Jonathan Bray — 2 April 2026
§1 Proposed Problem
Problem 1.1 (by Ever Ortega). Let An = {a1 , a2 , . . . , an } be a subset of positive real
numbers. Define ( )
X
S(An ) = x : A ⊆ An
x∈A
and let tn be the minimum of all possible cardinalities of S(An ) for An with n distinct
elements. Determine the set An when |S(An )| = tn .
Proof. Without loss generality assume a1 < a2 < · · · < an and let k = a1 . For n = 1, 2, 3
we have A1 = {k}, A2 = {k, a2 } and A3 = {k, a2 , k + a2 }. For n > 3 we will show that
An = {k, 2k, . . . , nk}.
P P
Consider the poset P = (P(An ), ≤) defined by the rule A ≤ B when a∈A a ≤ b∈B b
for all possible values of ai satisfying the total order. We also define the linear extension
P ′ = (P(An ), ≤∗ ) with the extra conditions
{a1 , ai } =∗ {ai }, 2 ≤ i ≤ n
{a2 , a3 } =∗ {a1 , a4 }
Note that these conditions imply that An is of the form {ik} and we denote such sets
A′n . Consider the following chain of P
{∅, {a1 }, {a2 }, . . . , {an }, {a1 + an }, . . . , {a1 + · · · + an }}.
This chain has n(n+1)
2 + 1 elements so |Q| ≥ n(n+1)
2 + 1 for every linear extension Q of P .
′ n(n+1) n(n+1)
Since |P | = 2 + 1 we conclude tn = 2 + 1. So the problem reduces to showing
that every linear extension Q of P with |Q| = tn is equal to P ′ .
Theorem 1.2
n(n+1)
P is a graded poset of length 2 + 1 with greatest element {a1 + · · · + an } and
least element ∅.
To prove this theorem we use the following lemmas. We define IX = {i : ai ∈ X}, for
X ∈ P and order the indices x1 < · · · < xm .
Definition 1.3. We say a function ϕ : IX → IY with X, Y ∈ P is special if it is injective
and x ≤ ϕ(x) for every x ∈ IX .
Lemma 1.4
Let X, Y ∈ P , then X ≤ Y if and only if there is a special function ϕ : IX → IY .
Proof. The if direction is trivial so assume X ≤ Y . Consider the following algorithm.
1
Jonathan Bray — 2 April 2026
Algorithm 1.5 — Set I1 = IY . For 1 ≤ i ≤ n set Ii = Ii−1 − {sup xi−1 } and if
Ii−1
sup xi exists define f (xi ) = sup xi , otherwise terminate.
Ii Ii
If this algorithm iterates through each xi then every element x ∈ IX has an image that
is greater than itself so f is special. Now suppose this algorithm terminates at i = k0 .
If we can show that there is some x ∈ IX such that there are more elements in the
set IX′ of elements in I greater than x than elements in the set I ′ of elements in I
X Y Y
greater than x then X ̸≤ Y . Intuitively this is because we could make the elements
in {ai : i ∈ IX ′ ∪ I ′ } arbitrarily large until the sum of the other elements becomes
Y
insignificant and also arbitrarily close to a N so that they are practically equal to N .
Since |{ai : i ∈ IX′ }| > |{a : i ∈ I ′ }| and the elements in both sets are close to N , we can
i Y
make the sum of the elements in the former greater then P those in thePlatter. Consequently
since the other elements are insignificant we can make a∈X a > a∈Y a so X ̸≤ Y . A
rigorous proof is left to the reader.
y1 y2 y3 y4 y5 y6 y7 y8 y9 y10 y11 y12 y13
x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11
Figure 1: green arrows connect xki with yli and the black lines connect xi with its image.
Notice that the sequence f (xi ) is increasing since otherwise
xi < xi+1 ≤ f (xi+1 ) < f (xi )
so the algorithm should have chosen f (xi+1 ) as the image of xi first, a contradiction.
Consider the sequence
xk0 , yl0 , xk1 , . . . xkn , yln
where yli = sup xki and xki+1 is the preimage of yli . Each of the yli must exist (since
IY
otherwise axki ∈ X can be made arbitrarily larger than any element in Y ) and must have
a preimage since otherwise the algorithm could have picked yli = f (xki ). This shows that
the sequence makes sense. The sequence (yli ) is also decreasing since yli −1 is an upper
bound of xki . So this sequence must end with f (xkn ) = yln as in 1. We show that all the
y > yln are in the image. Suppose y > yln is not in the image, we must have
yli+1 < y < yli
for some i but then
xki+1 ≤ yli+1 < y < yli
but at step ki+1 the algorithm should have chosen y as the image of xki+1 instead of yli a
contradiction. So the set IY′ of elements in IY greater than xkn are precisely the elements
y ≥ yln . On the other hand the set IX′ = {x ∈ X : x ≥ x }. Since elements in I ′ up to
kn X
xk0 −1 have distinct images in IY′ and there are aditional elements ≥ xk0 in IX ′ we have
2
Jonathan Bray — 2 April 2026
′ | > |I ′ |. Therefore X ̸≤ Y contradicting our assumption so the algorithm terminates
|IX Y
and the constructed f is special.
Lemma 1.6
Let X, Y ∈ P , then X ⋖ Y (X covers Y ) if and only if
1. IY is the set containing every element of IX except one xk , containing instead
xk + 1 or
2. IY is the set containing every element of IX in adition to the element 1 ̸∈ IX .
Proof. Suppose X ⋖ Y and |X| < |Y |, by lemma 1.4 we know that there exists a special
function ϕ : IX → IY . Let y ∈ Y be an element without a preimage. Assume y > 1,
there are two cases:
1. Y contains 1, 2, . . . , y
2. Y does not contain an element z, 1 ≤ z < y.
For the first case define Z = Y − {1}. Consider the special function ϕ′ : X → Z defined
as (
ϕ(x) + 1, x < y
ϕ′ (x) =
ϕ(x), x ≥ y.
(this requires y > 1). This shows that X < Z. We also obviously have Z < Y a
contradiction. The second case is easy, we just define Z to contain every element of
Y except y, containing instead z. Clearly X < Z < Y . This necessarily shows that
Y = X ∪ {1}. We now prove this is sufficient. Suppose that there is a Z with X ≤ Z ≤ Y ,
then there are two functions special functions ϕ1 : X → Z and ϕ2 : Z → Y . It is easy to
see that the composition ϕ2 ◦ ϕ1 must equal the identity I(x) = x since ϕ2 (ϕ1 (x)) can’t
equal anything less than x and if it equals anything greater then it couldn’t be injective
since more than one element would map to the greatest lelement ym ∈ Y .
x ≤ ϕ1 (x) ≤ ϕ2 (x)
However this would mean that Z is equal to X or Y showing that X ⋖ Y . This proof is
getting so long I leave it to the reader to complete the |X| = |Y | case.
We are now in a position to prove theorem 1.2.
Proof of theorem 1.2. We will show that the function ρ : P → N defined by
X
ρ(X) = x
x∈IX
is a rank function. If X ⋖ Y then by lemma 1.6 one of the two cases holds. In each case
ρ(Y ) = ρ(X) + 1 so ρ is a rank function. Clearly ∅ is the least element and {a1 + · · · + an }
is the greatest.
3
Jonathan Bray — 2 April 2026
Now remembering P ′ we see that the sum of the elements of X ∈ P ′ is simply kρ(X)
so all the elements with the same rank are equal. Similarly remembering how P ′ was
constructed if Q is a linear extension of P where all the elements with the same rank are
equal then Q = P ′ . So suppose Q is a linear extension of P not equal to P ′ , then there
are two elements X ̸= Y with ρ(X) = ρ(Y ). Consider the maximal chains
X0 , X1 , . . . , Xk , Xk+1 , . . . , X n(n+1)
2
and
X0 , Y1 , . . . , Yk , Yk+1 . . . , X n(n+1)
2
where i indicates the rank of Xi and X = Xk , Y = Yk . If X > Y then consider
X0 , Y1 , . . . , Yk , Xk , Xk+1 , . . . , X n(n+1)
2
otherwise if X < Y consider
X0 , X1 , . . . , Xk , Yk , Yk+1 , . . . , X n(n+1) .
2
In both cases the chain has length n(n+1)
2 + 2 > tn which proves that P ′ is the only linear
extension with P ′ = tn so we are done.