We successfully developed an in situ acoustic emission (AE) detection setup that allows recording of AE waveforms (triggered and streaming) and simultaneous x-ray diffraction and imaging on samples deformed at high pressure and high temperature (HPHT) conditions in the Aster-15 Large Volume Press at the synchrotron beamline station P61B. This high pressure AE detection system is a powerful tool to investigate AE phenomena from the HPHT chamber. Six commercial acoustic sensors, protected by a tungsten carbide support ring on each anvil of the same material, have excellent survivability throughout each successive experiment. By pulsing each sensor in succession, the average wave velocity through the anvils and cell assembly can be determined at any press load. The distance between the sensors is obtained by x-ray radiography and by logging the positions of each hydraulic ram. This provides a basis for accurately locating AE events in the sample. The feasibility of this AE detection setup was confirmed by compression and deformation test runs using several different self-designed AE sources in specialized assemblies. The present setup proves to be extremely efficient and accurate in measuring brittle processes in samples under HPHT. It is now available for applications for beam time and experiments without x rays at P61B. Combined with synchrotron x rays, in situ pressure, temperature, strain rate and stress, and phase changes can be monitored while recording AE activity. We provide a powerful tool to investigate the origin of earthquakes, for example, causing AE emissions due to brittle dehydration reactions or phase transformations in the Earth.

High pressure and high temperature (HPHT) generation is an important technique for synthesis of high-performance materials and simulating Earth’s interior environment.1–4 However, due to the large size of a typical Large-Volume Press (LVP) and a complex cell assembly surrounded by tungsten carbide (WC) anvils, it is difficult to access the sample under HPHT. Hence, in situ x-ray diffraction (XRD) and imaging at a synchrotron radiation facility5,6 and other in situ techniques such as electrical conductivity and acoustic emission (AE) detection have been developed. Particularly, AE detection of brittle behavior in samples under HPHT and high stress in an LVP has recently regained attention since its introduction at LVP beamlines at synchrotron facilities6,7 to investigate the origins and mechanisms of earthquakes. The addition of a new AE detection system at the beamline station P61B at PETRA III, DESY is the focus of this paper.

The first studies describing the brittle behavior of a rock sample under pressure were initially investigated with just a few transducers. Since at least four transducers are required for 3D relocation of events, investigating the origin of the AE events in the sample was not attempted for some time.8–10 In 1989, Meade and Jeanloz reported AE detection with diamond anvil cell with one transducer.11 In 2002 and 2004, Dobson et al. developed high pressure AE techniques to locate seismic sources in one dimension using two transducers but provided poor precision for 3D relocation.12,13 In 2006, Jung et al. developed a system with four transducers, which gave the possibility to carry out 3D relocation and waveform analysis.14 In 2007, de Ronde et al. reported an eight-transducer setup for multi-anvil deformation experiments. For the first time, to our knowledge, calculation of hypocenters of events in the sample assembly was performed using standard seismological algorithms.15 

Experimentally, there are several challenges for in situ AE detection and event relocation in a sample in the LVP due to the small size of the experimental assembly and many interfaces between the sample and the transducers. To accurately relocate AE in a sample one to several millimeters in size, the hypocenters should be resolved with submillimeter accuracy. Among the factors that determine the accuracy of three-dimensional hypocenter location (such as attenuation and uncertainty in the average wave speed), having more than the required number of transducers (i.e., >4) is most important for accurate relocation and ≥6 is important to carry out moment tensor inversion for focal mechanism analysis.16–19 To improve upon this, new studies would include complementary in situ x-ray data measured in the LVP at synchrotron facilities, coupled with sufficient AE sensors (6–8) to relocate AE events in a mm-sized sample. With the advantage of in situ x-ray diffraction and imaging, the phase transitions, lattice micro-strains, and the shape of the sample can be monitored while detecting AE, providing a direct explanation for the origin of AE.

The first group to combine AE detection and synchrotron radiation was that of Gasc et al. in 2011, who used a cubic-type AE system in the MAX80 press (at HASYLAB, DESY) with six transducers to relocate events and calculate the focal mechanisms. AE detection was coupled with simultaneous x-ray diffraction to estimate stress evolution in samples during cracking.16 More recently, Schubnel and co-workers20–22 followed by Ohuchi7,23 independently developed new AE detection systems combined with the D-DIA press at the APS, ANL, and SPring-8 synchrotron facilities, respectively. The D-DIA press offers the option to control deformation of the sample at constant displacement/strain rate at high pressures (<10 GPa). In 2020, Officer and Secco used six sensors in a walker-type configuration to investigate the transformational faulting in Fe2SiO4.24 In 2022, Officer et al.25 developed a double-difference relocation method for AE events using Hypo-DD.

The realization of HPHT AE detection at synchrotron facilities has become a powerful tool to investigate the origin of deep earthquakes. AE testing is used to detect elastic waves emitted from the propagation of micromillimeter-sized cracks in the kHz to MHz frequency range by piezoelectric transducers. In essence, each crack acts as a source with a corresponding moment magnitude MW. These miniature quakes in laboratory can be scaled up to earthquakes following the Gutenberg–Richter law with similar b-values between 0.5 and 1 in the frequency–moment magnitude distribution of events, implying a scale-invariant rupture process.26 Based on the source parameters extracted from the recorded waveforms, we can closely track event initiation, clustering, and propagation during HPHT deformation and transformation processes.7,20,27–29 In this way, the location, magnitude–frequency distribution, and focal mechanisms of events can be obtained in candidate materials deformed under HPHT in the LVP. Coupled with stress and strain measurements by x-ray diffraction and imaging, the mechanisms responsible for deep-focus earthquakes can thus be investigated in great detail.12,15,16,24

The objective of this paper is to present a newly developed AE detection system in the “Aster-15” 6-ram LVP at the endstation P61B, re-establishing the experimental technique at DESY. The AE detection system was previously briefly mentioned earlier by Farla et al.5 Here, we elaborate on the development and performance of this system. In order to achieve high location accuracy and moment tensor analysis, anvils with six sensors in cubic symmetry were independently designed. The wave speeds of the pressure-transmitting medium (PTM) and the positions of the six sensors were calibrated to provide a basis for accurate AE source location. In order to achieve deformation under HPHT using a quiet PTM, we calibrated a standard assembly for AE experiments. To test the reliability and accuracy of this system, different types of self-designed AE sources were used to calibrate this system with and without x rays. The present AE setup exhibits extremely high efficiency and accuracy in measuring cracks under HPHT, benefiting geoscience and materials research, with future prospects to carry out in situ AE detection at higher pressures (15 GPa), also coupled with synchrotron x rays.

Figure 1 shows a schematic representation of our AE system at P61B. The various components include a high pressure system (the LVP and high pressure assembly), high-flux white x-ray beam, x-ray diffraction detector system and radiography system, and AE detection and processing system. As shown in the schematic drawing of the AE detection setup, AE waves propagate through the PTM and are detected by piezoelectric ceramic sensors at the rear of tungsten carbide anvils. The details of this setup and results of various experiments are given below.

FIG. 1.

The schematic diagram of a portion of the MA6-6 assembly, acoustic emission system, and the x-ray beam path for imaging and diffraction at P61B. The upper left shows the tungsten anvils and supporting ring. The sensor is connected to AE amplifier and PCI-2 with coaxial cable. The lower part shows the high pressure assembly with the white x-ray beam for imaging and x-ray diffraction.

FIG. 1.

The schematic diagram of a portion of the MA6-6 assembly, acoustic emission system, and the x-ray beam path for imaging and diffraction at P61B. The upper left shows the tungsten anvils and supporting ring. The sensor is connected to AE amplifier and PCI-2 with coaxial cable. The lower part shows the high pressure assembly with the white x-ray beam for imaging and x-ray diffraction.

Close modal

A Hall-type LVP called “Aster-15” (Voggenreiter mavo LPQ-1500-1000) was used to achieve HPHT and controlled deformation. The LVP is permanently installed at the synchrotron endstation P61B at PETRA III, DESY (Hamburg, Germany). This press and the AE detection system can be operated both with and without synchrotron x rays. “Aster-15” is a cubic anvil apparatus with 6 independent hydraulic rams that can provide a combined load of 15 MN (5 MN per axis). As described by Farla et al.,5 the LVP supports various modes of compression for various assemblies. Here, mode 2 is typically used for anisotropic compression, e.g., rock deformation experiments, using the side ram No. 3 as the master ram to set the pressure profile to a target press load. The top and bottom rams can thus be controlled with constant displacement rate steps to deform a sample in the vertical direction perpendicular to the x-ray beam. Note, the oil pressures in all rams except the master ram are not controlled. However, the positions of all other rams must follow the position of the master ram with additional corrections for elastic frame bending toward higher press loads and, in the case of anisotropic compression, the additional displacement of the independently moving rams. The “MA6-6” assembly comprises six second-stage tungsten carbide anvils compressed by six steel first-stage anvils in the LVP and supports both isotropic and anisotropic compression modes, whereas the MA6-8 assembly featuring eight second-stage tungsten carbide anvils only supports isotropic compression but can be used to reach ultrahigh (30 GPa) pressures. For controlled deformation experiments described in this work, we only use the MA6-6 assembly.

Currently, the following anvils are available and compatible with the AE setup: 15 mm TEL (0.1–2 GPa), 12 mm TEL (0.5–4 GPa), and 9 mm TEL (up to 7 GPa), where TEL = truncation edge-length and the pressure range is given in brackets. At the moment, only one x-ray transparent anvil (c-BN) is available for in situ x-ray diffraction with a TEL of 12 mm [Fig. 2(c)]. Furthermore, a third stage of smaller WC and sintered diamond anvils can be tentatively used with AE detection; however, it was not tested until now. The third-stage assembly permits access to higher pressures with smaller TELs of 5 mm (2–10 GPa), 4 mm (up to 15 GPa), and 2.5 mm (up to 18 GPa) although tolerable stress is unknown. Using the MA6-6 compression geometry, the sample in a cubic PTM [Fig. 2(a)] can be directly subjected to a differential stress from the advancement of a pair of opposed hydraulic rams in the LVP. The maximum load that the anvils can withstand is greatly reduced by a WC support ring behind each anvil that protects the piezoelectric sensor for AE detection. The maximum load, based on the shape of the WC support ring with 35 mm outer diameter and 12 mm inner diameter, is calculated as 140 bars (1.13 MN per axis) for the steel first-stage anvils in the LVP with a hardness of 62 + 1 HRC. In fact, the steel first-stage anvils should be able to take a pressure of ∼2 GPa although we assume ∼1.7 GPa as the safe limit. The other unknown weakness is the slit in the WC ring for the sensor cable. Until now, no support ring was broken, as long as the surfaces are routinely cleaned. We use copper foils (0.1 mm thickness) to protect the anvil–ring–anvil surfaces.

FIG. 2.

Photographs of the HPHT assembly used to conduct AE experiments in this study. (a) The cell assembly includes components such as the boron–epoxy gasket, zirconia sleeve, crushable alumina piston, h-BN capsule, sample (glass beads, glass rod, or other sample), graphite heater, dense alumina piston, and cubic boron–epoxy PTM from left to right. The electrodes are not shown here and can be seen in Fig. S4. The schematic drawing of the assembly is shown below. (b) The brass supporting ring and tungsten carbide anvil and anvils after assembly (left). The anvils have a truncation of 12 mm. Six frames and anvils with gaskets after assembly (right). (c) The side view (left) and top view (right) of the whole assembly including the alignment frame and anvils. (d) A photo of the LVP with amplifiers on each ram and the whole assembly under press load.

FIG. 2.

Photographs of the HPHT assembly used to conduct AE experiments in this study. (a) The cell assembly includes components such as the boron–epoxy gasket, zirconia sleeve, crushable alumina piston, h-BN capsule, sample (glass beads, glass rod, or other sample), graphite heater, dense alumina piston, and cubic boron–epoxy PTM from left to right. The electrodes are not shown here and can be seen in Fig. S4. The schematic drawing of the assembly is shown below. (b) The brass supporting ring and tungsten carbide anvil and anvils after assembly (left). The anvils have a truncation of 12 mm. Six frames and anvils with gaskets after assembly (right). (c) The side view (left) and top view (right) of the whole assembly including the alignment frame and anvils. (d) A photo of the LVP with amplifiers on each ram and the whole assembly under press load.

Close modal

After several controlled deformation experiments, e.g., with TEL = 12 mm anvils in this study, it is possible to predict the PTM cube size and gasket thickness between each pair of anvils for any press load and, additionally, for the advancement of a pair of rams during anisotropic compression (Fig. S1). This is calculated using the logged displacement data of the rams from the press control software. In brief, while it is not possible to visually see the gap close between the first-stage and second-stage anvils, it is possible to reliably estimate at which ram position contact is made between all anvils and the cubic pressure medium of known initial size, e.g., 16 mm. The point where contact occurs is defined by the first intersection of the set point oil pressure and the actual oil pressure of the master ram (No. 3) and its position (the reference point) because the oil pressure of the master ram only increases when the anvils touch and start to compress the cubic PTM. Per experiment, the calculation of the cube size and gasket thickness as a function of press load using the reference point and the displacement data of the rams includes a correction for the small contribution of frame bending. This method of estimating the cube size under pressure was confirmed by imaging the anvil-to-anvil distance using x-ray radiography. It is therefore possible, also without using synchrotron x rays, to determine the distance between the AE sensors on the anvils at any press load during an experiment, which may improve the accuracy of AE event relocation.

In situ x-ray techniques offer crucial complementary information to an AE deformation experiment (Fig. 1). Imaging is done using x-ray absorption-contrast radiography. The white-beam x-ray microscope at P61B captures the white beam that passed through the sample. The maximum beam size at P61B is 2.2 mm (h) × 1.6 mm (v). A small fraction of the polychromatic x rays are converted into visible light rays using a GGG:Eu scintillator, magnified by a 5× or 10× objective and observed by the PCO.edge 5.5 MP sCMOS camera. As mentioned earlier, just like the anvil-to-anvil distance can be imaged, the length of sample along axial direction can be determined with μm precision during the various stages of a typical AE deformation experiment (i.e., for the cold compression, heating, and deformation). Because the beam size is usually smaller than sample length, scripted up/down movement of the press Z-stage can be performed. For example, the number of frames needed to cover a sample of about 5 mm length is about 9, depending on percentage of overlap chosen (e.g., 30%) and incident slit opening (here, 1.2 × 0.9 mm2). This opening is less than the maximum beam size in order to capture only the most homogeneous part of the beam. In addition, x rays only pass through the vertical anvil gap (see gasket thickness between the side rams in Fig. S1). Exposure time can be as low as 0.1 s, or as high as 1 s, depending on the absorption of the beam in the cell assembly and sample. Press travel time is 1–2 s for each increment and dead time is insignificant. Processing time to obtain a montage of all images is under a second. Therefore, in this study each Z-stage frame scan took up to half a minute. The montage of frames can be further processed to estimate the length changes, i.e., strain history and strain rate of the large sample.

The sample image data may also give information regarding the onset of cracking in the sample, which may coincide with the onset of AE production. It is known that mode-II and mode-III sliding and tearing cracks are generally not visible under pressure by absorption-contrast radiography. There is promise this will work using phase-contrast imaging, which is better exploited with a high-coherence and low-divergence undulator source and an x-ray microscope closer to the sample.30 Alternatively, the experimenter could add multiple thin metal foil layers in the sample and observe crack propagation through them, offsetting the layers. Furthermore, mode-I opening cracks/interfaces can be imaged directly using radiography but only at none to very low confining pressures.

Diffraction patterns, in the 30–160 keV range, can be collected during the different experiment stages using two high-purity germanium solid state detectors (Ge-SSD) by Mirion (Canberra) (Fig. 1). One detector is tilted at a given scattering angle (2θ) in the vertical plane (azimuth angle = 0°) to collect the diffracted x rays from the anvil gap (for MA6-6). A second detector is positioned at the same scattering angle but in the horizontal orientation (azimuth angle = 90°). To receive the diffracted x rays from the sample at 90° azimuth, one of WC anvils must be replaced by an x-ray transparent anvil, such as c-BN. One Ge-SSD is sufficient if only phase transformation and nucleation are monitored during the HPHT deformation experiment. Particularly, one may be interested in embrittlement processes resulting from phase transformations or dehydration/decomposition of hydrous phases. If the onset of AE production matches the observed change in the sample by x-ray diffraction, a scientific case can be made.

Additionally, the detector can be used to capture diffraction patterns of pressure–temperature (PT) calibrant to estimate the PT simultaneously in situ. This is sometimes preferred over using a thermocouple, which may disturb the stress-state of the sample and may increase the risk of a blowout at high pressures. Using a PT calibrant is a common and well-known practice,31 which offers a pressure resolution of 0.1 GPa (or better) and an uncertainty on the temperature as low as 50 °C, approaching the probable thermal gradient in the sample. Such a calibrant is typically a powder mixture of a highly compressible halide and a metal with large thermal pressure, with a mixing ratio such that both materials show reflections of similar intensities. One of the most compressible halides with a high melting point at high pressures is known to be CsCl, and a metal with the highest thermal pressure (α · K, volume thermal expansion times the isothermal bulk modulus) is known to be Pt. Such a combination would therefore make an ideal PT calibrant, at least before anisotropic compression (i.e., application of a deviatoric stress).

A second Ge-SSD can be used in combination with the other one to estimate stress in the sample (or piston) from changes in the lattice micro-strain recorded in the diffraction patterns from both detectors at 0° and 90° azimuth.5 This is only possible for controlled axisymmetric compression due to the lack of azimuthal detector coverage until a monochromator and area detector are installed at P61B. Notwithstanding, assuming the stress distribution in the sample is not too heterogeneous before embrittlement, a sudden drop in stress caused by the onset of a major crack can be estimated semiquantitatively during deformation by in situ XRD techniques and correlated to a burst of AE activity.20 

Although this study does not expand its scope toward employing energy-dispersive x-ray diffraction (ED-XRD) techniques and the use of geological samples, which may be candidates responsible for earthquakes in the Earth, we point out these techniques because it is already possible to carry out such experimental research at P61B.

For an AE experiment in the LVP, we are mainly concerned with the need to design a cool, low-noise assembly. The assembly requires good thermal insulation so that the sensors attached to the back of the anvils do not significantly heat up beyond 100 °C while the sample is at much higher temperature.

We pressed boron–epoxy cubes (amorphous boron powder and epoxy, 4:1 by weight) with 16 mm edge-length as PTM. A boron and epoxy ratio by weight of 8:1 was also attempted. However, due to low epoxy ratio, the cube becomes too brittle, generating many cracks (and AE) during an experiment. Here, we used 12 mm TEL anvils in combination with a 16 × 16 × 16 mm3 pressure medium with a maximum sample size of 5 mm in length and 2.5 mm in diameter [Fig. 2(a)]. The available pressure range for this anvil/cell assembly is 0.5–4 GPa. In principle, other combinations of cube size and anvil TEL are also available to access higher pressures.

Electrical resistive heating is performed using a graphite tube furnace inside the assembly [Fig. 2(a)]. Six anvils, each with sensor and support ring, are inserted into an aluminum alignment frame [Figs. 2(b), 2(c), S2(c), and S2(d)] to compress a cubic cell assembly [Fig. 2(d)]. Heat was generated with a DC power supply that can be controlled both manually and automatically. A traditional assembly uses 1 mm diameter Mo electrodes in ZrO2 insulation. However, this rod acts as a piston during controlled deformation and will easily punch through a molybdenum foil connected to the graphite heater, quenching the experiment. In order to avoid this problem, we designed a new kind of electrode shown in Fig. S3(a). We machined a star-shaped molybdenum foil and then wrapped it around the zirconia as shown in the Fig. S3(b). In this way, abrupt quenching can be avoided during controlled HPHT deformation. In addition, this electrode design further improves thermal insulation.

The samples in this study are typically all made of fused silica glass of various shapes and geometries. These samples were purposely used because they easily crack under an applied stress at pressure and therefore yield AE bursts that outline their shape. The simplest AE sources we tested are glassy beads of two different sizes, small (<100 μm) and large (>200 μm), wrapped in Ni foil. The shape of the capsules are small cylinders of a few mm in size. In addition, we used samples comprised of two wedge-shaped pistons of fused silica glass to restrict the origin of the AEs. In various experiments, the sliding surfaces were either grooved or smooth both with and without glassy beads glued on. We also changed the material of the pistons to alumina and solid molybdenum with glued-on glassy beads to make sure that AEs only come from the sliding interface between the pistons where the beads are crushed. In Table I, we report results from the most successful experiments.

TABLE I.

Experimental runs and conditions.

Run no.Load (MN)SampleActive sensorsPTMAE hitsaAE eventsaStrain (%)Disp. rate (μm/min)b
HH259 0.7 Glass beads (<100 μm) Pyrophyllite Yes/noise None 30 36.7 
HH264 1.1 Glass beads (>200 μm) Pyrophyllite Yes/noise None 30 18.3 
HH287 1.3 Glass beads (>200 μm) B-epoxy >60 000 4221 15 10 
HH289 1.7 Glass beads (<100 μm) B-epoxy >130 000 6800 15 10 
HH290 1.7 Glass beads (>200 μm) B-epoxy Yes/noise None ⋯ ⋯ 
HH292 1.1 Glass beads (<100 μm) B-epoxy >7 000 40 ⋯ ⋯ 
HH294 1.3 Fused glass wedge (FGW) B-epoxy >54 000 5118|1892 Shear 10 
HH295 1.3 FGW + h-BN B-epoxy 493 2|3 Shear 10 
HH308 1.3 FGW + beads (<100 μm) B-epoxy >24 000 3691 Shear 10 
HH321 1.3 Dense Al2O3 wedge B-epoxy >7 000 441 Shear 10 
HH367 1.3 Two layers glass beads B-epoxy >47 000 5655|133 n/a 10 
HH411 0.7 Mo piston + glass beads B-epoxy >1 000 88|71 Shear 10 
HH413 1.3 ⋯ B-epoxy 341 None 10 
HH420 1.3 ⋯ Pyrophyllite 616 None 10 
BT540 1.2 Glass beads (>200 μm) B-epoxy >94 000 5351 20 10 
BT681c 1.1 FGW and glassy beads B-epoxy >32 000 1760|483 33 
Run no.Load (MN)SampleActive sensorsPTMAE hitsaAE eventsaStrain (%)Disp. rate (μm/min)b
HH259 0.7 Glass beads (<100 μm) Pyrophyllite Yes/noise None 30 36.7 
HH264 1.1 Glass beads (>200 μm) Pyrophyllite Yes/noise None 30 18.3 
HH287 1.3 Glass beads (>200 μm) B-epoxy >60 000 4221 15 10 
HH289 1.7 Glass beads (<100 μm) B-epoxy >130 000 6800 15 10 
HH290 1.7 Glass beads (>200 μm) B-epoxy Yes/noise None ⋯ ⋯ 
HH292 1.1 Glass beads (<100 μm) B-epoxy >7 000 40 ⋯ ⋯ 
HH294 1.3 Fused glass wedge (FGW) B-epoxy >54 000 5118|1892 Shear 10 
HH295 1.3 FGW + h-BN B-epoxy 493 2|3 Shear 10 
HH308 1.3 FGW + beads (<100 μm) B-epoxy >24 000 3691 Shear 10 
HH321 1.3 Dense Al2O3 wedge B-epoxy >7 000 441 Shear 10 
HH367 1.3 Two layers glass beads B-epoxy >47 000 5655|133 n/a 10 
HH411 0.7 Mo piston + glass beads B-epoxy >1 000 88|71 Shear 10 
HH413 1.3 ⋯ B-epoxy 341 None 10 
HH420 1.3 ⋯ Pyrophyllite 616 None 10 
BT540 1.2 Glass beads (>200 μm) B-epoxy >94 000 5351 20 10 
BT681c 1.1 FGW and glassy beads B-epoxy >32 000 1760|483 33 
a

If two numbers are shown, then AE for compression and deformation applies; otherwise, it applies only for compression.

b

Per ram (×2 for opposing rams). General duration is 45–60 min.

c

In situ XRD AE experiment with 1 c-BN anvil.

To generate high stress in the sample, we placed dense and crushable Al2O3 (alumina) pistons on both ends of the sample. The sample in the assembly can be deformed with up to eight pre-programmed steps of displacement rates and duration for one or two pairs of opposite rams. They can be programmed to move in and out as well. The pressure–load relationship was previously calibrated at room temperature by the abrupt electrical conductivity changes induced by the Bi phase transitions (2.55, 2.7 and 7.7 GPa).32 AE signals were recorded during each experiment, including the cold compression, heating, deformation, quenching, and decompression stages. During the cold compression procedure, especially pre-pressing, many AE signals were recorded that originate from the crushing of the boron–epoxy cube and crushing of the gaskets. When oil pressure in the master ram is higher than 30 bars (16/12 assembly), no further AE signals were detected, demonstrating that our AE assembly is quiet under HPHT conditions. This feature ensures that the recorded AE signals during holding and deformation procedure did not originate from compaction of the HP assembly. According to previous reports of Gasc et al.,16 except for the crushable alumina, deformation of the assembly materials will not produce a significant amount of AE. Hence, most if not all, AE detected should come from the brittle sample inside.

The AE detection system for in situ studies of fracture and brittle behavior of materials at elevated pressures (and temperatures) was installed and commissioned in 2020 (Fig. 1). Below is a description of the purchased hardware and software provided by GMA/MISTRAS (Physical Acoustics Corporation, PAC in the USA).

This AE detection system operates with the following specifications. A six-channel system with 3× PCI-2 boards is built into a personal computer (PC) system. The bandwidth range is 0.1–3 MHz with a sampling rate up to 40 MS/s (25 ns interval) and a resolution of 18-bit A/D. AEwin software for six-channel data acquisition replays function with waveform recording and 3D location algorithm included [first threshold crossing (FTC)]. In addition, there are 6× 20/40/60 dB preamplifiers with internal filters mounted inside the LVP, which amplify the original signals from the sensors. Due to obvious reasons of radiation damage to electronics in the experimental hutch, we need 30 m long low-attenuation coaxial cables, which lead from the experimental hutch to the PC in the control hutch. Such long cables do not pose a problem. Each amplified signal from the sensors is processed by the three PCI-2 boards in the PC where analog to digital conversion takes place. There is a possibility to set up additional analog and front-end filters (Table S1). As mentioned, for experiments the preamplifiers are mounted on the insides of the hydraulic rams [Fig. 2(d)], with No. 1 and No. 2 covered with 3 mm thick lead plating to avoid damage by scattered x-ray radiation. The AEwin software allows data acquisition and real-time analysis of waveforms. Furthermore, the AEwin software automatically calculates all AE features of the waveform of each hit, which include MARSE (Measured Area under the Rectified Signal Envelope) energy, counts, duration, amplitude, peak and average frequency, and so on [Fig. 3(b)].

FIG. 3.

The main parameters that were used to determine the features of a waveform. (a) Three-dimensional location schematic picture. (b) The threshold based hit detection method to detect a transient hit and features used to identify a hit, where t1 denotes rise-time (from threshold to peak amplitude), t2 denotes peak definition time (PDT), t3 means duration time, t4 denotes hit definition time (HDT), and t5 denotes high lockout time (HLT). (c) The illustration of anvils and sensors location. (d) Three-dimensional location of six sensors (s1–s6), pressure medium (red box), and sample (yellow cylinder).

FIG. 3.

The main parameters that were used to determine the features of a waveform. (a) Three-dimensional location schematic picture. (b) The threshold based hit detection method to detect a transient hit and features used to identify a hit, where t1 denotes rise-time (from threshold to peak amplitude), t2 denotes peak definition time (PDT), t3 means duration time, t4 denotes hit definition time (HDT), and t5 denotes high lockout time (HLT). (c) The illustration of anvils and sensors location. (d) Three-dimensional location of six sensors (s1–s6), pressure medium (red box), and sample (yellow cylinder).

Close modal

In each channel, a threshold above the background noise is critical to ensure no signals are missed while each sensor does not trigger on the noise either. Tests have shown that the Aster-15 LVP generates surprisingly low-amplitude noise when operating. The AEwin software, monitoring the AE sensors, could be reliably configured to trigger with a first threshold crossing (FTC) at 26–30 dB using 40 dB gain preamplifiers. In other words, according to the relationship dBAE = 20 log(U/Ur) − P, where U is the threshold in volt, Ur (10−6 V) is the reference voltage, and P is the preamplifier gain (dB), the first threshold crossing using the AE setup at P61B can be as low as 0.002 V. Note, 60 dB amplification is also available for experiments that require more sensitivity. Although the noise is also amplified, in this case the FTC can be as low as 21 dB.

We exclusively used trigger-based AE hit acquisition to collect the waveform data on each channel (i.e., from each sensor). Continuous waveform streaming is also supported but lowers the maximum sampling rate on the other triggered channels. Since the system continuously buffers the acoustic signal, pre-trigger data of the waveforms before the first threshold crossing are also recorded per hit. AEwin can be used to constrain events (collection of six nearly simultaneous hits) and then be used to export the associated waveforms for further analysis. Alternatively, all raw hit data can be exported for data analysis from scratch, including event recognition and waveform processing to obtain AE characteristics of each hit. In this study, we permit AEwin to successfully recognize events from the six channels. However, we show that for more accurate event source relocation, better determination of the first P-wave arrivals is critical. Therefore, we reprocessed the waveforms of the exported events. More advanced crack mechanism analysis is possible when the sensitivity of each sensor is precisely calibrated (Fig. S4). Once the amplitudes of the sensors are calibrated, information of the integrated area and sign (+ or −) of the first arrivals can be used to carry out moment tensor inversion analysis to obtain the focal mechanisms. Even so, with a limited number of 6 sensors, moment tensor inversion is underdetermined and any result should be scrutinized with utmost caution.

Broad-bandwidth, high-frequency sensors, Micro-200HF, were purchased from the same company the AE system is purchased from (GMA/MISTRAS) and are chosen for the calibration experiments in this study. The Micro-200HF sensor has a good frequency response in the range 500–4000 kHz with a resonant frequency around 2500 kHz (Fig. S4). An ultrasonic broadband sensor is suitable here to characterize the possible range of frequencies generated by fractures propagating over distances ranging from micrometer to millimeter (typical range for grain size to sample sizes). Small size (9.5 mm in diameter and 11 mm in height) and wide temperature range (−65 to 177 °C) make this sensor an ideal candidate for the current application to deform samples at HPHT in the LVP. While expensive, these sensors are reusable. However, several stages of development were conducted to ensure the sensor on the back of each anvil does not break or detach during an experiment. One of the main challenges is the survival of all sensors throughout successive experiments, particularly during beam time. The replacement of a sensor on an anvil is time-consuming and each new sensor should first be calibrated before use.

The AE setup was optimized during the course of this study. In principle, the six sensors need to be protected from anvil compression by slightly taller tungsten carbide supporting rings (12 mm) behind the 35 mm diameter tungsten carbide anvils, with the sensor in the middle. For the first set of experiments until HH411, shown in Table I, the assembly was continuously improved based on an initial design shown in Figs. S2(a), S2(c), and S2(e). The optimized design, as confirmed by a high sensor survival rate after experiment HH411 and subsequent experiments [Table I, Figs. 2, S2(b), S2(d), and S2(f)], comprises a thin foil of Cu separating the support ring from WC anvil, which are held together by a surrounding brass ring with small screws in the corners. Each WC support ring is grooved to feed through the cable that connects the sensor to its amplifier. Each of the six sensors is attached to the back of an anvil using a special high temperature (200 °C) resistant adhesive (cyanoacrylate), which ensures that each sensor does not detach from its anvil during an HPHT experiment. Finally, all anvils with attached sensors are self-aligned in a frame machined from a single block of aluminum [Fig. 2(c)]. Since the sensitivity of the AE sensors must be similar, the sensitivities are checked before each experiment to ensure AE can be detected consistently. In addition, a full calibration can also be performed as shown in Fig. S4.

The locations of the six AE sensors are symmetric with respect to the center of sample, which constraints the location of AE source. Consequently, when the hypocenters of cracks occur closer to the center (0, 0, 0), the first P-wave arrival times will be very similar. The three-dimensional sensor arrangement allows 3D relocation of AE sources from the time difference, Δt, between the arrivals at each opposite sensor [Fig. 3(a)]. AEwin can do all the hit detection by first threshold crossing, AE, and event characterization in real time. Following an experiment, the data can be replayed in AEwin to analyze them in detail and can be exported for further processing, e.g., using MATLAB.

We briefly explain the operation of AEwin for 3D event relocation. For more information, please refer to the software manual provided by GMA/MISTRAS or PAC. AEwin requires some initial information on the positions of the sensors as well as the average wave speed of the AE through the medium (in our case, the WC anvil and B-epoxy PTM). The AE sensors are labeled with S1–S6 as shown in Figs. 3(c) and 3(d), corresponding to the ram numbering of Aster-15. For a typical experiment with a B-epoxy cube size of 15 mm under pressure (see Table I, Fig. S1) and anvil length of 37 mm, the sensor x, y, z coordinates in mm are S1 (0, 0, 44.5), S2 (0, 0, −44.5), S3 (44.5, 0, 0), S4 (−44.5, 0, 0), S5 (0, 44.5, 0), and S6 (0, −44.5, 0). Furthermore, the average wave speed for the 16/12 assembly with B-epoxy PTM was determined as 5522 m/s (Fig. S5) based on sensor pulse tests carried out under load for the experiments listed in Table I. The average wave speed can vary somewhat depending on the press load/cube size and components/sample used in the assembly, which means pulse tests should be carried out in every experiment. Since the triggered waveforms can be exported for reanalysis, manually changing the sensor positions in the software is not needed when increasing the load or starting controlled deformation by advancement of a pair of rams.

Fundamentally, the theory of AE location is a simple time–distance relationship given by the average velocity of sound waves. Each arrival time difference implies a difference in distance between two opposite pairs of sensors. The 3D relocation of AE events using multiple sensors (minimum requirement is 4) is calculated in AEwin using a modified multi-regression analysis algorithm, which is not revealed. However, the manual describes the basic computation as follows:

Δti,j=xixs2+yiys2+zizs2xjxs2+yjys2+zjzs2/v,χ2=i=16Δti,obsΔti,calc2,
(1)

where xi, yi, and zi are the coordinates of sensor i, xs, ys, and zs are the source position, ∆t is the difference in the arrival time of the signal on sensor i and sensor j. For sub-mm accuracy on the localization of events in 3D space, the survival of all six sensors is very important. Using five sensors may still give decent results, but they are less optimal. Using the minimum required number of four sensors does not permit accurate 3D relocation of events.

This study uses a MATLAB script with an example given in the supplementary material for reprocessing the triggered waveforms of events exported from AEwin. We also export the AE characteristics of each hit from each event [MARSE energy, number of counts, amplitude, (peak) frequency, and so forth] from AEwin. Hence, each event can be associated with a particular AE feature. The MATLAB script combines these exported data to reprocess and plot the relocated events and their AE features for each experiment. The MATLAB script borrows 3D relocation code from the work of Li et al.,33,34 a small script called “tradloc3D.m.” The function is expressed as follows:

function[x,y,z,t,e]=tradLoc3D(arrivalArray,v,sensors,distError,×minChannels,nChannels,initXYZ),

where arrivalArray accepts multiple arrival times as rows in an array, v is the average (isotropic) velocity, sensors is a matrix of the sensor locations, distError is a filter that has units same as sensors and velocity, minChannel and nChannels are the min number of channels allowed for localization and the number of channel available, and initXYZ is the initial guess of the event location for the minimization algorithm, and set at the origin 0,0,0. The script uses a MATLAB minimization algorithm called “fminsearch” to solve Eq. (1) for each event. Note, the MATLAB “fminsearch” algorithm may be too aggressive using the default parameters, such as too many iterations and too high termination tolerances. In other words, it can fail to recognize when it has converged and can continue futile iterations, leading to artifacts such as clustering. All data presented in this paper are reprocessed by the script with the following conditions for the MATLAB fminsearch solver: TolX: 2.45 × 10−4, MaxIter: 20, and MaxFunEvals: 1000.

In summary, we present an optimized AE detection system, a calibrated assembly, and AE data processing routines to support new and experienced users at the state-of-the-art ED-XRD beamline station P61B LVP. The assembly offers excellent reliability of all six sensors working during successive experiments. Furthermore, the commercial sensors can be recalibrated in-house after being glued on the back of anvils (Fig. S4), which is critical for fast 3D relocation with sub-mm precision without resorting to sophisticated cross correlation algorithms (e.g., Hypo-DD).

Briefly, Hypo-DD is an advanced seismological technique for very precise sub-mm relocation of events that requires there to be only a few main events (i.e., cracks) with many associated micro- to nano-cracks in rock samples, just like with real earthquakes sequences. The high accuracy of event relocation is obtained by cross-correlating each “lower amplitude” AE event with a main event. Since in this study the samples are made of a collection of glassy beads, each crack of a bead is a main event. Thus, we did not attempt to use such an algorithm in this study. Hypo-DD requires knowledge from an experienced seismologist to decide what parameters go into cross correlation of the waveforms, and so implementation can be complicated and time-consuming.25 

The internal temperatures at the sample position in two candidate assemblies were calibrated with a C-type (W90Re5–W74Re26) thermocouple. The power–temperature (P-T) relationships of these assemblies with a graphite heater of 6 and 4 mm outer diameter are shown in Fig. 4(a). A second thermocouple monitored the anvil temperature in both cases [Fig. 4(b)]. At a sample temperature of only 500 °C, the assembly with a large-diameter heater raised the anvil temperature to over 100 °C. In several test cases, the cyanoacrylate between the sensor and anvil softened too much, resulting in the detachment of the sensors from the anvils. Furthermore, it is well known that high temperatures cause the piezo-sensors to lose their sensitivity. We reduced the heater diameter to 4 mm and obtained a satisfactory anvil temperature (<80 °C) for a desired sample temperature of up to 800 °C without losing a sensor. This successful assembly is presented in Fig. 2(a). Based on several in situ experiments using this assembly design with a NaCl–Ni mixture as pressure–temperature marker, the estimated uncertainty of the sample temperature at 500–800 °C is estimated to be around ±50 °C.

FIG. 4.

(a) The chamber temperature as function of power for assembly with graphite heater of 6 and 4 mm. The heater wall thickness is 0.5 mm in both cases. (b) The evolution of the anvil temperature over time when the sample in the assembly is heated to temperatures of 500 and 800 °C.

FIG. 4.

(a) The chamber temperature as function of power for assembly with graphite heater of 6 and 4 mm. The heater wall thickness is 0.5 mm in both cases. (b) The evolution of the anvil temperature over time when the sample in the assembly is heated to temperatures of 500 and 800 °C.

Close modal

A major issue in studying AE at high pressure and temperature is the noise of the whole system, especially the crushing of the high pressure assembly. To achieve low AE rates and generally a quiet PTM, we tested pre-fired pyrophyllite and boron–epoxy (with an amorphous boron to epoxy ratio of 4:1) as PTM. A typical AE deformation experiment (Table I) consists of three different stages: 1. compression to target pressure (with rate about 12.5 MPa/min), 2. deformation (typically at −10 µm/min per hydraulic ram, results in ∼2 × 10−5 s−1 in a 4 mm sample), and 3. decompression to room conditions (with rate about 6.25 MPa/min).

The different behavior of boron–epoxy (HH413) and pyrophyllite (HH420) was compared as a function of oil pressure and time (Fig. 5). During the start of compression up to about 30 bars oil pressure, and at the end of decompression, we recorded many AE events. The AEs obtained during the initial compression stage come from the initial crushing of the cube and gaskets, whereas the AEs at the end of decompression come from the breaking apart of the cube due to stress release. We can see that the boron–epoxy PTM and preformed gaskets are somewhat quieter than pyrophyllite (Fig. 5, Table I). In addition, boron–epoxy cubes are extremely quiet during the deformation procedure, which ensures that many, if not most, AE events during deformation must come from the sample. Based on the above result, and the fact that boron–epoxy is much more x-ray transparent, we chose boron–epoxy (4:1) as the PTM for all test experiments (Table I).

FIG. 5.

The logarithm of acoustic emission hit rates, oil pressure, and displacement as a function of time during the typical experiment procedure with boron–epoxy and pyrophyllite as PTM.

FIG. 5.

The logarithm of acoustic emission hit rates, oil pressure, and displacement as a function of time during the typical experiment procedure with boron–epoxy and pyrophyllite as PTM.

Close modal

The noise of the LVP Aster-15 was also monitored. As previously discussed, the FTC for triggered AE detection can be set to a low value between 26 and 30 dB (0.002–0.003 V with 40 dB preamplifier gain). This is an exceptionally good result in comparison to previous studies (e.g., Gasc et al., Officer, and Secco16,24). An example of the background noise level in the LVP at high pressure load is given in Fig. 6. Notably, the FTC at 0.002 V is shown in the magnified portion of the waveforms for all six sensors. While the FTC method does not truly pick the first P-wave arrival as seismic algorithms do, the accuracy is extremely high. For the event shown in Fig. 6, the arrival times are nearly simultaneous on all six sensors, which suggests the crack must have taken place close to the sample center (i.e., at the center of the assembly). Furthermore, the waveforms here are interpolated with 2 ns interval from the original data with 25 ns interval (40 MS/s) prior to P-wave arrival picking (Table S1). We believe reprocessing the AE data this way offers the best possible accuracy using traditional 3D event relocation methods.

FIG. 6.

Typical trigger-based waveforms of six sensors over the entire triggering window (left). On the middle and right side, the magnified portions of the waveforms are shown with indication for the first P-wave arrival time (solid green vertical line) and first amplitude (dash red vertical line).

FIG. 6.

Typical trigger-based waveforms of six sensors over the entire triggering window (left). On the middle and right side, the magnified portions of the waveforms are shown with indication for the first P-wave arrival time (solid green vertical line) and first amplitude (dash red vertical line).

Close modal

In order to test the accuracy of our AE detection setup, several experiments were run with different self-designed AE sources, which are tabulated in Table I. For most tests, the compression, deformation, and decompression stages were carried out, and AE events were monitored during each stage. The locations of events were determined by our own MATLAB script (see supplementary material) based on the different arrival times obtained by the FTC method, selected just above the noise (e.g., at 26 dB). The AE hypocenters of all experiments are shown in Fig. 7. The circle size indicates the MARSE energy of an AE event. The gray shapes indicate the position and size of original AE source (i.e., sample outline). The amplitude distribution of each channel is shown by a histogram on the right side. For some experiments, a blank histogram is shown, which indicates the corresponding AE sensor was not working (HH308 and HH321). Photographs of the AE source/sample after an experiment are shown in the insets for each experiment, except for BT540, which was not recoverable (Fig. 7). Next, we describe each experiment in more detail.

FIG. 7.

The AE hypocenter locations of (a) large glass beads, (b) small glass beads, (c) fused glass wedge, (d) fused glass wedge + glass beads, (e) alumina wedge, (f) two layers of glass beads, (g) molybdenum wedge + glass beads, (h) large glass beads with beam. The gray area indicates the position and size of the sample. The size of the circles denotes the MARSE energy of events. “n” is the number of reprocessed events. The optical images of the self-designed AE sources are shown in the insets (if recovered). The amplitude distribution of each channel is shown at the right side with different colors.

FIG. 7.

The AE hypocenter locations of (a) large glass beads, (b) small glass beads, (c) fused glass wedge, (d) fused glass wedge + glass beads, (e) alumina wedge, (f) two layers of glass beads, (g) molybdenum wedge + glass beads, (h) large glass beads with beam. The gray area indicates the position and size of the sample. The size of the circles denotes the MARSE energy of events. “n” is the number of reprocessed events. The optical images of the self-designed AE sources are shown in the insets (if recovered). The amplitude distribution of each channel is shown at the right side with different colors.

Close modal

First, we wrapped in a Ni foil capsule glass beads with different diameters (≥200 µm in HH287, ≤100 µm in HH289) to be used as AE sources. The AE location results are shown in Figs. 7(a) and 7(b). The located AE events outline the shape of the glass beads sample. It appears that the MARSE energy of HH287 is higher than that of HH289, consistent with larger cracks forming in the larger glass beads.

The results from calibration experiments using (1) fused glass wedge (FGW) pistons [Fig. 7(c)], (2) FGW pistons + small glass beads [Fig. 7(d)], and (3) saw-cut alumina pistons [Fig. 7(e)] show the original outline of the saw-cut at 30° to the compression direction. With the serrated sliding surfaces of both FGW pistons in contact, numerous AE events were produced during the mutual movement of the FGW. It is clear that the FGW + small glass beads produce the most AE due to cracking of the additional glass beads. In contrast, with the addition of an h-BN soft layer between the FGW, the pistons barely produced AE and only a few cracks were detected despite obvious FGW piston displacements. The located source distribution is also compatible with the slope of the FGW. To rule out the possibility that AE events come from the breaking of the FGW body, we repeated an experiment with saw-cut solid molybdenum wedges with small glass beads glued to the sliding surfaces as shown in Fig. 7(g). We observe nearly the same location configuration as in the other experiments using brittle FGW. To further test the location accuracy of calculated AE, we performed a two-layer experiment with two AE sources separated by a soft h-BN block as shown in Fig. 7(f). As expected, we accurately identified the positions of the two layers of AE sources in the PTM cube. The distance between two clusters of AE sources matches the location of the two glass beads layers in the assembly, suggesting a high accuracy of our HPHT AE system.

A keen observer may notice in some samples with thousands of plotted hypocenters, a certain grid-like clustering of events [e.g., particularly HH289 and BT450, Figs. 7(b) and 7(g)]. This clustering can be artificially enhanced if the tolerances are too high and iterations too many using the MATLAB minimization algorithm “fminsearch” as discussed earlier. Clustering of hypocenters of events also shows up in AEwin. The exact origin of these artifacts is not known, but it could be the result of too many hits arriving simultaneously at the sensors, overloading the AE system. That is, under these unnatural conditions, each glassy bead acts as its own AE source, and as a result it becomes nontrivial to define which hits form true events. It entirely depends on how the algorithm defines an event, which AEwin does not disclose. These artifacts are unlikely to play a role in the cracking of rocks in situ, where AE activity is expected to be lower. Finally, any grid-like clustering does not seem to contribute large errors to the general locations of AE events in the samples. The overall shape of each cluster of AE events is in reasonable in agreement with the sample volume (i.e., gray area) in each experiment, with 0.5 mm precision (Fig. 7).

Based on these test runs, we can say that our AE detection system at P61B offers excellent quality AE data and can be used for AE investigations under HPHT to address scientific problems, such as the origin of earthquakes in the Earth’s mantle.

To test the capability of our high pressure AE system coupled with x-ray radiography and ED-XRD, we conducted two in situ AE experiments using the polychromatic x-ray beam of P61B. In the first experiment, BT540, we again used glass beads in Ni foil (25 μm foil thickness) as AE sources. Montages of multiple x-ray radiography images, making up the total sample length were obtained by x-ray radiography at several selected deformation points [Fig. S6(a)]. As indicated by the two arrows in the supplementary material figure, we show that the length of the sample gradually decreases with increasing strain due to the constant displacement rate of the hydraulic rams during anisotropic compression. The strain history of the glass bead samples at different time points is summarized in Fig. S6(b). The strain history is shown by a red line, which indicates a constant strain rate. The deformation time was 45 min using rams No. 1 and No. 2 at a constant displacement rate of −10 μm/min per hydraulic ram. The final strain, ε, is around 9.3% with a strain rate, ε̇, of 3.3 × 10−5 s−1. The AE events were recorded and the locations (n = 1275) are shown in Fig. 7(h). We also provide additional information on the typical peak frequency (around 550 kHz) produced by the cracking of the glassy beads in this experiment (BT540) (Fig. S7), which provides a reference for other experiments. Using x-ray radiography, the deformation process can thus be precisely tracked in situ, on other geological materials of interest also, which may produce AE.

An additional experiment using x rays, BT681 (Fig. 8), was carried out to demonstrate the advantage of combining AE detection, x-ray diffraction and imaging in a single experiment. It also serves to test the feasibility of using an x-ray transparent anvil by acquiring diffracted x rays from the sample through the anvil to a Ge-detector positioned at a horizontal scattering angle (2θ). This special anvil is made of c-BN (+Al2O3 binder) and appears black [see Fig. 2(c)]. This anvil will be beneficial for radial diffraction in the LVP for stress measurements using a monochromatic beam and an area detector35,36 as soon as a monochromator is successfully installed at the station in the future.

FIG. 8.

Experiment BT681. (a) ED-XRD spectra of the SiO2 sliding pistons and of rhenium acquired during the experiment at given pressure (∼1.5 GPa at 45 bars). At a particular heating power (189 W, ∼520 °C), the fused silica glassy pistons crystallized as seen by the appearance of many new reflections. Note that the gray and black dots above certain reflections indicate characteristic x-ray fluorescence of lead (detector shielding) and rhenium. (b) Reprocessed AE events relocated in the pistons and their environment for each step of the experiment. (c) Radiography images collected during each step of deformation and heating showing the displacement of the pistons. Note that sliding and tearing mode cracks are generally not expected to be visible by absorption-contrast imaging. See text for more information.

FIG. 8.

Experiment BT681. (a) ED-XRD spectra of the SiO2 sliding pistons and of rhenium acquired during the experiment at given pressure (∼1.5 GPa at 45 bars). At a particular heating power (189 W, ∼520 °C), the fused silica glassy pistons crystallized as seen by the appearance of many new reflections. Note that the gray and black dots above certain reflections indicate characteristic x-ray fluorescence of lead (detector shielding) and rhenium. (b) Reprocessed AE events relocated in the pistons and their environment for each step of the experiment. (c) Radiography images collected during each step of deformation and heating showing the displacement of the pistons. Note that sliding and tearing mode cracks are generally not expected to be visible by absorption-contrast imaging. See text for more information.

Close modal

In this experiment, the sample is comprised of two fused silica (SiO2) pistons cut at 45°, with glassy beads (200 μm) and rhenium powder lightly glued on the sliding surfaces. The assembly [Fig. 2(a)] was first isostatically compressed to an oil pressure of 45 bar in the master ram, No. 3, where the pressure between the pistons was estimated to be ∼1.5 GPa, by ED-XRD on rhenium and using its equation of state37 [Fig. 8(a)]. During this step and the following ones, AE activity was recorded while data for ED-XRD spectra (100 s) and radiography scans (30 s) were collected alternately. As expected, during initial compression the glassy beads were crushed, accompanied by many AE events relocated on the sliding surface [Fig. 8(b)]. Subsequently, for the first deformation step, the anisotropic compression of the assembly at room temperature caused the pistons to build up stress and slide. After 1 h, deformation was suspended and the pistons were subsequently heated to around 740 °C (250 W). See Fig. 4(a) for the power–temperature calibration curve using a 4 mm OD heater. During heating at about 520 °C (189 W), the pistons crystallized with the appearance of many SiO2 reflections. At the same time, a prolonged burst of AE activity was detected [Figs. 8(a) and 8(b)]. Presumably, stress relaxation at moderate temperatures coupled with crystallization favors rapid crack propagation. The AE events are observed to come from any location in the pistons [Fig. 8(b)]. At the highest heating power (250 W), AE activity subsided. After a quench, the now-crystallized pistons were again deformed and reheated once more to 300 W before quenching. Further AE activity was recorded but without large bursts. Note that, before deformation, the initial length of the combined pistons was ∼5 mm with a diameter of 2.5 mm as indicated by the gray shapes in Fig. 8(b). The pistons were pushed a total distance of 1.68 mm with a displacement rate of 7 μm/min per ram, twice for 1 h. Selected radiography scans are presented in Fig. 8(c).

In the above experiment, the c-BN anvil has the same dimensions as those of other WC anvils. However, due to the compositional difference (and the presence of a WC backing plate), the average wave speed of AE from the sample to the sensor is different for the c-BN anvil. The model used to locate AE assumes an isotropic velocity. Consequently, to account for the fact that the velocity is actually higher through the c-BN anvil, we modify the virtual position of the transducer on the c-BN anvil in the location model used by “bringing” it closer to the sample based on a calculation. First, pulse tests were performed to measure the velocity difference between carbide and c-BN anvils (see the supplementary material). The wave speed through a WC anvil is measured as 6166.67 m/s and the wave speed through the c-BN anvil is 7325.33 m/s. In addition, the wave speed through a 16 mm pure B-epoxy cube is about 3200 m/s, which will vary for an assembly with additional components. Given that the sensor to sample distances under load are generally well known (based on the displacements of the rams), we now calculate a “virtual” position for the sensor on the c-BN anvil by keeping the travel time in the WC anvils equal to that in the c-BN anvil. Thus, we calculated the virtual position of the c-BN sensor as 35.6 mm (see the supplementary material) although its real position is 44.5 mm as for the WC anvils. We then processed the AE hypocenters using the MATLAB script and the results are plotted in Fig. 8(b).

High-resolution imaging of the recovered samples should demonstrate that the AEs indeed come from the cracking and crushing of the glass beads. A scanning electron microscope (SEM, Tescan Amber X scanning electron microscope) at the DESY Nanolab was used for this purpose. After each experiment, the recovered boron–epoxy cubes were cut in half, parallel to the compression axis. Figures 9(a), 9(c), 9(e), and 9(f) show low-magnification SEM images of the recovered glass beads (HH289), fused glass wedge (HH264), and two layers of glass beads (HH367). In the glass beads sample, we can observe that some porosity reduction occurred by crushing during the compression and deformation procedure. Some glass beads were completely crushed to small pieces, whereas others preserved their original spherical shape (as shown by the arrows). We confirm this observation for all experiments with glassy beads samples. As shown in Fig. 9(h) for dense alumina wedge (HH321) and Fig. 9(i) for fused glass wedge (HH308), the sawtooth pattern was crushed with many cracks. The tips of the saw-cut pistons were crushed to fine powder around the sliding surface. Other parts of the piston did not crack seriously. The Mo pistons are noticeably uncracked [Fig. 7(g)] as expected. Hence, the only source of AE activity for HH411 could have come from the glass beads, which were glued on the sliding surfaces. All the features observed here are consistent with our AE relocation results, and show that, as expected, AE events can be relocated inside the sample with good accuracy, using traditional 3D relocation algorithms. As mentioned previously, AE data (of e.g., candidate rock samples) can be reprocessed by the experimenter using other, more advanced, seismic algorithms. Our AE detection system does not present any limitation to attempt this.

FIG. 9.

The SEM images of calibration sources after experiment. (a) and (b) Large and (c) and (d) small glass beads. (e)–(g) Two layers of glass beads, (h) dense alumina piston, and (i) fused glass wedge. The experimental run numbers are shown in each corner.

FIG. 9.

The SEM images of calibration sources after experiment. (a) and (b) Large and (c) and (d) small glass beads. (e)–(g) Two layers of glass beads, (h) dense alumina piston, and (i) fused glass wedge. The experimental run numbers are shown in each corner.

Close modal

We developed an acoustic emission (AE) detection setup compatible with in situ x-ray imaging and diffraction at the high-energy wiggler endstation P61B LVP. The system by GMA/MISTRAS can record triggered AE activity as well as stream full waveforms emanating from samples at HPHT conditions, deformed by anisotropic compression in the Aster-15 LVP. This setup is a powerful tool to investigate AE under HPHT and couple these acoustic measurements with high-flux x-ray diffraction and imaging. We calibrated the wave velocity of the pressure transmission medium, which provides a basis for accurately locating the acoustic emission sources in the sample. In order to achieve deformation under HPHT with low noise, we designed a new assembly for AE experiments. The efficiency of this AE setup was confirmed by the compression and deformation of various standard samples. Some samples were designed specifically to simulate a fault gouge. The hypocenters of the AE obtained in these experiments successfully allow imaging the fault plane. This AE setup therefore proved to be extremely effective and is available for users with and without the availability of synchrotron x rays at the P61B LVP station. We expect scientifically pioneering studies on (Earth) materials under high pressure, temperature, strain rate, and stress to investigate processes such as dehydration embrittlement and phase transformations in the Earth from which intermediate and deep earthquakes may originate. We offer thus a powerful tool to investigate mineral and material rheology at HPHT.

The supplementary material includes 7 figures (S1–S7), 1 table (S1), and the MATLAB script used to postprocess the AE waveform data exported from the acquisition software AEwin. The MATLAB script also replots the characteristics of each hit/event.

The authors gratefully thank discussions and technical support from Dr. Bhat and Mr. Sonntag at P61B, Tim Officer at APS (ANL, USA), and an internal reviewer at DESY. We would like to thank J. Arno and R. S. Snata for SEM sample preparation and SEM measurements. Finally, we thank anonymous reviewers for their helpful suggestions, which significantly improved the manuscript. We acknowledge DESY (Hamburg, Germany), a member of the Helmholtz Association HGF, for the provision of experimental facilities. This research was carried out at beamline P61B (Proposal No. I-20210227 for beam time) with support from the Federal Ministry of Education and Research, Germany (BMBF, Grant Nos. 05K16WC2 and 05K13WC2) and funding from POF4-6G3.

The authors have no conflicts to disclose. S.M. and R.F. designed and carried out the experiments. S.M. and R.F. co-wrote the manuscript with contributions from other co-authors.

Shuailing Ma: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (lead); Writing – original draft (lead); Writing – review & editing (equal). Julien Gasc: Formal analysis (supporting); Investigation (supporting); Methodology (supporting); Software (supporting); Validation (lead); Writing – review & editing (supporting). Robert Farla: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Resources (equal); Software (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available within the article and its supplementary material.

1.
T.
Irifune
,
A.
Kurio
,
S.
Sakamoto
,
T.
Inoue
, and
H.
Sumiya
,
Nature
421
,
599
(
2003
).
2.
A.
Chanyshev
,
T.
Ishii
,
D.
Bondar
,
S.
Bhat
,
E. J.
Kim
,
R.
Farla
,
K.
Nishida
,
Z.
Liu
,
L.
Wang
,
A.
Nakajima
,
B.
Yan
,
H.
Tang
,
Z.
Chen
,
Y.
Higo
,
Y.
Tange
, and
T.
Katsura
,
Nature
601
,
69
(
2022
).
3.
A.
Zerr
,
G.
Miehe
,
G.
Serghiou
,
M.
Schwarz
,
E.
Kroke
,
R.
Riedel
,
H.
Fueß
,
P.
Kroll
, and
R.
Boehler
,
Nature
400
,
340
(
1999
).
4.
T.
Ishii
and
E.
Ohtani
,
Nat. Geosci.
14
,
526
(
2021
).
5.
R.
Farla
,
S.
Bhat
,
S.
Sonntag
,
A.
Chanyshev
,
S.
Ma
,
T.
Ishii
,
Z.
Liu
,
A.
Néri
,
N.
Nishiyama
,
G. A.
Faria
,
T.
Wroblewski
,
H.
Schulte-Schrepping
,
W.
Drube
,
O.
Seeck
, and
T.
Katsura
,
J. Synchrotron Radiat.
29
,
409
(
2022
).
6.
T.
Yu
,
Y.
Wang
,
M. L.
Rivers
, and
S. R.
Sutton
,
C. R. Geosci.
351
,
269
(
2019
).
7.
T.
Ohuchi
,
X.
Lei
,
H.
Ohfuji
,
Y.
Higo
,
Y.
Tange
,
T.
Sakai
,
K.
Fujino
, and
T.
Irifune
,
Nat. Geosci.
10
,
771
(
2017
).
8.
C.
Meade
and
R.
Jeanloz
,
Science
252
,
68
(
1991
).
9.
H. W.
Green
,
C. H.
Scholz
,
T. N.
Tingle
,
T. E.
Young
, and
T. A.
Koczynski
,
Geophys. Res. Lett.
19
,
789
, (
1992
).
10.
T. N.
Tingle
,
H. W.
Green
,
C. H.
Scholz
, and
T. A.
Koczynski
,
J. Struct. Geol.
15
,
1249
(
1993
).
11.
C.
Meade
and
R.
Jeanloz
,
Nature
339
,
616
(
1989
).
12.
D. P.
Dobson
,
P. G.
Meredith
, and
S. A.
Boon
,
Science
298
,
1407
(
2002
).
13.
D. P.
Dobson
,
P. G.
Meredith
, and
S. A.
Boon
,
Phys. Earth Planet. Inter.
143–144
,
337
(
2004
).
14.
H.
Jung
,
Y.
Fei
,
P. G.
Silver
, and
H. W.
Green
,
Rev. Sci. Instrum.
77
,
014501
(
2006
).
15.
A. A.
De Ronde
,
D. P.
Dobson
,
P. G.
Meredith
, and
S. A.
Boon
,
Geophys. J. Int.
171
,
1282
(
2007
).
16.
J.
Gasc
,
A.
Schubnel
,
F.
Brunet
,
S.
Guillon
,
H.-J.
Mueller
, and
C.
Lathe
,
Phys. Earth Planet. Inter.
189
,
121
(
2011
).
17.
M.
Shigeishi
and
M.
Ohtsu
,
Constr. Build. Mater.
15
,
311
(
2001
).
18.
M.
Ohtsu
,
Res. Nondestr. Eval.
6
,
169
(
1995
).
19.
M.
Ohtsu
,
J. Geophys. Res.
96
,
6211
, (
1991
).
20.
A.
Schubnel
,
F.
Brunet
,
N.
Hilairet
,
J.
Gasc
,
Y.
Wang
, and
H. W.
Green
,
Science
341
,
1377
(
2013
).
21.
S.
Incel
,
L.
Labrousse
,
N.
Hilairet
,
T.
John
,
J.
Gasc
,
F.
Shi
,
Y.
Wang
,
T. B.
Andersen
,
F.
Renard
,
B.
Jamtveit
, and
A.
Schubnel
,
Geology
47
,
235
(
2019
).
22.
T. P.
Ferrand
,
N.
Hilairet
,
S.
Incel
,
D.
Deldicque
,
L.
Labrousse
,
J.
Gasc
,
J.
Renner
,
Y.
Wang
,
H. W.
Green
 II
, and
A.
Schubnel
,
Nat. Commun.
8
,
15247
(
2017
).
23.
T.
Ohuchi
,
Y.
Higo
,
Y.
Tange
,
T.
Sakai
,
K.
Matsuda
, and
T.
Irifune
,
Nat. Commun.
13
,
5213
(
2022
).
24.
T.
Officer
and
R. A.
Secco
,
Phys. Earth Planet. Inter.
300
,
106429
(
2020
).
25.
T.
Officer
,
L.
Zhu
,
Z.
Li
,
T.
Yu
,
D. R.
Edey
, and
Y.
Wang
,
Phys. Chem. Miner.
49
,
29
(
2022
).
26.
K.
Christensen
,
L.
Danon
,
T.
Scanlon
, and
P.
Bak
,
Proc. Natl. Acad. Sci. U. S. A.
99
,
2509
(
2002
).
27.
J. A.
Hudson
,
R. G.
Pearce
, and
R. M.
Rogers
,
J. Geophys. Res.
94
,
765
, (
1989
).
28.
F.
Waldhauser
,
Bull. Seismol. Soc. Am.
90
,
1353
(
2000
).
29.
Y.
Wang
,
L.
Zhu
,
F.
Shi
,
A.
Schubnel
,
N.
Hilairet
,
T.
Yu
,
M.
Rivers
,
J.
Gasc
,
A.
Addad
,
D.
Deldicque
,
Z.
Li
, and
F.
Brunet
,
Sci. Adv.
3
,
e1601896
(
2017
).
30.
B.
Yu
,
R. S.
Bradley
,
C.
Soutis
, and
P. J.
Withers
,
Philos. Trans. R. Soc., A
374
,
20160037
(
2016
).
31.
Y.
Fei
,
J.
Li
,
K.
Hirose
,
W.
Minarik
,
J.
Van Orman
,
C.
Sanloup
,
W.
van Westrenen
,
T.
Komabayashi
, and
K.-i.
Funakoshi
,
Phys. Earth Planet. Inter.
143–144
,
515
(
2004
).
32.
33.
Q.
Li
, MIT Rock Mechanics Group AE Code, https://www.mathworks.com/matlabcentral/fileexchange/72339-mit-rock-mechanics-group-ae-code, MATLAB Central File Exchange, 2022.
34.
B. Q.
Li
,
B.
Gonçalves da Silva
, and
H.
Einstein
,
Eng. Fract. Mech.
209
,
200
(
2019
).
35.
N.
Hilairet
,
Y.
Wang
,
T.
Sanehira
,
S.
Merkel
, and
S.
Mei
,
J. Geophys. Res.: Solid Earth
117
,
B01203
, (
2012
).
36.
R.
Farla
,
A.
Rosenthal
,
C.
Bollinger
,
S.
Petitgirard
,
J.
Guignard
,
N.
Miyajima
,
T.
Kawazoe
,
W. A.
Crichton
, and
D. J.
Frost
,
Earth Planet. Sci. Lett.
473
,
291
(
2017
).
37.
C.-S.
Zha
,
W. A.
Bassett
, and
S.-H.
Shim
,
Rev. Sci. Instrum.
75
,
2409
(
2004
).

Supplementary Material