ADVANCES IN

COLLOID AND
INTERFACE
Advances in Colloid and Interface Science SCIENCE
ELSEVIER 70 (1997) 171-210

Analysis of small-angle scattering data from colloids

and polymer solutions: modeling and least-squares
fitting I
J a n Skov Pedersen
Department of Solid State Physics, Riso National Laboratory, DK-4000 Roskilde, Denmark

Abstract

A n a l y s i s a n d modeling of small-angle s c a t t e r i n g d a t a from s y s t e m s consisting of

colloidal p a r t i c l e s or p o l y m e r s in solution a r e discussed. The analysis r e q u i r e s applica-
tion of l e a s t - s q u a r e s methods, and the basic principles of l i n e a r a n d n o n - l i n e a r least-
s q u a r e s m e t h o d s a r e s u m m a r i z e d with e m p h a s i s on applications in the a n a l y s i s of
s m a l l - a n g l e s c a t t e r i n g data. These include indirect F o u r i e r t r a n s f o r m a t i o n , square-root
deconvolution, size d i s t r i b u t i o n d e t e r m i n a t i o n s , and modeling. The inclusion of correc-
tions for i n s t r u m e n t a l s m e a r i n g effects is also discussed. The most common a n a l y t i c a l
e x p r e s s i o n s for model form factors a n d s t r u c t u r e factors a r e s u m m a r i z e d . A n e x a m p l e
of a n a l y s i s of s m a l l - a n g l e n e u t r o n and X-ray s c a t t e r i n g d a t a from block copolymer
micelles is given.

1. I n t r o d u c t i o n

The article concerns analysis of small-angle scattering data from

colloidal and polymer systems consisting of particles or molecules in a
solvent. Only systems with short range order and isotropic scattering
spectra, for which the scattering intensity is only a function of the

1
A premous version of this review was used as lecture notes in The Third European Summer School
on "Scattering Methods Applied to Soft Condensed Matter", Bombannes, France, 1996.
0001-8686/97/$32.00 © 1997 - - Elsevier Science B.V. All rights reserved.
PII: S0001-8686(97)00312-0
172 J.S. Pedersen/Adv. CoUoid Interface Sci. 70 (1997) 171-210

modulus of the scattering vector are considered. Both X-ray and neutron
scattering are treated, however, the more detailed discussion of the
program implementations, in particular those concerning instrumental
smearing, have emphasis on neutron scattering. It is the intention that
the notes can serve as a practical guide in analyzing small-angle
scattering data. They contain a relatively brief, but self-contained,
description of linear and non-linear least-squares methods with empha-
sis on the applications in the analysis of small-angle scattering data.
The notes also contain a large collection of form factor and structure
factors, which are convenient to have at hand when analyzing experi-
mental data.
Small-angle scattering data are usually analyzed either by model-
independent approaches or by direct modeling. Both of these approaches
require the application of least-squares methods. The model-independent
approaches may consist of a Fourier transformation of the experimental
scattering curve, which provides the pair distance distribution function
p(r) or, equivalently, the correlation function ~(r), where the relation is:
p(r) = r2~(r). The Fourier transformation is usually done by the Indirect
Fourier Transformation (IFT) method introduced by Glatter [1,2]. This
method can be applied for all systems for which the correlations have a
finite range. It has several advantages compared to a direct Fourier
transformation, as it allows corrections for instrumental smearing ef-
fects and it does not require extrapolations of the data. Thep(r) function
provides real-space information and comparisons to model calculation
may provide key information and give suggestions for the structure of
the particles [2,3]. After interpretation of the p(r) function it may be
possible to construct a model on an analytical form, which can be fitted
to the data. For particles with centro symmetry it is possible to go one
step f u r t h e r with the model-independent analysis and perform a
SQuare-root DEConvolution (SQDEC) ofp(r) as described by Glatter, so
that the radial scattering length density profile p(r) is obtained [4,5].
The information obtained by this procedure would of course also be
incorporated in later attempts to perform model fits to the data. The
indirect Fourier transformation requires the application of a linear
least-squares method, whereas the square-root deconvolution procedure
requires the application of non-linear least-squares methods.
For polydisperse systems, the aim of the analysis is to extract the size
distribution of the particles when a particular shape of the particles is
assumed. For very dilute systems, this can be done with a free-form size
distribution [6] by a linear least-squares method, which may include a
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 173

n o n - n e g a t i v i t y constraint [7]. For systems with a finite concentration of

particles w h i c h interact with a h a r d - s p h e r e potential, the size distribu-
tion can also be d e t e r m i n e d on a free-form, but this requires a non-linear
least-squares m e t h o d [7].
For systems w i t h a collection of monodisperse particles, t h e r e exists
a m e t h o d for d e t e r m i n i n g the shape of particles model i n d e p e n d e n t l y by
fitting directly to the scattering data. This requires t h a t the particles
have a n e a r l y homogeneous distribution of scattering length density. A
multipole expansion is used for the shape of the particles and the
coefficients of this expansion are d e t e r m i n e d by n o n - l i n e a r least-
s q u a r e s methods [8,9]. The method has been modified so t h a t it can be
used for s h a p e d e t e r m i n a t i o n of the components of 'two-phase' particles
[9], and it has been applied successfully to the 50S s u b u n i t of the
ribosome ofE. coli [10-12].
W h e n applying the least-squares methods, it is i m p o r t a n t t h a t the
scientist u n d e r s t a n d the basic principles, so t h a t the c o m p u t e r p r o g r a m s
do not entirely w o r k as a "black box". With some basic u n d e r s t a n d i n g it
is possible to avoid the most common pitfalls and to u n d e r s t a n d , w h y
the p r o g r a m 'reacts' as it does, and perhaps, if the applied procedure
fails, to choose a n o t h e r strategy which works better. Section 2 of the
p r e s e n t notes gives an overview of the least-squares methods. Some of
the most common available small-angle scattering model expressions
are s u m m a r i z e d in Section 3, w h e r e a s an example of modeling of
small-angle scattering d a t a from block copolymer micelles is given in
Section 3.

2. L e a s t - s q u a r e s M e t h o d s

T h e r e exist m a n y excellent books on least-squares methods. Two of

these are the book by Bevington [13] "Data Reduction and E r r o r Analy-
sis for the Physical Sciences" and the book by Press, F l a n n e r y , Teukol-
sky, and Vetterling [14] "Numerical Recipes". The book by Bevington
gives a good introduction as well as a description of the most common
methods w i t h o u t excessive use of m a t h e m a t i c s . The book by Press,
F l a n n e r y , Teukolsky, and Vetterling describes in addition some of the
n u m e r i c a l aspects and problems w h e n i m p l e m e n t i n g the method. It
should also be noted t h a t this book is very useful w h e n i m p l e m e n t i n g
complex model expressions as it contains a large collection of routines
for special functions.
 174 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

T h e l e a s t - s q u a r e s m e t h o d employs t h e c h i - s q u a r e d ()~U)function as a
m e a s u r e for t h e deviation b e t w e e n t h e e x p e r i m e n t a l d a t a a n d t h e model.
Let/exp(qi), i -- 1,...~N be t h e d a t a points m e a s u r e d for t h e i n d e p e n d e n t
v a r i a b l e qi. In a s c a t t e r i n g e x p e r i m e n t IeXp(qi) is t h e m e a s u r e d intensi-
ties a n d qi is t h e m o d u l u s of t h e s c a t t e r i n g vector. T h e c o u n t i n g statistics
will give rise to t h e statistical u n c e r t a i n t i e s ~5i o n t h e d a t a point IeXp(qi).
T h e chi-squared is defined as:

N
~2 = ~ ~Iexp(qi) --(~i(/m°d(qi))2 (1)
i=1

w h e r e Im°d(qi) is t h e model i n t e n s i t i e s w h i c h d e p e n d s on t h e p a r a m e t e r s
ai, i = 1,...,M. It is often c o n v e n i e n t to consider t h e reduced chi-squared
~2, w h i c h is given by:

~2
Z2r- N - M (2)

w h e r e N - M is t h e number of degrees of freedom. T h e o p t i m u m set of

p a r a m e t e r v a l u e s for a model is d e t e r m i n e d by m i n i m i z i n g Eq. (1). A fit
w i t h Zr2 = 1 is c o n s i d e r e d to be an ideal fit. Note t h a t for N >> M a fit
w i t h )~2 = 1 h a s 'on average' IleXp(qi) -/m°d(qi) I = (~i, w h i c h m e a n s t h a t
t h e d e v i a t i o n s are on a v e r a g e equal to t h e statistical u n c e r t a i n t i e s . A
m o r e rigorous discussion of t h e chi-squared function can be found in t h e
text books m e n t i o n e d above.

2.1 Linear Method

T h e m o d e l i n t e n s i t y function is linear, if it can be w r i t t e n as:

M
/mod(q) = ~ a~X:k(q) (3)
k=l

w h e r e Xk(q) is a set of basis functions. This is t h e case for t h e indirect

F o u r i e r t r a n s f o r m a t i o n [1,2] for w h i c h t h e d i s t a n c e d i s t r i b u t i o n function
is w r i t t e n as:
 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210 175

p(r) = ~. akBk(r) (4)

k=l

w h e r e Bk(r) are cubic b splines. These are bell-shaped functions, com-

posed of piecewise third-order polynomials. One has

M
/mod(q) = 4X f p(r) sin(qr) dr= ~ a~(k(q) (5)
qr
k=l

where

r) dr
Xk(q) = 4~ f Bk(r) sin(q
qr
(6)

Alternative linear approaches for performing the IFT have been

described by Moore [15], Svergun, S e m e n y u k and Feigin [16] and
Svergun [17]. Note t h a t the m a x i m u m entropy method, which also can
be used, does give rise to a linear problem [18].
The d e t e r m i n a t i o n of size distributions on a free form is also a linear
problem. In this case

/rood(q) = Ap2 f N(R)F 2(q,R)dR (7)

w h e r e Ap is the scattering length density contrast, N(r) is the size

distribution and F(q,r) is the form factor amplitude. For homogeneous
spheres:

4~ 3 3 [sin(qR) - qR cos(qR)]
F(q,R) = - ~ R (qR) 3 (8)

Setting:
M

N(R) = ~_~akBk(R) (9)

k=l

w h e r e Bk(R) are cubic [6] or linear [7] b splines, the model intensity
becomes
 176 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

M
/mod(q) = ~ a~Xk(q) (10)
k=l

with

Xk(q) = AP2 f Bk(R)F2(q,R)dR (11)

w h i c h d e m o n s t r a t e s t h a t this is also a linear problem.

The chi-squared function can be minimized by m a n y different meth-
ods, for example by m a k i n g a qualified guess on the values of the
p a r a m e t e r s and t h e n simply v a r y the p a r a m e t e r s one by one so t h a t
successively lower values of chi-squared are obtained. Such a simple
(but time consuming!) grid search method would work. It would, how-
ever, be b e t t e r to t a k e a d v a n t a g e of the general properties of the
chi-squared function and of the fitting function.
The m i n i m u m of chi-squared occurs w h e r e the partial derivatives of
(1) w i t h respect to a k are equal to zero:

~Z2 - 0 for k = 1,...,M (12)

~ak

E q u a t i o n s (1) a n d (3) give:

E Iexp(qi)- E aj~(qi) Xk(x i) = 0 for k = 1,...,M (13)

i=1 j=l

w h i c h is e q u i v a l e n t to the normal equations:

M

,T__., = 13k (14)

j=l

where
N
Xflqi~k(qi) (15)
i=1
 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210 177

and

N
Iexp(qi~k(q i)
~k = ~ 2 (16)
i=1 (~i

In m a t r i x notation

A. a = b (17)

w h e r e [A]ij = aij and [hi k = ~k"

The values for a k t h a t minimize chi-squared is t h u s d e t e r m i n e d by
solving a set of linear equations. It should be noted [14,19] t h a t the
equations should not be solved by n u m e r i c a l l y calculating the inverse
m a t r i x a n d multiplying it on both sides of the equations. If the equations
are close to being singular (which is not an u n u s u a l situation) the
n u m e r i c a l calculations will give an accumulation o f r o u n d - o f f e r r o r s and
the final solution for a k will not fulfill the original equation (17). It is
m u c h b e t t e r to use a more robust method like Gauss--Jordan elimination
w i t h pivoting [14,19] for solving the equations.
For e s t i m a t i n g the errors on aj a basic rule for accumulation of errors
is applied. For the function flxl,...,x N) of the p a r a m e t e r s x i with k n o w n
errors (~(xi):

N
(18)

This e q u a t i o n is valid if the ajs are i n d e p e n d e n t p a r a m e t e r s . If the

p a r a m e t e r s aj are considered to be functions of the observed intensities
IeXp(qi) so t h a t aj(IeXp(ql),...,Iexp(qN)) , t h e n

(19)
3Iexp(qi) °3
i=l

In order to calculate the q u a n t i t y in the brackets, the formal m a t h e -

matical solution to (17) is used:
178 J,S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

M M N t [ J~k[ jTeXp'q"i'~'qi"
(20)
k=l k=l i=1

The derivative of this equation w i t h respect to IeXp(qi) , w h i c h in this

context is considered to be a p a r a m e t e r , is t h e n calculated:
M
aa: Xk(qi)
a/~Xp(qi) - ~ [A-~Ij~ ~ (2i)
k=l

I n s e r t i n g this in (19) gives

M M

~2(a) = ~ ~ [A-lijk [A-1]jl ~ Xk(qi) Xl(qi) 1 (22)

k=l l=l
N

Noting t h a t [A]jk = ajk = ~_~ Xj(qi) Xk(qi)/~2i the final result is

i=1
,52(aj) = [A-1ljj (23)

w h i c h m e a n s t h a t the square of the errors are given by the diagonal

e l e m e n t s of A -1. If the reduced chi-squared Z2 at the m i n i m u m is larger
t h a n one, it is common to set
g2(aj) = Zr2[A-1]jj (24)

as this to a certain extent takes into account the short-comings of the

applied model function and/or systematic errors in the data.
U s i n g (24) the errors on, for example, p(r) can be e s t i m a t e d as
M M
(~2[p(r)] = ~ (~2(ak)Bk(r)2 = ~2 ~., [A-1]kkBk(r)2 (25)
k=l k=l

This equation has the short-coming t h a t it a s s u m e s t h a t the ajs are

i n d e p e n d e n t and t h u s neglects the covariances of the p a r a m e t e r s a N.
T a k i n g this properly into account [14] modifies the expression to
M M

~2[P (r)] =Xr2 E E [A-1]klBk(r)Bl (r) (26)

k=l l=l
 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210 179

F o r t h e indirect F o u r i e r t r a n s f o r m a t i o n [1] t h e n u m b e r of basis func-

tions in Eq. (4) u s u a l l y h a s to be quite large (30-60) in o r d e r to give a
sufficient r e s o l u t i o n for p(r). However, this often r e s u l t s in a n e a r l y
s i n g u l a r set of e q u a t i o n s a n d large oscillations in p(r). T h e f u n c t i o n is
expected to h a v e a relatively s m o o t h behavior, a n d t h e r e f o r e a s m o o t h -
n e s s c o n s t r a i n t on p(r) is applied. This is done in such a w a y t h a t t h e
n o r m a l e q u a t i o n s r e m a i n a l i n e a r set of equations. T h e e x p r e s s i o n
~2 + LN c is m i n i m i z e d i n s t e a d of X2, w h e r e N e is t h e m e a s u r e of t h e
M-1
s m o o t h n e s s , N c = ~(aj+l--aj) 2 + a ~ + a 2 [1,20], a n d ~. is a c o n s t a n t w h i c h
j=l
can easily be chosen by t h e point-of-inflection m e t h o d [1]. A l t e r n a t i v e
m e t h o d s for choosing ~ can be found in [21].
F o r d e t e r m i n a t i o n of size d i s t r i b u t i o n s it is physically r e a s o n a b l e to
a p p l y a n o n - n e g a t i v i t y c o n s t r a i n t for N(r) as well as a s m o o t h n e s s
c o n s t r a i n t . T h i s can be done by r e d u c i n g t h e n o r m a l e q u a t i o n s in a
s y s t e m a t i c way, so t h a t only those t h a t give n o n - n e g a t i v e v a l u e s are
kept. T h i s a p p r o a c h also gives a l i n e a r set of e q u a t i o n s [22].
T h e application of t h e s m o o t h n e s s c o n s t r a i n t h a s in b o t h t y p e s of
a p p l i c a t i o n s t h e c o n s e q u e n c e t h a t Eq. (26) c a n n o t be u s e d for a reliable
d e t e r m i n a t i o n of t h e errors on t h e d i s t r i b u t i o n functions. It can in t h e s e
cases be r e c o m m e n d e d to use t h e Monte Carlo m e t h o d [14,19]. In this
m e t h o d a large set of a d d i t i o n a l ' e x p e r i m e n t a l ' d a t a sets (typically NMC
= 50) a r e g e n e r a t e d from t h e original d a t a set by a d d i n g r a n d o m errors
to t h e original d a t a sets of t h e s a m e m a g n i t u d e as those of t h e original
e x p e r i m e n t a l data. T h e s e d a t a sets are a n a l y z e d a n d give t h e f u n c t i o n s
pi(r), i=1,..., NMC a n d t h e errors on p(r) are calculated as
NMC
1
(~2[p(r)] = ~ ~_, [Pi(r) - p ( r ) ] 2 (27)
i=l

T h i s a p p r o a c h a u t o m a t i c a l l y t a k e s into account t h e covariances a n d it

can of course also be u s e d for d e t e r m i n i n g t h e errors on o t h e r p a r a m e t e r s
d e r i v e d from aj. It s h o u l d be n o t e d t h a t t h e ' e x p e r i m e n t a l ' d a t a only
e n t e r t h e r i g h t - h a n d side of t h e n o r m a l equations. W h e n t h e G a u s s - J o r -
d a n e l i m i n a t i o n p r o c e d u r e is u s e d for solving t h e e q u a t i o n s , t h e different
r i g h t - h a n d sides can be t r e a t e d s i m u l t a n e o u s l y , a n d t h e e q u a t i o n s h a v e
to be solved only once [19].
 180 J.S. Pedersen/[Link]. 70 (1997)171-210

E x p e r i m e n t a l scattering d a t a are always influenced by i n s t r u m e n t a l

s m e a r i n g , and it m a y be necessary to include corrections for this in the
d a t a analysis. Owing to the finite resolution of the i n s t r u m e n t the
s c a t t e r i n g in a region a r o u n d the nominal scattering vector (q) is probed.
For small-angle n e u t r o n scattering, the distribution of scattering vec-
tors q can be described by a resolution function R((q),q) (see, for example
Ref. [23]), and the s m e a r i n g effects can be t a k e n into account in the
model scattering i n t e n s i t y by
M
/mod((q)) = ~ ak Xk((q}), w h e r e ~:k((q}) = ~ R((q),q) Xk(q)d q (28)
k=l
The e x p e r i m e n t a l d a t a which e n t e r the expression for chi-squared (1)
should be w r i t t e n as I((qi)) as t h e y are recorded for the nominal scatter-
ing vectors. For small-angle X-ray scattering d a t a recorded using a
long-slit camera, the s m e a r i n g is calculated using the u s u a l procedure,
w h i c h involves w e i g h t i n g functions (see, e.g. Ref. [1]). This procedure is
n u m e r i c a l l y more complicated as t h r e e integrals have to be calculated,
however, the result is in principle the s a m e as described by Eq. (28) as
it is the basis functions which are smeared.
In most small-angle n e u t r o n scattering experiments, the d a t a are
recorded using more t h a n one i n s t r u m e n t a l setting. E a c h setting corre-
sponds to a set of values of the wavelength, w a v e l e n g t h resolution,
collimation and sample-to-detector distance. Therefore each of the Nse t
settings has its own resolution function Ri((q),q), and it is therefore
convenient to write

~2 = N~''tE N(J) (lTP((qi)) ~.-i ~J °d((qi}) ) (29)

j=l i=1
w h e r e the index j refers to which d a t a set is considered. The model
i n t e n s i t y is t h e n
M
~i°d((q})=~akXk.j((q}), where Xkj((q})=~Rj((q),q)Xk(q)dq (30)
k=l
The corresponding changes to the n o r m a l equations ((13)-(16)) consist
in c h a n g i n g the s u m m a t i o n s over i = 1,...,N to a double s u m m a t i o n s over
j = 1,...,Nse t and i = 1,...JY(j) and changing qi to (qi).
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 181

It is often so t h a t n o t all d a t a sets are k n o w n on a n a b s o l u t e scale,

a n d f u r t h e r m o r e , even if t h e y are, t h e r e are typically small (systematic)
e r r o r s of a few p e r c e n t in t h e scale. As t h e errors from c o u n t i n g statistics
can be less t h a n one percent, t h e errors on t h e scales are q u i t e i m p o r t a n t
w h e n fitting t h e data. T h e scale factors of t h e d a t a sets s h o u l d t h e r e f o r e
be a d j u s t e d . S v e r g u n [24] h a s s u g g e s t e d a m e t h o d for a d j u s t i n g t h e scale
of one d a t a set w i t h respect to a n o t h e r , w h i c h is s o m e w h a t complicated:
t h e n o r m a l e q u a t i o n s for t h e p r o b l e m w i t h o u t t h e s m o o t h n e s s c o n s t r a i n t
plus a n e x t r a l i n e a r e q u a t i o n for t h e scale factor are solved by s i n g u l a r
v a l u e d e c o m p o s i t i o n [25]. However, a s i m p l e r a p p r o a c h can also be u s e d
[26]. F i r s t t h e m o s t reliable d a t a set is chosen so t h a t t h e o t h e r sets can
be scaled to m a t c h this one. T h e n t h e d a t a sets are scaled so t h a t t h e y
agree w i t h i n 10-25% in t h e overlap region, a n d t h e c o n s t a n t ~. is
d e t e r m i n e d by t h e point-of-inflection m e t h o d . T h e v a l u e o f k is k e p t fixed
while t h e scale factors are optimized. It s h o u l d be n o t e d t h a t it is only
t h e ~ks w h i c h d e p e n d on t h e scale factors, a n d t h e r e f o r e it is n o t
n e c e s s a r y to recalculate (and s m e a r by i n s t r u m e n t a l resolution) t h e aijs
a n d t h e basis f u n c t i o n s w h i c h e n t e r ~j. T h e scale factors can be a d j u s t e d
relatively fast to an accuracy b e t t e r t h a n 0.5% by a s i m p l e grid s e a r c h
c o m b i n e d w i t h a parabolic a p p r o x i m a t i o n of ~2 close to t h e m i n i m u m
[13,26]. If t h r e e or m o r e d a t a sets h a v e been recorded it m a y be
n e c e s s a r y to go t h r o u g h t h e d a t a sets several t i m e s a n d a d j u s t scale
factors u n t i l t h e p r o c e d u r e is converged. It can be r e c o m m e n d e d to
d e t e r m i n e t h e scale factors by this p r o c e d u r e before p e r f o r m i n g least-
s q u a r e s fit u s i n g a n a l y t i c a l models as it can save a lot of c o m p u t e r time.

2.2 Non-linear Method

If t h e fitting f u n c t i o n is n o t a l i n e a r function of t h e p a r a m e t e r s , t h e
l e a s t - s q u a r e s p r o b l e m is said to be non-linear. As for t h e l i n e a r m e t h o d ,
t h e c h i - s q u a r e d f u n c t i o n can s i m p l y be m i n i m i z e d by a grid s e a r c h in
w h i c h one m a k e s a qualified guess on t h e v a l u e s of t h e p a r a m e t e r s a n d
t h e n successively a n d r e p e a t e d l y optimizes t h e p a r a m e t e r s one by one.
However, this grid s e a r c h m e t h o d is v e r y t i m e c o n s u m i n g a n d it is often
a d v a n t a g e o u s to a p p l y m o r e a d v a n c e d m e t h o d s . D u e to t h e n o n - l i n e a r i t y
of t h e fitting f u n c t i o n all of t h e m e t h o d s r e q u i r e a set of s t a r t i n g v a l u e s
for t h e p a r a m e t e r s .
A m e t h o d , w h i c h is b e t t e r t h a n t h e grid s e a r c h a n d also relatively
s i m p l e is the gradient m e t h o d also k n o w n as t h e steepest descent m e t h o d
[13,14]. In this m e t h o d one calculates analytically or m o r e often n u m e r i -
182 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210

cally, t h e g r a d i e n t of chi-squared, ~'Z2]~aj. T h e search for a n e w set of

p a r a m e t e r values is done along the negative direction of the gradient.
All p a r a m e t e r s are c h a n g e d s i m u l t a n e o u s l y and the new values a n e w a r e
o b t a i n e d from the old set aold by

a n e w ----aol d + ~ a = a o l d + c o n s t a n t × b (31)

w h e r e b is the negative g r a d i e n t ([b] i = -~z2(aold)~ai ) and the constant

is a p p r o p r i a t e l y chosen, so t h a t Z2 decreases. The steepest descent
m e t h o d is effective relatively far a w a y from the m i n i m u m . As the
m i n i m u m is approached it becomes g r a d u a l l y more inefficient.
A m e t h o d which works b e t t e r close to the m i n i m u m is based on a set
of e q u a t i o n s similar to the n o r m a l equations for the linear problem.
These equations h a v e to be solved iteratively [13,14]. Let a i be an
e s t i m a t e of the p a r a m e t e r s which is not too far from the m i n i m u m of~ 2.
A m u l t i - p a r a m e t e r Taylor expansion of chi-squared a r o u n d a gives:

M M
~2Z___~2
z2(a')=z2(a)+2 ~at2. ( a ) ~ t i + 2 20aiOaj(a)Sai~i~tJ +''"
i=l ij=l

~ C - b . 5 a + 1 5a. A . 5a (32)
2

w h e r e = Da = a' - a and

_ ~2~2
c = z2(a) [b]i = OZ----~2(a) [Alij - (a) (33)
~ai " ~ai Oaj

U s i n g t h e approximation in the second line of (32) the g r a d i e n t of ~2

in a ' c a n be e s t i m a t e d as

DZ2 (a') = A. 5a - b (34)

Oa i

For a" = ami n this g r a d i e n t is equal to zero and one has the n o r m a l
equations

A. 5a = b (35)
 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210 183

w h e r e A a n d b are calculated at the old value of a. Thus, one obtains

(formally) a new e s t i m a t e for a by

anew = aold + 5a = aold + A -1 • b (36)

This m e t h o d can be applied successively until the changes in X2 are

negligible. However, t h e r e is one problem in applying this m e t h o d
directly as it requires t h a t both the first order and second order deriva-
tives are available. It is often only practically possible to calculate the
derivatives numerically. This is quite time consuming as the calculation
of the second order derivative requires the calculation of M(M-1)/2
t e r m s and each of these requires the evaluation of the fitting function
in N values. It is more efficient [13] to use a first order Taylor expansion
of the fitting function/m°d(q,a):

M ~/mod(q;a) M
/m°d(q;a') =/m°O(q;a) + E Oaj ~:}aj=-/m°d(q;a) + ~ 5ajXj(q;a) (37)
j=l j=l

w h e r e a' = a + 5a a n d

~/m°d(q;a)
Xj(q;a) - ~aj (38)

I n s e r t i n g t h e right-hand-side of(37) in the expression for chi-squared

(Eq. (1)), and considering [ieXp(q)-/mod(q;a)] as the 'experimental' data,
it is s t r a i g h t forward to see t h a t the n o r m a l equations for 5aj for the
linearized function are the s a m e as those for aj in the linear problem,
w i t h [iexp(q) _/mod(q;a)] instead ofleXp(q). Hence,

N Xd(qi;a)Xk(qi;a) and
N
~k=~_, [/eXp(q) _/mod(q;a)] Xk(qi;a)
(39)
i=l (~i i=1 ($i

w h e r e Xj(q;a) = 3Im°d(q;a)/3aj. A n u m e r i c a l calculation of these deriva-

tives r e q u i r e s only on the order o f M x N calculations.
The m e t h o d w i t h linearized fitting functions works best close to the
m i n i m u m of chi-squared, w h e r e a s the g r a d i e n t search is b e t t e r f u r t h e r
away. Ideally, t h e two methods should be combined, so t h a t t h e m e t h o d s
184 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210

are used in the r a n g e w h e r e they are best suited. T h e r e exists an

a p p r o a c h suggested by M a r q u a r d t [27] which smoothly combines the
two methods. A p a r a m e t e r k (not to be confused w i t h the prefactor of
the s m o o t h n e s s m e a s u r e in the IFT methods) is used for e n h a n c i n g the
diagonal e l e m e n t s of the m a t r i x [A]ij = aij so t h a t the n o r m a l equations
become

A' • 5a = b (40)

with

• = Ic~/j(1 + ~,) for i = j

(41)
aiJ [~j for iv j

For ~. small, the n o r m a l equations are basically u n c h a n g e d and the

m e t h o d is efficient close to the m i n i m u m . For ~. large, the n o r m a l
e q u a t i o n s reduce to a set of non-coupled equations:

~'qi] 5a/= ~j f o r j = 1,...,M (42)

w h i c h h a v e the solutions

1
(43)

Noting t h a t [~j is the negative gradient, a comparison w i t h Eq. (31)

shows t h a t the m e t h o d in this case is the s a m e as the steepest descent
m e t h o d w i t h the constant equal to 1/(~xzjj). Chi-squared will always
decrease for sufficiently large values of ~.. The c h a r a c t e r of the search
m e t h o d is t h u s d e t e r m i n e d by the value of ~., and ~. should be large far
a w a y from the m i n i m u m and small close to the m i n i m u m . M a r q u a r d t
[27] h a s designed the following procedure:
. Calculate ~2(a).
2. Set k = 0.001.
3. Calculate A and b.
4. Calculate A' for the c u r r e n t X and d e t e r m i n e 5a.
5. If ~2(a + 5a) > )~2(a), multiply )~ by 10 and r e p e a t (4).
6. Ifz2(a + 5a) < z2(a), divide ~ by 10, t a k e a + 5a to be the c u r r e n t
a and r e p e a t (3).
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 185

The iterations are continued until the decrease ofx 2 becomes negligible.
It should be noted that the normal equations, as for the linear problem,
should be solved using a numerically stable method like the Gauss-Jor-
dan elimination procedure [14].
The standard errors on the parameters aj can be estimated as for the
linear problem:

2 -1
(~(aj) = X r [ A ]jj (44)

However, this method only works if the fitting function is not strongly
non-linear and if the parameters are not strongly correlated. Another
and more reliable approach is to determine (~(aj) as the value for which
chi-squared increases by one, when aj is fixed at aj + g(aj) and the other
parameters are optimized [13]. If the reduced chi-squared is not equal
to one for the optimum values of the parameters, the increase in
chi-squared should be taken as X2 at the min imu m instead of one. The
advantage of this method is that it takes into account the correlation
between the parameters.
It should be noted that the method by Marquardt [27] for optimizing
chi-squared may not work for highly non-linear fitting functions with
significant correlations between the fitting parameters. For such prob-
lem it can be recommended to use the simple grid search in which the
parameters are optimized one by one repeatedly. This approach has the
advantage that it always works although it is very time consuming. The
grid search may be able to bring the parameter values sufficiently close
to the optimum values, that the method of Marquardt can be used.
The i n s t r u m e n t a l smearing effects should, as discussed for the linear
problem, be included in the calculation of the model function. For X-ray
scattering in the long-slit geometry this is done using the weighting
functions [28] and performing three numerical integrations. This is
quite time-consuming and makes it difficult to perform fits of compli-
cated model functions.
For small-angle neutron scattering the model function can be written
as

d¢y(q)
Im°d((q))= rJ R((q),q) ~-~ dq (45)

where d~(q)/d~2 is the ideal scattering intensity (the cross section, see
next section), and R((q),q) is the resolution function described in the
 186 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210

p r e v i o u s section. If several i n s t r u m e n t a l settings h a v e b e e n u s e d for

r e c o r d i n g t h e data, this of course h a s to be t a k e n into account.
T h e a c t u a l calculation of t h e i n t e g r a l in (45) is done n u m e r i c a l l y . F o r
s m a l l - a n g l e n e u t r o n s c a t t e r i n g e x p e r i m e n t s , t h e r e s o l u t i o n f u n c t i o n is
well a p p r o x i m a t e d by a G a u s s i a n function [23,29,30] a n d only a b o u t 10
p o i n t s is r e q u i r e d for calculating t h e integral. In an efficient i m p l e m e n -
t a t i o n [31] t h e r e s o l u t i o n function is calculated once a n d for all in t h e
b e g i n n i n g of t h e p r o g r a m , a n d saved for later use. In this w a y u n n e c e s -
s a r y r e p e a t e d calculations are avoided. It s h o u l d also be n o t e d t h a t
a l t h o u g h only 10 p o i n t s are r e q u i r e d in t h e convolution, it m e a n s t h a t
t h e cross section s h o u l d be e v a l u a t e d for 10 × N values, w h e r e N is t h e
n u m b e r of m e a s u r e d points. If t h e expressions for t h e cross section are
complex, t h e calculation t i m e s can be g r e a t l y r e d u c e d by m a k i n g u s e of
m a s t e r c u r v e s a n d i n t e r p o l a t i o n s [29].

3. M o d e l s

In t h i s section some of t h e m o s t c o m m o n expressions for form factors

a n d s t r u c t u r e factors will be reviewed. It is i m p r a c t i c a l to give all t h e
m a t h e m a t i c a l e x p r e s s i o n s in t h e p r e s e n t notes, a n d for s o m e of t h e
m o d e l s only t h e references to t h e original p a p e r s are given. T h e r e a d e r
is e n c o u r a g e d always to look into t h e original p a p e r s in order to check
t h e r a n g e of validity of t h e expressions (and for checking for possible
t y p i n g e r r o r s in t h e p r e s e n t notes).
T h e differential s c a t t e r i n g cross section dc(q)/d~2 of a s a m p l e can be
d e f i n e d as t h e n u m b e r of s c a t t e r e d n e u t r o n s or p h o t o n s per u n i t time,
relative to t h e i n c i d e n t flux of n e u t r o n or photons, per u n i t solid angle
at q p e r u n i t v o l u m e of t h e sample. T h e flux is t h e n u m b e r of n e u t r o n
or p h o t o n s p e r u n i t t i m e a n d per u n i t a r e a at t h e s a m p l e position. It is
c o n v e n i e n t to use t h e cross section as it does n o t d e p e n d on t h e form or
t r a n s m i s s i o n of t h e sample.
For a m o n o - d i s p e r s e collection of (spherically s y m m e t r i c ) particles
t h e s c a t t e r i n g cross section can be w r i t t e n as

d(~(q) _ nap2 V2 p(q) S(q) (46)

d~

w h e r e n is t h e n u m b e r d e n s i t y of particles, Ap is t h e difference in
s c a t t e r i n g l e n g t h d e n s i t y b e t w e e n t h e particles a n d t h e s o l v e n t / m a t r i x ,
 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210 187

V is t h e v o l u m e of t h e particles, P(q) is t h e particle form factor a n d S(q)

is t h e structure factor. T h e form factor describes t h e s t r u c t u r e of t h e
particles a n d fulfils P(q = 0) = 1. T h e s t r u c t u r e factor describes t h e
i n t e r f e r e n c e of s c a t t e r i n g from different particles a n d c o n t a i n s informa-
tion on t h e i n t e r a c t i o n b e t w e e n t h e particles. For v e r y d i l u t e s y s t e m s
S(q) = 1 a n d t h e s t r u c t u r e factor can be neglected.
A n a l t e r n a t i v e a p p r o a c h is to define t h e cross section p e r u n i t m a s s
of solute i n s t e a d of p e r u n i t v o l u m e of t h e sample. This cross section can
be w r i t t e n as

d(~m(q) _ 1 Ap2 M2 P(q)S(q) = Ap2 MP(q)S(q) (47)

dr2 M

w h e r e M is t h e m o l e c u l a r m a s s of a particle a n d Apm is t h e s c a t t e r i n g
l e n g t h difference p e r u n i t m a s s of t h e solute. T h e factor 1/M is t h e
n u m b e r of particles p e r u n i t m a s s of solute.
T h e above e x p r e s s i o n s [46,47] implicitly a s s u m e s p h e r i c a l s y m m e t r y
of t h e particle s h a p e a n d t h e interactions. For anisotropic identical
particles t h e cross section is

- AP2 V2 Fi(q,ei 12 + Z Fi(q,ei) Fj(q,ej) Sij(q , el, ej) (48)

ij

w h e r e t h e s u m s are over all particles in t h e s a m p l e a n d Fi(q,e i) is t h e

a m p l i t u d e of t h e form factor for t h e i t h particle w i t h o r i e n t a t i o n given
by t h e u n i t vector e i. T h e Sij(q,el,e j) f u n c t i o n s are t h e p a r t i a l s t r u c t u r e
factors w h i c h d e p e n d on orientations. Note t h a t t h e first t e r m in Eq.
(48) is j u s t t h e o r i e n t a t i o n a l a v e r a g e d form factor: P(q) - (F2(q))0 .
F o r particles w i t h only a small anisotropy, it can be a s s u m e d t h a t t h e
i n t e r a c t i o n s are i n d e p e n d e n t of t h e o r i e n t a t i o n a n d given by t h e a v e r a g e
size of t h e particles. This leads to t h e decoupling approximation [32].

do(q)
d~ - Ap2 V2p(q)[1 + [3(q)(S(q) - 1)] (49)

where

~(q) = (F(q)) 2 / (F2(q))o (50)

 188 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210

a n d S(q) is t h e s t r u c t u r e factor calculated for t h e a v e r a g e particle size.

F o r polydisperse s y s t e m s it is also not possible to write t h e s c a t t e r i n g
cross section as p r o d u c t of a form factor a n d a s t r u c t u r e factor. In this
case t h e s c a t t e r i n g cross section h a s t h e form:

de(q)
dg/- Ap2 D(R) V(R) 2 F(q,R) 2 dR

(51)

+ ~ f D(R)V(R)D(R')V(R')F(q,R)F(q,R')S(R,R',q)dRdR'
0 0

w h e r e D(R) is t h e n u m b e r size distribution, V(R) is t h e v o l u m e of a

particle w i t h r a d i u s R a n d form factor F(q,R), a n d S(R,R',q) are p a r t i a l
s t r u c t u r e factors.
For s y s t e m s w i t h small polydispersities, a d e c o u p l i n g a p p r o a c h simi-
lar to t h e one for anisotropic particles [32] can be used. It is a s s u m e d
t h a t i n t e r a c t i o n s are i n d e p e n d e n t of size. W i t h this one obtains:

d(~(q)
dr/ - Ap2 <V2p(q)> [1 + ~(q)(S(q) - 1)] (52)

w h e r e (VeP(q)) = ~ D(R)V(R)2F(q,R)2dR a n d

(53)

a n d S(q) is t h e s t r u c t u r e factor calculated for t h e a v e r a g e particle size.

N o t e t h a t Eqs. (51,52) can also be u s e d for slightly anisotropic particles,
if Fi(q,R) is replaced by (Fi(q,R)} o a n d Fi(q,R) 2 is replaced by (Fi(q,R)2}o .
It is also possible to use a local monodisperse approximation for
i n c l u d i n g t h e effects of polydispersity [7]. In this a p p r o a c h it is a s s u m e d
t h a t a particle of a c e r t a i n size., is always s u r r o u n d e d by particles w i t h
t h e s a m e size. Following this t h e s c a t t e r i n g is a p p r o x i m a t e d by t h a t of
m o n o d i s p e r s e s u b - s y s t e m s w e i g h t e d by t h e size distribution:

de(q) _ AP2 f D(R)V(R)2F(q,R)2S(q,R)dR (54)

dgl
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 189

in w h i c h it has been indicated t h a t the s t r u c t u r e factor is for particles

of size R. This approach works better t h a n the decoupling approximation
(52) for systems w i t h larger polydispersities and h i g h e r concentrations.

3.1. Form Factors

In the following it will be a s s u m e d t h a t the particles are r a n d o m l y

oriented in the sample so t h a t the theoretical form factors for anisotropic
particles have to be averaged over orientation. Note t h a t for spherical
objects t h e form factor can be w r i t t e n as P(q) = F2(q), w h e r e F(q) is the
a m p l i t u d e of the form factor.

(1) Homogeneous sphere

The form factor of a homogeneous sphere was calculated a l r e a d y in
1911 by Lord Rayleigh [33]. For a sphere with radius R:

3 [sin(qR) - qR cos(qR)]
FI(q,R) - (55)
(qR) 3

(2) Spherical shell

This form factor is obtained by subtracting the e m p t y core w i t h a
proper w e i g h t i n g by the volumes:

V(R1)FI(q ~ I) - V(R2)FI(q,R2)
(56)
F2(q) - V(R1 ) - V(R2)

w h e r e V(R) = 4~R3/3 and R 1 and R 2 are the outer and i n n e r r a d i u s of

the shell, respectively. An infinitely thin shell w i t h r a d i u s R has the
form factor F2(q)' = sin(qR)/(qR).

(3) Spherical concentric shells

This form factor is a generalization of the shell form factor. Let Ri, i
= 1JV be the radii of the shells and Pi be their scattering densities. With
this:
N
1
pl V(R1)FI(q,R1) + ~., (Pi - Pi-1)V(Ri)FI(q,Ri) (57)
i=2
190 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210

where M 3 is the scattering mass or scattering volume of the particle,

given by:
N
M3 = PlV(R1) + ~ V ( R i ) ( P i - Pi-1) (58)
i=2

(4) Particles consisting of spherical subunits

The expression was derived by Debye in 1915 [34]. For a particle
consisting of P subunits:

P
1 sin(qrij)
P4(q) = M2 ij=l
~ M3(i) M3(]') F3(q'i) F3(qJ) qrij (59)

where F3(q,i) and M3(i) are the form factor and scattering mass of the
ith particle, respectively, rij is the distance between the centers of the
ith and t h e j t h subunit, and

P
M 4 = ~ M3(i) (60)
i=1

(5) Ellipsoid of revolution

This expression was determined by Guinier [35]. The averaging over
orientations has to be done numerically. For the semi axes R~R,aR:

~12
P5(q,R,e) = ; F2[q,r(R,e,cz)l sin a da (61)
0

where r(R,e,a) = R(sin2a + e2cos2{~) 1/2. It is straight forward to generalize

(61) for concentric elliptical shells. Fl(q,r) has to be replaced by F3(q) in
which the volumes are V(R) = 4hER3/3. Note, that the different shells
can have different values of ~.. The form factor of an infinitely thin
elliptical shell is given by (61) with Fl(q,r) replaced by F2(qr)'.

(6) Tri-axial ellipsoid

For this object two numerical integrations have to be performed in
order to get the orientational average. For the semi axes a,b,c [36]:
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 191

¢/2 ~/2

P6(q,a,b,c)= 2 ~ ~ F2[q,r(a,b,c,a,~)] sin a d a d ~ (62)

0 0

where r(a,b,c,a,~) = [(a 2 sin2~ + b2cos2~) sin2a + c2cos2a] 1/2. The form
factor for the tri-axial ellipsoid can be generalized in the same way as
(61) for tri-axial ellipsoids consisting of concentric shells. The form factor
of an infinitely thin elliptical shell is given by (62) with Fl(q,r) replaced
by F2(qr)'.

(7) Cube and rectangular parallelepipedons

Two orientational averages have to be performed. For the edge
lengths a,b,c [37]:

P7(q,a,b,c) =
(63)
~12 ~12
2 ; ~ sin(qa sin~ cos~) sin(qb sin(~ cos~) sin(qc cos(x) sina d~ dR
0 0
qa sin(x cos[~ qb sina sin[~ qc cos~

(8) Truncated octahedra

The equations for an oriented particle were given by Hendricks,
Schelten and Schmatz [38]. The orientational average has to be done as
in Eq. (63).

(9) Faceted sphere

The equations for an oriented particle were given by Dubey [39] (see
also Ref. [40]). The orientational average has to be done as in Eq. (63).

(10) Cube with terraces

The equations for an oriented particle were given by Rodriguez,
GSmez Sal, Moreno, de Geyer, and Janot [41]. The orientational average
has to be done as in Eq. (63).

(11) Cylinder
The expression for a cylinder with radius R and length L was given
by Fournet [42].

~i212Bl(qR sina) sin((qL cosa)/212

Pll(q)= o L q-R~ (~cos--~ J sinad(z (64)
192 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

where B l(x) is the first order Bessel function. An expression for cylinders
consisting of concentric shells can be constructed by an approach similar
to the one used for spherical particles. The form factor of an infinitely
thin cylindrical shell with closed ends is given by (64) with the terms in
the square brackets replaced by 2B0(q R sin ¢z) cos [(qL cos a)/2].

(12) Cylinder with elliptical cross section

This expression was given by Mittelbach and Porod [43] for a cylinder
of length L and with cross-section semi axes a and b:
7 t / 2 ~t/2 ~-
2 ~ f ]2Bl(qr(a,b,¢,c~) sin((qL cos(x)/2)- dO sincx da (65)
P12(q) = ~ 0 0 L qr(a,b,¢,cO ((qL cosa)/2)

where r(a,b,O,a) = [a2sin20 + b2cos20]1/2 sins. An expression for cylinders

consisting of concentric shells with elliptical cross section can be con-
structed by an approach similar to the one used for spherical particles
and for tri-axial ellipsoids. The shells can have varying eccentricities
(a/b).

(13) Cylinder with spherical end-caps

The equations were given by Cusack [44]. Expressions for a shell
particle were also given.

(14) Infinitely thin rod

The expression was determined by Neugebauer [45]:

P14(q) = 2Si(qL)/(qL) - 4 sin2(qL/2)/(q2L 2) (66)

where
X

Si(x)=~ t - l s i n t d t (67)
0
and L is the length.

(15) Infinitely thin circular disk

The expression was determined by Kratky and Porod [46]"
2 [ Bl(2qR)-
P14(q) = q2R2 1 qR (68)
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 193

w h e r e R is the radius of the disk.

(16) Fractal aggregates

An empirical expression for a mass fractals consisting of spheres with
a radius R has been given by Teixeira [47]:

P16(q) = 1+ )D [1+ 1/~-q2~]~_1)/2 s i n [ ( D - 1) tan-l(q~)] F2(q) (69)

w h e r e D is the fractal dimension, ~ is a cut-off length for the fractal

correlations, and F(x) is the g a m m a function. (Note t h a t expressions for
fractal surfaces have been given by Bale and Schmidt [48].)

(17) Flexible polymers with Gaussian statistics

Flexible polymer chains which are not self-avoiding obey Gaussian
statistics. Debye [49] has calculated the form factor of such chains:

P17(q) = 2[exp(-u) + u - 1]/u 2 (70)

with u = (/?22)q2, w h e r e (R2) is the ensemble average radius of gyration

squared: (R~) = (Lb)/6, where L is the contour length and b is the
statistical (Kuhn) s e g m e n t length.

(18) Flexible self-avoiding polymers

Empirical expressions have been given by Utiyama, T s u n a s h i m a and
K u r a t a [50]. The p a r a m e t e r s should be taken as e = 0.176, t = 2/(1-~),
and s = 2.90 (see Ref. [51] which also contains a simple approximation).

(19) Semi-flexible polymers without self-avoidance

Numerical interpolation formulae for the K r a t k y - P o r o d model [52]
have been given by Yoshizaki and Y a m a k a w a [53]. These have recently
been corrected using results from Monte Carlo simulations [51].

(20) Semi-flexible polymers with self-avoidance

Numerical interpolation formulae have been given by Pedersen and
S c h u r t e n b e r g e r [51]. The results are given for R/b = 0.1, w h e r e R is the
cross section radius and b is the K u h n length. This corresponds to a
reduced binary cluster integral of 0.3, which is similar to the value found
for polystyrene in a good solvent.
194 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

(21) Star polymer with Gaussian statistics

The expression was given by Benoit [54]. For a star with f arms:

P21(q) _- fv22 [ v _ [ l _ e x p ( _ v ) ] + ~ [ l _ e x p ( _ v ) ] 2 ] (71)

with v = u2f/(3f-2), and u = <R2>q2, where <Rg2>is the ensemble average

radius of gyration squared of an arm.

(22) Star-burst polymer with Gaussian statistics

Expressions have been given by Burchard and Kajiwara [55] and by
H a m m o u d a [56]. The results are obtained by s u m m a t i o n of geometrical
series. However, it cannot be r e c o m m e n d e d to use these as the perform-
ance of t h e sums introduces singularities. It is better to use the expres-
sions before the sums are performed [57].

(23) Regular comb polymer with Gaussian statistics

Expressions have been given by Casassa and Berry [58].

(24) Arbitrarily branched polymers with Gaussian statistics

The form factor for P subchains can be written as [57]:

P24(q) = 1 p~P17(q,Li) + 2 ~ . piPj~(q,Li)~(q,Lj) exp(--q2diy) (72)

i>j
w h e r e <Re) in P17(q,Li) is calculated for Li, the contour length of the ith
subchain. Pi is the total excess scattering length of the ith subchain.
Furthermore,

w(q,L i) = [1 - exp(-u)]/u (73)

w h e r e u = <R2>q2. The p a r a m e t e r d 2. = Lijb , where Lij is the separation

in contour length between the starting points of the ith and the j t h
P

subchains. The seattering mass M24 is given by M24 = ~ Pi •

i=1

(25) Sphere with Gaussian chains attached

The expressions have recently been derived by Pedersen and Ger-
stenberg [59]. For a sphere with radius R and total excess scattering
length Ps with N c attached chains of contour length L:
 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210 195

2 2
P25(q) = M225[psF1 (q,R) + Ncp2cP17(q,L) +

Nc(N c - 1)p2cSec(q) + 2NcPsPcSsc(q) l (74)

with:
sin(qR)
Ssc(q) = Fl(q~)~g(q,L) - -qR (75)

and

Scc(q ) = ~( q,L )2 [ si~RR) ]2 (76)

The scattering mass is: M25 = Ps + NcPc, where Pc is the total excess
scattering length of a single chain.
For non-penetrating chains and R >> Rg, the form factor is approxi-
mately given by Eq. (74) with R in Eqs. (75,76) replaced R + Rg in the
sin(x)/x terms. (Rg is the root-mean-square radius of gyration of a chain.)

(26) Ellipsoid with Gaussian chains attached

An ellipsoid of revolution is considered with radius R, eccentricity ~,
and total excess scattering length Pe with N c chains of contour length L
and total excess scattering length Pc. The form factor is [57]:
1
P26(q) = ~ " ~ [PeF~
2 2(q,R) + Ncp2P17(q,L) +

Nc(N c _ 1)p2 Scc(q)ell + 2NcPsPeSes~l(q)] (77)

with:
n/2
sin[qr(R,e,a)]
Sse/~q) = ~g(q,L) f Fl[q,r(R,e,a )] sina da (78)
qr(R,e,~x)
0

and
n/2~
f Isin[zr(R,e,a)ll
.-~2

SecZJ(q)=~(q'L'b)2 o L qr(R,e,a) J sina d¢z (79)

where r(R,~,a) = R(sin2a + e2 cos2a) 1/2 and M26 = De + NcPc is the total
scattering mass.
196 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

For n o n - p e n e t r a t i n g chains and R and Re >> Rg, the form factor is

approximately given by Eq. (77) with r(R,e,a) in Eqs. (78,79) replaced
by r(R,e,a) + Rg in the sin(x)/x terms.

(27) Cylinder with Gaussian chains attached

A cylinder is considered with radius R, length Lcyl, and total excess
scattering length Pcylwith Nc chains of contour length L and total excess
scattering length Pc. The form factor is [57]:

P27(q) = ~ 217 [PcylF11(q,R

2 2 ) +Ncp2P17(q,L) +

Nc(N ~ _ 1)p2 Sccyc [q) + 2NcPcPcylSsc~q)]

cy (80)

with:

SCyl(q) = ~(q,L) x
(81)

2B l(qRsinc0 sin [(qLcos(~) / 21]

~/2 2Bo(qRsin~) cos [(qL cosc0 / 2] sins d(~
0 qRsin(x (qL cos(~)/2)
and
~/2
cyl (q) = ~(q,L,b) 2 ~ {2Bo(qR sins) co~[(qL cos~)/2]} 2 sin(z d(z
Scc (82)
0

The scattering mass is: M27 = Pcyl + NcPc.

For n o n - p e n e t r a t i n g chains and R and L >> Rg, the form factor is
approximately given by Eq. (80) with R replaced by R + Rg and L
replaced by L + 2Rg in the Bo(x) cos(y) terms in Eqs. (81,82).

3.2. Structure factors

There are only very few cases for which the structure factor can be
calculated analytically. Most of the available results have been obtained
from liquid state theory for particles with spherical s y m m e t r y interact-
ing with a spherically symmetric potential. The liquid state theory
combines the O r n s t e i n - Z e r n i k e integral equation with an approximate
closure relation t h a t relates the interaction potential to the direct
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 197

correlation f u n c t i o n (see e.g. Ref. [60]). If t h e e q u a t i o n s c a n n o t be solved

analytically, it is possible to obtain n u m e r i c a l r e s u l t s for t h e s t r u c t u r e
factor. In this case a closure relation can be chosen w h i c h gives t h e r m o -
d y n a m i c a l l y self-consistent r e s u l t s [61,62].

(1) Hard-sphere potential

T h e particles i n t e r a c t w i t h t h e h a r d - s p h e r e r a d i u s RHS a n d h a v e a
h a r d - s p h e r e v o l u m e fraction )l. T h e expressions for S(q) h a v e been
c a l c u l a t e d w i t h t h e P e r c u s - Y e v i c k a p p r o x i m a t i o n for t h e closure rela-
tion (see e.g. [63]):

1
(83)
Sl(q) = 1 + 24HG(RHs q) /(RHs q)

In t h i s e q u a t i o n :

G(A) = a ( s i n A - A cos A ) / A 2

+ ~(2A sinA + (2 - A 2) cosA - 2 / A 3 (84)

co, + co, + sin>, +

and

(~ = (1 + 2TI)2 /(1 -Ti) 4

=-6q(1 + ~ / 2 ) 2 / ( 1 -TI) 2 (85)

$ = TIc¢ 12

(2) Sticky hard-sphere potential

T h e m o d e l w a s i n t r o d u c e d by B a x t e r [64]. T h e particles h a v e a
h a r d - s p h e r e r a d i u s RHS a n d a delta function a t t r a c t i v e p o t e n t i a l at t h e
surface. T h e s t i c k i n e s s of t h e particles is given by t h e p a r a m e t e r x. T h e
e x p r e s s i o n s for t h e s t r u c t u r e factor can be calculated w i t h t h e P e r c u s -
Yevick a p p r o x i m a t i o n for t h e closure relation (see for e x a m p l e , [65,66]).
F o r a h a r d - s p h e r e v o l u m e fraction ~ a n d ~ = 2qRHs

1
S2(q) - A2 + B2 (86)
 198 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

where

{ (sinK-~ccos¢) (1-cosK) K ~_~}

A=1+1211 a rc3 +[3 ~c2 -12

and

(~/2-~sin~:+l-cos~) + ~ (K - sin ~c) X (1 - cos re)l

B = 12T1 a K3 I~2 12
J
Furthermore,

X = Min{6['~/n + 1 / ( 1 - TI)] + {36[x/n + 1 / ( 1 -1])]} 1/2}

~t = ~.~(1 - 13)

(1 + 2D - p)
a - (87)
(1 - T1)2
_ (-3T1 + p-)
2(1 - T1)2

(3) Screened Coulomb potential

T h e p a r t i c l e i n t e r a c t i o n is d e s c r i b e d by t h e h a r d - s p h e r e r a d i u s RHS
a n d a n i n t e r a c t i o n p o t e n t i a l g i v e n by:

CBZ2 exp [-K(r - 2RHs )]

V(r) - (88)
(1 + KRHs) 2

for r > 2RHs. C B = e2/(4ne), w h e r e e is t h e p e r m i t t i v i t y of t h e solvent, e

t h e e l e m e n t a r y c h a r g e , Z t h e n u m b e r of c h a r g e s p e r particle, a n d • is
the inverse Debye-Htickel length.
T h e s t r u c t u r e factor h a s b e e n c a l c u l a t e d in t h e M e a n - S p h e r i c a l
A p p r o x i m a t i o n (MSA) a n d t h e e x p r e s s i o n s c a n be f o u n d in Ref. [67]. T h e
r e s u l t s w o r k well for h i g h to m e d i u m c o n c e n t r a t i o n s , b u t t h e p a i r
c o r r e l a t i o n f u n c t i o n b e c o m e s u n p h y s i c a l for low c o n c e n t r a t i o n s . A b e t t e r
r e s u l t is o b t a i n e d b y a r e s c a l i n g of t h e h a r d - s p h e r e r a d i u s as d e s c r i b e d
b y H a n s e n a n d H a y t e r [68]. T h i s p r o c e d u r e is k n o w n as t h e R e s c a l e d
M S A (RMSA).
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 199

It is also possible to obtain structure factors from thermodynamically

self-consistent approaches [60,61,69]. However, these have the problem
that the integral equations have to be solved numerically (and itera-
tively) and this makes the applications quite time consuming.
(4) Hard-sphere potential, polydisperse system
Vrij [70] has given a solution for polydisperse hard spheres for an
arbitrary size distribution. It is relatively straight forward to implement
it in a computer program, as it only requires the calculation of certain
averages of products of form factors and trigonometric functions weighted
by the size distribution. The expressions can be found in the paper by Vrij.
(5) Sticky hard-sphere potential, polydisperse system
A solution within the Percus-Yevick approximation has been given
by Robertus, Phillipse, Joosten and Levine [71]. This solution is for
arbitrary size distributions. In order to obtain the solution a set of
equations quadratic in the parameters have to be solved numerically.

(6) Screened Coulomb potential, polydisperse system

The solution in the MSA has been given by Blum and Hoye [72] and
Blum [73]. Ruiz-Estrada, Medina-Noyola and N~ingele [74] have given
a procedure for making a computer implementation using the rescaled
mean-spherical approximation.
The implementation of a thermodynamically self-consistent (numeri-
cal) approach based on the closure relation of Rogers and Young [61] has
been described by D'Aguanno and Klein [60,75].

(7) Cylinders
The scattering can be calculated in the random phase approximation
(RPA) [76]. The structure factor for cylinders of length L and radius R,
with L >> R is [77]:
(d(~(q)~ = Pll(q)
(89)
[~)7 npc~yl(1 -FVPil(q))
where v is proportional to the concentration of the cylinders, Pcyl is the
total scattering length of a cylinder, and n is the number density.

(8) Solutions of polymers

The scattering can be calculated in the random phase approximation
(RPA). The structure factor is [76,78]:
200 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210

(~/ = P17(q)
; np2 (1 + ve17(q)) (90)

where v is proportional to the polymer mass concentration, to the

contour length and to the strength of the excluded volume interaction.
Pc is the total scattering lengtlh of a chain. The RPA is only a good
approximation for high polymer concentrations. At lower concentrations,
it can be recommended to use Eq. (90) with P17(q) replaced by Pls(q), and
with the parameters normalized, so that Eq. (90) gives the forward
scattering (intensity at q = 0) and the correlation length predicted by
renormalization group theory [79].

4. An application: block c o p o l y m e r micelles

In this section an example of an application is given. It involves both

indirect Fourier transformation, square-root deconvolution, modeling,
and least-squares optimization of a model. Both neutron and X-ray
small-angle scattering (SANS/SAXS) data are discussed.
The example concerns a block copolymer in a solvent. It is a tri-block
and consists of two end blocks of poly(ethylene oxide) (PEO) and a
c e nt ra l block of poly(propylene oxide) (PPO). The composition is
EO25PO40EO25 and the molecular weight is 4500 Da. The molecular
weights are 2300 Da for the PPO part and 1100 Da for each of the PEO
parts. The polymer is commercially available from Serva AG, Heidel-
berg, Germany, unde r the name P85.
The aggregation behavior of P85 in water has been studied by SANS
[59,80] and SAXS [81]. The P85 is present as single chains at low tempera-
tures (T < 5°C) and forms aggregates/micelles at higher temperatures. This
behavior is ascribed to the properties of PPO, which is hydrophilic at room
t e m p e r a t u r e s and becomes hydrophobic at higher temperatures. The
PEO is reasonably soluble for temperatures up to T = 70°C. The present
discussion is limited to the dilute region with well-defined micelles,
where the structure factor effects are small.
Measurements on a 0.5% wt. solution at T = 50°C were performed at
the SANS i n s t r u m e n t at Ris0 National Laboratory [82]. In order to
enhance the contrast and reduce the incoherent background, D20 was
used as the solvent. SAXS data have been recorded for a 1% wt. solution
T = 50°C by a conventional Kratky camera (see, e.g. Ref. [82]) by Glatter
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 201

101 , ' ' I ''''1 ' ' ' I ''''1 ' ' ' I '''

10 0

10 -1
(:7-
v

10 .2

10 .3

10 -4
10 -3 10 .2 10 -1 10 °
q (~-1)

Fig. 1. The small-angle neutron scattering data of 0.5% P85 in D20 (upper data) [59].
The curve is the fit by the analytical model including polydispersity of the micelles. Note
that the fitted curve is discontinuous where the data from different settings overlap due
to the instrumental smearing. The lower data are SAXSresults on similar micelles [81].
The curve is calculated for the results determined by fitting the SANS data.

et al. [81]. The SAXS d a t a w e r e corrected for i n s t r u m e n t a l s m e a r i n g

effects in connection w i t h an indirect F o u r i e r t r a n s f o r m a t i o n as de-
scribed in Section 2. The d e s m e a r e d SAXS s c a t t e r i n g curve is shown in
Fig. 1 t o g e t h e r w i t h the SANS data. Note t h a t t h e SANS d a t a h a v e been
r e c o r d e d using t h r e e different i n s t r u m e n t a l settings. Due to t h e differ-
e n t s m e a r i n g effects for t h e different settings, the d a t a do not coincide
in t h e overlap regions.
The SANS a n d SAXS s c a t t e r i n g curves are r e m a r k a b l y different. The
SANS d a t a h a v e a smooth decrease w i t h increasing q a n d r e s e m b l e t h e
s c a t t e r i n g form factor of a sphere. However, no oscillations a r e observed
(c.f. Eq. (55)). A q-2 s c a t t e r i n g is observed at large q, w h i c h is similar to
t h e s c a t t e r i n g from flexible chains obeying G a u s s i a n statistics (c.f. Eq.
(70)). In c o n t r a s t to the t h e SAXS d a t a h a v e large oscillations. The r a n g e
of s c a t t e r i n g vectors is not large e n o u g h to identify power law s c a t t e r i n g
a t large q.
A n indirect F o u r i e r t r a n s f o r m a t i o n of the d a t a sets h a v e been per-
formed [57,81], a n d the r e s u l t i n g distance distribution functions are
202 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

3.5 i i i , i i

3.0 2
2.5
1
2.0
v

"-K1. 5 ...........................
~......~......-.!.=.=....
0
1.0 I
I
0.5 -1
0.0 t t I /
0 50 1O0 150 0 20 40 60 80 100
r (~) r (,~)

Fig. 2. (a) Distance d i s t r i b u t i o n functions for t h e d a t a shown in Fig. 1. The full curve is
for t h e S A N S d a t a [57] and the b r o k e n curve is for the SAXS d a t a [81]. (b) Excess
s c a t t e r i n g l e n g t h d e n s i t y d i s t r i b u t i o n s for t h e d a t a shown in (a). The n o t a t i o n is t h e
s a m e as in (a).

s h o w n in Fig. 2(a). Corrections for i n s t r u m e n t a l resolution effects w e r e

included as described in Section 2.1. For the SANS d a t a the file scale
factors w e r e also adjusted. The SANSp(r) function has a bell shape w i t h
m a x i m u m at 55 .~ and a tail at larger r, w h i c h extends up to about 140
.&. The s h a p e is characteristic for homogeneous almost spherical parti-
cles. In contrast to the n e u t r o n p(r) function, the SAXS p(r) function h a s
m a x i m u m a r o u n d 90 A and a m u c h more a s y m m e t r i c t r i a n g u l a r shape,
w h i c h is characteristic for spherical shell shapes. Thus, the n e u t r o n s
'see' a h o m o g e n e o u s particle, w h e r e a s the X-rays 'see' a shell. This can
be explained by the quite different interaction of X-rays a n d n e u t r o n s
w i t h m a t t e r . The n e u t r o n s are i n t e r a c t i n g w i t h the nuclei, w h e r e a s
X-rays are i n t e r a c t i n g w i t h the electrons. For n e u t r o n s the scattering
l e n g t h of h y d r o g e n a n d d e u t e r i u m is quite different and the contrast is
provided by the difference in d e u t e r i u m density. It t u r n s out t h a t the
polymer h a s a scattering length density close to zero w h e r e a s D20 h a s
a large scattering length density: the scattering is basically from the
'holes' m a d e by the polymer in the D20. For X-rays the contrast is due
to the difference in electron density. Organic polymers h a v e electron
densities w h i c h are quite close to the electron density of water. There-
fore, the contrasts are very sensitive to the actual value of the specific
v o l u m e of the polymer. It seems reasonable from the p(r) functions alone
to a s s u m e t h a t the specific volumes are so t h a t m a i n l y a shell of
dissolved P E O chains is probed. Accurate density m e a s u r e m e n t s [83] as
a function of t e m p e r a t u r e for a series of P E O - P P O - P E O block copoly-
m e r s can be used for calculating the specific volumes of PEO and PPO
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 203

parts. Such calculations [59] show t h a t the contrast of the PPO is small
and negative, w h e r e a s the contrast of the PEO is l a r g e r and positive.
This confirms t h a t it is m a i n l y the P E O which is observed in SAXS. The
presence of dissolved P E O chains are also supported by the observed q-2
b e h a v i o r at large q in the SANS data. If a model should be constructed
from the available information, it should consist of a homogeneous and
spherical P P O core a n d a shell of dissolved P E O chains.
The p(r) functions indicate t h a t the particles h a v e an approximate
spherical s y m m e t r y and the oscillations in the SAXS scattering curve
show t h a t the size polydispersity is small. It is therefore r e a s o n a b l e to
apply t h e square-root deconvolution procedure to t h e p ( r ) functions [4,5].
In this procedure the excess scattering l e n g t h density distribution Ap(r)
is p a r a m e t e r i z e d by a set of step functions. The p(r) is calculated from
Ap(r) a n d least-squares fitted to the p(r) function from the indirect
F o u r i e r transformation. A smoothness constraint is applied and the
coefficients in the p a r a m e t e r i z a t i o n are optimized by non-linear least-
s q u a r e s methods. The r e s u l t i n g Ap(r) functions are shown in Fig. 2(b).
The function for the SANS d a t a shows the expected high density core
a n d a lower density shell. The function for the SAXS d a t a also shows
t h e expected behavior w i t h a small negative value in the core a n d a
l a r g e r positive value in the shell region. This is in perfect a g r e e m e n t
w i t h the calculated contrasts. Note t h a t the size of the core agrees quite
well for the two Ap(r) functions, but t h a t the functions do not go to zero
at the s a m e r value. The d i s a g r e e m e n t at large r is probably explained
by the fact t h a t the n e u t r o n s are not very sensitive to the low concen-
t r a t i o n p a r t of the shell w i t h the P E O chains.
The information from the m o d e l - i n d e p e n d e n t approaches suggests
t h a t the form factor given by Eq. (74), for a particle w i t h spherical core
a n d G a u s s i a n chains a t t a c h e d to the surface, is appropriate to use for
fitting t h e data. The n u m b e r of fitting p a r a m e t e r s in the model is quite
large a n d in order to reduce it, some of t h e p a r a m e t e r s w e r e e s t i m a t e d
from o t h e r m e a s u r e m e n t s . The excess scattering length densities w e r e
calculated from the composition of the P85 triblock copolymer a n d t h e
specific volumes, w h i c h can be derived from density m e a s u r e m e n t s [83].
The contour l e n g t h of the P E O chains w e r e calculated to be 90 A using
t h e p a r a m e t e r s given by Flory [85] and Aharoni [84].
The inter-particle interference effects are quite small for the low
concentrations (0.5-1% wt.), but as the chains extend into the solvent,
t h e y are not completely negligible. Therefore, the inter-micellar effects
w e r e included in the analysis of the d a t a using a h a r d - s p h e r e s t r u c t u r e
204 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

factor (Eq. (83)). As t h e conditions for t h e SANS a n d SAXS m e a s u r e -

m e n t s are n o t identical, it w a s not a t t e m p t e d to fit t h e two d a t a sets
s i m u l t a n e o u s l y . I n s t e a d a s t r a t e g y w a s chosen, in w h i c h only t h e SANS
d a t a w e r e fitted a n d t h e r e s u l t s w e r e u s e d for c a l c u l a t i n g a SAXS
s c a t t e r i n g curve w h i c h w a s c o m p a r e d to t h e recorded data.
T h e t h e o r e t i c a l s c a t t e r i n g curve w a s s m e a r e d by t h e i n s t r u m e n t a l
r e s o l u t i o n f u n c t i o n [23], w h e n t h e fit to t h e SANS d a t a w a s p e r f o r m e d .
F o r each s e t t i n g t h e a p p r o p r i a t e r e s o l u t i o n function w a s used. T h e
f i t t i n g w a s carried out as described in Section 2.2 by c o m b i n i n g grid
s e a r c h e s w i t h m o r e a d v a n c e d procedures.
It t u r n s o u t t h a t t h e oscillations in t h e model form factor a r e too
p r o n o u n c e d c o m p a r e d to t h e e x p e r i m e n t a l s c a t t e r i n g curve. Therefore,
size p o l y d i s p e r s i t y of t h e micelles w a s i n c l u d e d u s i n g a Schulz distribu-
tion for t h e a g g r e g a t i o n n u m b e r . Note t h a t s i m i l a r d i s t r i b u t i o n s are
p r e d i c t e d by t h e r m o d y n a m i c e q u i l i b r i u m t h e o r i e s for s u r f a c t a n t
micelles [86,87]. In a first a t t e m p t t h e core w a s a s s u m e d to be consti-
t u t e d solely by t h e P P O p a r t s of t h e chains. T h e d i s t a n c e from t h e
surface of t h e core to t h e s t a r t i n g points of t h e G a u s s i a n c h a i n s w a s a
fitting parameter.
T h e p o l y d i s p e r s e model fits t h e s c a t t e r i n g d a t a r e a s o n a b l y well. T h e
m a i n d e v i a t i o n s are in t h e region w h e r e t h e r e is a crossover in t h e
s c a t t e r i n g d a t a to t h e q-2 behavior. In order to achieve perfect a g r e e m e n t
in this region it w a s n e c e s s a r y to include a low d e n s i t y P E O shell a r o u n d
t h e core in accordance w i t h t h e s u g g e s t i o n by M o r t e n s e n a n d P e d e r s e n
[80]. It s h o u l d be n o t e d t h a t t h e excluded v o l u m e i n t e r a c t i o n b e t w e e n
t h e c h a i n s is n o t i n c l u d e d in t h e model. As t h e c o n c e n t r a t i o n of c h a i n s
is q u i t e large at t h e surface, it is likely t h a t t h e region closest to t h e P P O
core is m o r e h o m o g e n e o u s t h a n t h e region f u r t h e r away.
F o r a m o d e l w i t h a P P O core w i t h a low d e n s i t y shell of P E O a n d
dissolved P E O chains, t h e a v e r a g e a g g r e g a t i o n n u m b e r w a s Nagg -- 74
w h i c h c o r r e s p o n d s to a r a d i u s of t h e P P O core o f R = 40 .~. T h e a d d i t i o n a l
shell c o n t a i n s 23% of t h e P E O c h a i n s w i t h a w a t e r c o n t e n t of 77%. T h e
o u t e r r a d i u s of this shell is 49 A. T h e K u h n l e n g t h is d e t e r m i n e d to be
b = 10 A, w h i c h agrees well w i t h previous e s t i m a t e s [84]. T h e s t a r t i n g
p o i n t for t h e P E O c h a i n s is 0.25 × Rg a w a y from t h e surface of t h e P E O
core shell. T h e p o l y d i s p e r s i t y of t h e a g g r e g a t i o n n u m b e r is ~(Nagg)/Nagg
= 0.37, w h e r e g(Nagg) is t h e s t a n d a r d d e v i a t i o n of t h e Schultz distribu-
tion. T h e c o r r e s p o n d i n g polydispersity of t h e r a d i u s of t h e core is ~(R)/R
= 0.13. T h i s relatively large polydispersity could explain w h y t h e shear-
i n d u c e d single crystal BCC p h a s e observed at h i g h e r c o n c e n t r a t i o n
 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210 205

possesses only angular order and not positional order [88].

The corresponding scattering intensity of the resulting model for
X-rays was calculated and is shown in Fig. 1. The model intensity curve
reproduces the pronounced oscillation close to q = 0.1 ~-1 in the meas-
ured data and a reasonable qualitative agreement is obtained.
The analysis described in the present section demonstrates the
various steps in a typical analysis. The model-independent methods
were used for obtaining information on the shape of the micelles and
identifying the different components of the structure. A model was
constructed from the available information and fitted to the data using
least-squares methods. Detailed structural information was obtained on
the micellar structure from the analysis.

5. S u m m a r y and c o n c l u s i o n s

Analysis of small-angle scattering data for colloidal systems consist-

ing of particles or polymers in a solvent has been discussed. Only
systems with short range correlations have been considered. Systems
with long-range order have sharp Bragg reflections, which can be
resolution limited, and the analysis of the data can be quite different
from the analysis described in the previous section. In the cases where
the system possesses order similar to that of a single crystal, the
orientation of the sample also has to be considered. A complete descrip-
tion of the three-dimensional scattering geometry is required and this
is quite complicated [89].
A typical data analysis involves both a step with model-independent
analysis and a step in which a model on analytical form is fitted to the
data. The model-independent step can be an indirect Fourier transfor-
mation t h a t involves linear least-squares fitting. In some cases it can be
supplemented by a square-root deconvolution which gives the scattering
length density profile. The square-root deconvolution requires the ap-
plication of non-linear methods. A model may be constructed based on
the information obtained from the model-independent approaches, but
any additional available information on the systems should also be
considered. An example of an application of the model-independent
approaches and of model fitting was given in the previous section.
The indirect Fourier transformation [1] described in Section 2.1
concerns the three-dimensional distance distribution function, and it is
best suited for particles with shapes that are not too anisotropic.
206 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

However, a similar approach can be applied for obtaining cross-section

distance distribution functions for particles with large anisotropies
[2,90]. For infinitely long rods and infinitely large two-dimensional
planar structures the scattering from the overall dimension and the
cross-section separate. This means the scattering can be written as the
product of a power law and a Fourier transform of the cross-section
distance distribution function. For rods the power law is q-1 and the
basis function in the Fourier transform are zeroth order Bessel func-
tions. For planar structures, the power law is q-2 and the basis functions
are cosines. If the cross-section structure has centro symmetry, it is
possible to apply the square-root deconvolution procedure and obtain
the cross-section excess scattering length density distribution [4,5]. An
example with an application to planar structures has been given by
Maurer, Glatter and Hofer [91]. In this study the cross-section structure
of large lipid vesicles was determined by SAXS. Applications to rod-like
particles have been described by Schurtenberger, Jerke, Cavaco, and
Pedersen [92]. The cross-section structure of inverse lecithin micelles
with a small core of H20 and D20 in deuterated cyclohexane was
determined by SANS. The water and lecithin distributions were deter-
mined from the difference between the scattering length distribution of
the particles with H20 and D20.
The fact that the scattering from the cross-section and the scattering
from the large dimension(s) separate for particles with large anisot-
ropies can also be used for extending the available form factors. For
planar structure one has, if the cross-section dimension is small com-
pared to the overall size:

P(q) = P(q)" Pcs(q) (91)

where P(q)' is the form factor of an infinitely thin shell describing the
overall shape of the object and:

Pes(q) = ~ APcs(r) cos(qr) dr 2 (92)

where APcs(r) is the centro symmetric cross-section scattering length

density distribution function. P(q)" can be the form factor of an infinitely
thin spherical shell, elliptical shell, cylinder shell, or disk.
For particles with a local rod-like structure, the scattering can also
be written using an expression like Eq. (91), if the overall dimension is
large compared to the cross-section dimension. In this case
 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210 207

Pcs(q) = 2~ ~ APcs(r ) 2Bo(qr)r dr 2 (93)

w h e r e APcs(r) is a g a i n t h e centro s y m m e t r i c cross-section s c a t t e r i n g

l e n g t h d e n s i t y d i s t r i b u t i o n function. P(q)" can be t h e f o r m factor of an
i n f i n i t e l y t h i n rod (Eq. (66)) or a semi-flexible p o l y m e r c h a i n w i t h or
w i t h o u t excluded v o l u m e i n t e r a c t i o n s [51,93].
T h e c o n s t r u c t i o n of a model cross section m a y be a t e d i o u s job. F i r s t
t h e m o s t i m p o r t a n t f e a t u r e s of t h e s t r u c t u r e h a v e to be identified, so
t h a t a m o d e l s t r u c t u r e can be formed. B u t occasionally, a n a l y t i c a l
e x p r e s s i o n s a r e n o t available for t h e s c a t t e r i n g cross section of t h e m o d e l
s t r u c t u r e a n d it is n e c e s s a r y to c o n s t r u c t t h e m or calculate t h e m . This
w a s t h e case for t h e P85 micelles described in t h e previous section. W h e n
a n a l y t i c a l e x p r e s s i o n s are derived, it is often c o n v e n i e n t to u s e M o n t e
Carlo s i m u l a t i o n s [94] for c h e c k i n g t h e analytical e x p r e s s i o n s as it w a s
done for t h e P85 micelles [59]. In some cases it m a y be impossible to
o b t a i n a n a l y t i c a l e x p r e s s i o n s as it is for semi-flexible polymers. If
s c a t t e r i n g f u n c t i o n s are available, for e x a m p l e , f r o m c o m p u t e r s i m u l a -
tions for a large r a n g e of p a r a m e t e r s , t h e s c a t t e r i n g f u n c t i o n s can be
p a r a m e t e r i z e d a n d i n t e r p o l a t e d , so t h a t t h e y can be u s e d for least-
s q u a r e s fitting [51,53,93].
W h e n fitting e x p e r i m e n t a l data, it is n o t u n u s u a l t h a t t h e informa-
tion c o n t e n t of a d a t a set is quite low. T h e p a r a m e t e r s of t h e model are
t h e n v e r y c o r r e l a t e d a n d w h e n physical p a r a m e t e r s like t h e m a s s
d e n s i t y a r e c a l c u l a t e d from t h e fit p a r a m e t e r s it m a y give u n p h y s i c a l
values. S u c h p r o b l e m s can be g r e a t l y r e d u c e d if t h e d a t a are on a n
a b s o l u t e scale a n d t h e m o l e c u l a r p r o p e r t i e s of t h e c o n s t i t u t i n g mole-
cules are t a k e n into a c c o u n t in t h e model as c o n s t r a i n t s . If t h e d e n s i t i e s
a r e k n o w n from o t h e r t y p e s of m e a s u r e m e n t s this i n f o r m a t i o n can be
i n c l u d e d a n d u n p h y s i c a l solutions can be avoided. T h e i n f o r m a t i o n
c o n t e n t of t h e d a t a can also be e n h a n c e d by m e a s u r i n g for several
d i f f e r e n t c o n t r a s t s , or by m e a s u r i n g t h e s a m p l e s by b o t h X-ray a n d
n e u t r o n scattering. F o r n e u t r o n s t h e c o n t r a s t v a r i a t i o n can be done by
v a r y i n g t h e d e u t e r a t e d fraction of t h e solvent. If t h e particles h a v e
c o m p o n e n t s w i t h different composition, t h e c o n t r a s t v a r i a t i o n will pro-
vide a d d i t i o n a l i n f o r m a t i o n . T h e c o m b i n a t i o n of X-ray a n d n e u t r o n
s c a t t e r i n g is in s o m e cases v e r y efficient ([95], see also previous section).
T h i s is t h e case for s u r f a c t a n t s y s t e m s w i t h h e a v y a t o m s in t h e h e a d
group, for w h i c h t h e X-ray s c a t t e r i n g is a l m o s t e n t i r e l y from t h e h e a d
g r o u p region, w h e r e as t h e micelles a p p e a r to be h o m o g e n e o u s for
208 J.S. Pedersen/Adv. Colloid Interface Sci. 70 (1997) 171-210

n e u t r o n s c a t t e r i n g in D 20. T h e e x t r a i n f o r m a t i o n f r o m c o n t r a s t v a r i -
a t i o n is m o s t e f f i c i e n t l y t a k e n i n t o a c c o u n t i f t h e d a t a f r o m d i f f e r e n t
c o n t r a s t s a r e f i t t e d s i m u l t a n e o u s l y . In t h i s c a s e t h e s a m e p a r a m e t e r s
f o r t h e g e o m e t r y c a n be u s e d a n d o n l y t h e s c a t t e r i n g l e n g t h d e n s i t i e s
d e p e n d o n w h i c h c o n t r a s t is c o n s i d e r e d .

Acknowledgments

I t h a n k G~tz J e r k e , L i s e A r l e t h , Kell M o r t e n s e n , M i c h a e l G e r s t e n -
b e r g a n d P e t e r S c h u r t e n b e r g e r for t h e i r c o m m e n t s on t h e m a n u s c r i p t .

References

[1] O. Glatter, J. Appl. Cryst., 10 (1977) 415-421.

[2] O. Glatter, J. Appl. Cryst., 13 (1980) 577-584.
[3] O. Glatter, J. Appl. Cryst., 12 (1979) 166-175.
[4] O. Glatter, J. Appl. Cryst., 14 (1981) 101-108.
[5] O. Glatter and B. Hainish, J. Appl. Cryst., 17 (1984) 535-441.
[6] O. Glatter, J. Appl. Cryst., 13 (1980) 7-11.
[7] J.S. Pedersen, J. Appl. Cryst., 27 (1994) 595-608.
[8] H.B. Stuhrmann, Z. Phys. Chem. (Frankfurt), 72 (1970) 177-184, 185-198.
[9] D.I. Svergun, Acta Cryst. A 50 (1994) 391-402.
[10] D.I. Svergun, M.H.J. Koch and I.N. Serdyuk, J. Mol. Biol., 240 (1994) 66-77.
[11] D.I. Svergun, M.H.J. Koch, J.S. Pedersen and I.N. Serdyuk, J. Mol. Biol., 240
(1994) 78-96.
[12] D.I. Svergun, M.H.J Koch, J.S. Pedersen and I.N. Serdyuk, Proc. Natl. Acad. Sci.
USA, 91 (1994) 11826-11830.
[13] Data Reduction and Error Analysis for the Physical Sciences. McGraw-Hill, New
York, 1969.
[14] W.H. Press, B.P. Flannery, S.A. Teukolsky and W.T. Vetterling, Numerical Reci-
pes. University Press, Cambridge, 1989.
[15] P.B. Moore, J. Appl. Cryst., 13 (]980), 168-175.
[16] D.I. Svergun, A.V. Semenyuk and L.A. Feigin, Acta Cryst., A44 (1988) 244-250.
{17] D.I. Svergun, J. Appl. Cryst., 26 (1993) 258-267.
[18] S. Hansen and J.S. Pedersen, J. Appl. Cryst., 24 (1991) 541-548.
[19] D.I. Svergun and J.S. Pedersen, J. Appl. Cryst., 27 (1994) 241-248.
[20] J.S. Pedersen, S. Hansen and R. Bauer, Eur. Biophys. J., 23 (1994) 379-389.
[21] D.I. Svergun, J. Appl. Cryst., 25 (1992) 495-503.
[22] C.L. Lawson and R.J. Hanson, Solving Least Squares Problems. Prentice-Hall,
New Jersey, 1974.
[23] J.S. Pedersen, D. Posselt, and K. Mortensen, J. Appl. Cryst. 23 (1990) 321-333.
 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210 209

[24] D.I. Svergun, J. Appl. Cryst., 24 (1991) 485-492.

[25] D.I. Svergun, Private communication, 1993.
[26] J.S. Pedersen, Implemented in the program [Link], based on the method
in Ref. [1].
[27] D.W. Marquardt, J. Soc. Ind. Appl. Math., 11 (1963) 431-441.
[28] B. SjSberg, J. Appl. Cryst., 11 (1978) 73-79.
[29] J.S. Pedersen, J. Phys. IV France, Coll. C8 3 (1993) 491-498.
[30] J.G. Barker and J.S. Pedersen, J. Appl. Cryst., 28 (1995). 105-114.
[31] J.S. Pedersen, Implemented in the program [Link].
[32] M. Kotlarchyk and S.H. Chen, J. Chem. Phys., 79 (1983) 2461-2469.
[33] Lord Rayleigh, Proc. Roy. Soc. London, Ser. A, 84 (1911) 25-38.
[34] P. Debye, Ann. Phys. Leipzig 46 (1915) 809-823.
[35] A. Guinier, Ann. Phys., 12 (1939) 161-237.
[36] P. Mittelbach and G. Porod, Acta Physica Austriaca 15 (1962) 122-147.
[37] P. Mittelbach and G. Porod, Acta Physica Austriaca 14 (1961) 185-211.
[38] R.W. Hendricks, J. Schelten and W. Schmatz, Phil. Mag., 30 (1974) 819-837.
[39] P.A. Dubey, [Link]. Dissertation No. 9077. ETH Zfirich, 1990.
[40] P.A. Dubey, B. SchSnfeld and G. Kostorz, Acta Metall. Mater., 39 (1991) 1161-
1170.
[41] F. Rodriguez, J.C. GSmez Sal, M. Moreno, A. de Geyer and C. Janot, Phys. Rev. B
43 (1991) 7519-7526.
[42] G. Fournet, Bull. Soc. Fr. Mineral. Crist., 74 (1951) 39-113.
[43] P. Mittelbach and G. Porod, Acta Physica Austriaca 14 (1961b) 405-439.
[44] S. Cusack, J. Mol. Biol., 145 (1981) 541-543.
[45] T. Neugebauer, Ann. Phys. (Leipzig), 42 (1943) 509-533.
[46] O. Kratky and G. Porod, J. Colloid Sci., 4 (1949) 35-70.
[47] J. Teixeira, J. Appl. Cryst., 21 (1988) 781-785.
[48] H.D. Bale and P.W. Schmidt, Phys. Rev. Lett., 53 (1984) 596-599.
[49] P. Debye, J. Phys. Colloid Chem., 51 (1947) 18-32.
[50] H. Utiyama, Y. Tsunashima and M. Kurate, J. Chem. Phys., 55 (1971) 3133-3145.
[51] J.S. Pedersen and P. Schurtenberger, Macromolecules (1996) in press.
[52] O. Kratky and G. Porod, Rec. Trav. Chim. Pays-Bas, 68 (1949) 1106-1122.
[53] T. Yoshizaki and H. Yamakawa, Macromolecules, 13 (1980) 1518-1525.
[54] H. Benoit, J. Polym. Sci., 11 (1953) 507-510.
[55] W. Burchard, K. Kajiwara, D. Neger and W.H. Stockmayer, Macromolecules, 17
(1984) 222-230.
[56] J. Polym. Sci. B: Polym. Phys., 30 (1992) 1387-1390.
[57] J.S. Pedersen and M.C. Gerstenberg, unpublished, 1995.
[58] E.F. Casassa and G.C. Berry, J. Polym. Sci. Part A2, 4 (1966) 881-897.
[59] J.S. Pedersen and M.C. Gerstenberg, Macromolecules, 29 (1996) 1363-1365.
[60] B. D'Aguanno and R. Klein, Chem. Soc. Faraday Trans., 87 (1991) 379-390.
[61] F.J. Rogers and D.A. Young, Phys. Rev., A30 (1984) 999-1007.
[62] G. Zerah and J.P. Hansen, J. Chem. Phys., 84 (1986) 2336-2343.
[63] D.J. Kinning and E.L. Thomas, Macromolecules, 17 (1984) 1712-1718.
[64] R.J. Baxter, J. Chem. Phys., 49 (1968) 2770-2774.
[65] C. Regnaut and J.C. Ravey, J. Chem. Phys., 91 (1989) 1211-1221.
210 J.S. Pedersen /Adv. Colloid Interface Sci. 70 (1997) 171-210

[66] S.V.G. Menon, C. Manohar and K. Srinivasa Rao, J. Chem. Phys., 95 (1991)
9186-9190.
[67] J.B. Hayter and J. Penfold, Mol. Phys., 42 (1981) 109-118.
[68] J.-P. Hansen and J.B. Hayter, Mol. Phys., 46 (1982) 651-656.
[69] F.J. Rogers, J. Chem. Phys., 73 (1980) 6172-6178.
[70] A. Vrij, J. Chem. Phys., 71 (1979) 3267-3270.
[71] C. Robertus, W.H. Phillipse, J.G.H. Joosten and Y.K. Levine, J. Chem. Phys., 90
(1989) 4482-4490.
[72] L. Blum and J.S. Hoye, J. Stat. Phys., 19 (1978) 317-324.
[73] L. Blum, J. Stat. Phys., 22 (1980) 661-672.
[74] H. Ruiz-Estrada, M. Medina-Noyola and G. N/ingele, Physica A, 168 (1990)
919-941.
[75] B. D'Aguanno and R. Klein, Phys. Rev., A 46 (1992) 7652-7656.
[76] S.F. Edwards, Proc. Soc. London, 88 (1966) 265-280.
[77] T. Shimada, M. Doi and K. Okano, J. Chem. Phys., 88 (1988) 2815-2821.
[78] H. Benoit and M. Benmouna, Polymer, 25 (1984) 1059-1067.
[79] A. Takanishi and T. Ohta, J. Phys. A: Math. Gen., 18 (1985) 127-139.
[80] K. Mortensen and J.S. Pedersen, Macromolecules, 26 (1993) 805-812.
[81] O. Glatter, G. Scherf, K. Schill6n, and W. Brown, Macromolecules, 27 (1995)
6046-6054.
[82] J.S. Pedersen. In: H. Brumberger (Ed.), Modern Aspects of Small-Angle Scatter-
ing. Kluwer, 1995, pp. 57-91.
[83] J.K. Armstrong, J. Parsonage, B. Chowdhry, S. Leharne, K. L6hner and P. Laggner
J. Phys. Chem., 97 (1993) 3904-3909.
[84] S.M. Aharoni, Macromolecules, 16 (1983) 1722-1728.
[85] P.J. Flory. Statistical Mechanics of Chain Molecules. Wiley, New York, 1969.
[86] P.J. Missel, N.A. Mazer, G.B. Benedek, C.Y. Young and M.C. Carey, J. Phys.
Chem., 84 (1980) 1044-1057.
[87] M.M. Stecker and G.B. Benedek, J. Phys. Chem., 88 (1984) 6519-6544.
[88] K. Mortensen, W. Brown and B. Nord~n, Phys. Rev. Lett., 68 (1992) 2340-2343.
[89] P. Harris, B. Lebech and J.S. Pedersen, J. Appl. Cryst., 28 (1995) 209-222.
[90] J.S. Pedersen and P. Schurtenberger, J. Appl. Cryst, (1996) In press.
[91] N. Maurer, O. Glatter and M. Hofer, J. Appl. Cryst., 24 (1991) 832-835.
[92] P. Schurtenberger, G. Jerke, C. Cavaco and J.S. Pedersen, Langmuir 12 (1996)
2433-2440.
[93] G. Jerke, P. Schurtenberger and J.S. Pedersen, in preparation.
[94] S. Hansen, J. Appl. Cryst. 23 (1990) 344-346.
[95] L. Arleth, D. Posselt, D. Gazeau, C. Lampent, T. Zemb, K. Mortensen and J.S.
Pedersen, Langmuir (1996) submitted.