1xq+ a2xq 2 + a3xq 3 − xq4 ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ a1= 0 N(a0) = 1 a0(−aq 4+q2 0 + a q5+q4 3 a q4+q3+q2 0 + a q2+q 3 ) = 1 a1=−aq+10 a q2 3 a2=−aq 2+1 0 + a q3+q2 3 a q2+q+1 0 a3= aq 4 3a q3+q2+1 0 + a q2 3 a q3+q+1 0 − a q3+q2+q+1 0 a q4+q3+q2 3 4 Trq6/q2(λx) λ∈ F∗q6 3 a0x + a1xq+ a2xq 2 − xq3 ⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ N(a0) = 1 aq03+q+1+ a q3 2a q2 1a q+1 0 − a q 2a1= aq0 aq+12 =−a q3+q2+q+1 0 a q4 1 − a q 1 aq+11 = a2aq0+ a q2+q+1 0 a q3 2 3 Trq6/q3(λx) λ∈ F∗q6 2 a0x + a1xq− xq 2 ⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ a1= 0 N(a0) = 1 (aq0+ a q+1 1 )q 3 = aq05+q4+q3(a q 0+ a q+1 1 ) aq14a q3 0 + a q2 1(a q4 0 + a q4+q3 1 ) =− a1 aq+1 0 2 a0x− xq 2 Nq6/q2(a0) = 1 1 a0x− xq Nq6/q(a0) = 1 4. Application
As an application of Theorem 1.2 we are able to prove the following result on the splitting fieldofq-polynomials.
Theorem 4.1.Let f (x) = a0x+ a1xq+· · · + ak−1xq
k−1 − xqk
∈ Fqn[x] with a0 = 0
and let A be defined as in (1). Then the splitting field of f (x) is Fqnm where m is the
(multiplicative) order ofthematrixB := AAq· · · Aqn−1.
Proof. The derivativeof f (x) isnon-zeroandhencef (x) has qk distinct rootsinsome
algebraic extension of Fqn. Suppose that Fqnm is the splitting field of f (x) and let t
denote the order of B. Then the kernel of the Fq-linear Fqnm → Fqnm map defined as
x→ f(x) hasdimensionk over Fqand hencebyTheorem 1.2wehave
AAq· · · Aqnm−1 = Ik.
Since thecoefficientsofA areinFqn,thisisequivalenttoBm= Ik andhencet| m.On
theother hand
andhenceagainbyTheorem1.2thekerneloftheFq-linearFqnt→ Fqnt mapdefinedas x→ f(x) hasdimensionk over Fq. ItfollowsthatFqnm isasubfieldofFqnt from which
m| t. 2
AfurtherapplicationofTheorem1.2 isthefollowing.
Theorem4.2.Letn,m,s andt bepositiveintegerssuchthatgcd(s,nm)= gcd(t,nm)= 1
and s ≡ t (mod m). Let f (x) = a0x+ a1xq
s +· · · + ak−1xq s(k−1) − xqsk and g(x) = a0x+ a1xq t +· · · + ak−1xqt(k−1)− xqtk,wherea 0,a1,. . . ,ak−1∈ Fqm.The kernelof f (x)
consideredas alineartransformation ofFqnm has dimensionk if andonly ifthekernel
ofg(x) considered asalinear transformationof Fqnm has dimensionk.
Proof. DenotebyA thematrixassociatedwithf (x) asin(1).Byhypothesis,A∈ Fk×kqm
and it is the same as the matrix associated with g(x). By Theorem 1.2 the kernel of
f (x),consideredasalineartransformation ofFqnm,hasdimensionk ifandonlyif
AAqs· · · Aqs(nm−1) = Ik.
Sinces≡ t (mod m),wehave
AAqs· · · Aqs(nm−1) = AAqt· · · Aqt(nm−1) = Ik,
and, againbyTheorem 1.2, thisholds ifand onlyifthekernel ofg(x), consideredasa lineartransformationofFqnm,hasdimensionk. 2
Addendum
Duringthe“Combinatorics2018”conference,thefourthauthor presentedtheresults ofthispaperinthetalkentitled“Onq-polynomialswithmaximumkernel”.Inthesame conferenceJohnSheekeypresentedajointworkwithGaryMcGuire[8] inhistalkentitled “Ranksof LinearizedPolynomials and Roots of ProjectivePolynomials”.It turned out that, independently from the authors of the present paper, they also obtained similar results.
5. Appendix
Inthissectionwedevelopsomecalculationsregardingtherelationsonthecoefficients ofalinearizedpolynomialswithmaximumkernelpresentedinSections3.2,3.3and3.4, seealso[13].
(A1) ByEquation(11) withn= 4,s= 1 andk = 2,wegettheconditions
Σ : a0(aq 2 0 + a q2+q 1 ) = 1, a1=−aq+10 a q2 1 .
ByCorollary3.1, thesystemΣ isequivalentto thefollowingsystem Σ: ⎧ ⎪ ⎨ ⎪ ⎩ Nq4/q(a0) = 1, a0(aq 2 0 + a q2+q 1 ) = 1, a1=−aq+10 a q2 1 ,
whichcanberewrittenas follows
Σ: ⎧ ⎪ ⎨ ⎪ ⎩ Nq4/q(a0) = 1, aq12−1=− 1 aq+10 , aq+11 = aq02+q+1− aq0. Now considerthesystem
Σ∗:
Nq4/q(a0) = 1,
aq+11 = aq02+q+1− aq0.
Clearly, S(Σ)⊆ S(Σ∗),where S(Σ) andS(Σ∗) denotetheset of solutionsofΣ andΣ∗,respectively.Let(a0,a1)∈ S(Σ∗),thenbyusingthenormconditionona0
aq12−1= 1 aq03 − a q 0 q−1 = 1− aq+q0 3 aq03 q−1 = = 1− a 1+q2 0 1− aq+q0 3a q3−1 0 = 1− 1 aq+q30 1− aq+q0 3a q3−1 0 =− 1 aq+10 , i.e. (a0,a1)∈ S(Σ) and henceS(Σ∗)= S(Σ)= S(Σ).
(A2) From (11) withn= 5,gcd(s,5)= 1 andk = 3,wegetthefollowing conditions:
Σ : ⎧ ⎪ ⎨ ⎪ ⎩ a0(aq 2s 1 + a q2s+qs 2 ) = 1, a1=−aq s +1 0 a q2s 2 , a2=−aq 2s+1 0 − a q2s 2 a qs 1 a0. ByCorollary3.1, Σ isequivalentto Σ: ⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ Nq5/q(a0) = 1, a0(aq 2s 1 + a q2s+qs 2 ) = 1, a1=−aq s+1 0 a q2s 2 , a2=−aq 2s+1 0 − a q2s 2 a qs 1 a0,
Σ: ⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ Nq5/q(a0) = 1, a1=−aq s+1 0 a q2s 2 , −aq3s+q2s+1 0 a q4s 2 + a q2s+qs 2 a0= 1, a2=−aq 2s+1 0 + a q3s+q2s 2 a q2s+qs+1 0 .
Byraisingthethirdequationtoqsandmultiplyingbyaq2s+1
0 ,sinceN(a0)= 1,we
getthefourthequation.Therefore Σ, andhenceΣ,isequivalentto ⎧ ⎪ ⎨ ⎪ ⎩ Nq5/q(a0) = 1, a1=−aq s +1 0 a q2s 2 , −aq3s+q2s+1 0 a q4s 2 + a0aq 2s+qs 2 = 1.
(A3) ApplyingTheorem3.3withn= 5 andk = 2,wegetthatthepolynomialf (x) has maximumkernelifandonlyifitscoefficientssatisfy
Σ : Q0,5= a0(aq 2s 0 a q3s 1 + a qs 1 (a q3s 0 + a q3s+q2s 1 )) = 1, Q1,5= aq s 0 (a q3s 0 + a q3s+q2s 1 ) + a1(aq 2s 0 a q3s 1 + a qs 1 (a q3s 0 + a q3s+q2s 1 )) = 0, whichisequivalentto ⎧ ⎪ ⎨ ⎪ ⎩ Nq5/q(a0) =−1, a0(aq 2s 0 a q3s 1 + a qs 1 (a q3s 0 + a q3s+q2s 1 )) = 1, aq0s+1(aq03s+ aq13s+q2s) + a1= 0,
becauseofCorollary3.1andsinceaq02saq13s+ a1qs(aq03s+ aq13s+q2s)= 1 a0
.Theabove systemcanberewrittenas follows
⎧ ⎪ ⎨ ⎪ ⎩ Nq5/q(a0) =−1, aq03s+ aq13s+q2s =− a1 aqs +10 , aq13sa0q2s+qs+1− aq1s+1= aq0s, whichisequivalentto ⎧ ⎪ ⎨ ⎪ ⎩ Nq5/q(a0) =−1, a1qs+1+ aq0s= a1q3saq02s+qs+1, a1aq 4s+q3s+q2s 0 =−aqs +1a1 0 .
Ifthefirstandthesecondequationsaresatisfied,clearlyalsothelastoneisfulfilled, henceΣ isequivalent tothefollowingsystem
Nq5/q(a0) =−1,
(A4) By Theorem3.3, withn= 6,s= 1 andk = 2,weget Q0,6= a0(aq 2 0 (a q4 0 + a q4 +q3 1 ) + a q 1(a q3 0 a q4 1 + a q2 1 (a q4 0 + a q4 +q3 1 ))) = 1, Q1,6= aq 2 0 a1(aq 4 0 + a q4 +q3 1 ) + (a q+1 1 + a q 0)(a q3 0 a q4 1 + a q2 1 (a q4 0 + a q4 +q3 1 )) = 0, whichisequivalent to a0q2(aq04+ a1q4+q3) + aq1(aq03aq14+ aq12(a0q4+ aq14+q3)) = a1 0, a1 a0 + a q 0(a q3 0 a q4 1 + a q2 1 (a q4 0 + a q4+q3 1 )) = 0, i.e. a0q2(aq04+ a1q4+q3) + aq1(aq03aq14+ aq12(a0q4+ aq14+q3)) = a1 0, a0q3aq14+ aq12(a0q4+ aq14+q3) =− a1 aq+10 . ByCorollary3.1, theprevioussystemisequivalentto
⎧ ⎪ ⎨ ⎪ ⎩ N(a0) = 1, a0q2(aq04+ a1q4+q3) + aq1(aq03aq14+ aq12(a0q4+ aq14+q3)) = a1 0, a0q3aq14+ aq12(a0q4+ aq14+q3) =− a1 aq+10 , whichisequivalent to ⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩ N(a0) = 1, aq02(aq0+ aq+11 )q3 −aq+11 aq+1 0 = 1 a0, a0q3aq14+ aq12(a0q4+ aq14+q3) =− a1 aq+10 , henceitisequivalentto ⎧ ⎪ ⎨ ⎪ ⎩ N(a0) = 1, (aq0+ aq+11 )q3 = aq05+q4+q3(aq0+ aq+11 ), a1q4aq03+ aq12(a0q4+ aq14+q3) =− a1 aq+10 . (A5) By Theorem3.3with n= 6,s= 1 andk = 3,weget
⎧ ⎪ ⎨ ⎪ ⎩ a0Qq2,5 = 1, Qq0,5+ a1Qq2,5= 0, Qq1,5+ a2Qq2,5= 0, where Q0,5= a0(aq 2 1 + a q2+q 2 ), Q1,5= aq0a q2 2 + a1(aq 2 1 + a q2+q 2 ), Q2,5= aq 2 0 + a q2 2 a q 1+ a2(aq 2 1 + a q2+q 2 ),
hencewe obtainthefollowingsystem ⎧ ⎪ ⎨ ⎪ ⎩ a0(aq 3 0 + a q3 2 a q2 1 + a q 2(a q3 1 + a q3+q2 2 )) = 1, a1 a0 + a q 0(a q3 1 + a q3+q2 2 ) = 0, a2 a0 + a q3 2 a q2 0 + a q 1(a q3 1 + a q3+q2 2 ) = 0. ByCorollary 3.1itisequivalentto ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, a0(aq 3 0 + a q3 2 a q2 1 + a q 2(a q3 1 + a q3+q2 2 )) = 1, aq13+ aq23+q2 =− a1 a1+q0 , a2 a0 + a q3 2 a q2 0 + a q 1(a q3 1 + a q3+q2 2 ) = 0,
bysubstitutingthethirdequation intotheothersweget ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, a0(aq 3 0 + a q3 2 a q2 1 − aq2a1 a1+q0 ) = 1, aq13+ aq23+q2=− a1 a1+q0 , a2 a0 + a q3 2 a q2 0 − aq+11 a1+q0 = 0, i.e. ⎧ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, aq03+q+1+ aq23a1q2aq+10 − aq2a1= aq0, a2q+1=−a0q3+q2+q+1aq14− aq1, aq+11 = a2aq0+ a q2+q+1 0 a q3 2 .
(A6) Equations(11) withn= 6,s= 1 andk = 4 are ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ a0(aq 2 2 + a q2+q 3 ) = 1, aq0aq32+ a1(aq 2 2 + a q2+q 3 ) = 0, aq02+ aq32aq1+ a2(aq 2 2 + a q2+q 3 ) = 0, aq12+ aq32aq2+ a3(aq 2 2 + a q2+q 3 ) = 0,
which,byCorollary3.1,isequivalent to ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, a0(aq 2 2 + a q2+q 3 ) = 1, aq0aq32+ a1(aq 2 2 + a q2+q 3 ) = 0, aq02+ aq32aq1+ a2(aq 2 2 + a q2+q 3 ) = 0, aq12+ aq32aq2+ a3(aq 2 2 + a q2+q 3 ) = 0,
thusitcanberewrittenas follows ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, aq22+ aq32+q=a1 0, aq0aq32+a1 a0 = 0, aq02+ aq32aq1+a2 a0 = 0, aq12+ aq32aq2+a3 a0 = 0, and hence ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, a0(aq 2 2 + a q2+q 3 ) = 1, a1=−aq+10 a q2 3 , a2=−aq 2+1 0 − a q2 3 a q 1a0, a3=−aq 2 1 a0− aq 2 3 a q 2a0, i.e. ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ N(a0) = 1, a0(−aq 4+q2 0 + a q5+q4 3 a q4+q3+q2 0 + a q2+q 3 ) = 1, a1=−aq+10 a q2 3 , a2=−aq 2+1 0 + a q3+q2 3 a q2+q+1 0 , a3= aq 4 3 a q3+q2+1 0 + a q2 3 a q3+q+1 0 − a q3+q2+q+1 0 a q4+q3+q2 3 . References
[1]D.Bartoli,M.Giulietti,G.Marino,O.Polverino,Maximumscatteredlinearsetsandcompletecaps inGaloisspaces,Combinatorica38 (2)(2018)255–278.
[2]B.Csajbók,G.Marino,O.Polverino,ClassesandequivalenceoflinearsetsinPG(1,qn),J.Comb. Theory,Ser.A157(2018)402–426.
[3]P.Delsarte,Bilinearformsoverafinitefield,withapplicationstocodingtheory,J.Comb.Theory, Ser.A25(1978)226–241.
[4]U.Dempwolff, J.C.Fisher,A.Herman,SemilineartransformationsoverfinitefieldsareFrobenius maps,Glasg.Math.J.42 (2)(2000)289–295.
[5]E.Gabidulin,Theoryofcodeswithmaximumrankdistance,Probl.Inf.Transm.21 (3)(1985)3–16.
[6]R.Gow,R.Quinlan,Galoistheoryandlinearalgebra,LinearAlgebraAppl.430(2009)1778–1789.
[7]R.Gow,R.Quinlan,Galoisextensionsandsubspacesofalterningbilinearformswithspecialrank properties,LinearAlgebraAppl.430(2009)2212–2224.
[8] G. McGuire, J. Sheekey, Acharacterization of thenumber of roots of linearized and projective polynomialsinthefieldofcoefficients, https://arxiv.org/abs/1806.05853.
[9]S.Lang,Algebraicgroupsoverfinitefields,Am.J.Math.78(1956)555–563.
[10]R.Lidl,H.Niederreiter,FiniteFields,vol. 20,CambridgeUniversityPress,1997.
[11]O.Ore,Onaspecialclassofpolynomials,Trans.Am.Math.Soc.35(1933)559–584.
[12]J.Sheekey, Anewfamilyof linear maximumrank distance codes,Adv.Math.Commun. 10 (3) (2016)475–488.
[13]F.Zullo,LinearCodesandGaloisGeometries:BetweenTwoWorlds,PhDthesis,Universitàdegli StudidellaCampania“LuigiVanvitelli”,2018.