Initial Characterization of the Layer Interface for Graphite-Free Additive Friction Stir Deposition of AA7075

Abstract

Additive friction stir deposition (AFSD) is a novel friction stir technology. It is one of the most prolific solid-state metal deposition processes. In recent years, the aerospace and defense industries have increased their investment in the deposition of 7xxx aluminum alloys. This has allowed AFSDs of 7xxx aluminum to move from a laboratory environment to being tested in an industrial setting. This work strives to help move the AFSD of AA7075 toward an effective production environment by providing an initial characterization of the graphite-free layer interface. To the authors’ knowledge, this is the first graphite-free study to utilize both knub–scroll and scroll tools in AA7075. It is also the first study to compare how flat, knub, knub–scroll, and scroll influence layer mixing in graphite-free AA7075. The condition of the layer interface is particularly important to build direction properties. As many end users of AFSD desire isotropic properties, improving build direction properties is extremely important. This work looks at how external tool geometries and layer height impact the layer interface. The objective is to not only better characterize the layer interface but also to determine if a specific external geometry and or layer height could help facilitate a stronger layer interface. It was found that depositions made by the knub tool at a 2.5 mm layer height generated the most visually consolidated layer interface at an optical and SEM level. Under EDS analysis, the knub tool only saw a 12% variation between peak and background oxygen counts. EBSD scans also revealed a more consistent grain size distribution.

1. Introduction

1.1. History and Background on Additive Friction Stir Depostion (AFSD)

Additive friction stir deposition (AFSD) is a solid-state metal additive manufacturing (AM) process that was first developed by Aero Probe and later by its subsidiary MELD. AFSD utilizes spinning tools with a hole for feed material to pass through. Feed material is forced through the tool and effectively extruded onto some type of substrate material (typically an alloy like the one being extruded). As a result of the AFSD process, near-net-shape parts are produced. Some additional machining may be required depending on finish or part specifications, but the process produces parts that require significantly less subtractive machining when compared to other metal fabrication processes [1,2,3]. As a solid-state process, AFSD shares many of the advantages often associated with friction stir welding (FSW). AFSD has been pitched as a solution to replacing long lead time forgings, with the hope that parts produced during AFSD can be heat treated to proceed with forged-like properties. The AFSD process has been used to deposit or print several different alloys, many of which are extremely difficult to fabricate using fusion AM technologies. Magnesium alloys, aluminum alloys, steel alloys, copper alloys, and titanium alloys have all been fabricated to some extent using the AFSD process [4,5,6,7,8,9]. This paper and the proceeding work focus on AFSD with respect to 7XXX aluminum. It is important to note that all work presented here is conducted without the use of graphite or any other type of lubricant. Graphite lubricants are problematic for AFSD processes as they create unpredictability in both the as-deposited and heat-treated properties of a given build [10]. Up until recently, large-scale depositions of 7XXX aluminum using AFSD were unobtainable. However, thanks to work conducted by BYU and other universities, graphite-free depositions of 7XXX have been achieved and replicated successfully. This is, in part, thanks to novel tool designs developed at BYU [11] and further development of process parameters and machine capabilities. However, because the graphite-free development is new, much of the previous work regarding AFSD of 7XXX has been performed using graphite. While the information gathered is still valuable, there is still a need to establish and better understand the AFSD process without the use of graphite.

1.2. Initial Build and Transverse Direction Properties

Much of the previous work in 7XXX aluminum alloys is assumed to be performed with graphite. While this paper discusses graphite-free properties, it is still important to recognize and discuss the trends found previously. Avery et al. found that, in general, the tensile strength of the deposited material was about 50% of the base material. They also demonstrated that the vertical or XZ build direction performed worse when compared to the horizontal or XY direction [12]. Mason et al. also found that with 7050, the as-deposited properties were about 50% that of the base material. They also found that the vertical samples performed worse than the horizontal samples. However, not only were the vertical samples weaker in terms of UTS, but they also found that vertical samples had significantly less elongation than the horizontal samples [13]. To help bridge the gap between deposited and base material properties, some groups have attempted to heat treat the deposited material. While overall the UTS did generally improve (at the expense of lower % elongation), the UTS of the samples was still less than the base material [5,14].

Part of the motivation to remove graphite from the AFSD process was the hope that it would improve the overall properties of AFSD 7XXX aluminum alloys. Graphite was also seen as a problem during the heat treatment of AFSD builds and was often attributed to inconsistent properties and heat treatment results, which is why BYU developed novel tool designs that allowed for consistent graphite-free depositions in both AA7075 and AA7050 [11]. Due to this success, the authors have been able to gather the initial tensile properties of AA7075 in both the as-deposited and heat-treated (T7) conditions. Vertical properties are shown below in

Table 1. A table comparing vertical tensile samples of AA7075 gathered with and without graphite compared to the annealed and forged conditions.

Sample	UTS (MPa)	Elongation at Break (%)	Comparison to Forged UTS	Comparison to Forged Elongation
Graphite No HT Vertical 1 (1 mm layer)	283.27	2.05	62.26%	34.16%
T4V No Graphite W/HT (2.67 mm layer)	386.5	2.86	84.95%	47.67%
T5V No Graphite W/HT (2.67 mm layer)	205.06	1.47	45.07%	24.5%
T3V No Graphite No HT (2.67 mm layer)	322.96	16.28	70.98%	271.33%
T6V No Graphite No HT (2.67 mm layer)	351.34	13.7	77.22%	228.33%
T7V No Graphite No HT (2.67 mm layer)	389.7	13.3	85.65%	221.67%
7075-O	228	16	50.1%	266.67%
7075-T73Forged (AMS4141)	455	6	100%	100%

, and horizontal properties are shown below in

Table 2. A table comparing horizontal tensile samples of AA7075 gathered with and without graphite compared to the annealed and forged conditions.

Sample	UTS (MPa)	Elongation at Break (%)	Comparison to Forged UTS	Comparison to Forged Elongation
Graphite No HT Horizontal 1 (1 mm layer)	373.6	17.53	82.1%	292.17%
T1H No Graphite No HT (2.67 mm layer)	446.1	21.1	98.04%	351.67%
T3H No Graphite W/HT (2.67 mm layer)	591.58	11.13	130.02%	185.5%
T5H No Graphite W/HT (2.67 mm layer)	502.65	10.32	110.47%	172%
7075-O	228	16	50.1%	266.67%
7075-T73Forged (AMS4141)	455	6	100%	100%

.

The data collected show that in the as-deposited condition, removing graphite does significantly improve the UTS and percent elongation for both the vertical and horizontal build directions, so much so that when compared to the AMS 4141 specification on the properties of forged AA7075, an as-deposited horizontal sample was as high as 98% of the specification UTS and over 350% the necessary percent elongation. When horizontal samples were subjected to a T7 heat treat, we saw strengths improve to well above 100% while maintaining a percent elongation that was also above 100%.

The as-deposited vertical properties were also improved; however, they were not to the degree of the horizontal properties. While the vertical elongation was as high as 200%, in some cases, the UTS was only at most 85% of the specification. When looking at samples that were heat treated to a T7 temper, the results change as well. In the horizontal direction, UTS improved to beyond 100% of the spec at the expense of the percent elongation, which reduced to around 150% of the spec. The vertical samples that were heat-treated, however, were worse than the as-deposited vertical samples. UTS only marginally improved, and in one case, it dropped to 40% of the spec strength. The biggest decrease was in the percent elongation, where the as-deposited verticals were over 100% of the specification requirements; the heat-treated samples dropped to as low as 24.5% of the specification requirements. This discrepancy indicates that there is an issue with the vertical layer adhesion of the AFSD builds. Even though graphite is no longer a part of the system, there are still issues that need to be rectified to achieve more isotopic properties across both directions.

1.3. Attempts to Resolve Poor Vertical Properties

For AFSD to become a viable production process, isotropic properties are necessary. A large market for AFSD is replacing long lead time forging, and that becomes difficult if AFSD parts cannot achieve the properties of forged parts. Isotropic properties are particularly important for the aerospace industry, which is particularly interested in utilizing AFSD of 7XXX on future and current products. If forged properties can be achieved using AFSD, then AFSD gains a significant edge over forgings. The biggest advantage that AFSD would possess is time to manufacture. Whereas a forge house may take over a year to produce parts, an AFSD machine could produce a part in days or weeks. AFSD parts also do not require expensive dyes. Thus, AFSD allows for part modification and optimization as product design progresses. This could equate to lower costs and increased performance as parts that were originally forged can now be modified and made quickly. Instead of designing a product around a forged part, the said part can grow and adapt as the product design changes. AFSD allows this to be possible. It is also not just for future products but also existing products. Instead of waiting for a replacement for forged parts, that same part could be printed on an AFSD machine. This means that the same machine that makes new parts can also make replacement parts. This further condenses the supply chain. All this potential depends on the ability to improve vertical strengths in AFSD parts.

The authors have been able to find only one study looking at how to improve the vertical strengths of the AFSD process. Ghadimi et al. looked at how the vertical and horizontal strengths of AA6061 change by changing the layer height of the deposition. They ran depositions at 3 mm, 2 mm, and 1 mm layer heights. Their tool utilized positive features on the tool surface called knubs; the size of the knubs was not listed. They found little difference between the 2 mm and 3 mm layers but did see a drop in strength at the 1 mm layer. However, the standard deviation for the 1 mm layer was large, and some samples did perform as well as the 2 mm and 3 mm layer heights. The authors also indicated that graphite was used to coat the feed material [15]. Korganci et al. performed an extensive literature review of several materials, including aluminum. They reported properties of 6XXX, 5XXX, and 2XXX alloys [4]. A properties comparison with respect to 7XXX alloys was notably absent from this literature review. This has led the authors to believe that there has not been significant work conducted on improving the build properties of 7XXX alloys using the AFSD process.

Initially, graphite-free AA7075 depositions developed at BYU were performed with a layer height of 4 mm. Several tensile samples were gathered at this layer thickness. However, it was noticed that by reducing the layer height to 3 mm, the vertical strength was improved, but it still fell short of the horizontal strength. This improvement was hypothesized to be due to an increase in shoulder force during the deposition, which resulted in more pressure applied over the same area. However, it was also theorized that other factors were affecting layer adhesion. The two principal factors theorized were oxide growth between layers and poor diffusion bonding. It was thought that a lack of shoulder force (resulting in less pressure) and or not anatomically clean surfaces could be the cause of poor diffusion bonding.

1.4. Study Objective

The objective of this work is to characterize the AFSD layer interface in AA7075. To the authors’ knowledge, no work has been conducted to characterize the mixing or lack thereof of the layer interface during the AFSD process.

Initially, AFSD tools started out with flat shoulders or faces. While flat tools are the simplest to make, they do not mechanically engage with the deposited material beyond the top surface. Later tools with positive shoulder features, known as knubs, were also developed. Part of the reason that the knub tool was developed was to break up the oxides that form during depositions. It was also thought that knubs could improve the flowability of the aluminum during deposition [16,17,18,19]. It is important to recognize that the authors have only been able to find work involving knub tools while using graphite lubricants during deposition. This study will use flat and knub tools along two novel shoulder geometries: scroll and knub–scroll tools.

The scroll shoulder geometry is inspired by FSW tools. Scrolls were first developed in FSW to help improve the mixing of the weld nugget [20]. It was theorized that adding scrolls to the AFSD may help the material mix at the layer interface which would strengthen the diffusion bonding there. The knub–scroll is an attempt to see if combining both geometries can result in better performance than just one feature alone. The authors acknowledge this study will be the first time anyone has attempted to use a scroll or a knub–scroll tool design for AFSD of AA7075.

2. Materials and Methods

2.1. Machine Capabilities

Two machines capable of depositing material via AFSD were used as part of this study: a Bond RM2 (Bond Technologies, Elkhart, IN, USA) and a Meld L3 (Meld Manufacturing, Christiansburg, VA, USA). Both machines were set up to use 3/8 in feed material, which was the size of the material used for this work. Both machines are discrete feed machines.

2.2. Materials Used and Properties

The feed material used for all depositions was AA7075-T7. Feedstock was cut from an extruded plate using a water jet on the BYU campus. The feed material was then washed with clean water and wiped clean to wash any garnet from the water jet process. The composition of AA7075 is listed below in

Table 3. A table with the elemental composition of AA 7075-T7.

Element	Al	Zn	Mg	Cu	Cr	Si	Fe
Wt.%	86.9–91.4	5.6–6.1	2.1–2.5	1.2–2.0	0.18–0.28	Up to 0.4%	Up to 0.5%

.

2.3. Tooling

As mentioned before, four types of tools were used as part of this study. All tools were 38.1 mm wide with a 9.525 mm (0.375 in) square hole cut out of the center of each tool. All tools were made from hardened H13 tool steel. Three of the four tools had positive features that were being tested as part of the study. The other tool was a featureless or flat tool. One tool had 4 knubs placed around the shoulder section of the tool face. One tool had 4 positive “scroll” features analogous to the scrolls seen on FSW tools. The last tool utilized a combination of 2 knubs and 2 scrolls to create a combination knub–scroll tool. All tools were drafted 5 degrees with the draft extending 20 mm up from the face of the tool. All tools also had a 45-degree diffuser or chamfer cut at the exit orifice. CAD models of the tools and their features are shown below in Figure 1.

The depth of the positive features of the various tools is enumerated as part of

Table 4. A table enumerating the feature dimensions of each of the featured tools used as a part of this study.

Tool	Inner Knub Depth (mm)	Outer Knub Depth (mm)	Inner Scroll Depth (mm)	Outer Scroll Depth (mm)
Knub	1.088	1.075	N/A	N/A
Scroll	N/A	N/A	1.013	1.115
Knub–Scroll	N/A	1.11	0.51	0.58

below.

2.4. Study Parameters

The programmed layer heights started at 4 mm and were reduced by 0.5 mm increments until reaching a layer height or until the machine could no longer deposit to exceed the actuator force limit. Depositions were run using essentially the same parameters. The only difference between deposition programs is that scroll and knub–scroll tools ran at 350 rpm as opposed to 300 rpm with the flat and knub tools. Originally, the intention was that all tools were to be run at 350 rpm. However, it was found that at 350 rpm, both the knub–scroll and scroll tools produced excessive amounts of flash to be usable. By lowering the rpm to 300, this created depositions that were visually comparable to the flat and knub tools. Depositions were between 50–75 mm in length. Two-layer stacks were made for each programmed layer height with each tool. The actuator feeding speed for every deposition was 381 mm/pm. Table traverse speeds were determined volumetrically based on the programmed layer height. The scrolls on the scroll and knub–scroll tools were designed to work with the spindle-spinning CW. However, both the CW and CCW directions were tested. The flat and knub tools were only tested in the CW direction. This is because it was anticipated that there would be no significant difference between the CW and CCW directions.

Substrate plates were sanded with 80-grit sandpaper using a pneumatic hand sander. The objective of sanding the substrates was to remove the surface oxide of the aluminum. The removal of surface oxide helps to facilitate diffusion bonding between the 1st layer and the substrate by allowing for a more anatomically clean surface. Given that oxides can grow rather quickly on aluminum, each substrate was used no more than 24 h after being sanded. Once sanded, plates were wiped down with isopropyl alcohol until they were clean. The isopropyl alcohol was applied liberally to remove any leftover dirt or grime that could dirty the surface and impact diffusion bonding. One plate was used per tool. No graphite or any other lubricant was applied to the feedstock or in any other part of the feed system. An example of the experimental setup is shown in Figure 2.

2.5. Sample Preparation

Once depositions had been completed, cross-section samples were extracted from the center of the depositions. Cross-section samples were then cut using a wet bandsaw to be removed from the deposition and to fit into the mounting mold. These samples were then mounted in Bakelite using a LECO MX400 (LECO Corporation, St Joseph, MI, USA). Once mounted, samples were then sanded up to 1200 grit and then polished to a 1-micron finish via aluminum silica suspension on a LECO PX300 (LECO Corporation, St Joseph, MI, USA). Prior to optical examination, all samples were etched using the following Kellers etch solution: 2.5 mL HNO₃, 1.5 mL HCl, 1 mL HF, and 95 mL distilled water. Cross-section samples were then examined under a Keyence VHX-7000 (Keyence Corporation of America, Itasca, IL, USA) optical microscope. Optical images were all taken at a 50× magnification. The samples were polished again at 1 µm using aluminum silica before going to the SEM. All sanding and polishing were performed on a LECO PX 300 (LECO Corporation St Joseph, MI, USA) manual polisher. EBSD samples were polished to 0.05 microns via colloidal silica on a BUEHLER ViboMet 2 (BUEHLER Lake Bluff, IL, USA) vibratory polisher. To help determine the location of layer interfaces during EDS and EBSD scans, fiduciary marks were placed on or near the layer interface using a CLARK CM-402AT (SUN-TEC, Novi, MI, USA) microhardness tester. All SEM, EDS, and EBSD images were taken on an APREO C Low-Vac SEM (Thermo Fisher Scientific, Waltham, MA, USA). The SEM images and EDS scans were taken at a magnification of 3000×. The beam voltage was set to 20 kV, with the beam current being set as 0.40 nA. To ensure that oxide growth did not impact the results of the EBSD scans, both samples were placed under vacuum within 1 h of being removed from the vibratory polisher. A fiduciary mark and prior sample measurements were used to ensure that the EBSD scan occurred across the layer interface.

3. Results and Discussion

3.1. Deposition Appearance and Forces

A total of 34 depositions were made. A table with the images of all the depositions is shown below as

Table 5. A table showing all 34 depositions made as part of this study with all 6 tool types.

Layer Height	4.0 mm	3.5 mm	3.0 mm	2.5 mm	2.0 mm	1.5 mm	1.0 mm
Flat
Knub
CW Scroll						Deposition Exceeded Force Limit	Deposition Exceeded Force Limit
CCW Scroll					Deposition Exceeded Force Limit	Deposition Exceeded Force Limit	Deposition Exceeded Force Limit
CW Knub-Scroll							Deposition Exceeded Force Limit
CCW Knub-Scroll						Deposition Exceeded Force Limit	Deposition Exceeded Force Limit

.

Only the flat and knub tools were able to complete all depositions from 4 mm to 1 mm. The CW knub–scroll made it to 1.5 mm before failure. Both the CCW knub–scroll and the CW scroll were able to make it to 2 mm before failure. The CCW scroll performed the worst, only making it to 2.5 mm before failure. It is clear from the table that the tool type, direction, and layer height all had an influence on the deposition quality.

3.2. Analysis of Forces and Deposition Widths Across Tool Types and Layer Heights

It is important to note that due to the control method of the BOND RM2 machine, feed rates are reduced as the actuator force limit is reached. The rate is reduced in such a way as to attempt to keep the actuator force at or below the limit. This results in depositions being skinnier or underfed if they run at the force limit, which is why we generally see that as layer height starts to decrease, the deposition widths also decrease. A plot of the various deposition widths over the program or commanded (CMD) layer height is shown below in Figure 3.

Deposition width was measured as the widest possible point on the deposition (excluding flash). Most of the tools, except for the CW scroll, started out at or close to the tool diameter. The knub tool maintained the widest deposition widths across the largest range of CMD layer heights. The knub tool did not see a significant decrease in deposition width until a CMD layer height of 2 mm, whereas we start to see a significant decrease in deposition width at 3 mm for the flat, CCW knub–scroll, and CW knub–scroll tools and at 3.5 mm for the CCW scroll tool.

As mentioned previously, due to the BOND RM2 actuator force control scheme, actuator speeds will be reduced to maintain the force limit. As such, we can correlate the deposition widths to both the actuator and shoulder forces during the deposition. Figure 4, as shown below, presents the actuator forces, and Figure 5 presents the shoulder forces.

Apart from the CW scroll tool, the deposition width and shoulder force are generally proportionally related, and both are generally inversely proportional to the actuator force. As the actuator force approaches the force limit, both the shoulder force and deposition width start to decrease. It is clear from Figure 3 that one of the reasons for the knub tools prolonging deposition width is due to lower actuator forces compared to all other tools. The knub tool has the lowest overall actuator forces across all tools and at almost every CMD layer height, while the actuator forces deviate as layer height decreases; at 2.5 mm, the knub tool is almost 20% lower than the flat and CW knub–scroll, and it is almost 30% lower than the CCW spinning tools. The CW scroll is not much worse than the knub (only about 10% higher). However, as we see in

Table 6. A table enumerating the average grain diameters for layer 1, layer 2, and the layer interface for each tool; the relative change between layers 1 and 2 and the layer interface is also considered.

Tool	Average Grain Diameter Entire Sample (Micron)	Average Grain Diameter Layer 1 (Micron)	Average Grain Diameter Layer Interface (Micron)	Average Grain Diameter Layer 2 (Micron)	Average Layer 1 and Layer 2 Grain Diameter (Micron)	% Change Compared to Interface
Knub	1.904	1.739	1.924	1.724	1.732	9.978
Flat	1.804	1.688	1.649	1.634	1.661	−0.76

, the excessive flash and poor deposition surface finish are the most likely culprits for the lower actuator forces, while the CW knub–scroll tool had a lower average actuator force at 1.5 mm when compared to the knub tool. However, the deposition width of the 1.5 mm CW knub–scroll is much lower than the deposition width of the knub tool, still giving the knub tool an edge. Both the CCW knub–scroll and scroll tools demonstrated high actuator forces during deposition. This is likely due to the fact the scrolls were designed to run in the CW direction. So, while in the CW direction, the scrolls help to push out material away from the tool; when spinning in the opposite direction, the scrolls force material toward the center of the tool. This causes increased buildup at the center of the tool, which causes the force on the actuator to spike quicker than the other tools, which is, in part, why CCW tools are the first tools to fail as the CMD layer heights decrease. The lower forces exhibited by the CW scroll tool are likely due to the excessive flash creation during deposition. As the scrolls help to move material away from the center of the tool, this creates more space for the material to flow, which, in turn, reduces the overall actuator force. However, because there is nothing to constrain the ejected material, it manifests as a flash. This causes lower deposition widths and affects the viability of the deposition as well. In terms of actuator force, we see that the CW knub–scroll and flat tools perform almost identically. They also follow similar trends and magnitudes in terms of deposition widths. The only significant difference is in the shoulder force, where the CW knub–scroll tool presents with lower shoulder force after 4 mm deposition when compared to the flat tool.

As mentioned previously, pressure is an important aspect of diffusion bonding. The pressure during deposition is directly related to the shoulder force. Given that all tools used have the same diameter (38.1 mm), a change in shoulder force directly corresponds to a change in pressure. So, in terms of diffusion bonding, it is very plausible that higher shoulder forces help improve layer adhesion. This is where the flat tool performs better when compared to all other tools, particularly the CW knub–scroll and the knub tools. The CCW knub–scroll and scroll tools lose shoulder force quickly as layer height decreases, corresponding to actuator forces that sit around the machine limit. The CW scroll tool never presents with particularly high should forces. While the shoulder force does trend upward, the lack of deposition width still causes problems. We also see a general increase in the shoulder force on the knub tool as we decrease in layer height until approaching the actuator force limit at a layer height of 1 mm. Across all CMD layer heights, the flat tool generally exhibits higher shoulder force when compared to all featured tools; this may indicate better diffusion bonding from a pressure standpoint. As shown in Table 1, post-heat-treated samples made with a flat tool at a 2.67 mm layer height did not perform well. This indicates (as previously hypothesized) that diffusion bonding is not the only factor for effective vertical layer adhesion.

3.3. Analysis of Depostion Mixing Across the Layer Interface

As mentioned in the introduction, it was hypothesized that oxide formation between layers or along the layer interface may play a significant role in the strength of that deposition. As such, an investigation into the apparent mixing and presence of oxide was required. As part of the initial analysis, images were taken from the cross-section of each deposition.

3.3.1. Flat

Above in Figure 6, we see two distinct regions of interest. A darker region towards the top of the layer and the clearer region that exists at the layer interface (both between layers 1 and 2 and layer 1 and the substrate). Generally, as the layer height decreases, the clear interface gets small, presumably because as the layer height decreases, the shear effects of the process become more pronounced (the darker region). We do notice, however, that there appears to be a limit to how deep the darker region can penetrate. We notice that after 3 mm (Figure 6c), there does not appear to be more penetration of the darker region as the layer height decreases.

It is evident in Figure 6 that after we reach a 2.5 mm layer, we have essentially hit the peak penetration. What we see after 2.5 mm is that the clear region essentially stays about the same width, but the darker region gets smaller by virtue of the layer height decreasing. So, the darker region takes up less than the total layer height. We do not see significant evidence that the material in layer 2 is able to interact with the material in layer 1. In all the images, we clearly see the interface that indicates where each layer starts and ends. Layer interfaces are also generally to be horizontal across the sample. This indicates limited interaction between each of the layers, with the flat tool as our baseline as we move to the analysis of all the featured tools.

3.3.2. Knub

When looking at Figure 7, we see that the interfaces are more complicated than those of the flat tool. There appear to be three different regions within a layer. The topmost region is created by the knubs present on the tool. The other two regions appear like what was seen on the flat tool: a darker and clearer region. However, unlike flat tools, we see greater layer interaction at the interface, especially after we get below 3 mm in layer height. At the 2 mm layer, we start to see the interaction between the first layer and the substrate. We also see a more wavy or non-horizontal interface in the cross-sections of the knub tool. This becomes more apparent starting in Figure 7d. We also see the onion ring pattern begin to persist through the layer interface. Onion ring patterns are evidence of strong material mixing. As those onion ring patterns begin breaking through the layer interface, it is possible that it may result in a strong interface during tensile testing. This is because the onion ring penetration helps to essentially dissolve the layer interface such that there is no interface (or as much of an interface). Increasing the consolidation of the layer interface helps to reduce possible crack initiation sites. This is because the layer interface essentially starts to appear and possibly behaves like a bulk layer material, which is stronger than the interface.

After reaching a layer height of 1.5 mm, we start to see the effects of the actuator force approaching the force limit. While the layer interface looks well mixed, we see the deposition degradation along the top surface of layer 2. This degradation becomes more apparent in Figure 6g at the 1 mm layer height. Given that the knubs on the knub tool were just over 1 mm, it makes sense that these would present with some of the best overall mixing. You are almost able to restir the previous material when depositing the subsequent layer, which is great for interface mixing, but high actuator forces prevent the deposition from approaching the full tool width. However, it is still possible that at the 2 mm or 2.5 mm layer height, interface mixing is enough to create a stronger bond than that of the flat.

3.3.3. Knub–Scroll

CW

With the addition of the scroll in combination with the knubs, we immediately see an improvement in the interface mixing, as shown above in Figure 8. All cross-sections present very clear onion ring patterns, which is to be expected given that the scrolls on these tools were intended and designed to behave in a similar fashion to scrolls found on FSW tools. While the mixing is good, what detracts from these samples are the visible voids. Both the knub and flat tools did not present any visible voids at the magnification used. All samples, apart from the 3 mm sample (Figure 8c), have voids along the layer interface. Voids are undesirable between layers as they provide an explicit area for crack propagation to begin. A void, even a small void, is the perfect location for crack initiation during tensile loading. While some peripheral voids could be removed during post-deposition machining, any interior voids would still act as crack initiation sites. The voids effectively counteract the improved onion ring patterns created by the addition of the scrolls. It is possible that given the scroll geometry (size and shape), as well as the deposition parameters created, a condition where there was too much material flow exists. The deposited material was moved around so much to the extent that it had a hard time reconsolidating itself (especially on the periphery where the torque would be the highest). As such, it is also important to note that the parameters used for this study could be improved. Optimized parameters might be able to eliminate the voids seen in the cross-sections. It is also possible that the tool features themselves could be further optimized.

CCW

As mentioned previously, the CCW knub–scroll tool was one of the worst-performing tools. As shown above in Figure 9, after a layer height of 3 mm, the deposition quality deteriorated rapidly, coinciding with the high actuator forces. The CCW knub–scroll cross-sections also do not appear to be significantly better than the CW knub–scroll. While the CW knub–scroll does show signs of deterioration at the 2.5 mm layer height, it is not to the degree that is shown here with the CCW knub–scroll. As such, it does not appear that utilizing the knub–scroll tool in the CCW direction provides any significant benefit over using the tool how it was designed to operate, which is CW.

3.3.4. Scroll

CW

It is clear from Figure 10 above that the CW scroll tool has some of the worst-preforming cross-sections out of all the tools. Not only are voids present, but entire sections of the cross-section are not consolidated. Even at thicker layer heights, there is a significant lack of consolidation between layers. This is likely due to the scroll features acting unrestrained. In the CW direction, the scrolls are designed to spread out material. Without knubs to help restrain the material, we see that the CW scroll tool is mostly ineffective at creating a consolidated layer interface. While the layers themselves present with significant mixing due to the interaction of the scrolls during deposition, the scrolls are unable to create a layer interface that is completely consolidated across the width of the deposition.

CCW

As shown in Figure 11, the CCW scroll initially presents well-consolidated interfaces at the 4 mm and 3.5 mm layer heights. As shown previously in Figure 3, the CCW scroll had the highest actuator forces, even at thicker layer heights. Not only did this cause the CCW scroll tool to fail sooner when compared to other tools, but we also saw large voids along the interface after the 3.5 mm layer height. The 4 mm and 3.5 mm layer heights do show promise, but the significantly higher actuator forces narrow the operating window. As mentioned previously, further parameter and feature optimization could result in different outcomes than the one shown in this study. However, with all things being as equal as possible, the CCW scroll struggled in comparison to the knub tool and even the flat tool.

3.4. EDS Scans of Layer Interface

While the images taken on the optical microscope were informative, there was still the question of oxide presence at the layer interfaces. It was decided that EDS scans of the layer interface between layer 1 and layer 2 would be necessary.

These EDS scans primarily looked for the presence of oxygen on the surface of the sample. All samples were repolished at 1 micron to remove any etchant and any surface oxide that had built up. It was decided that the 2.5 mm cross-sections would be used for the EDS scans. This was conducted so that these scans could be used in comparison with previous tensile data, which were also gathered at a layer height closer to 2.5 mm. As mentioned in the method section, fiduciary marks were placed as close to the layer interface as possible. EDS scans were then performed at the point furthest into the sample, where the layer interface could be distinguished. If the layer interface could not be distinguished (other than with the presence of the fiduciary mark), then a scan was taken in approximately the middle of the sample across where the layer interface should be. If the layer interface was present across the entire width of the sample, then the scan was taken near the fiduciary mark. All EDS scans were performed using the general orientation and scan direction, as shown below in Figure 12.

3.4.1. EDS Scans of Flat Interface

After scanning the layer interface along the flat sample, a small void was found along the interface. The void was found approximately 3.36 mm from the edge of the sample. The SEM image and EDS are shown in Figure 13.

The plot shown in Figure 13b clearly demonstrates a sharp increase in oxygen along the void along the layer interface. This results in almost a 112% increase over the background oxygen counts. This suggests a large formation of oxide along that portion of the layer interface.

3.4.2. EDS Scans of Knub Interface

With regard to the knub tool, the authors could not find a clear indication of a void along the layer interface. As such, a scan was taken in approximately the middle of the sample, and the results are shown in Figure 14. The EDS scan shows little to no increase across the layer interface (approximately a 12% variation between peak and background counts). This may suggest that the interface is more fully consolidated as there is no increase in oxygen as the scan crosses over the interface. However, this represents one scan across one location. It is difficult to assume how these samples would perform in tensile tests based on just one scan. However, the interface is more consolidated when compared to any of the other tools, indicating that the knub tool has possibly greater potential to produce strong tensile and fatigue results.

3.4.3. EDS Scans of CW Knub–Scroll Interface

The last clear indication of the interface was located approximately 2.12 mm from the edge of the sample. A scan was performed there, with the results being presented in Figure 15. The CW knub–scroll interface presents with similar but slightly different behavior when compared to the flat. We do see a sharp increase in oxygen across the line scan. However, the increase occurs to the right of the interface on the second layer, which may indicate more oxide formation just to the right of the interface and not on the interface itself. The peak in oxygen count does result in a 43% increase over the background oxygen count. The more concerning issue, however, is the fact that the interface is still apparent. While it is not apparent across the whole sample, it still creates a possible crack initiation site, which could potentially negatively impact the performance of tensile properties.

3.4.4. EDS Scans of CCW Knub–Scroll Interface

Unlike the flat, knub, and CW knub–scroll, the interface was visible across the entire width of the sample. The scan and image shown in Figure 16 were taken close to the fiduciary mark that was placed approximately 6 mm from the edge of the sample. Like the CW knub–scroll, we do not see an increase in oxygen right on the interface; we see it to the right on layer 2. However, upon closer examination, we do see that one of the oxygen spikes occurs on a deviating line that stems from the layer interface. This results in almost a 100% increase in oxygen count across the interface. The large spikes at the end of the scan are caused by the white particle that is on the far right side of the scan.

3.4.5. EDS Scans of CW Scroll Interface

Like the CCW knub–scroll, the layer interface on the CW scroll was clearly evidenced across the entire width of the sample. However, as shown above in Figure 17, there is no sharp increase in the presence of oxygen when crossing the interface. The oxygen content does increase a little, but not to the extent that we see in the previous samples (only about a 30% difference between peak and background counts). There is a large spike at the end of the scan. Upon looking at the SEM images in Figure 17a, there does not appear to be any particle that the scan crossed over. As such, it is possible that the scan could detect an increase in oxide formation in that region, which is only 10 microns from the visible layer interface. However, like previous samples, the more significant concern is the fact that the layer interface was visible across the entire width of the sample, which suggests poor layer interface consolidation. Like the CCW knub–scroll, the scan for the CW scroll was also taken 6 mm from the edge of the sample.

3.4.6. EDS Scans of CCW Scroll Interface

For the CCW scroll, as shown above in Figure 18, the scan was taken approximately 2 mm from the edge of the sample, which was the furthest indication of the layer interface. This sample is interesting because the oxygen count decreased when crossing the interface. We do see a large spike just to the right of the interface, like what was observed on other samples. This results in almost a 73% increase in oxygen counts across the layer interface. While the layer interface was not apparent throughout the entire width of the sample, it still being apparent for a few millimeters is still problematic. The visible layer interface is particularly problematic when considering possible tensile performance.

3.4.7. EDS Scan Comparison

When we plot the approximate background oxygen against the estimated interface oxygen, we obtain the following chart, shown below as Figure 19.

We see that the flat and CCW scroll tools are the worst-performing out of all the other tools. When comparing the peak oxygen count to the background oxygen count, we see an increase of 100% or higher. The CW and CCW knub–scroll and the CW scroll perform better; however, they are still worse than the knub tool. The relative change in oxygen counts was lower than 100%, but even the lowest relative change (30% for the CW scroll tool) was still almost double the relative change of the knub tool (12%). The 12% relative change for the knub tool interface is easily within the noise of the machine and the natural variation of oxide grown across an aluminum surface. It is possible that the oxygen count could differ at other places along the whole knub sample. However, the fact that the middle of the sample shows no change is a positive indication of a stronger layer adhesion when compared to the other tools.

It is important to recognize that while the scans are compelling, they only represent one point along the entire width of the interface. To provide a more comprehensive and complete analysis, more data points should be taken. This would allow for a full oxide profile to be established across the sample. However, the initial scans do help us in our understanding of oxide growth along layer interfaces. This allows us to make initial tooling recommendations that will hopefully get us closer to achieving isotropic build properties.

We also need to consider the possibility that oxides could be pushed out toward the edge of a layer during deposition. However, if the oxides were always pushed toward the edge of the sample, those would theoretically be removed from the sample via post-deposition machining. The samples presented here are slightly cut widthwise in many instances for the sample to fit more comfortably into the mounting machine. If the oxides only congregated at the edge of the sample, then we would likely only see oxide at the edges of the sample. However, scans for all tools, except the knub, were taken between 2 and 6 mm from the edge of the sample. These scans still showed oxide despite not being on the edge. This work does not rule out the idea that some amount of oxide may be pushed to the edge. However, this work does show that significant oxide presence can be detected on the interior of a sample, indicating that oxides can form anywhere along the sample, which would negatively impact the adhesion of the layer interface.

3.5. EBSD

To help gain a better understanding of the grain size distribution of the layer interface, EBSD scans were performed on the 2.5 mm flat and knub cross-sections. Those samples were chosen specifically because, as shown in Figure 19, the flat and knub tools represent the two extremes of the spectrum. By picking the two extremes, we hoped to gain enough contrast to be able to gain an additional understanding of the layer interface. The resulting EBSD scans for the flat and knub samples are shown below in Figure 20 and Figure 21.

Each scan was cropped into sections. The flat tool had evidence of the layer interface in its scan, and as such, that area was sectioned specifically, with the remaining areas presumed to be layer 1 (above) and layer 2 (below). However, with the knub tool, there was no clear indication of the layer interface (despite the scan occurring over where the interface should be); as such, the scan was split into three equal parts, approximately representing layer 1, the layer interface, and layer 2.

It is evident from both Figure 20a and Figure 21a that there is a difference between the layer interfaces. The flat interface presented an area where it appeared that the grains were smaller and slightly more elongated when compared to the rest of the scan. This occurred approximately where the layer interface is presumed to be. We do see, however, from Figure 20e and Figure 21e that there is no large change in grain diameter, even in the various sections. There are some slight variations, and this becomes more apparent when we look at the average grain diameter for each section, shown below in Table 6.

Jeon et al. reported that the average grain diameter for the hot-rolled 7075 plate was approximately 34.2 microns [21]. As shown in Table 6, the two samples examined had an overall average grain diameter between 1.8 and 1.9 microns. That shows a significant reduction in grain size because of the AFSD process.

When comparing the flat and knub tools, we see that numerically, the grain sizes are similar across the entire samples (only a 0.1 µm difference). When we break down each of the sections, we see that there is some slight variation in the knub tool, while the flat tool, on average, stays constant through the width of the sample. While these average sizes are interesting, they do not quite paint the whole picture. In fact, the average grain size for the entire sample is small enough that it could fall within the noise of the machine. As such, the most useful analysis would be to consider the IPF. The IPF for the knub tools shows a more consistent distribution of grains. The middle section does have a larger average diameter, but looking at the entire sample, it is difficult to see a distinction between where layer 1 ends and layer 2 begins, whereas, with the flat sample, we do see indications of where the layers end and begin. The cluster of smaller grains creates a distinct boundary between the two layers. The variation is subtle, but subtle differences in microstructure can make a significant difference in mechanical properties [22,23].

The EBSD scans presented in this paper are by no means a comprehensive analysis. There is still significantly more work that can and needs to be conducted. However, these scans do support the previous findings of this paper. As there existed little to no indication of a layer interface when using a knub tool, likewise, when using a flat tool despite apparent macroscopic consolidation, there were visual indications of the layer interface at the microscopic level.

4. Conclusions

The objective of this work was to provide an initial characterization of the layer interface during the AFSD process using 3/8 in feed material. Various external tool geometries were also explored, with the intention of determining how changes in those geometries can affect the layer interface. Several graphite-free depositions were made across a variety of layer heights. Two of the tool geometries utilized in this study have never been used before, and depositions were successfully created. Not only did changing the external features of AFSD tools affect the process operating window, but they also impacted the layer interface at both the macroscopic and microscopic levels. Microscopy at the optical, SEM, and EBSD levels was successfully achieved. These results help us look at interface consolidation as various magnifications. We were also able to visually identify differences in the grain structure between the flat and knub tools thanks to the EBSD scans. The following conclusions were also drawn because of this study:

• The knub tool showed a significant improvement over the flat tool with regard to interface consolidation at the optical and SEM levels; it was also void-free. However, like all the other tools tested, knub tool depositions did lose layer width, and layer height decreased.
• The knub tool did not have a clear indication of the layer interface at 3000× magnification, and an insignificant 12% increase in oxygen was detected across the interface.
• The knub tool presented with the lowest overall actuator forces, and at the recommended layer height (2.5 mm), it was between 20 and 30% lower in force than the other tools (excluding the CW scroll, which had poor deposition quality), indicating it has the largest operating window for parameter optimization.
• All tools that utilized scrolls showed lots of material interface within the bulk of the material; however, they did not present fully consolidated layer interfaces when subjected to SEM.
• Both the CCW scroll and knub–scroll tools presented significantly higher actuator forces when compared to the same tools that ran in the CW direction.
• EBSD scans of the flat and knub samples provided visual indications that the knub tool had better interface consolidation. However, more data points need to be collected to further validate this finding.

Based on the conclusions, the authors recommend depositing with a knub tool at a layer height between 2.0 mm and 2.5 mm. Given the parameters used and the results of this study, a knub tool at a 2.5 mm layer height produced the most visually and quantitatively consolidated layer interface.

5. Future Work

As mentioned previously, there exists significant opportunity for work moving forward.

This study only explored one size and shape of both the knub and scroll features. Other shapes and sizes could be considered, and those will likely yield different results. It is possible that different knub geometries may be more effective than others and improve interface mixing and breaking up of oxides.

Parameter optimization is also a huge space for work to be conducted. In this study, all the parameters were kept as consistent as possible. This was conducted so that the performance and characterization of each tool could be correlated to the other tools used in this study. The authors make no claim that the parameters used in this study are the best and only parameters possible. As such, it is possible that an improved parameter set may allow for the knub–scroll tool to produce void-free depositions. There is also a need to go further to characterize the entire layer interface at the SEM level. More samples and more scans (particularly EDS and EBSD) will help strengthen the characterization of the layer interface of AFSD depositions. This could also help provide a more definitive answer to how interfacial oxygen affects the build direction tensile strength.

All the work here was performed with 3/8 in the feed material. As such, there exist many opportunities to perform similar studies with 1/2 in the feed material and tooling; 1/2 in the feed material may in and of itself also improve the layer interface. This could be due to expanding the surface area of the feed material in contact with the substrate during the deposition because of 1/2 in the feed material.