Unofficial archive — problems, solutions & results © IMC, reproduced with permission.

IMC / 2011 / Problems / Day 1, P5

IMC 2011 · Day 1 · P5

Let nn be a positive integer and let VV be a (2n1)(2n-1)-dimensional vector space over the two-element field. Prove that for arbitrary vectors v1,,v4n1Vv_1, \dots, v_{4n-1} \in V, there exists a sequence 1i1<<i2n4n11 \le i_1 < \dots < i_{2n} \le 4n-1 of indices such that vi1++vi2n=0v_{i_1} + \dots + v_{i_{2n}} = 0.

(Ilya Bogdanov, Moscow and Géza Kós, Budapest)

Solution (official)

Let V=aff{v1,,v4n1}V = \operatorname{aff}\{v_1, \dots, v_{4n-1}\}. The statement vi1++vi2n=0v_{i_1} + \dots + v_{i_{2n}} = 0 is translation-invariant (i.e. replacing the vectors by v1a,,v4n1av_1 - a, \dots, v_{4n-1} - a), so we may assume that 0V0 \in V. Let d=dimVd = \dim V.

Lemma. The vectors can be permuted in such a way that v1+v2,v3+v4,,v2d1+v2dv_1 + v_2, v_3 + v_4, \dots, v_{2d-1} + v_{2d} form a basis of VV.

Proof. We prove by induction on dd. If d=0d = 0 or d=1d = 1 then the statement is trivial.

First choose the vector v1v_1 such a way that aff(v1,v2,v3,,v4n1)=V\operatorname{aff}(v_1, v_2, v_3, \dots, v_{4n-1}) = V; this is possible since VV is generated by some d+1d+1 vectors and we have d+12n<4n1d + 1 \le 2n < 4n - 1. Next, choose v2v_2 such that v2v1v_2 \ne v_1. (By d>0d > 0, not all vectors are the same.)

Now let ={0,v1+v2}\ell = \{0, v_1 + v_2\} and let V=V/V' = V/\ell. For any wVw \in V, let w~=+w={w,w+v1+v2}\tilde{w} = \ell + w = \{w, w + v_1 + v_2\} be the class of the factor space VV' containing ww. Apply the induction hypothesis to the vectors v~3,,v~4n1\tilde{v}_3, \dots, \tilde{v}_{4n-1}. Since dimV=d1\dim V' = d - 1, the vectors can permuted in such a way that v~3+v~4\tilde{v}_3 + \tilde{v}_4, …, v~2d1+v~2d\tilde{v}_{2d-1} + \tilde{v}_{2d} is a basis of VV'. Then v1+v2,v3+v4,,v2d1+v2dv_1 + v_2, v_3 + v_4, \dots, v_{2d-1} + v_{2d} is a basis of VV.

Now we can assume that v1+v2,v3+v4,,v2d1+v2dv_1 + v_2, v_3 + v_4, \dots, v_{2d-1} + v_{2d} is a basis of VV. The vector w=(v1+v3++v2d1)+(v2d+1+v2d+2++v2n+d)w = (v_1 + v_3 + \dots + v_{2d-1}) + (v_{2d+1} + v_{2d+2} + \dots + v_{2n+d}) is the sum of 2n2n vectors, so wVw \in V. Hence, w+ε1(v1+v2)++εd(v2d1+v2d)=0w + \varepsilon_1 (v_1 + v_2) + \dots + \varepsilon_d (v_{2d-1} + v_{2d}) = 0 with some ε1,,εdF2\varepsilon_1, \dots, \varepsilon_d \in \mathbb{F}_2, therefore i=1d((1εi)v2i1+εiv2i)+i=2d+12n+dvi=0.\sum_{i=1}^{d} \bigl( (1 - \varepsilon_i) v_{2i-1} + \varepsilon_i v_{2i} \bigr) + \sum_{i=2d+1}^{2n+d} v_i = 0. The left-hand side is the sum of 2n2n vectors.

Similar problems

IMC 2004 · Day 2 · P12killeravg 0.2/10 · solved 1% · near-0 97% · disc 0.13
IMC 2010 · Day 2 · P9killeravg 0.5/10 · solved 2% · near-0 93% · disc 0.42
IMC 2021 · Day 2 · P8killeravg 0.0/10 · solved 0% · near-0 99% · disc 0.08
IMC 2024 · Day 1 · P3mediumavg 5.3/10 · solved 40% · near-0 30% · disc 0.55