J.
W ó j c i k ( W a r s z a w a )On the composite Lehiner numbers with prime indices, II
A. Schinzel has deduced from his Conjecture H (cf. [2], p. 188) a theorem about the so-called Lehmer numbers
where a, ft are roots of the trinomial: zi - L x,iz-\-M and L, M are rational integers.
More exactly, he proved the following two theorems:
I. I f LM Ф О, К = L — 4cM Ф 0 and none of the numbers —K L ,
~ 3 K L , — K M , —3 K M is a perfect square, or each of the numbers K , L is a perfect square, then there exists an integer Тс > 0 such that for every integer D Ф 0 one can find a prime q satisfying
(1) q\P s and ( s , D ) = 1,
II. Under the assumptions of I, Conjecture H implies the existence of infinitely many primes p such that P p(a,
/9)is composite.
The aforesaid conjecture reads as follows:
H. I f f l , f 2, . . . , f k are irreducible polynomials with integral coeffi
cients and the highest coefficients positive and such that f 1(x)f2(x) ... f k(x) has no fixed factor greater than 1, then for infinitely many integers x , fi(x) (i = 1, ..., Jc) are primes.
The aim of this paper is to prove the following theorems:
Th e o r e m
1. I f L M Ф 0,
(L , M) = 1, (\L\, \M\} Ф
< 2 ,1>, К =
= L — IM Ф 0, —K L is a perfect square and either 1Q\L or фЪМ and
± 5 3M are not perfect biquadrates, then there exists an integer ~k > 0 such that for every integer D ф 0 one can find a prime q satisfying
Pn(a, /9) = (an— j3n)[a — fi, if n is odd, (an—(3n) /a2 — /92, if n is even
where Jacobi's symbols of quadratic
character.
10G J. W ó j c i k
Theorem
2. I f L M Ф 0 , (L , M) — 1, <|L|, \M\) Ф ( 1 , 1 ) , <3, 1 ), К — L — 1M Ф 0, —3KL is a perfect square and either 27 f L or ± 7 M and ± 75M are not perfect sixth powers, then there exists an integer Тс > 0 such that for every integer T) Ф 0 one can find a prime q satisfying:
q\P(ą-i)lk and
q - i = i .
Theorem 3.
Under the assumption of Theorem 1 or 2, Conjecture H implies the existence of infinitely many primes p such that P p{a, fi) is com
posite.
The assumption (L , M) = 1 does not essentialy affect generality and simplifies a little the formulation of the theorems. Let us remark that Theorems 1 and 2 can be proved with g - 1
Те replaced throughout by g + 1
Те The proof is based on the arithmetic of Gauss’s field K( i) and of Eisensteins field K (
q). The only laws of reciprocity used are quadratic, cubic and biquadratic. As to the notation, we denote by A (a) the absolute norm of an element a of K( i) or K (
q). An integer a is called primitive if it is not divisible by a rational integer > 1, primary if a
~ 1 m od(l + ^)3 for aeK( i ) , a = —lm od 3 for a e l ( o ) (cf. [1], p. 152 and 187). are Jac°b i’s symbols of cubic and biquadratic char
acter, respectively. Tc2(a)
= k ( a ) , h 3(a)and &4(a) denote the square- free, the cube-free and the biquadrate-free kernel* of a, respectively
( а Ф
0). If a is primitive and
a'is its conjugate, then clearly (1) Nke{a) = ke(a)lce{a') = JceN{a) (e = 2, 3 or 4).
We begin with four elementary lemmas.
Lemma
1. For any biquadrate-free, primary, primitive integer aeK(i ) and for any rational integer g there exists a rational integer v(a, g) such that
(2) ((~ ~ a~~)) = i0 and И а ’ =
except for
g = 0m od4, A (a ) = 5 or 53;
(3) g = 2m od4, N (a) = 1;
g = 1 or 3m od4, A (a) is a perfect square.
P r o o f. The lemma is true if N{a) is a prime. Indeed, if Ж (a) > 5 , then
( Ж(a) —1)/4 > 1; thus for every value of
gthere exist at least two residue classes rmodiV^a) satisfying ((—]) = ia.
\\a//
Since N (a) is prime, at least one of them satisfies (r — 1, N {a)) = 1.
If Ж(a) = 5 and g Ф 0m od4, the existence of v(a, g) is obvious.
Suppose that Ж (a) is not a perfect square. If there exists a prime factory of Ж (a) such that
(4) <* = р1<*ц Р = Я ( р 1) > 5 , 1 = 1 or 3, then we define v(a,g) by the system of congruences
v{a, g)
— lmodJVr(a1),
v
\P
i,9£ +
Ж (cq)— 1
m od p .
If there exists no prime factor p satisfying (4), then we have
a—
Pi«i
j P — Ж{px) = 5, £ = 1 or 3
and either g ф 0mod4 or ax = q\a2, where q = Ж(qx) is a prime.
We put in the first case
( - I m o d N K ) , via, g) = {
I
v(Pi,#£)mod5, an d. in the second case
j — l m o d 5 N ( a 2),
via, q) — I
I v{qx, l)m od g.
Now, let Ж (a) be a perfect square and Ж (а) ф 1. We have to con
sider only the case g = 0 or 2 mod 4.
We put a = p\ax-, p = Ж(px) a prime,
— lmodJ\r(a1), v(Pi, g/2)modp.
Finally, if Ж (a) = 1 and g = 0mod4 we put v(a, g) = 1. It follows from the Chinese Remainder Theorem that the above systems of congruences are solvable and clearly any of their solutions satisfies (2).
Lemma 2.
I f a is any biquadrate-free, primary, primitive integer of K ( i ) , g arbitrary rational integer, and none of the exceptions (3) holds, then for any integer у > 1 there exists a positive integer m such that (5) m = 2y-f- lm o d 2 y+2,
and (m— 1, Ж {a)) — 1.
a
m
108 J. W ó jc i k
P r o o f. If we put
m = l - f 2 vmod2y+2, v(a, g)mo6.N(a),
it follows from the biquadratic reciprocity law that m satisfies (5).
Lemma 3.
For any square-free, primary, primitive integer aeK( o) and for arbitrary rational integers g, Ti there exists an integer v( a, g) such that
(6)
[
у ( д >9, h)
a
] Q°t y(g? N( a)
9, ty( - I f
and {v(a, g, h) — 1, N (a)) = 1 except for
# = 0 m od3, h = 0m od2, N (a) = 7, g ф 0mod3 or h ф 0m od2, N (a) = 1.
P r o o f. The lemma is true if N (a) is a prime.
Indeed, if N ( a ) > 7 , then (N(a) — l ) /6 > 1; thus for each value of g and h there exist at least two residue classes r m o d l(a ) satisfying
r
~N (a) ( - I f .
Since N (a) is a prime, at least one of them satisfies [г— 1,Ж (а)) — 1.
If N (a) — 7 and g =£ 0mod3 or h Ф 0mod2 the existence of v(a, g, h) is obvious. If N(a) is composite, and p — Ж( рг) is its greatest prime factor, we put
— lm o d N(a)/p, v(a, g, h)
v
Pi, 9, Ъ Щ* ) 1р- 12
m od p .
Finally, if N (a) — 1, g = 0m od3, h = 0mod2 we put v(a, g, h) = 1.
Lemma 4.
Let a, (3 be any primary, primitive integers of the field K (
q) such that a is cube-free and is square-free. Let g , h be any rational integers and у, 3 any positive integers. There exists a positive integer m such that
m = 2<5-b lm o d 2 <5+1, m = 3y-f lm o d 3 y+1,
(7)[ — 1 = ^ , Ш Ш = ( - 1)л and (m — 1, N (ap)\ — 1;
L
mJ
\ m 1except for
g = 0m od3, h = 2<5-1mod2, N(a) = 7 or 72, N (ft) = 7;
7 i^ 0 m o d 2 , N((3) — 1.
# ^ 0 m o d 3 , N(a) = l ,
P r o o f. Let
В — (Na, N/3), a = B xax, — D 2(3x, where
N ( Bx) =—
N ( B2)—
B , [N(oq),
N(/Si)) = 1. Further, let
dx=
( Bx,c^),
d2=
B xldx,a2 =
ax/dx;
<dx, d 2,
2 , 1 ) , if
N( dx) ^ N ( d 2),* ^4? S у Щ ) —
<<*а, dlf 1 , 2>, if N{ dx) < Sid, ).
Clearly N( ds) > N( d4) ‘, a = d\dla2, N ф) — N( d3) N(d4)N(fix) and by the assumptions about a and /3, the numbers N( d3), N(d4), N (a2) and N ({3X) are relatively prime in pairs. Moreover, a2 is cube-free and pri
mitive, and thus it can be represented in the form a2 = a\ a4, where (Na3, J a 4) = 1 and a3, a4 are square-free. We put
<«5, аб, С
у'd') — <«3 J a4 ? ^ , 1^>,
<a4 ? аз > 1 ^ ?
if JV(as) ^ ^ ( a 4), if W(a3) < N (a4).
Clearly ЗУ(а5) ^ N( a6), a2 = agCtg and (3Va5, JVa6) = 1 .
We shall define w by a system of congruences different for each of the following 6 cases.
1. N( d3) > 7 or N( d3) = 7, g ф 0m od3:
— 1 mod N (d4 a2 fix),
v(d3, eg, h + 2s^ { N { f ) ) - l ) + - (dt^?) ~-1- )modJV-(d,), 2<3 + lm o d 2 <5+1,
3y+ l m o d 3 y+1;
2. N (d3) — 7, g = 0mod3 and either Ж(| 31) > 1 or W(j31) = l ,
== 2<5_1 + lm o d 2 :
--lm odTV (a),
»(/?!, 0, h + 2 s- i { N(§1) + l) + l\ m oiN (ti1),
tyyi -— ~ ' '
2й+ 1 mod 2d+1, 3y+ ln a o d 3y+1;
3. 3V(c73) — 7, g = 0m od3, N (fix) = 1; h = 2d_1mod 2:
— lm o d W(ag),
у(а5,2 ^ !С ; 0)mod N( as), m ^ 2 mod 7,
2d + lm o d 2d+1,
3y+ lm o d 3y+1, where
п о J. W ó jc ik
4. N(dz) = 1, N( a 2) > 1 , and either N ((Зг) > 1 or h == 0m od2:
where
m =
— lm o d JV'(aJ),
»( f t , *, Ъ + 2"-! (J (ft) - 1 ) ) mod N (ft), r(a5, £#, l)m od (a5),
2a + lm o d 2a+1, 3y+ lm o d 3y+I,
5. N( d3) = N (a2)
= Omod 2:
(O, if ДГ(0 ) = 1 , (1 , if N ( P ) > 1;
= 1; g = 0m od3 and either N {f$x) > 1 or
where
m =
V ( f t , ® , Ь+ а а- 2 ( У ( f t ) - 1 ) ) m o d # ( f t ) ,
2a + lm o d 2a+1, 3y+ lm od 3y+1,
(0 , if N(P) = 1, j l , if
A =
It follows from the Chinese Remainder Theorem that the above systems of congruences are solvable, and from the quadratic and cubic reciprocity law and from Lemma 3 that any of their positive solutions satisfies (7).
P r o o f o f T h e o re m s 1, 2 and 3. For every pair L , M in question we shall give the value of Tc and construct two linear functions f 1(co),f2{x)
= — (fi(oc)~ l) satisfying the following conditions:
К '
(i) f i ( x )i f i { x ) have integral coefficients, are primitive, their leading coefficient divide 12TcKLM2 and
(8)
( f i ( l ) A ( l ) , 2 ) = 1,(ii) if q = /, (x) is a prime - f l l , then (9)
These conditions being satisfied, Theorems 1, 2 and 3 easily follow.
First, by condition (i) and Gauss’ s Lemma the polynomial f x(x)f2(x) is primitive of degree 2. Formula (8) implies that it has no fixed factor
> 1 .
Thus for every D ф 0, there exists an integer such that
(10) (/i(®o)/s(*o)> 12 ШЪМгП) = 1
Now by Dirichlet’s theorem, there exist infinitely many primes q == / 1(a?0)mod 12kKLM2l). Clearly, every snch prime q is of the form fx(x). For every such prime we have by (ii) divisibility (9); thus
because qi KL.
On the other hand,
4—~~ = y ( / i ( ® o ) - l ) = /a(® 0)modl>,
thus in view of (10) , 7>j == 1. Since by (8) f 1(x)f2(oc) has no fixed factor > 1 , Conjecture H implies that for infinitely many integers x, 0. = f i i x ) and V = h ( x ) are both primes. On the other hand, for p > N (a, ft) we have by inequality (5') of paper [3]
Thus, for p large enough, \Pp(a, /8)| > к р ф 1 = q and (11) implies that P p{a, fi) is composite.
1. Assume first that —K L is a perfect square. Since K L M ф 0 and (L , M) = 1 we have
= в
V2 S A ,
e ~ ~ W ~
where A is a primary, primitive integer of K( i ) , в4 = 1, t — 0 or 1, s = 0 or 1. Since <|£|, \М\)Ф < 2 ,1>, А ф 1. We put A = U2V~2, where у ^ 2, U is primary; у = 2, U — A , i i t = s = 0. U is not a perfect square if t — 1 or s — 1 (this is possible since A # 0 , 1 ) . Let e = N( U) . We put in Lemma 2: g = 2t — s, a = k4{U). Under the assumptions of Theorem 1 the conditions of this lemma are satisfied and by (1) there exists a positive integer m such that
(12) m — 2v + lm o d 2V+2, ^(*7) \\ _
.2 1~ ,m Jf amd 'm - 1 , fc4(e)) = 1.
We put Jc = 2у, / г(а?) = 2y+2k4l(e)x-\-m. Condition (i) is satisfied trivially. As to (ii) if q = f^ x ) and (q, e) = 1, we have by (12)
(13) g - 1
2y = 1 mod 4 and || —
;2 < — Ssince 11 — ) J, considered for r > 0 , is a character with conductor dividing u
4&4(e). We notice that if q — qxq2 is the decomposition of q into prime fac-
112 J. W ó j c i k
tors in К (i), we have (14) (UIU')iQ- m
Analogously
(15) ( UIU')(9~1)!i = ( ( y j j mod q2.
It follows from (14) and (15) that
mod qx.
(Z7/Z7')(a-1)/4 mod q.
Hence if q is prime we get by (13)
(16) (a/,? )« -Ч'2У = i(a+*)(«-4/*>'( = j«+» = i mod g , and conditions (ii) is satisfied. This completes the proof of Theorem 1 and of the first part of Theorem 3.
2. Assume now that — 3K L is a perfect square. Since K L M Ф 0 and (L , M) = 1, we have (if we neglect the order of a, ft):
a — &
q\ 1 e)5
A , ft
qu( 1 -
q2)
sVb*
A
where A is a primary, primitive integer of K (
q), = 1, t = 0 or l , s = 0 or 1. Since <|.L|, |Ж|> ^ < 1 ,1>, < 3 ,1>; А
Ф—1. We put A = Z72<3-1 = 7 зУ-1, where <5, у > 0, U, V are primary;
ó = l , U = A , if s = 0,
U is not a perfect square if s = 1 (this is possible since А
Ф0 ,1 ), у = 1, V — A , if t = s,
V is not a perfect cube if t ф s (this is possible since А
Ф0, ± 1 ) . Let e = N( U) , f = N{ V) . W e put in Lemma 4: g = r-f-2s, h = s, a — Jc3(V), ft = Tc(TJ). Under the assumptions of Theorem 2 the conditions of this lemma are satisfied and by (1) there exists a positive integer m such that
m = 2ó-f-1 mod 2a+1, m = 3y+ l mod 3y+1, (17)
Г М А 1 = ( M ) = ( _ ! ) • and ( т ~ 1 , Ц е Ш П ) = 1.
L m J \ m /
We put
Tc = 2a3y, f t (cc) = 2a+13y+Ifc(e)fc3( /) r + m .
Condition (i) is satisfied trivially. As to (ii), if q = f i ( o c ) > 0 and (q,fe) = 1? we have by (17)
0 — 1 0 — 1
(18) ^ - = l m o d 2 , - = lm o d 3 ■ И- ą A
J + 2 S\q ( - 1 ) %
XVl I e\
since I — I, I —I considered for r > 0 are cubic or quadratic characters with conductor dividing Nh3(V) — fc3(/) or 4&(e), respectively.
We notice that if q = qxq2 is the decomposition of q into prime factors in K (
q), we have
™ r a ■ [ а з - н а - ш —
Analogously
(20) ( UI U ' f - 1)/3 = j mod q2.
It follows from (19) and (20) that
( Z7/f7')(a_1)/3 = mod q.
Hence, if q is a prime ~/~M, we get by (18)
(21) = ( - 1 f Q~mS { s - t f ~ 1)l2d {U I U 'f ~ 1)l2
= ( — 1) j —J = 1 mod q and
(22) («/£)<«-1)/3V = ( ~ l f q- m %is- ł){q~m \ v / V T ~ 1)lb = е8~*[—] s l m o d ^ . It follows from (21) and (22) that
(а//?)(а~1)/2<5зУ = lm o d q,
i.e. condition (ii) is satisfied. This completes the proof.
References
[1] P. B a c h m a n .n , Die Lehre von der Kreistheilung und ihre Beziehungen zur Zahlentheorie, Leipzig 1872.
[2] A . S c h in z e l et W . S ie r p iń s k i, Sur certaines hypotheses concernant les nombres premiers, A cta Arith. 4 (1958), pp. 185-208.
[3] A . S c h in z e l, On the composite Lehmer numbers with prime indices, I , Prace Mat. this volume, pp. 95-103.
Prace Matematyczne IX. l