The Plane Built from Half-Open Intervals
The usual Euclidean plane has a countable dense set, and it also satisfies the Lindelof property. The Sorgenfrey plane is a useful contrast: it still has a countable dense set, but it is not Lindelof. The difference is detected by a particular uncountable subset, the anti-diagonal, whose points can each be isolated from all the others by open rectangles.
On \(\mathbb{R}\), let \(\mathbb{R}_{\ell}\) denote the lower-limit topology, whose basis consists of the half-open intervals \([a,b)\), where \(a<b\). The Sorgenfrey plane is \(\mathbb{R}_{\ell}\times\mathbb{R}_{\ell}\) with the product topology. A basis for this product topology consists of rectangles \([a,b)\times[c,d)\), where \(a<b\) and \(c<d\). In particular, a set is open if every point in it lies in one of these basis rectangles that is contained in the set.
A Countable Dense Subset
Consider \(Q=\mathbb{Q}\times\mathbb{Q}\), the set of points in the plane with both coordinates rational. The rationals are countable, and a product of two countable sets is countable: for example, enumerate pairs of natural numbers by increasing sum of their coordinates, and use this to enumerate pairs of rational-number indices. Thus \(Q\) is countable.
Proof. Let \(G\) be a nonempty open set in the Sorgenfrey plane. Choose a point of \(G\). Since the product rectangles form a basis, there are real numbers \(a<b\) and \(c<d\) such that \([a,b)\times[c,d)\subseteq G\). Each interval \([a,b)\) contains a rational: choose \(n\in\mathbb{N}\) with \(n(b-a)>1\), and let \(m\) be the least integer greater than or equal to \(na\). Then \(m\ge na\), and \(m<na+1<nb\), so \(a\le m/n<b\). Applying the same argument to \([c,d)\) gives a rational in that interval as well. The pair of rationals belongs to \([a,b)\times[c,d)\), so \(G\) meets \(\mathbb{Q}\times\mathbb{Q}\). Hence this countable set is dense. \(\square\)
Worked Example: A Rational Point in a Basic Rectangle
Consider the basic open rectangle \([1,3/2)\times[3,7/2)\). The rational point \((7/5,22/7)\) belongs to it: \(1\le 7/5=1.4<1.5=3/2\), and \(3\le 22/7\approx3.143<3.5=7/2\). Both coordinates satisfy the left-inclusive, right-exclusive membership conditions. This one rectangle illustrates how rational-coordinate points can occur in basic open sets; the theorem proves that every nonempty open set contains such a point.
The Anti-Diagonal Is Closed and Discrete
Now define the anti-diagonal \[ D=\{(x,-x):x\in\mathbb{R}\}. \] It is uncountable because different real numbers \(x\) give different points \((x,-x)\). To use it to test the Lindelof property, we need both its closedness and its discreteness in the Sorgenfrey plane. These facts follow directly from the half-open rectangle basis.
Proof. First let \((a,b)\notin D\), so \(s=a+b\ne0\). If \(s>0\), the open rectangle \([a,a+1)\times[b,b+1)\) contains \((a,b)\). Every point \((u,v)\) in this rectangle satisfies \(u\ge a\) and \(v\ge b\), so \(u+v\ge s>0\). No point of the rectangle lies on \(D\), since the coordinates of a point on \(D\) sum to zero.
If \(s<0\), set \(\delta=-s/4>0\). The rectangle \([a,a+\delta)\times[b,b+\delta)\) contains \((a,b)\). For every \((u,v)\) in it, \(u<a+\delta\) and \(v<b+\delta\), so \(u+v<s+2\delta=s/2<0\). Again the rectangle misses \(D\). Thus every point outside \(D\) has an open neighborhood disjoint from \(D\), which proves that the complement of \(D\) is open and \(D\) is closed.
To prove discreteness, fix \(x\in\mathbb{R}\) and consider the open rectangle \([x,x+1)\times[-x,-x+1)\), which contains \((x,-x)\). If a point \((t,-t)\in D\) lies in this rectangle, then its first coordinate gives \(t\ge x\), while its second coordinate gives \(-t\ge -x\), or \(t\le x\). Therefore \(t=x\). The rectangle intersects \(D\) in exactly \(\{(x,-x)\}\). This works for every \(x\), so \(D\) is discrete as a subspace. \(\square\)
Worked Example: Isolating One Anti-Diagonal Point
At the point \((2,-2)\in D\), take the rectangle \([2,12/5)\times[-2,-8/5)\). Its coordinates use the strict upper bounds \(u<12/5\) and \(v<-8/5\). If \((t,-t)\in D\) also lies in this rectangle, then \(t\ge2\) and \(-t\ge-2\), so \(t\le2\). Together these imply \(t=2\). The point \((2,-2)\) is in the rectangle, and the only anti-diagonal point in it is therefore \((2,-2)\) itself.
The same reasoning works for any positive rectangle width: the decisive conditions are the inclusive lower bounds \(t\ge x\) and \(-t\ge -x\), not the particular width chosen.
An Open Cover with No Countable Subcover
For each \(x\in\mathbb{R}\), define the open rectangle \[ U_x=[x,x+1)\times[-x,-x+1). \] The isolation calculation above shows that \(U_x\cap D=\{(x,-x)\}\). The complement of \(D\) is open because \(D\) is closed. These sets together form an open cover: \[ \mathcal{U}=\{\,\mathbb{R}^{2}\setminus D\,\}\cup\{\,U_x:x\in\mathbb{R}\,\}. \] Every point outside \(D\) belongs to \(\mathbb{R}^{2}\setminus D\), and every point of \(D\) belongs to its corresponding \(U_x\).
Proof. Suppose a countable subfamily of \(\mathcal{U}\) is chosen. If it does not contain \(\mathbb{R}^{2}\setminus D\), it covers no point outside \(D\), so it cannot cover the plane. If it does contain \(\mathbb{R}^{2}\setminus D\), that set covers no point of \(D\). Each remaining selected rectangle \(U_x\) covers exactly one point of \(D\). A countable collection of such rectangles therefore covers at most countably many points of \(D\). Since \(D\) is uncountable, some point of \(D\) is not covered. In either case, the chosen countable subfamily is not a cover. Thus \(\mathcal{U}\) has no countable subcover, and the plane is not Lindelof. \(\square\)
Worked Example: Checking the Cover at a Point off the Anti-Diagonal
The point \((0,1)\) is not in \(D\), since \(0+1=1\ne0\). The rectangle \([0,1/4)\times[1,5/4)\) is an open neighborhood of \((0,1)\). For any point \((u,v)\) in it, \(u\ge0\) and \(v\ge1\), so \(u+v\ge1>0\). Hence this rectangle misses \(D\), as required for \(\mathbb{R}^{2}\setminus D\) to be open at \((0,1)\). In the displayed cover, the point is covered by \(\mathbb{R}^{2}\setminus D\), while a point \((x,-x)\) on the anti-diagonal is covered by \(U_x\).
What the Example Shows
The Sorgenfrey plane is separable because a countable set meets every nonempty open set. It is not Lindelof because a particular open cover requires uncountably many of its listed rectangles to cover the anti-diagonal. These conclusions concern different requirements: density asks whether a single countable subset meets every open set, whereas the Lindelof property asks whether every open cover can be reduced to a countable subcover.
Common pitfall. One might argue, “The plane has a countable dense subset, so every open cover has a countable subcover.” That implication is false: the proof of density only shows that each nonempty open set contains at least one point of \(\mathbb{Q}\times\mathbb{Q}\). It does not select a countable collection of members from an arbitrary open cover that covers the entire space. In the cover \(\mathcal{U}\), the single set \(\mathbb{R}^{2}\setminus D\) misses every point of \(D\), and each \(U_x\) covers just one of those uncountably many points. The countable dense subset does not change this count of the cover’s members.
Check Your Understanding
Use the product basis and the open-cover construction to answer the following.
- Why does a basic rectangle \([a,b)\times[c,d)\) have to have \(a<b\) and \(c<d\)?
- Show directly that every nonempty basic rectangle in the Sorgenfrey plane meets \(\mathbb{Q}\times\mathbb{Q}\).
- For a fixed \(x\), determine exactly which points of \(D\) lie in \([x,x+2)\times[-x,-x+2)\).
- Explain why \(\mathbb{R}^{2}\setminus D\) alone cannot cover any point of \(D\).
- Why can no countable selection from \(\mathcal{U}\) cover \(D\)?