Consider the following diagram

diagram_representing_single_ordering

to represent \(b\leq a\). For instance, the diagram

diagram_representing_two_orderings

represents a situation where \(b\leq a\) and \(c\leq a\), but there is no particular relation between \(b\) and \(c\). Such a diagram is called a Hasse diagram.

Maximal and Minimal Elements

Definition 1 An element \(a\) of an ordered set \(A\) is a minimal element (resp. maximal element) of \(A\) if for all \(x\in A\), \(a\leq x\) (resp. \(a\geq x\)) implies \(x=a\).

A minimal element need not be unique. For example, in

diagram_representing_two_orderings

both \(b\) and \(c\) are minimal elements of the set \(\{a,b,c\}\). Mathematicians generally prefer such elements to be unique, so this situation is not particularly desirable.

Least and Greatest Elements

Definition 2 An element \(a\) of an ordered set \(A\) is a least element (resp. greatest element) of \(A\) if \(a\leq x\) (resp. \(x\leq a\)) for all \(x\in A\).

In the previous example, \(b\) and \(c\) cannot be least elements, since neither \(b\leq c\) nor \(c\leq b\) holds. By definition, a least element is unique. Moreover, the following holds.

Proposition 3 If \(A\) has a least element \(a\), then \(a\) is the unique minimal element of \(A\).

Proof

For any element \(x\) of \(A\), \(a\leq x\) holds. Thus if there exists \(x\in A\) such that \(x\leq a\), then by antisymmetry of \(\leq\), we must have \(x=a\). From this we see that \(a\) is a minimal element of \(A\).

If \(a'\) is another minimal element of \(A\) and \(a'\neq a\), then by the contrapositive of Definition 1, we must have \(a'\not\leq a\), which contradicts the fact that \(a\) is a least element. Therefore \(a'=a\).

Sometimes we need to consider a new element greater than all elements of an ordered set, or an element smaller than all elements. Such imaginary elements are conventionally denoted by \(\pm\infty\).

Proposition 4 Let \(A\) be an ordered set and \(A'=A\sqcup\{+\infty\}\). Then there exists an order relation on \(A'\) that extends the order relation defined on \(A\) and has \(a\) as its greatest element.

Proof

It suffices to add the elements of \(\bigcup_{x\in A}\left\{(x, +\infty)\right\}\) to the existing order relation.

Upper and Lower Bounds

Definition 5 Let \(A\) be a preordered set and \(X\) a subset of \(A\). If \(a\in A\) satisfies \(a\leq x\) (resp. \(a\geq x\)) for all \(x\in X\), then \(a\) is called a lower bound (resp. upper bound) of \(X\) in \(A\).

A set that has a lower bound (resp. upper bound) is called bounded below (resp. bounded above), and a set that is both bounded below and bounded above is simply called bounded.

Consider the following ordered set \(A=\{a,b,c,d,e\}\).

upper_and_lower_bounds

Then \(a\) is an upper bound of the set \(X=\left\{c,d,e\right\}\) but \(b\) is not. If we consider the set \(X'=\left\{d,e\right\}\), then both \(a\) and \(b\) are upper bounds of this set. From the above example, we see that a lower bound of a set \(X\) need not belong to \(X\), but if it does, then that element is the least element of \(X\).

Definition 6 Let \(A\) be an ordered set and \(X\subseteq A\). An element \(a\) of \(A\) is a greatest lower bound (or infimum) of \(X\) if it is the greatest element among the lower bounds of \(X\). Similarly, the least upper bound (or supremum) is defined.

If the supremum of \(X\subseteq A\) exists, we denote it by \(\sup_AX\), and the infimum by \(\inf_AX\). By definition, it is easy to verify that if \(X\subseteq A\) has a greatest element \(a\), then \(a=\sup_AX\).

Proposition 7 Let \(A\) be an ordered set and \(X\subset A\) have both a supremum and an infimum.

  1. If \(X\neq\emptyset\), then \(\inf_A X\leq\sup_A X\).
  2. If \(X=\emptyset\), then \(\sup_AX\) and \(\inf_AX\) become the least and greatest elements of \(A\), respectively.
Proof
  1. If \(X\neq\emptyset\), then there exists at least one element. Call it \(a\). By definition, \(\inf X\leq a\) and \(a\leq\sup X\), so by transitivity, \(\inf_AX \leq\sup_AX\).
  2. If \(X=\emptyset\), then for any element \(a\) of \(A\), both \(a\leq x\) and \(x\leq a\) hold for all \(x\in X\). Thus every element of \(A\) is both a lower bound and an upper bound of \(X\), and \(\sup_AX\) and \(\inf_AX\) become the least and greatest elements of \(A\), respectively.

Operations on Sets and Bounds

Now we examine the relationship between set operations and bounds.

Proposition 8 For two subsets \(X,X'\) of an ordered set \(A\), if \(\sup_AX,\sup_AX'\) are both defined and \(X'\subseteq X\), then \(\sup X'\leq\sup X\).

Proof

Let \(x\in X'\) be arbitrary. Since \(X'\subseteq X\), \(x\in X\). On the other hand, \(x\leq \sup X\) holds for any \(x\in X\), and thus \(\sup X\) is an upper bound of \(X'\). By definition, \(\sup X'\leq \sup X\).

Proposition 9 For an ordered set \(A\), consider families \((x_i)_{i\in I}\), \((y_i)_{i\in I}\) satisfying \(x_i\leq y_i\) for all \(i\in I\). If they all have suprema in \(A\), then \(\sup_{i\in I} x_i\leq \sup_{i\in I} y_i\).

Proof

For any \(i\in I\), \(x_i\leq y_i\) and \(y_i\leq \sup y_i\), so \(x_i\leq \sup y_i\) for all \(i\). By the minimality of \(\sup x_i\), we have \(\sup x_i\leq\sup y_i\).

Proposition 10 For an ordered set \(A\), an index set \(I\), and a covering \((J_k)_{i\in I}\) of \(I\), suppose \((x_i)_{i\in J_k}\) has a supremum in \(A\). Then \(\sup_{i\in I} x_i\) exists if and only if \(\sup_{k\in K}(\sup_{j\in J_k}x_j)\) exists, and the two values are equal.

Proof

Write \(b_k=\sup_{i\in J_k} x_i\). First suppose \((x_i)_{i\in I}\) has a supremum, call it \(a\). Then \(a\leq b_k\) holds for all \(k\). Also, if \(c\geq b_k\) holds for all \(k\), then for any \(x_i\), if \(i\in J_{k'}\), then \(b_{k'}\geq x_i\), and thus \(c\geq x_i\) for all \(i\). By minimality of \(a\), we must have \(c\geq a\), so \(a\) is the supremum and \(\sup_{i\in I}x_i=\sup_{k\in K}(\sup_{i\in J_k} x_j)\).

Conversely, if \((b_k)_{k\in K}\) has a supremum \(a'\), the proof can be completed in the same way.

Proposition 11 For the product \(A=\prod A_i\) of ordered sets \((A_i)_{i\in I}\) and a subset \(X\) of \(A\), let \(X_i=\pr_i X\). Then \(\sup_AX\) exists if and only if \(\sup_{A_i}X_i\) exists for each \(i\), and \(\sup_AX=(\sup_{A_i}X_i)\).

Proof

First suppose \(\sup_{A_i} X_i\) exists for each \(i\). It is clear that \((\sup_{A_i} X_i)_{i\in I}\) is an upper bound of \(X\). If \((c_i)\) were another upper bound of \(X\), then each \(c_i\) would be an upper bound of \(X_i\), so by minimality of \(\sup_{A_i}X_i\), we have \(c_i\geq\sup X_i\), and thus \((c_i)\geq(\sup X_i)_{i\in I}\).

Conversely, suppose \(\sup X=(a_i)\) exists. For all \(i\), \(a_i\) is an upper bound of \(X_i\). For if \(x_i\in X_i\), then there exists \(x\in X\) with \(x_i\) as its \(i\)th component such that \(x\leq (a_i)\). Now for any other upper bound \(a_i'\), define a new element \((c_i)\) by replacing the \(i\)th component of \((a_i)\) with \(a_i'\). Then \(c\geq a\), so \(a_i'\geq a_i\).

Remark For an ordered set \(A\) and \(X'\subseteq X\subseteq A\), only one of \(\sup_AX'\) and \(\sup_XX'\) may exist, or both may exist but with different values. For example, compare \(X'=\{x\in\mathbb{Q}\mid x < \sqrt{2}\}\) in each of the following sets:

  1. As a subset of \(X_1=\mathbb{Q}\), the supremum of this set does not exist, but it does exist in \(A=\mathbb{R}\). Thus even if \(\sup_AX'\) exists, \(\sup_{X_1}X'\) may not exist.
  2. Consider the set \(X_2=X'\cup \left\{2\right\}\). Then \(X'\subseteq X_2\subseteq X_1\) and \(\sup_{X_2}X'=2\), but \(\sup_{X_1}A\) does not exist.
  3. Finally, for \(X'\subseteq X_2\subseteq A\), both \(\sup_{X_2}X'\) and \(\sup_AX'\) exist but have different values.

Nevertheless, we can still prove the following.

Proposition 12 Let \(A\) be an ordered set and \(X'\subseteq X\subseteq A\). If both \(\sup_AX'\) and \(\sup_XX'\) exist, then \(\sup_AX'\leq\sup_XX'\). If \(\sup_AX'\) exists and belongs to \(X\), then \(\sup_XX'\) also exists and equals \(\sup_AX'\).

Proof

The set of upper bounds of \(X'\) in \(X\) is contained in the set of upper bounds in \(A\), and thus the supremum is larger. If \(\sup_AX'\) exists and belongs to \(X\), then it is clearly the supremum of \(X'\) in \(X\).


References

[Bou] N. Bourbaki, Theory of Sets. Elements of mathematics. Springer Berlin-Heidelberg, 2013.


댓글남기기