Introduction to

Small-Angle X-ray Scattering

Thomas M. Weiss
Stanford University, SSRL/SLAC,
BioSAXS beamline BL 4-2 BioSAXS Workshop, March 28-30, 2016
Sizes and Techniques
Diffraction and Scattering
 Scattering of X-rays from a single electron
Ie

Thomson formula for the scattered intensity

from a single electron
2
I0
1  cos( 2  ) 1
I e  r0
2
2
I0
2 r
2
e  15
r0  2
 2 . 817 10 m
I 0 : Intensity of incoming X - rays mc
2  : angle of observatio n

I e : Intensity of scattered X - rays

Classical electron radius

The Thomson formula plays a central role for all scattering calculations involving
absolute intensities. Typically calculated intensities of a given sample will be expressed
in terms of the scattering of an isolated electron substituted for the sample.
In small angle scattering the slight angle dependence (the so-called polarization factor)
in the Thomson formula can be neglected.
Interference of waves

• waves have and amplitude and phase

• interference leads to fringe pattern (e.g.
water waves)
• the fringe pattern contains the information
on the position of the sources (i.e. structure)
• in X-ray diffraction the intensities (not the
amplitudes) of the fringes are measured

“phase problem”

constructive destructive
 Scattering from two (and more) electrons
scattering vector q  k  k 0
two electrons
2 4  sin 
with k  s and q  q  2

  F (q )   f e exp( i q  r i )  f e (1  exp( i q  r ))
i 1
Note:
… generalized to N electrons F(q) is the Fourier
N Transform of the
F (q )   f e exp( i q  r i ) spatial distribution
i 1 of the electrons

… averaged over all orientations

N
sin( qr )
F (q )   fe
qr
i 1

using the continuous (radial) distribution  ( r ) of the electron cloud in an atom


sin( qr )
dr  ( r ) r f (0 )  Z
F (q )  f (q )   with
2

0
qr

atomic scattering factor

 Scattering from Molecules
The scattering amplitude or form factor, F(q), of an isolated molecule with N atoms can
be determined in an analogous manner:
N
i.e. the Fourier Transform
F (q )   f i ( q ) exp( i q  r i )
of the atomic distribution
i 1

The scattered intensity from the isolated molecule is then

N N


2
I (q )  F (q )  f i ( q ) f j ( q ) exp( i q  ( r i  r j ))
i 1 j 1

In solution:
N N
average over all orientations sin( qr ij )
sin( qr ij )
I (q )   fi (q ) f j (q )
qr ij due to solution average only
exp( i q  ( r i  r j ))  i 1 j 1
qr ij interatomic distances are
Debye formula measured, not atomic
coordinates
Scattering from Molecules

N N
sin( qr ij )
I (q )   fi (q ) f j (q )
qr ij
i 1 j 1

Each atomic distance rij in the molecule adds a

sinx/x like term to the scattering intensity

• small distance
low frequency in sinx/x
dominate signal at high q

• large distance
high frequencies in sinx/x
dominate the signal at low q
 Scattering Intensity


I ( q )  F (q ) F (q )  FT [  ( r )] FT [  (  r )]  FT [  ( r )   (  r )]

 (r )
Autocorrelation function

The measured scattering intensity is the spherically averaged Fourier

transform of the auocorrelation of the electron density of the particle
 Autocorrelation

 (r )   (r )   (  r )    ( r  u )  ( r ) dV u
Vu
 Autocorrelation

 (r )   (r )   (  r )    ( r  u )  ( r ) dV u
Vu
 Autocorrelation

 (r )   (r )   (  r )    ( r  u )  ( r ) dV u
Vu
 Autocorrelation

 (r )   (r )   (  r )    ( r  u )  ( r ) dV u
Vu
 Autocorrelation

 (r )   (r )   (  r )    ( r  u )  ( r ) dV u
Vu
 Autocorrelation

 (r )   (r )   (  r )    ( r  u )  ( r ) dV u
Vu

For a homogeneous particle Spherical average Characteristic Function

 (r )   r V  ( r )   (r )  0 (r )   (r ) /  (0 )
with  ( 0 )   V
2
 0 r V
“probability of finding a point
Pair distance distribution function: within the particle at a distance r
from a given point”

p ( r )  r  ( r )   Vr  0 ( r )
2 2 2
 Pair distance distribution function p(r)

The p(r) function represents the histogram of

distances between pairs of points within the
particle. Dmax is the maximum diameter in the
particle.

Measured scattering intensity Pair distance distribution


Dmax
p ( r )  4  I ( q ) qr sin( qr ) dq
0
 Scattering from model structures

Adopted from Svergun & Koch, “SAS studies of biological macromolecules in solution”, Rep. Prog. Phys. 66 (2003) 1735-1782, Fig. 5 (c)I
 Particles in Solution

For solution scattering we typically require the following characteristics:

• Monodisperse, i.e. identical particles

i j ( q )  i1 ( q ) j

• Uncorrelated, i.e. no inter-molecular interactions present

I (q )  n j
i j (q )
j

I ( q )  Ni 1 ( q )
 Background Scattering and X-ray Contrast
I solution ( q ) I solvent ( q ) I particle ( q )

• The solvent scattering background must be properly subtracted

to obtain the signal from the particles
• the contrast, that makes the particles “visible” for X-rays, is the
difference in electron density of the particle versus the solvent

(  (r )   s )
2
Protein solution scattering data

• weak level of scattering at small angles

• drops off quickly for higher angles
• due to low contrast scattering level of
background and sample is very similar
except for the lowest angles
• background and sample scattering need
to be measured with high accuracy

• a 1mg/ml solution of a globular protein

of the size of lysozyme (14kD) scatters on
the order of:
1 out of 106 incident photons

“…. one in a million!”

 X-ray Contrast and Contrast Variation

• change contrast by adding salts (e.g.

Substance Average Contrast CsBr), sucrose or glycerol to the solvent
(x1010 cm-2) • but that changes the chemical
Protein 2.5 environment for the particles
• other possibility to change contrast is
Nucleic Acid 6.7 anomalous scattering
Fatty Acid -1.1

Carbohydrates 4.5 Note:

Contrast variation is widely used in
neutron scattering, due to the
large scattering length difference
of hydrogen and deuterium
 Introducing the Radius of Gyration

Rg2 is the average electron density

 r  (r ) d r
2
weighted squared distance of the 2
scatters from the centre of the Rg 
object   (r ) d r

2
1 • Solid sphere radius R:
1 3
12 Rg = √(3/5) R
• Thin rod length L
Rg2 =(12+ 12+ 12+ 22+ 22+ 32 )/6=20/6
Rg = √(1/12) L
Rg=√3.333 = 1.82 • Thin disk radius R:
Rg = √(1/2) R
 The Guinier approximation
The low-q region of the scattering curve is characteristic
for the overall dimension of the particle.
1
Guinier region

I(q) 2
lim I  I 0 exp( 
2
q Rg )
q 0 3

I0 is proportional to Mw Radius of gyration:

0.1 0.2 0.3 0.4 0.5 size of the particle
q

“The Guinier Plot”

Plot ln I against q2 →Straight line, slope –Rg/3

Ln(I(q)
1 2
ln I  ln I 0 
2
q Rg
3

Deviation from the straight line in the

Guinier plot indicate intermolecular
interaction or aggregation 0.002 0.004 0.006 0.008
q*q
 The Guinier approximation

Plot ln I against q2 →Straight line “The Guinier Plot”

1 2
ln I  ln I 0 
2

Ln(I(q)
q Rg
3

N N
sin( qr ij )
Recall: I (q )   fi (q ) f j (q )
qr ij
i 1 j 1 0.002 0.004 0.006 0.008
q*q

2
 
thus I0    fi 
 i 
I0 
cMV
 partial
  0 
2

Na
i.e. the number of (excess)
electrons in the sample c: concentration : partial specific Vol.
M: molecular mass : prot. e-density
V: Volume 0: solvent e-density
 Radius of Gyration r  (r ) d r

2

2
Rg 
  (r ) d r

Alternatively to using the Guinier plot to determine

the Rg of the protein of can also use the following
experession involving the P(r) function:

D max
i.e. Rg equals the second moment

2
r p ( r ) dr of the electron density distribution
Rg  0
D max
as well as half the second moment
of the distance distribution
2  p ( r ) dr function
0

This is often better than using the Guinier plot as it involves the whole scattering curve
 Radius of gyration for proteins and viruses

Molecular Weight Rg (A)

Ribonuclease 12700 14.8

Lysozyme 14800 14.5
B-lactoglobulin 36700 21.7
BSA 68000 29.5
Myosin 493000 468
Brome Mosaic Virus 4.6 106 134
TMV 3.9 106 924
 Kratky analysis

• Kratky plot: I*q2 vs. q

• sensitive to morphology of particle
• sensitive to the compactness of a protein
• unfolded and folded states of proteins are
easy to distinguish

Hiller et al., Biomaterials, 24 (2003), Fig5

Example: folding of cytochrome C

Putnam et al., Quat.Rev.Bioph 40,3 (2007), Fig24

Akiyama et al., PNAS, 99, (2002)

 ab-initio structure determination
Envelop models
• using spherical harmonics to produce molecular
envelopes that fit the experimental scattering data

Bead models
• fitting the scattering data using bead as
scattering centers
• so-called dummy residues (scattering centers
representing the Ca atoms of the residues)

Can be extremely powerful particularly if combined

with (partial) crystal structures if available!

but be careful: Program packages:

• ATSAS from EMBL Hamburg (Svergun group)
you will always get a structure from these • IMP from UCSF (Sali group)
programs, but it doesn’t mean they make • SAXS3D from Stanford (Doniach group)
sense • SASTBX from LBL (Zwart group)
•…
 Experimental setup

2
I (q )  N F (q ) S (q )

guard slit 4  sin 

q 
detector

defining slit 2
sample q 
D

isotropic scattering signal:

the 2D detector image is
integrated to yield I(q) vs q
 BioSAXS instrument at SSRL BL 4-2
flightpath: 0.2 – 3.5m
detector

sample incident beam

• widely re-configurable instrument for Q-range Q= 0.003/Å ... 4.2/Å

• static and time-resolved solution scattering
• lipid/fiber diffraction
• grazing incidence scattering
• anomalous scattering
• variety of advanced sample environments
• solution scattering robot with attached analysis pipeline
• in-line size-exclusion chromatography setup
• stopped-flow mixer with low sample consumption
• humidity chamber for lipid studies
• high-throughput LCP screening setup

30
 Why do SAXS?
Structural information obtainable from SAXS
• Radius of gyration (globular, cross-sectional etc.)
• molecular weight (monomer, dimer, multimer ..)
• pair-distance distribution function
• low-resolution envelope of molecule and ab-initio
structures (about 1nm resolution)
• unfolded vs folded (Kratky plot)
• interaction potentials

Systems that can be studies by SAXS

• study protein at physiological conditions
• time-resolved studies possible (reaction kinetics)
• large protein complexes (no need for crystals)
• unfolded or partially folded proteins
• complex systems (protein-DNA, protein-lipid …)
Thank you