top of page
  • Writer's pictureMarabel Riesmeier

Understanding chemical substance (is harder than you might think)

Updated: Apr 12, 2022

What is a chemical substance? Whether you are a chemist, philosopher or historian of chemistry, or whether some other interest in chemistry has brought you to this blog, you probably have an intuitive sense of what the word “substance” means and the answer might seem so obvious that it is not even worth asking. In the following, I want to make you think twice about the systems you have been using to classify and categorise substances. I hope to show that answering the question “What is a substance?” is a lot less straightforward than you might think.

Molecular structure

Let us start with the property that is invoked most often when classifying chemicals: Molecular structure. A look at almost any chemistry textbook or scientific paper will show the ubiquity of structural formulas in chemistry. These formulas are often interpreted as depicting the structure of molecules, i.e. the smallest stable multi-atomic particles. Of course, not all chemical substances can be described using structural formulas. This is the case for salts and metals, for example, which could still be accounted for in a generous reading of structure that includes crystal structures besides structural formulas.

However, there are deeper problems with defining substances via structure. Even in substances usually considered to be molecular, the molecules are neither separate nor stable. For example, it is possible to describe a water molecule, but water as a substance is constantly in flux – molecules are present as fragments, exchange parts or aggregate to superstructures in which molecule-distinctions become meaningless. There is an ongoing debate as to whether a subdivision of matter into distinct molecules is consistent with quantum mechanical calculations of probability densities of electrons and nuclei (see e.g. Schummer 1998; Woolley 1978; Hendry 2016; Ochiai 2015). Furthermore, the actual microstructure of substance may depend upon a variety of things from circumstance and function (Tobin 2010) to impurities in solutions and crystals.

The various points at which structure breaks down show that it can be regarded as an idealisation at best. As an imprecise misrepresentation, it should not be regarded as the defining criterion for substance.

Macroscopic properties

Seeing the problems with structure as a basis for defining and distinguishing substance, we might try to start our search from the other extreme: we could attempt to base our understanding solely on macroscopic properties such as colour or boiling point. I take macroscopic properties to mean properties that can - at least in principle - be accessed using manual intervention and sensory perception. Of course, scientists have found many ways to aid and systematise their perception, from thermometers measuring heat to ultraviolet-visible spectroscopy measuring colour and chromatographic techniques measuring properties like miscibility. Importantly, all these techniques have in common that molecular structure does not need to be presupposed to interpret the measurements (although such interpretations are certainly possible).

There have been some attempts to come up with notions of substance that are not based on molecular structure. A relatively popular example of a system relying on intervention and observation is dividing substances by their behaviour during phase transition (Earley 2006, 847). The problem with this approach is that it would classify enantiomer and azeotropic mixtures as substance, although they are commonly regarded as mixtures of several substances. Schummer’s account of substance addresses these inconsistencies arguing that pure substances are those produced by purification methods (Schummer 1998, 139). Purification methods include all those based on phase transitions plus techniques such as separation of enantiomers using enzymes. Therefore, the approach accounts for a much broader range of substances. It excludes quasi-molecular species, the status of which can indeed be regarded as at least ambiguous. However, the status of chemicals only stable in solution remains unclear in Schummer’s account. On the one hand, they might be counted as impure substances, in which case the distinction between unstable substances and quasi-molecular species remains unclear. On the other hand, they may be regarded as pure substances because it is possible to distinguish them from the known solvent, in which case the boundaries of purification become unclear.

How molecular structure and macroscopic properties are intertwined

The biggest drawback of a definition of substance resting purely on macroscopic properties without reference to structure goes beyond the technicalities of a specific system. It falls short of capturing what modern chemistry can achieve and distinguish. There are several techniques that rely on a prior assumption of structural considerations for their interpretation. One of these is Nuclear Magnetic Resonance (NMR) spectroscopy, which can be used to illustrate the problem with omitting molecular structure when defining the notion of substance. This technique cannot be regarded as an extension of the senses as it relies on measuring perturbations in an electric field. The spectrum that is generated in the process maps resonance frequencies. A spectrum is characteristic of a (pure) substance in a known solvent. However, the information that can be obtained goes way beyond this. NMR spectra can be used to infer molecular structure, since the peak number, shape and location is dependent on where electrons are distributed in a molecule. It is commonly used to confirm the identity of newly synthesized substances. Without using molecular structure for interpretation, the largest part of the information that can be obtained using NMR is lost. In fact, the effort and expense of using NMR as a technique is pointless without the assumption of molecular structure. Similar arguments can be made about X-ray crystallography, some if not all forms of mass spectrometry and to an extent even about infrared spectroscopy. If structure is an abstraction or idealisation, we still measure it.

Not only do we measure molecular structure, it is also used with remarkable success when predicting chemical reactions and the properties of substances. Organic chemistry in particular would be close to non-existent without the omnipresent use of structural formulas. They are used when predicting complex pathways as well as making simple inferences about miscibility or boiling points. Even though molecular structure is a misrepresentation of what is going on at the microscopic level, it does seem to get at something in the world.

To summarise, I have shown that the notion of substance cannot simply be defined based on molecular structure, which is an imprecise idealisation at best. Yet, viewing substance simply as something that can be accessed through macroscopic observations and interventions also fails to capture a large part of the way modern day chemists understand substances. Is the truth somewhere in the middle? Perhaps. It remains yet to be understood how exactly scientists’ direct interaction with chemical substances combines with their idealisations to create a stable (or not so stable?) notion of substance.

Marabel Riesmeier is a PhD student in the Department of History and Philosophy of Science at the University of Cambridge (UK).


Earley, Joseph E. 2006. “Chemical ‘Substances’ That Are Not ‘Chemical Substances.’” Philosophy of Science 73 (5): 841–52.

Hendry, Robin Findlay. 2016. “Structure as Abstraction.” Philosophy of Science 83 (5): 1070–81.

Ochiai, Hirofumi. 2015. “Philosophical Foundations of Stereochemistry.” Hyle 21 (1): 1–18.

Schummer, Joachim. 1998. “The Chemical Core of Chemistry I: A Conceptual Approach.” Hyle 4 (2): 129–62.

Tobin, Emma. 2010. “Microstructuralism and Macromolecules: The Case of Moonlighting Proteins.”

Foundations of Chemistry 12 (1): 41–54.

Woolley, R. G. 1978. “Must a Molecule Have a Shape.” Journal of the American Chemical Society 100 (4): 1073–78.

222 views0 comments


bottom of page