A NEW APPROACH TO CREATIVE ERFUMERY
Kazuyoshi Tachikawa POLA Laboratories, POLA Corporation 27-1 , Takashimadai, Kanagawa-ku, Yokohama, 221 Japan
Key words : perfumery, method offragrance creation, GC/MS data, exponential equation, "Golden Sequence of Numbers", factor analysis,

Introduction

A perfumer combines dozens of aromas by trial and error to create a new compound. Since Jean Carles proposed "Carles' method" in 1961 [1], no new systematic method of combining aromas to a ratio has been proposed. If we can find common structure of good perfumes, it should be possible for us to use this structure to create new fragrances. The purpose of this paper is to report a finding, which can be used in creative perfumery. In this work, gas chromatographic data for 36 successful fragrances marketed around the world were first analyzed.

1. GC and GC/MS Analysis

Table 1. shows 36 sample fragrances selected to provide a reasonable spread of fragrance types such as bouquet, chypre and oriental. The ingredient data was obtained by gas chromatography, using a carbowax-type polar column, while the ingredient data showing very small peaks were neglected.

Table-1 Sample Fragrance

No,
Code
Odor Type
No,
Code
Odor Type
No,
Code
Oaor Type
No,
Code
Odor Type
1
GIOR
Bouquet Floral
10
HITS
Chypre Fruity
19
DSIM
Bouquet Fresh
28
CHO5
Bouquet Aldehyde
2
PARI
Bouquet Floral
ll
YSAT
Chypre Animal
20
ANAI
Bouquet Fresh
29
CALA
Bouquet Aldehyde
3
LAIR
Bouquet Floral
12
OPIU
Oriental Spicy
21
LILI
Bouquet Fresh
30
CALE
Bouquet Aldehyde
4
NIAN Bouquet Floral
13
SAMS
Oriental Ambery
22
LOUL
Bouquet Sweet
31
TRES
Bouquet Floral
5
FIDJ
Bouquet Floral
14
OBSE
Oriental Antbery
23
BEAU
Bouquet Green
32
BOUC
Bouquet Sweet
6
ETER
Bouquet Floral
15
COCO
Oriental Spicy
24
KNOW
Chypre Floral
33
CHAR
Bouquet Floral
7
LIZC
Bouquet Fruity
16
JOIY
Bouquet Floral
25
EDG I
Bouquet Fruity
34
CO93
Bouquet Green
8
MISD
Chypre Animal
17
MILL
Bouquet Floral
26
CALY
Bouquet Fruity
35
CHI9
Bouquet Green
9
CABO
Chypre Animal
18
NAHE
Bouquet Floral
27
ASHl
Bouquet Fruity
36
CO92
Bouquet Fruity

2. Data Analysis

In order to examine the distribution of the ingredients, the relative concentrations and their rankings were plotted on a semi-logarithmic graph. A typical result is given in Figure 1. In all cases, the best fit of the data is represented by a straight line with negative slope. The equation of regressiontrendline is given below.

Y=ae-bx (or Y=a/exp(bX) )    (1)

where Y is the concentration, X is the ranking, a and b are the constants

Fig-1 Semi-logarithmic Graph(case of GIOR)

Rank. Ingredients Conc. Rank. Ingredients Conc. Rank. Ingredients Conc.
1 Diethyl phthalate 432575 11 Hydroxycitronellal diethyl acetal 147999 21 Benzyl salicylate 62935
2 Benzyl acetate 391380 12 Helional 132606 22 Dipro 57966
3 Iso-E-super 257459 13 Linalool 128280 23 Iso-E-super 53570
4 ƒ¿-n-Amyl cinnamic aldehyde 223337 14 Linalyl acetate 119223 24 Geranyl acetate 49685
5 Phenyl ethyl alcohol 220596 15 Benzyl benzoate 100961 25 Iso-E-super 45591
6 Methyl anthranilate 208154 16 Lilial 86240 26 L.R.T.17 45134
7 Hydroxycitronellal 190034 17 Citronellol 83399 27 iso-Propyl myristate 31658
8 Hedione 180293 18 Geraniol 80172 28 Hexadecane 28741
9 Styrallyl acetate 179345 19 d-Limonene 74839 29 Lyral 26256
10 Hexyl cinnamic aldehyde 166777 20 Cinnamic alcohol 70611 30 Musk ketone 24486

Table 2. shows the constants(a), (b) and R-squared values(R^2); characteristics of a fragrance prescription must appear in these values of (b) and (R^2). The (b) values of MITS and DSIM are much larger than those of the other fragrances. With a larger coefficient (b), which then gives a steep slope on the graph due to a marked change in the ingredients, these special two fragrances have a simple and dynamic composition. R-squared value shows the reliability of the trend line. The minimum value for CALY is 0.88, therefore, the fit of the data to eq. (1) is excellent.

Table-2 constants(a)(b) and R-squared value

Regression equation Y=a/exp(bx)
No,
Code
a
b
R2
No,
Code
a
b
R2
No,
Code
a
b
R2
1
GIOR
765475
0.1386
0.9793
13
SAMS
575295
0.1296
0.9787
25
EDGI
197569
0.1588
0.9796
2
PARI
575329
0.1554
0.9852
14
OBSE
564637
0.1596
0.9699
26
CALY
1.00E+06
0.1673
0.8833
3
LAIR
733011
0.1651
0.9555
15
COCO 468331
0.1366
0.9848
27
ASHl
667124
0.1587
0.9725
4
NIAN
144591
0.1222
0.9527
16
JOIY
1.00E+06
0.1612
0.9740
28
CHO5
263954
0.1570
0.9872
5
FIDJ
565976
0.1419
0.9768
17
MILL
381495
0.1065
0.9496
29
CALA
195828
0.1325
0.9557
6
ETER
1.00E+06
0.1666
0.9861
18
NAHE
492298
0.1275
0.9432
30
CALE
473573
0.1302
0.9910
7
LIZC
1.00E+06
0.1240
0.9666
19
DSIM
777798
0.1990
0.9215
31
TRES
661189
0.1443
0.9177
8
MISD
598474
0.0961
0.9492
20
ANAI
474021
0.1086
0.9862
32
BOUC
337421
0.1324
0.9849
9
CABO
222091
0.1341
0.9838
21
LILI
673493
0.1204
0.9862
33
CHAR
117832
0.1155
0.9151
10
MITS
168860
0.2403
0.9764
22
LOUL
432084
0.1457
0.9770
34
CO93
386759
0.1611
0.9763
ll
YSAT
1.00E+06
0.1324
0.9865
23
BEAU
671798
0.1175
0.9813
35
CHI9
130717
0.1211
0.9306
12
OPIU
1.00E+06
0.1283
0.9804
24
KNOW
590172
0.0874
0.9753
36
CO92
294815
0.1468
0.9576

3. A Study on the Essential oils

A perfume is an artificial odor. A typical example of an exponential relationship in nature is the half-life data for radioactive substances. We analyzed several data reported in Perfumer & Flavorist by Dr.BrianM.Lawrence[2]. The results shown in Figure2. and Table 3. clearly indicate that the perfumes have a structure similar to natural oils. Do perfumers unconsciously follow the rule of nature?

Fig-2 Semi-logarithmic Graph (case of Jasmin oil)

Table-3 Coefficients and R-squared Values of Natural Oils

Essential OiIs
(-b)
R^2
Jasmin
-0.1958 0.9608
Rose
-0.1565 0.9106
Ylang Ylang
-0.0945 0.9257
Lavender
-0.0799 0.9237
Patchouli
-0.3003 0.8832

4. "Golden Sequence of Numbers

Our data show that the quantity of components arranged in descending order fit the sequence of numbers obtained from the exponential equation without exception. Therefore, our discovery suggests that the sequence of numbers can be applied to facilitate fragrance creation. If we associate odorants with these numbers, it may be possible to obtain an aesthetically pleasing fragrance; the number might be called the "Golden Sequence of Numbers (hereafter cited as GSN)" like the "Golden Section" which is a universal beauty proportion found in nature.

5. Factor Analysis

In using GSN for determining the composition of a fragrance, one problem is to decide which odorant should be associated with which number. A factor analysis was then applied to find a clue to placing the odorants. In this analysis, 34 significant aroma chemicals, generally used in fragrances, were regarded as variables. There are 36 samples (cases); 34*36 data matrix becomes an object of the analysis. Factor analysis is a tool that multi-dimensional data are converted into a small-dimensional data using few factors. If the ingredients of fragrance have not been selected at random, and have been set up as the clusters, the clusters could be extracted as the factors. Table 4. shows a varimax-rotated (Interpretation of factor becomes easy by the rotation ) factor matrix. In Table 4., it is seen that each factor corresponds to general odor classification of aroma chemicals; i.e., factors 1, 2,3,4,5 and 6 are identified as fresh note, rose note, jasmin note, floral note, woody-musky note, and sweet note, respectively.

Table-4 Factor Matrix (varimax-rotated)

variables
Factor1
Factor2
Factor3
Factor4
Factor5
Factor6
Linalyl P.A.
0.89 0.13
-0.10
-0.12
0.01
0.01
Llnalool
0.80 0.01
0.14
0.06
0.08
0.36
Hyd.CitD.A
0.74 0.07
0.01
0.02
0.34
-0.10
Citronellol
-0.06
0.91 0.09
-0.05
-0.ll
-0.03
Rose P
0.02
0.84 0.07
-0.1 5
0.12
-0.18
Ben. Sail.
0.01
0.06
0.82 0.00
-0.ll
-0.01
LRT 17
0.04
-0.13
0.75 0.35
0.03
-0.04
Eugene!
-0.06
0.10
0.74 -0.17
0.03
0.12
Helional
0.02
-0.04
-0.02
0.89 -0.01
-0.14
Lilial
-0.04
-0.07
0.15
0.79 0.16
-0.10
Musk T
0.ll
-0.05
-0.10
0.18
0.92 -0.02
Veti. Ace.
0.08
-0,04
0.03
-0.06
0.90 -0.1 2
Vanillin
0.02
0.03
0.00
-0.07
0.06
0.84
Coumatin
0.21
-0.ll
-0.05
-0.19
-0.15
0.81
Limonene
-0.04
-0.17
0.36
0.10
-0.13
0.71

6. Study of the Transformed data

Figure 3. shows a plot of the log of the summed concentrations for every class versus its rankings. A linear relationship was found as mentioned before for the single-odorant-level. This reminds us of a "fractar(self.resemblance structure) in that, irrespective of odorant-level, similar results demonstrating a linear line on a semi-logarithmic graph were found. The result of factor analysis provided a clue on how to associate aroma to GSN. Instead of assigning aroma to each number of GSN, we found it better to group GSN into clusters and associate them with notes. We can obtain the third geometrical progression by adding two individual geometrical progressions. Therefore, we can create fragrances by connecting several; bases individually fitting the exponential equation, hi this way, the bases can be used advantageously to achieve good harmony systematically. It has been described that the compositions of natural essential oils were expressed by exponential equation. Viewed in this light, these odorants can be regarded as a preblended base.

Fig-3 Semi-logarithmic Graph (classification level)

Rank. Class. Conc. Rank. Class. Conc.
1 JASM 1213213 7 GREN 204568
2 Fix 905253 8 SWET 127944
3 FLOR 685878 9 MUSK 43094
4 ROSE 525904 10 SPIC 16899
5 WOOD 525105 11 VANI 15446
6 FRES 351528

7. Procedure of Perfumery

The results of the factor analysis gave us an idea of what odorants to consider and how to place them on the GSN. Let us now look at a summarized procedure.
Step 1. To prepare the GSN
for example; 85.2, 72.6, 61.9, 52.7? 44.9, 38.3, 32.6, 27.8, 23
*these are obtained by putting b=0,16, a=100 on Y=a/exp(bX).
Step 2. To determine the base
Base 1 jasmin, Base 2:green, Base 3;fresh, Base 4:woody-musky, Base 5:floral(L), Base 6;sweet and each base have 4components aroma.
Step 3. To make a matrix

BASE 1

BASE 2

BASE 3

BASE 4

BASE 5

BASE 6

 Column-
Composition (%)

AROMA 1

85.2

72.6

61.9

52.7

44.9

38.3

63.1

AROMA 2

32.6

27.8

23.7

20.2

17.2

14.7

24.2

AROMA 3

12.5

10.6

9.1

7.7

6.6

5.6

9.2

AROMA 4

4.8

4.1

3.5

3.0

2.5

2.1

3.5

 Row-ompositionO

24.0

20,4

17.4

14.8

12.6

10.8

100.0

Step 4. To place aroma chemicals

BASE 1

BASE 2

BASE 3

BASE 4

BASE 5

BASE 6

AROMA 1

hedione

demethyl P.E.C.

lnalyl acetate

galaxol ide 50

lyral

benzy l benzoaie

AROMA 2

awl cin.ald.

dhydro nyrce.

linalool

vertofix cou

lilial

cinnamyl ace.

AROMA 3

benzyl acetate

dmethyl B.C.A.

d-1imonene

i so-E-SLpe r

cis-3-hex.sal i.

heliotrepine

AROMA 4

hexyl cin. aid.

styral lyl ace.

aldehyde c-9

bacdanol

a-terpi neol

vanllin

Step 5. Mixing
*Because GSN is geometrically progressive, there is no need to measure aroma one by one. To begin with, bases are made using the column composition; then they are mixed with the row composition.

We can obtain odor variations by the following
1) changing the coefficient(b) in step 1.
2) changing the selection of the base in step 2.
3) changing the aroma chemicals and their location in step 4.

8. Application

In order to check availability of the method developed, we made two compounds having the same value of coefficient (b); but one of them (sample A) just fit to GSN (R-squared value is 1.00) and the other (sample B) had small R-squared value (O.SO). Figure 4. and 5. are graphically showing for these samples. If it can be proven that sample A has aesthetically appropriate identity than sample B, using GSN to create a fragrance becomes useful. Subjective sensory information about [which of the samples were well-balanced?] was obtained from twenty-three perfumers. Sample A was preferred by nineteen perfumers, which was much higher than that of sample B which was 4 perfumers (l% significant difference).

Fig-4 Conc.-Rank.Graph for Sample A.


Fig-5 Conc.-Rank. Graph for Smple B

Conclusions

It surprises us that all the successful fragrance discovered by intuition and experiences have similar structure of composition, which is also found in natural essential oils. It has been shown that even the person who has little of talent or experience can create superior work by this method that utilized this construction. But perfumery is the art of creating pleasurable and meaningful odor. As always in perfumery, the perfumer must ultimately be guided by his experience and aesthetic judgment. The method is to be used, as a guide rather than to impose a fixed method of work.

Acknowledgments

We would like to thank perfumers of Soda Aromatic co.; LTD. for evaluating the application samples.

References

[1] Carles,J. Une methode de creation en pariumerie,Recherches, Dec 1961. English transl. A Method of Creation inPerfumery,Soap Perf, Cosm. 35:328-335(1962)[2] Lawrence,B.M.,Progress in Essential Oils, Perfumer & Flavorist,vol.16-20