Understand evaporation curve & longevity of your constituents

[Deleted member 26348570]

Public service announcement:

It is imperative in the learning process for perfumers to not only know the scent profile of an aroma chemical, as well as its effects in a blend, but to additionally, meticulously record each aromas longevity using a consistent method (i.e., on skin, on test strip, etc.). Not doing so will effectively render your perfumes inadequate and lacking in longevity.

For example, my first fragrance I developed did contain an attractive scent, although in hindsight it has terrible performance as there was no heart incorporated! Just volatile materials and heavy molecules.

Lastly, when measuring longevity. You must ensure you are evaluating only one of each class. If you sample too many materials of one family, it may reduce your olfactory’s ability to identify subtle aromas in said family group, thus skewing your results. Also, be sure to evaluate a consistent dilution or do not dilute materials when measuring longevity of an aroma chemical. It is inherent that a diluted material will not last as long as a pure unadulterated material.

 

[RomanB]

It's pointless to rely just on substantivity of separate substances since their evaporation dynamics is heavily affected by other components in mixtures. There is a UNIFAC method that allows to calculate activity coefficients for all constituents of the formula to predict how a complex mixture will evaporate. Since odor detection thresholds for many substances are not listed in literature, there is a way to estimate them. Needed properties like octanol/water partition coefficient could be found for example at chemspider. Together with UNIFAC prediction this gives a subjective proportion of perceived odor components during evaporation. Essential oils could be calculated using data from this database. That's how you can realistically estimate how would your materials perform in a mixture.

 

[Deleted member 26348570]

I appreciate your response with several citations. However, I personally do not have the time to assess the validity of your guidance. Maybe others do. While your argument may hold true, those who are beginning this journey must understand volatile vs non-volatile materials. Otherwise one would spend a substantial amount of time “attempting” to fix, or expand on the character of the perfume. Moreover, many online sources contain erroneous information about the evaporation of an aroma chemical.

But, that goes without saying this topic is one of MANY facets pertaining to the complicated nature of perfuming.

 

[xii]

Otherwise one would spend a substantial amount of time “attempting” to fix, or expand on the character of the perfume.

I spend all my blending time doing exactly that and I don't mind.

Additionally to the molecular dynamics affecting evaporation; which could essentially be modelled from vapour-liquid equilibrium curve of binary mixtures; there's the problem of some substances profoundly affecting perception of others. Only recently I found out, the hard way I'm afraid, PEA can be rendered virtually odourless at 20% in a blend.

 

[RomanB]

I appreciate your response with several citations. However, I personally do not have the time to assess the validity of your guidance. Maybe others do. While your argument may hold true, those who are beginning this journey must understand volatile vs non-volatile materials. Otherwise one would spend a substantial amount of time “attempting” to fix, or expand on the character of the perfume. Moreover, many online sources contain erroneous information about the evaporation of an aroma chemical.

But, that goes without saying this topic is one of MANY facets pertaining to the complicated nature of perfuming.

Parameters like vapour pressure or octanol-water partition coefficients are reliably predicted from molecular structure, so just use one model for all of the substances of your interest and you will get consistent results. Odor detection values are somewhat harder to obtain and they could differ for orders of magnitude in various sources, but again, if no reliable data is present, use one estimator.

While big companies use more advanced models, this one works surprisingly well. When you will get data of evaporation dynamics for all constituents of a formula, you could plot a so-called 100% stacked area chart to see how your perfume evolves in time. Here is my example:

Horizontal axis is time, vertical is a percentage of perception, coloured areas are ingredients. Not that time is shown non-linear, the beginning and ending are logarithmic, since I am interested in the fast evolution of top notes (hence the beginning is wide) and not so interested in slow evolution of base notes (hence the ending is compressed).

Then you can tweak quantities of your formula in a spreadsheet manually and watch how changes affect the overall picture. Or if you wish to create some specific scent profile, for example after 3 hours it should be like so and so, you can use optimisation algorithm to tweak quantities to get close to the desired result.

The main problems with this algorithm are:
It doesn't account for synergism/antagonism of substances (I doubt that even largest companies have a solution)
Chemical reactions which take place on skin, like decompositions of Schiff bases
Adaptation to certain odors like ionones is not accounted for

 

[mnitabach]

Parameters like vapour pressure or octanol-water partition coefficients are reliably predicted from molecular structure, so just use one model for all of the substances of your interest and you will get consistent results. Odor detection values are somewhat harder to obtain and they could differ for orders of magnitude in various sources, but again, if no reliable data is present, use one estimator.

While big companies use more advanced models, this one works surprisingly well. When you will get data of evaporation dynamics for all constituents of a formula, you could plot a so-called 100% stacked area chart to see how your perfume evolves in time. Here is my example:
View attachment 309746
Horizontal axis is time, vertical is a percentage of perception, coloured areas are ingredients. Not that time is shown non-linear, the beginning and ending are logarithmic, since I am interested in the fast evolution of top notes (hence the beginning is wide) and not so interested in slow evolution of base notes (hence the ending is compressed).

Then you can tweak quantities of your formula in a spreadsheet manually and watch how changes affect the overall picture. Or if you wish to create some specific scent profile, for example after 3 hours it should be like so and so, you can use optimisation algorithm to tweak quantities to get close to the desired result.

The main problems with this algorithm are:
It doesn't account for synergism/antagonism of substances (I doubt that even largest companies have a solution)
Chemical reactions which take place on skin, like decompositions of Schiff bases
Adaptation to certain odors like ionones is not accounted for

This all sounds fancy & predictive, but in practice the effects that aren't even in the slightest understood & thus are completely unpredictable from any theoretical basis whatsoever render this a complete waste of time for the actual practice of perfumery & nothing but a distraction from the tedious case-by-case empirical process of learning one's materials & how they interact in pairwise & higher-order combination.

 

[RomanB]

This all sounds fancy & predictive, but in practice the effects that aren't even in the slightest understood & thus are completely unpredictable from any theoretical basis whatsoever render this a complete waste of time for the actual practice of perfumery & nothing but a distraction from the tedious case-by-case empirical process of learning one's materials & how they interact in pairwise & higher-order combination.

The effects are well understood: certain molecules bind to certain receptors in olfactory mucosa. There are about 1000 types of receptors and only about a half is present in any particular human - that’s why odor perception is highly personal and vary. So, if you have molecular models for each receptor and odorant, you can model how they interact and compete with each other in access to said receptors. Computationally this is very laborious but possible - that’s already a question of resources, not of serendipity. There are databases where some receptor-odorant interactions are already calculated, but they are far from being complete.

As for dimensionality of odor space, there are only about 25 dimensions or so. They could be calculated from molecular structure of odorants.

 

[mnitabach]

The effects are well understood: certain molecules bind to certain receptors in olfactory mucosa. There are about 1000 types of receptors and only about a half is present in any particular human - that’s why odor perception is highly personal and vary. So, if you have molecular models for each receptor and odorant, you can model how they interact and compete with each other in access to said receptors. Computationally this is very laborious but possible - that’s already a question of resources, not of serendipity. There are databases where some receptor-odorant interactions are already calculated, but they are far from being complete.

As for dimensionality of odor space, there are only about 25 dimensions or so. They could be calculated from molecular structure of odorants.

Almost none of this is an accurate description of the state of empirical & theoretical understanding of human olfaction. Most importantly, and regardless if all of the above were accurate, there is essentially zero understanding of the physiological basis for perceptual interactions between multiple odor stimuli underlying perfumery effects such as suppression, exaltation, and the formation of accords (either pairwise or higher-order). Since these interaction effects are literally the entire basis of modern perfumery, any theoretical modeling scheme that fails to account for them is worthless either for predicting how a formula will perform or for guiding the composition of new formulas.