Pollutant Emissions Reporting and Performance Considerations for Ammonia-Blended Fuels in Gas Turbines

Today, despite serious climate change and energy security concerns, fossil fuel combustion represents the world's dominant mode of the primary energy conversion, as well as its greatest source of greenhouse gas (GHG) emissions [1]. Lower-cost renewable energy sources, such as solar and wind, remain fundamentally intermittent and often depend upon supplemental gas turbines to dispatch fossil-based power when renewable energy availability is low. This intermittency issue has led to significant momentum in policy and industry circles to replace fossil fuels with alternative carrier fuels like hydrogen (H₂) and ammonia (NH₃). Such fuels can be synthesized sustainably with excess renewable power, have high energy densities suitable for efficient transport and storage, and are suitable for on-demand use in gas turbines without contributing to carbon dioxide (CO₂) or methane (CH₄) emissions. Nonetheless, several important issues besides GHG emissions also arise in the context of a large-scale switch to “green” alternative fuels.

Among these, the main emphasis of this study is toward quantifying the potential air quality impacts associated with burning alternative fuels. In general, combustion releases a smattering of pollutants that are harmful for environmental and public health, and a variety of ordinances spanning international to local codes exist to limit such emissions. (For example, the US EPA's CFR Part 60 Subpart KKKK and European Council's Directive 2010/75/EU set emissions limits for stationary gas turbine operators [2,3].) Compared to fossil fuels, burning alternative fuels, such as H₂ and NH₃, release a smaller range of pollutants since these fuels do not contain the carbon atoms or impurities necessary to produce, for example, carbon monoxide (CO), particulate matter (PM_x), and sulfur oxides (SO_x). The possibility of eliminating such pollutants from combustion exhaust streams is a major draw for alternative fuels.

Nonetheless, not all pollutants are derived from fuel-borne carbon or impurities. Any air-breathing combustion system—even one burning an alternative fuel—produces nitrogen oxides (NO_x) due to spontaneous reactions of atmospheric nitrogen (N₂) and oxygen (O₂) that occur at high temperatures. Additionally, in the case of NH₃, fuel-bound nitrogen atoms are also partially converted to NO_x during combustion. Since NO_x is a poisonous and reactive gas that detrimentally affects human respiratory health and contributes to acid rain, there is widespread interest in quantifying the NO_x emissions from alternative fuels and comparing them to those of conventional fuels.

Even so, formulating a consistent and practicable framework for comparing NO_x and other combustion emissions from a fleet of distinct systems burning a variety of different fuels is not trivial. In general, combustion emissions assessments are ultimately concerned with the pollutant mass emitted per unit of delivered energy (e.g., ng NO_x/J), which is equivalent to the mass emission rate per unit of delivered power. However, quantifying combustion emissions on a mass-per-energy basis involves separate measurements of the exhaust gas composition, the exhaust flowrate, and the delivered power—a triad that is not always feasible to obtain in an operational system.

To circumvent this difficulty, the combustion community adopted a surrogate volumetric emissions reporting approach in the 1970s that is based only on gas analyzer measurements of the exhaust composition. In this now-standard framework, combustion emissions are quantified in terms of a pollutant's adjusted concentration after first drying the exhaust gas sample (i.e., removing all H₂O) and subsequently diluting (or concentrating) the dry sample to a reference O₂ concentration (typically 15% O₂ for gas turbines and 3% for industrial processes). In the U.S., the resulting emissions are reported on a volumetric basis in terms of the dry parts-per-million at the reference %O₂ (here 15%), a unit abbreviated as ppmvdr. In Europe, another functionally equivalent metric based on emitted pollutant mass per dried and O₂-referenced normal cubic meter (mg/drNm³) is often used. These metrics can be related via a simple conversion factor involving the pollutant molecular weight. For example, for NO_x,$1ppmvdrNOx=2.05mgNOx/drNm3$⁠. In either case, in the context of traditional fossil fuels, these adjustments tend to offset the effect of variations in fuel equivalence ratio, bypass air ratio, and steam injection on the resulting concentration values. This property manifests a robust correlation between a pollutant's dry, O₂-referenced volumetric concentration and its mass-per-energy production rate within a wide range of gas turbine systems firing hydrocarbon fuels—a fact that is exploited in numerous active emissions codes, including Refs. [2,3]. Crucially, however, this correlation is not fuel-independent [4,5].

On this subject, the authors of this study recently assessed approaches for quantifying combustion emissions from hydrogen–hydrocarbon fuel blends [6]. That work demonstrated that dry, O₂-referenced volumetric pollutant concentration values are strikingly improper for comparisons among high-hydrogen fuels and hydrocarbon fuels, with a penalty of 40% against hydrogen systems. More precisely, this 40% discrepancy was shown to be an artificial bias introduced indirectly by the drying and dilution adjustments rather than by any physical increase in the mass-per-energy emission rate. It is important to emphasize that these indirect effects are independent from changes to true pollutant production rates, and instead arise from how changes in other combustion or system parameters feed back into the variables used to quantify adjusted volumetric emissions. The direct effects that do follow from fuel- or condition-specific changes in the chemical pathways involved in pollutant formation are a separate issue that is not discussed here but has been considered elsewhere [7–9].

In this paper, we build from our earlier study of indirect emissions assessments from hydrogen fuel blends and consider how pollutant reporting practices based on dry, O₂-referenced volumetric concentration values affect combustion emissions interpretations from ammonia-based fuel blends. Importantly, ppmvdr values from systems firing NH₃ are not significantly biased by the drying and dilution adjustements relative to systems firing CH₄ [10]. Nonetheless, the present results show that other indirect influences still give rise to a 10% discrepancy against NH₃ when comparing ppmvdr values in ammonia and hydrocarbon fuel at an equal mass-per-work emission rate. Additionally, we consider how the practice of catalytically “cracking” NH₃ into its constituent N₂ and H₂ parts can lead to more significant ppmvdr disparities of up to 21% compared to hydrocarbons at a constant mass-per-work emission rate. In order to establish the source of these differences, a detailed accounting of the indirect influences confounding the relationship between ppmvdr and ng/J emissions values across a range of fuels is provided throughout the analysis. Following our discussion of these results, we conclude that the practice of quantifying combustion emissions using dry, O₂-referenced volumetric concentrations is generally inappropriate for the purpose of emissions comparisons across different fuels.

As mentioned above, emissions reports based on ppmvdr values are subject to a variety of indirect effects that confound the correlation between emissions quantified by ppmvdr and ng/J under different operating conditions. These include the drying and O₂-dilution adjustments already mentioned, as well as a variety of additional thermodynamic factors that relate the volumetric concentration of a generic minor exhaust product to the mass of that minor product emitted per unit of delivered energy. We seek to quantify these indirect effects. However, in a head-to-head comparison of real or simulated gas turbine systems firing different fuels, these varying indirect effects are convoluted by physical changes in emissions production rates (i.e., direct effects).

To avoid grappling with such complicated and uncertain influences in a series of detailed experiments or simulations, the approach adopted here and in our earlier work is much simpler. We compute the constant-pressure chemical equilibrium properties of a variety of fuel blends with a fixed adiabatic flame temperature within a simple Brayton cycle using ansyschemkin and the npss tool [11]. Conditions representative of an F-class gas turbine are considered with a compressor inlet temperature of 300 K, compressor discharge conditions of 17 atm and 700 K, a mixture adiabatic flame temperature of $Tad=1800$ K, and a turbine inlet temperature of 1600 K. Some sample data obtained for the (unblended) fuels considered in this study under these conditions are presented in Table 1. Nonetheless, it is worth remarking that the main results toward emissions comparisons are not sensitive to the specific conditions chosen, as shown in previous works [4–6].

where $nF/nP$ is the ratio of the reaction's fuel to product moles, $Δh¯c$ is the molar enthalpy of combustion (i.e., the molar lower heating value), and M_i is the molecular weight of species i.

The results of this study consist of assessments of indirect combustion emissions from propane (C₃H₈), hydrogen (H₂), ammonia (NH₃), and cracked ammonia (⁠$14N2+34H2$⁠) fuels blended with methane (CH₄) in ratios spanning 0% to 100%.

The most obvious indirect factors influencing the relationship between ppmvdr and ng/J emission values in different fuels are those of drying and O₂-dilution, which are quantified in Fig. 1. These indirect effects describe the relationship between ppmvdr and ppmv values. They occur explicitly because of the differences in the concentrations of steam and oxygen formed when burning hydrocarbon fuels and alternative fuels. For hydrocarbon fuels, heat is released when the fuel atoms are oxidized into both H₂O and CO₂. Conversely, alternative fuels like H₂ and NH₃ are oxidized simply into H₂O, leading to relatively higher concentrations of steam in the combustion products. Since we consider reactions where the equivalence ratio is adjusted to maintain a constant $Tad$⁠, the fuel composition also has a significant effect on the amount of generated H₂O and leftover O₂ in the combustion products through its heating value $Δh¯c$ and the resulting fuel-to-product mole ratio $nF/nP$⁠. The overall consequence of these differences in steam generation on the drying correction from Eq. (2) for various fuel blends is presented in Fig. 1(a), where the removal of higher proportions of steam from alternative fuels is shown to yield a larger drying correction compared to hydrocarbon fuels.

Besides such differences in steam generation, the properties of the reactant mixture also have significant effects on the reaction's consumption of oxygen. In addition to the effect of the fuel-to-product mole ratio already mentioned above, the presence or absence of carbon in the fuel also impacts the product O₂ concentration since the formation of CO₂ consumes more O₂ and releases less heat than the formation of H₂O. The net effect of these different influences on the leftover oxygen on the dilution correction from Eq. (3) is shown in Fig. 1(b) for the same range of fuel blends.

It is notable from Fig. 1(b) that the O₂-dilution correction for high-%NH₃ blends shows an opposite trend from that of high-%H₂ blends. This difference has an important consequence on the combined drying and dilution factor $A1A2$ shown in Fig. 1(c). While drying and O₂-dilution lead to a 40% difference in the ratio of ppmvdr values to true ppmv when comparing H₂ and CH₄ fuels, this same difference is only 2.1% when comparing NH₃ and CH₄. Hence, unlike in H₂–CH₄ blends, drying and O₂-dilution do not substantially affect the relationship between ppmvdr values and true ppmv values for NH₃–CH₄ fuel blends. Nonetheless, there are other sources of indirect effects that relate ppmv values to mass-per-work emissions, and these will later be shown to have significant implications for comparisons of emissions among NH₃–CH₄ fuel blends on the basis of volumetric concentrations.

Before advancing further, it is worth briefly commenting on the fact that exhaust from cracked ammonia fuel experiences a smaller drying and O₂-dilution correction (20% relative to CH₄) than from H₂ fuel. This is notable because cracking NH₃ yields H₂—it is equivalent to H₂ fuel with extra N₂ diluent. This difference in drying and O₂-dilution behavior arises primarily because the extra N₂ increases the fuel equivalence ratio required to reach $Tad$⁠, leading to less excess O₂. Hence, ppmvdr values from cracked ammonia combustion will generally be less affected by the drying and O₂-dilution corrections compared to pure H₂ (see Fig. 1(b)).

The next indirect effect considered accounts for how the thermodynamic properties of the mixture affect the relationship between the true (i.e., wet, undiluted) pollutant concentration in the exhaust and the mass of pollutant formed per unit of fuel energy input. As shown in Eq. (4), these thermodynamic effects arise explicitly due to variations in the fuel heating value and the fuel-to-product mole ratio. Note that the scaling of A₃ by the molecular weight of the pollutant is constant for a given pollutant species regardless of the fuel. To fix the molecular weight in this example, we consider the pollutant species to be NO_x, for which the approximation $MNOx=NNO2=46$ g/mol is typical.

It can be observed in Table 1 that $Δh¯c$ and $nF/nP$ both vary significantly across the considered fuels. However, the variation of A₃ for different fuels presented in Fig. 2(a) reveals that these effects tend to wash each other out such that the overall contribution of the indirect thermodynamic effect is relatively small. More quantitatively, the relative variation is only 3% between H₂ and CH₄ fuel and less than 1% between NH₃ and CH₄. Even less variation is noted in blends of these fuels. Hence, as shown in Fig. 2(b), these indirect thermodynamic effects play a relatively unimportant role in the correlations between ppmvdr and ng/J emissions values among these fuels.

For cracked ammonia–methane blends, on the other hand, Fig. 2(a) reveals that the thermodynamic correction experiences wider variations of about 7% and displays nonmonotonic behaviors as a function of the blend ratio. This nonmonotonicity of thermodynamic indirect effects manifests a corresponding nonmonotonic influence on the relationship between ppmvdr and ng/J values, as seen in Fig. 2(b). This is significant because it indicates that, unlike H₂ and NH₃ blends, which both experience the largest differences in indirect effects compared to CH₄ at the extreme of purity (i.e., 100% H₂ or NH₃), the “worst case” for cracked ammonia occurs at an intermediate blend ratio. Hence, this demonstrates that it is not generally sufficient to only evaluate indirect effects in pure fuels in order to bound the amplitude of possible indirect effects among the various blend ratios of those fuels. Depending on the thermodynamics of the mixture, the relationship between ppmvdr and ng/J may vary in a complex way as the blend ratio is varied.

The final indirect effect visible in Eq. (1) is the cycle efficiency, which describes the performance of the gas turbine. The cycle's thermal efficiency is a function of the fuel composition because the varying composition of the combustion products of different fuels leads to changes in the mixture specific heat, particularly due to changes in its H₂O content. Such changes control how much enthalpy the exhaust carries at a fixed turbine inlet temperature, which directly alters how much work the turbine can extract per unit of thermal energy input.

The thermal efficiencies resulting across the spectrum of considered fuel blends for the conditions described in Sec. 2 is presented in Fig. 3(a). These results indicate that the alternative fuels all experience an efficiency boost compared to methane. For H₂, this leads to an efficiency increase of less than 1% (a 2.1% relative increase). However, for NH₃, a more significant efficiency increase of almost 3% arises (a 7.4% relative increase). As indicated above, these rises in efficiency can be attributed to the increased specific heat of the NH₃ exhaust due to its higher steam content. As may be expected, the efficiency of the cracked ammonia cycles lies in between the efficiencies of the hydrogen and ammonia cycles.

Including these performance effects, the overall relationship between ppmvdr and ng/J emissions output can be quantified using Eq. (1). The resulting relationships for all of the considered fuel blends are shown in Fig. 3(b). These results indicate significant indirect effects for all of the alternative fuels considered. The most heavily affected fuel relative to CH₄ is H₂ with a relative discrepancy of 39% as reported in our earlier work [6]. However, both ammonia and cracked ammonia still do experience substantial indirect effects that penalize them relative to methane on the basis of ppmvdr values. For NH₃, this effect amounts to a 10% bias, while for cracked ammonia it is 21%.

As gas turbine systems begin to transition to green alternative fuels, significant interest has arisen toward assessing the potential air quality impacts of this decarbonization strategy. Nonetheless, the decades-old standard practice of quantifying combustion emissions using dry, O₂-referenced volumetric concentration measurements (i.e., ppmvdr) does not yield a robust or accurate comparative metric across all fuels. This paper has detailed how different fuels influence the relationship between a gas turbine system's ppmvdr values and its emitted mass-per-delivered energy emission rate. Placing particular emphasis on ammonia-based fuels, the individual drying, dilution, thermodynamic, and performance effects indirectly controlling this fuel-dependent relationship were quantified separately in order to establish mechanisms before assessing their combined influences.

The results of this analysis revealed that gas turbines burning ammonia fuels experience 10% higher ppmvdr emissions values compared to their hydrocarbon-fueled counterparts at an equal mass per delivered energy emission rate. This analysis was also repeated for turbines firing cracked ammonia fuel, which corresponds to H₂ fuel with additional N₂ playing the role of a diluent in the reaction. Compared to methane, cracked ammonia showed 21% higher ppmvdr values at the same mass-per-work emission rate. While these differences are both smaller than the 39% increase previously reported for hydrogen fuels [6], they are still significant for air quality assessments.

It is also interesting to note that the H₂–CH₄ and NH₃–CH₄ discrepancies are dominated by different indirect effects. For the H₂–CH₄ comparison, the dominant indirect effect stems from how drying and dilution corrections lead to a relative concentration of pollutants in hydrogen exhaust in comparison to hydrocarbon exhaust. For NH₃–CH₄ comparisons on the other hand, drying and dilution only contribute a bias of less than 3% compared to hydrocarbon fuels. The main driver of increased ppmvdr values in ammonia systems is instead mostly a performance effect driven by the efficiency boost linked to the increased specific heat of the ammonia exhaust compared to hydrocarbon exhaust.

Based on these results, we conclude that assessments of combustion emissions relying on volumetric emissions metrics are fundamentally inappropriate for comparisons of emissions involving alternative fuels. Nonetheless, there are a large number of recent studies available in the literature that have improperly compared pollutant emissions among traditional and alternative fuels solely on the basis of ppmvdr values. This is especially important today as these flawed reports—which we have shown exaggerate emissions from alternative fuels—are being cited in policy decisions that affect the penetration of renewable power systems. To avoid the possibility of further misinterpretations of emissions due to convolutions with indirect effects, we encourage the combustion community to organize a concerted shift toward robust metrics for comparing emissions in terms of the pollutant mass emitted per unit of energy delivered.

The authors thank Christopher A. Perullo from Turbine Logic for sharing cycle calculation data that were used in this paper.

• This project has received funding from the European Union’s Horizon 2020 Research and Innovation Programme under the Marie Skłodowska--Curie Grant Agreement No. 899987; Funder ID: 10.13039/501100007601.

A₁ =

drying factor

A₂ =

15% O₂-dilution factor

A₃ =

thermodynamic factor

χ_i =

volumetric (molar) concentration of species i

GHG =

greenhouse gas

$Δh¯c$ =

molar heat of combustion

ηT =

thermal efficiency

M_i =

molecular weight of species i

$nF/nP$ =

mole ratio of fuel to products

NPSS =

numerical propulsion system simulator

ppmv =

parts per million (volumetric)

ppmvd =

dry ppmv

ppmvdr =

dry ppmv at the reference O₂ concentration

$rO2$ =

reference O₂ concentration (15% in this study)

$Tad$ =

adiabatic flame temperature