1. A guide to Optimal Surfactants Professor Steven Abbott
This is my non-expert personal view of the field
It has been used as the basis for a number of tutorial talks
It is provided “as is” to stimulate discussion/debate
Please send feedback to
The website for the HLD-NAC software package is
HLD-NAC
A guide to Optimal Surfactants
Professor Steven Abbott
My thanks to, in alphabetical order:
Professors Acosta, Aubry, Salager and Sabatini
for their generous assistance in teaching me about HLD,
and to Professor Acosta and his group for their continuing support and guidance in developing the software and database.
Images from their papers are used with grateful acknowledgement

2. What I’m not talking about
My interest is in microemulsions for cosmetics, oil-field recovery, nanoparticles etc.
I have no particular interest in or knowledge of foams
I have no particular interest/knowledge in the dynamics of classical micelles and their CMC behaviour
I can’t talk about polymeric surfactants because they, too, are outside my knowledge base
This talk is pseudo-ternary-phase-diagram free

3. Just an ordinary scientist
I’m a formulator, not a surfactant scientist
I’ve read lots of surfactant theories
None of them provided a practical formulation tool
I want a rational choice of surfactants
and I suspect that surfactant suppliers would love to reduce their inventories …
I want to formulate efficiently, not trial-and-error
Surfactant formulation space is way too large
I want to rationally swap to “green” surfactants

4. Microemulsion terminology
Winsor Type I, o/w, lower (denser) phase
Winsor Type II, w/o, upper (lighter) phase
Winsor Type III, bicontinuous middle phase
Type I
Type III
Type II
In reality are bicontinuous
Generally a uniform size

5. I want to get this first time
EACN scan from Intelligent Formulation Green Surfactant project. Low EACN to left, high EACN to right
Water phase larger
Type I o/w
Oil phase larger
Type II w/o
Optimal Type III

6. I want: To be in the right phase, first time
Sometimes I want Type I, sometimes Type II, sometimes Type III
To have a high solubility parameter (SP)
ml water-in-oil/g surfactant
ml oil-in-water/g surfactant
Because surfactants are “bad” and “expensive”
(To be free from liquid crystal phases)
I’ve no solutions to this issue but high SP means low surfactant concentrations which will help

7. I want microemulsions because
I can do clever cosmetic formulations with them
But note that most published microemulsion cosmetic/pharma papers use them only because “nano is good”
These papers generally prove that if you shove a lot of surfactant onto skin you can get things to go through. Nothing to do with microemulsions

8. Not HLB
HLB really aren’t much use because they really only apply well to medium-sized, standard EO non-ionics
why should the % hydrophilic sorbitan be in any way equivalent to % hydrophilic EO?
With a choice of 2,000 surfactants, maybe 200 have your HLB, yet their properties are very different
And they don’t predict much beyond basic o/w or w/o, and often not very well
Don’t predict effects of salinity, oil, temperature

9. Not CMC Critical Micelle Concentration
Tells you a lot about its desire to escape from water and form micelles
But not a lot about making great w/o or o/w emulsions
Generally lower is better because its not too water soluble, and probably has long tails
the long tail is good for other reasons
CMC for extended surfactants are “meaningless”

10. Not CPP
CPP=Vtail/(Ltail*Ahead)

11. Not CPP
CPP=Vtail/(Ltail*Ahead)
Critical Packing Parameter seemed to be a good idea for distinguishing between o/w and w/o surfactants based on whether head or tail is bulkier
Theoretically bankrupt because microemulsion curvature is flat on the molecular scale …
… and one can go seamlessly from o/w to w/o with surfactants of very different CPP values
But they’re great for those interested in complex surfactant phases (hexagonal, laminar…)
And such phase behaviour in practice limits microemulsion practicality

12. CPP for phases in conc. surfactant
Any complete surfactant story would include HLD-NAC and the tendency to form liquid crystalline phases ...

13. But … Knowing the key parameters used in CPP is very important:
Head area, A
(Extended) Tail length, L
And there may be some routes from CPP to Cc (see later)
But so far that’s speculative

14. HLD Hydrophilic Lipophilic Difference
Apparently simple, but a lot of work to get it right from (in alphabetical order) Aubry, Sabatini and Salager
You get the Optimal Surfactant when HLD=0
lowest interfacial energy
highest SP (Solubility Parameter – not HSP)
ml/g i.e. ml of oil dissolved in water per g of surfactant
best cleaning, oil recovery and microemulsions

15. Same(ish) equation for i and non-i
EACN
Hexane = 6
Decane = 10
etc.
But
Benzene = 0
Limonene = 7
IPM = 13
Triolein = 16
i.e. they behave like an alkane with those numbers of C
Ionics
HLD = ln(S)-K.EACN-αΔT+Cc
Non-ionics
HLD=b.S-K.EACN+cΔT+Cc
S = Salt concentration, g/100ml
K= Constant ~0.17
EACN = Effective Alkane Carbon Number
α=temperature coefficient ~0.01. Note the - sign
ΔT=temperature difference from 25°C
b = Salt dependency for non-ionics ~ 0.13
c=temperature coefficient ~0.06. Bigger effect, opposite sign
Cc=Characteristic Curvature, unique to each surfactant

16. What does it mean? Ionics
0 HLD when ln(S)+ Cc = K.EACN + αΔT
as S goes up, Water solubility goes down
as Cc goes up, Water solubility goes down
as EACN goes up, Oil solubility goes down
as T goes up, Oil solubility goes down
because ionic is more soluble in water
The best surfactant for a hot cycle in a washing machine is often bad for a cold cycle – and vice versa

17. What does it mean? Non-Ionics
0 HLD when b.S+cΔT+Cc = K.EACN
As S goes up, Water solubility goes down
As T goes up, Water solubility goes down
EO is a strange beast so water solubility decreases with T
As Cc goes up, Water solubility goes down
As EACN goes up, oil solubility goes down

18. Why does it matter? - 1
If you’ve got an easy oil (e.g. heptane) and just want a modest amount of o/w or w/o microemulsion (e.g. for nanoparticle manufacture), then trial and error will get reasonable results without any theory
But …

19. Why does it matter? -2 Difficult oils have large EACNs
No problem – balance with Salt or a large +ve Cc
But high Salt is often not practical + small Type III
And large +ve Cc usually means v. long tail which generally means nasty liquid crystal phases and sticky messes
AOT is an exception because its Cc=2.5 (SDS=-2.3) so can formulate with EACNs 25 higher than SDS
Those two tails = lots of C but not too self-associative
The branches help bulk out and reduce self-association

20. Why does it matter? -3
To get large amounts of o/w, w/o or to get a whole, stable Type III 50:50 emulsion is very hard without theory
Especially if you want a high SP – best emulsion with least surfactant
For cosmetics, surfactants are often “bad”, so less surfactant is desirable.

21. The parameters
cΔT
No theoretical method to calculate c as EO solubilities are hard to understand
But c seems to be a good solid number
αΔT
No theoretical method to calculate α
No fundamental reason why different ionic headgroups should have the same α
insufficent data to know either way

22. The parameters (cont) EACN By definition OK for alkanes!
But where do EACNs for other oils come from?
My view is that HSP can predict these
“Like dissolves like” so compatibility between surfactant and oil will decrease EACN
But so far only found one good dataset (Aubry) to be able to test

23. The parameters (cont) K
Known to be smaller (~0.1) for Extended Surfactants than for “typical” surfactants (~0.17)
Why is K the value it is?
Can it be calculated?
Is it truly a constant?
Till there’s more data, assume that it is a constant

24. The parameters (cont) Cc Characteristic curvature
A –ve value means that for “normal” conditions the emulsion is curved around the oil, so it’s o/w Type I
A +ve value means curved around the water, so it’s w/o, Type II
Sounds a bit like CPP, but is simply an observed parameter, not something directly calculated (yet?) from shape/size
Sodium Di-octyl sulfosuccinate is +2.5
Sodium Di-hexyl sulfosuccinate is -0.9
why do 2 Me groups (+ branching) give such a big change?

25. Scans, not phase diagrams
*Hope for SP of 10
So 3% surfactant will dissolve 30g oil
20:60:20 result
A proper phase diagram varies
Oil, water, surfactant, brine, temperature
Need a 5D plot to visualise it
Life’s too short
So take 10 test tubes, 50:50 o/w, modest amount of surfactant*
And look for 2 (o+o/w) to 3(o+o~w+w) to 2 (w/o+w) phases as salt goes lowhigh

26. Scans not phase diagrams

27. More scans
Note the relatively large oil phase to the right of S*, showing a relatively large amount of water brought into the oil. I’m not clear why you don’t get the same phase volume to the left, though LC phases are part of the problem

28. The simple scan gives: S* - optimal salinity where HLD=0
You know EACN, you know ΔT=0 so you get Cc
And you choose the oil so the data apply to you
You don’t care if things would be better with hexane!
From the size of the middle phase you get SP
If you don’t get phases then put all tubes in a water bath and find T where you do
Especially good for non-ionics
That’s the theory – it’s never worked for me

29. Scans are: Tedious Error prone Limited by practical stock solutions
Therefore
High-throughput equipment is highly desirable!

30. High Throughput
Do in one day what a skilled technician can do in a week
Much more versatile because can weigh any amounts – no limits of stock solutions
Photo record/measure of phase volumes
A ChemSpeed Formax HT machine at VLCI

31. Measuring the parameters
Simple phase scans to find “optimal salinity” S* followed by simple algebra give you:
EACNs (find S* for same surfactant in different oils)
Ccs (find S* for same oil, different surfactant)
or extract more information from combined scans
DOE principles greatly reduce number of scans
Change proportion of unknown surfactant in known surfactant, and fit S* data to calculate Cc of unknown

32. Practical scans For experimental ease choose:
An EACN scan to get Cc of surfactant
Fixed S, varying EACN e.g. TolHexadecane = 116 or Tol Squalane =124
A surfactant scan using known Cc to get unknown Cc
Fixed S, Fixed EACN, vary known/unknown surfactant ratio
Salt scans are harder to do
adding solid salt is hard without a robot
solid salt to EACN or surfactant scans can explore other ranges if original scan fails to find HLD=0

33. Scanning for EACN Fix S and unknown oil, scan two known surfactants
A key step in building up the utility of HLD-NAC as too few real-world oils are known
Natural oils may show large variations due to differing levels of long chains and degrees of hydrolysis of glycerides
Customers might require EACN values from suppliers so they can match with correct surfactant (blend)

34. Guesswork Impossible to scan full range with high accuracy
So must guess Cc and plan scan around it
If guess is wrong then the scan helps narrow the range for future scans
As we build up more Cc values, there will be more prototypes for estimating Cc and more “right first time” scans

35. Planning
HLD-NAC software makes it easy to explore scan ranges and scan issues
A complex 2-surfactant situation optimal at 6% salt

36. QC via HLD
The “same” surfactant from different suppliers, and different batches from the same supplier might be different because of differing chain lengths, EO range, alcohol or acid levels
The Cc will be an accurate QC tool to show functional equivalence of different batchs
Can use small EACN steps around the Cc to get accurate values. Quick/simple test!

37. Calculating the parameters
Cc can come from CxEOy correlations
Some evidence of link of Cc to CCP
EACNs some rules of thumb, group contribution and, perhaps HSP
But till we have larger high quality databases, it’s hard to know if predictions are of great value

38. The Exxon model H/L = VH/VL = (VoH+VWH)/(VoL + VLo)
Rather like CPP, but Vol_Head/Vol_Tail
Vols made up of intrinsic VoH ,VoL (calculable) and associated VWH ,VLo (unknowable?)
High H/L = Hydrophilic, Big Head = Head outside = o/w
Low H/L = Lipophilic, Big Tail = Tail outside = w/o
But H/L is NOT a constant for a surfactant
Changes with salinity, T, oil
Robbins M.L. et al, J. Coll. Int. Sci. 124, , , 1988 & 126,

39. Exxon continued H/L changes with salinity
“dehydration” of the associated water
Curvatures & Volumes calculable
Many similarities with predictive capabilities of HLD-NAC
Phase volumes etc.
But couldn’t compute τr/δr
Little referenced since

40. Volume Fractions & Phase Volumes
A typical volume fraction scan
Same thing but change cation!
A lovely bit of work from Exxon

41. Exxon nukes CPP
Fig 3 of Paper II shows that small H/L ratio changes cause large S* shifts if L changes but small shifts if H changes
Because S* depends on (Head+Water)/(Tail+Oil)
Fig 4 shows how the real sizes change
τr = τH/ τL δr =δH/δL

42. HLD-NAC
We can do the HLD bit and get ourselves into the right formulation domain
But we can’t predict phase volumes, fish diagrams, viscosities …
Exxon were close to a predictive model, but they didn’t have the simple HLD and they couldn’t quite get the right parameters
Hence we need HLD-NAC

43. What about NAC? First, the Net Curvature Net Curvature=1/Roil-1/Rwater
If you’ve got w/o then Rwater is meaningful, what of Roil?
Give it a pseudo value = 3*Voil/As
i.e. the ratio of volume to area of surfactant
As is calculated from number of molecules * surface area per molecule
So what? See next slide

44. Total amount of water Net curvature also = -HLD/L
Derived from scaling theory of the chemical potential (HLD) with the Kelvin equation
Net curvature also = -HLD/L
L= extended length of surfactant tail
So, 1/Rwater=HLD/L-1/Roil
If Roil is very large then 1/Rwater=HLD/L or
Rwater=L/HLD
So a long tail means large solubility
A small HLD also means large solubility
That’s why long tails and small HLD are so good!
Of course you can always increase total water by adding more surfactant
or, for the same amount of water, more surfactant = smaller Rwater

45. Taming infinity At HLD=0 the solubility would be infinite
That’s why we need NAC – Net Average Curvature
NAC=0.5*(1/Roil+1/Rwater)
just the average of the two curvatures
NAC turns out to be equal to 1/ξ
ξ (chi) is the De Gennes “coherence length”
The longest length for which the surfactant “pallisade” can be considered to be a straight line
The longer it is, the larger the curvature
So Rwater is limited by the finite NAC

46. Everything now known The NAC limits the excesses of the NC
1/Roil-1/Rwater=HLD/L
1/Roil+1/Rwater=2/ξ
Calculations
L from the structure
HLD from S, T, EACN and Cc
Vw/As requires known surfactant area
If we knew ξ we’d know everything
But we don’t, however intuition isn’t bad
For large solubility:
HLD = small
L = large
As = large
ξ = large

47. Chi
Imagine the microsphere is made up of straight sections of length ξ
The bigger ξ, the bigger the sphere so the greater the solubility
ξ =a exp(2πk/kBT) – so stiff chains are good for a large ξ …
a=Tail+Boundary, k=stiffness, kBT - Boltzmann
… but if the interface is too stiff then it loves to be ordered and we have liquid crystal phases

48. What happens at 50:50 Oil:Water?
NC=0 (it’s neither o/w nor w/o)
NAC ≠0 (it’s limited by ξ)
So we have o/w curves and w/o curves
in other words we have a sinuous interweaving of the two phases
with straight lines of length ξ
typically 100Å
And there’s more...

49. More at 50:50
If you have a small(ish) amount of surfactant you’ll have 3 phases
If the SP of your surfactant is high (low HLD, large L, large As, large ξ) then the size of the middle, clear microemulsion phase is large
with essentially no surfactant in pure water or pure oil
This is what we look for in phase scans
1 That we actually have three phases
2 That the middle phase is large

50. Optimal insolubility
Any surfactant in the water or oil is doing nothing
So an optimal surfactant has zero solubility in the water and in the oil
And is therefore 100% at the interface
An impossible ideal, but a simple, clear goal

51. Proof of optimal insolubility
How much surfactant is needed to obtain 1 phase for 50:50 o:w
Surfactant solubility in oil phase

52. Other calculations Rd and Ld – radius & cylinder length
Vf - total dissolved volume fraction
Viscosity
Fish diagram
Vary % surfactant, calculate o/w and w/o volumes via ξ and therefore HLD and therefore T* for each transition
Can’t (yet) do the low % cutoff – needs surf. sol.

53. Calculating Shapes, Sizes, Viscosities
Simultaneous equations in Hn (NAC) and Ha (AC) give us true shape* of “drops” – cylinders of radius Rd and Length Ld
Correlates well with neutron scattering
Can calculate total Volume Fraction and Number Density from simple geometry
Viscosity can be predicted via “dilute rigid-rod theory”
N = Number Density of droplets
L = Length, d=Diameter of rods
cg = rigid rod concentration, ρ=µE density
*As De Gennes points out, “whenever R> ξ, the shapes must be strongly non-spherical”

54. Formulation: failure is normal
An average surfactant with an average “real world” oil will give no 3-phase at any sensible temperature or salinity
Cc is simply too far away
And high salinity tends to give narrow phases (Exxon)
Even if there are 3 phases, the SP is small
L too small, As too small, ξ too small, HLD not zero
So how do we choose the right surfactant(s)?

55. General Surfactant Design Rules
Make the tail and head large
But without making liquid crystal phases - hard!
Use mixtures
Long chains can be “balanced” by shorter chains
Impure is usually better than pure
Use Linkers
Long-chain alcohols as hydrophobic linkers
Sodium Mono/Dimethyl Naphthalene Sulphonate as hydrophlic linker
Works pretty well to increase L and As
But only pretty well...

56. Linkers Lipophilic linker – e.g. Octanol
Di-block linker – e.g. PEP5-EO5
No linker
Lipophilic & Hydrophilic linkers

57. Extended surfactants
So put a linker in between a conventional head and tail
PO chains are neutral compared to EO heads and to HC tails
Typical example C14PO8EO2SO4Na
C14 – a typical tail
PO8 – a long “neutral” linker section
EO2 – helps bulk up the head
SO4Na – typical anionic head

58. Problems Predicting ξ Predicting Cc Predicting EACNs
Predicting failure due to liquid-crystal phase formation instead of expected microemulsion
Uniting terminology and similar approaches (e.g. Exxon v HLD-NAC, %NaCl v g/100ml…)
Cranking out public-domain datasets

59. HLD-NAC software It’s only as good as the theory and the data
The theory seems better than anything else out there
Acosta has worked hard to supply data
Sasol have published some on extendeds
Two ways forward?
Groups take on measurement projects
See the Intelligent Formulation Green Surfactant Project
Surfactant suppliers supply data for everyone’s good
One supplier gains competitive advantage …

60. Acknowledgements
In alphabetical order, HLD-NAC is the fruit of work by Acosta, Aubry, Sabatini and Salager
Each has been generous in their discussions with me
Edgar Acosta at U. Toronto is now the person taking forwad HLD-NAC and the current version of the software owes a lot to his help
My colleagues at Syntopix, VLCI and Intelligent Formulation in the Green Surfactant project