In The Name of God

Regularity Chemistry

Abstract: A new method in order to obtain information about chemicals is presented. The method is statistical and results both predictive and descriptive information about chemicals. More than justification the method focuses on application. It seems not scientific, but in the instrumentalist approach to science, this is science. Also a suggestion to use AI in chemistry has been explained.

Introduction: Suppose a salesman who sell clothes in a store. He tries for a scientific method in order to predict his market. He records the color that people wear by month. For instance, the result is, black increases 30% in October, or red increases 25% in August. In this way, he manages his work and anticipates what to prepare monthly. The salesman could do something else. He could investigate the reason the people’s clothes colors change. For instance, there are ceremonies that cause to choose white colors that take place in some months. Or for any reason the death rate increases in October that causes the increase of color of black. Or deeper, the psychological effects of season change on tendency to wear special color in people. But maybe he doesn’t need them. He only needs to know what to prepare by month. The two approaches are two way of thinking about science in general in philosophy of science.  The first one is the theory of Regularity and the second one is the theory of Natural Laws. The Regularity theory claims that scientific investigation, at the end, is to discern some orders and regulatory in the phenomena and the Natural Laws theory claims that is the try to discover laws in the nature that really exist. Anyway, somehow the two approaches seem that can give us not just deeper understanding of the nature of science, but to suggest different ways to investigate the nature. The two possible approaches of the salesman, more generally could be explained by the two approaches to investigate the nature and phenomena as Regulatory approach and Natural Laws. I describe the two in an illustrative way.

 Natural Laws: In Natural Laws theory we presuppose reality of a theory that unifies all phenomena (for instance, quantum mechanics) and based on the theory we try to justify the phenomena or predict what that is going to happen, that is equivalent to imply some features to the subject (for instance, the compound will interact with this chemical or will not interact with that chemical.)  Suppose the following figure:

In this approach, all phenomena are emanations of a unified reality in different arrays of elements in different conditions. Though, in order to justifying the phenomenon, this is just to reduce it to the theory’s rules and principles and of course, to justifying a phenomenon, is to predict it. Because justification is to say, “it has to be so because of …” and prediction is to say “it will be so because of …” and they are the same.

Regularity: More than ontological, regularity approach is epistemological. It assumes the scientific theories not as true claims about the reality of the nature, but discerned regulatory, patterns or orders in the phenomena. Suppose the following Figure:

In the figure above, there are only relations between phenomena and there is no theory as reality of the nature. What we can say firmly is that the regulatory exists amongst phenomena and the rest are not definitely true. Those who defend the regularity approach claim that, what we call it, “theory” is not the real laws and the reality that exists, but just a matured regulatory or order we have discerned in the phenomena utilizing some theoretical elements and creatures that the theory presupposes they exist. For instance, they claim that electron doesn’t really exists and that is just a theoretical element that the theory works based on the assumption that it exists. What we really observe is the measurement’s results shown by lab equipment. No one has seen an electron and we just observe effects of the electron’s existence. The regularity philosopher claims that, what that really exists and could be relied on, is the relations between the records and different large scale observations in the experiments, and electron is just an imaginary object that is utilised to just structure the complicated regulatory amongst the phenomena. The regulatory that all scientists together have constructed. Metaphorically, a theory in Natural Laws theory is the reality behind the phenomena that makes them ordered, but in Regulatory theory, the theory is a mental construction before the phenomena that exists in mind, not in the nature. At the end they say that the theories also are a regulatory we have discerned in the phenomena and not necessarily they are about the nature and reality. In chemistry the two approaches about the nature can be illustrated as below:

The figure above is the Natural Laws approach in practice, that we reduce the phenomena to the theory’s principles and then we can find the phenomena relations to each other based on the theory. For instance, the compound shows itself in FTIR spectra in a specific area. Compound structure and spectra has a relationship based on the theory and reducing the two to the theory, we can discern the existence of the compound in a solution by looking at the solution spectrum.But the Regulatory approach is as follow:

In Regulatory approach, this is just to say the relation between them exists, and the theory is just to lead us to the regulatory. A historical example for opposition of the two approach is the periodic trends and patterns in the periodic table as periodic laws. For instance, as one progresses down a group on the periodic table, the ionization energy will likely decrease. Chemists knew that before in 18-century. But the justification that why there is so, was for the time the physicists delved material, and then they justified the periodic trends by structure of the atoms like, the number of valence electrons.  But for the purpose of practice, the 18-centry knowledge about materials also worked.

Suppose that a screen is showing a periodic pattern as following figure:

How to predict the patterns that will depict on the screen in future time?

There are two ways:

1) Laws of Nature: To open the screen and discover why that shows the pattern and to investigate the electrical devices inside the screen and figure out how the patterns are depicted on the screen, and then predict the next pattern that will be depicted on the screen.

2) Regulatory:  To only watch the screen and guess that the pattern is and then predict that the next pattern will be

In summary, the idea of Regulatory chemistry is just like in the above example instead of opening the screen, to discerning that the pattern is by just looking at the screen. But in chemistry, instead of screen we deal with chemicals and their features.

Regulatory Chemistry

Remaining on the surface, not going inside; increasing numbers of the evidences, not focussing in one case. This is the heart of the Regulatory chemistry. Regulatory chemistry is parted in three phases:

1) Gathering Information: We gather as more information as possible about the chemicals. So much information right now is available, though many research has been done in order to identify compounds and chemicals features. The information are atomic weights of compound, the 3D shape of the compound, the melting point, vaporization point, the elements they are contained, the materials they interact with, the materials they don’t interact with, NMR, FTIR, etc. The information will be sorted as follow:

Chemical 1: atomic weight, elements, etc.

Chemical 2: atomic weight, elements, etc.

…

…

…

2) Coordinization: Coordinization is to translate the information into numbers so that they are traceable and determinable. A true sample of coordinization is the 3D Cartesian coordination. Suppose the triple (1, 4, 9). If you have the zero point, you can easily appoint the point in the space correspond with the triple (1, 4, 9). In this way, we have coordinized the 3D space with the space , in a way that each member in the space , correspond with one point in 3D space. For our purpose, coordinization is to define a space in which each point is a chemical with its features, in a way that given the point coordination, all information about the chemical could be extracted. Equivalently, for each possible (more precisely imaginable) condition that a chemical could be in, there is a series of numbers and the condition is obtainable by the series. In this way, for each chemical and it’s features (information related to the chemical) there is a point in N-Dimension space, and equivalently, an N-Numbered series. Suppose the following figure:

At the end, the coordinization will code the information obtained from a chemical into an N-Numbered series as follow:

There are N features, and each feature can have some values, depend on the feature. For instance, a feature can be the FTIR spectrum of the chemical, that could be different graphs. Each graph is a series and each finite series can be represented by a number. (we will illustrate the encoding or coordinization process later in the sample project)

At the end we have a series as follow for each chemical:

3) Analyzation: Now we have structured the space. Every chemical with its features is a series. For m chemicals we have:

Each series can be defined as a point in the N-dimension space as a chemical and its features. The process of analyzation is something like this:

This is to guess the next value based on the previous records order. In this case:

 The process of analyzation is also the same. We can find the interior relation between different characteristics of the chemicals by supposing a function between them. Suppose the following:

In this way, given the feature of the chemical, we evaluate(more precisely, estimate) the unknown features. This can be written for all the features as follow:

There aregiven functions that relate the chemical features to each other. The function can have two parameters as follow:

Also it can have 3, 4, 5, … parameters. In this way, there are values for many functions, and the purpose is to find the relations between the chemical parameters. Also, it could be not to find the relation or function between different parameters, but just the equivalence feature of a special amount for a chemical feature.

You can ask, why the regularity that exists in current records have to maintain in the rest or future records? The question is the same which David Hume asked to reject the causal relation metaphysical necessity. But consider that the permanence of the regularity is the one that whole science is based on. And the more our records are, the results are more reliable.

A Sample Project: Chemicals FTIR Analysis

Each chemical has an individual unique FTIR spectra. The region 400-1500 in FTIR spectra is called  fingerprint for the chemicals and is unique for each chemical. The rest also contains information about the chemical and is more known by scientists. Suppose that we want to extract all information about a chemical by its FTIR spectrum. This is the purpose of the project. For the purpose, we assume a function that relate the FTIR spectra to the rest of the chemical features. For that we need to coordinize the FTIR spectra. This is a sample FTIR spectra:

There are  k points in the horizontal axis. Each point is a specific frequency and that means the spectrophotometer can distinguish K  frequencies. And the vertical axis has  J points. (for instance, in the figure above by the resolution 4,K is 35000000, and J  is 10000000.)It means the spectrophotometer can distinguish J  different absorbance degree for each frequency. Therefore, the FTIR spectrum could be represented by a series of different frequencies absorbance degree as follow:

Though the space is limited, the p dimension space could map on one-dimension space, or equivalently each FTIR spectra be represented by a number.

Thus, the chemicals features series would be as follow:

the purpose is to find the relations between FTIR spectra and the chemicals parameters. Also, it could be not to find the relation or function between them, but just the equivalence features of a special FTIR spectrum for a chemical feature.

Both of the spaces, FTIR spectrum and the rest of chemical features are finite spaces, thus capable to map to one-dimension space. In this way, the above series could be written as relation between two parameters, FTIR spectrum value, and chemical features value as follow:

The problem is to find the function F. In 2D, you can suppose that as a shape that we have m points of the shape, and the problem is to guess the shape. That’s like this:

Of course, the number of points we have in comparison with the number we don’t have is so few and the figure is just for illustration. But the problem is to guess the figure by this points, or equivalently to guess the function that maps the space of F to the space of C.

Also we just can find the value of C for an unknown F, and not to guess the function F, but just the function output for a given value or spectrum, that of course is an easier work.

To Use AI:  The AI itself will learn a relation between spectra and compound features. One needs to provide an adequate number of samples and let the machine train itself. The AI method is not explainatory per se, though. The term “blackbox” has beed dubbed for this situation: This is not essentially a problem. Using this method is as simple as setting a machine to learn from existing samples of compound samples and features pairs. The machine acts as a black box that one can use to implicitly pick up the relationship between the spectra and the compounds features. AI would learn and it can guess the next spectra equivalent chemical features, and that’s what we want.