EUROMAR 2011
• Why we need coherence selection
Coherence order and
coherence selection
James Keeler
Department of Chemistry
Why we need coherence selection
DQF COSY
DQ spectrum
NOESY
The spins don’t know what we want !
We want one out of many possibilities
Outline
• PDF of these slides available at
• Concept of coherence order
• Coherence transfer pathways (CTPs)
• Selecting a CTP with phase cycling
• Selecting a CTP with gradients
• Suppression of zero-quantum
coherence
Coherence order, p
Defined by phase acquired during
rotation by about z
ρˆ ( p ) ⎯rotate
⎯⎯ by φ about z
⎯⎯ ⎯→ ρˆ ( p ) × exp(−ipφ)
phase acquired is −pφ
different p separated by using this property
Further information
http://www-keeler.ch.cam.ac.uk/
• See also:
Understanding NMR Spectroscopy,
James Keeler (Wiley) [Chapt. 11]
Spin Dynamics. Basics of Nuclear
Magnetic Resonance, Malcolm Levitt
(Wiley)
Properties of coherence order
• takes values 0, ±1, ±2 …
0 is z-magnetization,
±1 is single quantum,
±2 is double quantum etc.
• only p = −1 is observable
• maximum/minimum value is ±N, where
N is number of spins
1
 Effect of pulses Coherence transfer pathway (CTP) Heteronuclear experiments
Indicates the desired coherence order
RF pulse all possible
p at each point HMQC
values of p
- which is why we need selection

special case: DQ spect. NOESY

DQF COSY
180° pulse separate p for each nucleus (pI, pS)
p −p note: always starts at p = 0 ends with p = −1 on observed nucleus
always ends at p = −1
pulse to S only affects pS

Frequency discrimination
- or, alternatively
and lineshapes in 2D
for absorption mode spectra must retain
p = ± 1 during t1: symmetrical pathways
N-type
Phase cycling
P-type

1. record two separate spectra:

echo or N-type: p = + 1 during t1
combine this with frequency discrimination
anti-echo or P-type: p = − 1 during t1
using ‘TPPI’ or ‘States’
2. combine to give absorption spectrum

2
 Pulse phase Receiver (rx.) phase Receiver phase
rx. phase

fixed rx.
phase
If the signal generated by the pulse
sequence shifts in phase, then this can
always be compensated for by shifting
rx. phase the receiver by the same amount.
follows
pulse phase

the phase of the spectrum depends on

the phase of the pulse

Phase cycling Effect of phase shift of pulse Selection of a single pathway

Selection of a pathway by repeating the

sequence with a systematic variation
of the pulse and rx. phases
How to design the sequence of phases,
- the phase cycle?
Pulse causes transfer from p1 to p2 +2 to −1, so Δp = –1 – (+2) = – 3
Change in coherence order Δp = p2 – p1
phase acquired by signal when pulse
If pulse phase shifted by Δφ phase shifted by Δφ is
acquired by signal is
− Δp × Δφ = 3 Δφ
−Δp × Δφ

3
 Four-step cycle Four-step cycle Four-step cycle
step pulse Δφ 3 Δφ equiv(3 Δφ) step pulse Δφ 3 Δφ equiv(3 Δφ) step pulse Δφ 3 Δφ equiv(3 Δφ)

1 0° 1 0° 0° 1 0° 0° 0°

2 90° 2 90° 270° 2 90° 270° 270°

3 180° 3 180° 540° 3 180° 540° 180°

4 270° 4 270° 810° 4 270° 810° 90°

Four-step cycle - other pathways Selected pathways

Pulse goes e.g. Δp = – 2 so − Δp × Δφ = 2 Δφ rx. phase [0°, 270°, 180°, 90°]
[0°, 90°, 180°, 270°]
step pulse Δφ 2 Δφ equiv(2 Δφ)
Pathway with Δp = −3 acquires phase 1 0° 0° 0° Δp= –3
[0°, 270°, 180°, 90°] 2 90° 180° 180°
rx. phase coherence
3 180° 360° 0° phase
If receiver phase follows these phases, 4 270° 540° 180°
contribution from the pathway will add up For Δp= –3, rx. phase follows coherence
Selected with rx. phases phase:
- but what about other pathways? [0°, 270°, 180°, 90°] ? all four steps add up

4
 Selected pathways Selectivity Combining phase cycles
rx. phase [0°, 270°, 180°, 90°] A four-step cycle designed to select
a particular value of Δp will also select
Δp + 4, Δp + 8 … and Δp − 4, Δp − 8…
Δp= –2
four-step cycle to select Δp = +1
rx. phase coherence
- all other pathways are suppressed step pulse Δφ1 − Δφ1 equiv(− Δφ1)
phase
(−4) −3 (−2) (−1) (0) 1 (2) (3) (4) 5 1 0° 0° 0°
2 90° −90° 270°
For Δp= –2, signal cancels on four steps
3 180° −180° 180°
selected in bold, suppressed in () 4 270° −270° 90°

Combining phase cycles Complete both cycles independently Tricks: 1

step Δφ1 − Δφ1 equiv(−Δφ1) Δφ2 2 Δφ2 equiv(2Δφ2) total
1 0° 0° 0° 0° 0° 0° 0°
2 90° −90° 270° 0° 0° 0° 270° 1. The first pulse can only generate
3
4
180°
270°
−180°
−270°
180°
90°
0°
0°
0°
0°
0°
0°
180°
90°
p = ±1 from equilibrium magnetization
5 0° 0° 0° 90° 180° 180° 180°
6 90° −90° 270° 90° 180° 180° 90°
four-step cycle to select Δp = −2 7 180° −180° 180° 90° 180° 180° 0°
8 270° −270° 90° 90° 180° 180° 270°
- no need to phase cycle this pulse
step pulse Δφ2 2 Δφ2 equiv(− Δφ2) 9
10
0°
90°
0°
−90°
0°
270°
180°
180°
360°
360°
0°
0°
0°
270°

1 0° 0° 0° 11 180° −180° 180° 180° 360° 0° 180°

12 270° −270° 90° 180° 360° 0° 90°
2 90° 180° 180° 13 0° 0° 0° 270° 540° 180° 180°
14 90° −90° 270° 270° 540° 180° 90°
3 180° 360° 0° 15 180° −180° 180° 270° 540° 180° 0°

16 270° −270° 90° 270° 540° 180° 270°

4 270° 540° 180°

5
 Tricks: 2 Tricks: 3 Tricks: 4
2. Group pulses together and cycle as
a unit 3. Only p = −1 is observable, 4. Don’t worry about high orders of
so it does not matter if other values of multiple quantum coherence
p are generated by the last pulse e.g ≥ 4.

- no need to phase cycle the last pulse, - they are hard to generate and likely
if a coherence order has been selected to give weak signals, especially if
All pulses: [0°, 90°, 180°, 270°] unambiguously before this pulse the lines are broad
Rx. for Δp = ±2: [0°, 180°, 0°, 180°]

Refocusing pulses: EXORCYCLE Axial peak suppression Examples: DQF COSY

Refocusing pulses cause p → −p z-magnetization which recovers by
symmetrical
relaxation during a pulse sequence is pathways in t1
made observable by last pulse
e.g. Δp = ±2
(single quantum) - leads to peaks at ω1=0: axial peaks
final pulse has Δp = −3 and +1
- easily suppressed using a two-step cycle - select using four-step cycle:
φ3 = [0°, 90°, 180°, 270°]
Pulse: [0°, 90°, 180°, 270°] 1st pulse: [0°, 180°] φrx = [0°, 270°, 180°, 90°]
Rx. for Δp = ±2: [0°, 180°, 0°, 180°] Rx. for Δp = ±1: [0°, 180°]
this is sufficient, as p can only be ±1 in t1

6
 DQF COSY (alternative) Examples: NOESY Examples: NOESY

symmetrical symmetrical
pathways in t1 pathways in t1

group first two pulses and select Δp = ±2 final pulse has Δp = −1 axial peak suppression also required
- select using four-step cycle: - select using four-step cycle: φ1 = [0°, 180°] φrx = [0°, 180°]
φ1 and φ2 = [0°, 90°, 180°, 270°] φ3 = [0°, 90°, 180°, 270°] Step 1 2 3 4 5 6 7 8
φrx = [0°, 180°, 0°, 180°] φrx = [0 °, 90°, 180°, 270°] Φ1 0° 0° 0° 0° 180° 180° 180° 180°
Φ3 0° 90° 180° 270° 0° 90° 180° 270°
this is sufficient, as p can only be −1 in t2 this is sufficient, as p can only be ±1 in t1 Φrx 0° 90° 180° 270° 180° 270° 0° 90°

Problems with phase cycling Field gradient pulses

• phase cycle must be completed: • the B0 field is made inhomogeneous

- unacceptably long experiment, for a short period (few ms)
especially for 2D/3D Gradient pulses
• coherences dephase, all signal lost

• cancellation of unwanted signals may • a subsequent gradient may rephase

be imperfect (especially for proton some of the coherences
detected experiments)

7
 Effect of a gradient Dephasing and rephasing Spatially dependent phase
gradient phase acquired by coherence p at
dephase position z in sample, after time t

φ(z) = −p × γ G z t

gyromagnetic ratio
gradient strength, G cm−1
active volume
rephase
off: sharp line on: v. broad line phase depends on position and p

Selection with a gradient pair Selection with a gradient pair Selection with a gradient pair

G1τ1 p
=− 2
G2 τ 2 p1

phase due to G1: φ1(z) = −p1 × γ G1 z τ1 φ1(z) + φ2(z) = −p1 γ G1 z τ1 − p2 γ G2 z τ2 G1τ1 −1 1

e.g. p1 = +2, p2 = −1 =− =
=0 G2 τ 2 +2 2
phase due to G2: φ2(z) = −p2 × γ G2 z τ2
G1τ 1 p if G1 = G2, τ2 = 2 τ1
refocusing condition: φ1(z)+ φ2(z) = 0 =− 2 alternatives
G2τ 2 p1 if τ1 = τ2, G2 = 2 G1

8
 Selection with a gradient pair Heteronuclear case Only one pathway selected

G1τ1 p only pS
=− 2 G1τ1 1
G2 τ 2 p1 changes =
G2 τ 2 2
G1τ1 −1 1 G1τ1 1
e.g. p1 = −2, p2 = −1 =− =− φ1(z) = −(pI γI+ pS γS)G1zτ1 = −(−γI+γS)G1zτ1 =−
G2 τ 2 −2 2 G2 τ 2 2
φ2(z) = −(pI gI+ pS γS)G2zτ2 = −(−γI−0)G2zτ2 can only select one of these pathways
refocusing: τ1 = τ2, G2 = −2 G1
G1τ1 1 - potential loss of sensitivity
‘−G’ means opposite sense of gradient =
G2 τ 2 (γS γI ) − 1 - problems in two-dimensional NMR

Refocusing pulses 180° in heteronuclear case Phase errors

DQF COSY

no coherence
on I spin

Ideal 180° causes p → −p

180° to I is acting as inversion pulse Offsets continue to evolve during gradients
Selected for all p by equal gradients Gradient pair ‘cleans up’ imperfect 180° - results in severe frequency-dependent
- ‘cleans up’ imperfect 180° - leaves S spin coherences unaffected phase errors

9
 Avoiding phase errors Selection of z-magnetization Examples: DQF COSY
add refocusing pulse / use an existing one
A gradient dephases all* coherences:
offset evolution - leaves behind only z-magnetization
refocused - simple and convenient
by 180° pulse
called a purge gradient or homospoil

more time • symmetrical pathways in t1 (no gradient)

efficient alternative *except homonuclear zero-quantum • extra 180° pulses to avoid phase errors
• loss of sensitivity

Examples: HMQC HMQC: refocusing condition HMQC: suppression of I spin

magnetization not coupled to S
I I
I
S S
S
P-type
(solid line)
(− γI G1 zτ1 − γS G1 zτ1 ) + (γI G1 zτ1 − γS G1 zτ1 ) + (γI G2 zτ 2 ) = 0
• separate expts. for P- and N-type • I magnetization dephased by 1st G1,
− 2γS G1 zτ1 + γI G2 zτ 2 = 0
• additional 180° associated with both G1 but rephased by second G1, and then
G1τ1 γ
• G2 in existing delay, so no phase error = I dephased by G2
G2 τ 2 2γS

10
 HSQC
• G1 is purge gradient
• extra 180° associated with G2
• G3 in existing spin echo
• can omit G2 and G3 (labelled samples)
An old, old problem in NMR
z-magnetisation and zero-quantum
coherence cannot be separated using
phase cycling or gradients
because
neither respond to z-rotations
i.e. both have coherence order, p,
of zero
Advantages and disadvantages
+ minimizes experiment time
+ excellent suppression, especially in
heteronuclear experiments with 1H obs.
Zero-quantum dephasing
- cannot select more than one pathway
→ possible loss of SNR
→ obtaining pure phase more complex
- phase errors
→ requires elaboration of sequence
- loss due to diffusion
Why is it a problem? Result: distorted multiplets in 2D
a 90° pulse converts z-magnetization
into in-phase magnetization along y
but converts ZQ into anti-phase
along x
the result is phase distortion and
unwanted peaks
z-magn. + ZQ z-magn. only
11
 Example: NOESY The z-filter but …
Sørensen, Rance, Ernst 1984 90°(y) 90°(−y)
t1 m t2
RF 90°(y) 90°(−y)
RF
G G
RF
wanted: z-magn. during m 2I1yI2z 2I1yI2x mixture of
G DQ and ZQ
→ in-phase, absorption multiplets
I1x −I1z I1x
½(2I1yI2x−2I1xI2y) ½(2I1yI2z−2I1zI2y)
unwanted: ZQ during m
everything else dephased ZQ
→ anti-phase, dispersion multiplets
‘J-peaks’ only in-phase magnetization survives Anti-phase component passes through

Make evolution dependent on

Zero-quantum evolution Zero-quantum dephasing
position
90° 90° 180 z
z As frequency is a function of position,
RF the zero-quantum coherence will
G dephase

position
The zero quantum evolves during Identical to dephasing in a conventional
τz at (Ω1− Ω2), the difference of gradient
the shifts

Macura et al 1981 this is the key … how to make 180° position dependent?
frequency
time

12
 Swept-frequency 180º z-filter with zero-quantum Typical parameters
suppression
normal spectrum
• swept pulse of duration 15 to 30 ms
apply gradient
RF
G • gradient 1 to 2 G cm−1
frequency position
swept 180º with gradient • dephasing rate depends on ZQ frequency

swept-frequency 180º pulse additional dephasing gradient • suppression of ZQ by factor of 100

(to make sure everything is dephased)
different parts experience pulse at different times

NOESY with zero-quantum NOESY results (strychnine) TOCSY

suppression
τm
conventional + ZQ suppression
t1 t1 DIPSI-2 t2
RF RF
G G

swept 180º with gradient isotropic mixing within z-filter

additional dephasing gradient ZQ dephasing needed before and

after mixing; unequal durations
NOE continues to build up throughout

13
TOCSY results (strychnine) TOCSY results (strychnine) Advantages of the z-filter
conventional + ZQ suppression conventional + ZQ suppress.
• excellent suppression
• no increase in experiment time
• simple to implement
• widely applicable
• negligible reduction in signal

Examples: HMQC Difference spectroscopy: HMQC

The End
select ΔpS = ±1 at first S pulse
and ΔpI = ±2 at 180° I pulse The cycle [0°, 180°] on first S pulse and
Phew! step
φS
1
0°
2 3 4 5 6 7
180° 0° 180° 0° 180° 0° 180°
8 rx. is just difference spectroscopy:
selects that part of the signal which
φI 0° 0° 90° 90° 180° 180° 270° 270°
goes via the S spin
φrx 0° 180° 180° 0° 0° 180° 180° 0°

14
 Difference spectroscopy

In heteronuclear experiments, a simple

two-step phase cycle (+x/−x) on the
pulse causing the transfer often suffices

- this is simply difference spectroscopy

15