HEPS中文
Figure 1: Bird’s eye view of HEPS complex buildings in the summer of 2022.
Accelerator status
The year of 2022 witnessed completion of several milestones in accelerator progress. Installation and high-power conditioning of the linac [1] were completed in the autumn -(). Almost 95% of the booster accelerator components (magnet, girder, and vacuum chamber) have been put into the booster tunnel. Another major milestone is the successful completion of a mock-up of a standard cell of the HEPS storage ring (). All of the magnets in this cell are now installed and aligned, all of the vacuum chambers have been connected together and inserted into the magnets. The mock-up assembly allowed the design and installation team to identify many necessary corrections, which have been integrated into the production process [2].
Figure 2: Installed HEPS Linac accelerator.
Figure 3: One cell “mock up” for the HEPS storage ring.
As a necessary measure for the coming beam commissioning, a high-level application framework based on Python, named Python accelerator physics application set (Pyapas), was proposed and has been developed [3]. By December 2022, the high-level applications for the injector had been developed, while and others are still ongoing.
Production of the “main series” of components for the HEPS accelerator is in good shape. Storage ring magnets are being tested in batches. All of the sextupoles and octupoles have been measured with a rotating coil system and met their specifications. Over 30% of the longitudinal gradient dipoles have been measured, and the integral field is tuned to less than 1E-4 deviation. About 18% of the quadrupoles and dipole-quadrupole combined magnets have been measured. The effect of corrector coils to the quadrupole field has been thoroughly studied. Over 50% fast corrector magnets have been tested and the bandwidth measured to be 4？kHz.
The manufacture of the more than 2,000 power supplies to power the main magnets and correctors has been completed and testing results are promising [4]. The digital controllers and homemade high-precision direct current-current transducer (DCCT) have been tested and formally used in the magnet power supply. As the core components of the magnet power supply, they ensure the performance of the output current. As an auxiliary system, the magnet temperature protection system has been developed, and the software and hardware development of a high-precision current stability test system have also been completed.
The key vacuum components of the storage ring, including the vacuum chambers, RF shielding bellows, and photon absorbers, are being fabricated, and two sets of the NEG coating facilities for the circular vacuum chambers and antechambers, respectively, have been built and tested. The pre-series 166.6 MHz superconducting cavities have passed vertical performance tests at 4K (), and the first jacketed cavity is being assembled with its high-power coupler, HOM absorber, and other ancillaries for the first cryomodule [5]. For the series 499.8 MHz superconducting cavities, all passed 4K tests and largely exceeded the design goal. All of the six 5-cell normal conducting cavities have passed the high-power tests, and three of them have been installed in the booster ring for the initial commissioning of HEPS. The factory acceptance tests of the two 500 MHz 260 kW solid-state amplifiers have been passed, while the production of the 166.6 MHz 260 kW solid-state amplifiers is underway. The in-house-developed second-generation low-level RF system has passed essential lab tests and the series production has been launched. The series RF interlock system has been produced and passed the performance tests. The manufacturing of the HOM absorbers is progressing well. The challenging brazing of the absorbing ferrite blocks is being followed up very carefully by means of sophisticated qualification procedures involving ultrasonic and RF power tests.
Figure 4: 166 MHz SRF cavity and vertical test results.
A total of 45 out of 288 storage ring girders have been surveyed. Meanwhile, the survey group has finished surveying the surface network, storage ring tunnel network, and synchrotron beamline network completely [6]. The cryogenics tanks are all in place, and the helium refrigerator has been delivered on-site.
All of the insertion devices (ID) were designed by our ID team. The in-air IDs, including 4 IAUs [7] and 2 IAWs, have been manufactured and are ready for tunnel installation. Mass production of the in-vacuum IDs, including CPMUs and IVUs, is in progress. The tuning of first CPMU and IVU has been realized (), and the batch measurement is ongoing. APPLE-Knot undulator (AK), a novel undulator to mitigate high head load, is successfully integrated, and magnetic performance optimization will begin [8]. The MANGO wiggler, a novel-type wiggler for symmetric and uniform photon beam distribution, is being built.
Figure 5: CPMU (cryogenic permanent magnet undulator, left) and IVU (ambient in-vacuum undulator) under magnetic measurement.
With distributed front-end data acquisition, server virtualization, and microservices application, the accelerator control system has been established to address high complexity on diverse controlled equipment and more than 200,000 PVs. The upper layer software development is based on EPICS and programming languages (Python, PyDM, CSS). The central control room design has been completed ().
Figure 6: Virtual reality snapshot of the HEPS central control room.
Mass production and installation of beam diagnostic equipment for the linac, booster ring, low-energy and high-energy beam transport lines have been completed as scheduled. Half of the equipment for the storage ring has been manufactured, tested, and is entering the pre-installation stage. The electronics beam position monitor (BPM), HEPS developed, had already proven its high performance through operations at the Beijing electron and positron collider (BEPCII). The beam position resolution of BPM electronics with ceramic BPF and cross-linked switch version can reach a few nanometers. Independent support combining of carbon fiber composite and super-invar was tested with the first-order frequency up to 98 Hz.
The main logic backboard of the FOFB system has been manufactured, and a joint test between the fast response power supply and the fast corrector has been done. The desktop test system of the FOFB has been set up successfully. Preliminary tests, such as high-speed serial data transmission, data processing, and feedback calculation, have validated the design. Key parts of the FOFB have been developed, and auxiliary functions, such as FPGA firmware remote configuration and remote power consumption monitoring, are in progress.
With more than 3,000 trigger requirements for the HEPS timing system, the HEPS timing system can provide 6 ns coarse delay and 10 ps fine delay, respectively. The delay and phase can automatically restore in the case of power failure or system restart. The challenge for the timing system is the requirement of high-precision synchronization in a long distance of 1.5 km, and the complex logic control, including the novel swap-out injection, synchronous of booster ramping, etc. A coincidence clock method is being adopted to separate the HEPS’ three injection process, and to realize the free combination of each process and any bucket injection between the booster and the storage ring. Most hardware and software systems have completed testing.
Beamline progress
In Phase I, there are 15 beamlines, divided into groups 1 and 2. Group 1, equipped with a bending magnet or in-air undulators or wigglers as sources, is planned to start commissioning in 2024. Group 2, with CPMU, IVU, AK and MANGO wigglers as sources, will start commissioning in 2025. Feature beamlines, like coherence and nano-focusing beamlines, are in group 2. Group 1 beamlines are under construction (). Group 2 beamlines are still under experimental station design optimization. Equipment for the 12 front ends for undulator beamlines has been delivered and is awaiting installation.
Figure 7: First FOE and experimental hutch under acceptance at the submicron focusing XAFS beamline.
In the beamline optical design, we developed a wave-optics simulation based on a coherent modes decomposition and a wavefront propagation model [9]. The simulation software, Coherence Analysis Toolbox (CAT), is used in an optical design of the coherent and nanoprobe beamlines. Coherence at different optic planes can be easily calculated, and the influence of surface errors in the optics is simulated. The effect of partial coherence on CDI experiments can be analyzed. Multilayer Laue lens (MLL) is a special diffractive optics, which could provide very high numerical aperture for hard X-ray nano-focusing. We proposed a single-order focus multilayer Laue lens [10] and a novel figuring method for the multilayer Laue lens [11] for our hard X-ray nano-probe beamline. The single-order focus MLL suppresses higher-order diffractions effectively, which will substantially increase the working distance. The figuring method will reduce the requirements for MLLs; i.e., the total thickness and layer placement accuracy of the multilayer structure, the main constraints on the development of the large aperture MLLs, thus making the fabrication of larger NA MLL with a longer working distance possible.
To address the high heat load issue on the X-ray mirror, we proposed a highly efficient thermal deformation optimization method for smart-cut mirrors over the entire photon energy range [12]. By optimizing the notches of water-cooled white-beam mirrors, the RMS of the curvatures of the thermal deformation of the white-beam mirror over the entire photon energy range is minimized. This considerably simplifies design of the water-cooled white-beam mirrors.
We have successfully achieved high-accuracy metrology of 18 delivered mirrors using a Flag-type Surface Profiler (FSP) developed by our X-ray optics group. As an example, the shape error result of a long flat mirror with 620 mm is 0.35 nm. The shape error results of diffraction-limited focusing elliptic K-B mirrors are 0.33 nm and 0.39 nm. During measurement, we found a mid-frequency surface errors issue and shared it with the manufacturer to improve their processing. A spatial frequency decomposition stitching interferometer, the Flag-type stitching interferometer, has been proposed and developed (), which can provide feedback for mirror processing with sub-nanometer accuracy. A new double-edged wavefront metrology has been developed that can achieve wavefront metrology of beamline optics, including crystals and CRLs, with nano-radian precision done at the current Beijing parasitic synchrotron on -BEPCII.
Figure 8: Flag-type stitching interferometer.
Silicon crystals are widely applied in monochromator and mirror studies. Wavefront preservation, vital for beamlines in 4th-generation light sources, depends on X-ray optics component quality, including surface quality of the silicon crystal. We developed a flexible chemical mechanical polishing method in our crystal fabrication lab to fabricate the wavefront preservation silicon crystal. Our crystal lab can now provide flat silicon crystals for crystal monochromators with 60～100 nrad RMS slope error in near cm scale area. In the meantime, our engineers developed a novel and cost-effective approach for polishing silicon channel-cut optics with a gap of a few millimeters [13]. With this so-called magnetically controlled chemical–mechanical polishing method, a high-quality surface with roughness of 0.614 nm RMS is achieved in a channel-cut crystal at current 7 mm gap. X-ray topography and rocking-curve measurements proved that the stress residual layer on the crystal surface was removed. For this achievement, a pseudo channel-cut monochromator (pCCM) for our protein crystallography beamline has been upgraded to a real CCM. X-ray Raman scattering (XRS) is a key method of our hard -X-ray inelastic scattering spectroscopy beamline. A huge XRS spectrometer is being manufactured; it needs 90 pieces of spherically bent crystal analyzers to realize largest solid angle coverage. Our team developed a new fabrication method to make a bubble-free analyzer. The first batch of 15 analyzers in diameter 100？mm has been fabricated. The net energy resolution of all analyzers reached 1 eV @ 9.7 keV (FWHM) with incident bandwidth 0.8 eV, and the focusing property was also validated.
Using DC magnetron sputtering technology, ultrafast laser etching, and the focused ion beam, two-dimensional focusing MLLs with 43(H)×63(V) μm thickness were fabricated in our multilayer fabrication lab. These were designed to focus the X-ray beam to 8 nm at 10 keV for our nanoprobe beamline. The accumulated layer position error (PV) is below ±5 nm and the RMS error is about 2.9 nm. The calculated focusing properties with optical error are 8.4(H)×8.2(V) nm based on the -Takagi–Taupin description.
Most of the X-ray mirror systems and the monochromators are designed by our opto-mechanic group. The first have been manufactured and delivered recently, including a white light mirror system, a double crystal monochromator, and a channel-cut monochromator. A novel X-ray transfocator scheme based on a push-push ratchet self-locking mechanism was proposed. Compared with the typical transfocator, it adopts the horizontal and vertical orthogonal layout driving scheme like playing the piano for the simplification of the overall system. This ultra-compact and stable design has been validated in a prototype and the transfocators based on this design will be widely used in beamlines for manipulation of the X-ray compound refractive lens and transmitting optical components like filters as well.
To address the sample alignment efficiency and automation in scanning tomography measurement, especially in ptychography, we developed an automatic tomographic alignment scheme based on genetic algorithm and human-in-the-loop software with excellent sub-pixel alignment efficiency [14]. In light of the big image data processing issue, we developed HEPSCT, a GPU-based reconstruction software for X-ray CT data processing and fast CT reconstruction. It provides image processing modules such as phase- -retrieval for holotomography, and preprocessing of raw data for 2D XANES and 3D XANES analysis. Basic operators for reconstruction processing are also included in HEPSCT; e.g., ring artifacts removal, automatic rotation axis correction, and automated and manual motion correction. In addition, multiple computed tomography reconstruction algorithms are provided in HEPSCT for users to choose from, such as FBP, Grid, SIRT, EM, and CGLS. The HEPSCT advantage is that it provides big data frameworks and fast reconstruction. The core algorithms for different image processing modules achieve 100% GPU acceleration based on multi-threading to achieve fast reading and data processing with IO, CPU, and GPU at the same time. The stand-alone version of HEPSCT is available for download [15].
The HEPS beamline software group is dedicated to the development, operation, and maintenance of unified experimental control and data acquisition software framework (Mamba, ) for high-throughput, multimodal, ultrafast, in-situ and dynamic experiments at HEPS. Mamba is designed to leverage HEPS’ critical aspects of scientific experiments throughout their whole life -cycles, such as automated experimental control (Control System and Mamba backend), online data acquisition and visualization (Mamba GUI), as well as high-throughput data streaming and processing (Mamba Data Worker, MDW) [16]. Ultimately, the system is intended to expand the current usage scenarios of Mamba into broader applications nationwide, looking forward to fully unleashing the capabilities HEPS has to offer through extensive inter-institution cooperation. The group also concentrates on the study and adoption of selected cutting-edge image processing algorithms at advanced light sources, with particular interest in artificial intelligence or machine learning algorithmic approaches that could be seamlessly integrated with Mamba. To cater to the unprecedentedly enormous data volume that could reach EB scale at HEPS, the group advocates that integration of Mamba and cutting-edge algorithms would best address the big data challenge anticipated in future [17].
Figure 9: The software framework Mamba’s overall architecture, constituting core parts such as backend, frontend (GUI), MDW, and algorithmic interfacing.
Phase I beamlines will produce more than 300PB/year raw data; this huge amount of data should be stored, analyzed, and shared efficiently. HEPS Computing and Communication Group (HEPS-CC), provides the IT R&D and services for the facility and science communities, including IT infrastructure, network, computing, analysis software, data preservation and management, and public services. DOMAS (Data Organization, Management and Accessing Software stack), aimed at automating the organization, transfer, storage, distribution, and sharing of scientific data from and for HEPS experiments, has been released in version 1.0 and open sourced. DOMAS provides the features and functions for the metadata catalogue, metadata ingestor, data transfer, data web portal, many application interfaces, and a workflow engine is included to define the data management pipeline according to the requirements. Daisy (Data Analysis Integrated Software System), designed for data analysis and visualization, has been released in version 1.0. The Daisy framework has a highly modular C++/Python architecture and is composed of four pillars: algorithm, workflow, workflow engine, and data store. The architecture supporting customized plug-in functions can easily access visualization tools and the Python-based scientific computing ecosystem. The computing system is designed and deployed in three types: Openstack, Kubernetes, and Slurm. Openstack combines the virtual cloud desktop protocol to provide users with remote desktop access services, and supports users in using browsers to access windows/Linux desktop, running commercial visualization data analysis software. Kubernetes manages container clusters, and starts multiple methodological container images according to user analysis requirements to provide users with the Daisy environment. Slurm is used to support HPC computing services and meet users’ offline data analysis needs.
Utilities progress
There are four 10 kV switching stations and 11 electric power distribution substations in the HEPS campus, with a total installed power capacity of about 68 MW and operational demand of about 39 MW. They are in service now. A chilled water plant for the cooling water system and HVAC has been completed, with total cold capacity of about 33 MW, and its commissioning was finished at the end of 2022 (). HVAC of HEPS was completed and became operational at the same time. The compressed air station, deionized water makeup system, and cooling water system for the Linac and the booster have been completed and put into operation.
Figure 10: Operational cooling towers.
According to the vibration levels and beam dynamics, the stability specifications of the HEPS site are set to be less than 25 nm (displacement RMS integral over 1–100 Hz) on operation. In the meantime, the slab amplification factor cannot be bigger than one. After the tunnel and the whole HEPS building were completed, vibration measurement indicated that the displacement RMS integral over 1–100 Hz on the slab was smaller than that on the plain ground, although vibration response on the slab was bigger than on the plain ground with some individual frequencies.
Groundbreaking for HEPS started in the summer of 2019, but it was soon delayed by the Covid pandemic outbreak in 2020. The progress of HEPS has been interwoven with the pandemic. As the zero-Covid policy is lifted, progress is expected to be facilitated. At MEDSI2023, to be held in Beijing in November, we anticipate informing participants that the HEPS with storage ring assembly will be nearly finished and some beamline installations will be completed.
Source： SR News