In a little-known paper published in 1896, Emil Fischer—the German chemist who would go on to win the 1902 Nobel Prize in Chemistry for synthesizing sugars and caffeine—said his laboratory had produced a crystal that seemed to break the laws of thermodynamics. To his puzzlement, the solid form of acetaldehyde phenylhydrazone (APH) kept melting at two very different temperatures. A batch he produced on Monday might melt at 65 °C, while a batch on Thursday would melt at 100 °C.
Colleagues and rivals at the time told him he must have made a mistake. Fischer didn’t think so. As far as he could tell, the crystals that melted at such different points were identical. A few groups in Britain and France repeated his work and got the same baffling results. But as those scientists died off, the mystery was forgotten, stranded in obscure academic journals published in German and French more than a century ago.
There it would probably have remained but for Terry Threlfall, an 84-year-old chemist at the University of Southampton, UK. Stumbling across Fischer’s 1896 paper in a library about a decade ago, Threlfall was intrigued enough to kick-start an international investigation of the mysterious crystal. Earlier this year in the journal Crystal Growth and Design, Threlfall and his colleagues published the solution: APH is the first recorded example of a solid that, when it melts, forms two structurally distinct liquids. Which liquid emerges comes down to contamination so subtle that it’s virtually undetectable.
A forgotten mystery
The quest began in 2008 when Threlfall, a fluent speaker of German and a keen student of the history of science, was searching the pages of the 140-year-old Berichte der deutschen chemischen Gesellschaft for interesting solid-state work relevant to his research on second-order phase transitions. After learning of the long-lost puzzle from Fischer’s paper, Threlfall followed the reported recipe and found that his own samples of APH melted according to the same peculiar pattern. One batch melted at around 60 °C, the other at 90–95 °C.
As Fischer knew 125 years ago, the laws of thermodynamics do not allow such a molecule. If a pair of solids have different melting points, then they must be structurally distinct. Yet all the modern structural analysis techniques that Threlfall and some colleagues tried on Fischer’s compound confirmed the 19th-century claim. X-ray diffraction, nuclear magnetic resonance, IR spectroscopy: All showed the crystals that behaved so differently were identical.
“For two years we wondered whether to believe the evidence of our own eyes and think that we needed to rewrite the laws of the universe, or to believe thermodynamics and think that we were simply incompetent experimentalists,” Threlfall says.
Piecing together the puzzle
The first clue for solving the mystery came from the way APH crystals are prepared. The molecule (C8H10N2) is made up of a benzene ring attached to a pair of nitrogen atoms, one of which is attached to a hydrogen atom and a methyl group that can point either up or down. Chemists make APH by dissolving solid acetaldehyde (a precursor for many useful chemical reactions and a compound found naturally in fruit) into aqueous ethanol and adding drops of liquid phenylhydrazine (also first made and characterized by Fischer, who used it in his seminal studies of sugars). If the mixture is chilled and stirred, jagged flakes and then thicker chunks of APH crystals start to appear.
According to reports from Fischer’s time, there were hints that impurities could play a role in the puzzling behavior of APH. Adding drops of an acid could steer the crystallization process toward the low-melting-point version of the molecule; with added alkali, the high-melting-point crystal would emerge. Threlfall confirmed that claim and found that he could convert between the two forms. The low-melting version could be made to melt at the higher temperature by exposing it to ammonia vapor. And the high-melting crystal just needed a whiff of acid to bring its melting point down.
That behavior seemed to suggest that the acid worked like rock salt does in lowering the melting point of water ice. But for salt to make a difference, a significant amount must be added—certainly enough to show up in a close examination of the ice’s structure. At as little as a thousandth of a molar equivalent, the quantities of acid or alkali needed to make the switch in APH were vanishingly small. Whatever contamination occurred did so with no detectable physical change to the crystal structure.
Threlfall got some important help from Hugo Meekes, a solid-state physicist at Radboud University in Nijmegen, the Netherlands. After hearing of a 2012 lecture that Threlfall had given about the conundrum, Meekes wondered if the solution might relate to a different, but equally curious, phenomenon called the disappearing polymorph problem. A scourge of drug companies, the problem manifests as the production of a solid that’s slightly but consequentially different from the desired product. The polymorphs are identical except for varying crystalline structures, which can give them different properties. In the late 1990s, for example, Abbott Laboratories learned that it had produced a less-soluble polymorph of its antiviral crystalline compound ritonavir.
The cause of disappearing polymorphs is disputed, but Meekes says it seems to come down to imperceptible contamination—perhaps a single molecule in the air can disrupt the process by seeding crystallization of the problematic form. “It sounds rather unbelievable, but it’s the only explanation,” he says. “We thought the situation with the APH must be something like this.”
But the APH case didn’t fit the pattern. The crystals of APH that melted at different temperatures weren’t polymorphs; they were identical. The researchers failed to find any other structural discrepancies either. For example, some molecules show different physical properties when their same atoms are arranged in different patterns, which is called isomerization. But both solid forms of APH contained the Z isomer, in which the methyl group points down.
Meekes too was stumped.
Enter Manuel Minas da Piedade, a solid-state physicist and thermodynamics researcher at the University of Lisbon, whom Threlfall met at a conference in 2011. After initially offering a hunch that led to another dead end, the Portuguese physicist did what many scientists do when faced with something that doesn’t add up: He went back to first principles. Because it is impossible for the same material to melt at different temperatures if the initial and final states are the same, he says, “either we don’t have the same crystal state, or the final state cannot be the same.”
Until then, all the tests performed by Threlfall and a growing number of interested colleagues had focused on solid APH, since differences in melting point typically stem from differences in the solid form. But, out of options on the solid front, in 2015 the researchers took a look at the liquids that emerged.
Back in the Netherlands, Meekes spun tiny tubes of the hot, molten APH in a solid-state NMR machine, once with the low-melting-point sample and once with the high-melting-point one. Occasional forays to temperatures higher than the delicate equipment’s 100 °C limit led to “frowning technicians,” Meekes says, but the risk was worth it. He discovered that the spectra of the two liquids were different. The same solid crystal was melting to form two liquids with distinct compositions—an unprecedented finding. “We think we have a clue as to what’s going on,” Meekes recalls telling Threlfall at a conference.
The difference, Meekes, Threlfall, and colleagues soon found as they probed further, comes down to isomerization, but only in the liquid phase. Although solid APH consists of solely the Z isomer, liquid APH also contains E isomer, in which the methyl group points up. In the liquid state, with the molecules spaced farther apart and therefore with more room to maneuver, APH can flit between the two forms, and it does so until it finds the most stable mix. That turns out to be a blend of about one-third of the Z isomer and two-thirds of the E form.
The relative amounts of each isomer at equilibrium are determined by the molecules’ Gibbs free energies, a measure of their thermodynamic potential. As the difference in Gibbs energy increases, so does the ratio of one isomer to the other. What makes APH so unusual, Threlfall says, is that the optimal isomer combination for liquid APH doesn’t match that of the solid form. “That the [solid] crystal is composed entirely of Z molecules shows that these must have a more favorable packing,” he says.
Tests showed that the high-melting solid crystal melted to a liquid that was also all Z. Then the Z-type molecules started to flip to E-type and continued until they hit that stable mix. But when the low-melting solid APH melted, it did so almost immediately to the stable mix of two-thirds E. The two liquids are different—and so the melting points are different—only because one represents an intermediate stage.
It was a melting-point suppression effect, just like salt and ice, but it was much larger than anyone on the team had thought possible. So what was behind it? Like the salt, they thought it must be an impurity. And like the disappearing polymorphs that plague the pharmaceutical industry, that impurity is too small to see or measure. Threlfall says hydrogen ions must be clinging to the surface of the solid crystal and catalyzing the shift from the Z form to the E form. To do so, those protons shift the electron density of the nitrogen atoms, which loosens the connection between nitrogen and carbon atoms in the APH molecules from a strong double bond to a weaker single one. The bond is therefore free to rotate, allowing a much more rapid switch between the Z and E forms.
With no acid present, the Z-form solid melts to Z-form liquid, and then this Z-form liquid starts the transition to E-form liquid until it reaches the stable 1:2 ratio. But when acid is there, the catalysis effect speeds the switch from Z form to E form, so much so that it happens as the solid melts.
Overall, the starting solid is the same, the finishing liquid is the same, and the amount of energy used is the same. The laws of the universe are safe. Gérard Coquerel, who works on thermodynamics and solid-state physics at the University of Rouen, France, and was not involved in the project, says it’s an important discovery that organic chemists and others who rely on melting points to help characterize compounds should take into account. “It shows that sometimes there is a need to be careful about what we consider as the melting point,” he says.
Fischer would have been delighted to see the answer, Threlfall says, and the 19th-century chemist would probably have understood it. Although the team’s work breaks genuinely new ground, Meekes cheerfully admits that the circumstances under which the melting-point suppression occurs are so specific that the research is unlikely to have useful applications. The team hasn’t even coined a name for the physical process by which identical solids can melt into distinct liquids. “If someone else wants to name it, then they can,” Threlfall says. “But if you ask me, the scientific literature is already cluttered with too many needless terms.”