The Boon Companion has been experimenting with long exposure times and intentional camera movement, of late. She was just about to discard the motion-blurred cyclist above as a failed experiment when something about the image caught my eye.
In the thirtieth-of-a-second exposure, the bicycle wheel has rolled a short distance. But why do the spokes look curved? Why don’t the curves point towards the centre of the wheel? And why is the effect only visible in the lower half of the wheel?
So I sat down to figure out the trajectory of a bicycle spoke as the wheel rolls along the ground. As you do.
Any point on wheel rolling across a flat surface without slipping follows a curve called atrochoid (from Greek trochos, “wheel”). I won’t pester you with the relevant equations (they’re at the other end of the link above). Here’s what the trochoid curves look like for the two ends of a radial spoke, spanning the distance between a wheel hub and a thick wheel rim:
The shape of the wheel is plotted in grey dashes, with a single vertical spoke marked. As the wheel rolls left or right, the ends of the spoke follow the curved trochoid trajectories, with successive positions marked at 20º intervals.
But bicycle wheels don’t (usually) have radial spokes, and I felt obliged, going into the problem, to look at the position of real bicycle spokes. Here’s a very common pattern:
Thirty-six spokes, laid out in what’s called a “three cross” pattern. Fundamentally, there are only two different spoke alignments in this pattern—leading and trailing.
For a wheel rolling from right to left, the spoke I’ve highlighted in red is leading, and the blue spoke is trailing. They’re simply mirror images of each other. One side of the wheel has nine leading spokes, spaced 40º apart, and nine trailing spokes in the same pattern. The other side of the wheel is laid out exactly the same way, but with the pattern rotated by 20º. The final result is called a “three cross” pattern because each trailing spoke crosses three leading spokes on its side of the wheel (and vice versa).
In this pattern, the anchor point for the spoke at the hub is offset 60º relative to its attachment at the rim. So to see the trajectory of a representative bicycle spoke, I need to slide the trochoid curve for the hub 60º out of alignment with the rim curve. Here its, with the spoke drawn in at 20º intervals:
This is the trajectory of a trailing spoke for a wheel rolling right to left, and a leading spoke for a wheel rolling left to right.
We can already get a hint of why some sort of spoke pattern shows up in the lower half of the wheel, but not the upper. In the upper half, the spoke is moving rapidly sideways, as it pivots across the top of the wheel; in the lower half it performs a sort of dipping motion, arcing downwards towards the point at which the wheel contacts the road, and then arcing back up again.
Now, I figure a cyclist moving at a reasonable speed for a shared-use path will rotate the wheels through about 30º during a thirtieth-of-a-second exposure. Here’s a more detailed trajectory for a trailing spoke (for a bicycle moving right-to-left) during a 30º rotation:
You can see that the spoke slides along itself, to some extent—different parts of the spoke occupy the same spatial position at different times. These are the only parts of the spoke that will show up as a dense “shadow” during a prolonged exposure that blurs the other parts. In the final photograph we’ll therefore see something like the arc I’ve sketched in red, while the rest of the spoke is smeared into a blur:
Something similar happens for a leading spoke as it passes through the same position:
So, actually, while the leading/trailing distinction slightly changes the details, both kinds of spoke produce a curved shadow during a prolonged exposure.
If we catch a spoke that’s higher in its trajectory, we get another arc:
Now we can put these images together, showing the curved shadows of several spokes at once. Each of them will lie on a different set of trochoid arcs, shifted laterally according to how far the spoke lies from the lowest point of its trajectory. Like this:
I’ve marked the visual centre of the wheel, for reference. Notice how the partial shadow arcs formed by the lower spokes seem to point below the centre, while the arcs of the higher spokes point above the centre. It happens because none of the spokes are radial, and because the centre of the wheel is never stationary, but shifts horizontally as the spokes sketch out their curved arcs.
So. Although I was baffled by the photograph when I first saw it, I began to get an inkling of what was going on as I made the first trochoid sketches, and was then pleasantly surprised by how things began to fall neatly into place as I added more detail. I’m hoping you’re as pleasantly surprised as I am.
Reflection: A transformation under which each point in a shape appears at an equal distance on the opposite side of a given line—the line of reflection.
It’s not often I have occasion to shout at the television, but a recent episode of the BBC’s long-running television series QI precipitated just such an outburst. The cause of my vexation was their answer to the question that forms the title of this post. The offending episode was the R Series: Reflections, and the explanation was an excellent approximation to gibberish, involving as it did some business about “The mirror doesn’t flip things around; we flip things around,” intoned by Sandi Toksvig as she stood in front of a mirror fiddled with a bit of card with the word BOSS written on it (see above). To be fair to Toksvig, it probably wasn’t her idea, and she did manage to deliver the entire farrago while wearing the sort of anxious expression people wear when they’re not entirely convinced by their own argument.
The answer to the question is really that it is ill-posed. Mirrors actually don’t reverse left and right, for the simple reason that mirrors have no way of telling left from right. They have no left-right asymmetry, in other words. The only asymmetry they do possess is in the plane of reflection—stuff in front of the mirror is the real world; stuff “behind” the mirror is the reflected world.
That’s what’s being described in the definition at the head of this post, which refers to a two-dimensional reflection, like this:
Here we have a “line of reflection”, corresponding to the mirror; a letter “B” in front of the mirror, representing the real world; and a reflected letter “B” behind the mirror. Every point on the reflected “B” is the same distance from the mirror as the corresponding point on the original “B”. So because the spine of the “B” is the farthest part from the mirror, its reflection also lies farthest from the mirror. Conversely, the curved parts of the letter lie closer to the mirror on both sides. And it’s that preservation of “near” and “far” on either side of the mirror plane which causes the reflection to be a reversed image of the original. If we travel from one side of the mirror to the other, we encounter in turn spine-curves-mirror-curves-spine. So, actually (and pace Toksvig), the mirror very much does “flip things around”. Indeed, many introductory geometry texts gloss the word “reflection” as “a flip”.
The same thing happens when a three-dimensional person stands in front of a real mirror:
The reflected image’s left and right (and head and feet) are pointing in the same direction as the real person’s. What has been flipped in the mirror image is the direction in which the nose and toes are pointing. So the mirror has reversed front and back, not left and right. If you lie down with your feet pointing at the mirror, the reflection will also have its feet pointing at the mirror—so on this occasion the mirror leaves your front and back, and left and right, in the same positions, but reverses you top to bottom. Only if you stand sideways on to the mirror does it truly reverse your left and right—but that’s because you’ve chosen to place one side of your body close to the mirror and the other far from it, not because the mirror has some magical ability to tell left from right.
So why do we always think the mirror image has reversed left and right, no matter how we orientate ourselves before the mirror? The answer, I think, lies in the single plane of symmetry in our own bodies. Our fronts are very different from our backs, our heads are very different from our feet, but our left side is very similar to our right side. So it’s very difficult for us to see the mirror reflection as having reversed front and back—instead, we see it as another person who has turned around to face us. In which case, their right hand now moves when we move our left hand, and vice versa. No matter what the orientation of the reflection, we always interpret it as a left-right reversed person, because a head-foot reversed person or a front-back reversed person is harder to conceptualize.
Remove the left-right symmetry, and we stop talking about mirrors reversing left and right.
This barber’s pole lacks a clear left-right distinction. So instead, we find ourselves saying that it “spirals the opposite way”. We’re still, apparently, unable to discern that what has been reversed is the front and back, but now we can’t blame a left-right switch either.
So if anyone ever asks you the title question, permit yourself the slightest of headshakes and the faintest of smiles, and say: “But mirrors don’t reverse left and right; they preserve near and far.”
A couple of weeks ago, I reviewed three books about the activities of 161 (Special Duties) Squadron, RAF, during the Second World War. For this post, I want to talk specifically about the cover of Hugh Verity’s memoir and personal history of 161 Squadron, We Landed By Moonlight (Revised Edition), published in 2000 by Crécy. It’s a marvellous book, as a source of both anecdote and historical record, and we should all be grateful to Crécy for keeping it in print—but it’s an odd cover.
The first thing that struck me about it is that the Westland Lysander on the cover is sporting South-East Asia Command roundels and flashes, putting it a very long way from 161 Squadron’s base in the south of England. The image in fact comes from this photograph of Lysander V9289, of 357 (Special Duties) Squadron, RAF, operating in Burma. Right kind of aircraft, right kind of duties, wrong continent. But that’s fair enough, given that 161 Sq. Lysanders flew almost entirely at night on secret missions, so tended to go unphotographed unless they actually crashed.
The Lysander image has been composited with a fine full moon, to produce an atmospheric cover image. (In fact, there are two versions of this cover from Crécy, both using the same Lysander and moon images—you can find the other in my link from the book title, above.) And it’s that moon image that really got me puzzling, and inspired this post. Here it is, in a larger and more contrasty version:
It definitely doesn’t look like our own familiar full moon:
But it doesn’t look like a random painting, either. And I found it naggingly familiar. At first, I wondered if it was a photograph of some other moon in our solar system, but then I began to recognize significant features. The curve of Mare Nectaris below the three linked blotches of Mare Serenitatis, Mare Tranquilitatis and Mare Fecunditatis settled it—it is a photograph of our Moon.
But there are three things wrong with it.
One is that it is mirror-reversed. The real Moon looks like this:
The second is that it is pretty much lying on its side. By my estimate, the north-south axis is tilted at about twenty degrees to the horizontal:
Now, you can see the Moon in this orientation, if you catch it just after moonrise in the tropics, when it’s moving almost vertically towards the zenith. But the farther north or south you go, the more tilted is the Moon’s trajectory as it rises, and the closer to vertical is its axis. Even when the Moon is as far into the northern sky as it ever gets, it can never be seen in that orientation anywhere in France, which was the 161 Sq. stamping ground.
But that’s a nitpick, really, because the striking thing about this view is that you can never see it from Earth. The paired dark blotches about halfway towards the upper rim of the Moon, above, are Mare Marginis and Mare Smythii, which (as the former name implies) sit right on the edge of the Moon’s disc when seen from Earth. The photograph used on Crécy’s cover has actually been taken by a spacecraft, somewhere over about 60ºE lunar longitude.
Once I’d figured all this out, I realized why the image looked naggingly familiar. This view is a classic, because it’s what successive Apollo astronauts saw as they departed the Moon towards Earth. At the conclusion of their lunar mission, they fired their main engine as they orbited over the far side of the Moon, and then came looping around the eastern hemisphere, pulling away in a long orbit back towards Earth. The same view was photographed several times, by several relieved astronauts, but I think the Moon on Crécy’s cover is this one:
It’s one of a series of departure photographs taken by Apollo 11 on 22 July 1969. If Verity landed his Lysander by that moonlight, he was a very long way from home!
In the northern hemisphere, the Harvest Moon falls on 1 October in 2020, which is what provokes this post. The Harvest Moon is defined as the full moon that occurs closest to the autumnal equinox, which fell on 22 September (in the northern hemisphere, in 2020). You can find many lists of “names of the full moons” on-line (there’s a rather marvellous compilation of lists here), but the Harvest Moon is the only one that’s defined by the date of the equinox, rather than the month in which it falls—about three times in four it occurs in September, but the rest of the time (as on this occasion) it drifts into October.
The other thing about the Harvest Moon is that it has real astronomical and historical significance. Like many other full moon names, it obviously derives from what’s going on in the seasonal cycle at the time it appears—but there’s a deeper significance, which is what I want to write about here.
To understand what’s special about the full moon around the autumnal equinox, and its relevance to harvesting crops, we need to talk a bit about the orbit of the moon.
As is well known, the Earth’s rotation axis is inclined to the plane of its orbit around the sun, by about 23½º. So the Earth’s rotation and its orbit define two planes, tilted relative to each other—the celestial equator, which is the extension of the Earth’s equatorial plane; and the ecliptic, which is the plane of the Earth’s orbit. So from the vantage point of the Earth, the sun moves around the sky along the ecliptic plane, from west to east, completing one revolution per year. Like this:
The two points at which the celestial equator and the ecliptic intersect have names with complicated astrological origins. The point on the celestial equator which the sun crosses when heading north is called the First Point of Aries. The point opposite that is The First Point of Libra. Both are symbolized by the zodiacal symbols for their corresponding constellations. These are the locations of the sun at the times of the equinoxes—it crosses the First Point of Aries at the March equinox, spends six months bringing summer to the northern hemisphere, and then crosses the First Point of Libra at the September equinox, on its way south for the southern hemisphere summer.
The moon orbits close to the ecliptic plane. For the purposes of this discussion, we can treat it as travelling in the ecliptic plane, and come back to the slight deviation later. So the moon moves (roughly) along the ecliptic from west to east, taking a month to make a full revolution. It also passes through the First Points of Aries and Libra, spending two week over the northern hemisphere, and two weeks over the south.
The moon makes one complete circuit of the celestial sphere every 27.3 days. If it moved at a constant rate along the celestial equator, it would therefore be about 13º farther west every day. The Earth would need to rotate correspondingly farther between successive moonrises and moonsets, making each moonrise and moonset occur about fifty minutes later than its predecessor. And that’s true on average for the real moon. But the fact that the moon’s orbit follows the ecliptic, and not the equator, introduces a subtle variation.
Here’s what happens to successive moonrises at 50ºN latitude, when the moon is passing through the First Point of Aries.
Its 13º displacement along the ecliptic has a northward component in this part of its orbit, which means that it lies closer below the horizon on successive nights than it would do if it were moving parallel to the equator. So the Earth has to rotate less far between successive moonrises, and the moon rises only slightly later each night. (The effect becomes more pronounced at higher latitudes, and less so at lower latitudes.)
But at moonset, the northward movement at the First Point of Aries serves to lift the moon farther above the horizon than it would otherwise be. So successive moonsets show longer delays than the average 50 minutes when the moon is in this part of its orbit.
The situation is reversed two weeks later, as the moon passes through the First Point of Libra. Now each successive moonrise at northern latitudes is delayed more than 50 minutes, like this:
And it should come as no surprise that the delay between successive moonsets is correspondingly shortened at this point in the moon’s orbit.
So although it all averages out over the course of a month, there’s a regular variation in the delay between successive moonrises (and moonsets) during that period. Here’s a chart of the delays for a representative period (September and October 2018) at 50ºN; I’ve marked the passages through the First Point of Aries:
So this happens every month. Why is it particularly relevant only once a year, on the Harvest Moon? Because full moons occur only when the sun is on the opposite side of the sky from the moon. Which means the only time we see a full moon passing through the First Point of Aries is when the sun is in the vicinity of the First Point of Libra—which, you’ll recall from the top of this post, happens during the September equinox. So in the northern hemisphere, at the time of the autumnal equinox, the full moon rises at almost the same time for several successive evenings, but sets more than an hour later each morning. And if you’re bringing in the harvest (as you do in temperate latitudes in the autumn), and you don’t have access to artificial outdoor illumination, then that’s hugely advantageous. For several nights, the full moon rises before twilight fades, and sets only when the morning sky is already bright. You can work around the clock, day and night, to get the crops in, in other words. Which is what’s going on in Mason’s painting at the head of this post.
Does the southern hemisphere have a Harvest Moon? It surely does. All the geometry flips over, so the First Point of Libra assumes the role that the First Point of Aries does in the northern hemisphere. Like this:
Full moons occur at this point when the sun is passing through the First Point of Aries—the March equinox, which is the autumnal equinox for the southern hemisphere. Isn’t that neat? (Well, I think it’s neat.)
There are a couple of subtleties, which mean not every Harvest Moon is the same. The first complicating factor is that the moon’s orbit does not lie perfectly in the ecliptic, but inclined to it at about 5º. The inclined orbit of the moon twists continuously in the ecliptic plane, completing one rotation every 18.6 years, under the influence of the sun’s gravity. This means that the tilt of the moon’s orbit sometimes subtracts from the angle between the ecliptic and the celestial equator, and sometimes adds to it, like this:
So we have “seasons” when the Harvest Moon is delayed even less than average on successive nights, and seasons (nine years later) when it is delayed more than average.
The other complicating factor is that the moon doesn’t orbit in a perfect circle—it moves in an ellipse, and it crosses the sky more slowly when it’s farthest from the Earth (its apogee), and more quickly when it’s closest (perigee). This ellipse twists around in the plane of the moon’s orbit with a rotation period of about 8.8 years. When the apogee aligns with the First Point of Aries, the delay between successive Harvest Moon moonrises is shortened even farther. Conversely, a few years later, the perigee will prolong the delay between successive moonrises.
It so happens that apogee is passing through the First Point of Aries in 2020, and we can see a noticeable effect on moonrises and moonsets. Here’s the delay graph for September and October 2020, again at 50ºN.
The slow movement of the moon at the First Point of Aries shortens the delay between successive moonrises, and between successive moonsets. Conversely, the fast movement at the First Point of Libra lengthens these same time periods. So the delay graphs are dented downwards at Aries, and shoved upwards at Libra—you can see this most clearly in the flattened tops on the moonset curve.
There are few places left in the world where any of this is relevant to the life of farmers, of course. But it’s a fine astronomical curiosity, I hope you’ll agree.
In my last two posts about rainbows, I discussed the formation of the primary and secondary rainbows, respectively, tracing their origins to specific light paths through falling raindrops.
The primary rainbow ray follows a path like this: For a raindrop at the apex of the rainbow arc, sunlight enters near the top of the drop, bounces once off the back, and then exits the bottom, descending towards the observer’s eye, making an angle of around 41.5º for the green ray shown.
For the secondary rainbow, sunlight enters near the bottom of the drop, is reflected internally twice, and then exits the front of the drop, descending towards the observer at an angle of around 51º.
These rainbow rays are special, representing the maximum or minimum angles of deflection of the incoming ray. And they are associated with a particular offset of the incoming ray in its interaction with the raindrop. I measured offset like this:The rainbow ray for the primary enters the raindrop at an offset of about 0.86; the secondary rainbow ray at about 0.95.
For more about these topics, in particular the importance of maximum and minimum deflections, I direct you back to my previous posts.
I finished my most recent post on this topic with a question: If a single internal reflection produces a primary rainbow, and two internal reflections produce a secondary rainbow, is there such a thing as a tertiary rainbow? And if so, where is it?
The path of the rainbow ray for a tertiary bow looks like this:
The three reflections carry it almost all the way around the raindrop, so that (in contrast to the primary and secondary bows) it leaves the drop heading away from the sun. This tells us that, to see a tertiary bow, we’d need to look towards the sun, rather than (as for the primary and secondary) towards the antisolar point.
Plotting my usual graph of light deflection and transmission against the full range of ray offsets with three internal reflections, I get this:
In contrast to my graphs for the primary and secondary, this one shows the deflection from the “solar point”—the position of the sun. The maximum occurs at an offset of about 0.97, reaching 42.5º for red light, and 37.7º for violet. The angular distance between red and violet is therefore about two-and-a-half times what we see in the primary bow. The light transmission is scaled to match my previous two graphs, and it shows that, because so much light is lost during multiple internal reflections, only about 1% of the tertiary rainbow ray survives to exit the drop. (But that’s equivalent to about 24% of the transmission for the primary rainbow ray, so not catastrophically dim.)
So it seems straightforward enough. We should see a broad, faint tertiary bow, about the same diameter as the primary but centred on the sun. Have you seen that? No, me neither.
The problem is that there’s a lot of light in the sky around the sun, particularly when rain is falling through the line of sight. Les Cowley at Atmospheric Optics calls this the “zero-order glow”, because it is formed from sunlight that passes through the raindrop without being reflected. So there’s always a directly transmitted, zero-order light ray parallel to the tertiary rainbow ray, like this:Much more light pours through the raindrop in the zero-order glow than survives through three internal reflections. This makes the tertiary rainbow very difficult to see.
But not impossible. There have been sporadic reports over the last few decades, by careful observers who knew what they were looking for. And finally, in 2011, a paper* appeared in the “Light And Color In The Open Air” edition of Applied Optics, entitled “Photographic evidence for the third-order rainbow”. The authors describe taking a photograph under favourable conditions (the sun obscured, a dark cloud in the region that would be occupied by the tertiary bow). The photographer could barely discern a hint of the tertiary rainbow—“only a faint trace of it at the limit of visibility for about 30 seconds”. But after a bit of image processing, a rainbow arc appeared in the resulting photograph. And, after careful analysis, the authors confirmed they had taken the first known photograph of a tertiary rainbow.
You’ll have discerned a pattern. Each successive rainbow (primary, secondary, tertiary) is produced by a rainbow ray which is increasingly offset from the centre of the drop, undergoing one more reflection that its predecessor. The increasing offset means that refraction is greater, with a larger difference between red and violet rays, leading to a broader rainbow. Each additional reflection along the light path means more light lost, and so a fainter rainbow.
The rosette shows the first twenty rainbows, produced by the first twenty internal reflections of light that enters the upper half of a water droplet. Light bounces around the drop clockwise, and the difference in deflection for successive rainbows quite quickly converges to settle at a little less than a right angle. So you can see that the primary (labelled “1” in the lower left quadrant) is followed by the secondary in the upper left quadrant, the tertiary in the upper right quadrant, and the quaternary in the lower right. The quinary rainbow (number five) brings the progression almost full circle, and appear just a little anticlockwise of the primary. I won’t bore you with the names for higher orders of reflection†, but you should be able to pick out how they go around again, and again, and again, becoming successively wider and fainter.
Because violet is refracted more than red, all the rainbows have the same layout, with violet always lying clockwise of red.
Rainbows in the lower left quadrant are being reflected downwards, and back towards the light source. So an observer would look for them as circles around the antisolar point. Because the red light is descending more steeply than the violet, red will appear on the outer edges of all these rainbows. Rainbows in the lower right quadrant are formed from light that has been reflected downwards and away from the light source—so the observer must look towards the sun to see them. In this position, violet light is descending more steeply than red light, so all these rainbows have violet on their outer edges.
Light in the top half of the diagram is all spraying upwards—these rainbows would not be visible to an observer standing below the drop. But we’ve neglected the light that enters the lower half of the raindrop, and bounces anticlockwise. It produces its own rosette of rainbows, identical to the one above, except mirrored in the horizontal plane, like this:
The reversed light path means that violet always lies anticlockwise of red for this family of rainbows. So antisolar rainbows with this light path have violet as their outer colour, and solar rainbows have red outermost.
So the rainbows we see in the sky come from a mixture of the two sets of possible light paths in the two diagrams above. So let’s stack them together, and switch from a view centred on the water drop to a view centred on the observer:
The sky is full of overlapping rainbows! You should think of the upper and lower “copies” of each rainbow as being linked by a vertical circular arc sticking out at right angles to your screen, which the observer sees as a (potential) circular rainbow.
My diagram has the sun directly behind the observer, in which case the rainbows in the lower part of the diagram would be invisible under normal circumstances, superimposed on the ground, and each rainbow would form a semicircular arc against the sky. But the sun is usually some distance above the horizon—as it climbs higher, it pushes the antisolar rainbows lower in the sky, but carries the solar rainbows higher. So antisolar rainbows generally form less than a semicircular arc, and when the sun is higher above the horizon than the radius of the rainbow they will drop entirely below the horizon. But solar rainbows will be lifted above the horizon by the sun, forming more than a semicircular arc, and at the extreme when the sun is higher above the horizon than the radius of the rainbow they can form complete circles. (Complete circles are also possible for antisolar rainbows, but only when the observer is looking down on water droplets suspended below the horizon, as from a plane flying over clouds.)
Are any of them visible to the naked eye? We know that the tertiary has been spotted very rarely. The quaternary sits right next to it, farther out in the zero-order glow, but additional light losses mean its rainbow ray transmission is just 15% of the primary. I don’t know of anyone who has seen it, but it has been photographed. In fact, the article‡ entitled “Photographic observation of a natural fourth-order rainbow” appeared in the same themed edition of Applied Optics as the tertiary rainbow report. (Early reports of the tertiary photographs had inspired the author to search out the quaternary using image stacking.)
The quinary bow has also been photographed. With rainbow-ray transmission sitting at just 10% of the primary, it is also partially obscured by the brighter secondary bow. But its green, blue and violet portion sit within Alexander’s Dark Band, giving reasonable hope of detecting those colours. And Harald Edens managed to (just) pick out the green stripe of the quinary rainbow in a photograph taken in 2012.
The sixth-order rainbow sits within the bright sky at the inner edge of the primary bow, and seems like a poor candidate for detection. The seventh is better separated from the zero-order glow than the third and fourth, but its rainbow ray transmission is only 6% of the primary. However, it seems likely that some of the higher order bows will yield to the sort of photographic techniques commonly employed in astrophotography these days. Watch this space. Meanwhile, for your delectation, I’ve appended some basic descriptive data for the first twenty rainbows, as featured in my diagrams above.
* Großmann M, Schmidt E, Haußmann A. Applied Optics50(28): F134-41 † Oh, alright, I will. Beyond quinary, five, the sequence goes senary, septenary, octonary, nonary, denary, undenary, duodenary … at which point we need to start making up names until we get to vigenary, twenty. ‡ Theusner M. Applied Optics50(28): F129-F133
Note: All values, here and in previous posts, are calculated using a refractive index for red light (wavelength 700nm) of 1.33141, and for violet light (wavelength 400nm) of 1.34451. See the page dealing with refractive index on Philip Laven’s excellent website, The Optics Of A Water Drop, for further information and references.
The First Twenty Rainbows
Outer Radius (º)
Transmission (% Primary)
* The order-11 and order-13 rainbows lie predominantly in the solar hemisphere, but extend slightly into the antisolar †The order-12 rainbow spans the solar point, and therefore overlaps itself—a small disc of blue-violet rainbow (radius 6.1º) is superimposed on a larger disc of green-yellow-red rainbow (radius 10.1º)
In my previous post about rainbows, I described how the light of the rainbow was reflected back to our eyes by falling water droplets. For a raindrop at the top of the rainbow arc, light follows a path that enters near the top of the raindrop, bounces off the back, and then exits from the bottom:
The angle between the incoming light ray and the reflected ray ranges from 42.4º for red light to 40.5º for violet light. All other light rays, entering the drop either closer to its centre or closer to its edge, are reflected back at smaller angles—and it’s their smeared and superimposed light which accounts for the white glow visible within the arc of the rainbow, above. I used the name “Offset” for the parameter that measures how close to central the incoming light ray hits the droplet. It’s measured like this:
And by plotting the deflection of light rays at various offsets, along with light transmission along the reflected pathway, I showed how the rainbow forms at the point of maximum deflection, corresponding to an offset of about 0.86:
I called the ray that follows this maximum-deflection route the “rainbow ray”. (See my previous post for much more detail.)
I also produced a little diagram of how light is lost from the rainbow ray each time it encounters a surface at which it is either reflected or transmitted, like this:
So the rainbow ray arrives at your eye containing only about 4.5% of the light that entered the water drop.
What I want to talk about this time is the fate of the light that undergoes a second internal reflection (labelled “0.5%” in the diagram above).
It’s possible for light from this second reflection to exit the drop when it next encounters the water-air interface, like this:
The second reflection takes the light to the front of the raindrop, and it exits on an upward course, crossing the path of the incoming ray. For this ray to reach the eyes of an observer looking towards the antisolar point (the centre of the rainbow arc), we have to flip the geometry upside-down, like this:
So now the incoming light enters near the bottom of the raindrop, and the reflected light is deflected downwards, towards an observer on the ground.
This light is the source of the secondary rainbow, which is larger and fainter than the primary rainbow I’ve been describing so far. (A secondary rainbow is dimly and partially visible in the photograph at the head of this post.)
Like the singly reflected light that forms the primary bow, the doubly reflected light that forms the secondary bow has its own characteristic angle of deflection, but this time (because of the flipped light-path described above) it’s the minimum angle of deflection from the antisolar point where light is concentrated to form the secondary rainbow ray:
I’ve kept the scale of the transmission curve the same, to allow comparison with that of the primary rainbow. And the range of the angle-of-deflection axis is the same, but it spans from 45º to 90º this time, rather than the 0º to 45º of the primary plot.
The minimum deflection occurs at an offset of about 0.95. The deflection is 50.4º for red light, and 53.8º for violet.
So there are several things going on with this secondary rainbow. Firstly, because the light enters closer to the edge of the raindrop, the refraction is greater, and that causes the separation between red and violet light to be greater—so the secondary rainbow has a width of 3.4º, compared to just 1.9º for the primary rainbow. Secondly, the fact the light-path is flipped over compared to the primary reverses the colour sequence—red is on the inside of the secondary bow, but on the outside of the primary. Thirdly, the additional reflection means more light is lost from the rainbow ray as it passes through the drop. This is partially offset by the fact that the rainbow-ray offset is closer to the offset of maximum light transmission—so the light transmitted into the secondary bow works out to be about 43% of that transmitted into the primary. Finally, because the rainbow ray occurs at a minimum deflection from the antisolar point, the rest of the sunlight entering the drop is reflected to greater angles than the rainbow ray—it lights up the sky outside the secondary bow.
So if we imagine a raindrop falling vertically towards the antisolar point, it at first sends a doubly reflected mixture of light, appearing white, towards the observer’s eye. When it has fallen to 53.8º from the antisolar point, it sends a relatively pure, doubly reflected violet light to the observer—the top of the secondary bow. As it falls from 53.8º to 50.4º, it reflects all the spectral colours in sequence, ending with red light at the inner rim of the secondary bow. Then, it quite literally goes dark. It is too low to send any doubly reflected light in the observer’s direction, but too high to send any singly reflected light. The region of sky between the secondary and primary bow is therefore noticeably darker than either the region outside the secondary, or inside the primary.
This dark region is called Alexander’s Dark Band*, in honour of the Greek philosopher Alexander of Aphrodisias, who described it in about 200AD.
After passing through the Dark Band, the raindrop lights up with singly reflected red wavelengths when it reaches 42.4º, and then runs through the spectral colours until it reaches violet at 40.5º. Below that point, it reflects a mixture of wavelengths (white light again) until it hits the ground.
The whole sequence looks like this:
The obvious question now is this: if the primary rainbow forms from singly reflected light, and the secondary from doubly reflected light, what happens to triply reflected light? Is there a tertiary rainbow? And if so, where is it?
The COVID-19 lockdown, in my part of the world, has produced an outpouring of children’s rainbow art—often stuck up in people’s windows, but sometimes sketched on the pavements, too.
I’ve been struck by the generally good command of spectral colours on display, with red on the outside and an appropriate progression towards violet on the inside. I was amused by the one above, which is a pretty flawless piece of artistry, undermined only by the position of the sun.
It reminded me that I had been planning to write about rainbow colours for years, ever since I wrote about converging rainbows back in 2015.
That was when I posted this diagram, showing the relationship between the sun’s position and the rainbow arc:
The rainbow is a complete circle of coloured light, centred on the antisolar point—the direction exactly opposite the position of the sun, which is marked by the shadow of your head. We usually only see an upper arc, because the area below the horizon usually contains too few raindrops along the line of sight to generate bright colours.
Every raindrop 42½º away from the antisolar point reflects red light towards our eyes; every raindrop at 40½º reflects violet light in the same way. And between those extremes, raindrops reflect all the spectral colours in turn, changing their apparent colour as they fall. But quite why that happens is a little complicated, and that’s what I want to write about this time.
Here’s the route that a ray of green light takes when it passes through a rainbow-forming raindrop and bounces back towards our eyes. Let’s call this particular trajectory the “rainbow ray” for short:
If the raindrop is falling through the top of the rainbow arc, the light enters near the top of the drop, is refracted as it crosses the air-water interface, then reflected from the back of the drop, then exits through the bottom of the drop, being refracted again as it moves from water back to air. The angle between the incoming ray and the outgoing is about 41½º for green light. At the sides of the rainbow, you have to imagine the diagram above lying horizontal—the ray enters the outward-facing side of the drop, and is reflected towards you sideways. And, obviously, the rest of the rainbow is formed by light paths that are more or less tilted between the horizontal and the vertical.
Violet light is refracted more than green light, which is in turn refracted more than red light. So a ray of white light (drawn in black below) is split into a fan of coloured rays as it enters the raindrop. These follow slightly different courses within the drop, and exit at different angles (exaggerated here for clarity):
So we see red at 42½º, and violet at 40½º, with green between. Simple.
But why choose to consider just those rainbow rays? What about all the light that enters the drop closer to its rim, or closer to its centre? I’ll call this general group of light rays “reflected rays”, of which the rainbow ray is only one example.
If I plot a couple of examples, you can see that rays entering the drop farther out than the rainbow ray are refracted more strongly, and end up exiting the drop at an angle less than the rainbow ray. Those that enter the drop nearer the centre are refracted less, but bounce back at a narrower angle from the back of the drop, and also exit the drop at an angle less than the rainbow ray. So for a given colour of light, it turns out that the rainbow ray is the light path that results in the maximum deflection angle from the antisolar point. All other reflected rays exit at narrower angles, and so should appear within the visible coloured arc of the rainbow itself.
How does that help, though? Why is the rainbow ray’s role as the maximum angle of deflection important?
To show why, I’m going to make a plot of what happens to all the reflected rays, using a parameter I’ll call Offset*, which works like this:
It’s just the proportional distance from the centre of the raindrop at which the ray enters—a zero offset means that the ray spears straight into the centre of the drop; an offset of one means the ray just grazes the edge of the drop.
So here’s how red and violet reflected rays are deflected, for the full range of offsets:
Deflection peaks at an offset of about 0.86, the location of the rainbow ray, with lesser deflection occurring on either side, as shown in the diagram above. Red and violet are at their maximum separation at the peak, and the rounded peak of the curves means that a lot of rays close to the rainbow ray end up reflected in the same part of the sky as the rainbow ray. You can see from the chart that all the rays with offsets from 0.75 to 0.95 end up within two degrees of the rainbow ray; a similar span from 0.2 to 0.4 is spread over fifteen degrees.
So, in the vicinity of the rainbow ray, there’s a lot of light in a small area of sky, and the spectral colours are well separated. Farther from the rainbow ray, the deflected light is smeared over a large area of sky within the rainbow arc, and the spectral colours are not well separated—all these other rays average out to a patch of white light filling the curve of the rainbow, with no colour separation. This bright area within the rainbow can often be strikingly visible, if the rainbow has dark clouds behind it.
One other thing contributes to the colour intensity of the rainbow—oddly, that’s that fact that some light is lost from the reflected rays every time they interact with an air/water or water/air interface. Here’s a diagram of how much light goes missing from the green rainbow ray as it passes through the raindrop:
The large proportion of light that shoots straight out the back of the drop, doubly refracted but without being internally reflected, creates a bright patch around the sun that appears whenever the solar disc is viewed through falling rain. Les Cowley at the excellent Atmospheric Opticssite has dubbed this the “zero-order glow”.
Interestingly, for a given ray each interaction with an interface results in exactly the same ratio of reflection to transmission (though not quite in my diagram, which features rounded figures). This is unexpected (at least to me), because the reflective properties of a water/air interface are generally different from those of an air/water interface; the former features the phenomenon of total internal reflection, for instance. But it turns out that the first passage through the air/water interface changes the angle of the ray just enough to make it interact with subsequent water/air interfaces in exactly the same way as its initial air/water encounter.
If I plot the final amount of transmitted light for all the different offset rays, and add it to my previous graph, it looks like this:
The transmission data are in brown, and refer to the new, brown axis on the right side of the chart. You can see that transmission starts to ramp up just as we get into the vicinity of the rainbow ray, boosting the brightness in the rainbow’s part of the sky. The peak of transmission does occur at very high offsets, beyond the rainbow ray, but in that region the angle of deflection changes very rapidly with slight changes in offset, which diffuses that light over a large arc.
The calculations I did to produce the transmission graph above involved Fresnel’s equations, so I had to track two different polarizations of light independently. For light reflecting from a surface between two transparent mediums, there’s a critical angle of incidence called Brewster’s angle, at which the reflected light becomes totally polarized. At that angle, the reflected light is entirely s-polarized; light polarized at right angles to this (p-polarized) is completely transmitted through the reflective surface. (Your polarizing sunglasses are designed to filter out s-polarized light reflected from horizontal surfaces, to reduce glare.)
The Brewster angle for an air/water interface is around 53º; for water/air it is about 37º. And it turns out that any light entering the water at an angle of incidence of 53º has its angle changed to 37º by refraction. So in the case of our raindrop, a ray that strikes the surface of the drop at 53º (corresponding to an offset of about 0.8) will continue through the drop and strike the water/air interface at 37º—it hits two Brewster angles in succession! This means that p-polarized light that hits the drop at Brewster’s angle is entirely refracted into the drop—none of it escapes by reflection from the air/water interface. But then it hits the back of the drop, and now none of it is re-reflected—it is all transmitted, again. So at an offset of 0.8, no p-polarized light gets into the reflected ray—it is all lost out the back of the raindrop.
So now if I mark up the total amount of transmitted light with its s- and p-polarized components, you can see that the light making up the rainbow will be strongly s-polarized, because the rainbow rays are pretty close to Brewster’s angle:
The resulting polarization follows the curve of the rainbow. Your polarizing sunglasses will largely block the light coming from the top curve of the rainbow, but will let light through from the sides. However, if you tilt your head, you’ll remove light from the sides of the rainbow, and bring the upper curve into view.
Here’s a nice short little YouTube video, by James Sheils, demonstrating how to make the lower curve of a rainbow appear and disappear using a polarizing filter:
And that’s it, for now. Some time in the future I’ll get around to discussing the secondary rainbow.
* The measure I’ve called Offset is sometimes called the “impact parameter”, a term borrowed from nuclear physics. While the analogy is strong, if you know its original application, I’m not sure the phrase itself helps with visualization, so I’m sticking with Offset in this and subsequent posts.
In short, taking every thing into consideration, the British empire in power and strength may be stated as the greatest that ever existed on earth, as it far surpasses them all in knowledge, moral character, and worth. On her dominions the sun never sets. Before his evening rays leave the spires of Quebec, his morning beams have shone three hours on Port Jackson, and, while sinking from the waters of Lake Superior, his eye opens upon the mouth of the Ganges.
Caledonian Mercury, 15 October 1821, page 4: “The British Empire”
It’s noticeable, when reading the above, that none of the places it mentions by name still belong to the United Kingdom. The British empire is now much reduced in size; in fact, its overseas possessions are confined to a scatter of places that few people could reliably place on a map:
Overseas Territories ● Anguilla ● Bermuda ● British Virgin Islands ● Cayman Islands ● Falkland Islands ● Gibraltar ● Montserrat ● Pitcairn Islands ● Saint Helena (with Ascension & Tristan da Cunha) ● Turks and Caicos Islands ● British Indian Ocean Territory ● South Georgia and South Sandwich Islands ● British Antarctic Territory (in abeyance under Antarctic Treaty)
Dependent territory ● Sovereign Base Areas of Dhekelia & Akrotiri (Cyprus)
If you’re one of the people who would have trouble placing these names on a map, here’s a map:
What stands out from the map above is that the UK still has the Atlantic, Caribbean and Mediterranean pretty well covered. There’s a solitary (and I do mean solitary) British possession in the Pacific, the Pitcairn Island group. (I’ve written about Pitcairn and its neighbouring islands a couple of years ago, when we were lucky enough to visit them.) And there’s another single possession in the Indian Ocean, the catchily named British Indian Ocean Territory (BIOT). BIOT occupies the whole of the Chagos Archipelago, and is inhabited entirely by British and American military personnel and contractors, based on the largest island, Diego Garcia. It used to be home to 2000 Chagossians, who were chucked out around 1970 to make way for the UK/US military installations. The poor Chagossians are still grinding through the courts attempting to get their homeland returned to them.
Anyway. Pitcairn and BIOT, which are a long way west and east of most UK territories, look like the key locations to examine when it comes to deciding whether the sun still “never sets on the British empire”. With Pitcairn’s time zone of GMT-8, and BIOT’s of GMT+6, there’s only ten hours of difference between the two territories, which should mean that the sun is visible from both locations for a couple of hours a day. But there’s a potential problem with the seasonal variation in day length—while BIOT sits close to the equator and won’t have much variation in the times at which the sun rises and sets, Pitcairn is south of the tropics, and so we can expect its sunsets to be noticeably earlier in June than they are in December.
So we’re going to need to plot daylight charts for the whole year. Here’s one for Greenwich:
Along the x-axis we have the months of the year, numbered from 1 to 12. On the y-axis, Greenwich Mean Time. The lower curve marks the time of sunrise, throughout the year, at Greenwich. The upper curve is sunset. The yellow area between the curves therefore represents the totality of daylight seen in Greenwich throughout the course of a year.
OK. Let’s superimpose the sunrise and sunset curves for Adamstown on Pitcairn, giving times in GMT:
The Pitcairn sunrise and sunset curves are in red, and Pitcairn daylight extends a long way through the Greenwich night. But sunset on Pitcairn always occurs before sunrise in Greenwich, so there’s a brief period when the sun is shining in neither location.
Will BIOT, with its sunrise earlier than Greenwich, fill the gap? Here are the BIOT curves (calculated for Diego Garcia) added in green:
It’s a close-run thing. Pitcairn’s midwinter sunset on 21 June 2020 comes just 38 minutes after BIOT’s sunrise. Here’s a south polar view of the Earth on that date, capturing the brief period when both territories are in sunlight:
But there’s no doubt the chart is full of daylight, and the sun still never sets on the British empire!
I had a photograph of my own to illustrate this post, but it was a bit rubbish. I was inspired to write about helium when I discovered the wreckage of a mylar-foil helium balloon, like the one pictured above, tangled in a gorse bush on the slopes of Newtyle Hill. It’s the second foil balloon I’ve discovered on the hill, and (like the first one) I stuffed it into my rucksack and carried it down for disposal. I took a photograph to illustrate what a non-biodegradable blot on the landscape these things are, but in the photo the balloon looked like just another bit of plastic debris.
The picture above is actually more useful, because it demonstrates the key fact about helium gas, the one thing that pretty much everyone knows about it, and the property from which many of its other interesting qualities derive—it’s lighter than air.
The reason it’s lighter than air is because its atoms are considerably less massive than the molecules that make up air. Helium is a monatomic gas, made up of individual atoms, and the mass of a single helium atom is about four daltons.* (For comparison, the mass of a common carbon atom is 12 daltons, and the commonest kind of hydrogen atom weighs in at around one dalton.) Air, on the other hand, is mainly composed of two diatomic gases, nitrogen and oxygen. Their molecules, N2 and O2, come in at a 28 and 32 daltons, respectively, giving air an average molecular mass of 29 daltons.
The fact that individual helium atoms have a low mass feeds into two other important properties of helium.
Firstly, its atoms are small—just a single electron shell containing two electrons. A small atom with tightly bound electrons is reluctant to redistribute its charge in response to nearby polar molecules. This means that its relatively immune to the intermolecular Van der Waals forces which cause atoms and molecules to transiently adhere to each other, which in turn means that helium gas isn’t very soluble.
Secondly, at any given temperature the atoms in helium gas move faster, on average, than the atoms or molecules of heavier gases. This is because temperature is a measure of the kinetic energy of gas particles, and kinetic energy scales with both velocity squared and mass. A low mass means velocity must be higher to produce the same kinetic energy. Since helium is only 4/29 the mass of an average air molecule, the mean velocity of its atoms is correspondingly higher by the square root of 29/4, or about 2.7.
So: helium is light, fast and not very soluble. I’ll come back to each of these as we go along.
Firstly, lightness. It turns out that, at equal temperature and pressure, equal volumes of different gases contain the same number of fundamental particles (to a good first approximation). So a litre of helium is only 4/29 the weight of a litre of air. The only less dense gas is hydrogen, which has diatomic molecules massing about two daltons. So both hydrogen and helium are so buoyant in air that they’re able to lift considerable additional mass as they rise—making them ideal fillers for balloons, large and small. Hydrogen, being half the mass of helium, is by far the better lifting agent, but it has one significant disadvantage:
That’s a photograph of the German dirigible “Hindenburg”, fatally aflame at Lakehurst, New Jersey, in 1937. Hydrogen is flammable; helium is not. In fact, helium is notoriously chemically unreactive, being the lightest of the so-called “noble gases” (the others are neon, argon, krypton, xenon and radon). All of these elements have full outer electron shells, rendering them almost completely chemically inert. Which is why modern balloons and dirigibles are filled with helium, not hydrogen.
Next, speed. The faster gas molecules move, the more readily they diffuse through a barrier—which is why a rubber balloon full of helium will lose its shape within a day, and why helium balloons are often made of less-permeable mylar foil, like the one in the photograph at the head of this post. (Because they’re not biodegradable, foil balloons are supposed to be used only indoors—my experience of finding two on the open hillside shows how well that rule is working in practice.)
The rapid movement of helium gas atoms also affects the speed of sound, because sound waves travel through a gas at a velocity roughly comparable to the average speed of the gas molecules. At 0ºC, the speed of sound in air is about 330m/s; for helium it’s 970m/s, almost three times faster. So if you have a resonant cavity full of helium, it will resonate at a frequency about three times higher than it would if filled with air. And that’s what causes the “duck voice” effect we hear when someone breathes a gas mixture containing helium. Their vocal cords vibrate at exactly the same frequency as usual—but the resonant gas cavities of their larynx and airways pick out and emphasize the higher-pitched harmonics of their voice.
Some people achieve this effect by taking a breath from a helium-filled party balloon, which is very much not a good idea, since it violates The Oikofuge’s First Law:
Never breathe anything that contains no oxygen
Breathing gas that contains no oxygen causes oxygen to leave your circulation and diffuse into the gas in your lungs—your circulating oxygen levels therefore fall very rapidly indeed, and a single deep breath can take you to the edge of unconsciousness.
To illustrate the duck-voice effect of someone breathing helium, here’s a recording of a saturation diver, breathing a helium/oxygen mixture in a pressurized underwater habitat:
Which leads us to wonder why deep divers breathe helium and oxygen (a mixture referred to as Heliox), rather than air.
The ambient pressure rises with depth underwater, by about one atmosphere for every ten metres of descent. To counterbalance this, divers must breathe gas at the ambient pressure. But the higher the pressure of gas we breathe, the more of it dissolves in our tissues—and it turns out nitrogen is an anaesthetic agent at high pressures. Its effects are detectable at depths as shallow as ten metres, where the pressure is twice that at the surface. And by the time divers descends to 30 or 40 metres (four or five atmospheres), their judgement becomes sufficiently impaired by nitrogen narcosis that they’re a potential danger to themselves and others.
So for deep diving, nitrogen has to go. But it can’t be replaced by pure oxygen, because oxygen is toxic at higher-than-normal pressures, damaging the lungs and causing convulsions. Indeed, the need to keep the partial pressure of oxygen in the breathing mixture close to what we’re used to at the surface means that, with increasing depth and pressure, oxygen must make up a lower and lower percentage of the breathing mixture by volume.
Helium is a good replacement for nitrogen, for several reasons. Firstly, its low solubility and chemical inertness mean that it doesn’t produce any anaesthetic effect. Secondly, because helium is less soluble than nitrogen, less of it dissolves in the diver’s tissues during a long dive at high ambient pressure, so there’s less of it to get rid of during decompression at the end of the dive, and therefore less risk of gas-bubble formation in the blood and tissues as ambient pressure decreases. Such bubbles are the cause of decompression sickness (“the bends”), and in order to avoid their formation, divers are forced to make their return to the surface slowly. But because helium dissolves in smaller volumes than an equivalent pressure of nitrogen, there’s less risk of bubble formation, and so a faster safe decompression.† And finally, the low density of helium comes into play again—because it’s less dense, it’s easier to breathe at high pressures.
Indeed, that last advantage is present even at one atmosphere of pressure. When a person’s airways are narrowed by disease or inflammation, air flow through the narrowed regions can shift from smooth, laminar flow to turbulent flow, which produces a higher resistance to flow through the airways and makes breathing more difficult. The transition from laminar to turbulent flow is determined, in part, by the density of the breathing gas. And, once turbulent flow occurs, the resistance to flow is higher for a denser gas. Substituting helium for nitrogen in the patient’s breathing gas drops its density by 60%, which delays the onset of turbulent flow, and causes less resistance to flow if turbulence occurs. That serves to reduce the work of breathing, decrease distress, and get a bit more oxygen into the patient—which is all good stuff.
So are there any disadvantages for divers breathing helium (apart from the funny voices)? There are. One is caused by that high average velocity of helium atoms—as well as conducting sound faster, helium is also more conductive of heat, with a thermal conductivity almost six times faster than nitrogen. Divers in a helium atmosphere find it harder to stay warm, and when submerged they lose heat to the water very quickly if they have helium filling their dry-suits. (So they often fill the insulating space in their dry-suits with argon, which has an even lower thermal conductivity than air.)
And finally, it turns out that the absence of anaesthetic effects with helium is actually a disadvantage for the deepest of dives. Below depths of about 150-300 metres (fifteen to thirty atmospheres of pressure), divers breathing Heliox develop a condition called High Pressure Nervous Syndrome (HPNS), associated with an apparent overactivity of the nervous system—tremors, muscle jerks, nausea, dizziness and cognitive impairment. No-one’s quite sure why this happens—it was at first blamed on a stimulant effect of helium that appeared only at high pressure, but it now seems more likely that it’s a direct pressure effect on nerve cell membranes, which are reduced in volume by such high ambient pressures. Ironically, the symptoms of HPNS can be damped down by introducing the sedative effects of nitrogen back into the mix, using a breathing mixture of nitrogen, helium and oxygen generically referred to as Trimix. Things get very technical at that point—not only must the ratio of helium and nitrogen be adjusted to minimize the effects of HPNS, but the proportion of oxygen in the mixture must be reduced with increasing depth, in order to limit the pressure of oxygen to a non-toxic level.
But there’s a problem with Trimix, which is that nitrogen at high pressures is difficult to breathe because of its density. What low-density gas could we substitute for nitrogen? Hydrogen, half the density of helium and a fourteenth as dense as nitrogen, turns out to be mildly anaesthetic at high pressures, and therefore it also limits the symptoms of HPNS.
But wait a minute, I hear you cry, glancing back at that photograph of the Hindenburg. Hydrogen is flammable. Can a breathing mixture containing hydrogen and oxygen be safe?
Well, yes it can. Remember that we have to wind down the proportion of oxygen in the breathing mixture as we go to greater depths, to keep the partial pressure of oxygen within safe limits. At thirty atmospheres pressure, a gas mixture containing just 1% oxygen provides an oxygen pressure equivalent to 30% oxygen at sea level—a little more than the 21% we’re used to, but within safe limits. Hydrogen/oxygen mixtures are flammable over a wide range of proportions—from 4% hydrogen in 96% oxygen to 95% hydrogen in 5% oxygen. But not at lower proportions of oxygen. So the low proportion of oxygen required for safety at great depth means that the hydrogen/oxygen ratio sits outside the flammable range. These Hydreliox mixtures are very experimental, but they’ve been used successfully, with 1% oxygen and roughly equal proportions of helium and hydrogen, at depths in excess of 500 metres.
And that’s about it for helium gas. Liquid helium, of course, has all sorts of interesting properties, but that’s perhaps a topic for another day.
* The dalton, also called the Atomic Mass Unit, is named after John Dalton, who first codified the idea that chemistry was due to atoms interacting with each other in a very systematic way.
† There’s an important distinction here, though. The advantages of helium’s lower solubility only appear in what’s called “saturation diving”—when divers stay at depth in pressurized habitats for long periods, so that their tissues become saturated with dissolved breathing gas. But divers who descend and then reascend relatively quickly (called “bounce diving”) are never at depth long enough for their tissues to become saturated with nitrogen. For them, helium paradoxically produces a worse risk of decompression sickness than nitrogen, because helium diffuses so much faster than nitrogen. The volume dissolved in the tissues rises very quickly initially, and in the short term may exceed what would be reached by nitrogen in the same time period. Like this:
The year 2020, newly begun as this post is published, is a leap year. I’ve written before about leap years, and how the occasional leap day added to the end of February keeps our calendar year synchronized with the seasons. For more on that topic, see my posts about February 30th and the Equinox.
But this year we are also fairly likely to observe a leap second. (I’ll come back later to the reason for that “fairly likely”.) A leap second is an additional second which will be added to either June 30th or December 31st, and it serves to keep our clocks synchronized with the rotation of the Earth.
The fundamental problem is that the Earth’s rotation is getting slower, primarily because the tidal bulges raised in the Earth’s oceans by the moon and sun generate friction in the ocean beds as the Earth rotates. Over the last couple of millennia the rate of slowing has averaged about 1.7 milliseconds per day, per century. Which sounds trivial, but it adds up to more than three hours over the last two thousand years. We can detect this problem if we look back at early records of astronomical events, particularly solar eclipses, which are visible from only a very limited region of the Earth’s surface. We know, for instance, that a solar eclipse was observed in Babylon early in the morning of 15 April, 136BC. But if we calculate back to the relative positions of the sun, Earth and moon on that date, and assume the Earth has rotated at a constant rate during the intervening centuries, we find the eclipse shadow sweeping through the Atlas Mountains of Morocco, 48.8º of longitude west of Babylon. The difference in longitude represents a three-hour lag in rotation—the naive calculation, ignoring the tidal slowing of the Earth’s rotation, has allowed Babylon to rotate out from under the eclipse track. That mismatch is one of the ways we know about the long-term slowing of the Earth’s rotation.
Here and now, we have three important ways by which we measure the passage of time. The first, and most important in everyday life, is by the rotation of the Earth. We define local noon as being the time at which the sun reaches its highest point in the sky, and we define a solar day as being the time between successive noons. Well, sort of. Because the Earth’s orbit is elliptical, and the Earth’s axis is tilted relative to its orbit, the elapsed time between successive noons varies during the course of the year. So we average those noon passages over a long time period in order to come up with a definition for the day—specifically, it’s a mean solar day.
This mean solar day is conventionally divided into the familiar hours, minutes and seconds, giving 24×60×60=86400 seconds per day. But you can see that there’s a problem with that, because seconds are of a fixed duration, established and defined as part of the Système International (SI) units in 1960. Whereas we now know that the length of the mean solar day is increasing as the Earth’s rotation slows.
Scientists were aware of this problem when the SI units were being defined, and decided they needed to use some other source, with a more fixed and regular motion, in order to define a constant second. Initially, they resorted to our second important means of time-keeping—the movement of the Earth around the sun. The length of the year is rather closer to being constant than the rotation period of the Earth. So the duration of the second was defined as being 1⁄31,556,925.9747 of a tropical year. (The tropical year is a measure of the passage of the seasons—it’s the year that our calendar strives to approximate with all those leap days.)
So that was fine, then. But not quite, because the tropical year is itself a little variable. So what was adopted as the standard was derived from a very specific (and sort of fictitious) tropical year, based on formulae given in Simon Newcomb’s astronomical opus, The Elements Of The Four Inner Planets And The Fundamental Constants Of Astronomy, published in 1895 and based on astronomical observations made between 1750 and 1892. Specifically, the tropical year on which the SI second was based was something produced by Newcomb’s formulae if you plugged in a precise time and date near the start of 1900. So there was no actual year that corresponded to the value used in the definition. And of course Newcomb, and the observers who provided his data, had never used a constant definition of the second. Their definition was based on a second that was exactly1/86400 of a mean solar day—so seconds, as defined in 1750, were a tiny bit shorter than seconds as defined in 1892. When Newcomb tabulated all these observational data and produced his summary formulae, he effectively averaged out the very slight drift in the value of the second over the observation period. Newcomb’s second, which became the SI second, was only ever exactly 1/86400 of a mean solar day somewhere around the year 1820. So even at the moment of its adoption in 1960, the SI second was slightly adrift from the duration of a mean solar day.
The astronomical definition of the SI second was always a bit unwieldy for general use. Fortunately, there was a third method of measuring time, which had been growing steadily more precise around the time the SI units were introduced—the atomic clock. So in 1967 the definition of the second was transferred to something you could actually measure in the laboratory—the behaviour of a particular kind of clock based on the element caesium. Thenceforth, the SI second was defined as:
… the duration of 9192631770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of the caesium 133 atom.
So now we have a precise and portable definition of the second, which carries over its duration from a previous astronomical definition, based on nineteenth-century observations. This is the basis for a standard called International Atomic Time (abbreviated TAI, for Temps Atomique International), which is based on pooled readings from multiple atomic clocks around the world.
But the Earth’s rotation is steadily lagging behind TAI. So to keep our everyday clocks in synchrony with the slowly lengthening mean solar day, we use a different time scale called Coordinated Universal Time (confusingly abbreviated UTC), which is the basis for Greenwich Mean Time and all the various time zones around the world.
UTC uses the same SI seconds as TAI, but every now and then needs to pause for a moment to allow the Earth’s rotation to “catch up”. Which is where the leap seconds come in—when required, at the end of a month, we add an extra second just before midnight, Greenwich Mean Time. A clock that displays such leap seconds reads 23:59:60 (as at the head of this post) before cycling through to 00:00:00. This leap second is added everywhere, simultaneously, so it occurs in the afternoon or evening in the Americas, but in the morning in Asia and Australia.
At present, the Earth’s rotation is slowing at about 1.4 milliseconds per day, per century (a little slower than the millennial trend I mentioned earlier). Since two centuries have elapsed since the second was exactly 1/86400 of the mean solar day, the day should now be 86400.0028 seconds long, which corresponds to almost exactly one extra second per year.
So why don’t we just schedule an extra second for every December 31st and have done with it? Because the Earth’s rotation rate varies irregularly, from day to day and year to year, around the long term mean rate of slowing. This happens because stuff (air, water, rock) is always moving around, sometimes shifting closer to the Earth’s axis of rotation, and sometimes farther away. If mass moves closer to the axis, the Earth speeds up a little; if mass moves away from the axis, the Earth slows down—this is caused by the same conservation of angular momentum that allows figure skaters and acrobatic divers to modify their rate of rotation by drawing in or spreading their arms. So earthquakes, glaciers melting and seasonal shifts in the air mass all contribute to the variability of the Earth’s rotation rate.
In the early days of the leap second (which was introduced in 1972) we did indeed have a leap second every year. But the Earth’s rotation rate has actually bucked the trend and speeded up a little of late, so leap seconds have become more sporadic—we had no leap seconds at all between 1999 and 2004, and the most recent was in 2016. The aim of the leap second is to keep UTC correct to within 0.9 seconds of the mean solar day, and the situation is constantly reviewed by the International Earth Rotation and Reference Systems Service, which issues a six-monthly Bulletin C, declaring or omitting a leap second at the end of each six-month period. Which is why I can’t (at time of writing) say for sure if we’ll have a leap second in 2020.
I’ll add a footnote as soon as I know. Meanwhile, have a Happy New Year.