Spin‐Noise‐Detected Two‐Dimensional Nuclear Magnetic Resonance at Triple Sensitivity

Abstract A major breakthrough in speed and sensitivity of 2 D spin‐noise‐detected NMR is achieved owing to a new acquisition and processing scheme called “double block usage” (DBU) that utilizes each recorded noise block in two independent cross‐correlations. The mixing, evolution, and acquisition periods are repeated head‐to‐tail without any recovery delays and well‐known building blocks of multidimensional NMR (constant‐time evolution and quadrature detection in the indirect dimension as well as pulsed field gradients) provide further enhancement and artifact suppression. Modified timing of the receiver electronics eliminates spurious random excitation. We achieve a threefold sensitivity increase over the original snHMQC (spin‐noise‐detected heteronuclear multiple quantum correlation) experiment (K. Chandra et al., J. Phys. Chem. Lett. 2013, 4, 3853) and demonstrate the feasibility of spin‐noise‐detected long‐range correlation.


Experimental Procedures
As mentioned in the main text, we modified the implementation of spin noise acquisition of previous publications [1][2][3] to avoid a hardware dependent artifact, which had been absent in our previous experiments. The artifact is generated, if one-dimensional spin noise spectra are recorded, noise block by noise block, while activating and deactivating the receiver channel for each block. The way onedimensional spin noise spectra are processed is as follows: All the noise blocks are Fourier transformed individually to give power spectra and are co-added after that (see Figure S1). Figure S1. Scheme of one-dimensional spin noise acquisition, where N long noise blocks are recorded. Depending on the required resolution each noise block can either be Fourier transformed as a whole before addition of the N resulting power spectra or split into overlapping sub-blocks [8] before Fourier transformation, allowing one to adjust resolution according to the sample properties. In order to assess the time dependent influence of the transmit/receive switch, the corresponding subblocks (A to E) of each of the N acquired original blocks were processes and co-added separately. By comparing noise spectra obtained from the different subblocks the time-dependent components can be identified.

Assessment of Switching Artifacts
As can be seen in Figure S5, the time-dependent component is strongest for sub-blocks taken from the beginning of each noise block and it decays within 7-20 ms (probe and sample dependent). This effect has been traced back to the activation of the transmit/receive (T/R) switch in the preamplifier. [5] The impedance change caused by the T/R switching apparently acts like a minute random pulse excitation. The strength and persistence of this effect is highly dependent on the probe and the type of preamplifier. It was found to be independent of the carrier frequency. Figure S5. Five 500 MHz 1 H spin noise spectra of a sample of 90% H2O / 10% D2O. The sub-blocks were extracted from the original noise blocks as indicated in Fig. S1 and processed accordingly. The original noise blocks consisted of 4096 data points, which were recorded over a period of 80 ms and then divided into 5 equally spaced overlapping blocks of 1024 data points (corresponding to 20 ms) as visualized in Fig. S1. Spectra A to E were obtained from the equidistant 1024 data point contiguous sub-blocks corresponding to delay times of 0 ms to 64 ms after the start of the original block. The proton spin noise signal of H2O is the negative peak at 4.7 ppm. The positive signals centered at 5.5 ppm are artifacts originating outside the spectrometer, most likely from a nearby TV station. The carrier frequency was set to 3.5 ppm.
For one-dimensional experiments this can be trivially accomplished by extending each noise block to the hardware maximum and discarding the noise recorded within 5×T2 * from the start of acquisition. For two-dimensional spin-noise-detected NMR spectra, where the acquisition of multiple non-contiguous individual noise blocks is required, the pulse programs have to be rewritten to prevent engaging the T/R switch. In Figure S6 we show a pulse program for a ctsnHMBC experiment, programmed such that the T/R switch is only activated once at the beginning of each burst of DBU-acquisition blocks. After the T/R switch is engaged, there is an additional delay of 20 ms, to allow for the artifact to decay. In order to avoid this potential source of interference in spin-noise-detected experiments, the programming of spin noise acquisition schemes was modified. Instead of activating the T/R switch for each noise block ( Figure S6, upper panel), the switch remains in the receive state continuously (in practice, until the hardware buffer is full, Figure S6, lower panel). Figure S6. The upper scheme shows the approach to transmit/receiver switch control used in previous experiments. Before each FID or noise block the switch is opened and after the acquisition the switch is closed again. The engagement of the switch however may cause a small pulse-like event at the beginning of the noise block, which shows up as an always positive distortion of the spin noise line shape, as shown in Fig. S2. This effect is only noticeable in spin noise experiments, due to the low amplitude of the phenomenon. Therefore, multiple FIDs or noise blocks are acquired with the switch continuously in the receive state, as shown in the lower scheme. This circumvents the problem if a delay ds of duration 5T2 * ) is inserted after the T/R switch is activated.
On the spectrometer systems we used here (Bruker Avance III) with the software version used (Topspin 3.5pl5), implementing acquisition as shown in the pulse program of Figure S6 requires one to set the "acquisition mode" to "analog" (i.e. disable the real time digital signal processor) to avoid digital filtering artifacts to occur when multiple noise blocks are concatenated. As a consequence of this "analogue" acquisition mode, the acquired signals contain a direct current offset, that may drift over the course of an experiment (20 h), and which may lead to a visible center artifact in the final spectrum. To mitigate this interference, the individual noise blocks are linearly offset corrected, i.e. their average is subtracted from each data point. There is one caveat with this approach: any genuine NMR signal at the center will be removed as well. The data presented here in this Supplementary Information were recorded on a Bruker 500 MHz Avance III spectrometer, equipped with a BBI probe. On our Bruker 700 MHz Avance III spectrometer with a TCI triple resonance cryoprobe, the effect is weaker and decays faster, but still might be problematic for spin noise experiments. In spin-noisedetected experiments using repetitive T/R switching with the acquisition and cross-correlation processing scheme outlined in the main text or applying the original acquisition scheme [1] the effect of the T/R switching artifact with is an increase of the uncorrelated noise as compared to the correlated spin noise signal, owed to the fact the T/R switch artifacts are not correlated between subsequently recorded noise blocks.

Spin Noise vs. Circuit Noise
While we have used the terms "pure spin noise" and "absorbed circuit noise" in previous work, [9] the distinction between spin noise and circuit noise may be misleading and is not as clear-cut as it might seem, in particular on high-Q probes. Due to the strong coupling between the RF-circuit and the spins, one observes a collective phenomenon. The major source of noise not originating from the nuclear spins in our setup is the preamplifier noise, which has been extensively covered in our recent paper. [10] As explained there, preamplifier noise causes line shape changes in the spin noise signals and, most importantly, quantitatively explains the frequency shifts in the tuning curves observed previously. [11][12][13][14][15][16] We assured that there was no additional source of noise in the experimental setup used here, as the spin noise response is not affected by changing the cabling between the cold (77 K) pre-amplifier and the "warm" (ambient temperature) pre-amplifier including replacement of the RF-pulse cable by 50 Ohm terminators.