B. Jaun, M.-O. Ebert: Structure determination by NMR 3.

1
specLroscopy:
PeLeronuclear shlfL correlaLlon
3.1 On the relative sensitivity of NMR experiments
The signal to noise ratio of NMR signals is proportional to:

(
excited
) (
detected
)
3/2
B
0
3/2
(number of transients)
1/2

As discussed in the preceding chapter, polarization transfer from
1
H (excited) to X (detected) leads to
a gain in sensitivity of up to
H
/
X.
The gain in sensitivity (see table below) is even larger for methods in
which protons are not only the excited nucleus, but also the detected one.

Relative sensitivity at 100% abundance without NOE
Combination of
nuclei
X excited
X detected
H excited
X detected
X excited
H detected
H excited
H detected
1
H/
31
P 1/10 1/4 2/5 1
1
H/
13
C 1/32 1/8 1/4 1
1
H/
15
N 1/300 1/30 1/10 1
Examples Inv. gated
13
C DEPT, INEPT PH-COSY HSQC, HMQC

Relative experiment times for identical S/N
Combination of
nuclei
X excited
X detected
H excited
X detected
X excited
H detected
H excited
H detected
1
H/
31
P 100 16 6.25 1
1
H/
13
C 1024 64 16 1
1
H/
15
N 100000 1000 100 1

The relevant relaxation time for
1
H-excited methods is T
1
of protons, which is usually much shorter
than T
1
of X-nuclei. Hence,
1
H-excitation has the additional advantage that more transients can be
acquired per time because only relaxation of protons has to be awaited in between transients.

B. Jaun, M.-O. Ebert: Structure determination by NMR 3.2
3.2 Introduction to 2D spectroscopy
Example: HSQC proton detected
1
H-
13
C-shift correlation via
1
J
CH

Basic HSQC Pulse sequence: double INEPT transfer from
1
H to
13
C and back.

Because HSQC is based on single quantum coherences of both,
1
H (I) and the heteronucleus S
(
13
C,
15
N etc.), it can be described correctly within the vector model of theory. In the following, the
vector model is used to explain how the chemical shift of the S-spin precessing during t
1
is modulating
the final I-spin signal acquired during t
2
. This is the basis of all n-dimensional NMR methods, but in
most of the 2D-experiments discussed in this course, multiple quantum coherence (for the discussion
of coherence see chapter 4) is involved in the mixing process, which makes its description within the
vector model impossible.


1
P
1
4 !
1
4 !
1
4 !
L
1
2
L
1
2
1
4 !
13
C
CA8 Cu
ACC
lnL1 lnL1
a
-x
8
A
L
1
/2
A
8

8

(n
S
/2)
x

8
A

A
= (U
S
+n!
lS
)-L
1
/2

8
= (U
S
-n!)-L
1
/2
L
1
/2
8
A

= n - U
S
-L
1

n
P

+cos U
S
L
1
-cos U
S
L
1
x
+cos U
S
L
1
-cos U
S
L
1
8
A
z
x x
S-Magneuzauon:
8
A

o
l
l
S
S
S =
13
C,
13
n
l =
1
P
n
1-4
: populauons aL
8olLzmann equlllbrlum
A = n
1
- n
2
8 = n
3
- n
4

n
4

n
2

n
3

n
1

8
A

o
l
l
S
S
n
4
+ A8'
n
2
+ AA'

n
3
- A8'
n
1
- AA'

aL ume a
(aer Lhe rsL lnL1 sLep)
aL ume b
B. Jaun, M.-O. Ebert: Structure determination by NMR 3.3



8
A

o
l
l
S
S
n
4
+ A8'
n
2
+ AA'

n
3
- A8'
n
1
- AA'

S-magneuzauon
aL ume b
A = 8 = n
3
-n
4
= n
1
-n
2
= S
o = = n
1
-n
3
= n
2
-n
4
= l

n
1
- AA' - n
2
- AA' = + cos U
S
L
1

AA' = S - (cos U
S
L
1
)/2

n
3
- A8' - n
4
- A8' = - cos U
S
L
1

A8' = S + (cos U
S
L
1
)/2


8
A
z
x
+cos U
S
L
1
- cos U
S
L
1
l-magneuzauon
aL ume b:
Z-magneuzauon of o: = n
1
- AA' - n
3
+ A8' = l + cos U
S
L
1
Ao' = (cos U
S
L
1
)/2
Z-magneuzauon of : = n
2
+ AA'- n
4
- A' = l - cos U
S
L
1
A' = - (cos U
S
L
1
)/2

8
A

o
l
l
S
S
n
4
- A'
n
2
+ A'

n
3
- Ao'
n
1
+ Ao'

(n
l
/2)
x
(n
l
)
y
x
o

z
x
l-magneuzauon
aL ume b
+cos U
S
L
1
-cos U
S
L
1
o

y
4!
lS
1
x

o
-
x

o
- x

o
-
(n
S
)
x
4!
lS
1
x

o
1he slgnal of ls:
l
= 2-cos U
S
L
1
-cos U
l
L
2
-cos n!
lS
l-magneuzauon
aL sLarL of
acqulsluon
Aer l1 wlLh quadraLure deLecuon ln boLh dlmenslons
a pure absorpuon cross peak aL u
S
|u
l
ls obLalned

!
lS
U
l
U
S
B. Jaun, M.-O. Ebert: Structure determination by NMR 3.4
Generalization:
The generic elements of 2D (and nD) NMR spectroscopy
Whenever the signal actually acquired during t
2
is also a function of the evolution time t
1
, the resulting
series of FIDs (one per t
1
increment) can be Fourier transformed for each time variable (t
2
and t
1
)
separately giving an 2-dimensional frequency spectrum. The basic idea of 2D-NMR-spectrocopy is to
generate a particular excited spin state during the preparation period (e.g. with a 90 pulse, or a
polarization transfer step), let the system evolve during time t
1
(the evolution period) then use a mixing
element to transfer magnetization (more general: coherence) between spins and finally detect x,y-
magnetization (single quantum coherence) during the detection period t
2
. The physical basis of the
mixing process is typically either J-coupling or dipolar cross relaxation. For each value of t
1
, the FID is
stored separately. The resulting array of FIDs is first Fourier transformed under t
2
to give an array of
spectra (the interferogram). In the next Fourier transform under t
1
, the vector constructed from the n
th

data point of each spectrum is then treated as a time-domain signal and is transformed under t
1
.



Series of FIDs, one tor each t
1

Interferogramm
Preparation Evolution Mixing Detection
t
2
t
m
t
1
B. Jaun, M.-O. Ebert: Structure determination by NMR 3.5

2D-spectrum

Quadrature detection in the indirect dimension t
1

Echo/anti-echo method (N/P-type selection)
This method uses phase cycles or gradients to select either positve (P-type) or negative (N-type)
coherence order during t
1
while eliminating the other one. Depending on whether the signal is
amplitude modulated or phase modulated by t
1
-evolution, the lineshape of echo-antiecho spectra can
be either pure absorption in both dimensions or mixtures of absorption and disperision. In the latter
case, the resolution is reduced and the spectra have to be plotted in absolute mode.
The hypercomplex method (RSH, States)
In the actual acquisition (t
2
), the receiver generates two signals by phase detection at 0 and 90, as in
1D-spectroscopy. After digitization the two arrays are stored separately and are used as the real and
imaginary parts in the complex FT. In the indirect dimension t
1
, the pulse sequence has to generate
two FIDs that are 90 out of phase of each other for each t
1
. These are

again stored separately in
order to be used as real and imaginary parts for complex FT under t
1
. The 90 phase-shifted signal is
obtained by changing the phase of the pulse after the evolution time by 90.

B. Jaun, M.-O. Ebert: Structure determination by NMR 3.6
For each t
1
value, FID1 and FID2 are stored in separate memory blocks. Together with the two
quadrature signals from channels A and B in t
2
, one obtains four memory blocks. Complex FT for both,
t
1

and t
2,
gives a 2D spectrum that is a matrix of four blocks: rr, ri, ir, and ii. Phase correction in each
direction mixes the two blocks in the corresponding dimension such that in the end, the rr block (the
one that is usually displayed and plotted) contains the 2D spectrum with absorption line shapes in both
dimensions.



TPPI (time-proportional phase incrementation)
Only one FID is acquired per t
1
value, but the t
1
increment is half of that used with the States method
(doubling the SW in
1
). This gives a single block with twice as many FIDs as in the hypercomplex
mode. Each time t
1
is incremented, the phase of a pulse is shifted by 90 relative to the receiver
phase. The spectrum obtained after a real FT in t
1
can be folded around
1
=0 and the reference is
shifted back into the middle of the
1
-domain. This procedure is analog to the Redfield method for
quadrature detection in 1D that was used by old Bruker spectrometers (<1985).


quadrature detection in t
2
q
u
a
d
r
a
t
u
r
e

d
e
t
e
c
t
i
o
n


i
n

t
1
FID
1
FID
2
FT(
2
)
complex
0 90
R
1
I
1
FID
2
FID
1
0 90
I
2
R
2
FID
spectrum
FT(
1
)
complex
RI II
RR IR
B. Jaun, M.-O. Ebert: Structure determination by NMR 3.7
The TPPI and States procedures are equivalent with regard to experiment time, S/N and digital
resolution. Certain artifacts, in particular so called axial peaks, appear at the center of the
1
domain
with States but at the edge of the spectrum with TPPI. Folding in
1
is also different in the two cases
(see folding with real and complex transforms in 1D). A forth scheme is called StatesTPPI. It
combines the advantages of the hypercomplex method with the preferred location of artifacts at the
edge of the spectrum and is therefore used in most cases today.

3.3 HSQC (Heteronuclear Single Quantum Correlation)
1
H-
13
C-correlation over one bond via
1
J
CH

Revisiting the HSQC experiment we can assign the elements of the pulse sequence as follows: The
preparation part is identical to INEPT (up to time point a). During t
1
,
13
C-magnetization precesses in x,y
while
1
H-magnetization is in z. Therefore, neither chemical shifts nor homonuclear couplings of
protons evolve during t
1
. The 180 pulse in the center of t
1
refocuses
1
J
CH
(therefore, the cross peaks
are not split by
1
J
CH
in the
1
-dimension of the final spectrum). At the end of the evolution period,
coherence is partially retransferred to
1
H-SQC, which is detected (reverse polarization transfer).
GARP-broadband
13
C-decoupling during t
2
collapses the cross peaks which would otherwise be split
by the heteronuclear coupling
1
J
CH
of 125-200 Hz in
2
(remember, it is the
13
C-satellites in the proton
spectrum, which are actually measured). Since broadband
13
C-decoupling of all
1
H-bearing carbons
(ca. 160 ppm) is demanding on the decoupler and heats the sample, in particular if it has a high
dielectric constant, GARP decoupling during acquisition may be left out at the cost of a more
complicated spectrum. HSQC without decoupling
13
C during t
2
is also the method of choice for
measuring
1
J
CH
.
Because only 1.1% of the protons are bound to
13
C, the 100 times stronger signal of the
12
C bound
protons has to be suppressed. If the selection is based on phase cycles, very high spectrometer
stability is required whereas with gradient-based selection only the signal of
13
C-bound protons
reaches the receiver.
Of the two coils of a heteronuclear probehead, the inner one, which is closer to the sample and
therefore more sensitive, is usually tuned to the detected frequency. For optimal X-detection, the inner
coil is tuned to X and the outer coil is used for
1
H-decoupling. For
1
H-detected heteronuclear
experiments, the opposite arrangement is used:
1
H on the inner coil and X on the outer coil (so-called
inverse or reverse probes)
The pulse sequence for HSQC shown above is the minimal version. Actual modern pulse sequences
are much more complicated as exemplified by the pulse sequence for multiplicity edited HSQC with
sensitivity enhancement and coherence pathway selection by gradients from todays Bruker library
shown below:
B. Jaun, M.-O. Ebert: Structure determination by NMR 3.8

HSQC gives clean absorption line shapes in both dimensions. However, the HSQC pulse sequence is
more complex and longer than that of HMQC and therefore more dependent on correct pulse
calibrations (modern versions use adiabatic composite pulses for
13
C-180). For the same reason,
HMQC often gives better results than HSQC for samples with broad lines (short T
2
).
Example for proton detected
1
H-
13
C shift correlation: Multiplicity edited HSQC




strychnine
N
O
O
H
H
H
H
N
B. Jaun, M.-O. Ebert: Structure determination by NMR 3.9
Coherence transfer pathway selection by gradients
Strychnine [c=50 mM] in CDCl
3
. 400/100 MHz. Acquisition: 2k(t
2
) x 256(t
1
), 2 scans/t
1
-increment. Optimized for
1
J
CH
=145 Hz. GARP-
13
C-CPD-decoupling in t
1
. Experiment time: 16 min. With adiabatic
13
C inversion pulses and
sensitivity enhancement. Processing: 1k x 1k, cos
2
window functions in t
2
and t
1
.





B. Jaun, M.-O. Ebert: Structure determination by NMR 3.10
3.4 HMQC (Heteronuclear correlation through Multiple
Quantum Coherence)
Pulse sequence of HMQC:

The first 90
1
H-pulse generates
1
H-SQC; the first 90
13
C-pulse transforms this partly into
heteronuclear double and zero quantum coherence. The
1
H-180 pulse in the middle of t
1
refocuses
the
1
J
CH
coupling and swaps ZQC (evolution with
C
-
H
) and DQC (evolution with
C
+
H
). At the
end of t
1
, the effects of
1
H chemical shifts during the two halves of t
1
cancel, and only
C
labeling
remains. Homonuclear
1
H-
1
H-couplings evolve during t
1
as well as t
2
. This homonuclear coupling leads
to a tilt in the cross peaks and reduces the possible resolution in the t
1
-domain. For small to medium
sized molecules, HSQC therefore gives better S/N for a given experiment time and better resolution in

1
than HMQC.

3.5 HMBC (Heteronuclear correlation through multiple
bonds)
1
H-
13
C-correl ati on over several bonds vi a
2, 3, ( >3)
J
CH
Pulse sequence: Long range variant of HMQC with low pass filter to suppress one-bond correlations.
Most
1
H-
13
C-scalar coupling constants over two and three bonds are in the 0 to 10 Hz range,
comparable to
1
H-
1
H couplings (see collection in compendium). Coupling constants
4
J
CH
over more
than three bonds are usually very weak and are rarely detected by HMBC with typical delays for
evolution of the long-range coupling of 70 ms (optimal for
2,3
J
HC
= 8 Hz). Because the 2-bond and 3-
bond coupling constants cover the same range 2-bond and 3-bond correlations cannot be
distinguished in HMBC spectra. Although less sensitive than HSQC and HMQC, HMBC is more
1
P
13
C
L
1
2
-x
1
2 !
CP
90
-x
180
-x
CA8 Cu
ACC
Lxchanges uCC and ZCC
8efocuses !
CP
ln L
1
heLeronuclear
uCC
and
ZCC
L
1
2
1
2 !
CP
90
-x
90
-x
heLeronuclear
uCC
and
ZCC
1
P magneuzauon
anu-phase wlLh regard
Lo
1
!
CP


B. Jaun, M.-O. Ebert: Structure determination by NMR 3.11
sensitive than INADEQUATE by several orders of magnitude and delivers the same type of
information: connectivity of the carbon skeleton including quaternary carbons. In contrast to
INADEQUATE, HMBC also correlates fragments across heteroatoms.

The following modifications are made to the HMQC sequence in order to measure long-range
correlations:
A so-called low pass filter eliminates signals of directly bound protons. It works because of the
difference of more than one order of magnitude between
1
J
CH
and
2,3
J
CH
. When the magnetization of
directly bound protons is in antiphase (after 1/(2
1
J
CH
) s), the small long-range couplings are still mostly
in phase. A 90 pulse at this time converts the magnetization due to directly bound protons into DQC
and ZQC but does not affect the signals of remote protons. A second, much longer delay brings the
long range coupling into antiphase. From thereon the sequence is identical to HMQC. Because
considerable signal is already lost due to relaxation during the long delay 1/(2
2,3
J
CH
), no attempt at
refocusing is made (that would require a second delay of equal length), and no
13
C-broadband
decoupling is applied during t
2
. The line shapes are mixed absorptive/dispersive because the
homonuclear
1
H-
1
H-coupling constants are comparable to those of heteronuclear long range
couplings. Therefore, the spectra have to be processed in magnitude or power mode and the cross
peaks appear tilted.

1
P
13
C
L
1
2
-x
1
2
n
!
CP
90

180

ACC
heLeronuclear
uCC
and
ZCC
L
1
2
90

90

heLeronuclear
uCC
and
ZCC
Low pass lLer:
ellmlnaLes slgnals
of dlrecLly bound 1P
90

1
2
1
!
CP
B. Jaun, M.-O. Ebert: Structure determination by NMR 3.12

Example for HMBC
Coherence transfer pathway selection by gradients
Strychnine [c=50 mM] in CDCl
3
400/100 MHz. Acquisition: 4k(t
2
) x 512(t
1
), 2 scans/t
1
-increment. Optimized for
2,3
J
CH
=8 Hz. Experiment time: 45 min.
Processing: 2k x 1k, sine window functions in t
2
and t
1
, absolute value plot.



B. Jaun, M.-O. Ebert: Structure determination by NMR 3.13
3.5 Proton detected H,X-COSY for abundant X nuclei
This
1
H-detected method is convenient for the correlation of protons with abundant X-nuclei such as
31
P. The pulse sequence is the direct heteronuclear analog of COSY with two simultaneous 90 pulses
on
1
H and X acting as the mixing pulse. At the beginning of the sequence, all natural
1
H-magnetization
is destroyed by a train of 180 pulses on
1
H. Therefore, only the chemical shift and coupling of X
evolve during the evolution time t
1
. The resulting 2D spectrum shows cross peaks in antiphase
absorption in both dimensions.

Example:
1
H,
31
P-COSY allows to correlate sequential sugar units in oligonucleotides through three-
bond couplings between the sugar protons and the
31
P of the phosphodiester linkage CHOPO
2
O-
CH-. In the example of a non-natural oligonucleotide with arabinopyranose sugars below, each
phosphorus shows cross peaks to the 2'-H of the preceding and the 4'-H
2
of the following sugar in the
sequence (plus a weak
4
J
HP
W-type cross peak to the 3-H of the following sugar).


1
P
31

-x
180

90

ACC
n
90

90

resaLurauon
of naLural
1
P magneuzauon
L
1