Review of the Impact of Hydrogen-Containing Fuels on Gas Turbine Hot-Section Materials

Decarbonization of the power-generation sector will require an “all-of-the-above” approach to ensure reliability, affordability, and stability of the electrical grid. With many countries and regions adopting 2030, 2040, and 2050 range goals, the reality of these targets hinges on optionality. Neither renewables, energy storage, carbon capture, nor low-carbon fuels singularly can achieve net-zero carbon emissions, especially in light of rapid and widespread electrification. Currently in the United States, approximately 20% of all final energy is electricity and this number is projected to double (∼43%) or triple (∼59%) by 2050 [1]. In addition, models around capacity needs for 2050 vary widely; however, some things are certain: increased capacity is necessary and current installation trends are not sufficient. It is predicted that wind and solar will need 160% to 480% growth in capacity, respectively, to help meet the demand while decarbonizing the energy sector. Compared with 2023 installation trends, year over year to 2050, the country will be 20–87% short of these expansion needs—note this does not include retirements over the next 26 years [2–4]. Thus, gas turbines will continue to be a vital part of the electricity generation portfolio, whether with carbon capture or utilizing low-carbon fuels.

One of the key components of a stable, decarbonized grid is the availability of clean dispatchable power. Today, gas turbine engines play a vital role in providing a range of grid services, including base-load power, frequency stability, and peaking. Going forward, their role in grid balancing will become more critical. California is a key example of the rapidly changing needs in demand response with high renewable penetration, as shown in Fig. 1. These curves show the demand for nonrenewable generation as a function of hour throughout a representative day for three different years: 2013, 2018, and 2023. The “duck curve” of 2018 has been replaced by the “canyon curve” of 2023, which shows the significant cycling necessary for nonrenewable power generation resources.

Moreover, as sunlight and wind availability are outside of human control, renewable energy cannot be the only means of generation. Likewise, the rate of additional energy storage capacity addition, including batteries and pumped hydro, is too slow to keep up with projected demand. Further, several regions require seasonal storage, which is not possible with modern battery technologies. Figure 2 is an example of a regional need for seasonal storage from the Electric Reliability Council of Texas (ERCOT), showing demand on peak summer and winter days. Note that ERCOT, as of the time of this publication, is the grid with the highest penetration of solar and wind capacity. From this chart, wind and solar provide a small part of the peak winter needs and only meet a portion of the summer demand. Hence, low-carbon, dispatchable generation is the key driver for meeting critical demand load levels.

The use of hydrogen-containing fuels, including hydrogen and ammonia, provides a low-carbon pathway for gas turbines to play these key roles in power generation. Figure 3 describes the carbon intensity of H₂ production based upon various production pathways. Some production pathways have very low carbon intensity, showing the potential role of these fuels for long-duration energy storage. For example, hydrogen could be produced using electrolysis when there is excess nuclear and/or renewable (wind or solar) energy available and then used for electricity production with gas turbines when demand is high.

Significant research is being done to understand the impact of these fuels on engine operability and emissions. Several reviews [7,8] discuss the implications for hydrogen and ammonia on combustion. In general, the higher flame speeds of hydrogen and lower flame speeds of ammonia can dramatically impact flame stabilization, which in turn alters operability limits and can change engine control procedures. As a first step, engines will likely use a blend of gases rather than neat fuels. For example, demonstrations of over 40% hydrogen blended into natural gas [9] have shown the potential for the use of hydrogen in the current fleet. Further, several researchers have proposed the use of blends of hydrogen and ammonia to overcome the aforementioned flame speed issues [10]. In all these cases, a wide spectrum of research, from fundamental chemistry to plant-level demonstrations, is considering the operability effects of these renewable fuels.

However, equally important for gas turbine operators is the implication of these fuels for material durability, lifing, and repair. Traditional life management consists of developing an understanding of the impacts of operation, design, fabrication practices, and metallurgy of materials on overall performance. In many cases, failures occur due to a compounding effect of two or more of these aspects. A change in fuel or fuel blending may have an impact on any of these features, motivating our need to understand the effect of fuel on engine materials. Life management practices are also often built around knowledge developed from field experiences and, to date, only a limited number of field studies have been conducted with hydrogen blending in industrial land-based gas turbines [9] and there are no field demonstrations of ammonia in these machines.

Currently, the most comprehensive review of the material issues associated with the introduction of hydrogen fuels in gas turbine engines was published by Stefan et al. [11]. This paper provides a literature survey on the implications of hydrogen on gas turbine material systems, including high-temperature alloys and coatings. However, certain key elements are not considered, including changes to radiative heat transfer and the impact of changes in operability on mechanical stresses. Further, the only fuel considered is hydrogen, whereas in this review we expand our consideration to hydrogen, ammonia, and blends of these fuels. A more recent review on both hydrogen and ammonia by Alnaeli et al. [12] focuses mostly on combustion phenomenon, without detailed consideration of the impact of fuel composition on materials systems.

This paper provides an overview of the potential materials issues that may arise from the use of hydrogen-containing fuels in power-generation gas turbines. The focus is on materials systems in engines that are in operation today where fuel blending or retrofits are being considered. Thus, next-generation, high-temperature materials like ceramic matrix composites are not considered. We begin by discussing the interaction mechanisms that arise from the use of these fuels, including changes to heat transfer, mechanical stresses, and material compatibility (oxidation, corrosion, etc.). We use first-order analysis of the potential interactions to understand the expected magnitude of each effect as the fuel composition changes. The final impact of these changes on current designs is unfortunately unknown, as the margin for material degradation that is incorporated into current engine designs is highly protected intellectual property of each turbine manufacturer. As such, we highlight the potential changes that may occur and leave it to the reader to infer the potential effects on a given design. For each mechanism, we provide a literature review of previous work on the impacts of these fuels on high-temperature metal alloys and thermal barrier coatings (TBC) typically found in power-generation gas turbines. Finally, we outline research and development topics necessary for the successful introduction of these fuels into the fleet.

In this section, we discuss the different ways in which fuels can impact heat transfer to hot section components. In this first-order analysis, we assume that the engines are running a typical Brayton cycle and we consider a baseline architecture to estimate the impact of fuel composition on heat transfer. The baseline is a GE 7 F frame engine with a combustor inlet condition of 408 °C (766.4 °F) and 1864.38 kPa (270.4 PSI) burning natural gas, which is modeled as methane. The 7 F was chosen given its prevalence in power generation in the United States [13].

Variations in fuel composition may affect convective heat transfer to hot-section components. Here, the heat capacity (c_p) and thermal conductivity (k) of the gases are sensitive to changes in the products of combustion. To estimate the magnitude of these changes, we calculate the equilibrium composition of different blends of hydrogen and natural gas as well as hydrogen and ammonia at a constant adiabatic flame temperature, 1526.85 °C (1800 K, 2780.33 °F); this condition is consistent with previous calculations by Breer et al. [14] focusing on the impact of fuel blending on emissions. At this condition, the equivalence ratio of the 100% CH₄ case is 0.508, of the 100% H₂ case is 0.424, and of the 100% NH₃ case is 0.542, representative of lean operation in a dry low-emissions combustion system [15]. Calculations are performed in Chemkin Pro [16] using GRIMech 3.0 [17] for mixtures of methane and hydrogen and the Mei 2019 mechanism [18] for mixtures of ammonia and hydrogen. For each fuel composition, the equivalence ratio in air is varied to maintain an adiabatic flame temperature within 2 °C of 1526.85 °C. From these compositions, the mixture-averaged c_p and k are calculated using ideal gas, linear blending rules at a temperature of 1526.85 °C. The heat capacity for each species was determined using the NASA curve fits [19] and the thermal conductivity was determined from the NIST Chemistry Webbook [20]. Where data were not available from the Webbook, linear extrapolation to the desired temperature was used.

Table 1 shows the results of the equilibrium calculations for H₂/CH₄ and H₂/NH₃ mixtures. Note that in both cases, only the major product species are provided. In these calculations, the mole fraction of N₂ is increased to account for all other minor species, ensuring that the sum of the mole fractions is unity. There are several interesting results from these calculations. First, the high heat capacity species, CO₂ and H₂O, vary significantly with fuel composition. In the case of H₂/CH₄ blends, CO₂ mole fraction goes to zero at 100% hydrogen, whereas the mole fraction of water increases from ∼6.5% to ∼11% as the fuel composition is varied from 100% CH₄ to 100% H₂. While no CO₂ is produced in any of the H₂/NH₃ flames, the trend in H₂O is opposite with H₂ content in these flames. The product mole fraction of H₂O decreases from ∼13% to ∼11% as fuel composition is varied from 100% NH₃ to 100% H₂. In general, the heat capacity and thermal conductivity trend with the mole fraction of H₂O in the products. The variation in these quantities is of the order of a few percent.

A canonical example is used to determine the change in heat transfer coefficient with fluid properties assuming that the flow velocity is held constant. The purpose of this analysis is not to provide realistic values for heat transfer coefficient in a combustor or turbine, but instead isolate the sensitivity of fundamental convective processes to gas composition. In these calculations, the heat transfer coefficient, h, is calculated for a laminar flat plate boundary layer and then normalized by h_baseline, which is at the 100% methane condition. Mixture-averaged densities and viscosities were used to calculate the mixture-averaged Prandtl number for all cases; density and viscosity values for each specie were obtained from the NIST Chemistry Webbook and linear extrapolation was used if the property values were not available at the desired temperature.

Figure 4 shows the variation in the ratio of the heat transfer coefficient to the baseline for mixtures of H₂/CH₄ and H₂/NH₃ as the fuel mole fraction of H₂ varies from 0 to 1. The mixture with the highest heat transfer coefficient is the 100% ammonia case, which is almost 4% higher than that of the baseline 100% methane case. Assuming a baseline wall temperature of 926.85 °C (1200 K, 1700.33 °F), this increase in the heat transfer coefficient could produce a 17 °C (62.6 °F) increase in wall temperature, resulting in the need for more cooling, material or design changes, or reduction of current part life if this increase falls outside the margin currently designed into the parts.

Few studies have considered the impact of fuel composition on convective heat transfer in hot section components. An early study by Lefebvre [21] indicated that convective heat transfer to the liner was difficult to measure and model; he provided a relatively simplistic model for heat transfer in straight pipes to capture the effects of mass flow rate, gas properties, and gas temperature on the wall temperature. More recently, a program at the National Energy Technologies Laboratory (NETL) studied next-generation materials and cooling systems for syngas-based, integrated gasification combined-cycle (IGCC) systems [22–26]. Syngas fuels can have up to 40% hydrogen in the fuel, with the remaining species including CO, CO₂, CH₄, and H₂O [27], making the results relevant to the high-hydrogen fuel blends under consideration here. These studies considered the impact of gas composition on vane heat transfer using CFD, including the use of a P-1 radiation model. The mixtures considered both oxy-fuel combustion, where the product gases have significant mole fractions of H₂O and CO₂, as well as a regular air-based cycle with syngas fuels. The results show that the mixtures with high levels of H₂O and CO₂ have significantly higher heat transfer coefficients, as can be seen in Fig. 5, which shows heat transfer coefficient for a turbine blade. A first-principles analysis, similar to the one done in this paper, shows very good agreement with the high-fidelity simulation, indicating that the reason for the increase in heat transfer coefficient is the change in heat capacity and thermal conductivity in the product gases.

Work by Chiesa et al. [28] used a correlation from Louis [29] to estimate the impact of the change in product composition on vane surface heat transfer, yielding similar conclusions to those in this work. A more fundamental study by Zou and Yuan [30] considered the change in gas composition on flat plate heat transfer. This experimental and computational study used a flat plate with a row of film cooling holes at the entrance of the channel, varying the water mole faction in air, static pressure, and static temperature; a baseline case showed very good agreement between the experiment and the simulation. The simulation used a gray gas model for the gas-phase radiation contribution. At all pressures and temperatures, increasing the mole fraction of H₂O in the mainstream increased heat flux to the wall. Correlations for the dependence of the non-dimensional heat transfer on all three parameters were developed. Finally, some studies mention changes to the heat transfer with variations in fuel composition or product composition but did not provide any further investigation or discussion, including work by Gazzani et al. [31] and Goke et al. [32].

Radiative heat transfer can account for a non-negligible amount of the total heat flux to the surface of gas turbine hot-section components. Work by Lefebvre showed the importance of radiation on combustor liner temperature in a study that considered the impact of fuel composition on gas turbine combustor liner temperature, pattern factor, and emissions [21]. This study used simplified scalings for radiation to the combustor liner to fit data from several engines, including a F101, a T56, and a J79. In general, liner temperature decreases with increasing hydrogen content of the hydrocarbon fuels, a result of the lower soot formation that occurs with less carbon in the fuel. As such, the authors point to flame radiation as one of the major sources of heat transfer to the wall.

In dry low NO_x (DLN) gaseous-fueled combustors, however, there is no soot formation, so the source of radiation to the wall is from the gases. A study by Johnson and Zhao [33] considered the radiative heat transfer to a TBC-coated wall from a methane/air flame operating at 40 atm. The participating species in the radiation calculation were H₂O, CO₂, and CO; the results showed that a line-by-line calculation of the gas-phase radiation produced more realistic wall heating than using a gray-body assumption for the gases. In particular, the wall heat flux from gas-phase radiation was significant (of the order of 10⁵ W/m²); this paper also discussed the particular impact on TBCs, which will be discussed in more detail later. This work, among others [34–36], shows that gas-phase radiation should not be neglected in considering heat transfer in gas turbine hot sections.

We used a spectral line survey to estimate the variation in radiative forcing that could result from changes to the composition of the combustion products. This analysis is a first-order approximation that does not take into account any absorption or scattering of light, although it is expected that scattering will be minimal at these conditions. Further, it does not take into account any radiation from solid surfaces or the shape factor of the combustor. It simply compares the total radiative intensity of one mixture versus another over a given range of infrared wavelengths to understand how radiative heat transfer could potentially change with mixture composition. In the mixtures and temperature under consideration, the three species that contribute to radiative heat transfer are CO₂, CO, and H₂O, as was done by Johnson and Zhao [33].

The line survey was calculated using the online tool Spectraplot [37] between the wavelengths of 2 and 8 microns. Spectraplot's line survey tool provides information about spectroscopic transitions over a large range of wavelengths; other tools on the website can provide much more detailed absorption and emission profiles over smaller wavelength ranges. However, for the purpose of this exercise, the line survey tool was appropriate for estimating bulk radiative intensity. The wavelength range was chosen based on potential damage mechanisms present for typical thermal barrier coatings. Work by Eldridge et al. [38] showed that the topcoat of yttria-stabilized zirconia TBCs (8YSZ) is up to 50% transmissive between the wavelengths of 2 and 8 microns, which can lead to significant heating of the bond coat and substrate below. While the transmission level varies with composition (13.5YSZ can be up to 90% transmissive), the wavelength range over which IR radiation passes through the topcoat is relatively similar for most TBC compositions. As such, this analysis will apply to a wide range of coating chemistries.

Figure 6 shows the output from Spectraplot of the baseline spectrum for the products of the 100% CH₄ flame at 1526.85 °C (2780.33 °F) and 1864.38 kPa (270.4 PSI), with H₂O in green and CO₂ in yellow; the levels of emission of CO are orders of magnitude smaller and so not plotted here. H₂O has wide bands of radiation over most of the wavelength range, whereas CO₂ has more concentrated regions of radiation with higher intensity. The strength of each of the bands is dependent on the mole fraction of each species as well as the pressure and temperature. In these calculations, the pressure and temperature are held constant and the mole fraction of the constituent species varies between cases.

For each mixture, the line strength of each of the lines was summed and plotted as a function of hydrogen mole fraction in the reactants, as shown in Fig. 7. The units of the summed line strength are arbitrary. The results show that across the board, flames with CO₂ in the products have much higher radiative power than those with just H₂O. Although it is not obvious from the spectra in Fig. 6, the density of CO₂ lines in the two bands around 2.8 microns and 4 microns is much higher than the density of H₂O lines across the entire range of wavelengths. As a result, the radiative power of CO₂ is higher than that of H₂O, making the gas-phase radiation of any H₂/NH₃ flame at a given temperature less than that of a flame that has any amount of CH₄. Thus, the radiative power of the H₂/CH₄ flames is largely controlled by CO₂, decreasing by a factor of three as the fuel mixture varies from 100% CH₄ to 100% H₂. The variation in the radiative power for the H₂/NH₃ flames is much smaller, decreasing slightly as the H₂O content in the mixture decreases.

The literature on radiative heat transfer in gas turbine hot sections is limited, making the literature regarding fuel composition effects even more limited. A series of studies from Tsinghua University [34,39–42] used both experiments and simulation to understand the role of gas composition on heat transfer, with explicit consideration of the role of radiation from CO₂ and H₂O. In these studies, a plate with a line of film cooling holes was tested and simulated at a range of gas compositions that mimic a variety of low-carbon fuel solutions, including natural gas, syngas, hydrogen, and oxy-fuel combustion of natural gas. The simulations include a discrete transfer model with a weighted sum of gray gases model to capture the spectral behavior of the gases. In all cases, the addition of the radiation to the heat transfer calculations significantly altered the temperature distribution on the wall and matched the experimental results more closely. Further, mixtures with higher levels of CO₂ and H₂O resulted in higher temperatures on the wall, a result of the increased heat transfer from radiation.

A number of studies in open flames have considered the level of flame radiation with different fuels, including hydrogen and ammonia [43–48]. There is an emphasis on the radiation heat transfer in ammonia/air flames because this mode of heat transfer is known to significantly impact the flame speed much more than in methane or hydrogen flames. Simulations by Nakamura and Shindo [46] showed that the flame speed of an ammonia/air flame could reduce by 3% at stoichiometric conditions when radiation is taken into account and the reduction is greater at both higher and lower equivalence ratios. This heat loss suppresses the thermal decomposition of NH₃ to H₂ in the flame, slowing the flame speed as a result.

Another set of studies consider the impact of wall heat transfer on flame behaviors, particularly for ammonia [45], although few of these consider gas turbine configurations specifically. For example, Cai and Zhao [49] explored the impact of wall thermal conductivity on the performance of a micropower system, which consists of a thin channel for reactions of hydrogen/ammonia and oxygen. Greater heat loss through the wall significantly impacted the system temperature and flame location in the system. In another study, Lam et al. [50] experimentally tested blends of natural gas and hydrogen at a range of fuel-splits in the piloted injector of a Siemens SGT-400 combustion system. The results showed lower injector wall temperature with increasing H₂ at a constant flame temperature; the reason given was the reduction in flame radiation with higher H₂O contents in the product gases.

Some studies have pointed to the role of radiation without any detailed consideration. For example, a review by Takeishi [51] indicated that the use of hydrogen or ammonia in place of natural gas would increase the amount of H₂O in the exhaust stream and hence the radiative heat transfer from H₂O, but does not discuss the relative strength of radiation from CO₂ versus H₂O. The results shown here indicate that the total radiative output from high-H₂O gases would be less than that with CO₂ at a constant flame temperature.

Put together, these heat transfer scalings show a mixed result as a function of fuel composition. Many of the effects are driven by the presence of H₂O in the combustion products. There are significant differences in the equilibrium mole fraction of H₂O for these fuel blends: 6.5% for stoichiometric CH₄ flames, 11% for stoichiometric H₂ flames, and 13% for stoichiometric NH₃ flames. When H₂ is blended into CH₄, the convective heat transfer has the potential to increase as more H₂O drives increases in heat capacity and thermal conductivity, but the radiative heat transfer can decrease significantly. When H₂ is blended into ammonia, both the convective and radiative heat transfer decrease with increasing H₂ as a result of decreasing H₂O in the products, but the effects are less dramatic than in blends of H₂ and CH₄.

The variation in product composition with fuel blend, even at a constant firing temperature, has interesting implications for operation of gas turbines. Figure 8 shows the power output of a F-class engine with different fuel blends at a constant firing temperature. Burning with H₂ has a higher power output because it has a higher energy per unit mass. Burning with ammonia has a higher power output because of the significantly higher fuel mass flow required to match the flame temperatures. As such, some of the heat transfer issues that arise with high-hydrogen fuels could be abated at a constant power output if the firing temperature was decreased. In this case, both convective and radiative heat transfer would decrease while maintaining the same load. For reasons like these, new fuels should be considered on a systematic level.

Because a change in the fuel composition will have an impact on both flame stabilization and heat transfer to the wall, these modifications will alter the temperature distribution in the hot section, resulting in changes to the mechanical stresses. The two main failure mechanisms of concern for these parts are creep and low-cycle fatigue [52]. Creep is a time-dependent plastic deformation process that occurs at elevated stresses and high temperatures, typically >50% of the melting temperature for gas turbine alloys. Low-cycle fatigue is driven by low-frequency load changes that cause oscillations in mechanical load. In gas turbines, the source of low-cycle fatigue is typically the cycling of the engine to vary load or the frequent starts and stops that might be experienced by a peaker machine.

Changes to the flame shape and heat transfer to the combustion liner could impact the mechanical stresses in this component. Several studies have shown the changes in flame shape with variations in fuel composition, which would drive the modification of liner temperature distribution. In swirl-stabilized flames, blends of hydrogen in natural gas show significant shortening of the flame [53]. For example, a study by Strollo et al. [54] blended hydrogen into natural gas at constant thermal power in an industry-relevant swirling fuel injector at atmospheric pressure. At each condition, the flame center of heat release was calculated from CH* chemiluminescence imaging, as shown in Fig. 9 (left). Variations in the H₂ content in NG from 10% to 20% by volume could result in an 18% decrease in flame height and a 5% decrease in flame width, concentrating the high-temperature region on the combustion liner.

Blending of ammonia into natural gas has the opposite effect as hydrogen, leading to slower flame speeds and typically longer flame structures [44]. For example, work by Elbaz et al. [55] in a swirl-stabilized flame with various blends of NH₃ and CH₄ in air at atmospheric pressure showed changes in the flame structure with fuel composition. Flame temperature distributions were measured for different conditions, including pure NH₃ and blends of 50% NH₃/50% CH₄ at a constant equivalence ratio. Figure 9 (right) shows the change in temperature distribution, where the 100% NH₃ flame shows slightly lower temperatures overall (as compared to 50% NH₃), including a 50 K decrease in peak temperatures in the central region of the flame.

Changes to the structure and temperature of the flame could also impact the pattern factor, or temperature distribution at the exit of the combustor and entrance of the turbine first vane. Studies by Lefebvre [21] showed this trend for jet fuels, where relative changes in the fuel properties were significantly less severe than they will be for different blends of hydrogen, ammonia, and natural gas. For example, variations between liquid fuels (JP4, JP8, and DF2) could drive a 20% change in the pattern factor, defined as $PF=(Tmax−T4)/(T4−T3)$⁠, where T_max is the maximum temperature in the profile, and T₃ and T₄ are the temperatures at the combustor inlet and exit, respectively.

Several studies have identified changes to pattern factor with variations in fuel composition for high-hydrogen fuels, particularly for microgas turbines [56–61]. Most of these studies are computational, as few annular or can-annular burner rigs exist. Many of these studies also report the change in temperature distribution on the combustor liner, which shows similar trends to the experimental studies discussed previously. One of the most comprehensive studies was recently published by Park et al. [62], who considered the impact of hydrogen and ammonia blending in natural gas on the combustor liner temperature distribution, pattern factor, emissions, and combustion oscillations of a flame using a 501 F can combustor operated at atmospheric conditions. Up to 30% H₂ and 20% NH₃ were blended into NG at a constant thermal load. The injector, liner, and transition piece were instrumented with thermocouples to determine the impact of fuel composition on wall temperature profile. Figure 10 shows the results of the hydrogen blending (top) and ammonia blending (bottom) tests. Significant variations in temperature were measured with both tests, resulting in almost 60 °C local temperature increases with H₂ and almost 50 °C local temperature decreases with NH₃.

Similarly, Fig. 11 shows the change in circumferential pattern factor, defined the same as in Lefebvre [21], with variations in fuel composition at 48% load. Companion CFD analysis and flame imaging in an optically-accessible rig showed that the flame shape changes, driven by changes in flame speed, were the root cause of these temperature variations. More compact flames with H₂ and more diffuse flames with NH₃ drove significant changes to liner temperatures and pattern factor.

Combustor liners are typically made of high-temperature polycrystalline wrought alloys like Hasteloy or Inconel, coated with an yttria-stabilized zirconia thermal barrier coating. Turbine vanes and blades are made from similar materials but are typically cast, often as directionally solidified or single-crystal structures, which are then coated in TBC for early-stage blades and vanes. Changes in temperature distribution of the magnitudes shown in the literature could have significant implications for the stress states on the parts [63]. If the overall part temperature of early-stage blades is increased by ∼50 °C, then the typical creep life of these components would be reduced by at least one order of magnitude. The simplest way to represent this effect is through the Larson-Miller parameter [64], which is a method for estimating the lifetime of a component based on changes in either operational time or temperature. The Larson-Miller parameter is represented by: $LMP=T ×[log(tR)+C]$⁠, where T is the temperature (in K), $tR$ is the rupture time (in hours), and C is a constant (typically ∼20 for most alloys).

Further, these changes in temperature distribution could have a significant impact on the combustor liner and turbine vane cooling schemes. In the combustor, a combination of liner cooling strategies, including back-side cooling, impingement cooling, and effusion cooling, are designed to accommodate the high levels of heat transfer to the liner in the region of the flame [65]. In the turbine, both internal cooling passages and external cooling features, including showerhead and film cooling holes, are designed to accommodate a particular range of pattern factors [66]. Given the changes in flame shape and temperature distributions with variations in fuel composition, the efficacy of these cooling strategies, and hence part life, could significantly deteriorate.

Finally, the use of hydrogen-containing fuel also introduces the potential for higher mechanical load on the turbine components. These changes can be realized in two different scenarios. The first scenario involves the variation of fuel composition from 100% NG to 100% H₂ with a constant turbine inlet temperature. Even though the inlet temperature is the same, the power generated by a blade row will increase proportionally with the increase in specific heat of the product gases with higher levels of H₂O. For example, the c_p of a 100% H₂ flame products as compared to a 100% NG at T_ad = 1800 K, as in Table 1, is 2.6% higher. Using specifications for a typical F-class engine (mass flow of 430 kg/s, combustor inlet temperature of 408 °C), the shift from CH₄ to H₂ fuel would result in a ∼3.8 MW increase in power for that stage, or a ∼30 kW increase in power per blade (assuming 90 blades). In addition to increased stresses, the actual stress profile and high stress locations may change because of the product composition. This variation could potentially lead to a need for design changes in early-stage blades to better redistribute stress or further considerations of changes in operating loads to account for the increased stress.

In the second scenario, hydrogen is used to increase the firing temperature of the machine, resulting in a turbine inlet temperature that could be as much as 200 °C (392 °F) higher than the baseline methane case. This increase in temperature also increases the specific heat, although the change in c_p with temperature is less than the change due to the composition. The increasing power per stage would depend on the number of stages and the distribution of loading between them, making an estimate of the increased stresses difficult. Regardless of the details, the loading would increase, resulting in higher stresses and reduced part lifetime.

Several material issues arise throughout the gas turbine hot section with the use of any fuel, but several known problems could become worse and new ones could arise in the presence of high levels of hydrogen-containing fuels. In this discussion, we focus on two different materials systems: high-temperature metal alloys used in the fuel injector, combustor liner, and turbine, and thermal barrier coatings. The fuel injector is discussed separately from the high-temperature alloys used for the combustor liners and turbines because of the significantly different thermochemical conditions experienced by this component. The material compatibility issues in the fuel delivery systems are significant but not discussed here; however, a recent paper by Pigon et al. [67] reviews these issues.

Fuel injectors see a variety of gas compositions and temperatures across their different parts. Compressor discharge air is delivered at moderately high temperatures through the main flow passage, whereas high-pressure fuel (sometimes heated) is fed through small inner passages that exit into the main air stream. Finally, high-temperature combustion gases interact with the face of the injector at the flame stabilization point, particularly on features like centerbodies. The particular focus of this section will be the inner fuel passages, as many of the issues experienced by the combustion-facing parts will be covered in the next two sections, depending on whether the injector face is coated or not. Typical fuel system conditions consist of fuel pressures between ∼2070 kPa (300 psi) on older B- and E-class machines to ∼3450 kPa (500 psi) on newer H-class machines, and fuel temperatures that range from ambient to ∼175 °C (350 °F). At these conditions, two main material degradation mechanisms exist for hydrogen-containing fuels: hydrogen embrittlement and nitridation.

Hydrogen embrittlement is a process by which hydrogen diffuses into the alloy and, through several possible mechanisms, makes the material more brittle [68]. It typically occurs in high-pressure, low-temperature conditions, although the temperature ranges for embrittlement of Ni-based superalloys is typically higher than for stainless steels [69]. The high pressures drive hydrogen diffusion into the alloy, allowing for internal embrittlement to occur. Studies have shown that embrittlement occurs in almost all Ni-based superalloys at room temperature and pressures above 34 MPa (4931 psi) [69], although the rates of embrittlement decrease as the pressure decreases due to slower rates of H-atom diffusion.

Figure 12 shows the impact of embrittlement as a function of pressure for six Ni-based superalloys used in aerospace and power-generation applications. Here, the hydrogen environment embrittlement (HEE) index, or the ratio of the notched tensile strength (NTS) of the specimen in hydrogen to the NTS of that same specimen in air or helium, is plotted versus hydrogen pressure at 22 °C. For most alloys, a precipitous drop in HEE occurs as pressure first increases from atmospheric before leveling off at higher pressures. The initial drop is driven by faster diffusion of hydrogen into the alloy, whereas the plateau is a result of hydrogen saturation in the material. The impact of temperature on embrittlement is a strong function of the alloy. Figure 13 shows the impact of temperature on the HEE based on area reduction ratio, which is a measure of tensile ductility for hydrogen versus helium or air environments, for three Ni-based superalloys at 51.7 MPa for Inconel 718 and 45 MPa for the EP alloys. Inconel 718, a common material for fuel injectors, is highly impacted at low temperatures, but not as susceptible to embrittlement as temperature increases. However, the minimum HEE varies with alloy, where the EP 99 and EP 741 are two alloys with relatively high-temperature embrittlement behaviors. However, these alloys are typically not used in fuel injectors.

Based on these results and the typical line pressures and temperatures of fuel in gas turbine engines, we would expect some reduction in fracture properties for 100% hydrogen service, depending on material. With fuel pressures of up to 4 MPa at atmospheric temperature, the HEE (Fig. 12) would be near 1 for Astroloy, Hastelloy X, and Haynes 188, 0.8 for Inconel 718, and near 0.5 for Rene 41. The temperature range of the fuel (20–175 °C) is within the range where embrittlement may be relevant, and so these results show that the choice of alloy for the fuel injector is quite important. One way to potentially avoid embrittlement issues in high-hydrogen fuels would be to heat the fuel. As shown in Fig. 13, HEE for Inconel 718 increases with temperature above 100 °C, where even moderate increases in fuel temperature could meaningfully reduce embrittlement. Typically, fuel heating also results in higher cycle efficiencies and so this strategy could have multiple benefits.

Nitridation is a process by which atomic nitrogen dissolves into the alloy and forms nitrides [70–72]. Several components of Ni-based superalloys have a propensity to form nitrides, including aluminum, titanium, and tantalum. That said, nickel-based alloys are typically more nitridation resistant than iron-based alloys. The rate at which nitridation occurs is proportional to the concentration of molecular nitrogen. In air atmospheres, molecular nitrogen is present in trace levels except at very high temperatures. In gaseous mixtures with ammonia, however, significantly more atomic nitrogen can be present at a wider range of temperatures, which increases the potential for nitridation. Further, the diffusivity of nitrogen into alloys increases significantly with temperature, which can both increase the rate as well as the penetration depth of nitridation [70]. Despite the moderately high pressures and potentially high levels of ammonia in fuel injectors, there is likely a low rate of nitridation at the temperatures present in the fuel-side surfaces of gas turbine fuel injectors. Most “low-temperature” nitridation studies in superalloys consider temperatures of ∼400 °C (752 °F), which is far above the temperature in even a heated fuel system.

Nickel-base superalloys are commonly used in gas turbine hot sections due to their exceptional high-temperature mechanical properties. Typical alloys include trade names such as Rene, Inconel, Hasteloy, PWA, and CMSX [52]. In the combustor and first turbine stages, these materials are typically coated with thermal barrier coatings, which protects them from direct exposure to the combustion gases. However, further downstream in the turbine, several uncoated stages still interact with combustion products at relatively high temperatures. This discussion focuses on these conditions.

The greatest potential hazard for these components is enhanced oxidation in the presence of higher water concentration in the combustion products of high-hydrogen fuels. Surface oxidation of nickel-base alloys is expected in post-combustion atmospheres in gas turbines. Such alloys, especially those with higher aluminum or chromium content, form a protective oxide scale on the surface of the part that can slow diffusion of gases into the metal and reduce the rate of internal corrosion [73]. The effect of higher H₂O concentrations on alloy oxidation behavior is dependent on the oxide chemistry of an alloy [74]. Most high-temperature nickel alloys typically form alumina or chromia scales and data show differing impacts of H₂O on oxidation rate, scale thickness, and scale adhesion depending on the oxide chemistry. Several studies have considered the impact of water-containing environments on oxidation of Ni-base superalloys [75–84], although the literature is not as extensive or complete as is needed to understand the impact of hydrogen fuels at gas turbine conditions. In particular, no fundamental studies of material oxidation have been identified that were conducted in atmospheres relevant to the products of high-hydrogen combustion; most studies consider different levels of water vapor in air.

A brief review by Pint et al. [85] highlighted several initial studies of a range of Ni-base superalloys and stainless steels in various H₂O-containing atmospheres. In general, alumina-forming alloys are less affected by the presence of H₂O as compared to chromia-forming alloys. For example, work by Pillai and Pint [86] examined the difference in mass change for alloys 740H, 282, X4, and 1483 during cyclic testing (100 h cycles for 500 h at 800 °C) in atmospheres of air with 10% and 60% H₂O. Alloy 282 experienced some increase in mass loss at the higher H₂O level, while X4 had a significantly lower mass loss at the higher H₂O level. For each material, two specimens were tested and there were significant differences between the behaviors of the two samples over the course of the testing. High-resolution imaging suggested that the structure and chemistry of the scale formation can have a significant impact on the internal oxidation rates.

Further, the impact that trace elements in the alloy have on scale adhesion can change the oxidation behavior as the presence of a stable scale can significantly reduce the incidence of internal oxidation [87]. Oxidation rates can increase with cyclic behaviors, which can also drive the cracking and spallation of the protective scales, leading to part damage. In addition to material loss, the potential for cyclic spallation behavior poses the risk of foreign object damage (FOD) in the hot section as scales are shed during cooldown.

Additionally, there can be a strong coupling between oxidation and the presence of material stresses [88]. Several studies [89–93] have shown more rapid oxidation rates in the presence of both tensile and compressive strain in superalloy materials. For example, work by Barnard et al. [89] considered the impact of stress on the isothermal and cyclic oxidation rate of two Haynes superalloys: alloy 75 and alloy 230. Figure 14 shows a summary of the results of the study after 1000 h of testing for both isothermal oxidation (iso) and cyclic (cyc) testing at two temperatures, where the cycles were performed at 100-hour intervals. Overall, the dog-bone samples under stress had a lower mass change than the unloaded samples. However, the rate of oxide layer growth was faster under stress, a result of the faster diffusion of Cr atoms to the surface to form oxide scales.

Inversely, the presence of oxidation can change the fracture behavior of a superalloy [94–97]. For example, fatigue experiments from Bache et al. [94] showed that fatigue failure occurred earlier for samples in air than in vacuum in a sample that was machined from a RR1000 turbine rotor. Inspection of the notched samples indicated that oxide formation in the notches initiated crack growth, resulting in poorer fatigue performance. Given the enhanced oxidation due to higher H₂O content as well as potentially higher mechanical stresses due to increased specific heats, changes to the fuel composition could have measurable impacts on both corrosion and mechanical failure modes.

The incidence of nitridation and hydrogen embrittlement in the combustor and turbine will likely not be any different with high-hydrogen fuels than they are for typical fuels. The rates of these mechanisms are all highly dependent on both the partial pressure of the attacking species (atomic nitrogen, atomic hydrogen, etc.) as well as the exposure time of the metal to these species. Assuming reasonable combustion efficiencies (>95%) and rarity of events like blowout where significant amounts of unburned fuel could propagate downstream past any coated parts, then the rate of these damage mechanisms should be negligibly small as compared to oxidation.

Despite the unfavorable conditions for nitridation, there are some initial studies on nitridation in ammonia flame environments. Work by Wang et al. [98] considered the nitriding effects of ammonia flames on iron-based metal walls, but the materials systems are more relevant for industrial systems rather than gas turbines. Recent work by Laws et al. [99] described an effort to design an experimental facility for understanding nitridation processes in combustion-relevant environments with high levels of ammonia.

Thermal barrier coatings are ultralow conductivity, multilayered ceramic coatings that reduce the heat transferred from the hot gas to the metallic part. The typical structure of a TBC includes a bond coat, typically made of mullite or a nickel- or cobalt-modified aluminide, that is applied to allow the topcoat, typically made of yittria-stabilized zirconia (YSZ), to better adhere to the substrate. The bond coat also provides corrosion and oxidation protection for the substrate. Between these two layers is a thermally grown oxide (TGO) layer that develops as the coating is subjected to high service temperatures. These coatings can be applied through various methods, including electron beam physical vapor deposition or air plasma spray processes. Regardless of the application method, both the material and structure are designed to minimize the thermal conductivity of the coating. Typically, the structure of the topcoat is optimized for both heat transfer and the difference in thermal expansion coefficient between the substrate and the TBC, allowing it to flex as the parts heat and cool.

The main damage mechanism to TBCs is failure of the thermally grown oxide layer between the topcoat and the bond coat [100]. The growth of this layer is driven by the diffusion of oxygen through the topcoat and the oxidation rate of the aluminum in the bond coat. As the layer forms, significant residual stresses can arise as a result of the difference in the thermal expansion coefficient between the substrate, the bond coat, and the topcoat; these stresses tend to increase with the thickness of the TGO [101]. These stresses can cause a number of potential failure mechanisms, all of which can result in cracking of the TGO and then cracking and spallation of the topcoat [102–104]. Stresses are exacerbated in parts with high levels of curvature, like turbine vane leading edges, leading to earlier failure in these critical locations [105]. Once the topcoat is cracked or spalled, significant oxidation of the bond coat and damage to the metal substrate are possible. The addition of Pt-modified coatings has been shown to extend the lifetime of the TBC [103], but are typically only used in aircraft engines, not ground-based turbines, and so are not discussed in detail here.

Thermally grown oxide growth and failure is a known durability issue in hot-section coatings; the coating is typically applied with the knowledge that it will be stripped and reapplied during regular service intervals [106]. However, the use of high-hydrogen fuels can accelerate the TGO growth rate due to the increase in H₂O concentration [104,107–110]. Fundamental studies of oxide formation rates identified two mechanisms by which H₂O accelerates oxide growth and damage. First, dissociation of H₂O to form oxygen radicals has a lower energy barrier than dissociation of O₂, resulting in more rapid oxidation [74]. Second, the presence of hydrogen and hydroxyl radicals can accelerate the oxidation kinetics through several pathways [74,111,112], particularly for aluminum oxides. Further, oxides formed in the presence of water have a higher volatility than those formed in oxygen atmospheres [74], resulting in lower damage thresholds in harsh hot section environments.

Figure 15 shows work by Zhou et al. [109], which plots weight gain as a function of time for a plasma-sprayed TBC with a NiCrAlY bond coat and a 7.5YSZ topcoat. The samples were exposed to two different atmospheres—O₂ and O₂ with 5% H₂O—at isothermal conditions (1050 °C) for up to 400 h. While the initial TGO growth rate is similar for the two atmospheres, the oxidation continues in the presence of H₂O, whereas it levels off after 100 h in the O₂ atmosphere.

Further, the presence of water can result in non-ideal TGO structure and the formation of spinels, which are mechanically weak layers of Cr, Co, and Ni oxides that grow between the topcoat and the alumina TGO [113]. Spinels typically form during the initial transient TGO growth [114,115] and can continue to grow at longer times [116–118]. Sullivan and Mumm [107,108] used isothermal oxidation experiments on several bond coats without their topcoats to better understand spinel growth. Figure 16 shows cross-sectional micrographs of the bond coat, alumina TGO, and spinel formation at four different levels of H₂O in the ambient after 300 h of exposure at 1125 °C. The thickness of the spinel layer increases significantly with H₂O concentration in the air. Further, it is evident that the spinel layer is much less uniform than the TGO layer, which can promote the wrinkled-interface failure mechanism of TBCs [100].

The composition of the substrate alloy may also play a role in the impact of H₂O on TGO growth rate. A study by Haynes et al. [119] compared the failure of a high-velocity oxy-fuel (HVOF) sprayed NiCoCrAlYHfSi bond coat and an air plasma sprayed 8YSZ topcoat, typical of power-generation gas turbine coatings, with two nickel superalloy substrates: superalloy 1483 and X4. The 1483 had a significantly shorter lifetime than the X4 in an environment of 10% H₂O in O₂, failing after 267 one-hour cycles at 1100 °C as compared to 380 cycles for the X4. Scanning electron microscopy (SEM) showed no significant difference in the thickness or structure of the TGO in both cases but X-ray elemental mapping using energy dispersive spectroscopy (EDS) identified significant differences in the composition of the oxide layers. The TGO on alloy 1483 showed lower levels of Al and higher levels of Ti and Cr, both of which have been shown to reduce the interfacial strength of the TGO [107,111,118]. These elemental differences could be connected back to the alloy itself, such that X4 is comprised of 7.5% Cr, 13.3% Al, and 1.2% Ti, and 1483 is comprised of 13.6% Cr, 7.3% Al, and 4.9% Ti. The lack of Al, which forms stronger TGOs, and abundance of metals that form weaker TGOs led to earlier TBC failure.

Finally, changes to the radiative emission from the gases could change the damage rates in TBCs. Our previous analysis showed that reduction of CO₂ levels in the product gases could reduce the radiative intensity of the gases, potentially resulting in less radiative heat transfer to the wall. Given the transmissivity of TBC topcoats in the midwave infrared spectrum [38] and the reduction in intensity in these same ranges, the radiative heat transfer to the TBC bond coats would likely decrease with the use of high-hydrogen fuels as compared to natural gas. The combination of these effects—increased oxidation damage and reduced radiative heat transfer—due to the presence of H₂O need to be considered together. The relative contribution of these effects is an important topic of future research.

However, one drawback of many of previous studies is that the effect of water addition is tested in air or oxygen, rather than in realistic combustion product compositions where the level of O₂ in the atmosphere would be substantially less. Lance et al. [120] tested the impact of replacing air with CO₂ on TBC lifetimes in mixtures with 10% H₂O and showed that CO₂ has little influence on the TBC lifetime with 100-hour cycles. However, further testing in realistic environments is necessary to better identify the effects of H₂O concentration in product gases on TBC behaviors and lifetimes.

This initial analysis shows that there could be significant implications for the high-temperature material systems found in gas turbine hot sections with the introduction of hydrogen and ammonia fuels. The impact of these fuels on existing hardware must be inter-rogated by each manufacturer given the unique operating and lifing margin that is inherent to each design. Incorporating these lessons learned and future research in this area into industrial design and analysis tools will be an important step toward integration of these new fuel blends into operation in existing machines. Further, use of these fuels will have implications for other parts of the power-generation system as well, including fuel delivery hardware and the heat recovery steam generators downstream of the engine [121].

Several lessons learned should be considered going forward. First, heat transfer from the gases to the walls can change significantly because of the change in product composition. The higher proportion of H₂O in the products of both ammonia and hydrogen combustion as compared to natural gas has two opposing effects. Convective heat transfer can be augmented as a result of the increase in heat capacity and thermal conductivity of the product gases. However, reduction of CO₂ in the product gases reduces the radiative emission at a given temperature. As such, more detailed research on the relative impact of these two modes of heat transfer on combustor liner temperatures, durability, and cooling strategies in realistic, reacting environments is critical to supporting the transition to high-hydrogen fuels.

These same changes to the product gas composition and temperature distributions will have significant implications for the mechanical stresses on hot-section components, particularly rotating components. Further research is necessary to understand the impact on part life and performance. For example, cycle-level analysis can help identify the ways in which these gas property changes can be leveraged to improve efficiency and power output. The results of these analyses can feed into better modeling of mechanical stresses in the rotating components. Further, better understanding of how fuel composition will alter gas temperatures and temperature distributions will provide input to research on improved cooling schemes and creep life prediction. Better component- and system-level research and integration is necessary to take full advantage of the benefits that can arise from the implementation of new fuels.

Material compatibility issues may be significant and a better understanding of material behaviors and processes for choosing appropriate materials for these new fuels and operating conditions will be critical. The initial analyses in this paper suggest that issues related to hydrogen embrittlement on the fuel-side of fuel injectors and enhanced oxidation of gas-side components could significantly reduce part life. However, many of the experiments done to understand corrosion and degradation in both metallic and coatings systems are not done at appropriate operating conditions or with relevant mixtures. Better experiments to understand the long-duration effects of high-hydrogen fuel combustion on these materials are required, and companion theory and simulation will be critical to understanding the engineering solutions necessary to enhance part life with these new fuels.

Finally, new advances in materials and manufacturing will support the transition to high-hydrogen fuels. Additive manufacturing (AM) has been a key enabling technology in development of novel combustion hardware (fuel mixers, burners, etc.) allowing for design freedom combined with simplified manufacturing of complex components. Manufacturers of high-hydrogen and fuel-flexible combustors are already taking advantage of technologies such as powder bed fusion AM. However, turbine components, traditionally made with superalloys, have seen less application of AM. The optimization of structural performance using directionally solidified (DS) and single crystal (SX) technology coupled with advanced TBCs and cooling is now reaching limits for the most advanced machines [122]. Only recently have industrial gas turbines seen the benefit of AM in the hot section with static guide vanes resulting in improvement in performance [123,124].

As discussed previously, the combustion of hydrogen-containing fuels will either increase temperatures or stresses in the hot-section structural components. To maintain current service intervals or improve durability under this increased demand for the structural vanes and blades, existing polycrystalline and directionally solidified superalloys will likely need to be upgraded to higher performing superalloys or alternatively take advantage of novel cooling and design features in AM coupled with new AM-specific superalloys that have the requisite high-temperature performance [125]. However, as demands for even higher efficiency are placed on new components where single-crystal superalloys may not provide adequate life, alternative material systems such as oxide dispersion strengthened materials or refractory alloys may provide new avenues if the challenges with processing can be overcome. Programs such as arpa-e'sultimate program are studying such approaches that address the significant processing challenges and new coatings that may be needed for these radically different material classes [126]. It is important to note that all these materials will still require improved coating systems for both thermal and environmental protection. As highlighted in this review, new coating strategies to mitigate spinel formation and delay TGO formation could provide improved durability with these high-hydrogen fuels.

Finally, ceramic matrix composites (CMCs) have already been demonstrated for robust operation in aerospace turbines and ultrahigh temperature CMCs could provide another alternative solution, but research is needed to further drive costs down [127]. Finally, it should be noted that not only the hottest section components will be impacted by increased temperature and/or loading; later stages of the turbine made today with more traditional cast superalloys will also need to be upgraded offering a plethora of opportunities for materials and manufacturing innovation.

The authors would like to thank Jim Harper, Tom Sambor, and Michael Caravaggio at EPRI, Chris Perullo at Turbine Logic, and Wenting Sun and David Wu at Georgia Tech for their discussions about the paper.

The authors attest that all data for this study are included in the paper.

C =

constant in the Larson Miller equation

LMP =

Larson Miller parameter

PF =

pattern factor

T =

temperature

c_p =

heat capacity

h =

heat transfer coefficient

k =

thermal conductivity

t_R =

rupture time