Figure 11.40

Figure 11.40

Furthermore, since the FID observed in each scan contains the large 12C-bound !H signal as well as the weak 13C-bound 1H signal, the receiver gain must be turned down to prevent this signal from overloading the digitizer, even though 99% of the signal will be canceled when the second FID is added to the sum in memory. Less amplification of the FID can result in lower signal-to-noise in the spectrum. Finally, subtraction of two FIDs is not always perfect, and subtraction artifacts will give vertical streaks in the spectrum (see Fig. 11.13). The solution to all of these problems is to use gradients in place of a phase cycle to eliminate the 12C-bound 1H signal. Because the "twist" imparted by a gradient is proportional to the magnetogyric ratio, a 1H SQC signal is twisted four times as much by a given gradient pulse as a 13 C SQC signal. Thus if a gradient of relative strength 4 is applied during the evolution time (i1), the 13C SQC signal (p = 1) will acquire a twist down the gradient axis (usually z) of 4 units (Fig. 11.40). A second gradient of relative strength —1 is applied after this 13C magnetization is transferred back to 1H SQC (p = 4) by the INEPT sequence, imparting an additional twist of —4 units, leading to a total of 0 units of twist:

Thus the signal will be perfectly "untwisted" and will be observed in the FID. The 12C-bound 1H signal cannot transfer from 13C SQC to 1H SQC during the INEPT sequence, so it will end up with a net twist at the beginning of the FID and will not be observed. With gradient selection of the coherence pathway, no 12C-bound 1H "streaks" are observed even with a single scan per FID. There is no harm in combining methods, so the phase cycle is always included even in the gradient experiment if you want to use more than one scan per FID.

11.7.4 Gradient-Selected HSQC with Phase-Sensitive Data Presentation

We saw with the gradient DQF-COSY experiment that the relatively long gradients 1 ms) allow chemical shift evolution that will produce large chemical-shift dependent phase errors in the final spectrum. In the sequence of Figure 11.40, the gradient placed in the second half of the t1 period will set a minimum value for t1 of twice the gradient time (and

recovery delay). Any time an FID is "started late" there will be chemical-shift evolution before it starts, leading to huge chemical-shift dependent phase errors ("phase twists"). In this case the phase errors will show up in F\, since we are starting the t\ FID late. The solution, as always, is to build a spin-echo with the gradient in one of the spin echo delays (Fig. 11.41). Because we have 13C SQC during the ti delay, we use a 13C 180° pulse in the center of the spin echo to refocus both 13C shift evolution and /-coupling evolution. The second gradient is already contained in a spin echo, so there is a big enough "gap" (1/(4/) = 1/(4 x 150 Hz) = 1.67 ms) to fit a typical gradient pulse (1 ms) and its recovery delay (200 |xs). Using the spherical operators to describe the coherence pathway, it becomes clear that the gradients must now be of the same sign to select the desired (echo) pathway. If echo-antiecho phase encoding is used in t1, the first gradient (relative strength +4) would alternate sign to select S+1° during t1 for the first FID (G1 = +4, echo signal) and to select S-1° during t1 for the second FID (G1 = -4, antiecho signal) acquired for each value of t1.

A different approach to destroying the 12C-bound 1H artifact is to use the gradient in a simpler way—as a "spoiler" that just kills all of the magnetization in the X-y' plane while our desired signal is "stored" briefly on the z axis. To do this we go back to our discussion of intermediate states in INEPT coherence transfer (Section 7.10) and recall that instead of using simultaneous 90° pulses on 1H and 13 C to effect coherence transfer, we can start with the 1H 90° pulse and then, after a short delay, complete the INEPT transfer with the 13C 90° pulse:

2Ix S

x Sz

90? 13C

z Sz

2Sv I

y Iz

(again we omit the factor of 4 reflecting our change to 13 C Mo as a standard of comparison because coherence will be transferred back to 1H in the end). The — 2IzSz product operator will not be affected by a gradient, nor will it undergo any evolution during the time of the gradient. 12C-bound 1H coherence cannot achieve this spin state, so any coherence it has at this point (Ix or Iy) will be destroyed. We call this a "spoiler" (or homospoil) gradient because the twist it produces is never untwisted in the pulse sequence—we are just getting rid of stuff. The same strategy can be used in the back transfer (Fig. 11.42), usually with a

n i/(4/)

1/(47) [1

0 0

Post a comment