molecules

Review
Perfume and Flavor Engineering: A Chemical Engineering
Perspective
Alírio E. Rodrigues * , Idelfonso Nogueira and Rui P. V. Faria

Laboratory of Separation and Reaction Engineering, LSRE-LCM, Department of Chemical Engineering,

Faculty of Engineering, University of Porto, 4200-465 Porto, Portugal; idelfonso@fe.up.pt (I.N.);
rui.faria@fe.up.pt (R.P.V.F.)
* Correspondence: arodrig@fe.up.pt

Abstract: In the last two decades, scientific methodologies for the prediction of the design, perfor-
mance and classification of fragrance mixtures have been developed at the Laboratory of Separation
and Reaction Engineering. This review intends to give an overview of such developments. It all
started with the question: what do we smell? The Perfumery Ternary Diagram enables us to de-
termine the dominant odor for each perfume composition. Evaporation and 1D diffusion model
is analyzed based on vapor-liquid equilibrium and Fick’s law for diffusion giving access to per-
fume performance parameters. The effect of matrix and skin is addressed and the trail of perfumes
analyzed. Classification of perfumes with the perfumery radar is discussed. The methodology is
extended to flavor and taste engineering. Finally, future research directions are suggested.

Keywords: perfume engineering; flavors and fragrances; perfumery ternary diagram; classification



of perfumes; perfumery radar; trail of perfumes; perfume performance; evaporation and diffusion of


perfumes; effect of matrix; flavor engineering
Citation: Rodrigues, A.E.; Nogueira,
I.; Faria, R.P.V. Perfume and Flavor
Engineering: A Chemical Engineering
Perspective. Molecules 2021, 26, 3095. 1. Looking Back: The Beginning of Perfume Engineering at LSRE. What Do We Smell?
https://doi.org/10.3390/
molecules26113095
In the late 1990s, research on perfume engineering at the Laboratory of Separation and
Reaction Engineering (LSRE) started with one of our members (AER) and a posdoc student,
Academic Editors: Manuela Pintado
Vera Mata. She was had PhD in porous media and wanted to be an entrepreneur. She was
and Patrícia Costa
interested in perfumes and we wanted to answer, with engineering tools, the question:
What do we smell? Can we predict it?
Received: 3 May 2021
Accepted: 20 May 2021
A perfume, according to Jean Carles [1], is a liquid mixture of top notes (first impact,
Molecules 2021, 26,
Published: 22 May 2021
x FOR PEER fresh), middle notes (main perfume character) and base notes (long-lasting) in solvents
REVIEW 2 of 23
(ethanol, water, matrix). The structure of a perfume is shown in Figure 1 as Carles’s pyramid.
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.

Copyright: © 2021 by the authors.

Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https:// .
creativecommons.org/licenses/by/
Figure 1.
Figure 1. The structure
structure of
of aa perfume
perfume represented
represented by
by Carles’
Carles’pyramid.
pyramid.
4.0/).

Perfume engineering is a branch of product engineering that caters to consumer

needs for a specific application or market by providing a new valuable product, following
the path: needs, ideas, selection and manufacturing
Molecules 2021, 26, 3095. https://doi.org/10.3390/molecules26113095
[3]. The product classification of
https://www.mdpi.com/journal/molecules
Raquel Costa et al. [4] is: commodities, specialty chemicals, formulated products, devices,
virtual chemical products, bio-based products and technology-based consumer goods.
Molecules 2021, 26, 3095 2 of 22

Molecules 2021, 26, x FOR PEER REVIEW 2 of 23

It should be noted that perfumed products appear in our everyday lives as fine fra-
grances (Chanel No. 5), air care (candles), fabric care (detergents), personal care (shampoos),
personal wash (bar soaps), home care (dish wash), etc.
The Flavors and Fragrances (F&F) industry has a large palette of essential oils and
fragrances (~104 ) to formulate products, mainly developed by perfumers. At the time
we started research in this area, formulation by trial and error (~1000 tests) implied long
development times (1–3 years) and production costs (perfumery raw materials, PRM, can
cost from $10 up to $50,000 per kg). The F&F industry involved businesses of $26 bn in
2017 [2]. The market share is around 33% in the USA, 28% in Western Europe, 28% in Japan
and China and 11% in the rest of the World. More than 78% of the market in 2017 was
shared by the top 10: Givaudan, Firmenich, IFF, Symrise, Mane SA, Frutarom, Takasago,
.
Sensient Flavors, Robertet SA, T. Hasegawa.
Figure 1. The structure
Perfume of a perfume
engineering represented
is a branch by Carles’
of product pyramid.
engineering that caters to consumer needs
for a specific application or market by providing a new valuable product, following the
Perfume engineering is a branch of product engineering that caters to consumer
path: needs, ideas, selection and manufacturing [3]. The product classification of Raquel
needs for a specific application or market by providing a new valuable product, following
Costa et al.needs,
the path: [4] is:ideas,
commodities, specialty
selection and chemicals,[3].
manufacturing formulated
The product products, devices,
classification of virtual
chemical products,
Raquel Costa et al. [4]bio-based products
is: commodities, and technology-based
specialty chemicals, formulated consumer
products,goods. Perfumes
devices,
can be easily
virtual chemicalrecognized
products,as formulated
bio-based products.
products Product engineering
and technology-based appeared
consumer goods.with the
shift in thecan
Perfumes evolution
be easily of the chemical
recognized industryproducts.
as formulated from bulk chemicals
Product to specific
engineering added-value
appeared
products (electronics,
with the shift flavors,ofcoatings,
in the evolution fragrances,
the chemical industryetc.).
from bulk chemicals to specific
added-value
Perfumeproducts (electronics,
engineering involvesflavors, coatings,
disciplines offragrances, etc.).
thermodynamics, transport phenomena
Perfume engineering
and psychophysics involves
(Figure 2) todisciplines of thermodynamics,
predict the odor of mixtures transport phenomenaevapora-
of fragrances,
and psychophysics
tion/release (Figure 2)diffusion
of fragrances, to predictandtheperformance
odor of mixtures of fragrances,
of perfumes, evapora-
as well as to predict
tion/release of fragrances, diffusion and performance of perfumes, as well
odor detection thresholds and classify perfumes into olfactory families (perfumery radar). as to predict
odor detection thresholds and classify perfumes into olfactory families (perfumery radar).

Figure 2.
Figure 2. Chemical
Chemicalengineering today:
engineering ChEChE
today: = M=P M
E 2[5]
2 2 and
P2 E [5]perfume engineering
and perfume as a branch
engineering as aofbranch of
product engineering [6].
product engineering [6].

2.2. The
The Perception
PerceptionofofOdors
Odors
The pyramid structure of a perfume mentioned above considers top notes as those
The pyramid structure of a perfume mentioned above considers top notes as those
giving the initial impact of fragrance; typically citrus, with green notes lasting 15–30 min
giving the initial impact of fragrance; typically citrus, with green notes lasting 15–30 min
on the skin. The middle notes are typically spicy, leather, or floral, giving the perfume
on the skin.
character Theand
(body), middle notes
lasting 3–4 hare
on typically spicy,
the skin. Base leather,
notes or floral,
are typically giving
amber the perfume
or musk,
character
giving the substantivity of the fragrance, and lasting more than 4 h on the skin. Theseor musk,
(body), and lasting 3–4 h on the skin. Base notes are typically amber
giving the substantivity
notes, together with solventsof(ethanol,
the fragrance, and lasting
water, matrix), more
stabilizers, than 4 h
colorants, onUV
and the skin. These
filters,
notes, together with
constitute the perfume.solvents (ethanol, water, matrix), stabilizers, colorants, and UV filters,
constitute the perfume.
The odor perception, from vapor to the nose, was addressed by Richard Axel [7] and
Linda B. Buck [8], who won the Nobel Prize in Physiology or Medicine in 2004 “for their
discoveries of odorant receptor and the organization of the olfactory system”. Odorants
in the air bind to odorant receptors; the odorant receptor cells in the nasal epithelium are
 Molecules 2021, 26, x FOR PEER REVIEW 3 of 23

The odor perception, from vapor to the nose, was addressed by Richard Axel [7] and
Molecules 2021, 26, 3095 3 of 22
Linda B. Buck [8], who won the Nobel Prize in Physiology or Medicine in 2004 “for their
discoveries of odorant receptor and the organization of the olfactory system”. Odorants
in the air bind to odorant receptors; the odorant receptor cells in the nasal epithelium are
activated
activated andand send
send electric
electric signals,
signals, which
which areare relayed
relayed in in glomeruli
glomeruli of of
thethe olfactory
olfactory bulb
bulb
and transmitted to higher regions of
and transmitted to higher regions of the brain. the brain.
TheThe process
process forfor
thethe perception
perception of of a perfume
a perfume cancan
be be divided
divided into
into four
four steps:
steps: first
first thethe
evaporation
evaporation of of
thethe perfume,
perfume, followed
followed byby diffusion
diffusion in in
thethe
airair until
until thethe olfactory
olfactory system
system is is
reached,
reached, where
where there
there is the
is the perception
perception of odor
of odor intensity
intensity and and odor
odor character.
character. TheThe first
first twotwo
steps are the domain of chemical engineering and the last two belong
steps are the domain of chemical engineering and the last two belong to psychophysics. to psychophysics.

2.1.2.1.
Odor Thresholds
Odor Thresholds
The Odor
The Detection
Odor Detection Threshold
Threshold(ODT) is theisminimum
(ODT) the minimum concentration of an odorant
concentration that
of an odorant
canthat
be can
detected by humans, or, according to ASTM (method E 679-91), is
be detected by humans, or, according to ASTM (method E 679-91), is the concen- the concentration
of tration
an odorous chemicalchemical
of an odorous at whichatthe physiological
which effect elicits
the physiological a response
effect elicits for 50%
a response for of
50%
theofpanelists. Human ODT in the air can vary by order of magnitudes
the panelists. Human ODT in the air can vary by order of magnitudes from 0.3 ppb from 0.3 ppb forfor
t-butyl
t-butylmercaptan,
mercaptan, 40 40
ppb
ppbforfor
menthol,
menthol,870870ppbppbforfor
formaldehyde,
formaldehyde, 15,000
15,000 ppb
ppbforfor
acetone
acetone
andand 141,000 ppb
141,000 ppbforformethanol
methanol [9–11].
[9–11].Recalling
Recalling a story
a storyof of
a Ph.D.
a Ph.D.student
student working
working onon
sweetening gasoline and using t-butyl mercaptan in our lab; when getting
sweetening gasoline and using t-butyl mercaptan in our lab; when getting on the crowded on the crowded
busbusto to
gogohome
home at at
rush hour,
rush a void
hour, space
a void spacewaswasnevertheless
nevertheless created around
created aroundthethe
student!
student!
The Odor Recognition Threshold (ORT) is the lowest odorant concentration
The Odor Recognition Threshold (ORT) is the lowest odorant concentration at which at which its its
recognition becomes possible. ODT can be measured in the laboratory
recognition becomes possible. ODT can be measured in the laboratory by a panel using by a panel using an
olfactometer as shown
an olfactometer in Figure
as shown 3.
in Figure 3.

Figure
Figure 3. Measurement
3. Measurement of of
ODTODT
byby a panel
a panel using
using an an olfactometer
olfactometer at LSRE.
at LSRE.

2.2.2.2.
Odor Intensity
Odor Models
Intensity Models
The simplest way
The simplest way to to
quantify
quantify thethe
smell of of
smell a fragrance molecule
a fragrance moleculei isi the Odor
is the Value
Odor Value
g
(OVi ) [12] defined as the ratio of its concentration in the gas-phase Ci and itsg odor detection
(OVi) [12] defined as the ratio of its concentration in the gas-phase Ci and its odor de-
threshold ODTi , i.e.,
tection threshold ODTi, i.e., g
Ci
OVi = . (1)
ODTiC g
A more realistic measure of the smell OV i = somehow,
i
which, . considers the saturation of the(1)
sensor (nose) is the odor intensity ψi defined, according
ODT
i to Stevens Power Law [13], as

A more realistic measure of the smell which,

g
!ni somehow, considers the saturation of
the sensor (nose) is the odor intensity y
ψi = i
Cdefined,
i
, according to Stevens Power Law(2)
[13],
ODTi
as
n
where ni is the power law exponent. æ Cig ö i
y =ç ÷ , (2)
2.3. Odor Character Model for a Mixture—Thei Strongest
è ODTi Component
ø Model
The simplest model to quantify the smell from a multicomponent mixture is to say
that we smell the component with the higher odor value or higher intensity, i.e.,

OVmixture = max{O Vi }
(3)
ψmixture = max{ ψi }
 2.3. Odor Character Model for a Mixture—The Strongest Component Model
The simplest model to quantify the smell from a multicomponent mixture is to say
that we smell the component with the higher odor value or higher intensity, i.e.,

OVmixture = max {OVi }

Molecules 2021, 26, 3095 . 4 of 22
(3)
ymixture = max {yi }
This is the idea behind the Strongest Component Model (SCM).
This is the idea behind the Strongest Component Model (SCM).
2.4. Sensory
2.4. Sensory Dose/Response Curve
Dose/Response Curve
The sensory
The sensory dose/response
dose/response curvecurve is
is aa qualitative
qualitative measure
measure by panel members
by panel members ofof the
the
odor intensity for different concentrations in the liquid mixture. The scale of intensity
odor intensity for different concentrations in the liquid mixture. The scale of intensity goes goes
from extremely
from extremely weak
weak (1)
(1) to
to strong
strong(7).
(7). Figure
Figure44shows
showsaadose/response
dose/response curve
curve indicating
indicating
the threshold and the saturation.
the threshold and the saturation.

Figure
Figure 4. Sensory dose/response
4. Sensory curvewith
dose/response curve with aa scale
scale of
of odor
odor intensity.
intensity.

Researchers from Kao Corporation [14] developed a database for 314 perfumery raw
Researchers from Kao Corporation [14] developed a database for 314 perfumery raw
materials (PRM) with Odor Intensity Standard Curves (OISC) showing the relationship
materials (PRM) with Odor Intensity Standard Curves (OISC) showing the relationship
between PRM odor intensity and gas concentration. They used the Labeled Magnitude
between PRM odor intensity and gas concentration. They used the Labeled Magnitude
Scale (LMS) of Green [15,16] with a range 0–100 with a quasi-logarithmic scale between
Scale (LMS) of Green [15,16] with a range 0–100 with a quasi-logarithmic scale between
labels: barely detectable (bd), 1.4; weak (w), 6.1; moderate (M), 17.2; strong (s), 35.4;
labels:
very barely
strong detectable
(vs), 53.3; and(bd), 1.4; weak
strongest (w), 6.1; moderate
imaginable, 100. Figure(M), 17.2; strong
5 shows (s),manznate,
OISC for 35.4; very
melonal, caryophyllene and cis-3-hexenyl isobutyrate. This treatment is an extension ofmel-
strong (vs.), 53.3; and strongest imaginable, 100. Figure 5 shows OISC for manznate, our
onal,
Molecules 2021, 26, x FOR PEER REVIEW caryophyllene and cis-3-hexenyl isobutyrate. This treatment is an extension
methodology in the sense that it links values of odor intensity to the sensory evaluation of
5 our
of 23
methodology
of panelists. in the sense that it links values of odor intensity to the sensory evaluation
of panelists.

Figure
Figure 5.
5. Odor
Odor Intensity Standard Curves
Intensity Standard Curves in
in LMS
LMSscale
scaleversus
versusthe
thelogarithm
logarithmofoffragrance
fragrance concen-
concentra-
tration in air (Reprinted with permission from Ind. Eng. Chem. Res. 2019, 58, 15036−15044. Copyright
tion in air (Reprinted with permission from Ind. Eng. Chem. Res. 2019, 58, 15036−15044. Copyright
2019, American Chemical Society).
2019, American Chemical Society).
2.5. Evaporation of Perfumes: Modeling Vapor-Liquid Equilibrium (VLE)
The first step to get the concentration of the fragrance in the gas-phase Cig is to re-
late the liquid composition given by the mole fraction xi with the partial pressure of com-
Molecules 2021, 26, 3095 5 of 22

2.5. Evaporation of Perfumes: Modeling Vapor-Liquid Equilibrium (VLE)

g
The first step to get the concentration of the fragrance in the gas-phase Ci is to relate
the liquid composition given by the mole fraction xi with the partial pressure of component
i in the gas-phase Pi = yi P with the Raoult’s law:

yi P = xi γi Pisat , (4)

where yi is the mole fraction of species i in the gas phase, P is the pressure, Pisat is the vapor
pressure of species i and γi is the activity coefficient of species i in the liquid phase. Taking
into account the ideal gas law, the concentration of component i in the gas phase is

g xi γi Pisat
Ci = , (5)
RT
and its odor value is: !
Pisat Mi

1
OVi = γi xi , (6)
ODTi RT

where Mi is the molecular weight. The key point is the prediction of vapor composition us-
ing group-contribution methods (UNIFAC, UNIFAC-D, ASOG, A-UNIFAC, COSMO-SAC)
for calculating the activity coefficients [17,18]. The prediction of the sensorially-dominant
odor was compared with the evaluation of perfumers with a correlation r2 = 0.806 [19].

2.6. Prediction of Odor Detection Threshold (ODT)

Molecules 2021, 26, x FOR PEER REVIEWThe task of predicting ODT is not easy, but a simplified vision of the process of odor6 of 23
perception will be helpful. We will consider that odorant molecules evaporate and reach
our nose (Pisat ), dissolve in the mucus layer (Ci,W ) and bind to specific olfactive receptors
(OR) located in the cilia (KOW ); above a certain concentration (ODT) are detected. A sketch,
(OR) located in the cilia ( KOW ); above a certain concentration (ODT) are detected. A
as shown in Figure 6, aids in understanding.
sketch, as shown in Figure 6, aids in understanding.

Figure 6. Simplified steps in odor detection: air and odorant molecules (C), mucus (B) and nasal
Figure 6. Simplified steps in odor detection: air and odorant molecules (C), mucus (B) and nasal
epithelium (A).
epithelium (A).

Introducingpartition
Introducing partitionequilibrium
equilibriumbetween
betweenthe
theconcentration
concentrationof
ofodorant
odorantininair
airand
and
P sat and equilibrium between the
Pisat
air/mucus ∝
water) KK air/mucus µ K air/water =
mucus Kair/water = i
mucus(assumed
(assumedas
aswater) and equilibrium between the
Ci,W RT
Ci,W RT
odorant concentrations in the mucus and bio-phase (assumed as octanol)
Kmucus/biophase µ Koctanol,water the ODT becomes:
K air,W Pi sat
ODTi µ = . (7)
K K C RT
Molecules 2021, 26, 3095 6 of 22

odorant concentrations in the mucus and bio-phase (assumed as octanol) Kmucus/biophase ∝

Koctanol,water the ODT becomes:

Kair,W Pisat
ODTi ∝ = . (7)
KOW KOW Ci,W RT

Data from literature for 121 odorant molecules were correlated by

log(ODTi ) = (0.97 ± 0.05)log( Pisat /KOW Ci,W RT ) + (4.2 ± 0.2), (8)

with a regression factor od r2 = 0.77 [20].

3. Perfumery Ternary Diagram (PTD)

An engineering tool to predict the smell of a liquid mixture was first presented by
Mata and Rodrigues [21–24] and can be illustrated for the simple case of a mixture of one
top note (A), one middle note (B) and one base note (C) in a solvent ethanol (S). The idea
comes from engineering ternary diagrams using mixture compositions on a solvent-free
basis and working with odor values (or intensity) for each component A, B, and C. For each
liquid composition, the odor values (or intensity) are calculated and using the Strongest
Component Model we can map the triangle in regions with different dominant odors. This
is the Perfumery Ternary Diagram (PTD), shown in Figure 7. The distance from a point in
the triangle to the AB line is the composition of C on a solvent-free basis xC0 = x +xxCB7+ofx 23
Molecules 2021, 26, x FOR PEER REVIEW
A C
where xA , xB and xC are the compositions in the whole perfume mixture. Similarly, the
distance from a point to AC gives the composition of B and the distance to the line BC
gives
to thethe composition
line of composition
BC gives the A. The lines separating odorseparating
of A. The lines zones are the Perfumery
odor zones areBinary Lines
the Perfum-
(PBL) where the OV of the two adjacent components are equal.
ery Binary Lines (PBL) where the OV of the two adjacent components are equal.

Figure7.7.The
Figure Theperfumery
perfumeryternary
ternarydiagram:
diagram:combining
combiningperfume
perfumepyramid
pyramidstructure
structurewith
withthe
theternary
ternary
phase diagram.
phase diagram.

Theidea
The ideawas
was extended
extended to quaternary
to quaternary (PQD)
(PQD) and and quinary
quinary mixtures
mixtures (PQ2D)(PQ2D) by
by Teix-
Teixeira
eira [25] to[25] to allow
allow visualization,
visualization, but canbut
be can be applied
applied to a mixture
to a mixture of any of
of any number number
compo- of
components. What is needed
nents. What is needed is just: is just:
(i) aa database
(i) database of of vapor
vapor pressure
pressureforforperfumery
perfumeryraw rawmaterials
materials(PRM);
(PRM);
(ii)
(ii) aa database
database of of ODT
ODT for forPRM
PRMand and
(iii)
(iii) aa tool
tool to
to calculate
calculate activity
activitycoefficients
coefficientsforforPRM.
PRM.
The
ThePTDPTDtooltoolallows
allowsus ustotoanalyze
analyzethetheeffect
effectofofnon-ideality
non-idealityofofperfume
perfumemixtures
mixtureson on
the 0s
theodor
odorzones
zonesasas well
wellas as
thethe
description of smell
description withwith
of smell odorodor
value or intensity
value from from
or intensity Steven
Ste-
law.
ven′sThelaw.effect
Theofeffect
base of
notes and
base fixatives
notes can be easily
and fixatives can bevisualized as illustrated
easily visualized in Figure 8.
as illustrated in
Figure 8. By changing the base note from vanillin (left) to tonalide (right) there is no region
where we smell tonalide; instead, we smell ethanol.
 components. What is needed is just:
(i) a database of vapor pressure for perfumery raw materials (PRM);
(ii) a database of ODT for PRM and
(iii) a tool to calculate activity coefficients for PRM.
Molecules 2021, 26, 3095 7 of 22
The PTD tool allows us to analyze the effect of non-ideality of perfume mixtures on
the odor zones as well as the description of smell with odor value or intensity from Ste-
ven′s law. The effect of base notes and fixatives can be easily visualized as illustrated in
Figure 8. By changing
By changing thenote
the base basefrom
notevanillin
from vanillin (left)
(left) to to tonalide
tonalide (right)(right)
there there
is no is no region
region where
where we smell
we smell tonalide;
tonalide; instead,
instead, we smell
we smell ethanol.
ethanol.

Figure 8. The effect of base note on odor zones on odor zones: left A—limonene (squares); B—
Figure 8. The effect of base note on odor zones on odor zones: left A—limonene (squares); B—gera-
geraniol (triangles); C—vanillin (circles); S—ethanol (losange); right—the base note C tonalide is
niol (triangles); C—vanillin (circles); S—ethanol (losange); right—the base note C tonalide is not
not perceived
perceived and ethanol
and ethanol is (Reprinted
is (Reprinted with permission
with permission from AIChEJ,
from AIChEJ, 2009,
2009, 55, 15.55, 15.Wiley
John John and
Wiley
and
Sons). Sons).

4. Diffusion and Performance of Perfumes

Molecules 2021, 26, x FOR PEER REVIEW
4. Diffusion and Performance of Perfumes
4.1. Perfume Performance
4.1. Perfume Performance
In perfumery, the performance of a perfume is evaluated using the words “impact”
In perfumery,
(immediate the performance
olfactory odor sensation),of a“tenacity”
perfume is(persistence
evaluated using the words
of fragrance after“impact”
some time
(immediate
near the olfactory
source), odor sensation),
“diffusion” “tenacity”
(efficacy of perfume(persistence
at some of fragrance
distance fromafter
the some time
source) and
“volume” (effectiveness of a perfume over time and distance). In chemical engi
near the source), “diffusion” (efficacy of perfume at some distance from
“volume” (effectiveness of a perfume over time and distance). In chemical engineering the source) and
terms, all we need is a model allowing the calculation of the concentration of p
terms, all we need is a model allowing the calculation of the concentration of perfume com-
components
ponents as a function
as a function of time and ofdistance,
time and Ci (distance, C (z, t) , and
z, t), and then iconverting then
that converting th
information
into odor value or intensity. This is shown in Figure 9 from Teixeira [26].
mation into odor value or intensity. This is shown in Figure 9 from Teixeira [26].

Figure
Figure 9.9. Perfume
Perfume performance
performance parameters
parameters (Reprinted
(Reprinted with from
with permission permission
AIChEJ, from AIChEJ,
2013, 59, 15. 201
John Wiley
John Wiley and
and Sons).
Sons).

The top notes will be perceived first (blooming phase) followed by the middle notes
The top notes will be perceived first (blooming phase) followed by the midd
(development phase) and then base notes will eventually show up in the lasting phase as
(development
sketched in Figurephase)
10. and then base notes will eventually show up in the lasting p
sketched in Figure 10.
 Figure 9. Perfume performance parameters (Reprinted with permission from AIChEJ, 2013, 59, 15.
John Wiley and Sons).

Molecules 2021, 26, 3095 The top notes will be perceived first (blooming phase) followed by the middle notes
8 of 22
(development phase) and then base notes will eventually show up in the lasting phase as
sketched in Figure 10.

Figure 10. Blooming, development and lasting phases.

Figure 10. Blooming, development and lasting phases.
4.2. A Simple 1D Diffusion Model
4.2. A Simple 1D Diffusion Model
A 1D perfume diffusion model [27] is developed for the gas–liquid system involving
Molecules 2021, 26, x FOR PEER REVIEW
a massA 1D perfume
balance in a diffusion model [27]
volume element is developed
of the gas phase with for the gas–liquid
thickness system
∆z (Figure
9 of 23involving
11)
a mass balance in a volume element of the gas phase with thickness Δz (Figure 11)
∂2 y2i
¶y ¶the
∂yi
2 yiaxial space coordinate, Di,air is
= D , (9)
= Di,air
In the above equations, t is the time ∂t i i,air
variable, z is∂z , (9)
¶t (10) represents
the diffusivity of species i in the air. Equation ¶z 2 the equilibrium at the gas-
with initial condition y =
liquid interface. For the liquid
i y =
phase
i0 0 and boundary conditions
the mass balance is: at t > 0
with initial condition yi = yi0 = 0 satand boundary conditions at t > 0
γdn
i Pii
sat γi P¶yi ni
yi = =satxDi i,air
= Alg i sat , for z = z0 (12) (10)
gdtP
P g PP¶z ∑ nni
yi = i i xi = i i z=0i i for z = z0
with initial condition ni = ni0 or å nini is the number of moles of
xi =Pxi0 , xi = xPi0 , where
yi = 0 for z = zim .
(10)
(11)
species i in the liquid phase, and Alg is the area of the liquid/gas interface.
yi = 0 for z = zm . (11)

Figure
Figure11.
11.Diffusion
Diffusiontube with
tube volume
with element
volume of thickness
element Δz.
of thickness ∆z.

The odor
In the evolution
above near the
equations, t issource for avariable,
the time perfumezmixture containing
is the axial limonene (A),
space coordinate, Di,air is the
geraniol (B) and vanillin (C) is shown in Figure 12.
diffusivity of species i in the air. Equation (10) represents the equilibrium at the gas-liquid
interface. For the liquid phase the mass balance is:

dni ∂y 
= Di,air Alg i  , (12)
dt ∂z z=0

with initial condition ni = ni0 or xi = xi0 , xi = xi0 , where ni is the number of moles of
species i in the liquid phase, and Alg is the area of the liquid/gas interface.
The odor evolution near the source for a perfume mixture containing limonene (A),
geraniol (B) and vanillin (C) is shown in Figure 12.

(a) (b)
Figure 12. Time evolution of OV near the source- we smell limonene first and geraniol after 1 h (a)
and iso-OV lines in a plot distance vs. time (b). Reprinted with permission from Chem. Eng. Sci. 2009,
64, 2570–2580, 2009, Elsevier.
 Figure 11. Diffusion tube with volume element of thickness Δz.
Molecules 2021, 26, 3095 9 of 22
The odor evolution near the source for a perfume mixture containing limonene (A),
geraniol (B) and vanillin (C) is shown in Figure 12.

(a) (b)
Figure
Figure12.12.
TimeTime
evolution of OV near
evolution of OV the near
source-
thewesource-
smell limonene first limonene
we smell and geraniolfirst
afterand
1 h (a)
geraniol after 1 h
and iso-OV lines in a plot distance vs. time (b). Reprinted with permission from Chem. Eng. Sci. 2009,
(a) and iso-OV lines in
64, 2570–2580, 2009, Elsevier.a plot distance vs. time (b). Reprinted with permission from Chem. Eng. Sci.
2009, 64, 2570–2580, 2009, Elsevier.
Evaporation lines are shown in PTD and PQD diagrams at z = 0 (near the source) for
Evaporation
a mixture of limonenelines
(A), are shown
geraniol (B),in PTD and
vanillin PQD
(C) and diagrams
ethanol at z = 0 (near
(S). Depending on the the source) for a
mixture
initial of limonene
perfume composition(A),thegeraniol (B), different
smell follow vanillin paths;
(C) andfor ethanol
example,(S).
if weDepending
start with on the initial
Molecules 2021, 26, x FOR PEER REVIEW
the mixture P1 we smell limonene first and thendifferent
geraniol (Figure 10 of 23
perfume composition the smell follow paths;13).for example, if we start with the
Diffusion models from a fixed source have been extended
mixture P1 we smell limonene first and then geraniol (Figure 13). to account for the radial
space dimension [28,29].

Figure
Figure13.13.
Evaporation lines of
Evaporation a perfume
lines mixture near
of a perfume the source
mixture near in PTD
the and PQD.
source Reprinted
in PTD with Reprinted with
and PQD.
permission from Chem. Eng. Sci. 2009, 64, 2570–2580, Elsevier.
permission from Chem. Eng. Sci. 2009, 64, 2570–2580, Elsevier.
5. Perfume Classification and Perfumery Radar
Diffusion models from a fixed source have been extended to account for the radial
Perfumery
space raw materials
dimension [28,29]. are classified in various categories by experts and are not
usual for people. A classification of fragrances may have different layers related to emo-
tions (culture, memories), look (color, texture), sensations (cool, dry, fatty, warm), aromas
5. Perfume Classification and Perfumery Radar
(floral, citrus, fruity, woody). There are classifications of fragrances, such as The Fragrance
Wheel Perfumery raw materials
by Michael Edwards [30] or theare
Dromclassified
fragrancein various
circle categories
[31]. Most by experts and are not
typical descriptors
usual
used in for people.
fragrance A classification
classification of fragrances
use a decreasing ordermay have
in the different
percentage inlayers related to emotions
an in-house
database
(culture, ofmemories),
2000 odorants: floral,
look rose, texture),
(color, diffusive, fruit, violet, green,
sensations (cool,musk, woody,
dry, fatty, herba- aromas (floral,
warm),
ceous,
citrus,citrus, spice,
fruity, jasmine,
woody). amber,
There arehoney, liquor, marine,
classifications leather moss,
of fragrances, tobacco.
such Exam-
as The Fragrance Wheel
ples of databases are those of Brechbill [32], Surburg and Panten [33], The Good Scents
by Michael Edwards [30] or the Drom fragrance circle [31]. Most typical descriptors used in
Company [34] and AIHA [35]. The 19 descriptors were used to choose eight olfactory fam-
fragrance
ilies classification
to be represented use aplot—the
in a radar decreasing order radar
perfumery in the(PR)—to
percentage
reduceinthe
anarbitrar-
in-house database of
iness of perfume classification. The Perfumery Radar (PR) methodology [36] involvesherbaceous,
2000 odorants: floral, rose, diffusive, fruit, violet, green, musk, woody, sev- citrus,
spice,
eral jasmine, amber, honey, liquor, marine, leather moss, tobacco. Examples of databases
steps:
(i)areClassification
those of Brechbill [32], Surburg
of pure fragrances and Panten
in j = 8 olfactory [33],
families: The
citrus, Good
fruity, Scents
floral, green,Company [34]
andherbaceous,
AIHA [35]. musk,The 19 descriptors
oriental, woody; were used to choose eight olfactory families to be
(ii) prediction of
represented inthe odor intensity
a radar for each
plot—the fragrance radar
perfumery i, OVi; (PR)—to reduce the arbitrariness of
N
OVj = å wi OVi , and normalization
perfume classification. The Perfumery Radar (PR) methodology
j [36] involves several steps:
(iii) calculation of the odor value for each family
(i) Classification of pure fragrances in j = 8 olfactory
i=1 families: citrus, fruity, floral, green,
OV
herbaceous, musk, oriental, woody;
j = L
j
(ii) OV
'
;
prediction of the odor intensity for each fragrance i, OVi ;
å j OV N j
(iii) calculation
j=1 of the odor value for each family OVj = ∑ wi OVi , and normalization
(iv) plotting the perfumery radar. i =1
0 OVj
OVj = L j ;
The weights wi j presented are needed when one PRM is allocated to more than one
∑ OV
family. The PR canj=1 be validated using GC-MS analysis of perfumes, family odor intensity

model and comparison with headspace and perfume classifications. Examples of PR for
feminine perfumes, such as L’Air du temps (Nina Ricci) and Addict-Eau de Toillete (Dior),
are shown in Figure 14, together with company classification Floral and Floral–Green and
Oriental and Oriental–Floral, respectively.
It is possible to combine the Perfumery Radar with a diffusion model and evaluate
Molecules 2021, 26, 3095 10 of 22

(iv) plotting the perfumery radar.

j
The weights wi presented are needed when one PRM is allocated to more than one
family. The PR can be validated using GC-MS analysis of perfumes, family11odor
Molecules 2021, 26, x FOR PEER REVIEW of 23 intensity
model and comparison with headspace and perfume classifications. Examples of PR for
feminine perfumes, such as L’Air du temps (Nina Ricci) and Addict-Eau de Toillete (Dior),
are shown
PRM in Figureand
in the database 14,uses
together with
scientific company
tools classification
to predict Floral
the odor space andof
instead Floral–Green
relying and
Oriental
on and Oriental–Floral,
the sensorial respectively.
perception of people.

Molecules 2021, 26, x FOR PEER REVIEW 11 of 23

PRM in the database and uses scientific tools to predict the odor space instead of relying
on the sensorial perception of people.

Figure 14. Perfumery radar of perfumes: left—L’ Air du temps (Nina Ricci); right—Addict-Eau de
Figure
Toillete14.(Dior)
Perfumery radar of(adapted)
(Reprinted perfumes: left—L’ Air du temps
with permission (Nina
from Ricci); right—Addict-Eau
Perfumery de tool for
radar: a predictive
Toillete (Dior) (Reprinted (adapted) with permission from Perfumery radar: a predictive tool for
perfume family classification, Ind. Eng. Chem. Res. 2010. Copyright 2010, American Chemical Society).
perfume family classification, Ind. Eng. Chem. Res. 2010. Copyright 2010, American Chemical Soci-
ety).
It is possible to combine the Perfumery Radar with a diffusion model and evaluate the
evolution of perfume performance as shown in Figure 15 for Gloria (Cacherel) at t = 0 and
t = 60 s. It is interesting to note that this perfume was classified by experts from various
companies as oriental–woody (Osmoz and Fragrantica.com), oriental–fresh (Scent Direct),
oriental–woody–floral (iPerfumer, Givaudan), floral–woody–amber (SFP), amber–rose
(LT&TS), floral–oriental
Figure 14. Perfumery (Perfume
radar of perfumes: Intelligence).
left—L’ The
Air du temps perfumery
(Nina radar correctlydepredicts the
Ricci); right—Addict-Eau
olfactory families of several commercial perfumes. It is flexible to inserting
Toillete (Dior) (Reprinted (adapted) with permission from Perfumery radar: a predictive new PRM in
tool for
the database
perfume and uses Ind.
family classification, scientific tools
Eng. Chem. Res.to predict
2010. the odor
Copyright space instead
2010, American ofSoci-
Chemical relying on the
ety).
sensorial perception of people.

Figure 15. Perfumery radar of Gloria (Cacherel) at time t = 0 (left) and after t = 60 s (right). (Reprinted
(adapted) with permission from Perfumery radar: a predictive tool for perfume family classification,
Ind. Eng. Chem. Res., 2010. Copyright 2010, American Chemical Society).

6. The Effect of Matrix and Skin

6.1. Effect of Matrix (Glycerine, Dipropylene Glycol, Skin Lotion)
It is expected that the odor of a liquid fragrance mixture will be influenced by the
Figure in
matrix Perfumery
15.which radar of[37–39]:
it is present Gloria (Cacherel)
ethanol inata time t = 0 (left)
perfume, and after
glycerine, t = 60 s (right).
dipropylene glycol(Reprinted
in cosmeticwith
(adapted) matrices.
permission The matrix effect wasradar:
from Perfumery analyzed in our lab
a predictive toolto forstudy
perfume the family
releaseclassification,
of
Figure 15. Perfumery
Origanum. majorana radar
L. of Gloria
from (Cacherel)
glycerine, at time t = 0 (left)
dipropylene glycoland
andafter t = 60a stopical
from (right). formulation
(Reprinted
Ind. Eng. Chem. Res., 2010. Copyright 2010, American Chemical Society).
(adapted) with permission from Perfumery radar: a predictive tool for perfume family classification,
–skin lotion. The supercritical fluid extraction with CO2 of the aerial part of O. majorana
Ind. Eng. Chem. Res., 2010. Copyright 2010, American Chemical Society).
6. The
was Effect
carried outofat Matrix
40 °C and and
10 Skin
MPa. The extract was incorporated in the matrix in 0.01%
(v/v). Dynamic headspace (DHS) coupled to GC-FID/MS was used to measure species con-
6. The Effect of Matrix and Skin Dipropylene Glycol, Skin Lotion)
6.1. Effect of Matrix (Glycerine,
centration over time. The odor profile in the presence of glycerine was initially character-
6.1. EffectIt isMatrix
expected that the odor of Glycol,
a liquidSkinfragrance mixture will be influenced by the
ized by ofthe (Glycerine,
fast release Dipropylene
of β-caryophyllene, sabinene,Lotion)
p-cymene, limonene, myrcene, li-
matrix in which
It isacetate
expected
it is present [37–39]: ethanol in a perfume, glycerine, dipropylene glycol
nalyl andthat the odor of afollowed
β-phellandrene liquid fragrance
by a high mixture
decreasewillinbe theinfluenced
next 5 h. by Thethefast
in
matrix cosmetic
releasein is
which matrices.
it isthe
due to present The matrix
[37–39]:nature
hydrophobic
effect was
ethanolofinthese analyzed
a perfume,
compounds
in
glycerine,our lab
with dipropylene
to study
low LogPow, glycol
the
i.e., low
release of
Origanum.
in affinity
cosmetic with majorana
matrices.
the polar The L. from glycerine,
matrixglycerine.
solvent effect was Fordipropylene
analyzed
example,inthe glycol
ourLogP
lab ow and from
to isstudy
4.5 forthe a topical
release of
limonene formulation
and
–skin
3.8 for lotion.
Origanum. TheL.supercritical
majorana
linalyl acetate. from glycerine, fluid extraction
dipropylene with
glycol andCO 2 ofathe
from aerial
topical part of O. majorana
formulation
–skin lotion. The supercritical fluid extraction with CO2 of the aerial part of O. majorana
was carried out at 40 °C and 10 MPa. The extract was incorporated in the matrix in 0.01%
(v/v). Dynamic headspace (DHS) coupled to GC-FID/MS was used to measure species con-
centration over time. The odor profile in the presence of glycerine was initially character-
ized by the fast release of β-caryophyllene, sabinene, p-cymene, limonene, myrcene, li-
Molecules 2021, 26, 3095 11 of 22

was carried out at 40 ◦ C and 10 MPa. The extract was incorporated in the matrix in 0.01%
(v/v). Dynamic headspace (DHS) coupled to GC-FID/MS was used to measure species
concentration over time. The odor profile in the presence of glycerine was initially char-
acterized by the fast release of β-caryophyllene, sabinene, p-cymene, limonene, myrcene,
linalyl acetate and β-phellandrene followed by a high decrease in the next 5 h. The fast
release is due to the hydrophobic nature of these compounds with low LogPow , i.e., low
affinity with the polar solvent glycerine. For example, the LogPow is 4.5 for limonene and
3.8 for linalyl acetate.
The odor intensity of the fragrance compounds was lower in dipropylene glycol (DPG)
as a consequence of its high retention ability. Sabinene was the most released compound
and after 2 h the headspace contained residual amounts of sabinene, myrcene, limonene,
β-phellandrene, γ-terpinene and β-caryophyllene. DPG can be a good solvent to prolong
the perceived scent of a fragranced product. In respect to the dipropylene glycol, the
dominant odor changed as the time increased: the mixture started to smell like linalyl
acetate (odor described as sweet, citrus, floral and woody) and then changed to myrcene
(odor described as peppery, terpene, spicy and balsam). Despite the lower headspace
amounts of myrcene compared to the remaining fragrance compounds, it will be more
perceived due to its low ODT value (4.5 × 10−2 mg/m3 ).
When the matrix is a skin lotion, the initial headspace is reached in sabinene, myrcene,
p-cymene, β-phellandrene and γ-terpinene and then a sharp decrease of concentration with
time. The most retained components over time were linalyl acetate and β-caryophyllene
but terpinen-4-ol increased after 5 h.

6.2. The Effect of Skin

The effect of the skin was studied in our lab with a Franz cell using skin prepared
from pig ears according to the protocol described elsewhere [40]. The Franz cell shown
in Figure 16 contains a donor compartment where the fragrance mixture is placed and a
Molecules 2021, 26, x FOR PEER REVIEW 13 of 23
receptor compartment separated from the donor by the membrane (skin). It allows the
study of the permeation of components through the skin and retention on the skin. At
the same time, the odor in the gas phase in the donor compartment can be followed. The
dC
Franz cell is equivalent to the Wicke–Kallenbach diffusion cell known in Chemical Reaction
Engineering [41]. Vr i,r = AK p.i (Ci,d - Ci,r ) (17)
dt

Figure 16.
Figure 16. Franz
Franzcell
cellfor
forpermeation
permeation studies of fragrance
studies mixtures
of fragrance (Reprinted
mixtures with permission
(Reprinted from
with permission from
Elsevier, Int. Journal of Biological Macromolecules 2020, 147, 150–159).
Elsevier, Int. Journal of Biological Macromolecules 2020, 147, 150–159).

The modeling of the system sketched in Figure 17 involves a mass balance for the
donor liquid solution
g

dCi,d ∂Ci 
− Vd = AK p,i (Ci,d − Ci,r ) − ADi,mix  , (13)
dt ∂z 
z =0
Molecules 2021, 26, 3095 12 of 22

with the initial condition (t = 0) Ci,d (0) = Ci,d,0 . In the above equation, A is the membrane
area, K p,i is the permeation coefficient of component i through the skin, Vd is the volume
of the donor compartment, Ci,d and Ci,r are the concentrations in the donor and receptor
Figure 16. Franz and
compartments cell for permeation
Di,mix studies of fragrance
is the diffusivity mixtures
of i in the (Reprinted
gas phase abovewith
thepermission
liquid in from
the
Elsevier,
donor Int. Journal of Biological Macromolecules 2020, 147, 150–159).
chamber.

Figure 17. Sketch

Figure17. Sketch of
of the
the Franz
Franz cell
cellsystem.
system.

The
The mass balancecoefficients
permeation in the gas phase abovefrom
calculated the liquid in the donor
infinite-dose compartment
experiments were is:
1.08 ×
10 cm/h, 8.25 × 10 cm/h and 2.15g × 10 cm/h for2 α-pinene,
−5 −6 −3
g limonene and linalool, re-
i ( z, t )
spectively. In all experiments, the∂Cfragrances were Ci
∂diluted in ethanol. This is illustrated in
= Di,mix 2 , (14)
Figure 18 for the linalool experiment, ∂z infinite-dose. ∂z
The model
with initial allowed
condition Ci (z,at0)the
g
= 0.same
Thetimeboundary to follow the concentration
conditions are: in the headspace
above the donor solution. Ultimately it would be important to separate the adsorption of
each fragrance onto the skin. The release ∂C g
is affected by the interaction between fragrances
i ( L, t )
z =
and ethanol and also by the vapor pressure ∂z L, of the = 0,
species. Vapour pressure and permea- (15)
tion coefficients in the skin were measured by Almeida et al. [42] for 14 PRM: camphor,
and at z = 0carvone,
carvacrol, the equilibrium
citronellol,at the gas–liquid
eucalyptol, interface
eugenol, geraniol, limonene, linalool, menthol,
menthone, tonalide, vanillin and α-pinene. Various models have been proposed to ad-
dress dermal absorption of fragrances g g Mi γi Pisat
Ci (0, t) =and Ci,eqdrugs
= [43–46]. xi . (16)
RT
The mass balance for the receptor compartment is:

dCi,r
Vr = AK p.i (Ci,d − Ci,r ) (17)
dt
The permeation coefficients calculated from infinite-dose experiments were 1.08 ×
10−5 cm/h, 8.25 × 10−6 cm/h and 2.15 × 10−3 cm/h for α-pinene, limonene and linalool,
respectively. In all experiments, the fragrances were diluted in ethanol. This is illustrated
in Figure 18 for the linalool experiment, infinite-dose.
Molecules 2021, 26, 3095 13 of 22
Molecules 2021, 26, x FOR PEER REVIEW 14 of 23

Figure18.
Figure 18.Cumulative
Cumulativeamount
amountofoflinalool
linaloolininthe
thereceptor
receptorcompartment
compartmentversus
versustime
timefor
forthe
theinfinite-
infinite-
dose experiment (Reprinted (adapted) with permission from Evaporation and permeation ofoffra-
dose experiment (Reprinted (adapted) with permission from Evaporation and permeation fra-
grance applied to the skin, Ind. Eng. Chem. Res., 2019, 58, 9644–9650. Copyright 2019, American
grance applied to the skin, Ind. Eng. Chem. Res., 2019, 58, 9644–9650. Copyright 2019, American
Chemical Society).
Chemical Society).
7. The
TheTrail
modelof Perfumes
allowed at the same time to follow the concentration in the headspace
aboveThe trail ofsolution.
the donor perfumes or sillageitiswould
Ultimately something we dealtowith
be important in our
separate theeveryday
adsorption life.ofIt
describes
each the onto
fragrance scented
the trail
skin.left
Theby the fragrance
release is affectedwearer.
by the It is determined
interaction by fragrances
between how long a
fragrance
and ethanoltravels
and alsoaway and
by the diffuses
vapor around
pressure thespecies.
of the wearer Vapour
[47]. Wepressure
tackledand
thispermeation
problem by
first analyzing
coefficients in thethe
skindiffusion of fragrances
were measured released
by Almeida et al.from a moving
[42] for 14 PRM:source [48].carvacrol,
camphor, First, we
started with
carvone, a 1D model
citronellol, considering
eucalyptol, eugenol,molecular
geraniol, diffusion
limonene,of a fragrance
linalool, molecule
menthol, (α-pi-
menthone,
nene) invanillin
tonalide, the air and
as the only mass
α-pinene. transport
Various modelsmechanism.
have been Considering an impermeable
proposed to address dermal
release of the fragrance m = k M C g the gas con-
absorption mass
boundary of fragrances
condition andand drugs [43–46].
constant i i i i
7.centration
The Trail in
ofaPerfumes
semi-infinite medium is:
Thet trail of perfumesì ïor sillage (t ))2
(z -isz0something (z + z0 (t ))2 üour everyday life. It
we deal with in
mimass ò the scented trail
1
Ci (z, t) =describes í exp-
left by the fragrance + exp-
wearer. ý dt by how long(18)
It is determined a
[p Di (taway
- t )]andîï diffuses i (t - t )the wearer4D - t )tackled
þ
1/2
fragranceti 2travels 4Daround i (t We
[47]. this problem by
first analyzing the diffusion of fragrances released from a moving source [48]. First, we
startedThe
withmass
a 1Devaporation rate canmolecular
model considering be calculated as of a fragrance molecule (α-pinene)
diffusion
mimass = µkimass
in the air as the only mass transport mechanism.
condition and constant release of the fragrance
Considering
M iC=ig k M C gantheimpermeable
gas
boundary
concentration
(19)
in a
i i i i
semi-infinite medium
where the mass is: coefficient ki takes into account film contributions from gas and
transfer
liquid sides. TheZtvalidation was performed ( in a diffusion tube, and a system2 ) was devel-
oped to move the
mass scented 1
source along the axial z0 (τ ))2 (Figure 19).
(z −direction (z +Results
z0 (τ )) showed that
Ci (z, t) = µi 1/2
exp − + exp − dτ (18)
experimental data 2 [
fitted
πD i ( t −
well τ )]
with the numerical − τ)
4Di (t simulation, 4Di (t −this
suggesting τ ) model as a
ti
valid tool to describe the trail of a fragrance released from a moving source for low Reyn-
oldsThe
number
mass of the order of
evaporation 10 can
rate (Figure 20).
be calculated as
In the case of a person walking at the speed of g
1.34 m/s in a room or corridor inside
a building, 3D models are required and µimassmass
= k itransport
Mi Ci (19)
of the perfume to the surrounding
air will be dominated by turbulent diffusion or eddy diffusion Dt which is two orders of
where the mass transfer coefficient ki takes into account film contributions from gas and
magnitude higher than molecular diffusion. For the 3D case the solution is:
liquid sides. The validation was performed in a diffusion tube, and a system was developed
to movet the scented 1 sourceïì along(rthe 0 (t ))
- raxial (r - r1 (t ))2 ü Results showed that
2direction (Figure 19).
= mimass ò
Ci (z, t) experimental data fitted1/2well exp- +
í with the numerical simulation, suggesting
exp- ý dt this model (20)
as
ti
a valid tool 8 [ to describe
p D i (t - t )] theî trail of a fragrance
ï 4D i (t - t ) 4D i (t
released from - t ) þ
a moving source for low
Reynolds number of the order of 10 (Figure 20).
where the source path is r0 (t) = [ x0 (t), y0 (t), z0 (t)] , r0 (t) = [ x0 (t), y0 (t), z0 (t)] and
T T

r1 (t) = [ x0 (t), y0 (t),-z0 (t)] .

T
Molecules
Molecules 2021,26,
Molecules2021,
2021, 26,3095
26, x xFOR
FORPEER
PEERREVIEW
REVIEW 1415of
15 ofof23
2223

Figure 19. Diffusion tube and moving source.

Figure
Figure19.
19.Diffusion
Diffusiontube
tubeand
andmoving
movingsource.
source.

Figure
Figure20.20.Simulated
Simulatedandandexperimental
experimentalgas gasconcentration profiles ofofa-pinene versus distance at t =
Figure 20. Simulated and experimental gas concentration
concentrationprofiles
profiles ofa-pinene
a-pineneversus
versusdistance
distanceatatt =
100 s of a source moving at 1.34 × 10 m/s
−2−2 and Dα-pin = 6.04 × 10 m
−6−6 2/s (Reprinted with
100 s of a source moving at 1.34 × 10 m/s and Dα-pin = 6.04 × 10 m /s (Reprinted with permission
2 permission
t = 100 s of a source moving at 1.34 × 10−2 m/s and Dα-pin = 6.04 × 10−6 m2 /s (Reprinted with
from
fromAIChEJ
AIChEJ2018,
2018,64,
64,2890–2897,
2890–2897,John
JohnWiley
WileyandandSons).
Sons).
permission from AIChEJ 2018, 64, 2890–2897, John Wiley and Sons).
As an example for
foran
aninitial
initialposition ofofthe moving source r0r(0,
0 (0,0,0,1.50
1.50m) m)isis shown
InAs ancase
the example
of a person walking position
at the speed theof moving
1.34 m/ssource
in a room or corridor shown
inside a
in
in Figure
Figure 21a, and
21a, and the
theconcentration
concentration profiles
profiles at t
at =
t =200
200s for
s a
for person
a person
building, 3D models are required and mass transport of the perfume to the surrounding with
with a fragrance
a fragrance α-α-
pinene
air willmoving
pinene moving
be atat1.34
dominated m/s
1.34bym/sare
areshown
turbulentshown ininFigureor 21b.
Figure
diffusion 21b. diffusion Dt which is two orders of
eddy
These
These models
models can
canhelp
helpthe
thefragrance
fragrance
magnitude higher than molecular diffusion. For the industry
industry totoachieve
3Dachieve the
case the desired
the desiredtrail
solution trailofoffragrance
is: fragrance
more
morequickly
quicklyandandefficiently.
efficiently.The
Thesearch
searchforfordevices
deviceswhich
whichcan canbebeused
usedtotomeasure
measurethe the
sillage of fragranced t products continues
sillage of fragranced products continues [49].
( [49]. )
1 (r − r0 (τ ))2 (r − r1 (τ ))2
Z
Ci (z, t) = µimass exp − + exp − dτ (20)
t
8[πDi (t − τ )]1/2 4Di (t − τ ) 4Di (t − τ )
i

where the source path is r0 (t) = [ x0 (t), y0 (t), z0 (t)] T , r0 (t) = [ x0 (t), y0 (t), z0 (t)] T and
r1 (t) = [ x0 (t), y0 (t), −z0 (t)] T .
As an example for an initial position of the moving source r0 (0, 0, 1.50 m) is shown in
Figure 21a, and the concentration profiles at t = 200 s for a person with a fragrance α-pinene
moving at 1.34 m/s are shown in Figure 21b.
 Molecules 2021, 26, 3095 15 of 22
Molecules 2021, 26, x FOR PEER REVIEW 16 of

(a) (b)
Figure
Figure 21.21.
3D 3D model-
model- moving
moving sourcesource at(a)
at 1.50 m 1.50
andmconcentration
(a) and concentration
profiles for 3profiles
values offor 3 values of diff
diffusivity
sivity evaluated
evaluated at and
at z = 1.60 m z = t1.60 ms and
= 200 t = 200 s with
(b) (Reprinted (b) (Reprinted withAIChEJ
permission from permission from
2018, 64, AIChEJ 2018, 6
2890–2897.
2890–2897.
John Wiley andJohn Wiley and Sons).
Sons).

These models
8. Flavor can help the fragrance industry to achieve the desired trail of fragrance
Engineering
more quickly and efficiently. The search for devices which can be used to measure the
sillage The methodology
of fragranced developed
products continuesfor perfume engineering can be extended to taste eng
[49].
neering or flavor engineering. We previously defined the odor detection threshold as th
8.minimum
Flavor Engineering
gas phase concentration at which an odorant is detected by the nose; and th
odorThe value OV as the
methodology ratio between
developed the concentration
for perfume engineering can of odorant
be extended in gas-phase
to taste en-divided b
gineering
the ODT. or Similarly,
flavor engineering.
we define Wethepreviously defined thethreshold
flavor detection odor detection FDTithreshold as
as the lowest liqu
the minimum gas phase concentration at which an odorant is
concentration of component i at which it is detected by the retronasal route; the flav detected by the nose; and
the odorFV
value value OV as the ratio between the concentration of odorant in gas-phase divided
i is then the ratio between the gas phase concentration and its FDT.
by the ODT. Similarly, we define the flavor detection threshold FDTi as the lowest liquid
The idea is to predict the sensory quality of flavored products based on their g
concentration of component i at which it is detected by the retronasal route; the flavor
value FVcomposition
phase with the help of psychophysical models and olfactory descriptors. Th
i is then the ratio between the gas phase concentration and its FDT.
first
Thetested
idea iscase of flavored
to predict products
the sensory qualitywas fruit juices
of flavored (peach,
products based lemon,
on theirmangogas phaseand pinea
ple). For each fruit juice, the headspace gas phase composition
composition with the help of psychophysical models and olfactory descriptors. The first was measured by chrom
tography.
tested case ofThe testsproducts
flavored were extendedwas fruittojuices
binary and lemon,
(peach, ternarymango mixtures and of fruit juices
pineapple). For[50]. Od
each
andfruit juice,radars
flavor the headspace gas phase composition
were constructed with families wasofmeasured by chromatography.
fruity, sweet, green, woody, fres
The tests citrus,
spicy, were extended to binary
fatty, ripe tones and ternary
and mixtures
validated byofafruit juices [50].
sensory Odor and
evaluation offlavor
consumers
radars
shown in Figure 22 for peach juice. It should be noted that the headspacecitrus,
were constructed with families of fruity, sweet, green, woody, fresh, spicy, of peach jui
fatty, ripe tones and validated by a sensory evaluation of consumers as shown in Figure 22
contained various compounds identified as ethyl butyrate (fruity, sweet, spicy) isoam
for peach juice. It should be noted that the headspace of peach juice contained various
acetate (sweet,
compounds identifiedfruity, ripe),butyrate
as ethyl benzaldehyde (woody,
(fruity, sweet, fruity,
spicy) isoamyl sweet),
acetate ethyl
(sweet,hexanoate
fruity, (swe
fruity,
ripe), green), limonene
benzaldehyde (citrus),
(woody, fruity, linalool
sweet), ethyl(citrus,
hexanoatesweet,
(sweet,woody). When limonene
fruity, green), one component
allocated to just one family, the weight is 1; when allocated
(citrus), linalool (citrus, sweet, woody). When one component is allocated to just one family, to three families the weigh
are
the 0.6, 0.3
weight is 1;andwhen 0.1.allocated
The ODTs for families
to three the six components
the weights areare 0.6,3.35 × 100.1.mg/m
0.3 and −4 3
The ODTs for ethyl b
for the six4.99
tyrate, components
× 10−1 forare isoamyl 10−4 mg/m
3.35 × acetate, 3 for ethyl butyrate, 4.99 × −2
6 for benzaldehyde, 1.5 × 1010for −1 for isoamyl
ethyl hexanate, 6.
acetate, × 10 −2 −3 ethyl hexanate, 6.19 × 10−1 for limonene and
× 10−1 6forfor benzaldehyde,
limonene and1.5 9.33 × 10for for linalool. The flavor detection thresholds FD
9.33 × 10−3for
(mg/kg) forthe
linalool.
same−six Thecomponents
flavor detection are thresholds
1.8 × 10−3, FDT
5.7 (mg/kg)
×i 10 −2, 5.3 × for
10the same
−1, 8.0 × 10six
−3, 2.1 × 10
components are 1.8 × 10 , 5.7 × 10 , 5.3 × 10 , 8.0 × 10 , 2.1 × 10 and 3.3 × 10−3 ,
3 − 2 − 1 − 3 − 1
and 3.3 × 10 , respectively.
−3
respectively.
A review on the biotechnological production of non-volatile flavors has been pu
lished recently by Paulino et al. [51]. Methodologies to advance the understanding of fl
vor chemistry have been proposed by Menis-Henrique [52] and the development of
model mouth discussed in detail by Panda et al. [53]. Encapsulation of flavors/aromas
food applications has been discussed by Gupta et al. [54].
Molecules 2021, 26, 3095 16 of 22
Molecules 2021, 26, x FOR PEER REVIEW 17 of 23

Molecules 2021, 26, x FOR PEER REVIEW 17 of 23

Figure22.
Figure 22.Odor
Odorand
andflavor
flavorradars
radarsfor
forpeach
peachjuice
juice(experimental;
(experimental;shaded
shadedarea–predicted).
area–predicted).Reprinted
Reprinted
with permission from Ind. Eng. Chem. Res. 2018, 57, 8115−8123 Copyright 2018, American Chemical
with permission from Ind. Eng. Chem. Res. 2018, 57, 8115−8123 Copyright 2018, American Chemi-
Society).
cal Society).

9. Looking
A reviewAhead
on the biotechnological production of non-volatile flavors has been published
recently More
by than
Paulinotwoetdecades
al. [51].have past since research
Methodologies on perfume
to advance engineeringof
the understanding started
flavorat
Figure
LSRE. 22.
chemistry Odor23
Figure
have and flavor
shows
been radars
the
proposed byfor
timeline peach
andjuice
the (experimental;
Menis-Henrique main[52] shaded
contributors
and the to area–predicted).
the developments
development Reprinted
of a model of
with
perfumepermission
design: from Ind.
perfumery Eng. Chem.
ternary Res. 2018,
diagram 57,
(PTD)8115−8123
and Copyright
extension to 2018, American
quaternary and Chemical
qui-
mouth discussed in detail by Panda et al. [53]. Encapsulation of flavors/aromas in food
Society).
nary systems
applications has(PQD
been and PQ2D),byprediction
discussed of [54].
Gupta et al. odor thresholds and VLE, performance of
perfumes—1D and 3D diffusion models—and trail of perfumes, classification of per-
9.9. LookingAhead
Looking
fumes—perfumery
Ahead radar, effect of matrix and skin, extension to flavor/taste engineering.
We MoreMorenot
have than
than two
two
listed decades
decades
many have
have
trainees pastsince
past
coming since
from research
research
France,on on perfume
perfume
Poland, Spain engineering
engineering
or started at
Brazil. started
atLSRE.
LSRE.Figure
Figure23 23shows
shows the the timeline
timeline andand the
the main
maincontributors
contributorstotothe thedevelopments
developments of
of perfumedesign:
perfume design:perfumery
perfumeryternaryternarydiagram
diagram(PTD)(PTD) and
and extension
extension to quaternary
quaternary andand qui-
quinary systems(PQD
nary systems (PQDand and PQ2D),
PQ2D), prediction
predictionofofodor odorthresholds
thresholdsandand
VLE, performance
VLE, of of
performance
perfumes—1D
perfumes—1D and 3D diffusion
and 3D diffusionmodels—and trail of perfumes,
models—and classification
trail of perfumes, of perfumes—
classification of per-
perfumery radar, effect
fumes—perfumery of matrix
radar, and
effect of skin,
matrixextension
and skin, to flavor/taste
extension toengineering.
flavor/tasteWe have
engineering.
not listed many trainees coming from France, Poland, Spain or
We have not listed many trainees coming from France, Poland, Spain or Brazil. Brazil.

Figure 23. Contributors for research in Perfume Engineering started by Alírio Rodrigues and Vera
Mata at LSRE.

Looking ahead our research directions are focused on:

(i) The importance of diffusivity on the trail of perfumes or sillage
Brahms and Benaim [47] consider that in a perfume there are top notes and base notes
and the heart (middle) notes come both from top and base notes molecules with fast dif-
Figure
Figure 23.(sprinters)
23.
fusivity Contributors
Contributors for
orfor
slowresearch
research in Perfume
in Perfume
diffusivity Engineering
Engineering
(long runners). Thisstarted
started
isby by Alí
anAlírio rio Rodrigues
Rodrigues
interesting and
view and
re-Vera
Vera
that
Mata
Mata at LSRE.
at LSRE.
quires a closer look at the measurements of diffusivities [55–57] and molecular design.

Looking ahead our research directions are focused on:

(i) The importance of diffusivity on the trail of perfumes or sillage
Brahms and Benaim [47] consider that in a perfume there are top notes and base notes
and the heart (middle) notes come both from top and base notes molecules with fast dif-
Molecules 2021, 26, 3095 17 of 22

Looking ahead our research directions are focused on:

(i) The importance of diffusivity on the trail of perfumes or sillage
Brahms and Benaim [47] consider that in a perfume there are top notes and base
notes and the heart (middle) notes come both from top and base notes molecules with fast
diffusivity (sprinters) or slow diffusivity (long runners). This is an interesting view that
requires a closer look at the measurements of diffusivities [55–57] and molecular design.
This reference to the importance of diffusivity goes back to the work of Mookerjee
et al. [58,59] on the Aura of Aroma. The Aura of Aroma® (AoA) technology consists
of sampling the molecules from the air surrounding a liquid fragrance by Solid-Phase
Micro-Extraction (SPME). The SPME needle should be as close as possible to the source,
without ever touching it, and the sampling period is typically comprised between 12 –1 h.
In one of the first references to the AoA, the authors state that it is almost exclusively
dependent on the species diffusivity, and independent from its “boiling point, molecular
weight, odour threshold, or odour value”. Nevertheless, our previous studies suggest
that VLE is expected to have an important role in the AoA. To test this hypothesis, the
composition of the vapor phase in equilibrium with a reconstituted orchid (dendrobium
superbum orchid) liquid fragrance was computed by applying modified Raoult’s Law and
estimating the activation coefficients of all the species through the UNIFAC model (details
can be found on the Supplementary Materials). The results were then compared with the
composition of the AoA reported by Mookerjee et al. for this fragrance (Table 1).

Table 1. Comparison between the Aura of Aroma® measured for a reconstituted orchid (dendrobium
superbum orchid) liquid fragrance and the gas-phase composition estimated from VLE.

VLE Gas-Phase
Species Oil (wt%) Aura of Aroma® (wt%)
Composition (wt%)
Benzyl acetone 0.02 0.03 0.17
Benzyl acetate 0.20 5.20 5.55
Linalool 2.20 34.10 57.05
Raspberry ketone 11.90 1.70 4.11
2-Tridecanone 0.02 5.50 0.04
2-Pentadecanone 69.00 33.50 25.62
Ethyl myristate 14.80 8.50 4.57

As stated by Mookerjee et al., the composition of the AoA is considerably different

from the composition of the liquid fragrance. For instance, the relative amounts of benzyl
acetone, benzyl acetate linalool and 2-Tridecanone increase from the fragrance to the AoA
(1.5, 26.0, 15.5 and 275 times, respectively). On the other hand, the relative amounts of
oxyphenolon, 2-pentadecanone and ethyl myristate decrease (by 0.1, 0.5 and 0.6) from
the liquid fragrance to the AoA. The gas phase composition values estimated from the
VLE were slightly different, but respected the same general trends. The largest deviations
relative to the AoA were observed for Linalool, 2-Tridecanone and 2-Pentadecanone.
Through this rather simple approach, one is led to believe that, although VLE calculations
alone do not describe the AoA accurately, they allow us to predict if the relative amount
of a species in the gas phase is expected to increase or decrease relative to the liquid
phase. Including the effect of the compound’s diffusivity would expectedly improve the
preliminary estimations obtained through the VLE.
A closer look at this topic is required using tools to track perfume composition and
sillage [60]. Related to this topic is the area of olfactive marketing where CFD tools
are important.
(ii) Formulation of perfumes and fragranced products
The methodology of perfumery ternary diagram (PTD) was a pioneering idea that
can be extended to fragrance mixtures of N components to find compositions delivering
a certain smell. It can be further elaborated to include the effect of skin on the evapora-
Molecules 2021, 26, 3095 18 of 22

tion of perfumes. The Perfumery Radar can be extended to other areas (wines) and the
methodology extended to taste/flavor engineering.
Zhao et al. [61] applied a model for the prediction of intensity and character of
fragrances, across three main consumer touchpoints for the laundry process. They used
headspace gas chromatography data with psychophysical models (Steven’s power law)
and tested with a trained sensory panel. The authors claim it is the first time that the
approach was validated under realistic conditions for a multi-component perfume mixture
diluted in a structured product base. The formulation of fragrance products to be applied
in detergents is described in the patent by Teixeira et al. [62] where specific terminology
(experimental velocity—distance traveled 60 cm divided by the speed of diffusivity— time
needed for olfactory detection at 60 cm of the mixing point) is used to classify fragrance
ingredients.
Fragranced products include microcapsules of polyurethane-urea with perfume as a
core material for textile applications such as perfumed suits [63] or eco-friendly microcap-
sules based on chitosan and Arabic gum [64], a research area at LSRE. Microencapsulation
of fragrances is also a way of increasing the life of fragranced products [65].
It is important to say that the engineering tools developed in our group have some
challenges to overcome: the prediction of the odor of chiral molecules and the prediction
of the odor quality of a molecule, which may change with its concentration. However,
having a database with the descriptors of enantiomers allow the use of the engineering
tools presented, and VLE methods such as COSMO-SAC [18] may help in predicting the
gas phase composition of chiral molecules from a liquid mixture.
It should be said that many odorant molecules are chiral. Leffingwell database collects
more than 400 enantiomers and their odors; about 60% of the pairs have similar smells and
40% smell differently. One example is limonene: R-limonene smells orange and S-limonene
smells lemon; another example is carvone: R-carvone smells mint and S-carvone smells
like caraway. Several attempts have been made over the years to predict the smell of
chiral molecules starting with the “shape” theory: “lock and key” between the odorant
and receptor based on Pauling and Delbruck idea [66]. The shape alone does not explain
reality, otherwise, virtually all pairs should smell different. Later, the pros and cons of the
vibrational theory of Wright [67,68] and Dyson [69], later taken by Turin [70] and reviewed
by Meierhenrich et al. [71] were also considered. Other researchers, such as Brookes et al.,
tried to correlate odor with molecular flexibility claiming that flexibility allows left and
right-handed molecules to be distinguished [72].
Another point to consider is when the odor quality changes with odorant concentration
as already reported by perfumers. One example is indole which smells floral at low
concentrations and putrid at high concentrations. Tentative explanation pointed out shifts
in the patterns of glomeruli activated by the odorant although other reports see no shift in
location and simply increase in the number of glomeruli activated at low concentration of
odorant [73]. Nevertheless, the engineering tools presented are still useful for predicting the
odor from a liquid mixture if a database with descriptors of chiral molecules are available.
(iii) Artificial intelligence (AI) and perfume design
The advent of fast computing, data digitalization, cloud data storage systems, and
several other tools from Industry 4.0 is changing several industrial paradigms. Among
these technologies, Artificial Intelligence (AI) is presented as the possibility for machines
imitating the intelligence and behavior of humans. AI is a key technology that can make
use of the increasing big databases to extract useful information and disrupt the way that
products are designed and developed. The concern about the potential of AI has also been
an increasing issue in the literature [74]. Big-data is an issue that is becoming common
in several sectors. Modern technology has allowed the generation and storage of a huge
amount of data, which represents a potential to be explored.
The perfume industry is not an exception in this process, and it is possible to note
a recent movement in this field towards the application of AI in perfume development.
For instance, predicting the relationship between the structure of a molecule and its odor
Molecules 2021, 26, 3095 19 of 22

(quantitative structure-odor relationship, QSOR), is a difficult task [75,76]. One of the first
reports in the direction is presented by Zhang, L. et al. [77], an AI model was trained to
learn fragrance molecule classification, building a computer-aided molecular screening tool.
The proposed model demonstrated a remarkable accuracy in performing the screening.
Yu et al. [78] present a comprehensive review about the application of computer-based
strategies in the design of experimental designs to flavor and sensory analysis. The referred
work shows the potential of these techniques, including Artificial Neural Networks (ANN),
for food flavor applications.
This topic has also caught the interest of multi-billionaire companies, such as ALPHA-
BET. Sanchez-Lengeling et al. [79] developed graph neural networks, in ALPHABET’s
subsidiary, Google, to quantify the relationship between molecular structure and odor
(quantitative structure-odor relationship, QSOR).
AI is already in the fragrances business, the International Business Machines (IBM)
working together with Symrise and O Boticário launched the project Phylira, to develop
an AI-based tool to create perfumes using their big databases. The support from IBM
on predicting natural language descriptions of mono-molecular odorants was published
recently [80]. In this same line, Givaudan launched, in 2019, the project Carto, an AI-
powered tool that brings science and technology to the development of new perfume
formulation. It uses the proprietary ingredients ‘Odour Value Map’ to maximize the
olfactive performance in the final formula.
It is interesting to mention the DREAM Olfaction Prediction Challenge-a crowd-
sourced competition to develop models that can predict how a molecule smells from its
physical and chemical features. Results from the DREAM Consortium were published
recently in Science [81].
Most recently, a step toward a more complex application of AI in the Flavour and
Fragrance field was presented by Zhang, X. et al. [82] and Santana et al. [83]. In Zhang’s
work, the authors proposed a strategy to develop AI models to be applied in an optimiza-
tion problem that searches to identify a fragrance formulation to deliver a certain odor.
On the other hand, Santana’s work presents one of the firsts application reports of a more
sophisticated AI tool, Deep Learning (DNN), to address dynamic-related problems found
in the perfume formulation. The DNNs are a group of machine learning techniques that use
types of more complex architectures of artificial neural networks to solve difficult problems.
The authors proposed a framework that makes use of DNN models and meta-heuristic
optimization (stochastic optimization algorithm) to systematically formulate fragrances,
considering their behavior through time and space.
This research line driven by Idelfonso Nogueira (LSRE) aims in the future to build a
novel smart Cyber–Physical System (CPS) for the on-demand design and production of
perfumes. It will be based on current emerging technologies: systems automation, artificial
intelligence and real-time optimization (RTO), and get them to work harmoniously in a
CPS enabled by the Internet of Things (IoT).

Supplementary Materials: The following are available online, Table S1: Physical properties of the
studied compounds; Table S2: UNIFAC sub-groups frequency table for the studied compounds.
Author Contributions: Conceptualization, A.E.R.; writing—original draft preparation, A.E.R. and
R.P.V.F.; writing—review and editing, A.E.R., R.P.V.F. and I.N.; supervision, A.E.R. All authors have
read and agreed to the published version of the manuscript.
Funding: This work was financially supported by the Base Funding-UIDB/50020/2020 of the Asso-
ciate Laboratory LSRE-LCM-funded by national funds through FCT/MCTES (PIDDAC). Financial
support of NORTE-01-0145-FEDER-000006f FCT–Fundação para a Ciência e Tecnologia under the
CEEC Institucional program is also acknowledged.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Molecules 2021, 26, 3095 20 of 22

Acknowledgments: One of us (A.E.R.) acknowledges the contribution of Miguel A. Teixeira for the
preparation of a short course on perfume engineering given at COBEQ 2014 in Florianopolis which
helped in the writing of this manuscript.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Carles, J. A method of creation in perfumery. Soap Perfum. Cosmet. 1962, 35, 328–335.
2. Leffingwell & Associates. Available online: http://www.leffingwell.com/top_10.htm (accessed on 6 April 2021).
3. Cussler, E.; Moggridge, G. Chemical Product Design, 2nd ed.; Cambridge University Press: Cambridge, UK, 2011.
4. Costa, R.; Moggridge, G.D.; Saraiva, P.M. Chemical product engineering: An emerging paradigm within chemical engineering.
AIChE J. 2006, 52, 1976–1986. [CrossRef]
5. Rodrigues, A.E.; Da Silva, F.A.; Granato, M.; Lamia, N.; Grande, C.; Gomes, P.S. Development of propane/propene separations:
From molecular simulation to process development. In Proceedings of the XXXIII Reunion Iberica de Adsorcion, Madrid, Spain,
22–24 September 2008.
6. Teixeira, M.A.; Rodríguez, O.; Gomes, P.; Mata, V.; Rodrigues, A.E. Perfume Engineering, Butterworth-Heineman; Elsevier: Oxford,
UK, 2013.
7. Axel, R. Scents and Sensibility: A molecular logic of olfactory perception (Nobel lecture). Angew. Chem. Int. Ed. 2005, 44,
6110–6127. [CrossRef]
8. Buck, L.B. Unraveling the sense of smell (Nobel lecture). Angew. Chem. Int. Ed. 2005, 44, 6128–6140. [CrossRef] [PubMed]
9. Devos, M.; Rouault, J.; Laffort, P. Standardized Olfactory Power Law Exponents; Editions Universitaires Sciences: Dijon, France, 2002.
10. Devos, M.; Patte, F.; Roualt, J.; Laffort, P.; van Gemert, L.J. Standardized Human Olfactory Thresholds; Oxford University Press, Inc.:
New York, NY, USA, 1990; pp. 1–165.
11. Van Gemert, L.J. Odour Thresholds. In Complications of Odour Threshold Values in Air, Water and Other Media (Edition 2003);
Oliemans Punter & Partners BV: Utrecht, The Netherlands, 2003.
12. Calkin, R.; Jellinek, S. Perfumery: Practice and Principles; John Wiley & Sons: New York, NY, USA, 1994.
13. Stevens, S.S. On the physiological law. Psychol. Rev. 1957, 64, 153–181. [CrossRef]
14. Wakayama, H.; Sakasai, M.; Yoshikawa, K.; Inoue, M. Method for Predicting Odor Intensity of Perfumery Raw Materials Using
Dose−Response Curve Database. Ind. Eng. Chem. Res. 2019, 58, 15036–15044. [CrossRef]
15. Green, B.G.; Shaffer, G.S.; Gilmore, M.M. Derivation and Evaluation of a Semantic Scale of Oral Sensation Magnitude with
Apparent Ratio Properties. Chem. Senses 1993, 18, 683–702. [CrossRef]
16. Green, B.G.; Dalton, P.; Cowart, B.; Shaffer, G.; Rankin, K.; Higgins, J. Evaluating the ‘Labeled Magnitude Scale’ for Measuring
Sensations of Taste and Smell. Chem. Senses 1996, 21, 323–334. [CrossRef]
17. Teixeira, M.A.; Rodriguez, O.; Mota, F.L.; Macedo, E.A.; Rodrigues, A.E. Evaluation of group-contribution methods to predict
VLE and odor intensity of fragrances. Ind. Eng. Chem. Res. 2011, 50, 9390–9402. [CrossRef]
18. Xavier, V.; Staudt, P.; Soares, R.D.P. Predicting VLE and odor intensity of matrices containing fragrances with COSMO-SAC. Ind.
Eng. Chem. Res. 2020, 59, 2145–2154. [CrossRef]
19. Teixeira, M.A.; Rodriguez, O.; Rodrigues, A.E.; Selway, R.; Riveroll, M.; Chieffi, A. Prediction and experimental measurement of
the odor intensity of fragrance mixtures. Ind. Eng. Chem. Res. 2013, 52, 963–971. [CrossRef]
20. Rodriguez, O.; Teixeira MARodrigues, A.E. Prediction of odor detection thresholds using partition coefficients. Flavor Fragr. J.
2011, 26, 421–428. [CrossRef]
21. Mata, V.G.; Gomes, P.B.; Rodrigues, A.E. Science behind perfume design. In Proceedings of the Second European Symposium on
Product Technology (Product Design and Technology), Groningen, The Netherlands, 15 September 2004.
22. Mata, V.G.; Gomes, P.B.; Rodrigues, A.E. Engineering perfumes. AIChE J. 2005, 51, 2834–2852. [CrossRef]
23. Mata, V.G.; Gomes, P.B.; Rodrigues, A.E. Perfumery ternary diagram (PTD): A new concept applied to the optimization of
perfume compositions. Flavour Fragr. J. 2005, 20, 465–471. [CrossRef]
24. Mata, V.G.; Gomes, P.B.; Rodrigues, A.E. A new methodology for the definition of odor zones in perfumery ternary diagrams.
AIChE J. 2006, 52, 2938–2948. [CrossRef]
25. Teixeira, M.; Rodriguez, O.; Mata, V.; Rodrigues, A.E. Perfumery Quaternary Diagrams for Engineering Perfumes. AIChE J. 2009,
55, 2171–2185. [CrossRef]
26. Teixeira, M.A. Perfume Performance and Classification: Perfumery Quaternary-Quinary Diagram (PQ2D) and Perfumery Radar.
Ph.D. Thesis, University of Porto, Porto, Portugal, 2011.
27. Teixeira, M.A.; Rodriguez, O.; Mata, V.G.; Rodrigues, A.E. The diffusion of perfume mixtures and odor performance. Chem. Eng.
Sci. 2009, 64, 2570–2589. [CrossRef]
28. Pereira, J.; Costa, P.; Loureiro, J.M.; Rodrigues, A.E. Modelling Diffusion of Fragrances: A Radial Perspective. Can. J. Chem. Eng.
2019, 97, 351–360. [CrossRef]
29. Almeida, R.N.; Rodrigues, A.E.; Vargas, R.M.; Cassel, D. Radial diffusion model for fragrance materials: Prediction and validation.
AIChE J. 2021. under review.
30. Edwards, M. The Fragrance Wheel. Available online: http://www.fragrancedirectory.info/usadirectory/ (accessed on 31
October 2012).
Molecules 2021, 26, 3095 21 of 22

31. DROM. Drom Fragrance Circle. Available online: http://www.drom.com/ (accessed on 31 December 2011).
32. Brechbill, G.O. A Reference Book on Fragrance Ingredients; Creative Endeavor Books: Hackensack, NJ, USA, 2006.
33. Surburg, H.; Panten, J. Common Fragrance and Flavors Materials-Preparation, Properties and Uses; Wiley-VCH: Weinheim, Ger-
many, 2006.
34. The Good Scents Company. Available online: http://www.thegoodscentscompany.com/ (accessed on 31 May 2010).
35. AIHA. Odor Thresholds for Chemicals with Established Health Standards; American Institute Hygiene Association: Fairfax, VA,
USA, 2013.
36. Teixeira, M.; Rodriguez, O.; Rodrigues, A.E. Perfumery Radar: A predictive tool for Perfume Family classification. Ind. Eng.
Chem. Res. 2010, 49, 11764–11777. [CrossRef]
37. Costa, P.; Velasco, C.; Loureiro, J.; Rodrigues, A.E. Effect of cosmetic matrices on the release and odor profiles of the supercritical
CO2 extract of Origanum majorana L. Int. J. Cosmet. Sci. 2016, 38, 364–374. [CrossRef]
38. Costa, P.; Teixeira, M.; Lièvre, Y.; Loureiro, J.; Rodrigues, A.E. Modelling fragrance components release from a simplified matrix
used in toiletries and household products. Ind. Eng. Chem. Res. 2015, 54, 11720–11731. [CrossRef]
39. Costa, P.; Teixeira, M.; Mestre, G.; Carneiro, L.; Loureiro, J.; Rodrigues, A.E. Predicting Vapor-Phase Concentrations for the
Assessment of the Odor Perception of Fragrance Chemicals Diluted in Mineral Oil. Ind. Eng. Chem. Res. 2017, 56, 8767–8777.
[CrossRef]
40. Almeida, R.; Costa, P.; Pereira, J.; Cassel, E.; Rodrigues, A.E. Evaporation and permeation of fragrances applied to skin. Ind. Eng.
Chem. Res. 2019, 58, 9644–9650. [CrossRef]
41. Wicke, E.; Kallenbach, R. Die Oberflachendiffusion von Kohlendioxyd in aktiven Kohlen. Kolloid Z. 1941, 97, 135–151. [CrossRef]
42. Almeida, R.N.; Hartz, J.G.; Costa, P.; Rodrigues, A.E.; Vargas, R.M.; Cassel, E. Permeability coefficients and vapor pressure
determination for fragrance materials. Int. J. Cosmet. Sci. 2021, 43, 225–234. [CrossRef] [PubMed]
43. Behan, J.M.; MacMaster, A.; Perring, K.; Tuck, K.M. Insight into how skin changes perfume. Int. J. Cosmet. Sci. 1996, 18, 237–246.
[CrossRef]
44. Vuilleumier, J.; Flament, I.; Sauvegrain, P. Headspace analysis study of evaporation rate of perfume ingredients applied onto skin.
Int. J. Cosmet. Sci. 1995, 17, 61–76. [CrossRef]
45. Schwarzenbach, R.; Bertschi, L. Models to assess perfume diffusion from skin. Int. J. Cosmet. Sci. 2001, 23, 85–98. [CrossRef]
46. Tibaldi, R.; ten Berge, W.; Drolet, D. Dermal absorption of chemicals: Estimation by IH SkinPerm. J. Occup. Environ. Hyg. 2014, 11,
19–31. [CrossRef] [PubMed]
47. Benaim, C.; Brahms, J. Sillage in perfumery. In Proceedings of the World Perfumery Congress, Nice, France, 5–7 June 2018.
48. Pereira, J.; Costa, P.; Coimbra, M.C.; Rodrigues, A.E. The trail of perfumes. AIChE J. 2018, 64, 2890–2897. [CrossRef]
49. Luetkenhaus, M.; Saint-Paul, A.; Singer, M.; Bensamou, S.; Carrasco, C. Device for Assessing the Sillage of Fragrance Materials.
U.S. Patent 10,794,798, 6 October 2020.
50. Monteiro, A.; Costa, P.; Loureiro, J.; Rodrigues, A.E. Flavour engineering—A methodology to predict sensory qualities of flavored
products. Ind. Eng. Chem. Res. 2018, 57, 8115–8123. [CrossRef]
51. Paulino, B.N.; Sales, A.; de Oliveira Felipe, L.; Pastore, G.M.; Molina, G.; Bicas, J.L. Biotechnological production of non-volatile
flavor compounds. Curr. Opin. Food Sci. 2021, 41, 26–35. [CrossRef]
52. Menis-Henrique, M.E.C. Methodologies to advance the understanding of flavor chemistry. Curr. Opin. Food Sci. 2020, 33, 131–135.
[CrossRef]
53. Panda, S.; Chen, J.; Benjamin, O. Development of model mouth for food oral processing studies: Present challenges and scopes.
Innov. Food Sci. Emerg. Technol. 2020, 66, 102524. [CrossRef]
54. Gupta, S.; Khan, S.; Muzafar, M.; Kushwaha, M.; Yadav, A.; Gupta, A.P. Encapsulation: Entrapping essential oil/flavors/aromas
in food, Chapter 6. In Encapsulations; Elsevier: Amsterdam, The Netherlands, 2016; pp. 229–268.
55. Tovar, T.M.; Mahlez, J.J.; Knox, C.K.; LeVan, D.M. 110th Anniversary: Molecular Structure Effects on Mass Transfer of C10
Hydrocarbons in BPL Activated Carbon. Ind. Eng. Chem. Res. 2019, 58, 15271–15279. [CrossRef]
56. Jeylisse Castañer, J.; Ramirez, C.A. Binary gas diffusivity estimates from transient, one-dimensional sublimation–diffusion
experiments in a spherical enclosure. Chem. Eng. Commun. 2018, 205, 1167–1192. [CrossRef]
57. Jaime, M.D.S.; Maisonet, S.; Ramírez, C. Effect of Liquid Phase Composition on the Experimental Determination of Binary Gas
Diffusivities in an Isothermal Stefan Diffusion Column. Chem. Eng. Commun. 2020, 12, 1658–1684. [CrossRef]
58. Mookerjee, B.D.; Trenkle, R.W.; Wilson, R.A. System for Perfume Creation Using Aroma Emission Analysis from a Living Fruit
and Flower in Close Proximity. U.S. Patent 5,367,899, 29 November 1994.
59. Mookerjee, B.D.; Patel, S.M.; Trenkle, R.W.; Wilson, R.A. Aura of aroma: A novel technology to study the emission of fragrance
from the skin. In Flavors and Fragrances; Swift, K.A.D., Ed.; Woohead Publishing Limited: Cambridge, UK, 2005.
60. Quellet, C.; Alchenberger, A. Method and Device for the Characterization of Perfume Composition. European Patent Application
EP 3521823 A1, 2019.
61. Zhau, W.; Doro, F.; Teixeira, M. A rapid approach to optimize the design of fragrances for fabric care products. Flavor Fragr. J.
2020, 35, 167–173.
62. Teixeira, M.; Unno, M.; Helweg, H.; Kay, L.M.; Shin, J.C. High Performing, High Impact Bloom Accord. International Patent
Application PCT WO 2018/071897 A1, 19 April 2018.
Molecules 2021, 26, 3095 22 of 22

63. Rodrigues, S.; Martins, I.; Fernandes, I.; Gomes, P.; Mata, M.; Barreiro, F.; Rodrigues, A.E. Scentfashion: Microencapsulated
perfumes for textile application. Chem. Eng. J. 2009, 149, 463–477. [CrossRef]
64. Sharkawy, A.; Fernandes, I.; Barreiro, M.; Rodrigues, A.E.; Shoeib, T. Aroma-loaded microcapsules with antibacterial activity for
eco-friendly textile application: Synthesis, characterization, release and green grafting. Ind. Eng. Chem. Res. 2017, 56, 5516–5526.
[CrossRef]
65. Lopes, S.; Pinheiro, C.; Costa, P.; Fernandes, I.; Barreiro, F.; Rodrigues, A.E. Chitosan-cellulose particles as deliver vehicles for
limonene fragrance. Ind. Crop. Prod. 2019, 139, 111407. [CrossRef]
66. Pauling, L.; Delbruck, M. The nature of the intermolecular forces operative in biological processes. Science 1940, 92, 77–79.
[CrossRef]
67. Wright, R.H. Stereochemical and vibrational theory of odour. Nature 1972, 239, 226. [CrossRef] [PubMed]
68. Wright, R.H. Odor and molecular vibration: Neural coding of olfactory information. J. Theor. Biol. 1977, 64, 473–474. [CrossRef]
69. Dyson, G.M. The scientific basis of odor. Chem. Ind. 1938, 57, 647–651. [CrossRef]
70. Turin, L. A Method for the Calculation of Odor Character from Molecular Structure. J. Theor. Biol. 2002, 216, 367–385. [CrossRef]
71. Meierhenrich, U.; Golebiowski, J.; Fernandez, X.; Cabrol-Bass, D. The molecular basis of olfactory chemoreception. Angew. Chem.
Int. Ed. 2004, 43, 6410–6412. [CrossRef]
72. Brookes, J.; Horsfield, A.; Stoneham, A. Odour character differences for enantiomers correlate with molecular flexibility. J. R. Soc.
Interface 2009, 6, 75–86. [CrossRef]
73. Laing, D.; Legha, P.; Jinks, A.; Hutchinson, I. Relationship between Molecular Structure, Concentration and Odor Qualities of
Oxygenated Aliphatic Molecules. Chem. Senses 2003, 28, 57–69. [CrossRef]
74. Clarke, R. Why the world wants controls over Artificial Intelligence. Comput. Law Secur. Rev. 2019, 35, 423–433. [CrossRef]
75. Chastrette, M.; Thomas-Danguin, T.; Rallet, E. Modelling the Human Olfactory Stimulus-Response Function. Chem. Senses 1998,
23, 181–196. [CrossRef] [PubMed]
76. Behrens, M.; Briand, L.; de March, C.A.; Matsunami, H.; Yamashita, A.; Meyerhof, W.; Weyand, S. Structure–Function Relation-
ships of Olfactory and Taste Receptors. Chem. Senses 2018, 43, 81–87. [CrossRef]
77. Zhang, L.; Mao, H.; Liu, L.; Du, J.; Gani, R. A machine learning based computer-aided molecular design/screening methodology
for fragrance molecules. Comput. Chem. Eng. 2018, 115, 295–308. [CrossRef]
78. Yu, P.; Low, M.Y.; Zhou, W. Design of experiments and regression modelling in food flavour and sensory analysis: A review.
Trends Food Sci. Technol. 2018, 71, 202–215. [CrossRef]
79. Sanchez-Lengeling, B.; Wei, J.N.; Lee, B.K.; Gerkin, R.C.; Aspuru-Guzik, A.; Wiltschko, A.B. Machine Learning for Scent: Learning
Generalizable Perceptual Representations of Small Molecules. arXiv 2019, arXiv:1910.10685.
80. Dário Gutiérrez, E.; Dhurandhar, A.; Keller, A.; Meyer, P.; Cecchi, G.A. Predicting natural language descriptions of mono-
molecular odorants. Nat. Commun. 2018, 9, 4979. [CrossRef]
81. Keller, A.; Gerkin, R.C.; Guan, Y.; Dhurandhar, A.; Turu, G.; Szalai, B.; Mainland, J.D.; Ihara, Y.; Yu, C.W.; Wolfinger, R.; et al.
Predicting human olfactory perception from chemical features of odor molecules. Science 2017, 355, 820–826. [CrossRef]
82. Zhang, X.; Zhou, T.; Ng, K.M. Optimization-based Cosmetic Formulation: Integration of Mechanistic Model, Surrogate Model,
and Heuristics. AIChE J. 2021, 67, e17064. [CrossRef]
83. Santana, V.; Márcio, M.; Loureiro, J.M.; Ribeiro, A.M.; Rodrigues, A.E.; Nogueira, I. Optimal fragrances formulation using a deep
learning neural network architecture: A novel systematic approach. Comput. Chem. Eng. 2021, 150. [CrossRef]