Scribd Logo
Upload

Welcome to Scribd!

  • 20 views

    Bruce Characerizing Subspaces

    Uploaded by

    niknavin
    The document summarizes a note that characterizes k-flats in geometric designs. It begins by defining geometric designs and examples of them, such as vector spaces and set systems. It then defines properties P(n,k,r,j) and Q(n,k,r,j) that sets of r-flats may satisfy. The theorem proved in the note is that if a geometric design G satisfies a certain condition and n is sufficiently large, then Q(n,k,r,j) characterizes the k-flats of G. This means a set R satisfies Q if and only if R is the set of r-flats contained within some k-flat. The proof uses counting

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Bruce Characerizing Subspaces

    Uploaded by

    niknavin
    20 views6 pages
    The document summarizes a note that characterizes k-flats in geometric designs. It begins by defining geometric designs and examples of them, such as vector spaces and set systems. It then defines properties P(n,k,r,j) and Q(n,k,r,j) that sets of r-flats may satisfy. The theorem proved in the note is that if a geometric design G satisfies a certain condition and n is sufficiently large, then Q(n,k,r,j) characterizes the k-flats of G. This means a set R satisfies Q if and only if R is the set of r-flats contained within some k-flat. The proof uses counting

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document summarizes a note that characterizes k-flats in geometric designs. It begins by defining geometric designs and examples of them, such as vector spaces and set systems. It then defines properties P(n,k,r,j) and Q(n,k,r,j) that sets of r-flats may satisfy. The theorem proved in the note is that if a geometric design G satisfies a certain condition and n is sufficiently large, then Q(n,k,r,j) characterizes the k-flats of G. This means a set R satisfies Q if and only if R is the set of r-flats contained within some k-flat. The proof uses counting

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    20 views6 pages

    Bruce Characerizing Subspaces

    Uploaded by

    niknavin
    The document summarizes a note that characterizes k-flats in geometric designs. It begins by defining geometric designs and examples of them, such as vector spaces and set systems. It then defines properties P(n,k,r,j) and Q(n,k,r,j) that sets of r-flats may satisfy. The theorem proved in the note is that if a geometric design G satisfies a certain condition and n is sufficiently large, then Q(n,k,r,j) characterizes the k-flats of G. This means a set R satisfies Q if and only if R is the set of r-flats contained within some k-flat. The proof uses counting

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 6

    JOURNAL OF COMBINATORIAL THEORY (A) 20, 398-403 (1976)

    Note

    Characterizing k-flats in Geometric Designs

    B. L. ROTHSCHILD*

    University of California, Los Angeles, California

    AND

    N. M. SINGHI**

    Colorado State University, Fort Collins, Colorado, and


    Tata Institute for Fundamental Research, Bombay, India

    Cummutzicated by R. C. Bose

    Received January 8, 1975

    Let V be an n-dimensional vector space over GF(q), and let ZJ be a


    k-dimensional subspace of I/. Let [y] denote the set of r-dimensional
    subspacesof U, and let C(n, k, ,;j) = (I[“,‘] n [:]I: W is an (n -,j)-
    dimensional subspace of V>. C(n, k, u,,j) is the set of cardinalities of
    “r-intersections” of U with all (n -j)-subspaces. In this paper we are
    concerned with the following question: Suppose S is a set of u-subspaces
    of V, and j S 1 = [I], the number of r-subspacesin a k-space over GF(q).
    Then if (1S n [‘:]I: W is an (n -,j)-subspace of V} C C(rz, k, r,j), must
    S = [F] for some k-subspace UC V? In [4, 51the answer was shown to be
    affirmative if I’ = I or j = 1. In this paper we show that the answer is
    affirmative for arbitrary Y and ,j provided that IZis sufficiently large. The
    minimum sufficient size for n is not decided here. The bound we obtain is
    clearly too crude.
    What we actually do here is to prove a theorem for somewhat more
    general objects than vector spaces.One specialcaseis that of graphs. In this
    case we get a theorem related to the Ulam reconstruction conjecture. In
    particular, we can reconstruct the complete graphs from only the number
    of edgesin the maximal subgraphs in certain cases.This also suggeststhe
    q-analog of the reconstruction conjecture.
    * Work partially supported by NSF Grant GP-33.580 and the Alfred P. Sloan Foun-
    dation.
    ** Work partially supported by NSF Grant 40583.
    398
    Copyright Q 1976 by Academic Press, Inc.
    All rights of reproduction in any form reserved.
    ~-FLATS IN GEOMETRIC DESIGNS 399

    Let G be a combinatorial geometry with rank function 01.We will follow


    Crap0 and Rota [l] for terminology and notation. G will be called a
    geometric design if all k-flats (closed sets of rank k) have the same cardi-
    i
    nality i;. We will have 0 = 0, = 1. Geometric designs were first defined
    by Edmonds, Murti, and Young [2], who called them simple matroid
    designs. Many examples of geometric designs are given in [2], and we list
    four of them here.
    (a) n-dimensional projective space over GF(q). The k-flats are
    (k - I)-dimensional subspaces, and we have E = (qk - l)/(q - 1).
    (b) n-dimensional affine space over GF(q). The k-flats are (k - l)-
    dimensional subspaces, we have E = qk--l, k 2 1, 0 = 0.
    (c) The subsetsof an n element set. The k-flats are the setsof size k,
    and k = k.
    (d) The geometric design G formed from a t - (v, k, X) design
    (see [l]) as follows. A t - (v, k, h) design D [4] is a pair (X, L) where X is
    a set with 1X j = zi, L is a family of subsetsof X such that for all B EL,
    1B / = k, and for any t-subsets Y C X, there are exactly h elementsA EL
    such that Y _CA.
    We define a geometric design as follows: the k-flats are all the
    -- k-subsets
    of X if k < t, the numbers of L if k = t, and X itself if k = t + 1. Thus,
    k=kifk<t,t+l=v=jXI.
    Now let G be a geometric design. We will denote by [“,I the set of r-flats
    contained in the set S. If S is a k-flat, then the usual argument of counting
    basesin vector spacescan be easily extended to show that /[“,]I is inde-
    pendent of the choice of k-flat, and that
    --
    r1=ILSII
    = k(k - i) ... (k - (r - 1))
    jq - 1) . .. (F - (T-1)) r>,l

    [I]. Note that r > k implies [z] = 0.


    Let G be a geometric design of rank ~1.Then CG(tz,k, r,j) (or just
    C(n. k, r,j)) will denote the set (0, [:I,..., [$?I). We say that a set R of
    r-flats of G satisfiesP(n, k, r, j) if
    (9 I R I = Cl
    (ii) / R n [“,]I E C(n, k, r, j) for every (12- j)-flat A of C.
    If G is any of examples (a), (b), or (c) above, and if R = [f] for some
    k-flat K, then R satisfiesP(n, k, r,,j).
    We will say that R satisfiesQ(n, k, r, j) if (i) above holds together with
    (iii) I R n [$]I = [:I or 1R n [:]I d [“;‘I for every (n -,j)-flat A
    of G.
    400 ROTHSCHILD AND SINGHI

    Obviously, P(n, k, r,j) implies Q(n, k, r,j). In any geometric design G


    if R = [f] for somek-flat K, then R satisfiesQ(n, k, r, j).
    If for some geometric design G a set R of v-flats satisfies Q(n, k, r, j)
    (respectively, P(n, k, u,j)) if and only if R = [f] for some k-flat K, then
    we say that Q(n, k, r, j) (respectively, P(n, k, r, j)) characterizes k-flats of G.
    Rothschild and van Lint [5] proved in examples (a) and (b) above that
    P(n, k, r,j) characterizes k-flats (except for q = 2, r = 1, j = n - 2 in
    case(b)) if Y = 1 orj = 1, and n - 2 2 j. In this note we will prove that
    if G is a geometric design of rank n satisfying the following condition

    then Q(n, k, v,j) characterizes k-flats of G provided that y1> r[f] + j - 1.


    We note that (a), (b), and (c) above satisfy (*). But not all goemetric
    designs do. For example, some t-designs (example (d)) give rise to
    geometric designsnot satisfying (“).

    THEOREM 1. Let G be a geometric design of rank 12satisfying (“1. Let


    n 3 r[:] + j - 1. Then Q(n, k, r, j) characterizes k-flats of G.
    First we prove two lemmas (seealso [2]).

    LEMMA 1. Let G be a geometric designof rank n. Let K be a k-Jut of G,


    O<k<n.Letj>O,andM=(SISisa.j-j?atofGwithK<S}.Then
    j A4 / is independentofthe choice of K and is given by 1A4 1 = ([~]/[~])[;J =
    th kj).
    Proof. This formula follows by counting bases as above for the
    formula for [z].

    LEMMA 2. Let G be a geometric designsatisfying (*). Let kl > k. Then

    (i) 3 <-L ifkbl


    GT’(k-1)

    (ii) k1 - (k, - 1) > K - (k - 1)

    (iii) T;, -(k,-l)>k-(k-l)@$=$&


    1 -
    k-FLATS IN GEOMETRIC DESIGNS 401

    Proof. (i) and (iii) are obvious, and (ii) is easily proved using the
    exchange property of geometries. -- (iv) follows ___
    from (i) and the elementary
    fact that --
    since k, > k, and thus, (k, --- 1) > (k - l), etc., we always have
    (ii1 - c)/(k, - 1) - c) < (I? - c)/((k - 1) - c) for all 0 < c < (k - 1).
    Proof of Theorem 1. Let R be a set of r-flats of G satisfying Q(n, k, r,,j).
    Let KR = {x 1 31 E R with x E I}, and let RR be the closure of KR (the
    smallest flat containing it). Let the rank of X, be ol(K,) = k, . We note
    that R C [?I.
    First, if k, < &, we have [F] = j R 1 < I[?]1 = $1 < [:I. Hence
    k = k, and R = [>I.
    So assume k, > k. Let (e, ,..., ek,) C KR be a basis for KR . Let
    M = {H 1His an (n -j)-flat of G such that the rank OL(H n KR) = k, - I}.
    If S is any (k, - 1)-flat of RR , then S is contained in exactly
    t(n, k, - 1, M - j) (n -j)-flats of G by Lemma 1. Exactly t(q kl , n - j)
    of these (n - j)-flats will contain KR . Thus

    I M I = (f(n, k, - 1, ’ -8 - ‘h k, >’ -j>) [klkh 11’ (1)


    For any HE M, let d(H) denote the number of I E R such that I C H, i.e.,
    d(H) = I[:] n R /. Not all the ei can be in H or else H n RR = iz, . Thus
    d(H) < [F]. By Q(n, k, r,j) we have d(H) < [“;‘I foqall HE M. Thus

    H;M4H) d I ~4I [” ; ‘] (2)

    By Lemma 1, any I E R is contained in exactly t(k, , r, k, - 1) (k, - l)-


    flats of K and thus in exactly t(k, , r, k, - l)(t(n, k, - 1, y1- j) -
    t(n, k, , n - j)) elements of M. Thus

    C d(H) = t(k l,r,kl-l)(f(n,Ic,--I,n-i)-f(~,kl,n-i))[~].


    HEM
    (3)
    If we take a basis for each I E R, their union certainly spans RR . Thus,
    k, < Y[:] and by the assumption of the theorem, n - j 3 k, - 1. Hence
    t(n, k, - 1, n -.j) - t(n, k, , n - j) > 0. Thus we can combine (2) and (3)
    and use Lemma 1 to obtain:

    (4)

    If r > 2, this contradicts Lemma 2(iv).


    402 ROTHSCHILD AND SINGHI

    __-Thus we assume I’ = 1. If for any HE h4 we have d(H) < [“;l] =


    (k - I), then the inequality in (2) is strict, and hence the inequality in (4)
    --
    is strict, contradicting Lemma 2, (i). Thus we must have d(H) = (k - 1)
    for all HE M, and equality in (4).
    We now proceed -- as in [5]. Let xH be the characteristic function of the
    set
    -- H. Then (k - 1)” I ~4 I = CHM (4H)Y = CHM (LR XHVY. So
    (k - 1)’ I A4 I = CLZ’ER am4 XHU) xfm = CM CHEAI (XHW +
    c I#Z’ER CHEM x&I) xd’) = (1 R 1 @I > 1, k - 1) + 1 R 1 (1 R 1 - 1) @I >
    2, k, - 1)) * (t(n, k, - 1, n -j) - t(n, k, , n -j)). Now using the formulas
    for / M 1 from (l), for t(n, k,.j) and for [J, we get:
    k, - 1

    (k)’ [k,:
    -(k--I)
    11 = k 1;1 [k,k’ I]
    --
    k(k - ‘)
    [ 1[k,:
    Jfl l]
    I 2I
    or
    __-
    (k - 1)” - -
    = k(k;- I) + It(F - l)(k, - l)((k, - 1) - 1)
    1 El& - 1)

    --
    Using the equality in (4) again, this gives E - (k - 1) = E, - (k, - l),
    contradicting Lemma 2(iii), and completing the proof of Theorem 1.
    We have shown that Q(n, k, r, j) characterizes k-flats provided
    y2 > r [E] + j - 1. This fact was used only once in the proof above in order
    to guarantee that k, - 1 < n - j. k, was the rank of the smallest flat
    containing all f-flats in R. But it is obvious that in order for k, to be
    anywhere near the value r[f], R would necessarily not satisfy Q(n, k, r, j).
    Thus the bounds could be easily improved.
    We will consider briefly the case of example (c), subsets of a set, and in
    particular the r = 2 case. This is the case of graphs. Namely, suppose
    we have a graph G on n vertices and (3 edges. Suppose every induced
    subgraph on YI - j vertices has either (5) edges or at most (“2’) edges. Then
    if y1is large enough, Theorem 1 guarantees that G consists of a complete
    graph on k vertices together with 12- k isolated vertices. If f(k,j) is the
    smallest y1 for which this property holds (i.e., Q(n, k, 2,j) implies that G
    has this structure), then Theorem 1 implies that f(k,,j) < O(k2) for
    fixed j.
    Even a rather crude argument can improve this to O(k). For suppose
    j - 1 of the vertices are isolated. Then using these in the complement of an
    y1- j set we see that any vertex with positive degree has degree at least
    k - 1 by property Q. This forces G to be the desired graph. On the other
    hand, suppose at most j - 1 vertices have degree 0. Then if we take
    II > (k + l)j, and divide the vertices into (k + 1) disjoint j-sets, each
    k-FLATS IN GEOMETRIC DESIGNS 403

    ,j-set must meet at least (k - 1) edges of G, by property Q. This requires


    more than (3 edges. Thus f(k,,j) < O(k). Since k vertices at least are
    required to have Q satisfied at all, we getf(k,j) = O(k).
    Similar arguments apply more generally. It may be true that for some c,
    f(kj) < k + c.

    REFERENCES

    1. H. H. CRAPO AND G.-C. ROTA, “Combinatorial Geometries,” MIT Press, Cam-


    bridge, 1970.
    2. J. EDMONDS, U. S. R. MURTI, AND P. YOUNG, Equicardinal matroids and matroid
    designs, in “Combinatorial Mathematics and Its Applications,” Second Chapel
    Hill Conference, 1970.
    3. D. R. HUGHES, On t-designs and graphs, Amer. J. Math. 87 (1965), 761-778.
    4. J. MACWILLIAMS, Error-correcting codes for multiple-level transmission, BeN System
    Tech. J. 40 (1961), 281-308.
    5. B. ROTHSCHILD AND J. VAN LINT, Characterizing fiinite subspaces, J. Combinatorial
    Theory, Ser. A 16 (1974), 97-110.

    You might also like