Diffusion Bonding and Transient Liquid Phase (TLP) Bonding of Type 304 and 316 Austenitic Stainless Steel—A Review of Similar and Dissimilar Material Joints

: Similar and dissimilar material joints of AISI grade 304 (1.4301) and AISI grade 316 (1.4401) austenitic stainless steel by solid state diffusion bonding and transient liquid phase (TLP) bonding are of interest to academia and industry alike. Appropriate bonding parameters (bonding temperature, bonding time, and bonding pressure) as well as suitable surface treatments, bonding atmosphere (usually high vacuum or protective gas) and interlayers are paramount for successful bonding. The three main parameters (temperature, time, and pressure) are interconnected in a strong non-linear way making experimental data important. This work reviews the three main parameters used for solid state diffusion bonding, TLP bonding and to a smaller degree hot isostatic pressing (HIP) of AISI grade 304 and AISI grade 316 austenitic stainless steel to the aforementioned materials (similar joints) as well as other materials, namely commercially pure titanium, Ti-6A-4V, copper, zircaloy and other non-ferrous metals and ceramic materials (dissimilar joints).


Introduction
Diffusion bonding is a solid-state welding technique capable of joining similar and dissimilar materials. The process operates on the principle of solid-state diffusion, wherein the atoms of two solid materials intersperse themselves over time at elevated temperature and pressure so that a highquality bond between the base materials is formed [1]. Solid state diffusion bonding is attractive for the joining of dissimilar materials since the formation of brittle intermetallic compounds (IMCs) as well as chemical segregation and accumulation of residual stress at the bond interface can be relatively low if compared to other welding techniques [2][3][4]. Compared with other joining techniques, diffusion bonding is expensive and only widespread in aerospace engineering. Besides being able to join dissimilar materials, diffusion bonding can be used for full cross-section joining that may be particularly interesting for holohedral joints, such as internal structures of plate-type heat exchangers [5,6]. Transient liquid phase (TLP) bonding differs from solid-state diffusion bonding in a way that a thin interlayer, that has a lower melting point than the base materials and can melt below the bonding temperature, is placed between the base materials that are to be joined. The interlayer element (or constituent of an alloy interlayer is used) diffuses into the base materials, causing isothermal solidification at the bonding temperature. As a result, a bond that has a higher melting point than the initial TLP bonding temperature is formed [7][8][9][10]. In vacuum brazing the interlayer (or filler material) that is placed between the base materials does not completely diffuse into the base materials, so that isothermal solidification does not occur and the resulting bond still has the melting point of the interlayer (filler) material. Additionally, no relevant pressure is applied in vacuum brazing allowing for much simpler equipment. For diffusion bonding, special and expensive equipment is required as parts have to be heated at high temperatures in a vacuum or protective gas atmosphere while applying considerable amounts of controlled pressure. It is worth noting that interlayers can also be used for solid-state diffusion bonding. This can for instance be useful if a large difference in coefficient of thermal expansion in two dissimilar materials has to be overcome. To be labelled a solid-state diffusion bonding process, melting of the used interlayer needs to be prevented.
Besides diffusion bonding, TLP bonding and vacuum brazing, hot isostatic pressing (HIP) can be differentiated. HIP compresses materials and can also be used for bonding materials by applying high temperatures of several hundreds to 2000 °C and isostatic pressure of several tens to 200 MPa at the same time. Argon is the most commonly used pressure medium. HIP applies isostatic pressure to materials using gas pressure, while in diffusion bonding and TLP bonding uniaxial pressure is applied using a ram. Figure 1 provides a brief schematic overview of how the aforementioned techniques, diffusion bonding and TLP bonding (left), vacuum brazing (center) and HIP (right) can be used to join two plates (yellow). AISI grade 304 (1.4301) and AISI grade 316 (1.4401) are the most commonly used austenitic stainless steels. Grade 316 stainless steel contains additional molybdenum (Mo) that gives it improved corrosion resistance. The low-carbon versions, 304L (1.4307) and 316L (1.4404) are used to avoid corrosion problems caused by welding. The "L" means that the carbon content of the alloy is below 0.03%. The low carbon, nitrogen-enhanced version of grade 316 steel is 316LN (1.4406). The nitrogen content in this steel provides solid solution hardening, and raises its minimum specified yield strength. The higher carbon variant of 316 is 316H (1.4919) which makes the steel more suitable for use in applications where elevated temperatures are present. Table 1 provides an overview of the typical chemical composition of 304, 304L, 316, 316L, 316LN and 316H austenitic stainless steels [11][12][13][14] that are discussed in this work.  [15][16][17], production of plate type heat exchangers or chemical reactors [18][19][20][21][22][23][24][25]  [34] present a systematic study for diffusion bonding of 304 stainless steel that puts particular focus on the deformation during diffusion welding. In a clever way, cone-shaped samples ("Gietzelt-Cones") consisting of several layers with slightly different areas were diffusion bonded so that different applied loads could be investigated with one sample diffusion bonded using the same force.
In this work we review similar and dissimilar material diffusion bonding and TLP bonding experiments conducted to join 304 and 316 austenitic stainless steel, including their low-carbon versions 304L, 316L and 316LN as well as the higher carbon variant 316H. Specifically we report on the different bonding parameters (bonding temperature, bonding pressure and bonding time). We further report the selected interlayers. Interlayers are particularly relevant if different materials with different coefficients of thermal expansion are to be joined [35,36]. Although diffusion bonding can be conducted in air [37][38][39][40] it is usually conducted in high vacuum or under protective gas atmosphere. We therefore do not explicitly report on the used atmosphere of the reported experiments. Besides the used atmosphere [1], surface treatments [41], sample geometries, furnace design, pre-and post-treatment of samples can have a significant impact on the quality of the diffusion bond. Reporting on all these different factors would go beyond the scope of this work and we thus decided to limit reporting to four factors: type of interlayer, (bonding) temperature, (bonding) time and (bonding) pressure that we found most relevant. We define bonding pressure, bonding temperature and bonding time as the parameters during the actual diffusion bonding process. Figure 2 provides a brief overview of the basic diffusion bonding process and the three reported parameters. In the basic diffusion bonding process, as we define it here, samples are fixed at a low pressure and heated with a certain heating rate until the desired bonding temperature is reached. When the desired bonding temperature is reached the pressure will usually be increased to the desired bonding pressure. Both bonding temperature and bonding pressure will be kept constant over the bonding time (sometimes referred to as holding time or soak time). After the defined bonding time, the pressure will usually be reduced (in Figure 2 it was reduced to the initial temperature used to fix the sample) and the temperature is gradually reduced to room temperature.
Fick's first law with the following temperature (T) dependent diffusion coefficient (D): is usually used to represent the diffusion bonding process (D0 is the frequency factor, Q the activation energy and R the gas constant). Atomic diffusion across the mating surfaces normally occurs before reaching the defined diffusion bonding temperature and similarly the process of diffusion still occurs while the sample cools down. Therefore, looking at bonding pressures, bonding temperatures and bonding times does provide a good overview, but does not tell the whole story of the conducted experiments. Song  Besides temperature, actual bonding pressure is challenging to thoroughly report on as well. In the basic diffusion bonding process that was introduced earlier, pressure is applied in a controlled manner using a ram. Not only different constant pressures but also varying the same pressure ("hammering") can have an effect on the quality of the diffusion bond. This may best be illustrated by works on impulse pressure diffusion bonding (IPDB), where the bonding pressure is increased and decreased several times over the course of the actual diffusion bonding time [47][48][49][50][51][52][53][54][55][56]. Diffusion bonding is also performed with vacuum furnaces that do not possess a ram and can thus not apply or control the applied pressure. Akhter et al. [57,58] and Munis et al. [59], for instance, report on diffusion bonding of 316L stainless steel and Zircaloy-4 samples in a vacuum furnace in which pressure was applied by putting the samples in a vice machine for 24 h before the actual heat treatment to introduce plastic deformation on the bonded surfaces. Samples were then wrapped in Kanthal wire that has a lower coefficient of thermal expansion than 316L and Zircaloy-4 so that pressure builds up during heating. Taking advantage of different coefficients of thermal expansion to generate pressure for diffusion bonding is not unusual [60][61][62] and can be complemented by clamp-like apparatuses that apply compression using screws [63][64][65][66][67]. If the sample geometry allows it, parts can also be press-fitted and put under pressure into a vacuum furnace where they are then diffusion bonded when the temperature is increased [68][69][70]. If no relevant pressure was applied [67,[71][72][73][74][75][76][77][78][79][80][81][82][83][84][85][86][87], the experiments were not reported here since we considered them brazing rather than diffusion bonding using the definitions introduced earlier. In brazing some pressure is often applied to fix the parts so that a definition between brazing and TLP bonding that only relates to applied pressure is not sufficient. An interesting case is for instance the experiments conducted by Xia et al. [88][89][90] who used a deadweight to apply 0.02 MPa pressure during the whole bonding process, so that these works on joining of 316L stainless steel with titanium and titanium alloy (Ti-6Al-4V) could be considered diffusion bonding studies, or more accurately TLP bonding, since melting interlayers were used that at least partly diffused into the base materials. Since the weight was so little, and more importantly the interlayers did not completely diffuse into the base materials, we did not consider them diffusion bonding or TLP bonding studies in this review. We consequently did also not consider the weight of the upper part of a diffusion bonding sample a dead weight or applied pressure.
It is worth noting that the applied pressure in diffusion bonding is used to bring the base materials closer together, and overcome oxide layers and asperities of the faying surfaces. The applied temperature, as indicated in Equation (1) has by far the largest effect on the diffusion bonding process. Other related diffusion bonding processes such as hybrid friction diffusion bonding (HFDB) of stainless steels [91,92] were also not considered in this review. It can also be argued that sintering is a form of diffusion bonding small kernels, or diffusion bonding is a form of sintering large plates. We do not argue either way and did not consider sintering experiments of 304 and 316 stainless steels [93][94][95][96][97][98][99][100][101][102] in this study.
Rather than being comprehensive, this work aims to provide a focused overview on conducted similar and dissimilar diffusion bonding experiments of grade 304 and grade 316 stainless steel. To reproduce experiments, we strongly recommend reviewing the relevant references to learn about the detailed parameters which were not reported here, such as surface preparations, pre-and postbonding treatment, that can have a significant impact on the overall quality of the achieved bond.

Materials and Methods
For this study more than 1200 papers published by various authors were reviewed using Scopus, Web of Science and Google Scholar. Four factors: interlayer, (bonding) temperature, (bonding) time and bonding (pressure) that we found most relevant were reported from nearly 140 studies. Figure 3 provides an overview of the considered scientific publications on similar and dissimilar diffusion bonding and TLP bonding of 304 and 316 stainless steel over the past 20 years. In detail, we reported the factors recommended by the authors. If the same result would be achieved with different bonding conditions, we would recommend the lower, easier and less costly parameter. If, for instance, the same bond quality, determined by tensile testing, could be achieved at 800 °C and 900 °C (P = const, t = const), then we would recommend using 800 °C. We further placed emphasis on the mechanical strength of the reported diffusion bonds, usually measured with tensileor shear tests. Depending on the application at hand, other parameters such as deformation not reported here may be equally important to those conducting the experiments.  Table 2 did unfortunately not report on the mechanical strength of the bond, so it is hard to quantitatively compare the achieved results.  Table 3 reports on similar diffusion bonding and TLP bonding experiments of 316 stainless steel. Just like similar bonding of 304 stainless steel this is a common process in industry so that the conducted experiments are generally concerned with additional influences or optimizations such as the impact on roll bonding on the diffusion bonding behavior [111] or the investigation of fatigue [112] and corrosion [112] behavior that go beyond "simply" reporting bonding parameters. A comparison is also difficult since most publications did not conduct mechanical tests that generate quantitative results, and if quantitative results are reported they are often not put in perspective by comparing them to the strength of the base material experiencing the same heat treatment.

Diffusion Bonding and TLP Bonding of 316 (Similar Joints)
Yeh and Chuang [113] provide excellent documentation for their study on diffusion bonding of 316 stainless steel rods. The researchers compare diffusion bonding of 316 stainless steel samples with and without interlayer (Dux 65: wt% 23.8 Cr, 5.9 Ni, 1.5 Mo, 1.1 Cu, 0.7 Si, 0.7 Mn, 0.14 N, 0.05 Al, 0.035 P, 0.03 C, 0.002 S, balance Fe) at different surface finishes. With a high surface finish, 99% of the bond strength of the base material could be confirmed through tensile testing (samples still fractured at the bond plane), while samples with lower surface finishes reached 38% of the base materials bond strength. Using the Dux 65 interlayer the diffusion bonds outperformed the base material with high and low surface finish and samples consequently fractured in the base material. These high bond strengths are not unusual and correspond well to the 94% ultimate tensile strength (566 MPa) of the base material reported for the solid-state diffusion bonding without the interlayer of 316L reported by Li et al. [114]. Similar results have also been reported by Mateus et al. [115]: ~550 MPa (also diffusion bonding of 316L without interlayer).  Table 4 reports on dissimilar joining by diffusion bonding and TLP bonding of 304 and 316 stainless steel to commercially pure titanium (cpTi). Titanium and its alloys show extraordinary mechanical properties and corrosion resistance. The widespread usage of titanium and its alloys is, however, limited as a result of the relatively high costs. To save costs titanium is frequently joined to more common and cheaper structural materials such as stainless steel, so that the individual components of both materials can be fully exploited and the overall cost of the structures can be reduced. Diffusion bonding and TLP bonding of titanium and titanium alloys finds particularly wide application in aerospace engineering [126] which can explain the relatively large number of experiments conducted and reported in Table 4. Compared with similar diffusion bonding and TLP bonding of 304 and 316, considerably lower temperatures were used. A large number of researchers reported detailed mechanical bond strengths measured using tensile tests for joining of 304 and cpTi so that we can attempt to compare the reported ultimate tensile strengths (UTS) with one another as was done in Table 5. Detailed references of the cited twelve studies are provided to give credit to the research-groups conducting them. Relatively high bond strength ranging from 70% to 100% of the weaker base material cpTi were reported without and with a number of interlayers. The reported UTS of the base material, cpTi Grade 1 and cpTi Grade 2 vary considerably so that we considered the relative bond strength measured in % UTS of the cpTi, that is the weaker base material, as the most relevant criteria. Cu-interlayers of 300 µm thickness seem to accomplish particularly good results. In general, samples with interlayers show higher bond strength than samples that were bonded directly. The study conducted by Shirzadi et al. [133] is an exception to this observation. In this study samples without an interlayer do in fact outperform the samples with a thin 0.5-1.1 µm Ni interlayer. In their study Shirzadi et al. [133] use gallium to remove the oxide layer prior to diffusion bonding so that this can be considered a special case.  [143] 304 Cu (300 µm) Gr 1 319 MPa 322 MPa 101% 3 1 As reported by study, 2 Gallium assisted diffusion bonding, 3 Sample ruptured in the cpTi base material at a UTS of 101% of the cpTi base material.

Diffusion Bonding and TLP Bonding of 304 and 316 to Ti-6Al-4V
Besides diffusion bonding and TLP bonding of 304 and 316 stainless steel and cpTi, a large number of publications looked into diffusion bonding of 304 and 316 stainless steel to Ti-6Al-4V (Grade 5). Ti-6Al-4V is the most used titanium alloys that due to its balance (as α-β alloy) has low density and excellent corrosion resistance. Diffusion bonding and TLP bonding parameters are reported in Table 6. Since a large number of researchers reported UTS and shear strength of their experiments, we also directly compared these results (UTS left and shear strength right) in Figure 4. It was noticed that among the 13 papers available that specifically bond 304 to Ti-6Al-4V, ten used interlayers to facilitate the atomic diffusion process and therefore acquire high quality bond strengths. Nevertheless, the solid-state diffusion bonding study by Ghosh et al. [147] achieved with 342 MPa the highest UTS reported for direct bonding of 304 to Ti-6Al-4V. Only the study by Song et al. [42] who joined 316L to Ti-6Al-4V using a Cu-Nb interlayer and a two-step diffusion bonding process mentioned in the introduction, reached a higher value (489 MPa). The reported UTS ranged from as low as 183 MPa [148] to the aforementioned high value reported by Song et al. [42]. The reasons for this may be found in the different user requirements for the bond. Chandrappa et al. [148] who reached the 183 MPa UTS, did for instance conduct joining experiments at a relatively low maximum temperature of 550 °C, while all others used bonding temperatures ranging from 750 to 1100 °C. With aluminum the researchers further used an interlayer that is notoriously difficult to join using diffusion bonding [149][150][151][152]. Reported shear strengths range from as low as 118 MPa to as high as 385 MPa. The highest reported value was again reached with a combination of 316L and Ti-6Al-4V as well as a Cu-Ni interlayer. The five studies with the highest reported shear strength all used copper interlayers.   Table 7 presents diffusion bonding and TLP bonding parameters of 304 stainless steel with Cu and CuZrCr that may be particularly relevant for parts in nuclear fusion applications. Akbar et al. [170] and Yilmaz [171]  further used a Gleeble 1500 thermo-mechanical material testing system. These systems heat samples by applying a high electric voltage directly between the two ends of the specimens [174][175][176][177]. In this way, Yuan et al. realized heating rates of 5 °C/s that are much higher than those of classic diffusion bonding or TLP bonding (max. 10 °C/min) where samples (often plate-shaped) are heated (ideally uniformly to avoid internal stress) through radiation from induction heating elements in the vacuum furnace. IPDB is nonetheless very promising as it may reduce costs for expensive interlayer materials or coating procedures during bonding. IPDB may further reduce the overall diffusion bonding time which is again beneficial.

Diffusion Bonding and TLP Bonding of 304 and 316 to Zircaloy and Zr-Sn-Nb
Zircaloy and other zirconium alloys such as Zr-Sn-Nb are widely used in the nuclear industry due to their favorable neutron cross sections, adequate mechanical properties and excellent corrosion resistance [186,187]. Table 8 lists reported dissimilar diffusion bonding and TLP bonding studies of 304 and 316 stainless steel to Zircaloy and Zr-Sn-Nb. Zircaloy-2 and Zircaloy-4 have a melting point of 1850 °C [188], that is higher than the melting points of copper (1085 °C) and titanium (1668 °C). Dissimilar diffusion bonding and TLP bonding of 304 and 316 to Zircaloy and Zr-Sn-Nb was subsequently (in most cases) conducted at higher temperatures than dissimilar bonding of stainless steel to copper and titanium. Only three studies reported mechanical bond properties, so a quantitative comparison between the studies is challenging. Bhanumurthy et al. [189] joined 304L stainless steel to Zircaloy-2 using sandwiched Ni-Cu-Nb interlayers of 30-70 µm thickness each. The accomplished bonds reached UTS of up to 450 MPa, and average UTS of roughly 400 MPa. Pan et al. [190] and Srikanth et al. [191] both joined 304L stainless steel to Zircaloy-4 reporting accomplished shear strengths of 166 MPa and 209 MPa, respectively.  Table 9 provides an overview of reported diffusion bonding and TLP bonding experiments of 304 and 316 stainless steels to other materials not discussed in the previous chapters. Magnesium alloys (AM60 and AZ31), mild steel (EN3B), low alloy steel (A533B), Nickel, titanium alloy 6246, stainless steels, cast iron and Kovar were bonded with 304 and 316 stainless steel. In addition, a number of ceramic materials (Al2O3, Ni3Al, Si3N4, Sialon, TiC, WC and ZrO2) were bonded to 304 and 316 stainless steel using diffusion bonding or TLP bonding. The large differences in the coefficient of the thermal expansion of steel and ceramics make joining challenging and is the reason that interlayers are often used [202,203].

Hot Isostatic Pressing (HIP) of 304 and 316 (Similar and Dissimilar Joints)
HIP was invented in 1955 for diffusion bonding applications in the nuclear industry [228]. Besides interfacial bonding, it finds much wider application now in upgrading castings, densifying pre-sintered components, consolidating powders [229,230] and post-treating parts produced by selective laser melting (SLM). HIP can also be used to eliminate voids at diffusion bonded interfaces [108]. HIP is often used for diffusion bonding of first-wall structures in fusion reactors. Most notably CuCrZr [231] and dispersion strengthened copper (DS Cu) [232] are bonded. Studies examining HIP of similar and dissimilar joints of 304 and 316 stainless steel are listed in Table 10.

Conclusions
This work provides an overview of similar and dissimilar diffusion bonding and TLP bonding of 304 and 316 stainless steel by reporting on the main process parameters: bonding temperature, bonding time and bonding pressure as well as used interlayer(s), if any. Diffusion bonding and TLP bonding are commercially used processes and it is noteworthy that reported parameters for identical material combinations can vary considerably from one another. To compare different studies, we put much emphasis on quantitively reported results such as tensile-and shear strength rather than qualitative results that are much more difficult to compare. This may overemphasize the importance of mechanical testing versus qualitative testing. We are also aware that this is in many cases an apple and orange comparison, since pre-and post-treatments are not taken into consideration and additional processing parameters, such as the sample geometry, can influence the bond quality. Nevertheless, both apples and oranges have features in common so that there may be something to learn from these comparisons. Most importantly we want to use this review to provide an overview of the conducted experiments on the market to assist upcoming diffusion bonding and TLP bonding studies of 304 and 316 stainless steel.
We believe that future studies on diffusion bonding and TLP bonding should aim to report quantitative results such as tensile-and shear strength so that the overall bond quality can be quickly assessed. We also urge researchers to report the experimental set-up used to generate this data, and ask them to follow standardized testing whenever possible. Since the strength of seemingly similar materials also showed considerable differences in the reported data, we further urge researchers to conduct similar experiments with the base material(s) used. The quality of the generated diffusion bonds can also show significant differences and we recommend a minimum of three samples for each data point. We are aware that diffusion bonding apparatuses often do not allow the production of samples suitable for tensile testing since a minimum height needs to be realized. Finding reliable correlations that allow a comparison between tensile-and shear-tested diffusion bonded and TLP bonded samples was identified as a promising field of research.
Similar and dissimilar diffusion bonding and TLP bonding of 304 and 316 stainless steel is already used commercially and future experiments should aim to further optimize existing bonding parameters. Technical parameters (temperature, time, pressure, interlayers, etc.) should be optimized to accomplish the same or better bond qualities with fewer resources, resulting in more economically and ecologically sound processes. Here, particularly the reduction of expensive interlayers and complicated surface treatments should be omitted as far as possible. In this regard creative techniques such as applied temperature gradients as proposed by Shirzadi and impulse pressure diffusion bonding (IPDB) are promising fields of future diffusion bonding and TLP bonding research.