Microstructural Characteristics of Frazil Particles and the Physical Properties of Frazil Ice in the Yellow River, China

: Frazil particles, ice crystals or slushy granules that form in turbulent water, change the freezing properties of ice to create “frazil ice”. To understand the microstructural characteristics of these particles and the physical properties of frazil ice in greater depth, an in situ sampler was designed to collect frazil particles in the Yellow River. The ice crystal microstructural characteristics of the frazil particles (morphology, size, air bubble, and sediment) were observed under a microscope, and their nucleation mechanism was analyzed according to its microstructure. The physical properties of frazil ice (ice crystal microstructure, air bubble, ice density, and sediment content) were also observed. The results showed that these microstructures of frazil particles can be divided into four types: granular, dendritic, needle-like, and serrated. The size of the measured frazil particles ranged from 0.1 to 25 mm. Compared with columnar ice, the crystal microstructure of frazil ice is irregular, with a mean crystal diameter less than 5 mm extending in all directions. The crystal grain size and ice density of frazil ice are smaller than columnar ice, but the bubble and sediment content are larger.


Introduction
The existence of frazil particles contributes greatly to the formation and evolution of river ice sheets [1]. Not only do they influence the ice physical properties, but they may also cause disasters such as ice jams, formed after the frazil ice accumulation, and ice discs, which can reduce the cross-sectional area of a river and further intensify riverbed scouring [2]. They can also raise upstream water levels, which may cause significant flooding [3,4] and the destruction of dykes [5]. In addition, frazil ice poses a threat to bottom fauna [6] and hydraulic structures [7].
Due to uncertainties associated with field experiments, it is extremely difficult to observe frazil particles directly, meaning that the quantitative results of field work are much less than those from laboratory research. Osterkamp [23] has reported on the nucleation of frazil particles, and, Schaefer [24], Osterkamp and Gosink [25] has took underwater images of frazil particles in a river channel. They observed particles of various sizes and gave brief descriptions of their crystalline shape; however, the specific observation method was not described and images of the cold, nucleated particles were not given. Field observations of frazil ice size were also conducted by Dubé et al. [26] and Kempema and Ettema [27]. A new type of diving imaging system, the FrazilCam provided the first quantitative measurements of the size distribution of frazil particles in a river [28], and, in recent years, upward-facing sonar devices have been used to estimate the size of frazil particles in rivers [29][30][31].
In addition, Omstedt and Svensson [32][33][34] used mathematical models to simulate the formation of frazil ice and grease ice on water surfaces. Later, a simulation model for the formation and evolution of ice in open channels was developed by Hammar and Shen [35]. More recently, frazil and anchor ice formation was theoretically simulated by Makkonen and Tikanmaki [36]. It is worth noting that the mechanisms proposed in this study partly differ from the conventional views in the literature.
In summary, due to the different longitudes and latitudes of rivers and the different meteorological and hydrological conditions in flow areas, the size and distribution of frazil particles very [37,38]. Although some researchers have produced images of frazil particles, there have been no detailed reports on their morphology, which, to some extent, affects the accuracy of numerical models of their formation and evolution. As such, it is necessary to measure the microstructural characteristics of frazil particles in different rivers to determine the properties of frazil ice (ice crystal microstructures, air bubbles, ice density, and sediment content) [39], which can affect the physical, mechanical, thermodynamic, and optical properties of ice [40][41][42][43].
Here, a field observation of the Yellow River ice was carried out and its findings are presented in this paper. The river is located in a mid-latitude region in China and is known as the river with the world's largest sediment content. In previous studies, the ice microstructure of the river ice was observed [44,45], but our study present the results of observing the microstructural and physical properties of frazil particles and frazil ice in the Inner Mongolia section of the Yellow River for 46 days during the winter of 2020-2021.

Technology of Unidirectional Freezing
When an ice block is artificially frozen, such as when a bucket filled with water is directly placed in a low-temperature environment, heat will be transmitted to the inside of the bucket from all directions. Ice crystals will also start to grow from the surface, bottom, and side walls of the bucket toward the inside. To achieve unidirectional ice freezing, a unidirectional freezing experiment was carried out in an indoor low-temperature laboratory by Li et al. [46]. According to the thermostatic technology proposed by Li, a freezing device to prevent ice growing from the plexglass bucket's side and bottom was designed ( Figure 1a). In order to ensure the plexiglass was in full contact with the water, we punched some holes through the bracket, besides, the plexglass had no bottom, i.e., a plexiglass tube was used. The water in the bucket and the plexiglass were the same medium. The working principle is that the water served as an insulation layer, and the plexiglass bucket was surrounded by water that was not directly in contact with the cold air. As the freezing point of water is 0 • C, as long as the water in contact with the side walls and bottom of the plexiglass bucket did not freeze, ice crystals in the plexiglass bucket will not grow from the side walls and bottom. It should be noted that, during the freezing process, the resistance wire was heated every 1 h to melt the ice in tube to release the pressure caused by freezing. Finally, a unidirectional frozen ice crystal may be obtained, as shown in Figure 1b

Preparation of Frazil Ice Materials
The field testing of frazil particles and frazil ice was carried out at Toudaoguai section in the Inner Mongolia of the Yellow River. Based on the field water extractor structure [47], a frazil particle sampler was designed and the process of frazil particles collection is shown in Figure 2. First of all, the frazil particle extraction position was selected and the position was marked by a GPS. Second, two holes were manually drilled: the ice hole upstream was used to collect the particles, while the downstream hole was used to pump water into a plastic bucket. Third, wooden brackets were fixed in the plastic bucket, and the plastic bucket was filled with water from the Yellow River. Then, the bottom of the frazil particle sampler was gently placed along the edge of the upstream hole into the bottom of the frazil particles. Next, a plexiglass tube was placed in the hole alongside a lead screw. In the following step, the plexiglass cylinder was slowly moved up and down to ensure that it was well sealed with the bottom of the sampler. Finally, the frazil particles were lifted out of the water and placed on the wooden brackets in the plastic bucket for subsequent observation or unidirectional freezing of frazil ice in a natural environment. During the winter of 2020-2021, this device was used in the Yellow River field experiments, and observations of the ice crystal microstructure at a depth of 40-54 cm showed that the ice in the sampler was formed with unidirectional freezing (Figure 11).

Preparation of Frazil Ice Materials
The field testing of frazil particles and frazil ice was carried out at Toudaoguai section in the Inner Mongolia of the Yellow River. Based on the field water extractor structure [47], a frazil particle sampler was designed and the process of frazil particles collection is shown in Figure 2. First of all, the frazil particle extraction position was selected and the position was marked by a GPS. Second, two holes were manually drilled: the ice hole upstream was used to collect the particles, while the downstream hole was used to pump water into a plastic bucket. Third, wooden brackets were fixed in the plastic bucket, and the plastic bucket was filled with water from the Yellow River. Then, the bottom of the frazil particle sampler was gently placed along the edge of the upstream hole into the bottom of the frazil particles. Next, a plexiglass tube was placed in the hole alongside a lead screw. In the following step, the plexiglass cylinder was slowly moved up and down to ensure that it was well sealed with the bottom of the sampler. Finally, the frazil particles were lifted out of the water and placed on the wooden brackets in the plastic bucket for subsequent observation or unidirectional freezing of frazil ice in a natural environment.

Observation Method of Frazil Particles Morphology
A digital image processing system, developed by Morris in 2003 [48] and improved and modified by Clark [14] and McFarlane et al. [28,49], was used to analyze the images taken in the series of experiments. A similar method of image processing can be found in [50]. Before obtaining the images, the frazil particles in the sampler needed to be separated. Through many experiments, we proved that the most appropriate temperature for separating frazil particles is between −4 and −8 °C. When the temperature is lower than −8 °C, the particles freeze together quickly when taken out and can not be separated easily. When the temperature is higher than −4 °C, the separated particles melt and easily evaporate. When the temperature was between −4 and −8 °C, we took some particles from the sampler and put them on a black acrylic plate with a ruler, where they were gently separated by tweezers. It is worth noting that while separating the frazil particles, it was necessary to throw away those that had been artificially damaged. Next, the separated frazil particles were placed on a microscope platform. Magnification was set at about 200× until they could be clearly observed. At the same time, the frazil particle images were recorded by a computer and the images were enhanced by Adobe Photoshop CS. Finally, the microstructures of the frazil particle images were statistically analyzed by a MATLAB algorithm. Figure 3 shows the process for the frazil particle observation.

Ice Crystal Microstructure and Air Bubble Observation Method
As shown in Figure 4a, the ice block in the sampler was obtained. In the field laboratory, where the temperature was below −6°C, ice slices were made by referring to the observation technique for ice crystal microstructure shown in [44]. The standard

Observation Method of Frazil Particles Morphology
A digital image processing system, developed by Morris in 2003 [48] and improved and modified by Clark [14] and McFarlane et al. [28,49], was used to analyze the images taken in the series of experiments. A similar method of image processing can be found in [50]. Before obtaining the images, the frazil particles in the sampler needed to be separated. Through many experiments, we proved that the most appropriate temperature for separating frazil particles is between −4 and −8 • C. When the temperature is lower than −8 • C, the particles freeze together quickly when taken out and can not be separated easily. When the temperature is higher than −4 • C, the separated particles melt and easily evaporate. When the temperature was between −4 and −8 • C, we took some particles from the sampler and put them on a black acrylic plate with a ruler, where they were gently separated by tweezers. It is worth noting that while separating the frazil particles, it was necessary to throw away those that had been artificially damaged. Next, the separated frazil particles were placed on a microscope platform. Magnification was set at about 200× until they could be clearly observed. At the same time, the frazil particle images were recorded by a computer and the images were enhanced by Adobe Photoshop CS. Finally, the microstructures of the frazil particle images were statistically analyzed by a MATLAB algorithm. Figure 3 shows the process for the frazil particle observation.

Observation Method of Frazil Particles Morphology
A digital image processing system, developed by Morris in 2003 [48] and improved and modified by Clark [14] and McFarlane et al. [28,49], was used to analyze the images taken in the series of experiments. A similar method of image processing can be found in [50]. Before obtaining the images, the frazil particles in the sampler needed to be separated. Through many experiments, we proved that the most appropriate temperature for separating frazil particles is between −4 and −8 °C. When the temperature is lower than −8 °C, the particles freeze together quickly when taken out and can not be separated easily. When the temperature is higher than −4 °C, the separated particles melt and easily evaporate. When the temperature was between −4 and −8 °C, we took some particles from the sampler and put them on a black acrylic plate with a ruler, where they were gently separated by tweezers. It is worth noting that while separating the frazil particles, it was necessary to throw away those that had been artificially damaged. Next, the separated frazil particles were placed on a microscope platform.
Magnification was set at about 200× until they could be clearly observed. At the same time, the frazil particle images were recorded by a computer and the images were enhanced by Adobe Photoshop CS. Finally, the microstructures of the frazil particle images were statistically analyzed by a MATLAB algorithm. Figure 3 shows the process for the frazil particle observation.

Ice Crystal Microstructure and Air Bubble Observation Method
As shown in Figure 4a, the ice block in the sampler was obtained. In the field laboratory, where the temperature was below −6°C, ice slices were made by referring to the observation technique for ice crystal microstructure shown in [44]. The standard

Ice Crystal Microstructure and Air Bubble Observation Method
As shown in Figure 4a, the ice block in the sampler was obtained. In the field laboratory, where the temperature was below −6 • C, ice slices were made by referring to the observation technique for ice crystal microstructure shown in [44]. The standard method for observing ice microstructures is the universal stage proposed by Langway [51], which Crystals 2021, 11, 617 5 of 13 was used to observe the ice crystal microstructure and air bubbles. In a dark environment, the ice slices were placed on a universal stage as shown in Figure 4b. The ice crystal microstructure was observed under polarized light and air bubbles were observed under natural light; meanwhile, the ice crystal microstructure and air bubble images were saved.
Crystals 2021, 11, x FOR PEER REVIEW 5 of 14 method for observing ice microstructures is the universal stage proposed by Langway [51], which was used to observe the ice crystal microstructure and air bubbles. In a dark environment, the ice slices were placed on a universal stage as shown in Figure 4b. The ice crystal microstructure was observed under polarized light and air bubbles were observed under natural light; meanwhile, the ice crystal microstructure and air bubble images were saved.

Ice Density and Mud Content Observation Methods
The measurement of ice density mainly includes mass/volume, submersion, specific gravity, freeboard, and ice thickness method [52]. Considering that the sediment content in ice also needs to be measured, a mass/volume method was used to obtain the ice density. Along the growth direction of the ice thickness, the ice sample was segmented by a bone saw into small test blocks with a height of 5-10 cm from the surface to the bottom. The test block was processed into a cuboid with a size of 10 × 10 cm × height. Then, the length and height of the test block were measured with a vernier caliper; its mass was measured with an electronic scale; and its density was calculated using the ρ = m/v formula. After the test block was melted in a plastic box, the dry sediment weight was measured after filtering and drying, and the sediment content per unit volume of ice was obtained.

Frazil Ice Nucleation
In the observation experiment of the frazil particles, the ice particles shown in Figure  5a and the materials wrapped in frazil particles, shown in Figure 5b, were locally enlarged under a microscope.
It can be seen in Figure 5a that the equivalent diameter of the enlarged ice particles is 1.2 mm. There were many microbubbles in the ice particles after local enlargement because of the highly turbulent natural water [53,54]. Makkonen and Tikanmaki [36], and Chow et al. [55] suggested that the nucleation of ice may be produced by turbulence in slightly supercooled cases, and Chow et al. [55] suggested that the bubbles that initiated nucleation did not disappear. As such, these bubbles were probably wrapped in the crystal during the nucleation of the frazil particles. The locally enlarged image in Figure  5b shows numerous sediment particles and other impurities. Due to the limited magnification of the microscope, it was impossible to measure the size of these particles accurately, but it could be determined that the equivalent diameters of these particles were less than 0.1 mm. McFarlane et al. [28] and Kempema et al. [56] found that the sediment concentration in the water decreases when frazil particles are formed. This may be caused by part of the sediment is wrapped in the frazil crystals, or by frazil particles being formed with sediment as a nucleus. In addition, Reimnitz et al. [22] observed that frazil flocs can collect more sediment.

Ice Density and Mud Content Observation Methods
The measurement of ice density mainly includes mass/volume, submersion, specific gravity, freeboard, and ice thickness method [52]. Considering that the sediment content in ice also needs to be measured, a mass/volume method was used to obtain the ice density. Along the growth direction of the ice thickness, the ice sample was segmented by a bone saw into small test blocks with a height of 5-10 cm from the surface to the bottom. The test block was processed into a cuboid with a size of 10 × 10 cm × height. Then, the length and height of the test block were measured with a vernier caliper; its mass was measured with an electronic scale; and its density was calculated using the = m/v formula. After the test block was melted in a plastic box, the dry sediment weight was measured after filtering and drying, and the sediment content per unit volume of ice was obtained.

Frazil Ice Nucleation
In the observation experiment of the frazil particles, the ice particles shown in Figure 5a and the materials wrapped in frazil particles, shown in Figure 5b, were locally enlarged under a microscope. Hence, we also suggested that frazil ice nuclei can be ice particles or other cold particles such as microbubbles, organic material, or sediment. In addition, Hanley [8], Daly [9], Osterkamp and Gosink [25], and McFarlane et al. [28] also suggested that ice crystals and snow that have fallen from the air into rivers could act as seed crystals.

Microstructural Characteristics of Single Frazil Particles
The area, perimeter, major axis, and minor axis data for each frazil particle were extracted by MATLAB, and the equivalent diameter, slenderness ratio, and roundness of each frazil particle were calculated using the following formulae.
where D is the equivalent diameter; Sr is the slenderness ratio; R is the roundness; S is the It can be seen in Figure 5a that the equivalent diameter of the enlarged ice particles is 1.2 mm. There were many microbubbles in the ice particles after local enlargement because of the highly turbulent natural water [53,54]. Makkonen and Tikanmaki [36], and Chow et al. [55] suggested that the nucleation of ice may be produced by turbulence in slightly supercooled cases, and Chow et al. [55] suggested that the bubbles that initiated nucleation did not disappear. As such, these bubbles were probably wrapped in the crystal during the nucleation of the frazil particles. The locally enlarged image in Figure 5b shows numerous sediment particles and other impurities. Due to the limited magnification of the microscope, it was impossible to measure the size of these particles accurately, but it could be determined that the equivalent diameters of these particles were less than 0.1 mm. McFarlane et al. [28] and Kempema et al. [56] found that the sediment concentration in the water decreases when frazil particles are formed. This may be caused by part of the sediment is wrapped in the frazil crystals, or by frazil particles being formed with sediment as a nucleus. In addition, Reimnitz et al. [22] observed that frazil flocs can collect more sediment.
Hence, we also suggested that frazil ice nuclei can be ice particles or other cold particles such as microbubbles, organic material, or sediment. In addition, Hanley [8], Daly [9], Osterkamp and Gosink [25], and McFarlane et al. [28] also suggested that ice crystals and snow that have fallen from the air into rivers could act as seed crystals.

Microstructural Characteristics of Single Frazil Particles
The area, perimeter, major axis, and minor axis data for each frazil particle were extracted by MATLAB, and the equivalent diameter, slenderness ratio, and roundness of each frazil particle were calculated using the following formulae.
where D is the equivalent diameter; S r is the slenderness ratio; R is the roundness; S is the area; M 1 is the length of the minor axis; M 2 is the length of major axis; and l is the perimeter. Based on the analysis and the threshold segmentation of the shape, diameter, slenderness radio, and roundness of the frazil particles in the Yellow River, the frazil particles can be divided into four types as shown in Figure 6 and Table 1, i.e., granular, dendritic, needle-like, and serrated.  [28] who found the percentage of the major frazil particles to be between 61 and 87%. The serrated and needle-like frazil particles had similar percentages of around 10%. The percentage of dendritic frazil particles was the lowest with less than 2%. As shown in Figure 6e, the structure of the dendritic frazil particles was obviously unstable. Under the action of turbulence, dendrites are easily broken. In addition, our findings were similar to those of Schaefer [24]. Until the diameters of the frazil particles reached 2 mm, it was rare to find secondary protruding structures developing around the edges. As the frazil particles grew, though, such secondary structures appeared, and may formed the dendritic and serrated particles, as shown in Figure 6e-g.  The diameter, slenderness ratio, and roundness values for the granular, serrated and dendritic frazil particles were fitted by a log-normal distribution; however, the content of dendritic frazil particles was too low, and the observation of dendritic frazil particles was insufficient, which did not allow a reasonable fit for the data; thus, the fitting of data for dendritic frazil particles was not performed. All of the characteristic parameter distributions were well described by a log-normal distribution, which was consistent with the measurements of many previous studies [14,20,28,49]. As can be seen from Table  1, the distribution range of slenderness ratio and roundness were similar, and their  Granular frazil particles are shown in Figure 6a-d, and are similar to an ellipse. Compared with other types of frazil particles, their equivalent diameters were the smallest, but the slenderness ratio and the roundness were the largest. The equivalent diameters of the serrated frazil particles shown in Figure 6g,h were larger, and the equivalent diameter of the largest serrated frazil particle can be up to 25 mm. A needle-like frazil particle is shown in Figure 6f. Its slender shape results in the lowest slenderness ratio and roundness, which were both less than 0.2. The dendritic frazil particle shown in Figure 6e features a larger grain perimeter, which leads to smaller roundness, and a larger area results in a larger equivalent diameter than the granular and needle-like frazil particles.
In the research of frazil particles mentioned in the introduction of this paper, the field measurements of the major particle size ranged from 0.1 to 6 mm, and the laboratory measurements of that ranged from 0.022 to 5.5 mm. Here, the mean equivalent diameter of the granular frazil particles was 3.2 mm, and the standard deviation was 2.1 mm, which was similar to previous observations, especially the mean diameter of 3.15 mm and the standard deviation of 2.5 mm observed by Morse and Richard in the field [37].
The percentages of different particles were also calculated, among which the percentage of granular frazil particles was the largest (77.8%), indicating that the majority of frazil particles observed in the field were granular. In previous studies, the majority of particles were described by researchers as disk-shaped. In this study, we found that there were still a large number of particles where the boundaries were not smoothly disk-shaped, thus were described as granular. As such, the percentage of major frazil particles in the Yellow River were consistent with the laboratory measurements of McFarlane et al. (2015) [20], who found that the percentage of the major frazil particles was 75%, and the field measurements of McFarlane et al. (2017) [28] who found the percentage of the major frazil particles to be between 61 and 87%. The serrated and needle-like frazil particles had similar percentages of around 10%. The percentage of dendritic frazil particles was the lowest with less than 2%. As shown in Figure 6e, the structure of the dendritic frazil particles was obviously unstable. Under the action of turbulence, dendrites are easily broken.
In addition, our findings were similar to those of Schaefer [24]. Until the diameters of the frazil particles reached 2 mm, it was rare to find secondary protruding structures developing around the edges. As the frazil particles grew, though, such secondary structures appeared, and may formed the dendritic and serrated particles, as shown in Figure 6e-g.
The diameter, slenderness ratio, and roundness values for the granular, serrated and dendritic frazil particles were fitted by a log-normal distribution; however, the content of dendritic frazil particles was too low, and the observation of dendritic frazil particles was insufficient, which did not allow a reasonable fit for the data; thus, the fitting of data for dendritic frazil particles was not performed. All of the characteristic parameter distributions were well described by a log-normal distribution, which was consistent with the measurements of many previous studies [14,20,28,49]. As can be seen from Table 1, the distribution range of slenderness ratio and roundness were similar, and their log-normal distribution results also showed a high degree of similarity. The log-normal distribution fitting results for the diameter and roundness of granular frazil particles are shown in Figure 7.

Microstructural Characteristics of Accumulated Frazil Particles
Frazil particles are free-floating particles in early development. As their number increases, they agglomerated together with weakly bonding in turbulent streams. Some

Microstructural Characteristics of Accumulated Frazil Particles
Frazil particles are free-floating particles in early development. As their number increases, they agglomerated together with weakly bonding in turbulent streams. Some representative frazil particles that agglomerated together are shown in Figures 8-10. Under the effect of turbulence, varying agglomeration angles were formed and with an increase in the number of agglomerations, the maximum angle between two adjacent frazil particles gradually decreases and should not exceed 150 • ; for three frazil particles, the angle should not exceed 120 • . When multiple frazil particles aggregated, the maximum angle between two adjacent frazil particles should not exceed 90 • . The shape and size of agglomerated frazil particles will change. An increase in agglomerated frazil particles will also produce an increase in the size, roundness, and slenderness ratio. Finally, due to the influence of temperature and submersion in water, the particles will gradually freeze together to form frazil ice.

Microstructural Characteristics of Accumulated Frazil Particles
Frazil particles are free-floating particles in early development. As their number increases, they agglomerated together with weakly bonding in turbulent streams. Some representative frazil particles that agglomerated together are shown in Figures 8-10. Under the effect of turbulence, varying agglomeration angles were formed and with an increase in the number of agglomerations, the maximum angle between two adjacent frazil particles gradually decreases and should not exceed 150°; for three frazil particles, the angle should not exceed 120°. When multiple frazil particles aggregated, the maximum angle between two adjacent frazil particles should not exceed 90°. The shape and size of agglomerated frazil particles will change. An increase in agglomerated frazil particles will also produce an increase in the size, roundness, and slenderness ratio. Finally, due to the influence of temperature and submersion in water, the particles will gradually freeze together to form frazil ice.

Microstructural Characteristics of Accumulated Frazil Particles
Frazil particles are free-floating particles in early development. As their number increases, they agglomerated together with weakly bonding in turbulent streams. Some representative frazil particles that agglomerated together are shown in Figures 8-10. Under the effect of turbulence, varying agglomeration angles were formed and with an increase in the number of agglomerations, the maximum angle between two adjacent frazil particles gradually decreases and should not exceed 150°; for three frazil particles, the angle should not exceed 120°. When multiple frazil particles aggregated, the maximum angle between two adjacent frazil particles should not exceed 90°. The shape and size of agglomerated frazil particles will change. An increase in agglomerated frazil particles will also produce an increase in the size, roundness, and slenderness ratio. Finally, due to the influence of temperature and submersion in water, the particles will gradually freeze together to form frazil ice.

Microstructural Characteristics of Accumulated Frazil Particles
Frazil particles are free-floating particles in early development. As their number increases, they agglomerated together with weakly bonding in turbulent streams. Some representative frazil particles that agglomerated together are shown in Figures 8-10. Under the effect of turbulence, varying agglomeration angles were formed and with an increase in the number of agglomerations, the maximum angle between two adjacent frazil particles gradually decreases and should not exceed 150°; for three frazil particles, the angle should not exceed 120°. When multiple frazil particles aggregated, the maximum angle between two adjacent frazil particles should not exceed 90°. The shape and size of agglomerated frazil particles will change. An increase in agglomerated frazil particles will also produce an increase in the size, roundness, and slenderness ratio. Finally, due to the influence of temperature and submersion in water, the particles will gradually freeze together to form frazil ice.

Ice Crystal Microstructure and Air Bubble in Frazil Ice
The ice crystal microstructure and air bubbles are shown in Figure 11. The microstructure of the ice corresponds to the original block of ice shown in Figure 5a, which shows that an ice thickness of 34 cm is an obvious boundary; the color of the original ice block at 0-34 cm is turbid; and the color of the original ice block at 34-54 cm is transparent. From the microstructure of the ice, it is obvious that the ice crystal microstructure at 0-34 cm is completely different from that at 34-54 cm.
Frazil ice was found in the 0-34 cm range. No mater whether the ice crystal microstructure was perpendicular or parallel to the ice surface, the shapes of ice crystals were extremely irregular. The grain sizes of most of the ice crystals were small, accompanied by a small number with a large grain size, and there were overlapping crystals at the grain edge. This was related to the formation of frazil ice formed by freezing frazil particles through the connection with Yellow River water. Due to the different shape of frazil particles, the ice crystals may not form one individual frazil particle, but may instead contain overlapping parts of other frazil particles. As such, the crystal microstructure of frazil ice is messy. In addition, some crystals in the frazil ice were inclined at certain angles, but the direction of Crystals 2021, 11, 617 9 of 13 the tilt was irregular, which may be related to the effect of turbulence under the ice cover of the Yellow River during the freezing period. According to the classification of river ice and lake ice by Michel and Ramseier [57], this ice structure belongs to the P3 category. To further analyze the characteristics of the ice crystal grain size and bubble content, the ice crystal microstructure and air bubble images in Figure 11 were processed by Adobe Photoshop CS and MATLAB, and, their ice crystal grain size and bubble content were measured. The variation in the mean equivalent diameter of the ice crystals and bubble content in ice along the depth was obtained as shown in Figure 12.
It can be seen that the mean equivalent diameters of frazil ice crystals ranged from 2.2 to 2.6 mm and fluctuated slightly according to depth. The equivalent diameters of columnar ice ranged from 3.6 to 24.8 mm, which was slightly larger than that previously reported in other regions [44], and obviously increased with depth. With the growth of columnar ice, ice thickness gradually increased, and thus heat conduction became weaker. The growth rate of ice crystals became slower, such that the ice crystals had sufficient time to develop. These ice crystals restricted one another and competed for growth. The number of ice crystals constantly decreased as their equivalent diameter of the ice crystals grew larger; however, the frazil ice here was formed by a combination of ice particles and Yellow River water, where the equivalent diameters of frazil ice were mainly related to the size of the frazil particles; and thus, the equivalent diameters of frazil ice are small.
Regarding the bubble content in ice, the bubble content in frazil ice varied from 3.72 to 5.68% and fluctuated at depth larger than that of columnar ice; however, the bubble content in columnar ice varied from 1.63 to 2.07% and decreased gradually with depth. This was consistent with the statistical results of Zhang et al. [59] for Yellow River ice from 2015 to 2020, which suggested that the bubble content of frazil ice is 5.29 ± 2.57% The ice crystals in the range of 34-37 cm were regular. It could be seen that the ice crystals perpendicular to the ice surface were close to rectangles, while the ice crystals parallel to the ice surface were granular with clear boundaries. Although their sizes were small, they were uniform and there were no large ice crystals. This represents the transition stage from granular ice to columnar ice, and the ice crystals gradually extended along the direction of heat conduction. With the increase of ice thickness, a preferred growth orientation of some crystals appeared due to geometric selection. The growth in this preferred direction extended past the growth in other directions, causing the formation of columnar crystals [58], which can be seen in Figure 11 when the ice thickness is 37-54 cm. The ice crystals perpendicular to the ice surface continuously extended along the direction of heat conduction, while the ice crystals parallel to the ice surface were nearly elliptical and featured clear boundaries. According to the classification of river ice and lake ice by Michel and Ramseier [57], this ice structure belongs to the S1 category.
To further analyze the characteristics of the ice crystal grain size and bubble content, the ice crystal microstructure and air bubble images in Figure 11 were processed by Adobe Photoshop CS and MATLAB, and, their ice crystal grain size and bubble content were measured. The variation in the mean equivalent diameter of the ice crystals and bubble content in ice along the depth was obtained as shown in Figure 12. the water, forming a air-water mixing layer where rapid and sudden freezing captures most of the air. After the ice sheet is formed, bubbles dissolved in water are released and rise to the downward-moving freezing front (the ice-water interface) to be incorporated into the ice. As shown in Figure 5a, a large number of bubbles were wrapped in ice particles during supercooling, which is when the bonding force between ice crystals is lower, and, this caused frazil ice pores increase. This is the main reason that there are more bubbles in frazil ice than in columnar ice.

Ice Density and Sediment Content in Ice
The variation in ice density and sediment content with depth is shown in Figure 13. Note that the density values of the frazil and columnar ice were both lower than that of pure ice (917 kg/m 3 ) [40]. The density of frazil ice ranged from 856.4 to 880.3 kg/m 3 , which was less than that of columnar ice, which ranged from 891.1 to 896.0 kg/m 3 ; however, the sediment content of frazil ice ranged from 0.538 to 0.854 kg/m 3 , which was larger than that of columnar ice, which ranged from 0.063 to 0.104 kg/m 3 . These results were also consistent with the statistical results of Zhang et al. [59]. The density and sediment content of frazil ice fluctuated strongly, but the general trend was a gradual increase with depth. It was also found that the trend for frazil ice density was similar to the trend for sediment content in frazil ice. When the sediment content in ice was large, the ice density was also relatively large. In contrast, the sediment content in the columnar ice was very low, but the density of the columnar ice gradually increased with depth. This is because ice density is not only affected by sediment content, but also by bubble content. The bubble content of frazil ice was larger and its density was lower. The influence of sediment content on columnar ice density was not significant; and the bubble content of columnar ice was very low; thus, the ice density of columnar ice was greater than that of frazil ice. It can be seen that the mean equivalent diameters of frazil ice crystals ranged from 2.2 to 2.6 mm and fluctuated slightly according to depth. The equivalent diameters of columnar ice ranged from 3.6 to 24.8 mm, which was slightly larger than that previously reported in other regions [44], and obviously increased with depth. With the growth of columnar ice, ice thickness gradually increased, and thus heat conduction became weaker. The growth rate of ice crystals became slower, such that the ice crystals had sufficient time to develop. These ice crystals restricted one another and competed for growth. The number of ice crystals constantly decreased as their equivalent diameter of the ice crystals grew larger; however, the frazil ice here was formed by a combination of ice particles and Yellow River water, where the equivalent diameters of frazil ice were mainly related to the size of the frazil particles; and thus, the equivalent diameters of frazil ice are small.
Regarding the bubble content in ice, the bubble content in frazil ice varied from 3.72 to 5.68% and fluctuated at depth larger than that of columnar ice; however, the bubble content in columnar ice varied from 1.63 to 2.07% and decreased gradually with depth. This was consistent with the statistical results of Zhang et al. [59] for Yellow River ice from 2015 to 2020, which suggested that the bubble content of frazil ice is 5.29 ± 2.57% and that of columnar ice is 2.57 ± 1.09%. The formation of bubbles in ice is related to the formation of ice sheets, and is primarily produced in two ways [60,61]. In the initial stage of freezing, powerful wind and water turbulence bring a large amount of air into the water, forming a air-water mixing layer where rapid and sudden freezing captures most of the air. After the ice sheet is formed, bubbles dissolved in water are released and rise to the downward-moving freezing front (the ice-water interface) to be incorporated into the ice. As shown in Figure 5a, a large number of bubbles were wrapped in ice particles during supercooling, which is when the bonding force between ice crystals is lower, and, this caused frazil ice pores increase. This is the main reason that there are more bubbles in frazil ice than in columnar ice.

Ice Density and Sediment Content in Ice
The variation in ice density and sediment content with depth is shown in Figure 13. Note that the density values of the frazil and columnar ice were both lower than that of pure ice (917 kg/m 3 ) [40]. The density of frazil ice ranged from 856.4 to 880.3 kg/m 3 , which was less than that of columnar ice, which ranged from 891.1 to 896.0 kg/m 3 ; however, the sediment content of frazil ice ranged from 0.538 to 0.854 kg/m 3 , which was larger than that of columnar ice, which ranged from 0.063 to 0.104 kg/m 3 . These results were also consistent with the statistical results of Zhang et al. [59]. The density and sediment content of frazil ice fluctuated strongly, but the general trend was a gradual increase with depth. It was also found that the trend for frazil ice density was similar to the trend for sediment content in frazil ice. When the sediment content in ice was large, the ice density was also relatively large. In contrast, the sediment content in the columnar ice was very low, but the density of the columnar ice gradually increased with depth. This is because ice density is not only affected by sediment content, but also by bubble content. The bubble content of frazil ice was larger and its density was lower. The influence of sediment content on columnar ice density was not significant; and the bubble content of columnar ice was very low; thus, the ice density of columnar ice was greater than that of frazil ice.

Conclusions
The ice conditions of the Yellow River are complex, and it is necessary to understand the physical properties of Yellow River ice, which is of great significance for further study. The microstructural characteristics of frazil particles and the physical properties of frazil ice in the Yellow River were observed in this work, and the main conclusions are summarized below.
(1) Frazil ice nuclei can be ice particles or other cold particles, such as microbubbles, organic material, or sediment particles; (2) Based on their shape, diameter, slenderness ratio, and roundness, the frazil particles may be divided into four types as seen in Table 1, where each has its own characteristic parameters; (3) A large number of frazil particles may agglomerate in turbulent streams. As the number of agglomerated particles increases, the maximum angle between two adjacent frazil particles will gradually decrease; (4) Following the unidirectional freezing technology summarized in the laboratory, unidirectionally_frozen frazil and columnar ice were obtained. Compared with the columnar ice, the ice crystal structure of frazil ice is disorderly; the crystal size is small; and the diameter is 2.2-2.6 mm; (5) Ice density is affected by both the bubble and sediment contents. The nucleation mechanism of frazil particles, coupled with the large amount of sediment in the Yellow River, resulted in significantly larger bubbles and sediment content in frazil ice than in columnar ice.

Conclusions
The ice conditions of the Yellow River are complex, and it is necessary to understand the physical properties of Yellow River ice, which is of great significance for further study. The microstructural characteristics of frazil particles and the physical properties of frazil ice in the Yellow River were observed in this work, and the main conclusions are summarized below.
(1) Frazil ice nuclei can be ice particles or other cold particles, such as microbubbles, organic material, or sediment particles; (2) Based on their shape, diameter, slenderness ratio, and roundness, the frazil particles may be divided into four types as seen in Table 1, where each has its own characteristic parameters; (3) A large number of frazil particles may agglomerate in turbulent streams. As the number of agglomerated particles increases, the maximum angle between two adjacent frazil particles will gradually decrease; (4) Following the unidirectional freezing technology summarized in the laboratory, uni-directionally_frozen frazil and columnar ice were obtained. Compared with the columnar ice, the ice crystal structure of frazil ice is disorderly; the crystal size is small; and the diameter is 2.2-2.6 mm; (5) Ice density is affected by both the bubble and sediment contents. The nucleation mechanism of frazil particles, coupled with the large amount of sediment in the Yellow River, resulted in significantly larger bubbles and sediment content in frazil ice than in columnar ice. Data Availability Statement: All the data are contained within this article.