Influence of composite admixtures on the freezing resistance and pore structure characteristics of cemented sand and gravel

: CSG shows low freezing resistance due to its poor cementing properties, and adding admixtures is the main method to improve its freezing resistance. In this paper, experiments were designed based on the Taguchi method and on the freezing resistance and pore structure characteristics of CSG after adding admixtures. After 20 freeze-thaw cycles, the results showed that both the compressive strength and dynamic elastic modulus of CSG decreased. The area occupancy of more harmful pores (>200 nm) in the NMR test increased, and the fractal dimension of the pore structure in the SEM images also increased. Based on the above indicators, it was concluded that L9 (sodium dodecyl sulfate, graphene oxide, nanosilica powder, and water-reducing agent) has the best freezing resistance. Experiments were conducted on L9 to verify that L9 can significantly improve the freezing resistance of CSG materials under different water-binder ratios, sand ratios, and cement dosages.


INTRODUCTION
Cemented sand and gravel (CSG) is a new kind of environmentally material for dam construction. It is made by simply mixing a few cementing materials into the riverbed sand and gravel, excavated waste materials, or other rock-based materials that are easy to obtain near a dam site, making it cost-effective and environmentally (1)(2)(3)(4). L. Guo et al. (5) introduced fractal theory to study the effect of aggregate properties on CSG properties, quantifying aggregate gradation and shape. The results shown that the closer the aggregate gradation to the standard gradation, the better the performance of the CSG; as the proportion of polygonal aggregates increased, the worse the performance of the CSG. However, due to the small amount of cement, the CSG also reflects the shortcomings of its own material, its compressive strength is small, and its frost resistance is poor. As far as cement-based materials are concerned, an appropriate amount of additives can improve their mechanical properties and durability. Sanrı-Karapınar, I., et al. (6) used κ-carrageenan/ PVA/eggshell nanostructures as a novel biodegradable and homogeneous nanostructure in cement compositions, which improved the compressive strength and flexural strength of cement-based materials. Yildirim, M., & Derun, E. (7) studied the compressive and flexural strengths, thermal properties and pore structure of two types of boron waste and different amounts of CuO nanoparticles-modified mortars. It was observed that nano-CuO addition in the range 2%-2.5% can improve mechanical properties. Hernández, EF, et al. (8) investigated the mechanical performance and durability of concrete with water/cement (w/c) ratios of 0.30 and 0.60 containing cactus mucilage and brown marine seaweed extract solutions. It was found that when the water cement ratio was 0.60, the compressive strength of concrete after 120 days increased by 20% compared with the control group. Silva,Bruna,et al. (9) investigated the freeze-thaw resistance of lime mortar with different types of admixtures (water reducers, viscosity modifiers, air-entraining agents, and water repellents). Results showed that the water-reducer (superplasticizer) based on polynaphthalene sulfonate and the water-repellent based on silicone resin were the admixtures that more significantly improved the resistance to freeze-thaw, followed by the air-entrainers based on sodium dodecyl sulfate and sodium alfa-olefin sulfonate. Behfarnia, K., & Salemi, N. (10) investigated the frost resistance and mechanical properties of concrete containing nano-silica and nano-alumina. The results showed that the frost resistance of concrete containing nano-Al 2 O 3 was better than that of concrete containing the same amount of nano-SiO 2 . However, its compressive strength is lower than that of concrete containing the same amount of nano-SiO 2 . As a special cement-based materials, CSG can also improve its freeze-thaw resistance durability by adding appropriate amount of admixtures. In cold regions, freeze-thaw damage is the main factor causing the failure of a CSG dam and greatly shortens the service life of the dam. Some experts believe that there is a strong correlation between indoor freeze-thaw tests and actual field engineering (11)(12)(13). Therefore, it is of particular importance to study how to improve the freezing resistance of CSG materials through laboratory experiments.
At present, there are few studies on improving the freezing resistance of CSG materials, as the main focus has been mixing the proper amount of cementing materials and the enhancing effect of a single admixture. Xu Long (14) studied the effects of the type and content of traditional sodium alpha-olefin sulfonate (AOS) air-entraining agent and lightweight ceramsite (clay sand, polystyrene particles, and rubber powder of different fineness) on the freezing resistance of CSG, and the results showed that the freezing resistance of CSG increases with the increase in the content of AOS air-entraining agent, and light ceramsite can improve the freezing resistance of CSG to a certain extent. Feng Wei and Jia Jinsheng et al. (15) studied the freezing resistance of CSG under the action of air freezing and water freezing after the addition of admixtures (water-reducing agent and air-entraining agent). For CSG with an additive, the impermeability level at 180 d can reach W12, and it was destroyed in over than 75 freezing-thawing cycles in water but not destroyed in over 350 cycles in air. Matsumura K et al. (16) introduced air freezing and water freezing in CSG freeze-thaw tests, and their results showed that the freezing resistance of CSG with air freezing is approximately 10 times that of water freezing. Chen S, Zheng Y et al. (17) established a CSG freeze-thaw damage evolution model, which can predict the freezing resistance of CSG. G. Jin et al. (18) studied the influence of cement content on the strength and durability of CSG through freeze-thaw tests. As a result, the strength and durability of CSG are reliable when the amount of cement is over 0.8 kN/m 3 . Yeon K S et al. (19) studied the freeze-thaw resistance of CSG per unit of cement content. The above research shows that adding a single admixture can indeed improve the freezing resistance of CSG, but the effect is limited. Therefore, this paper investigates the effect of mixing multiple admixtures on the freezing resistance and pore structure of CSG and seeks a compound admixture to improve the freezing resistance of CSG.

Experimental materials
In this experiment, P•O 42.5 Portland cement produced by Henan Tianrui Co., Ltd. was selected, and its technical indicators are detailed in Table 1. The fly ash used was Class I ash from the Yulian Power Plant, and its technical performance meets the relevant requirements in GB/T1596-2017 fly ash used for ce ment and concrete; see Table 2 for details. The fine aggregate was produced by Tanghe County Xinmiao Sand Co., Ltd., with a fineness modulus of 2.94 and a bulk density of 1,625.0 kg/m 3 . The coarse aggregate used was natural pebbles dug out from a riverbed. Containing a large amount of mud, the pebbles were cleaned before the experiment, and those with particle sizes of 5-20 mm and 20-40 mm were screened. The mixing water was tap water from the laboratory.
Five kinds of admixtures were selected, namely sodium dodecyl sulfate, triterpenoid saponins, graphene oxide, nano silicon powder, and polycarboxylate superplasticizer. Their characteristics are shown in Table 3.

Mixing ratio
The basic mixing ratio of CSG is shown in Table 4, where the water-binder ratio is 1.0 and the sand ratio is 0.3.
Due to multiple admixtures, this experiment was designed using the Taguchi method (20,21). Five factors were selected, namely sodium dodecyl sulfate (A), triterpenoid saponins (B), graphene oxide (C), nano silicon powder (D) and polycarboxylate superplasticizer (E), and four levels were selected for each factor to design the L16 (4 5 ) orthogonal arrays, as shown in Tables 5 and 6, where L1 is the blank group.

Preparation and curing of the test pieces
The specimens were 100 mm × 100 mm × 100 mm for the uniaxial compression test and 100 mm × 100 mm × 400 mm for freeze-thaw test. First, add the weighed cementitious materials and aggregate into HJW100 mixer in sequence, and stir for 1min. Next, add the mixture of compound admixture and water to the blender and stir for 2min. Finally, the mixture is filled into steel moulds of corresponding size. Put the mold with the material on the vibrating table and vibrate for 60s. After pouring, the specimens were placed indoors for 24 hours and then demolded. The demolded specimens were transferred to a standard curing room (temperature 20±2ºC, humidity 95%) for curing for 28 days.

Experimental methods
The experiment was performed in accordance with the standard for test methods of concrete physical and mechanical properties (GB/T50081-2019) (22), test code for hydraulic roller compacted concrete (DL/T 5433-2009) (23), standard for test methods of longterm performance and durability of ordinary concrete (GB/T 50082-2009) (24) and other standards.

Compressive strength test
The compressive strength test adopted a WAW-1000 electrohydraulic servo universal testing machine. The specimens was a 100 mm×100 mm×100 mm specimens before and after freezing and thawing. During the test, the specimens was continuously and uniformly loaded at a speed of 5 kN/s until it fractured, and the peak failure load P was recorded to calculate the compressive strength.

Freezing resistance test
The freeze-thaw test was performed with three specimens as a group. First, a 100 mm × 100 mm ×  400 mm specimens that had been cured to 24 days of age was soaked in water at 20 °C ± 3 °C for four days. Second, the specimens was removed, and the surface moisture was wiped off. We weighed the initial mass and the initial transverse fundamental frequency of the specimens, put the specimens into the test box, and tap water was added until the level was 20 mm higher than the top surface of the specimens. Finally, the test box containing the specimens was put into a freeze-thaw box containing a freeze-thaw solution with a temperature of -25 °C to 20 °C, and the parameters of the freeze-thaw machine were set: the cycle lasts 4.0 h, the cooling lasts 2.5 h, and the heating lasts 1.5 h. After 20 freeze-thaw cycles, we measured the mass and the transverse fundamental frequency of the specimens again, and calculated the relative dynamic elastic modulus and mass loss rate through Equation [1] and [2] to characterize the freezing resistance of the CSG.
The relative dynamic elastic modulus was calculated using the following formula: [1] Where is the relative dynamic elastic modulus (%) of the CSG specimens after the N-th freeze-thaw cycle; is the dynamic elastic modulus(MPa) of the CSG specimens after the N-th freeze-thaw cycle; is the dynamic elastic modulus(MPa) of the CSG specimens before the freeze-thaw cycle test; is the shape calculation parameter(kg/m 2 ); is the mass of the specimen(kg); is the transverse fundamental frequency (Hz) of the CSG specimens after the N-th freeze-thaw cycle; is the initial transverse fundamental frequency (Hz) of the CSG specimens before the freeze-thaw cycle test.
The mass loss rate was calculated using the following Equation [2]: [2] Where is the mass loss rate (%) of the specimens after the N-th freeze-thaw cycle; is the mass (g) of the specimens before the freeze-thaw cycle test; and is the mass (g) of the specimens after the N-th freeze-thaw cycle.

Scanning electron microscopy test
A broken specimens from compression tests was randomly selected for scanning electron microscopy (SEM) examination. Before the SEM test, the sample was polished and ground to a size of approximately 10 mm and fixed on the sample base with conductive adhesive for metal spraying. The spraying thickness was 200-300 Å. During the test, a Quanta-650 Environmental Scanning Electron Microscope (ESEM) and an Apollo-X Spectrometer was used. working conditions: Test in BSE mode, the acceleration voltage was 25kV, the SEM scanning time was 45 seconds, and the EDS scanning time was 3~5min. The sample magnification was 500~5000×.

Pore structure analysis 2.4.4.1 Porosity test
A MesoMR12-060H-I nuclear magnetic resonance device was used in the test. First, the specimen with an age of 28 days for 24 hours vacuum saturation, and then the specimen was put into the coil of the nuclear magnetic resonance instrument. NMR relaxation measurements were carried out by a nuclear magnetic resonance analysis system. After the Carr-Purcell-Meiboom-Gill (CPMG) sequence attenuation signal data were collected, the data inversion was performed to obtain the T2 spectrogram which was then converted into a pore size distribution diagram. The pore size was measured between 50 nm and 200 μm.

Pore structure characterization based on SEM images
With reference to the technology adopted by Liu et al (25) pore and crack analysis system (PCAS) software was used to obtain the pore binary characteristics. The threshold segmentation method was used to binarize the electron microscope images, and the different gray values of pores and hydration products in the images were used as the basis for distinguishing. Select an SEM image and put it into the software to convert it into a binary image by setting an appropriate threshold. The first step is to set a larger threshold. At this time, many elements that are originally hydration products are identified as pores and appear white. And then reduce the threshold appropriately until the fine hydration products in the pores are just identified, which is a more appropriate threshold at this time. The user determines the threshold for the relationship between pores and hydration products. The microscopic image is usually processed three times in different periods to reduce subjective errors. When the average value of the difference does not exceed 2 degrees, the average value of the three selected thresholds is obtained, which is the final selected threshold. For each series of experimental samples, at least 3 SEM images were recorded, and the average value was taken as the value of digital image quantitative processing. In this study, the SEM image with a magnification of 1000 times was selected as the original image processed by PCAS. Pores of 0~25 μm were collected. The distribution characteristics of irregular pores were quantified by fractal dimension (26). An example of pore structure feature extraction is shown in Figure 1.

Compressive strength
The compressive strength of the CSG before and after freezing and thawing is shown in Figure 2. Compared with the blank group (L1), the pre-freezing compressive strength of the CSG test pieces after being mixed with composite admixtures all decreased to varying degrees. The reasons for this phenomenon are as follows: (1) After the air-entraining agent was introduced, a large number of tiny, closed and uniformly distributed polar bubbles were formed inside the CSG test pieces, which were beneficial to the workability of the mixtures, but reduced the effective compressive area, resulting in different degrees of strength loss of the test pieces mixed with admixtures; (2) both the water reducing agent and the air-entraining agent can reduce water (27); thus, in this study, the actual water-binder ratio was large and interfered with the hardening process of the CSG test pieces. Finally, microcracks appeared on the test pieces, reducing their strength. In accordance with the design requirements of CSG compressive strength specified in the Technical Guideline for Cemented Granular Material Dams (28) (SL 678-2014), the strength loss rate should be less than 20%; that is, the design strength should reach 8 MPa. Figure 2 shows that the test groups L2, L3, L6, L9, L10, and L13 met the compressive strength requirements.
To clarify the influence of composite admixtures on the mechanical properties of CSG, the strength loss rates of the L2, L3, L6, L9, L10, and L13 groups that met the requirements of the specification were analyzed and calculated, as shown in Figure 3. The compressive strength loss rates before and after freezing and thawing were all lower than that of the blank group (22.50%), indicating that adding composite admixtures is beneficial to improving the freezing resistance durability of CSG. Figure 2 and 3 show that, when the range of strength loss is met, the strength loss rate of L9 before and after freezing and thawing is the smallest, at only 0.92%, and its admixtures include 0.008% sodium dodecyl sulfate, 0.04% graphene oxide, 2.0% nano silicon powder, and 0.40% water reducing agent.

Relative dynamic elastic modulus
The dynamic elastic modulus is a main indicator that reflects the internal structural performance of a material. For freezing resistance, the smaller the relative dynamic elastic modulus is, the more obvious the effect of freeze-thaw damage is (29). Similarly, for CSG, the relative dynamic elastic modulus can also characterize its freeze-thaw damage (30). Figure 4 shows changes in the relative dynamic elastic modulus of CSG with different admixtures. Compared with the blank Group L1 (62%), the relative dynamic modulus of CSG with admixtures in the test groups was significantly higher. The relative dynamic elastic moduli of L5, L7, L8, L12, and L13 all exceed 80%, indicating that they have better freezing resistance. However as seen in Figure 2, after 20 freeze-thaw cycles, none of them meets the design standard of compressive strength. Therefore, the relative dynamic elastic modulus cannot be selected as the only indicator to evaluate the freezing resistance of CSG, and other indicators need to be considered together (31)(32)(33).
shows that the harmless pores, less harmful pores and harmful pores are concentrated in the first peak, and the more harmful pores are mainly in the second and third peaks. The pore size distribution curves of the L5, L11, L12, L13, L14, and L16 groups before and after freezing and thawing have relatively larger changes. After freezing and thawing, the first peak is approximately 0.025%, and the second and third peaks increase. For example, the second peak value of the L13 group before freezing was 0.006%, and the pore diameter was 1.17 μm; after freezing, the second peak value rose to 0.0084%, and the pore size increased to 1.45 μm. The pore size distribution curves of L9 and the rest of the test groups in the first peak are basically the same and show little change before and after freezing and thawing, approximately 0.02%, and the changes of the second and third peaks are different, but both move to the right. For example, the second peak pore size of the test pieces in the L9 group before freezing was 1.35 μm; the corresponding value after freezing was 1.45 μm.
The area occupancy of more harmful pores was calculated according to the pore size distribution diagram, as shown in Figure 6. After 20 freeze-thaw cycles, the area occupancy of more harmful pores of CSG increases to varying degrees, and especially in the L5, L11, L12, L13, and L14 groups, it increased by 48.93%, 49.87%, 30.45%, 36.03%, and 75.50%, respectively, which verifies the above analysis. The reason for such phenomena is that at the beginning of freezing and thawing, harmless, less harmful and harmful pores with small sizes are internally filled with pore water, which may generate a frost heaving force due to the differential pressure inside and outside during freezing and thawing. In such cases, the pore wall is damaged, the pore size increases, and many pores are changed to more harmful pores. More harmful pores themselves are large in size and will become even larger under the frost heaving force until the test pieces undergo freeze-thaw damage.
The CSG test pieces experienced a gradually changing damage process under the action of water freeze-thaw cycles, and the damage pattern of its pore structure is shown in Figure 7. With the increase of the number in freeze-thaw cycles, the pore water in the test piece undergoes repeated volume expansion, resulting in the gradual enlargement of the pores and the increase in the degree of irregularity until penetrating cracks develop. The number of damage points inside the specimen increases, which leads to a decrease in its compressive strength.

Fractal characteristics of pore structure in SEM images
CSG images obtained before and after freezing and thawing in the SEM test are shown in Figure 8. Due to the large number of test groups, only partial SEM

NMR pore analysis
NMR technology has been widely used in the study of microscopic pore structures and damage mechanisms of rocks and concrete (34)(35)(36). Figure 5 summarizes the pore size distribution of the test pieces before and after freezing and thawing obtained by NMR. Academician Wu Zhongwei (37) classified pores of 0~0.02 μm into harmless pores, pores of 0.02~0.05 μm into less harmful pores, pores of 0.05~0.2 μm into harmful pores, and pores above 0.2 μm into more harmful pores.   images are shown. Figure 8(a) shows that the microstructure of the blank Group L1 before freezing is needle-shaped, and the surface is flat and has cracks. The energy spectrum analysis reveals that the needle-like substance is ettringite, because it contains Si, Al, Ca and other elements (38,39). Figure 8(b) shows the morphology of the blank Group L1 after freezing. In the needle-like substance fractures, the pores gradually expand into connected cavities, and the compactness of the test pieces decreases. Since the selected SEM image size is 100 µm, combined with the pore size distribution of NMR, it is considered a more harmful pore. Figure 8(c), (e) and (g) show the pre-freezing microstructures of the L5, L12, and L13 groups, which have obvious changes compared with Figure 8(a). This is because different admixtures have different enhancement mechanisms (40,41), and they can improve the cementing effect of the gel or the   weak area of the aggregate interface to different degrees. After freeze-thaw damage, the deterioration of the pore structure in Figure 8(d), (f) and (h) changes compared with Figure 8(b). The connected pores are significantly reduced and gradually become independent pores with smaller sizes, macroscopically indicating the improvement of the freezing resistance of the test samples. This is the same as the pore size distribution law obtained by NMR experiments. Through EDS analysis, the hydration products of CSG mainly included C-S-H gel, AFm and AFt. After 20 freeze-thaw cycles, the needle-shaped C-S-H gradually became thicker and even lumpy, resulting in no cross-linking at the terminal, so that the voids could not be filled, and the compactness and freezing resistance were reduced.
To further clarify the changes in the internal pore structure during the freezing and thawing process, we calculated the fractal characteristics of the pore structure on the selected SEM images before and after freezing and thawing using PCAS software. The results are shown in Table 7.
It can be seen in the table that the fractal dimension of pores improves to varying degrees after 20 freeze-thaw cycles. Among them, the fractal dimension before and after freezing and thawing in the L5, L11, L12, L13, and L14 groups changes more distinctly, which indicates that the pores deteriorate after freezing and thawing. The fractal dimension change rate of the L9 group before and after freezing and thawing was the lowest, at only 0.30%; meanwhile, its compressive strength was also better. Therefore, the fractal dimension of 1000 times SEM images can be used to characterize the freeze-thaw characteristics of CSG.

OPTIMAL GROUP VERIFICATION TEST
In summary, the macroscopic mechanical properties of the L9 group are the best, and the strength loss rate after 20 freeze-thaw cycles is only 0.92%. The micropore structure shows that the area occupancy of more harmful pores is the lowest after freezing and thawing. To verify that the composite admixtures in the optimal Group L9 improved the freezing resistance of CSG, further experiments were performed at different water-binder ratios, sand ratios, and cement contents, during which the compressive strength was used as an indicator to evaluate how the admixtures enhanced the freezing resistance of CSG. The mixing ratios and test results are shown in Table 8. Table 8 shows that the compressive strength of test group L9 mixed with composite admixtures is significantly higher than that of blank group L1, and after 20 freeze-thaw cycles, the compressive strength loss rate ranges within -3.5%~2.7%, which is lower than that (22.5%) of blank group L1 and the effect of a single admixture. Even when the water-binder ratio, sand ratio and cement content are changed, admixtures (containing sodium dodecyl sulfate, graphene oxide, nano silicon powder, and water reducing agent) in the L9 group can significantly enhance the freezing resistance of CSG, Which can improve the service life of the CSG structures in severely cold areas.

CONCLUSIONS AND PROSPECTS
In this paper, based on the macroscopic mechanical properties and microscopic pore structure characteristics, the effects of composite admixtures on the freezing resistance durability index strength, relative dynamic elastic modulus, harmful pore occupancy and fractal dimension of CSG were discussed. The following conclusions were drawn: (1) After 20 freeze-thaw cycles, both the compressive strength and dynamic elastic modulus of CSG decrease. The comparison shows that the strength of L9 (containing sodium dodecyl sulfate, graphene oxide, nano silicon powder, and water reducing agent) meets the design requirements, and its freezing resistance (more than 75% relative to dynamic elastic modulus, lowest strength loss rate) is also good. Such results were confirmed through further experiments, and it was found that under the experimental conditions of different water-binder ratios, sand ratios and cement content, composite admixtures in the L9 group can significantly improve the freezing resistance of CSG.
(2) The NMR test shows that the harmless pores, less harmful pores and harmful pores are mainly in the first peak with a small area occupancy and have little influence on the freezing resistance of CSG with composite admixtures. The area occupancy of more harmful pores (>200 nm) is the main factor affecting the freezing resistance.
(3) Upon analysis of the fractal characteristics of the pore structure of CSG, the fractal dimension increases after 20 freeze-thaw cycles, indicating that the pore structure has deteriorated. Among them, the fractal dimension after freezing in the L9 group is the smallest, signifying that the pore structure of CSG deteriorates less and the freezing resistance remains good.
(4)This article explores the effect of composite admixtures on the freezing resistance of CSG and obtains the optimal composite admixture after comprehensive analysis of various indicators, but it does not studied the enhancement mechanism of a single admixture on the freezing resistance of CSG.