{"id":2869,"date":"2020-10-19T15:33:24","date_gmt":"2020-10-19T13:33:24","guid":{"rendered":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/?page_id=2869"},"modified":"2022-11-19T15:01:12","modified_gmt":"2022-11-19T14:01:12","slug":"quasi-uniform-spaces-i-pervin-quasi-uniformities-pervin-spaces","status":"publish","type":"page","link":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/?page_id=2869","title":{"rendered":"Quasi-Uniform Spaces I: Pervin quasi-uniformities, Pervin spaces"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">A <em>uniform space<\/em> is a natural generalization of the notion of a metric space, on which completeness still makes sense [1].  The non-Hausdorff variant of this is called a <em>quasi-uniform space<\/em>, and the purpose of this post is to introduce some of the features of those spaces, with a particular stress on a now very classical construction due to William Pervin [3].  It is rather puzzling that I managed to avoid the subject of quasi-uniform spaces in something like the 7 years that this blog existed!<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The definition of quasi-uniform spaces<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Let <em>X<\/em> be a set.  A binary relation <em>R<\/em> on <em>X<\/em> is a subset of <em>X<\/em> \u00d7 <em>X<\/em>.  Binary relations compose: <em>R<\/em> o <em>S<\/em> is the set of those pairs (<em>x<\/em>,<em>z<\/em>) such that (<em>x<\/em>,<em>y<\/em>) \u2208 <em>R<\/em> and (<em>y<\/em>,<em>z<\/em>) \u2208 <em>S<\/em> for some <em>y<\/em> in <em>X<\/em>.  Note that <em>R<\/em> o <em>R<\/em> \u2286 <em>R<\/em> if and only if <em>R<\/em> is transitive.  A relation <em>R<\/em> is reflexive if and only if it contains the diagonal \u0394 \u225d {(<em>x<\/em>,<em>x<\/em>) | <em>x<\/em> \u2208 <em>X<\/em>}.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">A <em>quasi-uniformity<\/em> <strong><em>U<\/em><\/strong> on <em>X<\/em> is a filter of reflexive binary relations on <em>X<\/em> that satisfies the following relaxed transitivity condition: for every relation <em>S<\/em> in <strong><em>U<\/em><\/strong>, there is a relation <em>R<\/em> in <strong><em>U<\/em><\/strong> such that <em>R<\/em> o <em>R<\/em> \u2286 <em>S<\/em>.  Explicitly, it is a family of binary relations on <em>X<\/em> such that:<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>(reflexivity) for every <em>R<\/em> in <strong><em>U<\/em><\/strong>, for every <em>x<\/em> in <em>X<\/em>, (<em>x<\/em>,<em>x<\/em>) is in <em>R<\/em>;<\/li>\n\n\n\n<li>(filter 1) <em>X<\/em> \u00d7 <em>X<\/em> is in <strong><em>U<\/em><\/strong>;<\/li>\n\n\n\n<li>(filter 2) for every <em>R<\/em> in <strong><em>U<\/em><\/strong>, every binary relation <em>S<\/em> on <em>X<\/em> such that <em>R<\/em> \u2286 <em>S<\/em> is in <strong><em>U<\/em><\/strong>;<\/li>\n\n\n\n<li>(filter 3) for all <em>R<\/em>, <em>S<\/em> in <strong><em>U<\/em><\/strong>, <em>R<\/em> \u2229 <em>S<\/em> is in <strong><em>U<\/em><\/strong>;<\/li>\n\n\n\n<li>(relaxed transitivity) for every <em>S<\/em> in <strong><em>U<\/em><\/strong>, there is a relation <em>R<\/em> in <strong><em>U<\/em><\/strong> such that, for all points <em>x<\/em>, <em>y<\/em>, <em>z<\/em> in <em>X<\/em> such that (<em>x<\/em>,<em>y<\/em>) \u2208 <em>R<\/em> and (<em>y<\/em>,<em>z<\/em>) \u2208 <em>R<\/em>, we have (<em>x<\/em>,<em>z<\/em>) \u2208 <em>S<\/em>.<\/li>\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\">A <em>quasi-uniform space<\/em> is a set with a quasi-uniformity <strong><em>U<\/em><\/strong>.  The relations <em>R<\/em> in <strong><em>U<\/em><\/strong> are called <em>entourages<\/em>, a French word that means surroundings.  (And also environment, neighborhood, relatives, acquaintances.  As most French words, the word also exists in English.)<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The primary example is given by hemi-metric spaces.  Given any hemi-metric <em>d<\/em> on <em>X<\/em>, one defines a quasi-uniformity <em><strong>U<\/strong><sub>d<\/sub><\/em> as follows.  A <em>basic entourage<\/em> is a relation of the form [&lt;<em>r<\/em>] \u225d {(<em>x<\/em>,<em>y<\/em>) | <em>d<\/em>(<em>x<\/em>,<em>y<\/em>) &lt; <em>r<\/em>}, where <em>r<\/em> varies over the positive real numbers (namely, <em>r<\/em>&gt;0).  Then <em><strong>U<\/strong><sub>d<\/sub><\/em> is the family of relations that contains some basic entourage [&lt;<em>r<\/em>].<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">It is interesting to see why relaxed transitivity holds.  Let <em>S<\/em> be in <em><strong>U<\/strong><sub>d<\/sub><\/em>, so <em>S<\/em> contains some basic entourage [&lt;<em>s<\/em>], with <em>s<\/em>&gt;0.  Define <em>r<\/em>\u225d<em>s<\/em>\/2 and <em>R<\/em> as [&lt;<em>r<\/em>].  For all points <em>x<\/em>, <em>y<\/em>, <em>z<\/em> in <em>X<\/em> such that (<em>x<\/em>,<em>y<\/em>) \u2208 <em>R<\/em> and (<em>y<\/em>,<em>z<\/em>) \u2208 <em>R<\/em>, we have <em>d<\/em>(<em>x<\/em>,<em>y<\/em>)&lt;<em>r<\/em> and <em>d<\/em>(<em>y<\/em>,<em>z<\/em>)&lt;<em>r<\/em>, so <em>d<\/em>(<em>x<\/em>,<em>z<\/em>)\u2264<em>d<\/em>(<em>x<\/em>,<em>y<\/em>)+<em>d<\/em>(<em>y<\/em>,<em>z<\/em>)&lt;<em>s<\/em>, showing that (<em>x<\/em>,<em>z<\/em>) is in <em>S<\/em>.  In other words, relaxed transitivity stems from the triangular inequality.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">By the way, the way we have built <em><strong>U<\/strong><sub>d<\/sub><\/em> is a practical way of defining a quasi-uniformity.  A <em>base (of entourages<\/em>) is any non-empty family <strong><em>B<\/em><\/strong> of binary relations on <em>X<\/em> satisfying:<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>(reflexivity) for every <em>R<\/em> in <strong><em>B<\/em><\/strong>, for every <em>x<\/em> in <em>X<\/em>, (<em>x<\/em>,<em>x<\/em>) is in <em>R<\/em>;<\/li>\n\n\n\n<li>(filter 3) for all <em>R<\/em>, <em>S<\/em> in <strong><em>B<\/em><\/strong>, there is a relation <em>T<\/em> in <strong><em>B<\/em><\/strong> such that <em>T<\/em> \u2286 <em>R<\/em> \u2229 <em>S<\/em>;<\/li>\n\n\n\n<li>(relaxed transitivity) for every <em>S<\/em> in <strong><em>B<\/em><\/strong>, there is a relation <em>R<\/em> in <strong><em>B<\/em><\/strong> such that <em>R<\/em> o <em>R<\/em> \u2286 <em>S<\/em>.<\/li>\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\">Then the set of binary relations that contain some element of <strong><em>B<\/em><\/strong> forms a quasi-uniformity.  This is the quasi-uniformity <em>generated by<\/em> <strong><em>B<\/em><\/strong>.  This is what we just did, with <em><strong>B<\/strong><\/em> consisting of the relations [&lt;<em>r<\/em>], <em>r<\/em>&gt;0.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Before we go on, I should mention the original notion of uniformity.  Given any binary relation <em>R<\/em> on <em>X<\/em>, let <em>R<\/em><sup>op<\/sup> be its converse: (<em>x<\/em>,<em>y<\/em>) is in <em>R<\/em><sup>op<\/sup>  if and only if (<em>y<\/em>,<em>x<\/em>) is in <em>R<\/em>.  A quasi-uniformity <strong><em>U<\/em><\/strong> is a <em>uniformity<\/em> if and only if it satisfies the following extra condition:<\/p>\n\n\n\n<ol class=\"wp-block-list\" start=\"6\">\n<li>(symmetry) for every <em>R<\/em> in <strong><em>U<\/em><\/strong>, <em>R<\/em><sup>op<\/sup> is in <strong><em>U<\/em><\/strong>.<\/li>\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\">The prime example of a uniformity is <em><strong>U<\/strong><sub>d<\/sub><\/em> provided that <em>d<\/em> is a <em>pseudo<\/em>-metric now.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Quasi-uniform spaces and topological spaces<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">There is a traditional way to extract a topology from a quasi-uniformity <em><strong>U<\/strong><\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Given any point <em>x<\/em> in <em>X<\/em>, and any binary relation <em>R<\/em> on <em>X<\/em>, let <em>R<\/em>[<em>x<\/em>] denote the set {<em>y<\/em> \u2208 <em>X<\/em> | (<em>x<\/em>,<em>y<\/em>) \u2208 <em>R<\/em>}; namely, the <em>image<\/em> of <em>x<\/em> by <em>R<\/em>.  We say that a <em><strong>U<\/strong>-neighborhood <\/em>of <em>x<\/em> is a set of the form <em>R<\/em>[<em>x<\/em>] for some entourage <em>R<\/em> \u2208 <strong><em>U<\/em><\/strong>.  Next, we call <em><strong>U<\/strong>-open<\/em> any subset of <em>X<\/em> that is a <em><strong>U<\/strong><\/em>-neighborhood<em> <\/em>of each of its points.  This forms a topology, which I will call the <em>induced topology<\/em>, or the <em>topology induced by<\/em> <strong><em>U<\/em><\/strong>.  One also says that the quasi-uniformity <strong><em>U<\/em><\/strong> is <em>compatible<\/em> with a topology if and only if that topology is the induced topology.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For example, if <em>d<\/em> is a hemi-metric, a <em><strong>U<\/strong><sub>d<\/sub><\/em>-neighborhood of <em>x<\/em> is simply a subset of <em>X<\/em> that contains the open ball <em>B<sup>d<\/sup><sub>x<\/sub><\/em><sub>,&lt;<em>r<\/em><\/sub> for some <em>r<\/em>&gt;0.  The topology induced by  <em><strong>U<\/strong><sub>d<\/sub><\/em> is therefore the open ball topology of <em>d<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">By the way, for every <em>x<\/em> \u2208 <em>X<\/em>, the <em><strong>U<\/strong>&#8211;<\/em>neighborhoods <em>R<\/em>[<em>x<\/em>] (<em>R<\/em> \u2208 <strong><em>U<\/em><\/strong>) are neighborhoods of <em>x<\/em> in the topology induced by <strong><em>U<\/em><\/strong>.  Indeed, consider the set <em>U<\/em> of all the points <em>y<\/em> \u2208 X such that there is an <em>S<\/em> \u2208 <strong><em>U<\/em><\/strong> such that <em>S<\/em>[<em>y<\/em>] is included in <em>R<\/em>[<em>x<\/em>].  This is open in the induced topology underlying <strong><em>U<\/em><\/strong>, by definition.  In fact, it is easy to check that <em>U<\/em> is the interior of <em>R<\/em>[<em>x<\/em>] in that topology.  Additionally, <em>x<\/em> is in <em>U<\/em>, because one can take <em>R<\/em> for <em>S<\/em> in that case.  Hence <em>R<\/em>[<em>x<\/em>] indeed contains an open set <em>U<\/em> that contains <em>x<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">A natural question arises: which topologies arise as topologies induced by quasi-uniformity?  The question was solved by Cs\u00e1sz\u00e1r [2]: all topologies whatsoever!  The simplest known argument is due to Pervin [3], and is as follows.  There are in general several quasi-uniformities that induce the same topology, and we will see an example below; another way of stating the following theorem is to say that there is at least one.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Theorem (Cs\u00e1sz\u00e1r [2], Pervin [3]).<\/strong>  Let <em>X<\/em> be a topological space.  For every open subset <em>U<\/em> of <em>X<\/em>, let <em>R<sub>U<\/sub><\/em> be the binary relation {(<em>x<\/em>,<em>y<\/em>) | <em>x<\/em> \u2208 <em>U<\/em> \u21d2 <em>y<\/em> \u2208 <em>U<\/em>}.  The finite intersections of relations <em>R<sub>U<\/sub><\/em>, <em>U<\/em> \u2208 <strong>O<\/strong><em>X<\/em>, form a base <strong><em>B<\/em><\/strong> of entourages of a quasi-uniformity on <em>X<\/em>, whose induced topology is exactly the topology of <em>X<\/em>.  This quasi-uniformity is the <em>Pervin quasi-uniformity<\/em> of the topology of <em>X<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Proof.<\/em> We check the axioms for a base of entourages.  First, (reflexivity) and (filter 3) are clear.  In order to verify (relaxed transitivity), we verify the much stronger statement that <em>R<sub>U<\/sub><\/em> o <em>R<sub>U<\/sub><\/em> \u2286 <em>R<sub>U<\/sub><\/em> for every open subset <em>U<\/em> of <em>X<\/em>.  That inclusion just means that if <em>x<\/em> \u2208 <em>U<\/em> \u21d2 <em>y<\/em> \u2208 <em>U<\/em> and if <em>y<\/em> \u2208 <em>U<\/em> \u21d2 <em>z<\/em> \u2208 <em>U<\/em>, then <em>x<\/em> \u2208 <em>U<\/em> \u21d2 <em>z<\/em> \u2208 <em>U<\/em>.  The relation <em>R<sub>U<\/sub><\/em> o <em>R<sub>U<\/sub><\/em> \u2286 <em>R<sub>U<\/sub><\/em> will play an important r\u00f4le later in this post.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">We now consider an arbitrary element <em>S<\/em> \u225d <em>R<sub>U<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>U<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub> in <strong><em>B<\/em><\/strong>.  We look for an element <em>R<\/em> of <strong><em>B<\/em><\/strong> such that <em>R<\/em> o <em>R<\/em> \u2286 <em>S<\/em>, and we simply define <em>R<\/em> as <em>S<\/em>.   Every pair (<em>x<\/em>,<em>z<\/em>) in <em>R<\/em> o <em>R<\/em> is such that there is a point <em>y<\/em> such that (<em>x<\/em>,<em>y<\/em>) and (<em>y<\/em>,<em>z<\/em>) are in <em>R<\/em>.  Hence (<em>x<\/em>,<em>y<\/em>) and (<em>y<\/em>,<em>z<\/em>) are in <em>R<sub>U<\/sub><\/em><sub><sub><em>i<\/em><\/sub><\/sub> for every <em>i<\/em>, 1\u2264<em>i<\/em>\u2264<em>n<\/em>, which shows that (<em>x<\/em>,<em>z<\/em>) is in <em>R<sub>U<\/sub><\/em><sub><sub><em>i<\/em><\/sub><\/sub> o <em>R<sub>U<\/sub><\/em><sub><sub><em>i<\/em><\/sub><\/sub> \u2286 <em>R<sub>U<\/sub><\/em><sub><sub><em>i<\/em><\/sub><\/sub> for every <em>i<\/em>, 1\u2264<em>i<\/em>\u2264<em>n<\/em>, hence that (<em>x<\/em>,<em>z<\/em>) is in <em>R<\/em>=<em>S<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Let <strong><em>U<\/em><\/strong> be the quasi-uniformity generated by <strong><em>B<\/em><\/strong>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Given any point <em>x<\/em> of <em>X<\/em>, and any basic entourage <em><em>S<\/em><\/em> \u225d <em><em>R<sub>U<\/sub><\/em><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em><em>R<sub>U<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub><\/em>, the image <em>S<\/em>[<em>x<\/em>] is equal to the intersection of the opens sets <em><em>U<sub>i<\/sub><\/em><\/em> that contain <em>x<\/em>.  It follows that every <strong><em>U<\/em><\/strong>-open set is a neighborhood of each of its points, hence is open in the original topology on <em>X<\/em>.  Conversely, let <em>U<\/em> be any open set in the original topology on <em>X<\/em>.  For every <em>x<\/em> in <em>U<\/em>, <em><em>R<sub>U<\/sub><\/em><\/em>[<em>x<\/em>]=<em>U<\/em> is a <strong><em>U<\/em><\/strong>-neighborhood of <em>x<\/em> included in <em>U<\/em>, so <em>U<\/em> is <strong><em>U<\/em><\/strong>-open.  \u2610<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">You may have noticed the similarity with Wilson&#8217;s theorem (Theorem 6.3.13 in the <a href=\"https:\/\/www.cambridge.org\/gb\/knowledge\/isbn\/item7069109\/Non-Hausdorff%20Topology%20and%20Domain%20Theory\/?site_locale=en_GB\">book<\/a>), which says that every countably-based topological space is hemi-metrizable.  You may have noticed that the proof strategy is very similar, too!<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Indeed, each open subset <em>U<\/em> of <em>X<\/em> gives rise to a binary relation <em>R<sub>U<\/sub><\/em>, but that relation has many properties.  First and foremost, it is a preordering.  Indeed, <em>R<sub>U<\/sub><\/em> o <em>R<sub>U<\/sub><\/em> \u2286 <em>R<sub>U<\/sub><\/em>, as we have shown in the proof above, and that means that <em>R<sub>U<\/sub><\/em> is transitive; and every entourage is reflexive by definition.  Being a preordering, it induces a hemi-metric <em>d<sub>U<\/sub><\/em>, defined by <em>d<sub>U<\/sub><\/em>(<em>x<\/em>,<em>y<\/em>) \u225d 0 if (<em>x<\/em>,<em>y<\/em>) \u2208 <em>R<sub>U<\/sub><\/em>, and 1 (or \u221e) otherwise.  The open ball topology of that hemi-metric contains only <em>U<\/em>, the empty set, and the whole space.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Now remember that the topology defined by a countable family of hemi-metrics is hemi-metrizable (Lemma 6.3.11).  What we have just said shows the following.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Proposition.<\/strong>  Every topology is defined by a family of hemi-metrics.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><em>Proof. <\/em>Just take the hemi-metrics <em>d<sub>U<\/sub><\/em>, where <em>U<\/em> ranges over the open subsets.  \u2610<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">As an aside, the topologies defined by a family of pseudo-metrics (symmetric hemi-metrics) are exactly the completely regular topologies, and are also exactly the topologies induces by uniformities.  Let me allow not to say why this month, and to postpone it to a later post.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Morphisms<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">At this point, one may be tempted to think that there is no point in inventing a new notion (quasi-uniform spaces), since quasi-uniform spaces and topological spaces seem to be one and the same thing.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The difference is in <em>morphisms<\/em>.  While the morphisms in the category <strong>Top<\/strong> of topological spaces are the continuous maps, the morphisms from <em>X<\/em> to <em>Y<\/em> in the category <strong>QUnif<\/strong> of quasi-uniform spaces are the <em>uniformly continuous maps<\/em>, namely the maps <em>f<\/em> : <em>X<\/em> \u2192 <em>Y<\/em> such that (<em>f<\/em>\u00d7<em>f<\/em>)<sup>\u20131<\/sup> (<em>S<\/em>) is an entourage of <em>X<\/em> for every entourage <em>S<\/em> of <em>Y<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Every uniformly continuous map <em>f<\/em> is continuous for the underlying induced topologies.  Indeed, let <em>V<\/em> be any open subset of <em>Y<\/em>.  For every <em>x<\/em> \u2208 <em>f<\/em><sup>\u20131<\/sup>(<em>V<\/em>), we need to show that there is a neighborhood of <em>X<\/em> whose image by <em>f<\/em> is included in <em>V<\/em>.  We use the fact that <em>f<\/em>(<em>x<\/em>) is in <em>V<\/em>, and the definition of the induced topology on <em>Y<\/em>, to obtain an entourage <em>S<\/em> in <em>Y<\/em> such that <em>S<\/em>[<em>f<\/em>(<em>x<\/em>)] \u2286 <em>V<\/em>.  Let <em>R<\/em> be the entourage (<em>f<\/em>\u00d7<em>f<\/em>)<sup>\u20131<\/sup> (<em>S<\/em>).  Then <em>R<\/em>[<em>x<\/em>] is a neighborhood of <em>x<\/em>, by definition of the induced topology on <em>X<\/em>, and every element of its image by <em>f<\/em> is of the form <em>f<\/em>(<em>y<\/em>) with (<em>x<\/em>,<em>y<\/em>) \u2208 <em>R<\/em>; so (<em>f<\/em>(<em>x<\/em>),<em>f<\/em>(<em>y<\/em>)) is in <em>S<\/em>, by definition of <em>R<\/em>, showing that <em>f<\/em>(<em>y<\/em>) is in <em>S<\/em>[<em>f<\/em>(<em>x<\/em>)], hence in <em>V<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For quasi-metric spaces <em>X<\/em> and <em>Y<\/em>, with respective quasi-metrics <em>d<sub>X<\/sub><\/em> and <em>d<sub>Y<\/sub><\/em>, a uniformly continuous map <em>f<\/em> : <em>X<\/em> \u2192 <em>Y<\/em> between the corresponding quasi-uniform spaces is a map such that for every <em>s<\/em>&gt;0, (<em>f<\/em>\u00d7<em>f<\/em>)<sup>\u20131<\/sup> ([&lt;<em>s<\/em>]) contains a basic entourage [&lt;<em>r<\/em>] for some <em>r<\/em>&gt;0.  In other words: for every <em>s<\/em>&gt;0, there is an <em>r<\/em>&gt;0 such that for all points <em>x<\/em>, <em>y<\/em> such that <em>d<sub>X<\/sub><\/em>(<em>x<\/em>,<em>y<\/em>)&lt;<em>r<\/em>, <em>d<sub>Y<\/sub><\/em>(<em>f<\/em>(<em>x<\/em>),<em>f<\/em>(<em>y<\/em>))&lt;<em>s<\/em>.  This is the usual definition of uniform continuity in (quasi-)metric spaces.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">It is well-known that there are continuous maps between metric spaces that are not uniformly continuous, for example the map <em>x<\/em> \u21a6 1\/<em>x<\/em> on the positive real numbers.  Hence that also holds in the realm of (quasi-)uniform spaces.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Rather remarkably, uniform continuity and continuity coincide on Pervin quasi-uniformities arising from topologies.  Indeed, let <em>X<\/em> and <em>Y<\/em> be two topological spaces, and consider them with their Pervin quasi-uniformities.  Let <em>f<\/em> : <em>X<\/em> \u2192 <em>Y<\/em> be a continuous map.  For every basic entourage <em>S<\/em> \u225d <em>R<sub>V<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>V<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub> in <em>Y<\/em>, (<em>f<\/em>\u00d7<em>f<\/em>)<sup>\u20131<\/sup> (<em>S<\/em>) is the set of pairs (<em>x<\/em>,<em>y<\/em>) of points of <em>X<\/em> such that <em>f<\/em>(<em>x<\/em>) \u2208 <em>V<\/em><sub>1<\/sub> implies <em>f<\/em>(<em>y<\/em>) \u2208 <em>V<\/em><sub>1<\/sub> and &#8230; and <em>f<\/em>(<em>x<\/em>) \u2208 <em>V<\/em><sub><em>n<\/em><\/sub> implies <em>f<\/em>(<em>y<\/em>) \u2208 <em>V<\/em><sub><em>n<\/em><\/sub>, in other words (<em>f<\/em>\u00d7<em>f<\/em>)<sup>\u20131<\/sup> (<em>S<\/em>) is the basic entourage <em>R<sub>U<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>U<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub> where <em>U<\/em><sub>1<\/sub>\u225d<em>f<\/em><sup>\u20131<\/sup>(<em>V<\/em><sub>1<\/sub>), &#8230;, <em>U<\/em><sub><em>n<\/em><\/sub>\u225d<em>f<\/em><sup>\u20131<\/sup>(<em>V<\/em><sub><em>n<\/em><\/sub>).<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">What all that means, categorically, is:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>There is a (forgetful) functor <em>U<\/em> : <strong>QUnif<\/strong> \u2192 <strong>Top<\/strong> that maps every quasi-uniform space to the underlying topological space (with the induced topology), and every uniformly continuous map <em>f<\/em> to itself, seen as a continuous map.<\/li>\n\n\n\n<li>There is a functor <em>Perv<\/em> : <strong>Top<\/strong> \u2192 <strong>QUnif<\/strong> that maps every topological space to itself, seen as a quasi-uniform space with the Pervin quasi-uniformity of its topology.  It maps every continuous map <em>f<\/em> to itself, since <em>f<\/em> is uniformly continuous as well, as we have just seen.<\/li>\n\n\n\n<li>The composite functor <em>U<\/em> o <em>Perv<\/em> is the identity functor on <strong>Top<\/strong>: indeed, that is what the Cs\u00e1sz\u00e1r-Pervin theorem above states.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">The composite functor <em>Perv<\/em> o <em>U<\/em> does not seem to have any notable feature, and in particular there does not seem to be any adjunction involving <em>Perv<\/em> and <em>U<\/em>.  Notably, the Pervin quasi-uniformity of a topology <strong><em>O<\/em><\/strong> is in general neither the smallest nor the largest quasi-uniformity that induces <strong><em>O<\/em><\/strong>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Here is a counterexample, again due to Pervin [3].  Consider the quasi-metric d<sub><strong>R<\/strong><\/sub> on the real numbers: d<sub><strong>R<\/strong><\/sub>(<em>a<\/em>,<em>b<\/em>)\u225dmax(<em>a<\/em>\u2013<em>b<\/em>,0).  That induces a quasi-uniformity, whose basic entourages are [&lt;<em>r<\/em>] = {(<em>a<\/em>,<em>b<\/em>) \u2208 <strong>R<\/strong> \u00d7 <strong>R<\/strong> | <em>a<\/em>&lt;<em>b<\/em>+<em>r<\/em>}, <em>r<\/em>&gt;0.  The topology induced by this quasi-uniformity is the open ball topology of d<sub><strong>R<\/strong><\/sub>, which is the Scott topology on <strong>R<\/strong>.  The Pervin quasi-uniformity of this Scott topology has basic entourages of the form <em>R<sub>U<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>U<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub> where each <em>U<\/em><sub><em>i<\/em><\/sub> is of the form ]<em>a<\/em><sub><em>i<\/em><\/sub>, \u221e[.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>No such entourage contains any quasi-metric entourage [&lt;<em>r<\/em>]: any pair (<em>a<\/em>,<em>b<\/em>) where <em>b<\/em> is strictly larger than every <em>a<\/em><sub><em>i<\/em><\/sub>, and <em>a<\/em>\u2265<em>b<\/em>+<em>r<\/em> is in <em>R<sub>U<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>U<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub> but not in [&lt;<em>r<\/em>].<\/li>\n\n\n\n<li>In the other direction, we claim that no quasi-metric entourage [&lt;<em>r<\/em>] contains any Pervin entourage <em>R<sub>U<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>U<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub> (where <em>U<\/em><sub><em>i<\/em><\/sub> = ]<em>a<\/em><sub><em>i<\/em><\/sub>, \u221e[).  if <em>n<\/em>=0, this is clear, since the empty intersection is the whole of <strong>R<\/strong> \u00d7 <strong>R<\/strong>.  Otherwise,  let \u03b5&gt;0 be chosen such that \u03b5&lt;<em>r<\/em>.  Then the point (<em>a<\/em><sub>1<\/sub>+\u03b5,<em>a<\/em><sub>1<\/sub>) is in [&lt;<em>r<\/em>], but not in <em>R<sub>U<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>U<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub>, in fact not even in <em>R<sub>U<\/sub><\/em><sub><sub>1<\/sub><\/sub>.<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">It follows that the two (Pervin and quasi-metric) quasi-uniformities on <strong>R<\/strong> are incomparable.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">One may, at least, attempt to identify what kinds of quasi-uniform spaces can be obtained by applying <em>Perv<\/em> to arbitrary topological spaces.  I do not know the answer to that question.  However, if we are prepared to replace topological spaces by sets equipped with a lattice of subsets (not necessarily closed under arbitrary unions), <em>Perv<\/em> generalizes naturally, and the answer to the corresponding question is known, as I will now argue.  The corresponding quasi-uniform spaces are called <em>Pervin spaces<\/em>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Pervin spaces<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">As far as I know, Pervin spaces were introduced by Mai Gehrke, Serge Grigorieff and Jean-\u00c9ric Pin, see [4].  They form a generalization of the Pervin quasi-uniformities introduced in the Cs\u00e1sz\u00e1r-Pervin theorem: we now define the Pervin quasi-uniformity of any family of subsets of a set <em>X<\/em> whatsoever, not necessarily a topology.  The formula is the same: for any subset <em>B<\/em> of <em>X<\/em>, <em>R<sub>B<\/sub><\/em> is the set of pairs (<em>x<\/em>,<em>y<\/em>) such that <em>x<\/em> \u2208 <em>B<\/em> implies <em>y<\/em> \u2208 <em>B<\/em>.  Accordingly, the Cs\u00e1sz\u00e1r-Pervin theorem has the following generalization.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Theorem.<\/strong>  Let <strong><em>F<\/em><\/strong> be a family of subsets of a set<em> X<\/em>.  The finite intersections of relations <em>R<sub>B<\/sub><\/em>, <em>B<\/em> \u2208 <strong><em>F<\/em><\/strong>, form a base <strong><em>B<\/em><\/strong> of entourages of a quasi-uniformity <strong><em>U<\/em><\/strong> on <em>X<\/em>, whose induced topology is exactly the topology <strong><em>O<\/em><\/strong> generated by <strong><em>F<\/em><\/strong> on <em>X<\/em>. This quasi-uniformity is the <em>Pervin quasi-uniformity<\/em> of the family <strong><em>F<\/em><\/strong>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Proof.  The proof that <strong><em>U<\/em><\/strong> is a quasi-uniformity is as in the Cs\u00e1sz\u00e1r-Pervin theorem.  Notably, we need to observe that <em>R<sub>B<\/sub><\/em> o <em>R<sub>B<\/sub><\/em> \u2286 <em>R<sub>B<\/sub><\/em>, namely that <em>R<sub>B<\/sub><\/em> is a transitive binary relation on <em>X<\/em> for every <em>B<\/em> in <strong><em>F<\/em><\/strong>. This implies that every basic entourage <em>R<sub>B<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>B<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub> is transitive as well.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Given any point <em>x<\/em> of <em>X<\/em>, and any basic entourage <em><em>S<\/em><\/em> \u225d <em><em>R<sub>B<\/sub><\/em><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em><em>R<sub>B<\/sub><\/em><sub><em>n<\/em><\/sub><\/em>, the image <em>S<\/em>[<em>x<\/em>] is equal to the intersection of the sets <em><em>B<sub>i<\/sub><\/em><\/em> that contain <em>x<\/em>. It follows that every <strong><em>U<\/em><\/strong>-open set is a neighborhood of each of its points (relative to the topology <strong><em>O<\/em><\/strong>), hence is open in <strong><em>O<\/em><\/strong>. Conversely, let <em>U<\/em> be any open set in <strong><em>O<\/em><\/strong>. For every <em>x<\/em> in <em>U<\/em>, there is a finite intersection <em><em>B<\/em><\/em><sub>1<\/sub> \u2229 &#8230; \u2229 <em><em>B<\/em><\/em><sub><em>n<\/em><\/sub> of elements of <strong><em>F<\/em><\/strong> that contains <em>x<\/em> and is included in <em>U<\/em>.  Let <em><em>S<\/em><\/em> \u225d <em><em>R<sub>B<\/sub><\/em><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em><em>R<sub>B<\/sub><\/em><sub><em>n<\/em><\/sub><\/em>.  Then <em>S<\/em>[<em>x<\/em>] = <em><em>B<\/em><\/em><sub>1<\/sub> \u2229 &#8230; \u2229 <em><em>B<\/em><\/em><sub><em>n<\/em><\/sub> is a <strong><em>U<\/em><\/strong>-neighborhood of <em>x<\/em> included in <em>U<\/em>, so <em>U<\/em> is <strong><em>U<\/em><\/strong>-open. \u2610<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In the proof, we have noticed that the Pervin quasi-uniformity <strong><em>U<\/em><\/strong> of <strong><em>F<\/em><\/strong> has a base of <em>transitive<\/em> entourages (namely, those of the form <em>R<sub>B<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>B<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub>).  The quasi-uniformities with that property are called&#8230; <em>transitive<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Here is another property of <strong><em>U<\/em><\/strong>.  Given any entourage <em>R<\/em> in <strong><em>U<\/em><\/strong>, <em>R<\/em> contains some basic entourage <em>R<sub>B<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>B<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub>.  For every subset <em>I<\/em> of {1, &#8230;, <em>n<\/em>}, let <em>C<sub>I<\/sub><\/em> be the set obtained as the intersection of the sets <em>B<\/em><sub><em>i<\/em><\/sub> with <em>i<\/em> \u2208 <em>I<\/em>, and of the complements of the sets <em>B<\/em><sub><em>i<\/em><\/sub> with <em>i<\/em> \u2209 <em>I<\/em>.  The family of all the sets <em>C<sub>I<\/sub><\/em>, where <em>I<\/em> ranges over the subsets of {1, &#8230;, <em>n<\/em>}, forms a finite cover of the whole space <em>X<\/em>.  Moreover, <em>C<sub>I<\/sub><\/em> \u00d7 <em>C<sub>I<\/sub><\/em> is included in <em>R<sub>B<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>B<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub>, hence in <em>R<\/em>.  Indeed, for every pair (<em>x<\/em>,<em>y<\/em>) in <em>C<sub>I<\/sub><\/em> \u00d7 <em>C<sub>I<\/sub><\/em>, for every index <em>i<\/em>, if <em>i<\/em> \u2208 <em>I<\/em> then (<em>x<\/em>,<em>y<\/em>) is in <em>R<sub>B<\/sub><\/em><sub><sub><em>i<\/em><\/sub><\/sub> because <em>y<\/em> is in <em>B<\/em><sub><em>i<\/em><\/sub>, and if <em>i<\/em> \u2209 <em>I<\/em> then (<em>x<\/em>,<em>y<\/em>) is in <em>R<sub>B<\/sub><\/em><sub><sub><em>i<\/em><\/sub><\/sub> because <em>x<\/em> is <em>not<\/em> in <em>B<\/em><sub><em>i<\/em><\/sub>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">A quasi-uniformity <strong><em>U<\/em><\/strong> on a set <em>X<\/em> is <em>totally bounded<\/em> if and only if for every entourage <em>R<\/em> \u2208 <strong><em>U<\/em><\/strong>, there is a finite cover of <em>X<\/em> such that, for every element <em>C<\/em> of that cover, <em>C<\/em> \u00d7 <em>C<\/em> is included in <em>R<\/em>.  We have just argued that the Pervin quasi-uniformity of a family <strong><em>F<\/em><\/strong> of subsets is always transitive and totally bounded.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">It is instructive to see what this notion of  &#8216;totally bounded&#8217; means when restricted to a hemi-metric space <em>X<\/em>, <em>d<\/em>.  For every subset <em>C<\/em> of <em>X<\/em>, for every <em>r<\/em>>0, <em>C<\/em> \u00d7 <em>C<\/em> is included in [&lt;<em>r<\/em>] if and only if, for all <em>x<\/em>, <em>y<\/em> in <em>C<\/em>, <em>d<\/em>(<em>x<\/em>,<em>y<\/em>)&lt;<em>r<\/em>.  Every such subset <em>C<\/em> is either empty, or included in the open ball with center <em>x<\/em> and radius <em>r<\/em> under the symmetrized pseudo-metric <em>d<\/em><sup>sym<\/sup>, where <em>x<\/em> is any given point of <em>C<\/em>.  Hence, if <em>X<\/em>, <em>d<\/em> is totally bounded qua quasi-uniform space, then for every <em>r<\/em>>0, there is a finite cover of <em>X<\/em> by (sets included in) open balls of radius <em>r<\/em> under <em>d<\/em><sup>sym<\/sup>.  In other words, <em>X<\/em>, <em>d<\/em> is totally bounded qua hemi-metric space (compare with Definition 6.7.4 in the <a href=\"https:\/\/www.cambridge.org\/gb\/knowledge\/isbn\/item7069109\/Non-Hausdorff%20Topology%20and%20Domain%20Theory\/?site_locale=en_GB\">book<\/a>).<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Conversely, if <em>X<\/em>, <em>d<\/em> is totally bounded qua hemi-metric space, then for every <em>r<\/em>&gt;0, <em>X<\/em> has a cover by open balls <em>B<\/em> of radius <em>r<\/em>\/2 under <em>d<\/em><sup>sym<\/sup>, and each of these open balls <em>B<\/em> satisfies <em>B<\/em> \u00d7 <em>B<\/em> \u2286 [&lt;<em>r<\/em>].  Therefore <em>X<\/em>, <em>d<\/em> is totally bounded qua quasi-uniform space.  In summary, &#8216;totally bounded&#8217; really means the same thing as the usual notion on hemi-metric spaces.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Let us return to Pervin quasi-uniformities.  The following is due to Mai Gehrke, Serge Grigorieff and Jean-\u00c9ric Pin (April 2012; no, I will not give any more precise reference, because the document I have says that it is not finished and therefore must remain confidential\u2014it is fairly easy to find it on the Web nonetheless).  The proof I will give is different, though.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Proposition.<\/strong>  A quasi-uniformity <strong><em>U<\/em><\/strong> on a set <em>X<\/em> is the Pervin quasi-uniformity of a family of subsets of <em>X<\/em> if and only if it is transitive and totally bounded.  In that case:<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><strong><em>U<\/em><\/strong> is the Pervin quasi-uniformity of the family <strong><em>F<\/em><\/strong> of subsets of <em>X<\/em> of the form <em>R<\/em>[<em>E<\/em>] \u225d {<em>y<\/em> \u2208 <em>X<\/em> | \u2203<em>x<\/em> \u2208 <em>E<\/em>, (<em>x<\/em>,<em>y<\/em>) \u2208 <em>R<\/em>}, where <em>R<\/em> ranges over the transitive entourages in <strong><em>U<\/em><\/strong> and <em>E<\/em> ranges over the subsets of <em>X<\/em>;<\/li>\n\n\n\n<li><em><strong>F<\/strong><\/em> is a subbase of the topology <strong><em>O<\/em><\/strong> induced by <strong><em>U<\/em><\/strong>;<\/li>\n\n\n\n<li><strong><em>U<\/em><\/strong> is also the Pervin quasi-uniformity of the lattice <strong><em>L<\/em><\/strong> of finite unions of finite intersections of elements of <strong><em>F<\/em><\/strong>, and <strong><em>L<\/em><\/strong> is a base of <strong><em>O<\/em><\/strong>.<\/li>\n<\/ol>\n\n\n\n<p class=\"wp-block-paragraph\">Proof.  We have already proved the &#8216;only if&#8217; direction.  Let us assume that <strong><em>U<\/em><\/strong> is transitive and totally bounded.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">We first note that every element of <strong><em>F<\/em><\/strong> is open (in <strong><em>O<\/em><\/strong>).  Indeed, let us pick any such element.  It is of the form <em>R<\/em>[<em>E<\/em>], where <em>R<\/em> is a transitive entourage, and <em>E<\/em> is a subset of <em>X<\/em>.   For every point <em>y<\/em> of <em>R<\/em>[<em>E<\/em>], every point <em>z<\/em> \u2208 <em>R<\/em>[<em>y<\/em>] is such that (<em>y<\/em>,<em>z<\/em>) \u2208 <em>R<\/em>, and since (<em>x<\/em>,<em>y<\/em>) \u2208 <em>R<\/em> for some <em>x<\/em> in <em>E<\/em> and <em>R<\/em> is transitive, (<em>x<\/em>,<em>z<\/em>) is also in <em>R<\/em>, showing that <em>z<\/em> is in <em>R<\/em>[<em>E<\/em>].  This shows that <em>R<\/em>[<em>y<\/em>] is a neighborhood of <em>y<\/em> included in <em>R<\/em>[<em>E<\/em>].  Since <em>y<\/em> is arbitrary in <em>R<\/em>[<em>E<\/em>], <em>R<\/em>[<em>E<\/em>] is open in <strong><em>O<\/em><\/strong>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">We claim that, given any element <em>B<\/em>\u225d<em>R<\/em>[<em>E<\/em>] of <strong><em>F<\/em><\/strong>, <em>R<sub>B<\/sub><\/em> is in <strong><em>U<\/em><\/strong>.  (Remember that <em>R<\/em> is assumed to be transitive, and in <strong><em>U<\/em><\/strong>, here.)  Every pair (<em>y<\/em>,<em>z<\/em>) in <em>R<\/em> is in <em>R<sub>B<\/sub><\/em>: if <em>y<\/em> is in <em>B<\/em>=<em>R<\/em>[<em>E<\/em>], then (<em>x<\/em>,<em>y<\/em>) is in <em>R<\/em> for some <em>x<\/em> in <em>E<\/em>, hence (<em>x<\/em>,<em>z<\/em>) is also in <em>R<\/em>, since <em>R<\/em> is transitive; so <em>z<\/em> is in <em>B<\/em>.  This shows that <em>R<\/em> is included in <em>R<sub>B<\/sub><\/em>.  By (filter 2), <em>R<sub>B<\/sub><\/em> is in <strong><em>U<\/em><\/strong>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">A similar argument shows that <em>R<sub>A<\/sub><\/em> is in <strong><em>U<\/em><\/strong> for every <em>A<\/em> in <strong><em>L<\/em><\/strong>.  Let us write <em>A<\/em> as the finite union (over <em>i<\/em>=1&#8230;<em>m<\/em>) of the finite intersection (over <em>j<\/em>=1&#8230;<em>n<sub>i<\/sub><\/em>) of elements <em>B<sub>ij<\/sub><\/em> \u225d <em><em>R<sub>ij<\/sub><\/em><\/em>[<em>E<em><sub>ij<\/sub><\/em><\/em>] of <strong><em>F<\/em><\/strong>.  Let <em>R<\/em> be the intersection of <em>all<\/em> the relations <em><em>R<sub>ij<\/sub><\/em><\/em>, when <em>i<\/em> and <em>j<\/em> vary.  We claim that <em>R<\/em> is included in <em>R<sub>A<\/sub><\/em>.  As above, this will entail that <em>R<sub>A<\/sub><\/em> is in <strong><em>U<\/em><\/strong>.  Given any pair (<em>y<\/em>,<em>z<\/em>) in <em>R<\/em> such that <em>y<\/em> is in <em>A<\/em>, there is an index <em>i<\/em> such that for every <em>j<\/em> (1\u2264<em>j<\/em>\u2264<em>n<sub>i<\/sub><\/em>), <em>y<\/em> is in <em>B<sub>ij<\/sub><\/em>=<em><em>R<sub>ij<\/sub><\/em><\/em>[<em>E<em><sub>ij<\/sub><\/em><\/em>].  Hence for every <em>j<\/em>, there is a point <em>x<em><sub>j<\/sub><\/em><\/em> in <em>E<em><sub>ij<\/sub><\/em><\/em> such that (<em>x<em><sub>j<\/sub><\/em><\/em>,<em>y<\/em>) is in <em><em>R<sub>ij<\/sub><\/em><\/em>.  Since (<em>y<\/em>,<em>z<\/em>) is in <em>R<\/em>, hence in <em><em>R<sub>ij<\/sub><\/em><\/em>, and since <em><em>R<sub>ij<\/sub><\/em><\/em> is transitive, (<em>x<em><sub>j<\/sub><\/em><\/em>,<em>z<\/em>) is also in <em><em>R<sub>ij<\/sub><\/em><\/em> for every <em>j<\/em>.  This shows that <em>z<\/em> is in the intersection of the sets <em>B<sub>ij<\/sub><\/em>=<em><em>R<sub>ij<\/sub><\/em><\/em>[<em>E<em><sub>ij<\/sub><\/em><\/em>], 1\u2264<em>j<\/em>\u2264<em>n<sub>i<\/sub><\/em>, hence in <em>A<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">In the converse direction, we claim that every entourage <em>R<\/em> (in <strong><em>U<\/em><\/strong>) contains some basic Pervin entourage <em>R<sub>B<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>B<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub>, where each <em>B<sub>i<\/sub><\/em> is in <strong><em>F<\/em><\/strong>.  Since <strong><em>U<\/em><\/strong> is transitive, and up to the replacement of <em>R<\/em> by a smaller entourage, we may assume that <em>R<\/em> is transitive.  We now use total boundedness.  Let {<em><em>C<\/em><\/em><sub>1<\/sub>, &#8230;, <em><em>C<\/em><\/em><sub><em>n<\/em><\/sub>} be some finite cover of <em>X<\/em> such that each product <em><em>C<\/em><\/em><sub><em>i<\/em><\/sub> \u00d7 <em><em>C<\/em><\/em><sub><em>i<\/em><\/sub> is included in <em>R<\/em>.  For each <em>i<\/em>, <em>R<\/em>[<em><em>C<\/em><\/em><sub><em>i<\/em><\/sub>] is in <strong><em>F<\/em><\/strong>, by definition.  We let <em>B<sub>i<\/sub><\/em> \u225d <em>R<\/em>[<em><em>C<\/em><\/em><sub><em>i<\/em><\/sub>].  For every pair (<em>x<\/em>,<em>y<\/em>) in <em>R<sub>B<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>B<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub>, we claim that (<em>x<\/em>,<em>y<\/em>) is in <em>R<\/em>.  Since {<em><em>C<\/em><\/em><sub>1<\/sub>, &#8230;, <em><em>C<\/em><\/em><sub><em>n<\/em><\/sub>} is a cover, there is an index <em>i<\/em> such that <em>x<\/em> is in <em><em>C<\/em><\/em><sub><em>i<\/em><\/sub>, hence in the larger set <em>B<sub>i<\/sub><\/em> = <em>R<\/em>[<em><em>C<\/em><\/em><sub><em>i<\/em><\/sub>].  (Note that <em><em><em>C<\/em><\/em><sub><em>i<\/em><\/sub><\/em> is included in <em>R<\/em>[<em><em>C<\/em><\/em><sub><em>i<\/em><\/sub>] because <em>R<\/em> is reflexive.)  Now (<em>x<\/em>,<em>y<\/em>) is in <em>R<sub>B<\/sub><\/em><sub><sub><em>i<\/em><\/sub><\/sub>, so <em>y<\/em> is also in <em>B<sub>i<\/sub><\/em>.  It follows that there is a point <em>z<\/em> in <em><em>C<\/em><\/em><sub><em>i<\/em><\/sub> such that (<em>z<\/em>,<em>y<\/em>) is in <em>R<\/em>.  Therefore (<em>x<\/em>,<em>z<\/em>) is in <em><em>C<\/em><\/em><sub><em>i<\/em><\/sub> \u00d7 <em><em>C<\/em><\/em><sub><em>i<\/em><\/sub>, hence in <em>R<\/em>.  It follows that (<em>x<\/em>,<em>y<\/em>) is also in <em>R<\/em>, since <em>R<\/em> is transitive.  This finishes to show that <em>R<sub>B<\/sub><\/em><sub><sub>1<\/sub><\/sub> \u2229 &#8230; \u2229 <em>R<sub>B<\/sub><\/em><sub><sub><em>n<\/em><\/sub><\/sub> is included in <em>R<\/em>.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">All this establishes item 1, and also item 3.  Item 2 is a consequence of the previous Theorem.  \u2610<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Warning.<\/strong>  The previous proposition does <em>not<\/em> show that <em><strong>U<\/strong><\/em> is the Pervin topology of the topology <strong><em>O<\/em><\/strong>, only of its subbase <em><strong>F<\/strong><\/em>, or of its base <strong><em>L<\/em><\/strong>.  The latter is closed under finite unions, but not under arbitrary unions in general.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">There is much more that one can say about quasi-uniform spaces&#8230; enough to fill a <a href=\"https:\/\/www.cambridge.org\/gb\/knowledge\/isbn\/item7069109\/Non-Hausdorff%20Topology%20and%20Domain%20Theory\/?site_locale=en_GB\">book<\/a>, quite certainly.  One should certainly read Hans-Peter K\u00fcnzi&#8217;s introduction to the subject [5], which is already quite encyclopedic!  I was sad to hear that he had recently passed away (see <a href=\"https:\/\/www.nrf.ac.za\/media-room\/news\/obituary-professor-hans-peter-k%C3%BCnzi\">this page<\/a>).  I had met him once or twice; the last time in Leicester, UK, in 2016, and I certainly enjoyed the lively breakfast we had there one morning with his many students and postdocs.<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>Nicolas Bourbaki.  <a href=\"https:\/\/www.springer.com\/gp\/book\/9783540339366\">Topologie g\u00e9n\u00e9rale<\/a> (\u00e9l\u00e9ments de math\u00e9matique), chapitre 2.  Springer, 2007.  (Many other, previous editions, too.)<\/li>\n\n\n\n<li>\u00c1kos Cs\u00e1sz\u00e1r.  Fondements de la topologie g\u00e9n\u00e9rale.  Gauthier-Villars, Paris, 1960.<\/li>\n\n\n\n<li>William Joseph Pervin.  Quasi-Uniformisation of Topological Spaces.  <a href=\"https:\/\/link.springer.com\/journal\/208\"><em>Mathematische Annalen<\/em><\/a>&nbsp;147:316\u2013317, 1962.<\/li>\n\n\n\n<li>Jean-\u00c9ric Pin.  <a href=\"https:\/\/www.mat.uc.pt\/~wdual\/slidesPin.pdf\">Pervin Spaces<\/a>.  Slides of a talk given in Coimbra, Portugal, September 2016.<\/li>\n\n\n\n<li>Hans-Peter Albert K\u00fcnzi.  An Introduction to Quasi-Uniform Spaces.  In <em><a href=\"https:\/\/dx.doi.org\/10.1090\/conm\/486\">Beyond Topology<\/a><\/em>, pages 239-304, volume 486 of the <em>Contemporary Mathematics<\/em> series, American Mathematical Society, 2009, edited by Fr\u00e9d\u00e9ric Mynard and Elliott Pearl.<\/li>\n<\/ol>\n\n\n<div class=\"wp-block-image\">\n<figure class=\"alignright is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/wp-content\/uploads\/2016\/08\/jgl-2011.png\" alt=\"jgl-2011\" class=\"wp-image-993\" width=\"60\" height=\"83\"\/><\/figure>\n<\/div>\n\n\n<p class=\"wp-block-paragraph\">\u2014 <a href=\"https:\/\/www.lsv.ens-paris-saclay.fr\/~goubault\/?l=en\">Jean Goubault-Larrecq<\/a> (October 19th, 2020)<\/p>\n","protected":false},"excerpt":{"rendered":"<p>A uniform space is a natural generalization of the notion of a metric space, on which completeness still makes sense [1]. The non-Hausdorff variant of this is called a quasi-uniform space, and the purpose of this post is to introduce &hellip; <a href=\"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/?page_id=2869\">Continue reading <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_crdt_document":"","footnotes":""},"class_list":["post-2869","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/pages\/2869","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=2869"}],"version-history":[{"count":41,"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/pages\/2869\/revisions"}],"predecessor-version":[{"id":5895,"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=\/wp\/v2\/pages\/2869\/revisions\/5895"}],"wp:attachment":[{"href":"https:\/\/projects.lsv.ens-paris-saclay.fr\/topology\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=2869"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}