This article is available at www.springerlink.

com

Optimization of an Alkylpolyglucoside-Based Dishwashing

Detergent Formulation
I. Bozetine T. Ahmed Zad C. E. Chitour
J. P. Canselier

Abstract The aim of this work was to formulate and

optimize the washing performance of an alkylpolyglucoside-based dishwashing detergent. The liquid detergent was
formulated with five ingredients of commercial origin:
anionic (linear sodium alkylbenzenesulfonate and sodium
laurylethersulfate), nonionic (C12C14 alkylpolyglucoside)
and zwitterionic (a fatty acid amide derivative with a
betaine structure) surfactants, and NaCl for viscosity control. In addition to the plate test, other properties were
investigated including cloud point, viscosity, and
emulsion stability. Statistical analysis software was used to
generate a central composite experimental design. Then, a
second order design and analysis of experiments approach,
known as the Response Surface Methodology, was set up
to investigate the effects of the five components of the
formulation on the studied properties in the region covering
plausible component ranges. The method proved to be
efficient for locating the domains of concentrations where
the desired properties were met.
Keywords Alkylpolyglucoside
Dishwashing detergent

Formulation
Response surface methodology

Washing performance

I. Bozetine
T. Ahmed Zad
C. E. Chitour

Departement de Genie Chimique, Ecole Nationale
Polytechnique, 10, Avenue Pasteur (Hassen Badi),
BP 182, El-Harrach, 16200 Algiers, Algeria
J. P. Canselier (&)
Laboratoire de Genie Chimique, ENSIACET-INP Toulouse,
BP 1301, 31106 Toulouse, France
e-mail: [email protected]

Introduction
Liquid detergents play very important roles in our daily
lives for personal care, household surface care, and fabric
care. For many reasons, mainly because they dissolve more
rapidly than powdered detergents and are easier to dose,
liquid detergents have gained an increasing popularity [1].
All other factors, surface to be cleaned, soil, water
hardness, and temperaturebeing equal, cleaning performance is a function of the concentrations and types of the
active ingredients that are delivered into the cleaning bath.
Among these active ingredients, the nonionic surfactants
alkylpolyglucosides (APGs) have been known for a long
time [2, 3]. As a result of their interesting properties and of
the development of new and efficient technologies, they
have been incorporated into detergents, namely dishwashing liquids, since the 1990s [49]. APGs are currently
manufactured by several large detergent companies [10,
11]: they are derived from renewable raw materials,
namely glucose and fatty alcohols (coming from vegetable
oils). Their physical-chemical properties in pure aqueous
solution [10, 1214] or mixed with other surfactants [15,
16], then often showing synergistic effects [6, 17], have
been extensively investigated. In addition to their mildness
to the skin, APGs are readily and rapidly biodegradable
when discharged into the aerobic aquatic environment [18,
19] and even possess detoxifying properties [20]. They are
among the most resistant non-ionic species to alkaline
hydrolysis [5, 20].
In the present study, the goal was to formulate a hand
dishwashing detergent with four major ingredients of
commercial origin.
Statistical analysis software was used to generate a
central composite experimental design and to investigate
the effects of the five components on the following selected

foam end point was reached, that is when half the

surface of the water was covered with a thin layer of
foam. Usually, the numbers of plates washed when
liquor not covered with foam becomes visible and when
almost complete destruction of foam occurs are Np - 1
and Np ? 1, respectively. This so-called foam end
point is considered as the point least subject to errors
in judgment by operators [1, 23]. A few parameters and
conditions need to be specified:

properties: washing performance according to ISO 4198

standardized procedure [21], viscosity, cloud point and
emulsion stability. Besides good washing performances, a
viscosity of at least 300 cSt and a cloud point lower than
5 C were desired. Only quantitative aspects of the formulation, on a physical chemical basis, were considered,
although psychological and economical aspects may also
be important regarding the detergent market.
Experimental
Materials
Ingredients: Except for NaCl (analytical grade), all the
ingredients: anionic, nonionic and zwitterionic surfactants
of commercial origin (Henkel-Cognis GmbH), were used
without further treatment.
The trade names, chemical compositions, physical states
and approximate costs of the ingredients are given in
Table 1.
Soil: Mixture of vegetable (palm and sunflower) oils
(product of Cevital), chosen according to local culinary
habits.
Surface: 20 cm diameter dinner plates.
Methods
Experimental procedures are very close to those described
in [22] with other components. The formulas were prepared
in 100 g quantities. Ingredients were mixed in the
increasing order given in Table 1 at room temperature and
homogenized after addition of distilled water up to 100 g.
The pH-value was adjusted to 6.57.5 by addition of citric
acid.

The dishwashing test was carried out according to the

ISO 4198 guide for comparative testing of performances [21]. The result of the test was taken as the
number of plates, Np, which could be cleaned until the

Table 1 Ingredients of
formulated liquid detergents and
their physical states and
approximate costs

Compound
(commercial name)
1 Maranil Paste A55

Preparation of soiled plates: 150 lL of vegetable oil

were transferred to a dry, clean ceramic plate using an
automatic pipette. The oil was then spread uniformly on the
plate.
Preparation of dishwashing solution: 2.4 mL of liquid
detergent were first transferred to a 35 cm diameter wash
pan of plastic material. Three liters of water (280 ppm
hardness) were then poured into the pan using a 1 L glass
flask from a height of 20 cm above the center of the pan in
order to produce the initial foam. A detergent concentration
of 0.8 mL/L was thus used. The dishwashing solution
temperature was approximately 25 C.
Washing procedure: A 7 cm 9 10 cm dish mop was
used for this operation. The soiled dishes were washed one
at a time, both front and back, using a rotating motion with
the dish mop while keeping the dish half submerged in a
given angular position (ca. 45) with respect to the bottom
of the wash pan. This wash process took around 30 s,
including the rinsing step and the visual inspection of the
plate. The plates that had been washed and rinsed during
the test were placed in racks and allowed to dry.

Kinematic viscosity (m) was measured at 25 using

Ubbelohde viscometers.
The cloud point, tcloud, was determined by cooling
the samples in 20-mL glass test tubes in an ice bath and
the appearance of turbidity was taken as the cloud
point. The reproducibility was 1 C. The clear
point, tclear, was observed when the sample became
transparent while coming back to room temperature. In

Chemical structure, class (concentration in wt.%,

physical state)

Current cost (/kg

approximate)

Linear sodium dodecylbenzenesulfonate,

0.55

anionic surfactant (5456% pasty)

2 Texapon NSO

Sodium laurylethersulfate,

0.40

anionic surfactant (26.527.5% liquid)

3 Glucopon 600CS
UP

Alkylpolyglucoside,

4 Dehyton K

Fatty acid amide derivative with betaine structure,

zwitterionic (2932% pasty)
NaCl (crystalline solid)

5 Sodium chloride

1.65

nonionic surfactant (5053% thick liquid)

0.70
0.28

Europe and North America, recommended values of

cloud point and clear point are lower than 5 C
and lower than 10 C, respectively [1].
The emulsifying power of the samples towards grease
(emulsion stability) was illustrated by the following test.
20 g of a solution of known composition were introduced
into a 30-mL vial (2.5 cm 9 9.5 cm) and 0.2 g of
vegetable oil was added to the vial. The vial was
stoppered and submitted to an oscillatory motion (180
back and forth) twenty-five times at an approximate rate
of one rotation per second. The vial was then permitted to
stand for 5 min. Nephelometric readings (intensity of the
light scattered at 90) were taken using a WTW Turb 555
turbidimeter after 1, 3 and 5 min. and expressed in NTU.
Linear regressions gave the predicted readings for 3 min,
which were taken as the turbidity values, s, higher for
more stable emulsions [24].

Table 3 Composition and

measured properties of samples
for the central composite design

Run N

Table 2 Concentrations of ingredients

Compound: commercial name,
abbreviation

Concentrations: reduced
and actual (wt.%)
2

?1

?2

Maranil Paste A55, M

10

13

16

Texapon NSO, T

10

12

Glucopon 600CS UP, G

Dehyton K, D

NaCl, NaCl

0.3

0.6

0.9

1.2

Statistical Methods
The second-order experimental design, a central composite
design with four center points, is shown in Table 3. All

Ingredients and reduced concentrations

Response variables

NaCl

Np

-1

-1

-1

-1

15

-1

-1

-1

-1

-1

-1

-1

-1

m (cSt)

tcloud (C)
2

tclear (C)

11

23

18

13

25

15

16

-1

7
9

-1

-1

25

581

10

-2

-1

-1

-1

-1

19

18

-1

-1

36

24

11

-1

-1

-1

35

112

13

-3

-1

-1

42

35

18

-4

-1

-1

-1

-1

24

10

16

15

10

-1

-1

29

46

17

-3

11

-1

-1

28

169

17

12

-1

-1

33

45

15

-1

13

-1

-1

32

18

11

16

14

-1

-1

29

12

16

-2

15

-1

-1

43

20

17

-3

16

47

442

13

-2

7
8

17

-2

25

19

17

-1

18

28

32

23

-1

19

-2

22

23

12

-1

20

33

22

15

-2

21

-2

39

10

17

-2

14

22

29

419

13

23

-2

31

12

15

24

35

135

-1

25

-2

22

10

18

-4

26

23

63

14

-3

27 (C)

31

24

13

-2

28 (C)

30

26

14

-3

29 (C)
30 (C)

0
0

0
0

0
0

0
0

0
0

27
26

23
24

13
13

-2
0

6
6

analyses were performed with the help of Statistica

(Design of Experiment module) [25], which provides a set
of options to allow the user to optimize multiple response
variables interactively, given the current model. For convenience, analyses were performed using coded
independent variables (-2, -1, 0, ?1, ?2) rather than
actual values. In model quadratic equations, only some of
the terms are statistically significant. The Pareto diagram
sorts the effect (absolute values) of the parameters (independent variables) on the responses (dependent variables).
Unless otherwise stated, the threshold value was fixed at
5% (p \ 0.05).

Results and Discussion

Fig. 2 Contour plots

(M = D = NaCl = 0)

for

predicted

number

of

plates

The actual and corresponding reduced concentrations of

the ingredients are reported in Table 2. Along with the
reduced concentration values of the five ingredients, the
response variables (number of plates washed, viscosity,
turbidity, cloud point and clear point) are indicated in
Table 3.
Plate test: For a 5%, threshold on the Pareto chart, only
Texapon, with its linear term, shows a significant effect on
Np. As can be seen for p = 0.15, Texapon, Glucopon and
Maranil exhibit a statistically significant effect on the
number, Np, of cleaned plates (Fig. 1). The model for this
response is:
Np 29:37 4:50T 2:33G 2:33M

with a rather low correlation coefficient R2 of 0.47. In fact,

a lot of factors, associated with the details of the experimental procedure and which cannot be accounted for,
affect detergent performance. For instance, the exact
quantity of soil deposited cannot be known with high
precision as some soil is always left on whichever device is
used to spread it onto the plate. This in turn will have an
effect on the foaming power and hence the plate number.

Fig. 1 Pareto chart of effects for the plate test (L stands for linear)

Fig. 3 Predicted versus actual values for number of plates (Np)

The contour plots for Np (projection of the response

surface) predicted from the model are given in Fig. 2. From
this plot, one can easily see that the best washing performances are obtained with increasing surfactant
concentrations.
The precision of the Np model can be best visualized by
a plot of actual versus predicted values (Fig. 3).
Residuals are differences between the observed values
and the corresponding ones predicted by the model and
thus represent the variance that is not explained by the
model. The better the fit of the model, the smaller the
values of residuals. Figure 4 shows the residuals plot.
Analysis of variance for the Np model is given in
Table 4.
Obviously this result is specific to a particular application. Since there is wide variation in soils and consumer
dishwashing habits, for a given detergent formula, the use
of other types of soil (e.g. those described in the ASTM
D4009 standardized procedure) [23] and washing procedure would lead to different values of Np. It is not likely
that a general equation could be obtained to represent the

Fig. 5 Pareto chart of effects on viscosity (L stands for linear and Q

for quadratic)

Fig. 4 Residuals plot for the number of plates (Np)

Table 4 Anova table for the number of plates (Np)
Sum of
squares

Degrees of
freedom

F-value

p-value

(1) Maranil (L)

130.667

23.05882 0.017189

(2) Texapon
(L)

486.000

85.76471 0.002664

(3) Glucopon
(L)

130.667

23.05882 0.017189

Lack of fit

830.633

23

6.37315 0.075848

Pure error

17.000

1,594.967

29

Total sum of
squares

performances of a family of detergents containing various

ratios of the same ingredients.
Viscosity The viscosity of a liquid detergent is very
important for its dispersability on dilution [26]. The viscosity of a light duty liquid detergent is typically in the
range of 100500 cSt.
At p = 0.05 level, three ingredients were found to show
significant effects on viscosity (Fig. 5).
Excluding the other parameters, the viscosity can be
described by the following equation:

Fig. 6 Contour plots of predicted viscosity (M = D = T = 0)

m 42:345 87:44G 46:94G2 56:32NaCl 46:58G:D

70:55G:NaCl 46:07D:NaCl

with a correlation coefficient R of 0.92.

The contour plots of predicted viscosity from the above
model are given in Fig. 6. Viscosity values reach 300 cSt
and more in the upper right corner of the plot (high concentrations of Glucopon and NaCl).
Emulsion Stability. Turbidity measurements were performed in order to measure product ability to emulsify
grease (see Experimental).
At p = 0.05 level, the Pareto and contour charts for
turbidity are given in Figs. 7 and 8, respectively. The
model equation for turbidity reads:

Fig. 7 Pareto chart of effects on turbidity (L stands for linear and Q

for quadratic)

s 13:82 1:55M2  1:35D 1:09M:D 1:08D:T

 0:94NaCl 1:04T:G  1:86T:NaCl
2

with a coefficient of determination R of 0.72.

Turbidity does not vary very much throughout the
investigated concentration range. The higher (desirable)

Table 5 APG-based dishwashing detergent: limit and compromise

formulas
Ingredient

Concentrations (wt.%)
In limit formula

In compromise formula

Reduced Actual Reduced

Actual

Maranil Paste A55

13

10

Texapon NSO

10

0.5

Glucopon 600CS UP 1

0.5

Dehyton K

NaCl

0.9

0.9

Property
Np (plate number)
Fig. 8 Contour plots of predicted turbidity (T = G = NaCl = 0)

values obtained at the lower left corner of the plot, corresponding to low concentrations of Maranil and Dehyton
(out of the range studied) are not reliable. Therefore, the
emulsifying power will not be critical for the formulation.
Cloud point: The cloud point is the temperature at which
the product begins to turn cloudy upon cooling. At p = 0.1
level, only two ingredients were found to exhibit significant
effects on the cloud point. The contour chart for the cloud
point is given in Fig. 9. The model for the cloud point can
be described by the following equation:
tcloud 1:33  0:58T  0:58D

with R2 = 0.73.
In fact, in the investigated concentration range, the
products should exhibit cloud points lower than 0 C. On
the other hand, except for samples 9, 13 and 21, the clear
points do not exceed 10 C.

Conclusion
Finally, we may conclude that:

The liquid detergent we have formulated possesses

acceptable properties nearly throughout the investigated composition range,
Practically, all the properties improve for higher
surfactant concentrations (of course, at the expense of
cost).
Emulsion stability (expressed as turbidity) and cloud
point are not very critical. When choosing the optimum
for the product formulation, the two main properties to
be considered are the cleaning properties and the
viscosity.

Obviously, the active matter concentration of the finished detergent is probably the most important parameter
as it has a direct influence on both physical and

m (cSt)

38.5
396

32.8
190

s (NTU)

14.44

12.21

t (C)

-2.5

-1.6

performance characteristics. The active matter concentration also, of course, determines the cost of the finished
formulations. In the present case, contour plots and desirability profiles help to achieve the best solution.
The addition of NaCl increases the viscosity of the
mixture, showing the same trend as the addition of surfactants (Eq. 2). So, from Eqs. 1 to 4, it is interesting to
calculate the properties (responses) of the detergent formulated according to the experiment n 16, where all the
ingredients are at their maximum concentrations, i.e. ?1 in
reduced value (Table 2). The obtained values, reported in
Table 5, largely exceed the initial fixed goals (m C 300 cSt,
Np C 30 and tcloud B5 C), but correspond to a high ratio
of active matter (*14%) and to the highest cost (within the
limits defined above): 0.234 /kg.
In order to reduce ingredient consumption, it will be
advantageous to lower the surfactant amounts while
maintaining NaCl (cheapest component) at its highest level
so that viscosity remains high. Keeping the fixed goals in
mind with regard to cleaning power, a second, compromise formula, together with the properties herewith
associated, is presented in Table 5. With T = G = 0.5,
M = D = 0, therefore ca. 11% active matter, except for
the kinematic viscosity, the properties appear satisfactory
at a lower cost of 0.19 /kg. To reach the desired viscosity,
it would be sufficient to increase the amount of NaCl up to
1.26% (reduced value: 2.21, making the hypothesis that our
model remains valid). Emulsion stability would be a little
less (s = 10 NTU) and the cost would then be 0.194 /kg,
slightly higher than the previous one.
The method proved to be efficient for sketching out the
domains of concentrations where the desired properties are
met. Unfortunately, the models should probably not be
applied to similar products from different suppliers. It is
well known that the chemical composition of commercial

Fig. 9 Contour plots of predicted cloud point (M = G = NaCl = 0)

surfactants (which are always mixtures) varies with the

details of the manufacturing process and even from batch
to batch, although the global specifications are met.
Finally, let us note that this choice does not take ingredient
cost as an optimization criterion and should be modified if
a quality-to-cost optimum formulation was searched for.
Acknowledgments The authors wish to thank Henkel-Cognis
(Algeria) for supplying all the components used in this study.

References
1. Cahn A (1996) Liquid detergents: an overview. In: Lai KY (ed)
Liquid detergents, surfactant science series, vol 67. Marcel
Dekker, New York, pp 119
2. Ames GR (1960) Long-chain derivatives of monosaccharides and
oligosaccharides. Chem Rev 60:541553
3. Egan PA (1989) Surfactants from biomass. Chem Tech, pp 758
762
4. Andree H, Middelhauve B (1991) Possibilities of the use of alkylpolyglucosides in detergents and cleaning liquids. Tenside
Surfactants Deterg 28(6):413418
5. Lomax E (1994) Alkylpolyglucosides for personal care, laundry
and liquid detergent applications. Speciality Chem Mag
14(1):2124, 26
6. Nieendick C, Schmid KH (1995) Alkyl polyglycosides: a new
generation of surfactants for the use in manual dishwashing
agents. Jornadas Comite Esp Detergencia 26:6367
7. Cottrell JS, Rennie GK (1996) Light duty cleaning compositions
showing acceptable mildness, cleaning and foaming, WO
9635770
8. Balzer D (ed) (2000) Nonionic surfactants: alkylpolyglucosides,
surfactant science series, vol 91. Marcel Dekker, New York
9. Soldanski HD, Adomat C, Buisker D (2007) Skin-protecting
dishwashing detergents. WO 2007028571
10. von Rybinski W, Hill K (1998) AlkylpolyglucosideEigenschaften und Anwendungen einer neuen Tensidklasse. Angew
Chem 110(10):13941412
11. Bognolo G (2004) Tensioactifs non ioniques: mise en oeuvre
industrielle. Techniques de lIngenieur, Doc J 2(265):16, Paris
12. Balzer D (1991) Alkylpolyglucosides, their physico-chemical
properties and their uses. Tenside 28(6):419427

13. Nilsson F (1996) Alkylglucosides physical-chemical properties.

Inform 7(5):490497
14. Bognolo G (2004) Tensioactifs non ioniques: proprietes: tableaux
comparatifs. Techniques de lIngenieur, Comp J 2(266):18,
Paris
15. Nickel D, Nitsch C, Kurzendorfer P, von Rybinski W (1992)
Interfacial properties of surfactant mixtures with alkylpolyglucosides. Progr Colloid Polym Sci 89:249252
16. Hofmann R, Nickel D, von Rybinski W (1994) Grenzflacheneigenschaften und Phasenverhalten von Fettalkoholsulfaten.
Tenside Surfactants Deterg 31(1):6366
17. Florescu S, Leca M, Rob M, Zahanagiu F, Stanciu I, Filipescu L
(1997) Synergistic effects in binary mixtures of alkylpolyglucosides with anionic and nonionic surfactants. Chimica Oggi 15(9/
10):4346
18. Samdani G, Shanley A (1991) Surfactants: super molecules.
Chem Eng (international ed.) 98(3):37, 39, 41, 43
19. Garcia MT, Ribosa I, Campos E, Sanchez Leal J (1997) Ecological properties of alkylpolyglucosides. Chemosphere
35(3):545556
20. Bognolo G (2004) Tensioactifs non ioniques: mise en oeuvre
industrielle, Techniques de lIngenieur J 2(265):19, Paris
21. ISO 4198 (1984) Surface active agentsdetergents for hand
dishwashingguide for comparative testing of performance,
international
organization
for
standardization,
http://www.iso.ch/iso
22. Ahmed Zad T, Bensari L, Benmaza K, Chitour CE, Canselier JP,
Pierlot C (2007) Response surface methodology as an approach to
the optimization of a dishwashing detergent. Tenside Surfactants
Deterg 44(2):94101
23. ASTM D4009-92 (2006) Standard guide for foam stability of
hand dishwashing detergents
24. Simion FA, Warscjewski DW, Zyzyck LA (1990) U.S. Patent
4,923,635 to Colgate-Palmolive Co
25. Statistica software, kernel version 5.5, StatSoft Inc.,
http://www.statsoft.com (19842006)
26. Drozd JC, Gorman W (1988) Formulating characteristics of high
and low 2-phenyl linear alkylbenzene sulfonates in liquid detergents. J Am Oil Chem Soc 65:398

Author Biographies
Isma Bozetine is a former post-graduate student in Chemical Engineering at the Ecole Nationale Polytechnique, Algiers.
Toudert Ahmed Zad, D.Sc., is a professor in the Chemical Engineering Department, Ecole Nationale Polytechnique, Algiers. He has
published 10 research papers in international journals mainly in the
field of Surfactants and Formulation Technology.
Chems-Eddine Chitour, D.Sc., is Professor and Head of the Fossile
Energy Valorization Laboratory at the Chemical Engineering
Department, Ecole Nationale Polytechnique, Algiers. He has published an appreciable number of research papers and several books on
Energy, Petroleum Engineering and Physical Chemistry of Surfaces.
Jean Paul Canselier, D. Sc., just retired, has been a Lecturer at the
School of Chemistry and Process Engineering of Toulouse (ENSIACET). He has more than 25 years experience in the domain of
surfactants and dispersed systems (mixed micelles, microemulsions,
surfactants in chemical engineering operations, formulation, etc.). He
has authored or co-authored around seventy papers and book chapters,
mainly in this research area.