Conceptual design of a macromolecular diffractometer for the Jülich high brilliance source

Neutron macro-molecular diffractometer (MMD) is a powerful tool capable of directly revealing the positions of hydrogen atoms within protein structures.^1,2 Unlike x-ray crystallography, which struggles to discern hydrogen atoms due to their weak scattering power, neutron crystallography offers a unique advantage in visualizing hydrogen atoms.^2,3 In recent years, artificial intelligence (AI) has made remarkable strides in predicting protein structures, exemplified by AlphaFold2.⁴ Although AI-based programs like Alpha-Fold excel in predicting overall protein folds, they face limitations in accurately determining the protonation states of amino acid residues, particularly hydrogen atoms.⁵ Moreover, the neutron datasets from those diffractometers are used to calibrate quantum mechanics/molecular mechanics (QM/MM) calculation approaches.⁶ 

Currently, available neutron macromolecular diffractometers are mostly built at spallation or reactor neutron sources by using tapered or ballistic guide systems.^7–9 Since the typical sample size for a macromolecular diffractometer is very small, typically between 0.01 and 1 mm³,^10,11 the beam from the moderator has to be transported onto a small beam spot to optimize the signal-to-noise ratio enabling the collection of high-resolution single-crystal datasets.^12,13 For these existing neutron macromolecular diffractometers, collimation systems consisting of slits and collimation tubes are installed close to the sample to tailor the beam size and divergence at the sample position.^7,8,14

As a novel approach to neutron research facilities, High-Current Accelerator-driven Neutron Sources (HiCANSs) are currently being developed for use as future neutron sources. The HBS project, designed at the Jülich Center of Neutron Science (JCNS) in Germany, presents one such facility.¹⁵ According to the current design, the linear proton accelerator will accelerate proton beams to 70 MeV for distribution toward three independent target stations of the HBS to produce neutrons. Each of these target stations receives the same average proton beam power of 100 kW and can be operated with different frequencies. The total neutron yield for a power of 100 kW is ∼10¹⁵ s⁻¹ as calculated with MCNP6.1 and the ENDF/B-VII database.¹⁶ Each target-moderator-reflector (TMR) unit is surrounded by a shielding block of about 4 m diameter. A duty cycle of 1.6% and two source repetition rates of protons at the target stations of 96 Hz (τ_proton = 167 µs) and 24 Hz (τ_proton = 667 µs) were adopted to serve the requirements for various instruments and achieve the same average power at the target. A high-density hydrogen-rich thermal moderator will be used to moderate the primary neutrons into the thermal energy range, and a cryogenic moderator will be embedded in the thermal moderator for cold neutron instruments.¹⁷ A so-called one-dimensional “finger” moderator system has been developed using liquid para-hydrogen to realize high brilliance cold neutron beams at HBS.¹⁸ The surface area of such a moderator has a diameter of 2–3 cm. These moderators feature an enhanced emission toward the surface normal of the moderator. Thus, although the primary source strength (time-averaged) is two to three decades below the source strength of research reactors, simulations from the current design show that the HBS has a higher peak brightness compared to continuous neutron sources such as BER II, ORPHEE, and a comparable time average brightness compared to spallation neutron sources.¹⁶ As the HBS source does not produce particles of more than 70 MeV energy and guide cross-sections are much smaller than on existing neutron sources, the radiation background at the sample and detector will be much lower. Due to its time structure, low background, and flexible, high-brilliance moderator setups, HBS is well suited for instruments studying small samples, such as neutron macromolecular diffractometers.

By Liouville’s theorem, the neutron brilliance at the sample can never be higher than that at the moderator surface. Therefore, it is important to maximize the brilliance transfer of the neutron beam for small samples. Neutron-reflecting guides and supermirrors make it possible to transport neutrons over a large distance. Straight and tapered neutron guides are commonly used in scattering instruments. Focusing optics such as parabolic, elliptical, and double elliptical neutron guides have also been applied for focusing neutrons to small samples.^19–22 Micro-focusing concepts that have been successfully used in x-ray optics, such as Kirkpatrick–Baez mirrors and Montel optics, have also been adopted for efficient neutron transport on small samples.^23–25 However, the limitations of these optics due to the existing of coma aberration and lack of flexibility in adjusting neutron phase space are known. In recent years, the concept of the SELENE guide was proposed to deliver neutrons to a small area with a low background.²⁶ A SELENE guide can be described as two mirror-inverted Montel guides in two dimensions, and the two Montel guides share the focal point in the medium central plane. The point-to-point focusing of the SELENE guide system offers several advantages. First, it filters out all unwanted neutrons at the beginning of the guide system. Second, it allows for precise control over the properties of the phase space at the sample position, leading to a significant reduction in background noise.²⁷ Currently, SELENE optics have been applied at the neutron reflectometer AMOR at PSI and are planned to be used in the reflectometer ESTIA at ESS.^28,29 The point-to-point focusing of the SELENE guide system offers several advantages. First, it filters out all unwanted neutrons at the beginning of the guide system. Second, it allows for precise control over the properties of the phase space at the sample position, leading to a significant reduction in background noise.²⁷ Currently, SELENE optics have been applied at the neutron reflectometer AMOR at PSI and are planned to be used in the reflectometer ESTIA at ESS.^28,29

In this work, a concept for a macro-molecular diffractometer instrument is proposed for HBS. The performance of the instrument was evaluated by using VITESS Monte Carlo simulations.³⁰ A flexible tunable neutron beam with a low background can be achieved at the mm scale sample by coupling SELENE guides with the compact high brilliance moderator considered at HBS. The instrument is expected to have comparable performance to existing instruments at medium power spallation and reactor neutron sources.

Targeting a 2 Å wide wavelength band (2 to 4 Å), the instrument length (from moderator to detector) at the 96 Hz target station is fixed at 20.6 m. In contrast, the instrument length would extend to 82.4 m at the 24 Hz target station. Considering engineering feasibility, constructing a long-distance SELENE optical instrument poses challenges.³¹ Therefore, we have selected the 96 Hz target station. The schematic outline of the macromolecular diffractometer is shown in Fig. 1. The instrument is equipped with a liquid para-hydrogen moderator placed at the 96 Hz target station of HBS, which gives a time-of-flight (tof) frame of 10.4 ms and a neutron pulse length τ_neutron = 252 µs, providing a resolution contribution of the moderator δλ ≈ 0.05 Å resulting in wavelength resolution δλ/λ between 1.45% and 2.81% across the band. Within the wavelength band used, this pulse length is largely wavelength-independent.

2D SELENE optics is used to transfer neutrons to the sample. The beamline is kinked around the source and kinked back at the sample position, i.e., the focal point of the ellipse. A slit is positioned after the moderator to define a virtual neutron source. By varying the size of the virtual source, one can adjust the dimensions of the neutron beam on the sample. The divergence of the incident beam has to be tuned to the value needed for the measurement. The beam divergence is defined by a second slit (Slit_2 in Fig. 1) set before the entry of the first part of the SELENE guide, and another (Slit_3 in Fig. 1) is set before the sample to further repress unwanted neutrons transferred by the guide. The beam definition is done far upstream of the sample by Slit_1 and Slit_2 to maximize the signal-to-noise ratio. Moreover, the instrument should allow to tailor the divergence to the unit cell size of the sample, for example, a low divergence beam is required for the sample with a large unit cell. A frame overlap chopper will be set outside the shielding block, and a wavelength band chopper will be set about halfway between the two parts of the SELENE guides. SELENE optics are sensitive to a stable alignment. It has been shown that the required mechanical stability can be achieved by state-of-the-art technology for a distance <30 m between the virtual source and the sample.^29,31

The macromolecular diffractometer requires a detector with specifications such as high efficiency in the wavelength range above 2 Å, high spatial resolution to have a sufficient angular resolution for a short sample to detector distance, which is required to cover a large solid angle, modest time resolution and count rate capabilities, small dead areas, and low sensitivity for gamma rays. The detectors will cover a range of 2θ = 10°–170°. Both, up-to-date scintillators and gas-based neutron detectors, can satisfy the spatial and temporal resolution of the MMD.³¹ In addition, a newly developed neutron imaging detector with advanced event-mode data acquisition is a potential candidate. The detector uses a ⁶LiF: ZnS scintillator (>95% ⁶Li enrichment) with an MCP image intensifier and a TPX3Cam optical camera and can achieve a high-resolution time-of-flight image with a good signal-to-noise ratio and large field of view.³² 

For the SELENE optics chosen for MMD at HBS, the maximum divergence Δθ_max is determined by the guide shape and can be estimated as $Δθmax=θmax−θmin≈2b(c−d)aa2−(c−d)2$⁠, in which θ_max and θ_min are the maximum and minimum divergence of the neutron entering the SELENE guide. b is the short axis of the SELENE guide, a is the long axis of the SELENE guide, $c=a2−b2$ is the half distance between the two focal points, and d is the distance from the focal point to the guide entry/exit. The SELENE guide is kinked with $θkink=θmax+θmin2$ around the virtual source, and the sample is kinked back with −θ_kink at the sample position. The guide system for MMD was designed to provide a beam on the sample position with a maximum divergence of Δθ = 0.7° (FWHM) in both the horizontal and vertical planes, illuminating a sample area of 1 × 1 mm².

One of the advantages of the HiCANS sources is the comparably low radiation level close to the source, which allows the first slit to be positioned at 50 cm distance from the moderator. The distance from the virtual source to the sample is 4c = 19.6 m. The length of each of the two SELENE guides is set to 6 m, the semi-minor axis b of the guide is optimized to 4.2 cm, and the guide coating is m = 3. For the geometrical optimization, we fixed the virtual source size to 1 × 1 mm², matching the phase-space requirements at the sample plane with 1 × 1 mm² sample cross section, and Δθ = 0.7° collimation, as shown in Fig. 2. The ratio of the neutrons arriving at the sample area to the total neutron transferred to the sample plane is >90%, and a clean and continuous phase space is achieved at the sample area.

The brilliance transfer of the optimized guide system is shown in Fig. 3. The brilliance transfer, which is defined as the ratio of the brilliance at the sample to that at the moderator surface, provides a quantitative measure of how well the guide system transports neutrons.³³ It shows that the optimized SELENE guide achieves a brilliance transfer between 50% and 80% in the required bandwidth range.

By changing the size of the Slit_1, the neutron beam size at the sample position can be flexibly controlled in a certain range. Figure 4 shows the neutron spatial distribution at the sample position with different sizes of Slit_1. The beam size on the sample can be adjusted flexibly from 0.25 × 0.25 mm² to 1.25 × 1.25 mm² with a uniform spatial distribution. When the virtual source size becomes larger, the neutron beam at the sample position is gradually distorted. Figure 5 shows the neutron beam intensity integrated along the horizontal axis as a function of the vertical divergence and position for different virtual source sizes with θ = 0.7°. Figure 5 also shows the distortion of the divergence when the size of the virtual source >1.5 × 1.5 mm².

The asymmetry both in the angular and position distributions can be removed by Slit_3 between the sample and the downstream mirror. It should be noted that this slit removes only the tails of the distribution, while the main absorption and, hence, background creation take place at the upstream slit_2. Both slits together can provide a low and symmetric divergence profile, which is required for the collection of diffraction data from single crystals with larger unit cells. Figure 6 shows the divergence distribution at the 1 mm² sample with different collimations θ selected by Slit_3. Figure 7 shows the neutron beam intensity integrated along the horizontal axis as a function of the vertical divergence and position for different θ at the 1 mm² sample area. It shows that the tuning of the divergence does not distort the beam spot distribution significantly.

Figure 8 shows the neutron flux at the sample area with different Δθ when chopper systems are taken into account. The simulated results were also compared with the simulated neutron flux of the MaNDi beamline at the first target station of SNS. On MMD at HBS, the calculated neutron flux through a 1 mm² sample cross section is 3.25 × 10⁷ n/s/cm² for 0.7° divergence and 1.10 × 10⁷ n/s/cm² with 0.38° divergence within 2–4 Å wavelength range, respectively. On MaNDi,⁸ the neutron flux at 1 mm² sample area is 1.3 × 10⁷ n/s/cm² with 0.38° divergence and 4.5 × 10⁷ n/s/cm² with 0.8° divergence for all neutrons between 2.1 and 4.2 Å. Namely, with the same fixed beam divergence of 0.38°, the flux calculated of the MMD at HBS is about 84.6% of on MaNDi. Since the reciprocal space resolution is dominated by wavelength resolution contribution at large scattering angles close to backscattering while dominated by the divergence at the small scattering angles, MaNDi has a much higher resolution in reciprocal space at large scattering angles for the reason that MaNDi has excellent wavelength resolution better than 0.15%. While the reciprocal space resolution of the ManDi and MMD at HBS is 7.6% and 7.96% at 2θ = 10°, respectively. The neutron flux value of the MMD at HBS is comparable with instruments at reactor neutron sources, such as BIODIFF at the FRMII which has a flux of 7.6 × 10⁶ n/s/cm² and wavelength resolution of 2.5% with 0.7° divergence for 4 Å.^34,35 Here, one has to keep in mind that BIODIFF is a monochromatic instrument and some flux of the MMD contributes more to redundancy by measuring low-resolution Bragg reflections again and again at different wavelengths throughout the given wavelength band of 2 Å.

To test the performance of the proposed MMD instrument in a realistic scenario, we used realistic structures from the protein data bank in order to simulate scattering patterns for different unit cell axes using VITESS 3.4. In the virtual experiment, the neutrons at the sample were set with 0.38° divergence within the 2–4 Å wavelength range, and the distance from sample to detector is 1 m. The mosaic spread used in the virtual experiment was zero. We assumed a crystal sample with a volume of 1 mm³ and a scattering resolution of 1.5 Å. This resolution is among the best measures for protein crystals with neutron diffraction as deposited in the protein database.³⁶ To probe, which reflections can still be distinguished in a time of flight experiment, we limit the wavelength band from 3.0 to 3.05 Å, which is the time of flight contribution to the overall resolution from the full pulse.

To simulate the diffraction from the sample, we used the VITESS module “sample_singlecryst.” This was fed with an hkl-file list, which was calculated by the program phenix³⁷ using the cns-output option. The input for each of these calculated hkl-files was the atomic structure of an HIV epitope scaffold (3ru8.pdb)³⁸ for 90 Å unit cell dimensions, human acidic chitinase (2ybt.pdb)³⁹ for 150 Å unit cell dimensions, and aldolase (1f2j.pdb)⁴⁰ for 209 Å unit cell axis. Phenix was set to output always hkl-lists in the space group p212121 in order to have only extinctions on the major axis. The simulated diffraction patterns for the different unit cell axes from the MMD at HBS onto a 4π spherical detector located 0.5 m from the sample are shown in Fig. 9.

The diffraction patterns at 90 and 150 Å unit cell size are clearly resolved by the MMD instrument. However, at 209 Å unit cell size, overlapping of Bragg reflections starts to occur, especially at phi angles of 0° and −90°. This unit cell size seems to be the limit for the MMD instrument presented here. A resolution limit of up to 200 Å unit cell axis would cover about 80% of all existing x-ray structures deposited in the pdb. The assessment of the resolution limit provided here lacks, of course, the consideration for real crystal shapes and different orientations of the crystal with respect to the beam.

In this work, a neutron diffractometer for macromolecular crystals was designed for the Jülich High Brilliance Source. By coupling SELENE guides with a compact high brilliance neutron moderator at HBS, a low background tunable neutron beam can be achieved at a small sample. With the optimized guide geometry, the neutron flux at the 1 mm² sample area is 3.25 × 10⁷ n/s/cm² for 0.7° divergence and 1.10 × 10⁷ n/s/cm² for 0.38° divergence. This will allow investigations on protein crystals with large unit cells. Virtual experiments indicate that the designed instrument is able to resolve well the diffraction pattern from the sample with a unit cell axis of up to 200 Å.

The strength of the instrument concept is the definition of the phase space far upstream of the sample area and the clean restriction of the illuminated area. Together with the prevention of the direct line of sight (inherent to the SELENE guide concept) and the reduced high energy background (owing to the lower source strength and lower particle energy), such a type of instrument offers a very high signal-to-noise ratio. This instrument concept could also be envisaged for other instruments restricted to small sample volumes, such as dedicated instruments for high-pressure or very high magnetic fields. Such instruments would also profit from the free access to the sample area as all the beam-defining elements are far upstream. This work is the first important step toward the design of macromolecular diffractometers for the Jülich High Brilliance Source. Further work on the instrument design will include modeling the effects of misalignment and waviness of SELENE optic systems.

We thank the “2019 Helmholtz – OCPC – Program for the involvement of postdocs in bilateral collaboration projects” for their financial support that enabled this important study. We thank Dr. Andreas Ostermann (TUM) for helpful discussions and for providing the McStas file of the instrument BIODIFF. We also thank Dr. Jochen Stahn, Dr. Artur Glavic, and Dr. Zamaan Raza for helpful discussions.

The authors have no conflicts to disclose.

Z. Ma: Conceptualization (lead); Investigation (lead); Software (lead); Writing – original draft (lead); Writing – review & editing (equal). K. Lieutenant: Formal analysis (equal); Software (equal); Writing – review & editing (equal). J. Voigt: Investigation (equal); Writing – review & editing (equal). T. E. Schrader: Data curation (equal); Software (equal). T. Gutberlet: Funding acquisition (lead); Project administration (equal); Resources (lead).

The data that support the findings of this study are available from the corresponding author upon reasonable request.