Wednesday, November 23, 2016

N Rays: A Case Study in Pathological Science

The first decade of the twentieth century was an exciting time in experimental physics. The recent past had seen major discoveries: Röntgen's discovery of X-rays in 1895, Becquerel's discovery of radioactivity in 1896, Thomson's discovery of electrons in 1897, et cetera. Many more discoveries were to come. In 1903, the newest ray was announced by a French physicist named Prosper-René Blondlot: N-rays. If you've never heard of N-rays, there is a simple reason for that: they turned out not to exist.

Prosper-René Blondlot
When Blondlot is remembered at all these days, it is for his "discovery" of N-rays. That's a shame because he seems to have been a respected researcher who did some solid experimental work in physics. In a series of experiments around 1891, Blondlot confirmed that the speed of electrical signals in wires was nearly equal to the speed of light and he was the first to measure the speed of radio waves. This man was not a scientific lightweight by any means, and one would not think he was a man who would be easily deceived in matters of physics. And yet he seems to have deceived himself.

This arouses my curiosity. How could a respectable and competent man of science find himself in such a peculiar rabbit hole? I took a weekend to scour the record to see if I could find some clues, and found a strange mix of the reasonable and the wishful that it wasn't always easy to untangle.

The story of N-rays begin with Blondlot's attempts to determine whether X-rays could be polarized. The rationale of his experiment was reasonable and interesting. In fact, he begins with a supposition that Charles Barkla was to use in 1904 in the experiment that did definitively prove that X-rays could be polarized.  To understand, let's take a moment to review what polarization is.

X-rays - along with radio waves, microwaves, infrared, visible and ultraviolet light, and gamma rays - are electromagnetic waves. In general, electromagnetic waves are produced anytime an electric charge such as an electron accelerates. (We'll ignore quantum physics for the moment) The acceleration of the charge produces an electric field that is changing in a particular direction and that disturbance in the electric field propagates through space (along with its perpendicular magnetic field, thus the term electromagnetic). Notice that that electric field has a particular direction it oscillates - up and down, right and left, etc. Most radiation sources will emit the waves in all the random possible orientations, but under the right circumstances you can have all the electromagnetic waves vibrating in the same direction. This set of waves with a common alignment of their electrical fields are referred to as being polarized.

A sketch of Blondlot's experimental setup is shown below. HH' is the vacuum tube which is the source of the X-rays, and cc' is a spark gap that is set up so that a spark will leap across the tiny gap between the ends. The way that the vacuum tube produces X-rays is that electrons are accelerated at high speed from the cathode H towards the anticathode H'. When the electrons strike the anticathode they shed a lot of kinetic energy as they slow down abruptly in a process called Bremsstrahlung. The energy goes into the X-rays.

Blondlot's idea here has two interesting roots. First, he makes the guess that the X-rays produced by the vacuum tube are already polarized. "I asked myself whether 'X' rays emitted by a focus tube are not polarized as soon as emitted... For each ray is generated from a cathode ray, and the two rays define a plane; thus, through each ray emitted by the tube a plane passes, in which, or normally to which , the ray may well have special properties, this being, in fact, an asymmetry characteristic of polarization." This supposition, as it turns out, is entirely correct, and formed a key part of Barkla's later experiment proving the polarization of X-rays, although the apparatus was otherwise quite different.

The second item of interest is the use of the spark-gap to probe for polarization. From my 21st-Century perspective it it was not immediately obvious to me why Blondlot would even think this would work. It turns out that he was adapting a technique used by Heinrich Hertz to demonstrate the polarization of radio waves years earlier.

With radio waves, the apparatus worked quite well. If the spark was parallel to the electric field in the plane of polarization, a strong spark was produced, but rotate the spark-gap ninety degrees so that it is perpendicular to the plane of polarization and no spark is produced. This is what Blondlot hoped to achieve with X-rays.

But was it reasonable to suppose that what would work for radio waves would also work for X-rays? It was well known that X-rays behaved differently from ordinary visible light. They were not subject to refraction or reflection in the same way as light, and the shadows cast by X-rays (the shadows of one's bones, say) were much sharper than a shadow cast by light. Such leading lights as J. J. Thomson, discoverer of the electron, had already been proposing by 1898 that X-rays were very thin pulses.


This is not far from the truth, since X-rays are very short wavelength radiation and the wavelength plays a large role in determining how the rays interact with matter. It seems to me that Blondlot should have anticipated this as a problem for his spark-gap setup, since the gap must have been huge compared to the "thin pulse" of the radiation, but he does not mention it. He sees, in fact, precisely what he expected to see.

Is this, then, an instance of confirmation bias? It seems likely. Bear in mind that these observations depended on observing the change in brightness of a flickering spark. There was no quantitative data here - no measures of voltage changes or adjusting the size of the gap - simply human visual perception.

So far, this is not so bad. We have a basis for experiment and a result that could be tested more rigorously. Even if his conclusion is flawed, I would argue Blondlot is not actually doing bad science. This is the point where the train really goes off the rails however.

Bizarrely, even though X-rays were known not to refract or reflect in the fashion of visible light, Blondlot checks to see if the plane of polarization can be rotated by passing through sugar or quartz. He concludes that such a rotation does indeed occur, and then decides to check whether the rays can be refracted after all. He determines that such a refraction can indeed be detected, but then makes what seems to me a sudden and unaccountable reversal.

It is as if he suddenly remembered all the previous research on X-rays and decided that his new results must be evidence of something utterly new instead of applying to the Röntgen rays. He claims to have discovered a "new species of light", which he christens N-rays in honor of his hometown of Nancy.

Bear in mind that the only evidence of this discovery is the naked-eye observation of changes in brightness of a flickering spark in a dark room. The physical reasoning underlying the investigation was that the electric field of the X-rays would increase the discharge across the spark-gap if aligned parallel to the spark gap. What followed seemed more an exercise in free-association than an investigation of physics. 

He first replaced the spark by a flame. Then he replaced the flame with an incandescent wire (determining that the N-rays increased its brightness but not its temperature), replaced the incandescent wire with sheets of metal heated to a dull red glow, and replaced these with a sheet of white paper illuminated just enough to stand out as a faint gray blur in a dark room. A luminous clock dial illuminated with N-rays gained just enough brightness that you could make out the circle of its face and almost distinguish the hands. By the end of the investigation, the N-rays aren't even directed at the glowing object but at the eyes looking at the glowing object. And still an increase in brightness was claimed. Weirdly, by this point, he seems to have abandoned any pretense of finding a plausible physical mechanism for the increase in brightness, much less look for alternative explanations like confirmation bias or the unreliability of human perception. He is just examining all these properties of the new phenomenon, with data that is entirely gleaned at the very threshold of visual perception.

Attempts to replicate the results of the N-ray study were widely attempted. 120 scientists published papers on N-rays, some of them were certainly capable and respected researchers and not all of them were French. But the scientists who could not reproduce the N-ray detection included such prominent physicists as Rayleigh, Langevin, Rubens, and Drude. These were not challenges to be taken lightly. Ultimately, the American physicist Robert W. Wood went to Nancy himself to observe the experimentalists in action. He recounted the visit and his conclusions in a letter to the journal Nature in the September 29, 1904 issue.

Robert W. Wood
Wood was first of all unable to detect the changes in brightness with his own eyes. Then he played tricks during the experiments, removing the aluminium prism from the middle of the experimental apparatus, and observing that the readings reported in the darkened room were no different. Finally, he observes of the photographic evidence: "It appears to me that it is quite possible that the difference in the brilliancy of the images is due to a cumulative favoring of the exposure of one of the images, which may be quite unconscious, but may be governed by the previous knowledge of the disposition of the apparatus." Wood concluded, "After spending three hours or more in witnessing various experiments, I am not only unable to report a single observation which appeared to indicate the existence of the rays, but left with a very firm conviction that the few experimenters who have obtained positive results have been in some way deluded."

Some writers have claimed that Blondlot went mad and died after the exposure of N-rays as a figment of the experimentalist's imagination, but he did not. He retired from teaching in 1910 and lived until 1930. He seemed to have spent his years in normal pursuits for an emeritus professor: writing, working on new editions of his books, and a handful of public speeches.

What went so wrong here? How was an experienced and talented experimental physicist so deluded? Many suggestions have been made that have the ring of truth about them: personal ego and ambition, national and regional pride, the strong desire for a novel discovery, overeager assistants. I do think it is very important not to overlook the biggest factor her: science is hard. In a way, you could argue that this was not pathological science so much as science working the way it is supposed to. A discovery is announced, the community rushes to replicate the discovery, the idea is discarded and forgotten when it is seen not to hold up.

I suspect that if we look at other discoveries - the ones that did hold up, like Becquerel's discovery of radioactivity - with the same critical eye, we may be surprised to find that the early days of studying a true phenomenon are not so different from the exploration of the imaginary "discovery". That would be a good subject for another time.

Thursday, December 1, 2011

Dante and Galileo

 ..."For the comprehending, philosophy
Dante Alighieri
Serves in more places than one to demonstrate
How Nature takes her own course from the design
Of the Divine Intelligence and Its art.
Study your Physics well, and you'll be shown
In not too many pages that your art's good
Is to follow Nature insofar as it can,
As a pupil emulates his master; God
Has as it were a grandchild in your art."
- Inferno, XI, 93-101
When I first learned that one of Galileo Galilei's earliest professional achievements was delivering a lecture on the size and structure of Hell as described in Dante's Inferno, I was intrigued to say the least. I had always thought of the Commedia as a work of epic poetry and theological allegory, rich in poetry and symbolism, certainly, but I don't think I had ever considered that it might have been taken seriously in terms of natural philosophy - what we today would call Science. Apparently, however, it was taken rather seriously - seriously enough that a talk on the subject was integral to the early advancement of Galileo's career. I wanted to explore further this intriguing connection between the greatest of the medieval writers and one of the founders of modern science. What did Dante have to say that was of scientific interest and what influence had Dante had on the work and thought of Galileo?
Some time later, I came across this video on the topic, which emphasized Dante's impact on Galileo's thinking about scaling. 
The idea is that the physical structure of an object, whether it is an animal or a bridge, has to depend on its size. As it says in the video, you can't just double the size of a horse to get an elephant. A double-sized horse wouldn't weigh twice as much, but eight (two cubed) times as much, so its legs would have to get thicker to support the additional weight. Changing the size of the structure also requires a change in its shape. So giant Lucifer chewing on Judas, Brutus and Cassius in his three mouths wouldn't be able to stand if he had the shape of a human being. His muscles would collapse and his bones break under his immense weight.
I am eager to read Mark Peterson's book Galileo's Muse, which is about the influence of arts and humanities on Galileo's thinking, but my local public library doesn't have it yet. Peterson, however, does have a paper about precisely this issue on arXiv, Galileo's Discovery of Scaling Laws. It was fascinating reading.
It turns out that although Galileo's lecture on the structure of Hell was a rhetorical and political success, he got the science exactly wrong. He defended the view of his fellow Florentine Antonio Manetti that Dante's Hell could be a cone roughly 3000 miles across at the top and extending to a point at the center of the Earth and that it could be covered by a ceiling of earth and rock 400 miles thick. The essence of Galileo's argument was that since these proportions were more than adequate to an architectural structure like Brunelleschi's Dome, they should work equally well if scaled up by a factor of 100,000.

The problem of course is that the weight of the roof won't increase linearly like its span and thickness. If you scale up its linear dimensions by 100,000, its volume (and therefore its weight) will increase by a factor of 100,000 x 100,000 x 100,000, one quadrillion times! There is no way a structure like that could support its own weight. Peterson suggests that Galileo realized his mistake soon after the talk, and subsequently developed a whole new theory of scaling which he was later to present in his Dialogues Concerning Two New Sciences.

The most fascinating thing to me is Peterson's argument that Galileo would develop this new theory and then sit on it for fifty years before publishing it. It highlights just how different the intellectual culture of Renaissance Italy was from the scientific culture of today. Galileo apparently developed the theory in order to defend himself in case anyone started poking holes in his original talk. Since he never needed it for that purpose, he didn't publish this revolutionary piece of science until nearly the end of his life.