Effect of syngas fuel compositions on the occurrence of instability of laminar diffusion flame

The paper presents a numerical investigation of the critical roles played by the chemical compositions of syngas on laminar diffusion flame instabilities. Three different flame phenomena – stable, flickering and tip-cutting – are formulated by varying the syngas fuel rate from 0.2 to 1.4 SLPM. Following the satisfactory validation of numerical results with Darabkhani et al. [1], the study explored the consequence of each species (H2, CO, CH4, CO2, N2) in the syngas composition. It is found that low H2:CO has a higher level of instability, which however does not rise any further when the ratio is less than 1. Interestingly, CO encourages the heat generation with less fluctuation while H2 plays another significant role in the increase of flame temperature and its fluctuation. Diluting CH4 into syngas further increases the instability level as well as the fluctuation of heat generation significantly. However, an opposite effect is found from the same action with either CO2 or N2. Finally, considering the heat generation and flame stability, the highest performance is obtained from 25%H2+75%CO (81 W), followed by EQ+20%CO2, and EQ+20%N2 (78 W).


Introduction
An increase in the demand for energy worldwide encourages the research and development towards the production and utilisation of clean and sustainable energy [2]. Syngas, defined as the mixture of hydrogen (H 2 ) and carbon monoxide (CO), is one of the promising options in this respect ( [3,4]). It could be produced from various forms of feedstock such as food waste, biomass waste, coal, and agricultural waste [5e9], and [10]. However, this benefit has an impact on the diversity it presents in its fuel compositions. Depending on the production process and feedstock, syngas contains not only H 2 and CO but also methane (CH 4 ), carbon dioxide (CO 2 ), nitrogen (N 2 ) and other impurities at various percentages ([8e11]). Therefore, the design of a combustion system that is suitable for the diverse syngas is challenging. Inevitably, a microscopic understanding of the combustion characteristics to be affected by the diversity in syngas composition is fundamentally important for utilising this gas fuel effectively.
In the recent years, the characteristics of syngas diffusion flame have been studied in the several papers. This flame type being the energy source is also used in various applications at different scales, e.g. in industrial furnace, cooking hob, and external combustion chamber. The fundamental of this flame type in which the fuel and oxidiser streams are supplied into combustion separately is also similar to combustion in a gas turbine engine. A diffusion flame could also be categorised based on the flow regime of fuel stream as laminar or turbulent ([12e17], and [18]). These studies principally focused on the investigation of impact of syngas composition on the flame temperature, flame appearance, heat release, and emission.
The knowledge gap is identified from the review of literature for the flame condition when the fuel and oxidiser are supplied at an improper ratio that has the potential to destabilise the flame movement to cause instability such as 'flickering' and 'tip-cutting'. The difference between the phenomena of stable, flickering, and tip-cutting flames are illustrated in Fig. 1. Briefly, a flickering flame e as presented in the figure e only moves upward and downward forming an oscillative cycle, while its tip is cut at the body when it is a tipcutting flame. Due to this, the instability level of a tip-cutting flame will be higher than that of a flickering flame. Both these phenomena were studied in the several pieces of published literature for the diffusion flames of methane (CH 4 ), propane (C 3 H 8 ), mixture of both species (CH 4 þC 8 H 8 ), and hydrogen [19]. However, as reported in our earlier work that the diversity of syngas composition strongly affects the stable diffusion flame characteristics [20]. Thus, the composition is expected to play a significant role in the flame instability and this paper for the first time addresses this issue.
An in-depth understanding of flame instability is crucially important since this phenomenon directly affects the performance and consistency of the heat generation of flame. This work aims to fulfil the scientific knowledge gap by critically examining the impacts of the syngas fuel compositions on the instability of a buoyancy drive laminar diffusion flame. Also, the impact of these phenomena on the flame temperature and heat generation capability is particularly focused. Further, attention is paid on the flickering and tip-cutting phenomena with the influence of the variation of the flow rates of both fuel and co-flow air. The co-flow burner of [21] which was designed specifically for studying the tip-cutting and the flickering flame is modelled numerically in this paper. The validation of newly generated model is firstly presented, followed by the study of the impact of syngas composition on flame instabilities.

Model formulation
Detailed geometry of the co-flow burner modelled in the paper was already presented in Darabkhani and Zhang of [21]. In brief, the inner diameter of the fuel and co-flow air tube of the burner is 4.57 and 37.8 mm respectively, whereas the position of the co-flow air exit is 4 mm lower than the fuel exit. The i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 7 5 7 3 e7 5 8 8 shape of the fuel inlet tube is a tapered nozzle which assists the stabilisation of flame.
The burner is modelled by creating an axisymmetric domain according to the stated geometrical design, and presented in Fig. 2. The domain covers the volume above the fuel and air exit, with the height and width extended to 204 mm and 32.9 mm, respectively. It is worth noting that the symmetric axis is on the right-hand side of the domain.
The fuel and co-flow air inlets, at the bottom of the domain, are both defined as a velocity inlet. The top and left-hand sides of the boundary are both assinged a pressure boundary where the temperature and pressure are set to 293 K and 1 atm respectively. The species on these boundaries are air with a composition consisting of 79% N 2 and 21% O 2 . Lastly, the fuel and air tubes are defined as a no-slip wall boundary.
The generated mesh, as illustrated in Fig. 2, is designed particularly considering the interaction and chemical reactions to occur between the fuel and co-flow air streams, and also between the fuel/air stream and the fuel/air tube. The higher density zone, therefore, locates at the edge of the fuel and air tubes. The non-uniform mesh is developed by a hyperbolic function stretching from the smallest cell to the highest one. In the validation stage, the mesh dependency test is also performed considering three different levels of mesh resolution. The fine, medium and coarse mesh resolutions have the smallest cell size of 1.25E-4, 2E-4, and 3.33E-4 respectively, with the corresponding cell numbers of 39,560, 19,720, and 6720. Among them, the mesh level providing a good comparative result with a reasonable computational cost will be selected for the study stage.

Governing equations and numerical techniques
The numerical simulation employed is transient, thus the time step for each simulation case is computed based on the convective courant number, which is set as 1. The governing equations for the axisymmetric domain are presented below [22,23]: where u is the velocity component (ms À1 ), t is the time, r is the mixture density,x is the axial coordinate, r is the radial coordinate, the subscription x and r refers to the direction,m is the dynamic viscosity, p is the static pressure, and rg x is the gravitational body force. The term V, u ! is expressed as: Species transport equation i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 7 5 7 3 e7 where R i is the production rate of species i, J i is the diffusive heat flux which relates to the molecular diffusion and thermal diffusion, and Y i is the mass fraction of species i. The expression of J i is defined according to the Fick's law as where D i is the molecular diffusivity coefficient of species i, D i;t is the thermal diffusion coefficient for species i.
where q j is the energy flux, and Q r is the thermal energy production rate. The total energy (E) is computed based on the enthalpy (h) of the mixture as The dynamic viscosity (m) of the mixture is the combination of the same definition of each individual species based on their mass fraction. At the species level, the dynamic viscosity ðm i Þ of species i is defined by interpolating the gas property data for the major species such as H 2 , CO, CH 4 , CO 2 , N 2 , O 2 , and H 2 O with a constant value for the other species [22]. The thermal conductivity (lt) is computed according to the kinetic theory for an individual species then combined with the method proposed by Mathur-Saxena [23]. It considers the translation contribution ðC v;tran Þ, rotation contribution ðC v;rot Þ and vibration contribution ðC v;vib Þ as follows: The model considers the thermal radiation, and its transport equation is solved by the Discrete Ordinates Method (DOM) [24]. The radiative property e.g. absorption coefficient is calculated by the weighted sum of gray gas model (WSGG), in which the species H 2 O and CO 2 are assumed to dominate the cloud emission and absorption. According to the setup of the simulation domain, the medium is assumed to be optically thin, and the combustion product gases (H 2 O, CO 2 and N 2 ) occupy the majority of the volume. The optical path length ðSÞ supplied into the WSGG model is calculated as Volume of the domain surfacce area of domain (11) Further, the transient segregated implicit unsteady solver is selected for solving the continuity, momentum, species, and energy equations with the boundary conditions described in the previous sections. An inbuilt algorithm 'operator splitting' in STAR CCM þ CFD software is also utilised for formulating the species transport equation, and the details of this algorithm are available in Ref. [24]. The DRM22 [25] chemical mechanism, which consists of 22 species and 104 chemical reactions is imported to the model for processing the reaction pathway of each species. This chemical reaction mechanism is capable of predicting the combustion of syngas with the reasonable result comparing to the experimental one [26]. The sundials CVODE ODE solver, developed and introduced by Ref. [27], is selected for computing the reaction rates and chemical reactions.
The ignition is set to occur at the area above the fuel tube. The cells around this area are set to have a temperature of 1800 K for the first 0.5s to trigger the combustion. The time step is defined according to the convective courant number (ðCCNÞ) as The maximum value of CCN is monitored at every timestep, and this parameter is maintained at 1 ± 0.05 by increasing or decreasing the time-step through a user-defined function (UDF). The cell having the smallest size (an area close to the fuel tube) is found to have the maximum value of CCN while the value is 30e50% lower at the reaction zone of the flame. By applying this setup, the time-step is adjusted automatically by UDF which becomes stable once the maximum CCN satisfies the preferable range.

Data extraction
According to the study of [1], the analysis to be performed relates to the several important parameters, e.g. (i) the magnitude of flame oscillation (L f ), (ii) the oscillation wavelength (l), and (iii) the average oscillating flame height (H f -ave).
The parameters L f and l are already illustrated in Fig. 1, where L f is defined as the difference between the maximum and minimum flame height (H f -max -H f -min); and l is the difference between H f -max and tip-cutting position. Further, H f -ave is calculated from the average flame height during an instability cycle ((H f -min þ H f -max)/2). Thus, according to this, there are three parameters to be required for processing the simulation datasets: H f -max, H f -min, and tip-cutting position. Further, the values of L f and l are greater than zero for a tipcutting flame, whereas a flickering flame has the value of L f > 0 and l ¼ 0, and a stable flame has both L f and l equalling to zero.
The flame front line, where the stoichiometric combustion occurs, is drawn by a zero-temperature gradient method introduced by Ref. [16]. The position where the flame front line crosses the axis represents the flame height (H f ). The value of H f is monitored and plotted every time-step, and this action forms the profile of this parameter changing over time, and the instability cycle is observed from this profile. The value of L f , l and H f -ave are consequently calculated from the profile. Three different types of profile are obtained from the stable, flickering, and tip-cutting flames, and the values of H f -max, H fmin, and tip-cutting are extracted as presented in Fig. 3.
The same approach i.e. the average flame length (H f -ave) is also applied for the flame temperature (T) and the total chemistry heat release (Q). These parameters are, therefore, i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 7 5 7 3 e7 5 8 8 monitored at every time-step and subsequently, their profiles and changes over the time are extracted for the respective maximum and minimum values to finally compute the average flame temperature (T-ave) and the total chemistry heat release (Q-ave). By the same principle as L f , which is the difference between the maximum and minimum value, the fluctuations in T and Q are examined.

Validation of the generated model
The combustion condition and experimental results of Darabkhani et al. [1] are used as a reference at this stage, and accordingly the flame of CH 4 is formulated at the fuel rate of 0.2e0.35 SLPM (standard litre per minute). The flow Reynolds number (Re f ) consequently stays between 61.02 and 106.78, while the Froude number (Fr) between 1.249 and 3.845. The flames are buoyant dominated based on Ref [28]. The co-flow air rate is varied from 0 to 20 SLPM, where the ambient air acts as an oxidiser at 0 SLPM.
The values of L f , l, and H f -ave of the simulated flames are compared with the experiment results, as shown in Fig. 4. The dash lines connecting the values of L f , l, and H f -ave at each fuel and air rate are drawn to form the profiles of these parameters. All the CFD models, with the three different mesh resolutions, are shown to be capable of formulating the similar profiles of L f , l, and H f -ave when comparing with the experimental data. The same types of flame (stable, flickering, or tip-cutting) are predicted by the fine and medium mesh models at the same fuel and co-flow air rates as the experimental result. However, the coarse mesh model predicts the flickering and stable flames at the lower co-flow air rate than those of the fine and medium mesh models. To further clarify this, the values of L f and l approach to zero at a lower co-flow air rate than the one observed from the other mesh resolutions. The profiles of L f , l, and H f -ave are slightly overpredicted or underpredicted by the fine and medium mesh models. This difference, however, seems to be insignificant, so both the mesh levels are capable of providing a satisfactory result. Nevertheless, the medium mesh model is selected for the further study stage, since the performance of this model was comparable to that of the fine mesh model. Also it required lower computational resources and time to process the resulting data.

Impacts of the syngas fuel compositions
The validated CFD model (medium mesh) is now utilised to study the impacts of the syngas compositions on the occurrence of instability in the flames. At this stage, the fuel flow rate injected is varied from 0.2 to 1.4 SLPM, and combustion occurs with the ambient air as an oxidiser (i.e. the co-flow air rate is fixed to 0 SLPM). Various syngas compositions are defined to project the impacts of each species, as shown in Table 1. They are divided into two groups; (i) pure syngas (H 2 þCO) and (ii) syngas containing other species such as CH 4 , CO 2 , and N 2 . In the first group, the ratio of H 2 :CO concentration is varied to generate the syngas fuels having a different quantity of H 2 , e.g. 1:3 (CO-rich), 1:1 (EQ), and 3:1 (H 2 -rich). Whereas, the ratio is kept equal (1:1) for the second group, but the percentage of the additional species (CH 4 , CO 2 , and N 2 ) is varied from 10% to 20%.
The Reynolds number (Re f ) and Froude number (Fr) of the fuel stream are computed according to the range of the fuel rate, Re f becomes 10e70 for hydrogen, and 22e367 for the other fuels. The value of Fr calculated varies between 2 and 103, but this is significantly higher than the one determined in the validation stage. Thus, the flame requires another level of analysis and in this context, 'Flame Froude number ðFr f Þ', is introduced ([29e31]): where the parameters related to the stoichiometry mixture fraction (ðY s Þ), air/fuel density ratio (x), diameter of the fuel injector ðD f Þ, adiabatic flame temperature ðT ad Þ, velocity of fuel stream ðUÞ, gravitational acceleration ðgÞ, and ambient temperature ðT ∞ Þ. The flame with Fr f < 5 is defined as a buoyantdominated diffusion flame, while one with Fr f > 5 is a momentum-dominated flame. Fr f calculated from all the studied fuels is significantly lower than the critical value at  i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 7 5 7 3 e7 5 8 8 every fuel rate and thus considered as the buoyant diffusion flame.

Flame characteristics and heat generation
The temperature contour plot is initially analysed for projecting the appearance and movement of the flame during the instability cycle. At 0.2 SLPM, the stable flame is obtained from all the studied fuel compositions. The temperature contour plot representing the appearance of the stable flames of H 2 , H 2 -rich, EQ, and CO-rich is presented as an example in Fig. 5. While the structure of the stable flames at this fuel rate is similar, the temperature is affected by the fuel compositions. The high-temperature zone (>1800 K) is located on the midaxis and top of the flame and due to this, the impact of fuel composition on the flame structure is weak. However, the stable flame transforms to flickering at a relatively higher fuel rate, and the transition between the flame types occurring at a different fuel rate depends on the fuel composition. The movement and structure of the flickering flames are illustrated in Fig. 6 for H 2 , H 2 -rich, EQ, and COrich at 0.5 SLPM. As shown in the figures, the flame length increases to its maximum then drops back to the minimum value e this is typically the case occurred at all the fuel compositions. However, the flame structure is different, relying on the concentration ratio of H 2 :CO. For example, for the flames of H 2 and H 2 -rich with H 2 :CO > 1, the position of the high-temperature area is on the flame front and above the fuel tube. While for EQ and CO-rich with H 2 :CO 1, it is on the top part of the flame. The structure of the EQ syngas flames mixed with either CH 4 , CO 2 , or N 2 at 10% and 20% is also found to be the same as that of the EQ syngas, since all of them have H 2 :CO ¼ 1. Moreover, throughout the instability cycle, the high-temperature area expands and reduces both upward and downward along the flame front line. Also, the oscillation of the maximum flame temperature is observed; and the temperature escalates when the flame height increases.
Interesting, by increasing the fuel rate further, the flickering flame experiences a higher level of instability and subsequently becomes a tip-cutting flame. The fuel rate at which this transition occurs again strongly depends on the fuel composition, as shown in Fig. 7. For instance, the flame length escalates to its maximum value, where the tip cut occurs at the body of the flame. Following that, the flame fluctuates and consequently, its length decreases to the minimum value. The flame temperature, however, oscillates throughout the instability cycle and the high-temperature area is above the fuel tube at the beginning of the cycle. It is found that the hightemperature area is closer to the fuel tube for the flame of syngas with the higher H 2 :CO. This area moves along the flame front line to the axis at the same level where the tipcutting occurs. During the movement, the temperature of this area reduces, and the new high-temperature area is formed after the tip-cutting at the original position.
The average flame length (H f -ave) of all the studied fuel compositions is compared in Fig. 8a. The profiles of H f -ave is found to be almost linear and directly proportional to the fuel rate used. Particularly, H f -ave is higher when the fuel rate supplied is larger, with a significant role played by the species of CH 4 and CO. For instance, the syngas with a lower H 2 :CO has a higher value of H f -ave at the same fuel rate. Also, the EQ syngas mixed with CH 4 has a higher H f -ave than the EQ and EQ mixed with either CO 2 or N 2 . Further, the role of CH 4 is stronger than CO as the profiles of EQþ10%CH 4 and EQþ20% CH 4 have a higher H f -ave than the CO-rich case at a given fuel rate. Conversely, the higher percentage of inert N 2 leads to a lower value of H f -ave as seen from the profiles of EQþ20%N 2 and EQþ10%N 2 compared to the EQ case. Having CO 2 in the i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 7 5 7 3 e7 5 8 8 fuel composition surprisingly escalates the value of H f -ave when the flame is unstable, as seen from the profiles of EQþ10%CO 2 and EQþ20%CO 2 .
As discussed above, the flame temperature fluctuates during the instability cycle. Therefore, an average flame temperature (T-ave) is calculated for the different cases and presented in Fig. 8b. The profile of T-ave, escalates along with an increase of the fuel rate, but relies strongly on the H 2 :CO ratio. The flames of H 2 -rich and H 2 have a non-linear profile with a higher increasing rate at the lower fuel rate. On the other hand, the pattern is different for syngas having H 2 :CO 1. The profile of T-ave of EQ, CO-rich, EQ mixed with either CH 4 , CO 2 , or N 2 are also non-linear at the fuel rate lower than 0.5 SLPM but linear at the fuel rate above. This further implies that H 2 plays a significant role in the flame temperature, and the addition of CH 4 , CO 2 , or N 2 causes the reduction of T-ave. The impact of CO 2 is found to be stronger than N 2 and CH 4 at the same percentage of addition. Also, the level of impact of these species is stronger when the additional percentage of these species is higher, and the result is consistent for both the stable and unstable flames. The total flame chemistry heat release (Q total ) is computed by combining the heat released from the chemistry reaction at every grid on the simulation domain. The average total chemistry heat release (Q-ave) is calculated from the maximum and minimum of Q total , as presented in Fig. 8c. The pattern of Q-ave profile of all the fuel compositions is almost linear, and the higher average heat is generated at a higher fuel rate. The lower H 2 :CO of concentration percentage and the higher concentration of CH 4 in fuel composition encourage the heat generation capability. The impact of CH 4 is stronger than CO and H 2 . The comparable role of CO 2  are observed as both the species cause the reduction of Q-ave at an equal level when they are added to syngas at the same additional percentage.

Critical conditions
The type of flame at a specific fuel and air rate identified based on the value of L f and l, are presented in Table 2. Worth noting that the critical condition in this context refers to the transition occurring between the flame types such as stable/flickering and flickering/tip-cutting. The highest fuel rate at which a flame maintains its stable condition is also termed as a stable/flickering critical fuel rate ( _ v s/f ), while the highest fuel    and 20% of additional percentage, respectively. However, diluting CO 2 or N 2 at 10% and 20% conversely reduces the sensitivity level as the value of _ v f /t increases to 0.7 and 0.8 SLPM. The comparable role of CO 2 and N 2 on the critical fuel rate is again emphasised, and the dilution also rises the value of Re f /t to (i) 161 and 245 for 10% and 20% of CO 2 , and (ii) 141 and 195 for 10% and 20% of N 2 , respectively.

Flame instabilities
The parameters L f and l express the level of flame instability and tip-cutting phenomenon. Comparing the value of L f between the flames of different fuel composition, the higher value of L f represents the higher level of instability as the flame moves to the larger range. Similarly, the higher value of l implies the higher level of tip-cutting phenomenon since the distance between the maximum flame length and tip position is longer. For that, the profiles of L f and l of all the studied compositions are compared in Fig. 10aeb, and as seen the profile pattern of these parameters is similar for all the fuel compositions. Also, they are directly proportional to the fuel rate and it is reported that the higher level of instability and tip-cutting phenomenon escalate at a higher fuel rate. Specifically, the value of L f is found to be above zero at the fuel rate higher than _ v s/f . Similarly, l > 0 at the fuel rate higher than that of _ v f /t . Nevertheless, the values of L f and l of each fuel composition are different at the same fuel rate, and this finding, therefore, further reveals the impact of diversity in the fuel composition of syngas/producer gas on these parameters. The results again emphasise the comparable role of CO 2 and N 2 , while an addition of species CH 4 significantly increases the level of instability and tip-cutting phenomenon. Furthermore, the frequency range also relies directly on the fuel composition, as shown in Fig. 11aeb. More specifically, a higher H 2 :CO ratio in the syngas composition affects the lower frequency at the same fuel rate. The range is 12.5e13.2 Hz for CO-rich, which is higher than the one obtained from EQ (11.8)-(12.3) Hz), H 2 -rich syngas (10.8e12 Hz) and H 2 (9.6e11.8 Hz). The flame of EQ mixed with CO 2 has a higher F than EQ mixed with N 2 or CH 4 at the same additional i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 7 5 7 3 e7 5 8 8 percentage and fuel rate supplied into combustion. The frequency range of EQ mixed with CH 4 is 10.9e12.3 Hz, EQ mixed with CO 2 is 12.5e13.7 Hz, and EQ mixed with N 2 is 12.3e13.2 Hz. By comparing the range of F, the addition of CO 2 and N 2 rises the value of F, but an opposite direction is found for the addition of CH 4 .

Fluctuation of flame temperature and heat release
The fluctuations of the flame temperature (dT) and total chemistry heat release (dQ) compared in Fig. 12aeb are affected directly by the level of flame instability reported above. Similar to the profiles of average flame temperature (Tave ), dT relies strongly on H 2 :CO and becomes prominent for the flames of H 2 -rich (H 2 :CO > 1) and H 2 . This is also influenced by the fuel rate as seen for the H 2 -rich flames when the fuel rate >8 SLPM. However, for the CO-rich flame (H 2 :CO < 1), its magnitude drops but escalates again for the syngas flame having H 2 :CO ¼ 1. More specifically, in terms of a quantitative comparison, at the maximum studied fuel rate (i.e. 1.4 SLPM), the H 2 flame has the highest dT (95 K) followed by the H 2 -rich (81 K), CO-rich (61 K), and syngas (27e40 K) flames. Consequently, a conclusion could be drawn for this that H 2 playing an important role in destabilising the flame temperature.
The impacts of other species on dQ are also illustrated as it is reported that CH 4 significantly increases the fluctuation level in the heat generation capability, since the dQ profiles of EQ mixed with CH 4 is clearly above of the other fuel compositions. The role of CO 2 and N 2 are again comparable; as the value of dQ of EQ mixed with either CO 2 or N 2 is similar at the same additional percentage. Both the species reduce the fluctuation in the heat generation capability since the profile of EQ mixed with either of these species is below the profile of EQ. Again, at the maximum studied fuel rate (1.4 SLPM), the flame of EQþ20%CH 4 has the highest level of fluctuation (145 W) followed by EQþ10%CH 4 (116 W) and H 2 (91 W) respectively. The lowest level is observed from EQ mixed with either CO 2 or N 2 at 20% as~39 W.

Optimum heat generation
For an unstable flame, the values of Q-ave and T-ave are the average value without consideration of any instability. Thus, it is unreasonable to conclude that the higher value of these parameters would refer to the higher heat generation capability since they do not consider any fluctuation of the total  i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 7 5 7 3 e7 5 8 8 chemistry heat release (dQ) and the flame temperature (dT). Therefore, the flame having a lower value of dQ and dT, but with higher Q-ave is preferable. By this perception, the combustion condition that provides the maximum consistency heat (dQ ¼ 0 and dT ¼ 0) is thus the optimum heat generation condition. This condition also refers to Q-ave at the stable/ flickering critical fuel rate ( _ v s/f ). The comparison of Q-ave at this condition of all the studied fuel compositions is shown in Fig. 13. Additionally, the flame temperature (T-ave) at this condition is also presented along with Q-ave in this figure.
The highest optimum heat generation is reported by the CO-rich flame as 81 W at the flame temperature of 2153 K. This result is subsequently followed by the heat released from the flames of EQ, EQþ20%CO 2 , and EQþ20%N 2 . At this condition, the respective temperature of flame is 2118 K, 1945 K, and 2016 K. Consequently, the species such as CO, CO 2 , and N 2 encourage the higher optimum heat generation of syngas. The value of _ v s/f also represents the consumption of fuel for the optimum heat generation condition. The consumption of both CO-rich and EQ are 0.4 SLPM, while it is 0.5 SLPM for EQþ20% CO 2 , and EQþ20%N 2 . The lower consumption for the same amount of heat is the benefit of having CO in the syngas composition. In contrast, the flame of EQþ20%CH 4 has the lowest optimum heat generation (55 W at 2052 K) so adding CH 4 to syngas has a negative impact on the consistent heat generation.

Discussion
The stable, flickering and tip-cutting flames of syngas with various fuel compositions are formulated to project the impacts of fuel composition on the occurrence of flame instability. The key findings, especially considering the roles of each species obtained from the analysis are discussed below:The reaction mechanism selected is allowed the analysis of an impact of each species in the syngas composition. The studied fuel composition has single and multiple combustible species such as H 2 , CO and CH 4 , which have different reaction rate. The results obtained from the fuel having different volume fraction of these species are then expected to be different. Both H 2 and CO, which are the two major species in the syngas fuel composition, play a significant role in various aspects. For example, the analysis showed the advantage of having a higher concentration of CO over H 2 in the syngas composition. Though syngas with a lower H 2 :CO has a higher level of instability, the level has an upper limit when H 2 :CO ¼ 1. In addition, syngas with a lower H 2 :CO has a higher average total chemistry heat release with low fluctuation in the heat generation (lower dQ). With regard to CH 4 , it has a negative impact on the flame instability, but CO is the most desirable among the combustible species in the syngas composition. However, this finding is obtained by considering only the heat generation capability and flame instability and is, therefore, not included in the emission formation. On the other hand, H 2 plays a significant role in increasing the average flame temperature (higher T-ave) and consequently, encourages the fluctuation of this parameter.
For a stable diffusion flame, the higher concentration of CH 4 in the syngas composition is suggested since this species enhances the heat generation capability of the flame and also lowers the emission formation [20]. In an unstable flame, however, this species promotes the level of instability as well as the fluctuation in heat generation. Also, further adding CH 4 to the syngas gas composition raises the level so the flame becomes unstable when the fuel rate increases. Therefore, it is less desirable among the combustible species in the syngas composition when the oxidiser supplied into combustion is limited and the flame is unstable. The sufficient rate of oxidiser stream is necessary to maintain the flame stability and to potentially utilise the benefit of CH 4 on the heat generation capability effectively.
As a non-combustible species, the higher concentration of CO 2 and N 2 in the syngas composition decreases the capability in heat generation. Nevertheless, an addition of these species into the syngas composition has a benefit in the reduction of instability level and fluctuation in heat generation. Furthermore, the result points to the comparable role of CO 2 and N 2 on various aspects, e.g. the flame instability level (L f ), the average total chemistry heat release (Q-ave), and the fluctuation in heat generation (dQ). Diluting either of these species into the syngas composition is one of the options which could increase the flame stability. Nevertheless, this method requires further analysis for optimising the proper additional percentage, which balances between the reduced heat generation capability and the increased stability level.
The study result obtained in this work could also be compared with the study of [20] where the stable flame of syngas/producer gas is investigated extensively on various aspects. It is found that the effect of H 2 , CO, CH 4 , CO 2 and N 2 on the flame length, flame temperature and heat generation are in the same direction for the stable and unstable flames. The exception is observed only from the impact of CO 2 on the flame length. The higher concentration of this species results in the reduction of flame length (H f ) for a stable flame but increases the average flame length (H f-ave ) in an unstable flame. The result in this work also emphasises the finding of [20] for the relation between Q-ave and T-ave. The flame generating i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 7 5 7 3 e7 5 8 8 the higher Q-ave is not necessary to have a T-ave. For instance, the flame of EQþ20%CH 4 generates a higher Q-ave than COrich syngas at every fuel rate although its profile of T-ave is lower. As a result, the direction of the result found from the stable flame is also valid for the unstable flame.
All the studied flames are formulated from the same fuel rate (same velocity) and diameter of the fuel injector. Therefore, the critical Reynolds numbers for both the types of transition (e.g. Re s/f and Re f /t ) rely strongly on the fluid property (e.g. dynamic viscosity and density) according to the definition of Re s/f and Re f /t .
The simulation is processed on the axisymmetric domain. Thus, the impact of possible three-dimension vortex interactions e.g. vortex stretching which could lead to the extended computed flame is excluded. However, this limitation does not play a significant role since the laminar flow is studied in this work.
Lastly, further study is suggested for the impact of syngas composition on the elimination of flame instability by supplying a larger amount of oxidiser. The different instability level of the flame of various syngas composition would demand a different amount of oxidiser to eliminate. The impact of each species in syngas composition on this aspect is significant, and the result could be the guideline for reducing and eliminating the flame instability on various combustion applications.

Conclusion
A CFD model was formulated for studying the impacts of the diversity in the syngas composition on the occurrence of the flame instability in a laminar co-flow diffusion flame. The stable flame was observed at a low fuel rate which, however, transformed to a flickering and tip-cutting flame respectively at a higher fuel rate. The effect of species in the syngas composition on this phenomenon was also examined and the key conclusion is as follows: -Syngas with a lower ratio of H 2 :CO has a higher level of instability; nevertheless, the level does not rise further when the ratio is lower than 1. Adding CH 4 to syngas significantly increased the instability level, and the flame was more sensitive to the fuel rate. The opposite was found when syngas was diluted with CO 2 and N 2 . -CH 4 and CO concentrations in syngas have a significant impact on the increase of heat generation capability. CH 4 has a stronger role than CO in this aspect; however, higher concentration of CH 4 in the fuel composition also promotes the fluctuation in the heat generation significantly. This suggests that CO is preferable than CH 4 as the higher concentration of CO in syngas could reduce the level of heat generation fluctuation. -The effect of H 2 on the increase of flame temperature was significant, and this species also encouraged the fluctuation level of flame temperature. -The effect of H 2 , CO, CH 4 , CO 2 and N 2 on the flame length, temperature, and heat generation was mostly in the same direction for the stable and unstable flames. An exception was obtained from the effect of CO 2 on the flame length.
The higher concentration of this species reduced the flame length (H f ) in the stable flame but increased the average flame length (H f-ave ) for the unstable flame.
Finally, a diffusion flame, as an energy source, and also the development of production process of syngas for highperformance fuel composition, provide a higher level of flame stability and heat generation. Therefore, it is anticipated that the results presented in the paper have the potential benefit in the effective design of a combustion system for the utilisation of syngas and hydrogen fuels.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.