(Note that the possible f(x) is not unique.) We now prove that
f(x)=(x2+1)(x2+2)(x2−2)(x2+7)
satisfies the conditions. It is easy to see that f(x)=0 has no rational real root.
1. First we prove: for every power of an odd prime pα, there exists an integer mpα such that
pα∣f(mpα).
In fact, we can prove the following lemma by induction:
Lemma 1. For every odd prime p, there exists rp∈{−1,−2,2}, such that for all α≥1, there must exist mpα∈Z such that pα∣mp2−rp.
a. First, for α=1, note that:
- If p=4k+1, then −1 is a quadratic residue modulo p;
- If p=8k+3, then −2 is a quadratic residue modulo p;
- If p=8k+7, then 2 is a quadratic residue modulo p.
Combining the above, there must exist mp∈Z and rp∈{−1,−2,2}, such that p∣mp2−rp.
b. Suppose the proposition holds for α=k. When α=k+1, take mpk+1=mpk+tpk, where t is to be determined. Note that we have
mpk+12−rp=(mpk+tpk)2−rp≡(mpk2−rp)+2mpktpk(modpk+1).
By the induction hypothesis, pk∣mpk2−rp, so it suffices to take
t≡−pkmpk2−rp(2mpk)−1(modp),
to make pk+1∣mpk+12−rp hold (Note: clearly p does not divide mpk, so (2mpk)−1 exists.)
Combining the above, Lemma 1 is proved, and hence the original proposition holds when n is a power of an odd prime.
2. Next we prove: for every power of 2, 2α, there exists an integer m2α such that 2α∣f(m2α).
In fact, we can prove the following lemma by induction:
Lemma 2. For all α≥1, there must exist m2α∈Z such that 2α∣mp2+7.
a. First, for α≥3, take m2α=1.
b. Suppose the proposition holds for α=k≥3. When α=k+1, take
m2k+1=m2k+t2k−1,
where t is to be determined. Note that we have
m2k+12+7=(m2k+t2k−1)2+7≡(m2k2+7)+m2kt2k(mod2k+1).
By the induction hypothesis, 2k∣m2k2+7, so it suffices to take
t≡2km2k2+7(mod2),
to make 2k+1∣m2k+12+7 hold.
Combining the above, Lemma 2 is proved, and hence the original proposition holds when n is a power of 2.
- From the two lemmas above, we know that for all 1≤i≤s there exists an integer mpiαi such that piαi∣f(mpiαi).
- Also, for any polynomial f(x) with integer coefficients, and any two distinct integers u,v, it must be that
u−v∣f(u)−f(v).
Hence, as long as we let m satisfy the system of congruences
m≡mpiαi(mod piαi)∀i=1,2,…,s
we then have
f(m)=[f(m)−f(mpiαi)]+f(mpiαi)≡[m−mpiαi]+0≡0(modpiαi)
holding for all i=1,2,…,s; in other words, n∣f(m).
Combining the above, we know that f(x) satisfies the requirements of the problem.