Comparison of the effects of a sulfuric acid environment on traditionally manufactured and additive manufactured stainless steel 316L alloy

Additive Manufacturing 23 (2018) 272–286

Contents lists available at ScienceDirect

Additive Manufacturing journal homepage: www.elsevier.com/locate/addma

Comparison of the effects of a sulfuric acid environment on traditionally manufactured and additive manufactured stainless steel 316L alloy

T

⁎

Jacob T. Millera, Holly J. Martina, , Edward Cudjoeb a b

Department of Civil/Environmental and Chemical Engineering, Youngstown State University (YSU), One University Plaza, Youngstown, OH, 44555, United States Department of Mechanical and Industrial Engineering, Youngstown State University (YSU), One University Plaza, Youngstown, OH, 44555, United States

A R T I C LE I N FO

A B S T R A C T

Keywords: Stainless steel Laser powder bed fusion Sulfuric acid Acid corrosion Hydrogen embrittlement

The effects on the surface and mechanical properties of stainless steel AISI316L dogbones created using either traditional manufacturing (TM) or laser powder bed fusion (LPBF) exposed to 0.75 M sulfuric acid solution over 2184 h were studied. General corrosion was not a major form of corrosion, based on surface feature changes, surface roughness, and mass loss for either method. No change to the mechanical properties occurred for the TM samples. Both tensile stress and strain decreased for the LPBF samples. The decrease was caused by hydrogen embrittlement, due to the formation of large brittle particles, as demonstrated by scanning electron microscopy.

1. Introduction Stainless steels, which are steel alloys containing a minimum of 11% chromium by weight [1], are incredibly versatile because they have the strength of typical steel products while also remaining resistant to many corrosive environments. The chromium passivates in the presence of oxygen to form a thin layer of chromium oxide, which protects the metal underneath [2]. This film reforms even if the surface of the metal is scratched or damaged. Within the various grades of stainless steel, the 300 series contains at least 16% chromium and 6% nickel by weight [3]. The specific grade in this research, SS 316L, contains 16–18% chromium, 10–14% nickel, 10% nitrogen, 2–3% molybdenum, 2% manganese, and 0.3% carbon by weight, as compared to 0.8% by weight in 316 [4]. SS 316 is used in environments where both strength and corrosion resistance are required, such as aerospace, pharmaceuticals, cutlery, and marine applications [5], while SS 316L is used for improved performance under corrosive environments, especially where potential leaching of carbon from the steel would decrease strength [1]. Because of the mechanical properties and corrosion resistance of stainless steel, various grades are currently being investigated for use in Additive Manufacturing (AM), which can build complex parts layer-bylayer [6]. AM has experienced vast improvements in recent years, by decreasing porosity and improving mechanical properties of a variety of metals [6]. Previously, the parts produced were more porous, with altered mechanical properties, including reduce tensile stress, strain, and brittle fractures [7,8], caused by differences in laser power, laser speed, and powder production processes [9,10]. Within AM, there are multiple

⁎

methods of deposition, including Laser Beam Melting (LBM), Electron Beam Melting (EBM), and Laser Metal Deposition (LMD) [6]. Within the LBM group is laser powder bed fusion (LPBF), which uses a metal powder laid out in a very thin layer, usually around 20–100 μm thick [6,11]. A laser above the stage is used to heat the powder until it nearly melts and then solidifies [6,12]. After the powder solidifies and before another layer of powder is laid down, all of the loose powder is recycled back into the printer, the stage moves down, another powder layer is applied on top of the solid layer, and the laser is activated to melt the new powder layer. This process is repeated, building the item layer by layer, with the geometries specified in the SolidWorks or ComputerAided Design (CAD) file. As with all AM methods, LPBF gives manufacturers the ability to print out a single metal part with intricate geometries that would be impossible to produce using traditional machining techniques, as well as eliminating nearly all wasted material typically produced in traditional machining processes [13]. LPBF differs from Selective Laser Melting (SLM), which is also a type of LBM, because the parts are made in an environment of inert gas [14]. One of the major drawbacks of AM stems from the manufacturing process, since the metal powder is not melted and solidified in the same way as traditional manufacturing methods. Specifically, due to the rapid solidification and cooling, the molecules in the metal are arranged differently, in higher energy state, resulting in high residual stresses within the metal [15–17]. With LPBF and other LBM methods, the initial layer contains melt pools comprised of supersaturated metal, since the elements could not diffuse to different locations due to the rapid cooling [18,19]. However, because the laser heats up the metal powder

Corresponding author. E-mail address: [email protected] (H.J. Martin).

https://doi.org/10.1016/j.addma.2018.08.023 Received 19 March 2018; Received in revised form 15 August 2018; Accepted 16 August 2018 Available online 18 August 2018 2214-8604/ © 2018 Elsevier B.V. All rights reserved.

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Fig. 1. Dog Bone dimensions used to produce LPBF samples and TM samples, with all dimensions in millimeters. Dimensions are scaled down from ASTM E-8 standards [34].

resistance and the formation of chromium oxides in place of iron oxides [31,32]. Hydrogen embrittlement of stainless steel in sulfuric acid environments was also a problem, where larger grains and thicker grain boundaries as compared to the laser peened surface, allowed for hydrogen to diffuse into, and collect within, the grain boundaries, resulting in a more brittle surface [33,34]. The purpose of this research, then, is to determine how exposure to a sulfuric acid solution would affect samples as-produced using LPBF compared to TM samples. The weight loss, surface effects, mechanical properties of tensile stress and strain, and fracture surfaces were determined on as-printed or as-purchased 316L using an immersion environment containing 0.75 M sulfuric acid over 2184 h. The total time of exposure was chosen as 2184 h, instead of the standard 1000 h, to ensure that the effects of the corrosive environment would be realized, as stainless steel is a slow corroding metal in sulfuric acid [30]. The ultimate goal of this research, then, is to determine how the long-term exposure to sulfuric acid affected the behavior of the stainless steel, with respect to both the surface and the mechanical properties.

as it is melted, the layers underneath the laser also experience heating, allowing for diffusion of the elements to different locations, causing higher concentrations of these elements along the edges of the melt pools [18–20]. This preferential diffusion can lead to higher rates of corrosion, as the elements that are added to reduce corrosion of stainless steel, such as chromium or molybdenum, can concentrate along the edges of the melt pool [18,20]. The research on the corrosion behavior of stainless steels produced by additive manufacturing is just beginning. With respect to the corrosion behavior, samples produced by AM methods exposed to chloride containing environments have shown to be slightly more noble than traditionally manufactured samples [20], with more susceptibility to pitting [20,21], higher pitting potentials (Epit), indicating less susceptibility to pitting [22], lower repassivation potentials (Erep), indicating difficulty in initiating repassivation for the AM samples [20–22], and lower break-down potentials (Eb), indicating an easier break down of the oxide film [20,21]. With this combination of behavior, the breakdown of the oxide film occurring sooner and repassivation taking longer, the metastable pits that form prior to the pitting potential degrade the surface prior to true pit initiation and growth, leading to greater corrosion degradation of the AM samples [20–22]. While the study of corrosion of 316L AM parts in chloride containing environments is minimal and ongoing, the study of corrosion of 316L AM parts in sulfuric acid is non-existent. Typically, stainless steels are not considered good materials of construction to use with sulfuric acid environments, especially as temperatures increase [23,24]. However, because the stainless steel could be located in an environment where sulfur oxidizing bacteria (SOB) exist, sulfuric acid could be produced, leading to microbial influenced corrosion [25], even when the stainless steel is not being used for sulfuric acid transportation. In addition, in the case of AM samples, the rough surface [26] can provide protection for the microbes, allowing them to reproduce, metabolize more frequently, and produce larger amounts of corrosive by-products, including sulfuric acid [25]. In addition, corrosion from SOBs is typically very localized and unpredictable, with large variations in the damage caused, which is far different from an abiotic system, where no living organisms are present and the corrosive concentrations are essentially constant [27]. For traditionally manufactured steels, sulfuric acid corrosion has been investigated, which showed that different grades experienced different effects at low concentrations of 1–2 M, but were similar at high concentrations of 6 M [28], due to the inability of the stainless steel to passivate in the higher concentrations [29]. Iron oxides were the main corrosion by-product [30], with high amounts of chromium and low amounts of manganese leading to better corrosion

2. Materials and methods 2.1. Sample preparation Twenty-seven samples of dimensions shown in Fig. 1 in the dog bone shape modified from ASTM E-8 [35] were cut from traditionally manufactured (TM) AISI316L stainless steel (OnlineMetals.com) or printed using laser powder bed fusion (LPBF). The TM metal was used as received, with no changes made to the metal besides creating the dog bone shape. The TM dogbones were cut from a 4.76 mm thick, 12.7 mm wide, and 1219.2 mm long (0.1875″ × 0.5″ × 48″) flat rectangular bar (OnlineMetals.com). For LPBF, gas atomized stainless steel 316L powder, with a mean particle size of 30 μm, supplied by the vendor Renishaw was used. The chemical analysis for both metals are shown below in Table 1, along with the ASTM standard. A Renishaw 250AM printer (Gloucestershire, United Kingdom), located at America Makes, with a continuous wave Ytterbium fiber laser running in an inert atmosphere of Argon or Nitrogen, was used [38]. The laser ran at 200 W in a modulated operation pulsed with a TTL trigger, with a beam diameter of 70 μm at the powder surface, a scanning speed of 590 mm/sec, and an exposure time of 110 μs [38]. This produced a build rate of approximately 20 cm3/hr [38]. After printing, the excess material in the form of supports for the parts was removed with small vice grips, ensuring that the grip thickness nor grip width were compromised. The samples were then measured with calipers to ensure that the 273

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

data with error bars based on one standard deviation. Prior to scanning electron microscopy (SEM), samples were cleaned by placing the broken dog bones in acetone and sonicating for ten minutes in a sonicator (M2800 Bransonic series, Branson Ultrasonics Corporation, CT). The samples were then removed and allowed to air dry for 20 min before sealing in plastic bags prior to SEM imaging. Single axial tensile testing was performed on all of the samples using an INSTRON model 5500R equipped with a 150 kN load cell running at an extension rate of 1 mm/min until fracture (Bluehill 3, version 3.61, Illinois Toolworks, IL). At each time period chosen, three samples per production method were broken, producing an average stress and strain, with error bars based on one standard deviation. The time periods used for tensile testing were t0 = 0 h, t1 = 24 h (1 day), t2 = 96 h (4 days), t3 = 168 h (7 days), t4 = 336 h (14 days), t5 = 672 h (28 days), t6 = 1176 h (49 days), t7 = 1680 h (70 days), and t8 = 2184 h (91 days), as shown in Table 2. Optical microscopy of the corroded surfaces prior to tensile testing was performed with an inverted light was used to take multiple images of the resulting surface at 5× and 10× magnification (Nikon SMZ800; μscope essentials, version 22.1 × 65; PixelLink, Ottawa, ON, Canada). Scanning electron microscopy (SEM) of the fractured surfaces was performed using either a JOEL JSM-IT300LV or a JIB 4500 Multi-Beam System (JEOL, MA). For analysis, the samples were placed on doublecoated carbon conductive tape, placed vertically on the stage with the fractured side of the each sample facing upward, and placed in the SEM. The working distance was maintained between 10 and 25 mm, with magnifications of 33x, 500x, 5000x, and 10,000×. Determination of the density of the samples occurred after tensile testing. Both ends of the fractured samples were weighed using the same balance as before. The two ends were then placed in 25 mL graduated cylinders filled with 15 mL of water, with the initial volume change recorded. Water was then added to the cylinders until the cylinders were completely filled, to prevent the small amount of evaporation and condensation that was seen to previously occur, unfortunately skewing previous results. The cylinders were covered using Parafilm and the samples were left for 72 h, to allow for any potential penetration into any existing pores or voids. After the 72 h, the samples were removed, the total remaining volume was recorded, and the samples were weighed immediately while wetted. The difference in dry mass versus wetted mass and the determined volume were used to calculate density. A NANOVEA model MC/NA equipped with a NANOVEA optical pen CL3/MG70 (NANOVEA, Inc., Irvine, CA) was used to obtain the surface roughness values of the TM and LPBF control samples and samples corroded over 2184 h. The optical pen ran at a depth of 143/300 and an intensity of 88/100, measuring the reflected light from the sample surface. Prior to analysis, the samples were loaded onto the stage, viewed under the microscope to ensure that no surface voids were present, and then moved to the optical pen for scanning. Mechanical 3D scan v1.3.0. software (NANOVEA, Inc., Irvine, CA), was used to perform the scan. The scan parameters are as follows: scan direction was in the x-axis, scan type was area x and y lengths were set to 0.1 mm with a resolution of 0.5 μm. Each sample took roughly 45 min to scan. The data collected was analyzed using the Analysis 3D v1.3.0 software (NANOVEA, Inc., Irvine, CA). The surface roughness (RA) and root mean square (RQ) values were obtained by drawing horizontal, vertical, and diagonal lines across scanned areas with no black spots or holes. The average value between these lines was what was recorded as the roughness and root mean square values.

Table 1 Chemical Composition of TM and AM samples, with ASTM standard values for comparison.

Balance

TM [37] (% by Weight) Balance

LBPF [38] (% by Weight) Balance

16.0–18.0 10.0–14.0 2.0–3.0 2.0 Maximum 0.75 Maximum – – 0.045 Maximum – 0.10 Maximum 0.03 Maximum 0.03 Maximum

16.88 10.15 2.03 1.06 0.550 – – 0.028 – 0.017 0.019 0.002

17.898 10.330 2.220 1.363 0.499 0.133 0.024 0.017 0.007 – – –

Element

ASTM A240 SS 316L [36] (% by Weight)

Iron (Fe) Chromium (Cr) Nickel (Ni) Molybdenum (Mo) Manganese (Mn) Silicon (Si) Copper (Cu) Vanadium (V) Phosphorus (P) Niobium (Nb) Nitrogen (N) Carbon (C) Sulfur (S)

Notice that, for both sets of samples, all composition values are within the range of the ASTM standard. For the LBPF samples, though, the XRF used to collect the chemical composition cannot detect elements below sodium (Na), thus skewing the information regarding carbon and nitrogen [38].

measurements matched the dimensions in the STL file, as shown in Fig. 1. Besides the support removal, no additional processing was performed on the samples. 2.2. Corrosion testing Twenty-four of the samples of each production method were completely immersed in covered 3-quart Pyrex dishes filled with 2 liters of a 0.75 M sulfuric acid solution [25], using 96% sulfuric acid (Acros Organic, New Jersey) and deionized water and maintained at room temperature and humidity. The solution was removed and replaced every seven days, to ensure both an adequate number of corrosive ions and to combat any potential evaporation of the water. Three dog bones each of TM and LPBF were exposed to the test environment for a given amount of time, removed, rinsed with deionized water, dried, and analyzed. Following analysis, the dog bones were then placed back into the test environment for additional time. The remaining twenty-one samples were exposed to the test environment for a given amount of time, removed, rinsed with deionized water, and then subjected to tensile testing; the three samples that were analyzed were the last samples to undergo tensile testing. The samples were exposed to the sulfuric acid environment for t0 = 0 h, t1 = 24 h (1 day), t2 = 96 h (4 days), t3 = 168 h (7 days), t4 = 216 h (9 days), t5 = 336 h (14 days), t6 = 504 h (21 days), t7 = 672 h (28 days), t8 = 840 h (35 days), t9 = 1008 h (42 days), t10 = 1176 h (49 days), t11 = 1344 h (56 days), t12 = 1512 h (63 days), t13 = 1680 h (70 days), t14 = 1848 h (77 days), t15 = 2016 h (84 days), and t16 = 2184 h (91 days), as shown in Table 2. While normal exposure time for corrosion testing is usually 1000 h, extended times were chosen to ensure that the behavior of the stainless steel in the corrosive environment would be adequately documented. 2.3. Analysis Following each time exposure, the dog bones were analyzed for changes in mass, thickness, and using optical microscopy for changes in surface features. The dog bones were weighed prior to testing and following each exposure on a laboratory grade scale (SI-234, Denver Instrument, NY). Thickness measurements were taken on each sample prior to testing and following each exposure using standard calipers (Harbor Freight). Prior to weighing and measuring, the samples were rinsed with deionized water to remove any sulfuric acid solution and allowed to dry. Measurements for all figures were averages from the

3. Results Fig. 2 shows micrographs of the traditionally manufactured (TM) (Fig. 2a) and laser powder bed fusion (LPBF) (Fig. 2b) samples before being exposed to the sulfuric acid solution (top row) and after 2184 h of corrosion (bottom row). Scale bars are provided to allow for accurate 274

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Table 2 Immersion and Tensile Testing Plan for Traditionally Manufactured and Additive Manufactured Samples. Time

Control (0 h) 24 h (1 day) 96 h (4 days) 168 h (7 days) 216 h (9 days) 336 h (14 days) 504 h (21 days) 672 h (28 days) 840 h (35 days) 1008 h (42 days) 1176 h (49 days) 1344 h (56 days) 1512 h (63 days) 1680 h (70 days) 1848 h (77 days) 2016 h (84 days) 2184 h (91 days)

Total Number

Weighed

Tensile Testing

H2SO4 Remaining

TM

LPBF

TM

LPBF

TM

LPBF

TM

LPBF

24 21 18 15 15 15 12 12 9 9 9 6 6 6 3 3 3

24 21 18 15 15 15 12 12 9 9 9 6 6 6 3 3 3

6 6 6 3 3 6 3 6 3 3 6 3 3 6 3 3 3

6 6 6 3 3 6 3 6 3 3 6 3 3 6 3 3 3

3 3 3 – – 3 – 3 – – 3 – – 3 – – 3

3 3 3 – – 3 – 3 – – 3 – – 3 – – 3

21 18 15 15 15 12 12 9 9 9 6 6 6 3 3 3 0

21 18 15 15 15 12 12 9 9 9 6 6 6 3 3 3 0

Arrangement of samples throughout the 2184 h exposure time. Twenty-four samples were initially available, with three samples immediately removed for control weighing and tensile testing. The remaining samples were placed in H2SO4. At each point outlined in Section 2.2, three samples from each method were removed and examined before being returned to H2SO4. At each point outlined in Section 2.3, three samples from each method were removed and broken, resulting in fewer samples remaining in H2SO4. Samples were not broken at each time period that samples were examined.

respectively) and after corrosion (Fig. 3g and h, respectively). When looking at the graphs, the y-axes for the two TM samples are much smaller, ranging from -1.5 to 2 μm, than the y-axes for the two LPBF samples, ranging from -30 to 30 μm for the control and -40 to 100 μm for the corroded sample. From these graphs, the RA and RQ were determined for each of the four surface types, as shown in Table 3. The TM samples exhibited a small amount of smoothing, as the RA value dropped from 0.6054 μm for the control sample to 0.5356 μm for the sample after 2184 h of corrosion, while the RQ value dropped from 0.5779 μm to 0.4043 μm, respectively. The LPBF samples exhibited a large increase in the surface roughness, with an RA value of 9.2103 μm prior to corrosion and 19.5774 μm after corrosion and RQ values of 129.9945 μm and 696.1570 μm, respectively. The TM samples were smoother than the LPBF samples both before and after corrosion.

measurement of surface features. Prior to corrosion, the TM samples were much smoother than the LPBF samples, with lines indicative of the rolling process, while the LPBF samples were much rougher than the TM samples, with surfaces indicative of the fusion process, which ultimately leads to the rough texture appearance. After corrosion, the TM samples were still much smoother, with lines indicative of the rolling process still present. However, the number of lines present were greatly reduced, indicative of surface corrosion. For the LPBF samples, the surfaces were still rough, although they appeared smoother than before corrosion, indicative of the fusion process. Fig. 3 shows the 3-D and 2-D surface roughness scans. The scans are arranged to show the 3-D then 2-D scans of TM samples prior to corrosion (Fig. 3a and b, respectively) and after corrosion (Fig. 3c and d, respectively) and LPBF samples prior to corrosion (Fig. 3e and f,

Fig. 2. Surface microscopy of the TM (a,c) and LPBF (b,d) samples prior to corrosion (top row) and after 2184 h of corrosion (bottom row). Notice the smoother surface for the TM surface, with the lines indicative of rolling, as compared to the rougher surface of LPBF, indicative of the fusion process. After corrosion, both surfaces appeared to show signs of surface corrosion. For the TM samples, notice that some of the lines had disappeared, while, for the LPBF samples, the larger peaks appeared to be were reduced (see Fig. 3).

275

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Fig. 3. Surface roughness 3-D and 2-D graphs of the TM (a–d) and LPBF (e–h) samples prior to corrosion (a–b, e–f) and after 2184 h of corrosion (c–d, g–h). Notice that the y-axes for the TM samples are much lower than the y-axes for the LPBF samples, regardless of before or after corrosion. In addition, notice that the y-axes for the LPBF samples are much smaller for before corrosion than after corrosion. These axes changes indicate that the 2184 h greatly increased the surface roughness of the LPBF samples, contrary to the appearance in Fig. 2. For both environments, corrosion changed the surface, resulting in a smoother surface for the TM samples and a rougher surface for the LPBF samples.

276

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

MPa as a function of exposure time. For the TM samples, no real change is seen in the ability to withstand stress over time; the samples at the end of the exposure time break at the same point that the control samples did. For the LPBF samples, there is a gradual decrease in the ability to withstand stress over time. The data point at 1680 h does skew the trend, causing it to appear to be greater than it likely is. The stress for this data point, and its large errors bars, were likely caused by a small amount of torque being accidently placed on one sample when it was placed in the grips. Fig. 7 shows the strain of both the TM and LBPF samples in mm/mm as a function of exposure time. For the TM samples, no real change is seen in the strain experienced over time; the samples at the end of the exposure time have the same elongation that the control samples did. For the LPBF samples, there is a gradual decrease in the ability to withstand strain over time. The data point at 1680 h does skew the trend, causing it to appear to be greater than it likely is. The strain for this data point, and its large errors bars, were likely caused by a small amount of torque being accidently placed on one sample when it was placed in the grips. Fig. 8 shows representative stress-strain curves of both the TM and LBPF samples prior to exposure and after 2184 h of exposure. When looking at these stress-strain curves, the main difference between the TM and LBPF samples are the type of failure. TM samples experienced ductile failure, as deduced by the gradual decreasing in stress after a strain of 0.6 mm/mm, while the LBPF samples experience brittle fracture, as deduced by the sudden, drastic decrease in stress after a strain of 0.3 mm/mm. The other main difference to notice is that, for the LBPF samples, the stress withstood decreased after 2184 h of exposure, which is to be expected. For the TM samples, though, the stress withstood appeared to have increased after 2184 h. While this is the appearance, the actual results demonstrated that the stress was similar for the control and the 2814 h samples. As these were representative graphs, and not all of the graphs collected, the actual averaged data was more comparable between control and 2814 h samples. Table 4 shows the average stress and strain data for each of the exposure times when the dog bones were broken. When looking at Table 4, the stress experienced by the TM samples did not decrease, nor did the strain. In fact, there appears to be a slight increase in both stress and strain withstood from the control to the final exposure, which is further confirmed by the percent difference. This increase is explained

Table 3 Average Roughness (RA) and Root Mean Square Roughness (RQ) values for the TM and LPBF samples before and after corrosion, including differences. Surface Type

RA

RQ

TM Before Corrosion TM After Corrosion LPBF Before Corrosion LPBF After Corrosion TM Roughness Change LPBF Roughness Change

0.6054 μm 0.5356 μm 9.2103 μm 19.5774 μm −0.0698 μm 10.3671 μm

0.5779 μm 0.4043 μm 129.9946 μm 696.1570 μm −0.1736 μm 566.1624 μm

Notice that, for the TM samples, both the RA and RQ values decreased over 2184 h of exposure to the sulfuric acid environment. The differences, though, are small, indicating that the corrosive environment had minimal effect on the surface. For the LPBF samples, though, both the RA and RQ values greatly increase, more than doubling and quintupling, respectively. The corrosive environment had a degrading effect on the surface, preferentially degrading various spots along the surface, potentially following the melt pools.

Fig. 4 shows the mass loss in grams of both the TM and LPBF samples as a function of exposure time in hours. Both production types show positive linear trends, although the TM trend is within the error bars, so may not be an accurate trend. The errors bars are one standard deviation in either direction. While the error bars do look similar in height, this is due to the small changes in weight. The LPBF samples lost 0.016 g, while the TM samples lost 0.006 g, a percent difference between the two sample types of 91%. Fig. 5 shows the density of both the TM and LPBF samples in g/mL as a function exposure time in hours. Both production times show no real change in density over time, with any perceived change being within the error bars. The density of the TM samples was determined to be 8.30 ± 0.12 g/mL, while the density of the LPBF samples was determined to be 8.44 ± 0.12 g/mL. These values were consistent after immediately reading the water displaced by the samples and after reading the water displaced by the samples after 72 h. The error bars are one standard deviation in either direction. When looking at the graph, the error bars do appear to be large, but this is due to the values chosen for the y-axis. In fact, when looking at the error bars at the largest point, after 2184 h of corrosion, the density varies from 7.5 g/ mL to 9.0 g/mL. Fig. 6 shows the tensile stress of both the TM and LPBF samples in

Fig. 4. Mass loss in grams as a function of exposure time in hours. Notice that the TM samples experience much less mass loss than the LPBF samples (abbreviated 3D). Also notice that, while both experience a positive linear trend with respect to mass loss, the TM trend lies within the error bars, so it may not be an accurate trend. In addition, all error bars are one standard deviation in each direction.

277

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Fig. 5. Density in g/mL as a function of exposure time in hours. Notice that the TM samples and the LPBF samples (abbreviated 3D) do not change density throughout the exposure time. All error bars are one standard deviation in each direction.

of exposure (Fig. 9i). When looking at the images, there are no distinct differences between Fig. 9a and i, with respect to either the fracture surface or the sample surface. This indicates that corrosion was not prevalent enough to change either of the surfaces after 2184 h. In addition, one can see signs of necking in all nine images, where the surface edge is visible beyond the fracture surface. Fig. 10 shows the SEM images of the TM samples at 500x magnification as time 0 (control) (Fig. 10a), after 24 h of exposure (Fig. 10b), after 96 h of exposure (Fig. 10c), after 168 h of exposure (Fig. 10d), after 336 h of exposure (Fig. 10e), after 672 h of exposure (Fig. 10f), after 1176 h of exposure (Fig. 10g), after 1680 h of exposure (Fig. 10h), and after 2184 h of exposure (Fig. 10i). When looking at the images, as with Fig. 9, there are no distinct differences between Fig. 10a and i, further indicating that corrosion did not affect the behavior of the TM samples. In addition, all nine of the samples have similar void numbers

by the sample choice. The largest samples, with respect to weight, were chosen for the longest exposure time. These slight differences in weight likely affected the stress and strains, as more material was present, so more load and extension were necessary to reach failure. When looking at the LBPF samples, both the stress and strain decreased from the control to the final exposure, as further confirmed by the percent difference. Even if the data point at 1680 h, were torque was accidently applied, were removed, the decreasing trend in stress and strain would still occur, indicating that the environment had a negative effect on the mechanical behavior of the LBPF samples. Fig. 9 shows the SEM images of the TM samples at 33x magnification as time 0 (control) (Fig. 9a), after 24 h of exposure (Fig. 9b), after 96 h of exposure (Fig. 9c), after 168 h of exposure (Fig. 9d), after 336 h of exposure (Fig. 9e), after 672 h of exposure (Fig. 9f), after 1176 h of exposure (Fig. 9g), after 1680 h of exposure (Fig. 9h), and after 2184 h

Fig. 6. Stress in MPa as a function of exposure time in hours. Notice that the TM samples did not experience much change in stress as a function of exposure time, while the LPBF samples (abbreviated 3D) experienced a decrease in stress as a function of exposure time. Also notice that all error bars are one standard deviation in each direction. The large error bars of the LPBF samples at 1680 h was likely due to a small amount of torque accidently applied to one of the samples, causing a much earlier failure than the other two.

278

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Fig. 7. Strain in mm/mm as a function of exposure time in hours. Notice that the TM samples did not experience much change in strain as a function of exposure time, while the LPBF samples (abbreviated 3D) experienced a decrease in strain as a function of exposure time. Also notice that all error bars are on standard deviation in each direction. The large error bars of the LPBF samples at 1680 h was likely due to a small amount of torque accidently applied to one of the samples, causing a much earlier failure than the other two.

Fig. 8. Stress-Strain graphs for TM and LPBF samples (abbreviated 3D) prior to exposure (control) and after 2184 h of exposure. Notice that the LPBF samples have much lower strain values than the TM samples, indicating brittle fracture and ductile fracture, respectively. Also notice that the LPBF sample exposed for 2184 h has a lower stress value than the LPBF control.

same area as the sample surface. Fig. 12 shows the SEM images of the LBPF samples at 500x magnification as time 0 (control) (Fig. 12a), after 24 h of exposure (Fig. 12b), after 96 h of exposure (Fig. 12c), after 168 h of exposure (Fig. 12d), after 336 h of exposure (Fig. 12e), after 672 h of exposure (Fig. 12f), after 1176 h of exposure (Fig. 12g), after 1680 h of exposure (Fig. 12h), and after 2184 h of exposure (Fig. 12i). When looking at the images, the LBPF samples fractured with much rougher edges than the TM samples. In addition, after 24 h exposure (Fig. 12b) and continuing through 2184 h (Fig. 12i), various sizes of particulate are removed from the surface during fracture. Cracking at 168 h (Fig. 12d) is seen, while very large particulate (Fig. 12h and i) is dislodged after 1680 h.

and sizes, as indicated by the black holes present on each of the surfaces. Fig. 11 shows the SEM images of the LBPF samples at 33x magnification as time 0 (control) (Fig. 11a), after 24 h of exposure (Fig. 11b), after 96 h of exposure (Fig. 11c), after 168 h of exposure (Fig. 11d), after 336 h of exposure (Fig. 11e), after 672 h of exposure (Fig. 11f), after 1176 h of exposure (Fig. 11g), after 1680 h of exposure (Fig. 11h), and after 2184 h of exposure (Fig. 11i). When looking at the images, the LBPF samples fractured with much rougher edges than the TM samples. The samples fractured at steep angles, as seen in Fig. 11c and f where the arrows follow the slope, and at sharp angles, as seen in Fig. 11d and e where the arrows indicate the focal point. After 1176 h, material begins to fall out of the samples upon fracture, as shown by the arrows in Fig. 11g, h, and i. The samples with the longest exposure times, Fig. 11h and i, show a dramatic increase in the size and number of particulate missing from the surface. In addition, there is no sign of necking on any of the images, with the fracture surface starting in the

4. Discussion The amount of time chosen to expose the different sample types to the sulfuric acid solution was set for 2184 h. As the standard amount of 279

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Table 4 Compiled Stress and Strain Data for TM and LPBF Samples. Production Method

Exposure Time (hours)

Stress (MPa)

Percent Difference

Strain (mm/mm)

Percent Difference

TM TM TM TM TM TM TM TM TM

0 24 96 168 336 672 1176 1680 2184

574.06 542.89 572.80 569.64 596.51 590.50 565.26 562.30 616.92

± ± ± ± ± ± ± ± ±

25.595 33.669 13.082 8.646 8.901 20.468 36.349 35.905 58.424

5.43 % 0.22% 0.77% −3.91% −2.86% 1.54% 2.05% −7.47%

0.535 0.555 0.578 0.575 0.549 0.577 0.573 0.555 0.592

± ± ± ± ± ± ± ± ±

0.007 0.053 0.006 0.005 0.011 0.021 0.017 0.032 0.026

-3.74% -8.04% -7.48% -2.61% -7.85% -8.22% -3.73% −10.65%

LBPF LBPF LBPF LBPF LBPF LBPF LBPF LBPF LBPF

0 24 96 168 336 672 1176 1680 2184

565.95 518.78 509.66 505.10 507.19 485.80 499.96 349.54 471.94

± ± ± ± ± ± ± ± ±

20.902 10.254 18.773 26.366 15.264 9.388 15.756 128.342 73.367

8.33% 9.95% 10.75% 10.38% 14.16% 11.66% 38.24% 16.61%

0.302 0.304 0.339 0.346 0.336 0.324 0.323 0.207 0.237

± ± ± ± ± ± ± ± ±

0.024 0.018 0.024 0.009 0.023 0.006 0.007 0.087 0.076

-0.66% −12.25% −14.57% −11.26% -7.28% -6.95% 31.46% 21.52%

Notice that, for the TM samples, the average stress and average strain did not change for the entire 2184 h of exposure. For the LBPF samples, though, both the average stress and average strain decreased overall. Even if the data point for 1680 h, where torque was accidently added, were removed, the decreasing trend for both stress and strain would remain. This is further confirmed by looking at the percent differences values. The TM samples hover around 0, with increases and decreases in the percent differences for both stress and strain. The LPBF samples, though, for stress consistently increase in percent difference, while strain increases in percent difference at the end of the study.

Fig. 9. SEM Images of the TM samples at 33x magnification. a) is the control, b) is after 24 h of exposure, c) is after 96 h of exposure, d) is 168 h of exposure, e) is after 336 h of exposure, f) is after 672 h of exposure, g) is after 1176 h of exposure, h) is after 1680 h of exposure, and i) is after 2184 h of exposure. Notice that for all of the samples, there is distinct necking, as the fracture edge is narrower than the sample edge. Also notice that there are no distinct differences in the appearance of the fracture surface or the sample surfaces. 280

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Fig. 10. SEM Images of the TM samples at 500x magnification. a) is the control, b) is after 24 h of exposure, c) is after 96 h of exposure, d) is 168 h of exposure, e) is after 336 h of exposure, f) is after 672 h of exposure, g) is after 1176 h of exposure, h) is after 1680 h of exposure, and i) is after 2184 h of exposure. Notice for all of the samples that voids, indicated by the black spots, appear. However, the voids present are in similar numbers and sizes. Also notice that there are no distinct differences in the appearance of the fracture surface.

(Fig. 2d), although the roughness appears to have lessened slightly as compared to the control (Fig. 2b). However, the optical microscopy image is deceiving, likely due to the depth of field, as the roughness of the surface greatly increased (Fig. 3e–h). In fact, the roughness of the LPBF samples doubled with respect to the average roughness and more than quintupled with respect to the mean square root roughness (Table 3). The depth of field with respect to the optical microscopy image would cause the appearance of a smoother surface, as the top of the peaks and the bottom of the valleys both could not be focused, causing one to appear “fuzzy” and the overall appearance to be smoother. Overall, unlike the TM surfaces, there appears to be a dramatic effect on the LPBF surfaces due to the sulfuric acid environment, potentially as the sulfuric acid attacked along the melt pool edges. The observations made in Figs. 2 and 3 are confirmed when the mass loss of the two sample types are compared. Fig. 4 compares the cumulative mass lost over the 2184 h of exposure time between the TM and LPBF samples. The TM samples lost a total of only 0.004 g, which would further support the findings of the Fig. 3 (a–d) and Table 3, where the roughness decreased only slightly, meaning only a very small amount of stainless steel was removed from the surface. The LPBF samples lost 0.014 g, meaning the rate of mass loss was more than three times higher in the LPBF samples. In addition, as shown in Fig. 3 (e–h) and Table 3, the roughness greatly increased, meaning more of the stainless steel was removed from the surface due to the corrosive environment. The general corrosion rate, though, of both samples would be considered low, which was not originally expected. Originally, it was thought that the LPBF samples would likely exhibit more corrosion due

exposure time for a corrosion study is only 1000 h and stainless steel is inherently resistant to corrosion due to the alloying elements present, this limited amount of time may not have allowed the effects of the corrosive environment to be easily observed. Therefore, the standard amount of exposure time was more than doubled to ensure the mechanisms of corrosion taking place would cause enough damage to characterize them. When looking at Figs. 2 and 3, micrographs of the traditionally manufactured (TM) samples (Fig. 2a and c) and laser powder bed fusion (LPBF) samples (Fig. 2b and d) show the surface differences at time 0, the control (Fig. 2a and b), and after 2184 h (Fig. 2c and d). When looking at the control surfaces, one can see that the surface produced using the TM method contain lines indicative of the rolling process (Fig. 2a), while the surface produced using the LPBF method is rougher, indicative of the powder laying and melting process (Fig. 2b). Fig. 3 confirms these differences, with a smoother surface for the TM samples both prior to (Fig. 3a-b) and after corrosion (Fig. 3c–d) as compared to the LPBF surfaces (Fig. 3e–h). When looking at the corroded surfaces, one can see that both corroded surfaces have some change as compared to the control, which is further confirmed in Fig. 3. For the TM surfaces, the quantity and length of the rolling lines are reduced after 2184 h (Fig. 2c) compared to the control (Fig. 2a). This reduction in rolling lines are further demonstrated (Fig. 3a–d) with a reduction in both the RA and RQ values, as shown in Table 3. However, overall, there is minimal surface corrosion with respect to the TM surfaces, as the reduction in smoothness is only 0.0698 μm from before to after corrosion. For the LPBF samples, the surface is still rough in the micrograph 281

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Fig. 11. SEM Images of the LBPF samples at 33x magnification. a) is the control, b) is after 24 h of exposure, c) is after 96 h of exposure, d) is 168 h of exposure, e) is after 336 h of exposure, f) is after 672 h of exposure, g) is after 1176 h of exposure, h) is after 1680 h of exposure, and i) is after 2184 h of exposure. For images (c) and (f), the samples fractured on an angle, where the arrows are following the slope. For images (d) and (e), the samples fractured with very steep angles, where the arrows indicate where the focal image of the micrograph is located. For images (g), (h), and (i), the arrows indicate holes and missing material, with the holes getting larger from (g) to (i). Notice that as exposure time increases, the fracture surfaces become rougher, with more material missing. Also notice that there are no signs of necking, with the fracture surface appearing to be in line with the sample surface.

relatively small, due to the y-axis values chosen, with the largest range of 7.5 g/mL to 9.0 g/mL and are likely the result of the reading of the water meniscus. The minimal differences in density were attributed to two aspects of the LPBF method. The first is that the LPBF method melts the powder, forming melt pools that would reduce void formation on the surface [18,19]. In addition to melting the powder, the laser would also heat some of the underlying layers, allowing for some diffusion of elements that also would help remove any potentially formed voids [19]. This would result in a similar density between the TM samples and the LPBF samples. While similarities in surface corrosion and densities exist between the two manufacturing types, that ends with the mechanical behavior. When looking at Fig. 6, one can see that the initial tensile stress withstood by both surfaces is similar, but the initial strain in Fig. 7 withstood by both surfaces is much less for the LPBF samples as compared to the TM samples. The differences in manufacturing methods explain the reduction in strain withstood by the LPBF samples, as well as the fracture behavior. During LPBF production, the molecules of the LPBF samples do not have time to settle into the lowest energy confirmation the same way as the TM samples [15–17]. Instead of all of the molten metal cooling together at a constant rate, the LPBF samples solidified in layers that were independent of one another [15–17,19]. Each thin layer of powder is heated rapidly when contacted by the laser. The molten powder only remains in the liquid phase for a short time before solidifying, and it can only interact with the thin solid layer of material

to a higher porosity allowing for penetration of the corrosive ions. It was expected that, because the LPBF process melts the metal powder used in creating the samples in very thin individual layers, voids are left between each of these layers as they are unable to arrange into their lowest energy conformation, which in turn would decrease the initial density of the LPBF samples. The particles can only reach their lowest energy conformation when given enough time for all of the available particles to interact with one another [15–17], as well as a slow cooling rate, which is characteristic of the TM production method. However, as research on TM samples has shown, low concentrations of sulfuric acid, like the 0.75 M used in this research, are not as destructive as high concentrations of sulfuric acid [28,30,32]. Thus, the low rate of corrosion is in line with TM processes. When a low corrosion rate was seen, the density of the two methods was determined to see if the density was different between the samples or if the density changed as the samples were exposed to the corrosive environment. Fig. 5 compares the density of both samples following the exposure of the samples to the corrosive environment. One can see that the density of both types of samples were basically the same and were unaffected by the exposure time. In addition, the density was unaffected by the reading times, where the amount of water displaced by the samples were the same after immediately placing the samples in water and after 72 h of the samples soaking in the water. There do appear to be large discrepancies between the samples, especially from the control to the samples at 2184 h of corrosion. These discrepancies are actually 282

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

Fig. 12. SEM Images of the LBPF samples at 500x magnification. a) is the control, b) is after 24 h of exposure, c) is after 96 h of exposure, d) is 168 h of exposure, e) is after 336 h of exposure, f) is after 672 h of exposure, g) is after 1176 h of exposure, h) is after 1680 h of exposure, and i) is after 2184 h of exposure. Notice that the fracture surface changes between each of the exposure times, In (a), the surface appears mostly smooth, while just 24 h later, in (b) the surface has distinct holes where particulate may have fallen out. By (e), though, larger pieces of material are missing, while in (h) and (i) entire pieces of material have been removed.

withstand. Since there was minimal mass removal, which also means minimal area removal, another explanation must exist for the reduction in mechanical properties of the LPBF samples, but not of the TM samples. Fig. 8, which is a representative graph showing the control curve for the two methods and the curve for 2184 h exposure for the two methods, begins to demonstrate what is occurring when examining the TM samples as compared to the LPBF samples. Fig. 8 confirms the data in Fig. 6, that the stress withstood by the two manufacturing methods are nearly identical. In addition, Fig. 8 confirms the data in Fig. 7, that the strain withstood by the LPBF method is less than the strain withstood by the TM method. Fig. 8 also demonstrates why the LPBF method withstands less strain, as the sudden drop in tensile stress is indicative of a brittle fracture, and why the TM method withstand more strain, as the gradual decrease in tensile stress is indicative of ductile fracture. When comparing the two lines for TM (black), one can see that there appears to be an increase in tensile stress and strain from the control to the 2184 h of exposure, which is not expected. This occurred because the samples chosen to undergo the 2184 h of exposure were the largest samples, so a slight increase in tensile stress and strain is expected, as the weight and area of the 2184 h specimens were slightly higher than the control specimens. In addition, this is a representative graph, showing just one of the three samples data. When comparing the two lines for LPBF (grey), one can see that a drop in tensile strength and strain occurred from the control to the 2184 h of exposure, which is expected when exposed to a corrosive environment. However, as previously stated, general corrosion is likely not the cause for this drop, as

beneath it [19]. This holds true for each subsequent layer of powder, resulting in minimal interaction of the powder within the LPBF samples, although the heat does conduct to the layers beneath forming a heat affected zone [19]. Reduced interaction of molecules and faster cooling times also encourage the formation of potential slip planes within the LPBF samples [15–17]. Without the chance to interact and settle into a lower energy conformation, residual stress can build up within the sample [15–17]. All of these factors lead to differences in the mechanical behavior between the TM and LPBF samples, but does not fully explain the differences in mechanical property behavior as a function of exposure time. The mechanical behavior between the TM samples and LPBF samples also changes after 2184 h of exposure. Specifically, both the tensile stress (Fig. 6) and strain (Fig. 7) decrease for the LPBF samples as compared to the TM samples, which did not change at all. For both the LPBF and TM samples, since there was no reduction in mass (Fig. 4), there would be no reduction in area, meaning the stress applied to the control samples should be the same stress applied to the samples after 2184 h. While this is true for the TM samples, it is not true for the LPBF samples. Therefore, the reduction in tensile stress seen must be due to corrosion, and its reductive effects on mechanical properties. When combining the minimal surface changes (Fig. 2) and mass loss (Fig. 4) with the reduction in mechanical properties (Figs. 5 and 6), general corrosion alone cannot be the only explanation. This is because general corrosion would affect both the surface condition, mainly through the removal of metal on the surface, and the mechanical properties, by reducing the area, thereby reducing the stress the samples could 283

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

is 1.07, while the elongation ratio is 1.11. Susceptibility to environmentally assisted cracking is indicated by the distance from unity of the ratios; the lower the ratio from unity (1), the higher the susceptibility [39]. For the LPBF samples, the distance from unity is 0.167 and 0.215, respectively, while for the TM samples, the distance from unity is -0.07 and -0.11, respectively, which show the ratio is above unity. Looking at these numbers, the LPBF samples are more susceptible to environmentally assisted cracking, specifically hydrogen embrittlement. In addition, all three samples fractured at stress and strain values lower than the control, further indicating that hydrogen embrittlement had occurred [40]. In order to demonstrate that hydrogen embrittlement was the cause of the decrease in mechanical properties of the LPBF samples, scanning electron microscopy (SEM) micrographs of the fracture surfaces were collected, as shown in Figs. 9–12. Figs. 9 and 10 showing the surfaces of the TM samples at 33x magnification and 500x magnification, respectively, and Figs. 11 and 12 showing the surfaces of the LPBF samples at 33x magnification and 500x magnification, respectively. When looking at Fig. 9, one can see that the nine fracture surfaces, ranging from the control to 2184 h of exposure, show no real differences. The samples all show necking, confirming the ductile fracture of Fig. 8. In addition, all of the fracture surfaces appear to be close to the same size, which would appear to validate the behavior seen regarding tensile stress (Fig. 6) and strain (Fig. 7). Lastly, minimal changes around the edges with no distinct corrosion features, further confirm the surfaces seen in Fig. 2, as well as demonstrate that any mass loss (Fig. 4) would be minimal, as there are no missing areas on the surfaces. When looking at Fig. 10, the surfaces again all look very similar, with small voids present on all of the surfaces. The size and quantity of the voids appear very similar for all nine of the fracture surfaces, indicating that the void formation was caused by plastic deformation during tensile testing. Overall, Figs. 9 and 10 demonstrate that no discernible corrosion occurred on the samples, confirming the results seen in Figs. 4,6,7, and 8. Fig. 11 shows the changes to the interior of the LPBF samples as exposure time to the sulfuric acid solution increases. When looking at the control through 1176 h of corrosion, there are minimal changes occurring, although the surfaces of the samples all show different variations of roughness. However, after 1176 h of corrosion, the fracture surfaces become much more jagged with large particles missing from the internal portion of the structure. Unlike Fig. 9, which showed distinct necking, Fig. 11 shows minimal changes in thickness, indicating a more brittle fracture occurred. In addition, the fractures occurred at different angles, demonstrating that there was not a preferred fracture behavior or direction, unlike the TM samples, with necking before fracture. Fig. 12, though, shows that the large particles of detached material begin appearing at 672 h, with very large particles appearing after 1176 h. Both Figs. 11 and 12 help explain the behavior seen in Figs. 6–8, where tensile stress and strain decrease, with minimal mass loss (Fig. 4). The tensile stress and strain both decrease as the particles detach from the surface. Particle detachment also continues the brittle fracture mechanics, since failure occurs suddenly when the particles detach. The large particles of detached material confirm the mechanical behavior seen in Figs. 6–8, demonstrating that hydrogen embrittlement had a large effect on the LPBF samples. Ultimately, the diffusion of hydrogen led to the formation of large, brittle particles in the internal structure of the LPBF samples, leading to reduced mechanical properties (Figs. 6–8) as the particles separated from the samples during tensile testing, as seen in Figs. 11 and 12. In addition to explaining the decreases in mechanical properties and the formation of large areas of missing material, hydrogen embrittlement also explains why the surface features only minimally changed (Fig. 2 and 3), density was unchanged (Fig. 5), and the mass loss minimal (Fig. 4). Hydrogen embrittlement would not visibly affect the surface, since general corrosion or pitting corrosion would change the surface

minimal weight loss occurred. In order to ensure that the tensile strength and strain did not actually increase over time, Table 4 was compiled with the exact averages and standard deviations for both the TM and LPBF samples. When looking at the TM data, one can see that there is not much variation in the stress, ranging from a minimum of 542.89 ± 33.669 MPa at 24 h to a maximum of 616.92 ± 58.424 MPa at 2184 h, a difference of roughly 74 MPa. This difference is tensile stress can come from differences in weight and thickness, as the percent difference is 7.47% from the control value. In addition, there is not much difference in strain, ranging from a minimum of 0.535 ± 0.007 mm/mm at the control to 0.592 ± 0.026 mm/mm at 2184 h, a difference of roughly 0.057 mm/ mm. Again, this difference, which amounts to 10.65% from the control value, can very well be attributed to the differences in weight and thickness. Table 4 also shows the differences in stress and strain for the LPBF samples. Unlike the TM samples, where the difference fluctuates throughout the table, the LPBF samples see a general downward trend, with a starting value of 565.95 ± 20.902 MPa for the control and an ending value of 471.94 ± 73.367 MPa for the 2184 h of exposure, which is a difference of approximately 94 MPa, or more 16.61% from the control value. The strain shows a similar downward trend, with a starting strain of 0.302 ± 0.024 mm/mm for the control and a strain of 0.237 ± 0.076 mm/mm for the 2184 h of exposure, which is a difference of approximately 0.065 mm/mm, or 21.52% from the control value. The differences in tensile stress and strain would likely not be attributed to just differences in weight and thickness, especially since these were printed to specifications versus machined to specifications. One point to notice, though, in looking at Table 4, as well as Figs. 6 and 7, is that the data point at 1680 h appears to be lower than the 2184 h of exposure for both tensile stress and strain. While this data point is lower, it also has a large standard deviation. This data point was skewed by torsion being applied to one of the samples accidently. Because of the applied torsion, the sample broke prematurely, skewing both the tensile stress and strain. However, while the data point is skewed, it does not affect the overall trend, since both the tensile stress and strain at 2184 h are reduced as compared to the control values for the LPBF samples. The decreasing trend for both tensile stress and strain is the result of corrosion, but not of general corrosion due to the limited change in mass and surface features. While general corrosion causes the surface changes, the mechanical property changes are likely due to hydrogen embrittlement, which is known to happen to stainless steel 316L in sulfuric acid [33]. Hydrogen embrittlement occurs when hydrogen ions, released during the corrosion reactions, diffuses into the metal and reacts with compounds within the sample, creating brittle hydrides and carbides [4,33,34]. The TM samples did not experience this diffusion, likely because of the gradual cooling process allowing for elements to diffuse and form grains with grain boundaries. With the LPBF samples, melt pools are supersaturated with elements which may not be bound with neighboring elements in a desirable way, opening up the opportunity for hydrogen to bond with and then create brittle hydride particles [18]. The longer the exposure time, the more hydrogen ions could diffuse, creating more and larger hydride and carbide particles within the body of the samples, thereby reducing the mechanical properties, as seen in Figs. 6–8. This is further confirmed by comparing the ultimate tensile strength ratio and the elongation ratio, values modified from ASTM G129 [39]. ASTM F519 designates that, when two or more samples fracture at a sustained load during the exposure time, then hydrogen embrittlement is indicated [40]. Using the ratios from ASTM G129, the test environment value is divided by the control environment value; in this case, the sulfuric acid environment after 2184 h of exposure and the control value prior to any testing, respectively. In the case of the LPBF samples, the ultimate tensile strength ratio is 0.833, while the elongation ratio is 0.785. In the case of the TM samples, the ultimate tensile strength ratio 284

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

decreased the mechanical properties.

features, but hydrogen embrittlement would only create hydride compounds, which would not be visible at the lower magnifications of the optical microscopy without sanding, polishing, and etching. In addition, while hydrogen diffusion would create the brittle hydrides within the body of the samples, it would not result in a change in the material of the samples, thereby leaving the density and mass unchanged. The slight reduction in mass loss does indicate that there is some general corrosion of the surface occurring, but overall, hydrogen embrittlement is the main mode of destruction for the LPBF samples. Hydrogen embrittlement, though, only affected the LPBF samples, while the TM samples overall seemed immune to the corrosive environment, indicating that, for a situation where microbially influenced corrosion may occur, the traditionally manufactured method may be more suitable for producing parts. The main corrosion mechanism for the LPBF samples is likely hydrogen embrittlement. When it comes to previous research on TM samples that are charged with hydrogen, the grain boundaries act as trap sites for hydrogen [34,41,42]. These sites allow for either the collection of hydrogen, ultimately creating hydrogen blisters [33] or an ability to react with the elements within the grain boundaries, creating carbides and hydrides [4,34]. In addition, as the precipitates within the grain boundaries bond with hydrogen, the shape of the precipitates change, stopping grain boundary sliding and reducing strength [34]. LPBF samples do not have grain boundaries in the same way that TM samples do. Instead, as the laser initially melts the powder and then the metal rapidly cools, melt pools containing supersaturated elements are left behind [18]. The laser ultimately heats up the underlying layers. This heat does not cause the underlying layers to re-melt, but does allow for some diffusion of elements, creating locations of chromium and carbon precipitates [18,19]. Like the TM-charged samples, then, the melt pool edges act as trap sites for the hydrogen, leading to brittle carbide and hydride compounds. Since the precipitates that lead to these compounds are beneath the surface, as the heat effects are caused by the laser heating the underlying layers, the hydrogen effects would be seen internally. The hydrogen bonded precipitates likely change shape, as was the case with the charged TM samples, leading to a reduction in strength and the hydrogen bonded precipitates falling out of the surrounding material. Lastly, the hydrogen bonded precipitates would not cause a change in failure mechanism [33], but would cause the failure to occur at lower stresses.

Data availability The raw/processed data required to reproduce these findings cannot be shared at this time as the data also forms part of an ongoing study. Acknowledgements Financial support from Youngstown State University (YSU) Graduate School and University Research Council are gratefully acknowledged. Support from America Makes, which provided 316L powder and Renishaw 250AM printer, and Rodrigo Enriquez Gutierrez for making the samples are also deeply appreciated. Support from Dr. Dingqiang Li for training in the theory and operation of the Scanning Electron Microscopes at YSU is also gratefully acknowledged. Cutting and milling of the traditionally manufactured samples by John Dodson is also deeply appreciated. References [1] W.D. Callister, D.G. Rethwisch, Materials Science and Engineering: An Introduction, John Wiley, Hoboken, 2014. [2] R.V. Shankar, L. Singhal, Corrosion behavior and passive film chemistry of 216L stainless steel in sulphuric acid, J. Mater. Sci. 44 (2009) (2009) 2327–2333. [3] D.R. Askeland, W.J. Wright, The Science and Engineering of Materials, Cengage Learning, Boston, 2016. [4] E. McCafferty, Introduction to Corrosion Science, Springer, New York, 2010. [5] H. Alves, D.C. Agarwall, H. Werner, Alloys Suitable for Phosphoric Acid Applications, NACE International, San Diego, 2006. [6] D. Herzog, V. Seyda, E. Wycisk, C. Emmelmann, Additive manufacturing of metals, Acta Mater. 117 (2016) 371–392. [7] R. Li, Y. Shi, Z. Wang, L. Wang, J. Liu, W. Jiang, Densification behavior of gas and water atomized 316L stainless steel powder during selective laser melting, Appl. Surf. Sci. 256 (2010) 4350–4356. [8] J. Kruth, M. Badrossamay, E. Yasa, J. Deckers, L. Thijs, J.V. Humbeeck, Part and material properties in selective laser melting of metals, 16th International Symposium on Electromachining (ISEM XVI), (2010) https://lirias.kuleuven.be/ handle/123456789/265815. [9] R. Rai, J.W. Elmer, T.A. Palmer, T. DebRoy, Heat transfer and fluid flow during keyhole mode laser welding of tantalum, Ti-6Al-4V, 304L stainless steel and vanadium, J. Phys. D Appl. Phys. 40 (2007) 5753–5766. [10] A. Simchi, Direct laser sintering of metal powders: mechanism, kinetics and microstructural features, Mater. Sci. Eng. A 428 (2006) 148–158. [11] I. Yadroitsev, A. Gusarov, I. Yadroitsava, I. Smurov, Single track formation in selective laser melting of metal powders, J. Mater. Process. Tech. 210 (2010) 1624–1631. [12] C. Kamath, B. El-dasher, G. Gallegos, W. King, A. Sisto, Density of additivelymanufactured, 316L SS parts using laser powder-bed fusion at powers up to 400 W, Int. J. Adv. Manuf. Tech. 74 (2014) 65–78. [13] P. Lykov, S. Sapozhnikov, I. Shulev, D. Zherebtsov, R. Abdrakhimov, Composite Micropowders for Selective Laser Sintering, Metallurgist 59 (2016) 851–855. [14] B. Zhang, L. Dembinski, C. Coddet, The study of the laser parameters and environment variables effect on mechanical properties of high compact parts elaborated by selective laser melting 316L powder, Mater. Sci. Eng. A 584 (2013) 21–31. [15] J. Cherry, H. Davies, S. Mehmood, N. Lavery, S. Brown, J. Sienz, Investigation into the effect of process parameters on microstructural and physical properties of 316L stainless steel parts by selective laser melting, Int. J. Adv. Manuf. Tech. 76 (2015) (2015) 869–879. [16] P. Mercelis, J.P. Kruth, Residual stresses in selective laser sintering and selective laser melting, Rapid Prototyp. J. 470 (2016) (2016) 170. [17] X. Lou, M. Song, P.W. Emigh, M.A. Othon, P.L. Andresen, On the stress corrosion crack growth behaviour in high temperature water of 316L stainless steel made by laser powder bed fusion additive manufacturing, Corros. Sci. 128 (2017) 140–153. [18] J.R. Trelewicz, G.P. Halada, O.K. Donaldson, G. Manogharan, Microstructure and corrosion resistance of laser additively manufactured 316L stainless steel, JOM 68 (2016) 850–859. [19] M. Cabrini, S. Lorenzi, T. Pastore, S. Pellegrini, D. Manfredi, P. Fino, S. Biamino, C. Badini, Evaluation of corrosion resistance of Al-10Si-Mg alloy obtained by means of Direct Metal Laser Sintering, J. Mater. Process. Tech. 231 (2016) 326–335. [20] M. Zietala, T. Durejko, M. Polanski, I. Kunce, T. Plocinski, W. Zielinski, M. Lazinska, W. Stepniowski, T. Czujko, K.J. Kurzydlowski, Z. Bojar, The microstructure, mechanical properties, and corrosion resistance of 316 L stainless steel fabricated using laser engineered net shaping, Mater. Sci. Eng. A 677 (2016) 1–10. [21] Y. Sun, A. Moroz, K. Alrbaey, Sliding wear characteristics and corrosion behaviour of selective laser melted 316L stainless steel, J. Mater. Eng. Perform. 23 (2014) 518–526. [22] G. Sander, S. Thomas, V. Cruz, M. Jurg, N. Birbilis, X. Gao, M. Brameld, C.R. Hutchinson, On the corrosion and metastable pitting characteristics of 316L

5. Conclusions Stainless steel 316L produced using two different methods, traditionally manufactured (TM) and laser powder bed fusion (LPBF), were studied to understand the corrosion behavior in 0.75 M sulfuric acid, to mimic microbial influenced corrosion. The general trends were observed: 1 After 2184 h of exposure, for the TM samples, the mass loss was nonexistent, while the surface roughness values slightly decreased, indicating general corrosion did not occur. For the LPBF samples, the mass loss was minimal, while the surface roughness values increased drastically, indicating general corrosion did occur, but only slightly. 2 The tensile stress and strain remained nearly constant for the TM samples, which demonstrated ductile failure, but decreased for the LPBF samples, which demonstrated brittle failure. 3 SEM micrographs of the fracture surfaces further confirmed no changes for the TM samples, while showing large particles missing from the surface of the LPBF samples. In addition, SEM micrographs confirmed the different modes of failure between the two manufacturing types. 4 Hydrogen embrittlement was determined to be the main form of corrosion on the LPBF samples, likely due to the manufacturing process allowing hydrogen ions to diffuse more easily into the samples, creating brittle carbide and hydride precipitates that 285

Additive Manufacturing 23 (2018) 272–286

J.T. Miller et al.

[23] [24]

[25]

[26] [27]

[28]

[29] [30] [31] [32]

[33] E. Agyenim-Boateng, S. Huang, J. Sheng, G. Yuan, Z. Wang, J. Zhou, A. Feng, Influence of laser peening on the hydrogen embrittlement resistance of 316L stainless steel, Surf. Coat. Tech. 328 (2017) 44–53. [34] J. He, G. Han, S. Fukuyama, K. Yokogawa, A. Kimura, Effect of hydrogen on dynamic precipitation of carbide in type 304 stainless steel during creep process, Acta Mater. 45 (1997) 3377–3388. [35] ASTM E8-16a, Standard Test Methods for Tension Testing of Metallic Materials, ASTM International, West Conshohocken, PA, 2016www.astm.org. [36] ASTM A240-15b, Standard Specification for Chromium and Chromium-Nickel Stainless Steel Plate, Sheet, and Strip for Pressure Vessels and for General Applications, ASTM International, West Conshohocken, PA, 2015www.astm.org. [37] TA Chen International, Inc, Mill Test Report, Jeng-Teh, Tainan, TWN, 2014. [38] T.J. Kneen, Characterizing the High Strain Rate Mechanical Behavior of Stainless Steel 316L Processed by Selective Laser Melting, Youngstown State University, 2016. [39] ASTM G129-00, Standard Practice for Slow Strain Rate Testing to Evaluate the Susceptibility of Metallic Materials to Environmentally Assisted Cracking, ASTM International, West Conshohocken, PA, 2000www.astm.org. [40] ASTM F519-13, Standard Test Method for Mechanical Hydrogen Embrittlement Evaluation of Plating/Coating Processes and Service Environments, ASTM International, West Conshohocken, PA, 2000www.astm.org. [41] G.M. Pressouyre, I.M. Bernstein, A kinetic trapping model for hydrogen-induced cracking, Acta Metall. 27 (1979) 89–100. [42] J.P. Hirth, Effects of hydrogen on the properties of Iron and steel, Metall. Trans. A. 11A (1980) 861–890.

stainless steel produced by selective laser melting, J. Electrochem. Soc. 164 (2017) C250–C257. R. Turton, R.C. Bailie, W.B. Whiting, J.A. Shaeiwitz, Analysis, Synthesis, and Design of Chemical Processes, Prentice Hall, Upper Saddle River, 1998. M. Lindgren, S. Siljander, R. Suihkonen, P. Pohjanne, J. Vuorinen, Erosion-corrosion resistance of various stainless steel grades in high-temperature sulfuric acid solutions, Wear 364-365 (2016) 10–21. S. Okabe, M. Odagiri, T. Ito, H. Satoh, Succession of sulfur-oxidizing Bacteria in the microbial community on corroding concrete in sewer systems, Appl. Environ. Microb. 73 (2007) 971–980. C.P. Ma, Y.C. Guan, W. Zhou, Laser polishing of additive manufactured Ti alloys, Opt. Laser Eng. 93 (2017) 171–177. D. Dierksen, P. Kühner, A. Kappler, K.G. Nickel, Microbial corrosion of silicon nitride ceramics by sulphuric acid producing bacteria Acidithiobacillus ferrooxidans, J. Eur. Ceram. Soc. 31 (2011) 1177–1185. R.T. Loto, C.A. Loto, potentiodynamic polarization behavior and pitting corrosion analysis of 2101 duplex and 301 austenitic stainless steel in sulfuric acid concentrations, J. Fail. Anal. Preven. 17 (2017) 672–679. J.R. Kish, M.B. Ives, J.R. Rodda, Anodic behaviour of stainless steel S43000 in concentrated solutions of sulphuric acid, Corros. Sci. 45 (2003) 1571–1594. R. Wang, An AFM and XPS study of corrosion caused by micro-liquid of dilute sulfuric acid on stainless steel, Appl. Surf. Sci. 227 (2004) 399–409. C.A. Della Rovere, J.H. Alano, J. Otubo, S.E. Kuri, Corrosion behavior of shape memory stainless steel in acid media, J. Alloy Compd. 509 (2011) 5376–5380. V.S. Rao, L.K. Singhal, Corrosion behavior and passive film chemistry of 216L stainless steel in sulphuric acid, J. Mater. Sci. 44 (2009) 2327–2333.

286