A uniaxial load frame for in situ neutron studies of stress-induced changes in cementitious materials and related systems

The analysis of a material under a sustained load below its ultimate strength is relevant to a multitude of engineering applications. For example, concrete experiences creep as a result of sustained loading, whereby the viscoelastic nature of the material causes ongoing permanent (plastic) deformation as a function of time.¹ Wood and polymers also experience creep at ambient temperature,^2,3 whereas metals tend to be prone to creep at elevated temperatures.⁴ For many materials experiencing ongoing plastic deformation due to sustained loading, the underlying chemistry and physics responsible for the macroscopically measured strain remains elusive. However, these details are extremely important since creep of a material in infrastructure (i.e., bridges, buildings, and dams) can limit its lifetime and therefore lead to costly maintenance and repairs or even demolition and replacement.

With the rapid advancements in accelerator physics and target development, spallation-derived neutrons are enabling the discovery of time-dependent material processes due to significant improvements in beam flux.⁵ Although synchrotron-derived X-rays are extremely appropriate for the investigation of fast dynamics in many materials, neutrons enable for the analysis of large samples where X-ray attenuation and multiple scattering are issues. Furthermore, due to the non-monotonic behavior of neutron scattering strength with increasing Z-number, neutrons can provide enhanced contrast for elements adjacent on the periodic table. Hence, neutrons are very appropriate for the analysis of engineering materials, including metals and concrete. Previous neutron studies on cementitious materials have elucidated the water dynamics during hydration (formation) using quasi-elastic neutron scattering (QENS),^6,7 the atomic structural rearrangements that occur upon heating using in situ and ex situ neutron pair distribution function (PDF) analysis,^8,9 and the local structural changes that occur during formation of alkali-activated materials (AAMs) using in situ neutron PDF analysis.¹⁰ 

The Nanoscale Ordered Materials Diffractometer (NOMAD)¹¹ at the Spallation Neutron Source (SNS) offers the world’s highest flux neutron powder diffraction and PDF capability, expanding the scope for in situ and in operando materials structure studies.^12–15 Here, the design of a novel load frame is presented for use on this neutron scattering beamline, with an emphasis on maximizing the scattering angle coverage to optimize data collection for PDF.¹⁶ Details on the frame components and design reasoning are provided. Finally, preliminary total scattering data are given showing subtle but significant structural changes occurring during creep of an alkali-activated slag paste sample.

Schematics of the load frame are given in Figs. 1 and 2(a), where the overall dimensions are 328 mm × 229 mm × 72 mm (12.9″ × 9″ × 3″). The sample (i.e., mechanically hard cement paste cylinder of 10 mm diameter and 30 mm length) is mounted between the ram and platen. Rotation of the lead screw at the top of the frame leads to compression of the chrome-silicon steel die spring and a uniaxial compressive load being exerted on the sample cylinder. A photograph of the load frame is given in Fig. 2(b).

Material selection for the components of the load frame was carried out to reduce the secondary scattering effects of the sample environment and minimize the activation of the frame due to neutron exposure. Specifically, components that were in the vicinity of the sample (ram and platen) and within the scattering solid angle (vertical posts) were made from 7075 aluminum. This alloy of aluminum was chosen (i) due to the reduced iron concentration compared with 6061 aluminum and (ii) for its strength and hardness. The vertical posts were reduced in the cross-sectional area within the projected solid angle detector coverage. Furthermore, alignment of the frame in the instrument was such that the sample cylinder was centered in the neutron beam with vertical posts oriented at 90° and 270°, perpendicular to the beam direction.

Other components of the load frame were generally made from 6061 aluminum, apart from the spring (chrome-silicon steel), 18-8 stainless steel spherical washer for the mounting load platen, and certain metal components along the load shaft (440C stainless steel bushing, carbon steel thrust bearing, 18-8 stainless steel washer, 360 brass ACME nut, and 1018 carbon steel ACME lead screw). Steel was generally avoided in the vicinity of the direct and scattered neutron beam. All materials were sourced from McMaster Carr and altered as necessary prior to assembly of the load frame.

The spring calibration was performed on an Instron model 5567 materials testing system. Three parts of the uniaxial load frame were mounted in the compression test space of the Instron testing system; the die spring with the two pilot bushings at either end (see Fig. 3). Note that the dowel pins protrude from each pilot bushing. The pins can be accessed while the spring and bushings are assembled in the uniaxial load frame housing.

The spring and pilot bushing assembly was compressed in the Instron, where the contraction of the length of the spring was measured continuously while the spring assembly was steadily compressed at a rate of 25 mm/min until a target load was reached. A plot between load (measured in Newton) and contraction in spring length (in mm) is shown in Fig. 4.

As can be seen in Fig. 4, the relationship between the induced load and contraction is almost linear. This linear relationship suggests that, for the ranges (0–15 mm) tested here, the employed spring undergoes uniform compression and is therefore reliable for providing a given load on the sample. The only non-linear region is the one at the very beginning (0–1 mm) which is due to the clearance allowed between the spring assembly and the Instron upper platen prior to imposing spring compression. Finally, the slope of the linear region, which corresponds to the spring constant of the employed spring, was determined to be 165.2 N/mm. The dowel pins in Fig. 3 were used to design a custom aluminum gauge which allows a user to induce a desired amount of compressive load on any given sample. While at target load, the gauge was fabricated to closely fit between the pins. The gauge is designed to enable spring contraction repeatability and can therefore be used to reach the target load while the spring and pilot bushings are assembled in the uniaxial load frame housing.

Calibration scans of the sample environment loaded in the diffractometer were required to enable detector-pixel mapping, background subtraction, and correct normalization needed for the measured data. The instrument calibration procedure includes masking of detector pixels that are shadowed or otherwise corrupted by the specific sample environment components. The following calibration scans were obtained (i) empty load frame, (ii) vanadium rod in the load frame, (iii) empty vanadium can in the load frame, and (iv) diamond powder in the vanadium can in the load frame. Furthermore, silicon powder (NIST standard reference material) was measured in the vanadium can in the load frame, and a comparison of the total scattering pattern and corresponding neutron PDF dataset with the equivalent data obtained from the standard setup of the instrument (linear sample changer) is shown in Fig. 5. These data show that after instrument calibration, the data quality obtained on the load frame is equivalent to that associated with general measurements conducted on the instrument.

After these calibration scans were performed, the cement sample was mounted in the load frame and lowered into the instrument without an imposed load for acquisition of a dataset of the sample prior to viscoelastic relaxation. The sample was a 7 day old sodium silicate-activated blast furnace slag paste cylinder with a H₂O/slag wt. ratio of 0.440 and a Na₂O wt. % of 4 (relative to the weight of slag). D₂O was used during sample synthesis to minimize incoherent scattering during the measurement; hence, the equivalent D₂O/slag wt. ratio was 0.489. The cylinder was wrapped tightly in aluminum foil to prevent drying from occurring and as a containment barrier of the sample. Details on sample synthesis can be found in Ref. 17, where the paste was initially cured for 24 h in a plastic tube (10 mm diameter, 30 mm length). After demolding, the sample was then wrapped tightly in the aluminum foil and stored in a high humidity environment (wet paper towel inside a ziplock bag) to prevent drying until it was mounted on the load frame.

Immediately after the 90 min first scan of the paste cylinder was complete [see Fig. 2(b) for the photograph of the uniaxial load frame with the mounted sample], the load frame was removed from the instrument and housed within a plastic shielding system to contain the activated sample should it be damaged under applied stress and fall out of the load frame. A uniaxial stress was imposed on the sample via rotation of the lead screw at the top of the load frame until the spring compressed by a known distance in the vertical direction (calibrated a priori) that corresponded to 22 MPa of uniaxial compression on the sample (1731 N in Fig. 4). Once the desired stress level was reached, the load frame was allowed to rest for 10 min to ensure that the sample was tightly (and safely) held between the ram and the platen. After this allotted time period had expired, the frame was lowered back into the instrument and data were continuously acquired for 24 h.

The sustained uniaxial compression will cause the sample to contract in dimension, which leads to a decrease in the applied load according to the spring constant and extent of contraction. Laboratory-based creep tests combat this decrease in load by periodically measuring the load and correcting the value if found to be more than 2% from the designed value [according to the American Society for Testing and Materials (ASTM C512) standard]. Here, selection of the spring characteristics, specifically the spring constant, was carried out to avoid the need to readjust the load throughout the measurement. To the best of our knowledge, direct data on creep shrinkage in alkali-activated slag paste are not available in the literature, but several studies on alkali-activated slag concrete have been reported (∼600–900 microstrain recorded at 100–180 days after loading, load imposed at 28 days after mixing).^18,19 Creep of cement paste is significantly higher compared to concrete due to the absence of sand and gravel, with experimental and modeling results revealing that for concrete with ∼70 vol. % aggregates the extent of creep shrinkage in the concrete is approximately one-fifth of that in the corresponding cement paste.²⁰ Taking this ratio into account together with the creep shrinkage reported for alkali-activated slag concrete and the spring constant (165.2 kN/mm), the corresponding load decrease is ∼23 N (1.3% of initial load). Although the above calculation was based on samples loaded 28 days after mixing, and after 180 days of sustained loading, this approximate calculation indicates that there is minimal change in the applied stress as the alkali-activated slag paste experiences creep.

Selected raw total scattering patterns for the sample are presented in Fig. 6, prior to subtraction of the incoherent scattering from H/D. Incoherent scattering is apparent from the large elevated background present in Fig. 6 along with the downward slope of the data at higher Q. Approaches for the removal of the incoherent contributions have been outlined in the literature.^21,22 The figure shows that the changes due to creep in the paste are very subtle, but the difference plot clearly shows that atomic structural changes are occurring. Specifically, the difference plot contains multiple instances of sudden increases/decreases in intensity across a wide range of momentum transfer (Q) values, indicating that alterations to the structure are happening across length scales (atomic to nanoscale). Comparison of the difference plot with the scattering pattern from liquid D₂O (not shown here) reveals that the changes seen during sustained loading are not attributed to the loss of D₂O. The simulated neutron scattering pattern of deuterated Na-containing 14 Å tobermorite is also given in Fig. 6,²³ where it is clear that there are some similarities between this simulated pattern and certain peaks in the experimental data. The tobermorite family of crystal structures (e.g., 11 Å tobermorite, 14 Å tobermorite, clinotobermorite) possess structural similarities to the main binder phase in alkali-activated slag, sodium-containing calcium-alumino-silicate-hydrate (C-(N)-A-S-H) gel and therefore are often used to aid in data interpretation. Although it is beyond the scope of the current article, a detailed analysis of the data after careful data reduction will shed light on the fundamental material mechanisms responsible for macroscopically measured creep in these sustainable cements.

Here, a custom-build load frame is outlined to enable for analysis of the structure (atomic, nano, and larger length scales) of engineering materials whilst under sustained compressive loading conditions. The load frame is a scaled-down version of a creep apparatus used for macroscopic analysis of concrete samples. The frame is primarily made from aluminum metal, with components shielded from primary/scattered neutrons consisting of brass and steel. Preliminary data are shown for an alkali-activated slag paste cylinder subjected to 40% of its ultimate compressive strength, revealing that there are subtle but important structural changes at the atomic/nanoscale level occurring to the material as a result of the sustained load. Future experiments using this apparatus may include creep studies on Portland cement-based systems, wood, and polymers. Additionally, given that engineering materials are often exposed to extreme environment conditions during their lifecycle, analysis of the material in non-ambient temperature environments (hot and cold) whilst being under load would enable researchers to explore material behavior under certain environmental conditions.

The authors would like to acknowledge the School of Engineering and Applied Sciences machine shop at Princeton University for the fabrication of components for the load frame and Genevieve Martin at Oak Ridge National Laboratory for the photograph used in Fig. 2. This work was supported by NSF through the MRSEC Center (Grant No. DMR-1420541). The research used resources at the Spallation Neutron Source, a DOE Office of Science User Facility operated by the Oak Ridge National Laboratory.