SPATIAL AND TIME RESOLVED STUDY OF TRANSIENT
PLASMA INDUCED OH PRODUCTION IN QUIESCENT
CH4-AIR MIXTURES
by
Charles Cathey
A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING)
December 2007
Copyright 2007 Charles Cathey
Abstract
This thesis presents an experimental study into the utility of transient plasma ignition (TPI) of hydrocarbon fuel-air mixtures and the basic physics behind said phenomena. Transient plasma has several advantages over traditional spark ignition, consistently demonstrating reductions in ignition delay and extended lean burn capability. It is postulated that high energy electrons generated during the electrical discharge are responsible for electron impact dissociation, ionization, and fragmentation of molecules in the medium resulting in radical formation that drives the combustion process. The experiments described in this thesis illustrate TPI’s effects and attempt to understand the underlying physics by looking at transient plasma induced production of radicals, ignition, and flame propagation.
Using optical diagnostic techniques the ability of a transient plasma to populate a cylindrical discharge volume with the hydroxyl radical (OH) is analyzed during combustion of stoichiometric quiescent CH4 – air mixtures. Ground state OH production is analyzed using planar laser induced fluorescence (PLIF). Broadband transitions of OH near 309 and 314 nm were also used to monitor transient plasma production of OH* via optical emission spectroscopy (OES). A high speed camera was used to image ignition and flame propagation in the chamber, providing spatial and temporal resolution over the entire combustion event.
Transient plasma successfully ignited the CH4 - air mixture, populating the discharge volume with radicals. Mean OH number densities produced by the
x discharge were found to reach ~ 4·1014 molecules – cm-3 and decay within 100 µs of the plasma. Ignition under these conditions was found to occur approximately 1 ms after the discharge along the anode, creating a cylindrically expanding flame front.
TPI was found to be a truly volumetric ignition source creating multiple ignition kernels whose proximity to the anode is consistent with the region of highest field and thus maximum radical density. Additionally the rapid decay of the initial OH (as seen by the pump beam) with respect to the time of ignition suggests that the anode itself may have an important role in sustaining and enhancing combustion kinetics.
xi Chapter 1 Introduction to Transient Plasma
Introduction
There is a great interest in the combustion community for non-equilibrium
(Te >> Tgas) plasma devices that accelerate ignition and stabilize combustion.
Specific applications are in the development and implementation of non-equilibrium plasma ignition sources for supersonic and hypersonic aircraft and low emission automobiles. Using the non-equilibrated plasma in first stages of arc development for ignition, subsequently known as transient plasma ignition (TPI), great reductions in ignition delay a factor of 2 to 9, and the extension of the lower flammability limit for a variety of fuels has been demonstrated.
Equilibrated Atmospheric Pressure Plasmas
The most familiar ignition methodology uses an equilibrated (Te = Ti) or nearly equilibrated plasma in the form of an arc or spark which is commonly used in both automobile and aircraft engines. For this type of discharge, ignition is achieved by local heating of the gas, which increases the dissociation rate, and reactions of chain branching and propagation.
The arc discharge is the classic example of a quasi-equilibrated plasma, even though they are often far from equilibrium. The arc cathode has a relatively low cathode fall voltage of about 10 eV, corresponding to the ionization potential.
Additionally electron emission is via thermionic and field effects. The high current typical in arcs results in the cathodes and the gas being subjected to large amounts of joule heating. The high amount of current produced leads to erosion and evaporation
1 of the cathode. However, there is an alternate way to ignite the mixture in the form of non-equilibrated plasmas.
Non-equilibrated Atmospheric Pressure Plasmas
The basic difference between equilibrated (thermal) plasmas and non- equilibrated (non-thermal) plasmas is in the ionization mechanisms of the two processes. The ionization process in non-equilibrated plasmas is dominated by electrons impacting with “cold” non-excited atoms and molecules. Additionally in non equilibrated plasmas the majority of the energy goes into creating highly energetic electrons instead of heating of the gas. These electrons produce excited species through impact dissociation, excitation, and ionization of background gas molecules in the system. Another area of distinction is that non-equilibrated plasmas seem to be very selective as to the allowed plasma chemical reactions. A common application would be treatment of exhaust gasses, where the radicals created via the non thermal plasma, are able to oxidize, or decompose the pollutant molecules.1
Whereas pollution control with a thermal system (plasma torch) heat the entire gas system in order to destroy the pollutants.
Electric Breakdown of Gases
The subsequent sections describing the breakdown process are based on
Fridaman and Razier.2,3 When an electric field is applied some primary electrons near the cathode will form and drift towards the anode. As the electrons drift towards the anode, the gas will ionize and an avalanche will form. Electrical breakdown in a gas is usually preceded by an electron avalanche, where a 2 multiplication in the number of primary electrons occurs over in a cascade of ionization. The ionization in an avalanche is described by the Townshend ionization coefficient α, which refers to the electron production per unit length. The
Townshend ionization coefficient is related to the ionization rate coefficient ki (E/n0) and electron drift velocity vd by:
ν i 1 1 ki (E / n0 ) α = = ki (E / n0 )n0 = (1-1) ν d ν d µe E / n0
Where vi is the ionization frequency, µe is the electron mobility, and E/no is the applied electric field over the neutral gas density, also called the reduced electric field. Each primary electron produced near the cathode creates exp(αd)-1 positive ions in the gap. The total electronic current is the electronic current at the anode, because of the absence of ionic current, which gives the Townsend formula:
i exp(αd) i = 0 (1-2) 1− γ[exp(αd) −1]
γ is the secondary electron emission coefficient, and depends on the cathode material, sate of the surface, type of gas, and E/n0. The current will not sustain as long as the denominator in equation (1-2) is positive. As the applied electric field is increased the Townsend α coefficient will also increase and eventually will push the denominator to zero, and transition to self sustained current, or breakdown occurs.
Thus the simplest breakdown condition can be expressed using (1-2) as:
⎛ 1 ⎞ γ[exp(αd) −1] = 1 , or αd = ln⎜ +1⎟ (1-3) ⎝ γ ⎠
It is convenient to re-write equation 1 in terms of α in the following way:
3 α ⎛ B ⎞ = Aexp⎜− ⎟ (1-4) p ⎝ E / p ⎠
Where A and B are experimental parameters. In air A = 15 1/ cm Torr, and B = 365
V/cm Torr. Combining equations (1-3) and (1-4) gives what is commonly called the
Paschen curve.
B( pd) V = (1-5) break C + ln( pd)
⎛ ⎞ ⎜ ⎟ ⎜ A ⎟ Where C = ln⎜ ⎟ , which accounts for the weak influence of secondary ⎜ ⎛ 1 ⎞ ⎟ ⎜ ln⎜ +1⎟ ⎟ ⎝ ⎝ γ ⎠ ⎠ electron emissions.
The Townshend breakdown formula is only applicable for pd < 4000 Torr cm, or at atmospheric conditions, d < 5 cm. In larger gaps, and higher pressures spark breakdown occurs. While the Townsend breakdown is quasi-homogenous, spark breakdown occurs across a narrow channel, is not directly related to electrode phenomena, and has very high current currents. The method of spark breakdown is based on the concept of the streamer.
Electron Avalanches
The electron avalanche that occurs when an electric field is applied is a critical step in the breakdown process. In electronegative gasses, like air, electrons will have a tendency to attach to oxygen molecules. Accounting for electron
4 attachment, the rate of change of the total number of electrons Ne, positive N+ and negative N- ions in an avalanche moving in the direction x is: dN dN dN e = ()α − β N , + = αN , − = βN (1-6) dx e dx e dx e
Where α is the Townshend ionization coefficient and β is the Townsend attachment coefficient. Assuming that the avalanche starts from only the primary electrons the number of charged particles can be found using equation 6 as:
α β N = exp[(α − β )x], N = (N −1) , N = (N −1) (1-7) e + α − β e − β −α e
The electrons in the avalanche will move in the x direction with the applied electric field at their drift velocity, vd = µeEo. However, free diffusion (De) of the electrons will also be taking place, spreading the electrons out radially (direction r) from the x direction. Accounting for drift and diffusion, the electron density in the avalanche is given as:
2 2 1 ⎡ (x − µe E0t) + r ⎤ ne (x,r,t) = exp⎢− + (α − β )µe E0t⎥ (1-8) 3 / 2 4D t ()4πDet ⎣⎢ e ⎦⎥
Equation (1-8) shows that the avalanche electron density will decrease with distance from x as diffusion occurs. Taking into account the Einstein relation between electron mobility and free diffusion the avalanche head radius, rA will also increase with time, and distance of propagation along x (x0).
x0 4Te rA = 4Det = 4De = x0 (1-9) µe E0 eE0
5
Using equation (1-8) the space distribution of charge for the negative and positive ion densities during avalanche propagation are:
t t n (x,r,t) = αµ E n (x,r,t')dt' , n (x,r,t) = βµ E n (x,r,t')dt' (1-10) + ∫ 0 0 e − ∫ 0 0 e 0 0
A simplification for positive ion space distribution can be found using equations (1-
10) and (1-8) for densities close to the x axis, in the absence of attachment as t → ∞.
α ⎡ r 2 ⎤ n+ (x,r) = 2 exp⎢αx − 2 ⎥ (1-11) πrA (x) ⎣ rA (x)⎦
Equation (1-11) tells us that the ion a concentration in the tail of the avalanche is growing along the x axis is dictated by the multiplication and exponential increase in the electrons.
Figure 1.1: Avalanche propagation with a drift velocity, vd towards the anode under the influence of an external applied electric field, Eo,
6 Figure 1.1 depicts an avalanche propagating towards the anode. It is important to note, that if you were to image the avalanche you would see a wedge shape, where the visible radius is growing linearly, in contrast with the square root of x as you would expect from equation 9.4 Knowing that the visible radiation can be determined by the absolute density of the exited species and is proportional to the exponential factor Φ from equation (1-11), and grows with x, equation (1-11) can be modified to the following; explaining the linearity of a potential image.
2 r (x) 4Te x 4Teα 2 = αx − ln Φ , r(x) ≈ rA (x) αx = αx = x (1-12) rA (x) eE0 eE0
An important change in avalanche behavior takes place when charge amplification, exp (αx), becomes large. When this occurs significant space charge is created, which posses its own electric field, Ea, which should be added to the applied external field E0. This field occurs because a dipole is created between the electrons at the head of the avalanche and the ions which have not moved. The dipole has a characteristic length of 1/α, which is the distance the electrons move prior to ionization, and has a charge of Ne ≈ exp (αx). For example, in atmospheric air, breakdown occurs at approximately 30 kV/cm, and α ≈ 10 cm-1, which means that the characteristic ionization length can be estimated to be .1 cm.
7
Figure 1.2: Electric field distribution during the electronic avalanche, dipole induced field, E’, and applied electric field, E0 are shown (left), and combined field, E, (right).
Figure 1.2 depicts the electric field distortion due to the avalanche induced dipole. Observe that the induced dipole adds constructively to the applied field, actually making the total electric field stronger in front of (and behind) the avalanche head, which will further accelerate ionization. Also note that inside the avalanche or between the separated charges the total electric field is lower so ionization is slowed relative to the applied field. Another important part of the space charge induced field is the radial component. At a critical distance, αx, close to the avalanche radius, the electric field of charge, Ne ≈ exp (αx), will equal the value of the applied field, E0.
In air the avalanche radius, rA ≈ .02 cm, and the critical value when the avalanche field is comparable to the applied electric field, αx =18 (meets the Meek criteria for streamer formation discussed in the following section). When αx ≥14, the radial
8 growth of the electron repulsion drift exceeds the diffusion effect, accounting for this gives the following expression for the avalanche radius:
3e ⎛αx ⎞ 3 Ea r = 3 exp⎜ ⎟ = (1-13) 4πε 0αE0 ⎝ 3 ⎠ α E0
This speed of the transverse avalanche size restricts the electron density to its maximum value as seen below.
ε αE n = 0 0 (1-14) e e
The broadening of the transverse avalanche slows down rapidly once it reaches the characteristic ionization length, where the induced electric field is approximately the same as the applied external field (equation 1-13).
The avalanche will eventually reach the anode, sinking into the electrode, and leave behind its mostly ionic tail in the discharge gap. As there are no or few electrons present in the gap, the total electric field is comprised of the applied external field and an ionic charge image in the anode (see Figure 1.3). The total electric field near the anode is less than the applied external field, however further from the anode it will exceed the applied external field. The total electric field in air
(atmospheric pressures) will be a maximum at the characteristic ionization distance,
1/α ≈ .1 cm from the anode.
9
Figure 1.3: Electric field distribution as avalanche reaches anode, dipole induced field, E’, and applied electric field, E0 are shown (left), and combined field, E, (right).
Streamer Theory
The breakdown mechanism of the transient plasma is based on streamer theory, which is modified version of Townshend breakdown theory. The breakdown mechanism of the transient plasma is based on streamer theory originally developed by Loeb5, Raether,6 Meek and Craggs.7 A highly simplified version of streamer theory begins with a single electron in an electric field. In the electric field the electron gains enough energy to ionize an atom or molecule, resulting in two (or more) free electrons, which in turn are accelerated and ionize more atoms (or molecules) resulting in an electron avalanche. An avalanche to streamer transition will take place when the internal field of the avalanche becomes comparable to that of the external applied field, meaning αd is sufficiently large.
10 The discharge gap for transient plasma ignition applications is relatively small, and there is a significant overvoltage present, this leads to formation of cathode-directed streamers, in which the streamer propagates from the anode towards the cathode. A primary electron avalanche creates a positively charged trail that the streamer will follow back across the gap towards the cathode. Secondary avalanches are produced by photons close to this trail. The electrons produced are pulled into the trail by the field. The secondary electrons produced intermix with the primary avalanche ions to form a quasi-neutral plasma (streamer channel).
Additionally they excite atoms so that new photons are emitted. The secondary electrons enhance the positive charge at the tip of the streamer head, thus attracting electrons and initiating another wave of secondary avalanches. This is how the streamer propagates across the gap, followed by its thin quasi-neutral plasma channel
(Figure 1.4).
Figure 1.4: Cathode directed streamer propagation (left), enhanced electric field near streamer head (right). 11
The streamer can be though of as a thin conducting needle where electric field at the tip of the streamer is quite large expediting streamer propagation towards the cathode. The electron drift velocities near the streamer head (high field) are on the order of 108 cm/sec, which is an order of magnitude greater than typical electron velocities during the initial avalanche (107 cm/sec). The streamer channel is approximately .01 to .1 cm, which corresponds to the maximum size of the avalanche head (~ ionization length 1/α). The plasma density in the streamer channel is ~ 1012 to 1013 cm-3, which matches the maximum electron concentration in the head of the primary avalanche. There are also a large number of secondary avalanches resulting from photons that are emitted during the ionization process.8
Once the streamers bridge the gap between the anode and the cathode, the plasma channel will become highly conductive and an arc will form. The transient plasma
(pulse corona discharge) that is used for our combustion application is short enough
(<100 ns) such that it will not allow an arc to form.
In order for a streamer to form, the space charge induced field of the avalanche, Ea, must be on the order of the applied external field, E0. Mathematically this is means:
e ⎡ ⎛ E ⎞ ⎤ E = exp α⎜ 0 ⎟* x ≈ E (1-15) a 2 ⎢ ⎜ ⎟ ⎥ 0 4πε 0 rA ⎣ ⎝ p ⎠ ⎦
If the avalanche head radius, rA, is assumed to be the ionization length (1/α), with a gap d between the electrodes the requirement for streamer formation can be found:
12 ⎛ E ⎞ 4πε E α⎜ 0 ⎟ * d = ln 0 0 ≈ 20 (1-16) ⎜ ⎟ 2 ⎝ p ⎠ eα
8 Ne = exp (αd) ≈ 3·10 (1-17)
Which gives the Meek breakdown condition, αd ≥ 20.
In electronegative gasses, the electron attachment properties of the gas will slow the streamer down, and can greatly increase the needed electric field for streamer formation. To account for this the ionization coefficient, a, in equation (1-
16) is replaced by α – β. However, in air, with gaps less than 15 cm, the electric fields required by the Meek criteria are relatively high meaning α >> β, so attachment can be neglected.
Figure 1.5: The streamer head (left), steamer channels (center), and an optically attenuated arc (right) are all seen above. Images were taken using an ICCD camera with a .5ns exposure. Unpublished work by C. Cathey, F. Wang, and Y. Sun.
Observe in Figure 1.5 the different zones of the electric discharge. The main production of radicals from the streamers takes place in the streamer head (hundreds of Td) where electron mean energies can be 5-10 eV and up to 15 eV.9 Note that once the streamers bridge the gap the faint plasma channels that trail the streamer head become highly conductive which is seen in the increase in their brightness. The
13 electron density in the channel is determined by the rate of electron-ion recombination, and electron attachment and adds only heats the gas by 10s of K and thus contribute little to radical production. Another thing to note is the marked difference in the discharge volume between the arc and the streamer process. While the streamers are not spatially homogenous as in a low pressure glow discharge, the electrode geometry can be structured in such a way to ensure quasi-homogenous streamer production over the entire discharge volume. It is interesting to note that as the pressure is reduced, the streamer head increases in size, and branching reduces, which will eventually transition to a glow like discharge.
It should be mentioned that the times scale for the two processes are significantly different. The streamers occur on the nanosecond time scale, whereas the arc is a much longer process occurring over hundreds of microseconds to milliseconds. This is useful in that formation of the transient plasma discharge and its products can be easily separated in time from the combustion kinetics.
Corona and Pulse Corona Discharge
At high pressures, the corona discharge is the commonly observed non- equilibrated plasma. Typically it is a low power weakly luminous discharge that appears on sharp points, or edges where field enhancement is occurring. Thus it is a spatially non-uniform discharge (apposed to a glow discharge) where a strong electric field, luminosity, and ionization occur all near one electrode. Physical examples you may be familiar with are the buzzing you hear around high voltage transmission lines, and “Saint Elmo’s Fire” that occurs on the masts of ships.
14 In order to get around the low power characteristic of the corona, but still retain its ability to generate radicals throughout a large volume without significant heating of the gas; a pulsed corona methodology is used. This allows high power discharges and increases the number of radicals produced. The total time for avalanche development, avalanche to streamer development, and streamer propagation between the electrodes is typically 100-300 ns. By using short ( < 100 ns) high voltage pulses we are able to effectively couple the energy into a non- equilibrated plasma without a transition from streamers to sparks. The pulse generator methodology that used for this is outlined in detail in Chapter 3.
As mentioned previously corona generation generally requires a non-uniform electric field. This can be produced using a coaxial geometry, which his the geometry that is used in all of my work. In a coaxial geometry there are two electrodes, the outer wall, and one located centrally. Typically for safety reasons the center electrode is the anode and the outer shell is cathode. The cylindrical geometry was chosen largely for ease of installation on the PDEs which will be discussed in subsequent sections. For this simple geometry the electric field prior to breakdown is described in Figure 1.5.10
V E = ⎛ R ⎞ x ln⎜ ⎟ ⎝ r ⎠
Figure 1.6: Where x is the radial distance from the center axis, r is the inner electrode radius, and R is the gap distance.
15 This equation depicted in Figure 1.6 gives the electric field prior to the plasma disturbing it. It is apparent that the maximum electric field occurs when x equals the inner electrode radius, r. The effective generation of particles mainly takes place in the active corona volume, which for most applications is typically near the electrode that sees the high electric field. In our situation this is the anode (center electrode). This is an important area in that most of the excitation and chemical reactions take place in this zone. It should be noted that our typical electrode is a threaded rod, so there is significant field enhancement above and beyond the smooth anode in the coaxial geometry case.
Electric field nonuniformity facilitates the production of streamers. A nonuniformity in the field will reduce the breakdown voltage for a given distance between electrodes. The breakdown condition outlined in equation (1-17) actually does not require αd, but the integral:
x2 N = α(E)dx (1-18) e ∫ x1
A(E) has a strong exponential dependence (see equation (1-4)), thus there is a significantly larger values of the integral using a nonuniform field than a uniform field.
In a typical continuous corona discharge this volume is quite small, and is characterized as having low currents and a very low discharge power. However, by over-volting the gap in a pulsed mode the size of the active volume is increased.
This is due to the fact that the streamer head moves across the space between electrodes, and scales in size and energy with the electric field. Thus by pulsing the 16 corona you are able to extend the active volume such that it will fill the discharge gap. In addition, using a pulsed corona allows you to significantly over-volt the system without breakdown, increasing the power of the discharge. It should be noted that in this regime where streamers are bridging the gap, a pulse that is too long will cause a spark channel to form, and the pulsed corona discharge will collapse into an arc. Thus, pulsed power must be matched to the discharge conditions.
17 Chapter 1 Endnotes
1 R. Hippler, S. Pfau, K. H. Schoenbach (Eds), Low Temperature Plasma Physics, Wiley-VCH, New York, 2001).
2 A. Fridman, and L. A. Kennedy, Plasma Physics and Engineering, Taylor and Francis, New York, 2004.
3 Y. P. Razier, Gas Discharge Physics, 1991.
4 Y. P. Razier, Gas Discharge Physics, 1991.
5 L. B. Loeb, Basic Processes of Gaseous Electronics, University of California Press: Berkeley (1960).
6 H. Raether, Electron Avalanches and Breakdown in Gases, Butterworth & Co: London (1964).
7 J. M. Meek, J. D. Craggs, Electrical Breakdown of Gases, Wiley: New York (1978).
8 Y. Raizer, Gas Discharge Physics, Springer: New York, 1991.
9 E M van Veldhuizen and W R Rutgers, “Pulsed positive corona streamer propagation and branching,” J. Phys. D: Appl. Phys. 35 (2002).
10 A. Fridman, and L. A. Kennedy, Plasma Physics and Engineering, Taylor and Francis, New York, 2004.
18 Chapter 2 Combustion Theory & Applications
Introduction
Ignition is one of the most promising applications for transient plasmas. The time scales involved (sub 100 ns) make characterization of the streamers difficult.
However, one large advantage of the transient nature of the streamers is that the plasma kinetics can be treated separately from the combustion kinetics which occurs over microsecond to millisecond time scales. Thus allowing the researcher to treat the two events independently, this is quite unlike a spark which often will last several milliseconds or more, making it difficult to decouple the effects of the plasma from combustion kinetics.
It is important to understand and refine the theoretical basis for transient plasma ignition. Reductions in ignition delay, lean burn capability, reductions in soot production, and the ability to ignite higher mass flow rates are all effects that have been demonstrated. Understanding the physics of the process will allow the researcher to better tailor TPI to a particular application. The following material will outline the general theory of TPI, discuss experimental results, and the practical aspects necessary to successfully field a TPI system.
Transient plasma Combustion Physics
So why is the production of radicals important to ignition, and reducing ignition delay? Using non-equilibrium plasmas allows the researcher some control over the electron energies, such that they can change the energy branching in the plasma from heating and excitation of low-energy vibrational/rotational levels, to
19 excitation of highly energetic electronic states, and ionization. Key to combustion of hydrogen and hydrocarbons are the formation of free radicals, O•, OH•, and H•.
Combustion chemistry is quite complex, however, by using a simplified model for hydrogen-oxygen the importance of the free radicals can be illustrated. Traditionally the chain is initiated via reaction (1), where the key product is H•.
H2 + O2 → H• + HO2•, (chain initiation) (2-1)
The H• now attacks the O2 to produce two radicals, O•, and OH•.
H• + O2 → O• + OH•, (chain branching) (2-2)
The products now interact with fuel in equations (2-3), and (2-4).
O• + H2 → OH• + H•, (chain branching) (2-3)
OH• + H2 → H2O + H•, (chain propagation) (2-4)
Now counting radicals produced from a single cycle through the chain you see 3 H• atoms are formed. Looking carefully at the reactions it is seen that (2-2), and (2-3) generate 2 OH• molecules who in turn form 2 H• atoms via (3), where the third H• atom is also formed. These 3 H• atoms now go through the cycle again and produce 9 H• atoms, then 27 H• atoms, then 81 H• atoms, and so on, leading to exponential growth of the radical pool size and hence the reaction rate.1 Both chain branching and chain propagation reactions are important in increasing the number of free radicals and atoms in the system. Additionally they are fast reactions and they do not lead to the re-combination of the radicals which sustains the chain mechanism of combustion. Another important point is that each cycle produces one molecule of 20 water which releases heat into the system. As the branching increases exponentially the rate of heat release also increases exponentially, and will eventually surpass the rate of heat removal resulting in an explosion. While not all combustion reactions are chain branching, even possessing a small number of chain branching reactions can have a significant result on the time it takes for combustion to occur.
Now understanding the exponential growth of free radicals in a system, what would happen if the number of radicals initially present was increased greatly? For example if the initiation reaction (2-1) is replaced with the electron impact dissociation reactions of (2-5), and (2-6). For a hydrogen molecule this occurs via
3 + 3 + b Σu , and a Σg electronically excited states, and for the oxygen molecule it occurs
3 + 3 - via A Σu and B Σu , with thresholds of ~7 eV. Note the electron impact dissociation cross section increases with vibrational level, thus the threshold decreases for vibrationally excited molecules.2 This gives credence to the possibility of multi step ionization in addition to single electron ionization.
H2 + e → H• + H• + e, (chain initiation) (2-5)
O2 + e → O• + O• + e, (chain initiation) (2-6)
Reactions (2-5) and (2-6) are examples of what is believed to occur when a non-equilibrated plasma in the form of a fast ionization wave or a transient plasma is applied to the mixture. The reduced electric field, E/N, of the discharge determines the mean electron energy. With a reduced electric field of about 100-300 Td
(Townsend, 1 Td = .33V/ (cm – Torr) ), the reaction rate constant is several orders of magnitude larger than of reaction rate constant in (1). These conditions facilitate 21 electronic excitation and ionization in the mixture. Figure 2.1 demonstrates the differences in reaction rate constants versus temperature.3 The large difference in reaction rate values implies that initiating the reaction via electron impact is a more efficient and faster process than via traditional thermal (arc) methods.
Figure 2.1: Autoignition reaction rate constants, the horizontal lines are produced via electron impact at 100 Td, and 200 Td.
There is a lot of ongoing work in trying to model both the non-thermal plasma and its effects on the combustion process.4, 5 Preliminary work has shown that short high electric field conditions can have a significant effect on the electron energy distribution. Figure 2.2 depicts a simulation done in Kinema showing a highly non-maxwellian electron distribution under short, high field conditions. What is key is that that as the electric field is applied the probability of electrons in energies around 10 – 15 eV or more increases greatly. This is important in that at these energy levels the probability of molecular dissociation increases significantly.
While this simulation is not a true representation of a streamer it does lend support to 22 the idea that a large pool of radicals is created throughout the discharge volume which will then drive the combustion process.
Figure 2.2: After applying a high voltage (E/N = 410×-15 Vcm2) the electron energy distribution function is highly non-Maxwellian for a few nanoseconds. 300K, Equivalence ratio = 1.0, 50 species including electron, neutrals, radicals, positive and negative ions. Nearly 200 reactions including electron impact dissociation, excitation, ionization, chemical reactions for neutrals, radicals, and ions. Simulation by Chunqi Jiang using Kinema.
The radical density is seen to rise sharply during the high field process.
Consider a methane – oxygen fuel mixture, one possible reaction pathway is high energy electrons dissociate gasses through electron impact.
1 3 O2 + e → O( D) + O( P) + e (2-7)
CH4 + e → CH3 + H + e (2-8)
OH is then generated via rapid chemical reactions, one of which is listed below.
1 -10 3 CH4 + O( D) → CH3 + OH (Rate coefficient, kf = 1.4×10 cm /s) (2-9)
The geometry of the discharge area is such that ignition is achieved with a high degree of spatial uniformity over a large volume relative to traditional spark 23 ignition. The short timescale of the pulse ( < 100 ns) prevents formation of an arc, and a voluminous array of streamers is used for ignition. Chapter 3 outlines a series of experiments to prove experimentally that energetic electrons in the highly non- equilibrated electron energy distribution of the streamers cause dissociation of hydrocarbon chain molecules, producing active radicals throughout the ignition volume. This leads to the possibly of multiple ignition kernels in the discharge volume. Contrary to what is expected it was found that ignition occurred only along the anode. So volumetric ignition does occur, however, it is not over the entire plasma discharge volume.
In theory a further advantage of TPI is that a smaller fraction of the electrical energy goes into thermal heating of the mixture. This is because the electric pulse is short enough such that the ions will not have time to respond to the field. Thus the gas temperature should be cold relative to the electron temperature and change little when pulsed. However if the transient plasma heats the gas significantly this is definitely an indication of a thermally based ignition. Thermal effects are important because chemical reaction rates in general increase exponentially as a function of temperature (Arrhenius equation); a good rule of thumb is the reaction rate will double for every 10 degree Celsius increase in temperature.
Figure 2.3 illustrates the spatial differences between the spark and the transient plasma discharge. Observe that the discharge volume for the transient plasma is easily tailorable, and occurs over a much larger volume.
24
Figure 2.3: Transient plasma discharge (left), and spark discharge (right) to a metal mesh from a central anode. Observe the coaxial geometry allows for a voluminous array of streamers over the anode length.
Traditionally ignition (traditional spark) relies on a thermal plasma in the form of an electric arc to ignite the fuel/air mixture. Where the predominate ignition mechanism is twofold; in the forms of local joule heating and local radical production. The arc locally is able to create the radicals necessary to breakdown the fuel and its extremely high temperature is able to help accelerate the chemical reactions to ensure that combustion is achieved. As you can now surmise if we are correct about the energies in the streamer head, the ignition methodology for transient plasma is quite different from a spark and is an attractive approach to many combustion applications.
Pulse Length Effects
The pulse length and applied voltage for a transient plasma are ultimately determined by the pressure and gap distance. However, once breakdown conditions have been determined using a long pulse, a shorter pulse can be used in conjunction with a larger applied field. While this can increase insulation requirements, it has the added advantage of increasing the reduced electric field of the system. Figure 2.4
25 depicts ignition delay as a function of reduced electric field. Most effective gas excitation, for ignition initiation, is in the range of E/n~300Td, which coincides with the E/n number at streamer head.
Figure 2.4: Ignition delay as a function of reduced electric field.6
The short pulse allows an increase in the applied electric field which will positively affect several related plasma parameters that control the rate of radical production. The electron drift velocity, vd, and the mean electron energies both have strong dependences on the applied electric field.
− eE vd = = µe E (2-10) mvm
3π eEl e E ε = ≈ (2-11) 4 2m 2m n σ M tr M
26 Where vm is the effective collision frequency for momentum transfer, l is the mean path length (has an inverse relation to n), m is the electron mass, and M is the ion mass. An increase in E will further skew the mean electron energy to higher levels, thus increasing the probability of electron impact dissociation reactions, and ionization.
(a)
27
Figure 2.5: Reduced electric field, E/n, as a function of distance from a smooth anode in a coaxial geometry with a gap of 50 mm at p = 1 atm (a), and p = 10 atm (b).
Figure 2.5 depicts possible reduced electric field conditions as a function of pressure with a constant gap of 50 mm. A coaxial geometry is used and is modeled after the combustion chamber used in later parts of this thesis. The anode is assumed to be smooth for these calculations, however, for the experiments described in subsequent chapters a threaded rod is used which further enhances the electric field near the anode. It is apparent that there is large benefit of increasing the applied electric field, which can be accomplished by going to a shorter pulse. Under the conditions depicted a typical PDE with a 75ns, 60 kV pulse will operate at 1 atm, giving an E/n of ~ 400 Td near the anode. By utilizing a shorter pulse higher overvoltages can be achieved without arc formation, which helps maintain high reduced electric field values near the anode. At the operating voltage and pulse 28 lengths used and described in this paper, the high electric field in the streamer head
(~300 Td) is the primary source of radicals. The shorter pulse can enable a larger applied electric field, and potentially can result in a Fast Ionization Wave (commonly found in low pressure systems), where the electrons in the primary avalanche are the primary radical producing source.
Transient Plasma Decomposition of Fuels
One use of transient plasmas or pulsed coronas is to break or crack hydrocarbon molecules into other stable structures that are potentially more reactive and burn easier than the original molecule. Generally this is done via a high repetition rate dielectric barrier discharge7 to keep the power requirements down, however, the mechanism for radical creation and cracking the fuel is the same as what is postulated for combustion reactions ignited via transient plasma (dissociation via electron impact). Initially it was though that the decomposition of the fuel into other hydrocarbons was a key step in the combustion process, however, the following experiment disproved this hypothesis.
In an attempt to characterize transient plasma’s ability to decompose fuel an experiment was performed with Aamir Abid, in Professor Hai Wang’s lab using a gas chromatograph (GC). A line-type pseudospark switched pulse generator charged by a Glassman HV DC supply was used to create a 55ns, 60 kV electrical pulse. The pulse was used to create a transient plasma in 101.6 mm ID, 203.2 mm long cylindrical stainless steel chamber. The anode was a 8-32, 75mm long threaded rod.
The chamber contained a methane – helium mixture at ambient temperature and a 29 pressure of 2 atm. Once the plasma was applied to the mixture, it was cycled through the GC to determine how much, if any, decomposition occurred.
Additionally a stainless steel insert was used to reduce the inner diameter of the chamber to 75 mm. This was done to determine if increasing the reduced electric field would increase the percentage of fuel decomposition. The experimental setup is seen in Figure 2.6.
Figure 2.6: Experimental CH4 decomposition setup using a gas chromatograph.
The results seen in figure 2.7 (a) – (d) indicate that transient plasma does not decompose the methane into other hydrocarbons; at least greater than 1 ppm (lower range of sensitivity of gas chromatograph detector). This result refutes the idea that a single transient plasma pulse is able to decompose the fuel into stable hydrocarbons that burn faster than the original fuel. Additionally there was no effect seen when the reduced electric field was increased by adding the insert to reduce the discharge gap. It is important to note that this experiment is not able to measure the amount of 30 radicals produced by the transient plasma (due to their transient nature), and is only able to see stable molecules. In order to see radicals another diagnostic is necessary,
Chapter 4 outlines a planar laser induced fluorescence experiment to look at OH production induced by the transient plasma.
300 300 TRACE GC-Channel 2 TRACE GC-Channel 2
CH4_pure_092006_1.dat Pure_CH4_092106_1.dat Area Area 250 250 CH4
200 200 ts l lts o o 150 iv 150 lliv i ill M M
100 100 CO2
50 50
0 0 0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24 Minutes Minutes (a) Pure CH4 trace –no discharge (b). Pure CH4 trace –corona discharge
300 300 TRACE GC-Channel 2 TRACE GC-Channel 2
CH4_sparkplug_092006_1.dat Pure_CH4_insert_092106_1.dat Area Area 250 250
200 200 lts o 150
iv 150 ill Millivolts M
100 100
50 50
0 0 0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24 Minutes Minutes (c)Spark-plug discharge –CH4 trace (d)Pure CH4 trace – corona discharge, with insert Figure 2.7: GC trace of a methane – helium mixture, exposed to a transient plasma. Figures provided by Aamir Abid.
Automobile Internal Combustion Engines
It is well understood in the combustion community that combusting under a
8 lean condition will result in higher thermal efficiency and reduced NOx production.
In an ideal internal combustion engine constant volume combustion occurs at the minimum cylinder volume under a high compression ratio. With spark ignition lean 31 combustion does not occur rapidly enough to approach this ideal constant volume cycle. A way to get around this condition is to use a multipoint ignition source, so ignition occurs over a volume such as transient plasma ignition can provide. In addition a voluminous ignition source may have the potential to reduce flame turbulence during the ignition process. There is a requirement for a certain amount of turbulence during the ignition process to accelerate flame mixing and burning rates, at the expense of heat loss to the walls. A spatially extensive ignition method may be able to reduce the turbulence requirement, thereby increasing the thermal efficiency of the system. Both the lean burn capability and the volumetric effect of the transient plasma gives it an inherent advantage over spark ignition, in particular, the reduction of NOx production. By combusting fuel in lean conditions NOx production is reduced, which is important as NOx is an important factor in the production of smog.
There are several key differences in the operating conditions of an automobile internal combustion engine (AICE) versus a PDE. The first is that the pressure in the combustion chamber. In an AICE the pressure generally peaks at 10 atm, whereas in a PDE the pressure is generally between 1 - 1.5 atm. The other major difference is the discharge gap length. The gap in an AICE is on the order of
~5 mm (assuming a geometry similar to figure 2.9), whereas the gap for a PDE varies from about 25 – 50 mm. Under these small gap, high pressure conditions a pulse length of 75 ns will have a tendency to arc over or form a surface discharge.
The surface discharge is especially undesirable in that it will result in selective heating of the ceramic which may cause it to fracture. 32 It should be noted that the high pressures in automobile gas engines there will be a significant amount of branching. This is indicative of the reduction in the mean free path of the electrons. The increase in the number of collisions means that the effective electron energy will decrease relative to the same pulse at lower pressures.
For transient plasma to be effective the electron energy must be such that it can effectively dissociate the molecules in the mixture. In order to help mitigate the lower electron energies a higher field needs to be applied without the gap breaking down into an arc or equivalently increase the reduced electric field, E/n.
Nissan Experiment
A collaborative effort between USC and Nissan research laboratories used both a long pulse ( 85 ns ) and a short pulse ( 20 ns ) in a single cylinder iso-octane engine. Our experimental setup (Figure 2.8) consisted of a Glassman HV DC supply charging the pulse generator used to generate the transient plasma. The 85 ns, 44kV,
60 mJ pulse is generated by a pseudospark switched, lumped element Blumlein pulse generator. The 20 ns, 52 kv, 57.2 mJ pulse is generated by a solid state opening switch pulse generator (Chapter 3). A complete description of this experiment can be found in Cathey, et. al.9
33
Figure 2.8: Nissan experimental setup, testing transient plasma ignition in a single cylinder engine burning iso-octane.
The spark plug (80 mJ) for a single cylinder gasoline engine was replaced with the electrode pictured in Figure 2.9 and is based off a coaxial design to maximize the discharge volume and sits in a small recessed chamber in the cylinder.
For this experiment the pressure in the engine at top dead center was typically around 10 atm and was adjusted to a maximum of 14 atm, with a gap of 6.4 mm.
Several electrodes were used with various discharge volumes, however, the data presented is from a 12 mm long, 2.2M-.45 threaded rod.
Electrode design to meet the conditions (pressure, gap, feed through) is not a trivial process. In order for the electrode to interface with AICE it had to mount to a traditional 14 mm spark plug feed through. In order to achieve this a small
34 cylindrical chamber threaded to fit a 14 mm feed through and tapped to receive a 10 mm electrode was developed. The electrode would essentially sit recessed from the main cylinder in the small volume this created. Additionally the electrode had to maintain a coaxial grounding structure with an OD of ~ 25.4 mm” to feed through the engine head to reach the sparkplug interface.
In a spark plug based ignition system the ground is essentially the engine block, and the HV cabling is unshielded. The lines are also resistive to dampen any ringing in the relatively long pulse (100s of microsecond to milliseconds). In a transient plasma system any resistance (or inductance) in the lines will load down the pulse, additionally the short pulse in relation to the line length of a ~ 1 m, cause them to start to act as antennas, necessitating the need for shielding. Recall that the actual electrode had to fit a 10 mm spark plug interface. The high voltages associated with the transient plasma made porcelain which is traditionally used for spark plugs not feasible. The porcelain plugs developed were significantly overvolted and the insulation would eventually break down. In order to mitigate this handmade plugs out of Shapal were made, which has a higher breakdown voltage ( ~ 70 kV/mm) and is easily machined.
It should be noted that the high pressures in automobile gas engines there will be a significant amount of branching as seen in Figure 2.9. This is indicative of the reduction in the mean free path of the electrons. The increase in the number of collisions reduces the effective electron energy, however, this can be mitigated with an increase in the applied electric field. In order to prevent arc formation at the higher voltage a shorter pulse is needed. The utility of the 85 ns pulse is reduced by 35 the small gap, as it will have a tendency to break down in the form of a surface discharge at voltages over 44 kV. Under these conditions the 85ns, 60 mJ pulse performance was similar to that of a commercial spark plug. In order to help mitigate this higher field needs to be applied without the gap breaking down into an arc. One way to solve this problem is to go to a shorter pulse, in our case a 20 ns pulse.
Figure 2.9: Transinet plasma discharge in air (top), and the electrode schematic (bottom left). The gap is on the order of 6.4 mm. The anode is a 2.2mm diameter .45 thread/mm threaded rod that is 12mm long. Image at bottom right is plasma discharge from electrode taken w/ a ICCD, 700 ns gate, at 10 atm.
Although the effect of initial radical production is not well characterized, it is expected that the higher fraction of energetic electrons during the transient phase
36 allows the researcher some control over the electron energies, such that the energy branching in the plasma from heating and excitation of low-energy vibrational/rotational levels, to excitation of highly energetic electronic states, and ionization will be enhanced at the beginning of the pulse. Key to combustion of hydrocarbons (RHs) are the formation of free radicals, O•, OH•, and H•.
Traditionally the chain is initiated via reaction (2-12), where the key product is H•.
RH + O2 → R• + HO2•, (chain initiation) (2-12)
H• + O2 → O• + OH•, (chain branching) (2-13)
O• + RH → OH• + R•, (chain branching) (2-14)
OH• + RH → H2O + R•, (chain propagation) (2-15)
Looking carefully at the reactions it is seen that there is an exponential growth of the radical pool size and hence the reaction rate. Facilitated by the growth of the free radical pool, the hydrocarbon free radicals undergo further reactions and are eventually oxidized to CO2 and H2O, leading to heat release and a rise in temperature. It is important to note that the reaction rate constants usually have an exponential dependence on temperature via Arrhenius’ Law,
E (− a ) k = Ae RuT (2-16) where Ru is the universal gas constant, Ea is the activation energy, and A is the pre- exponential factor. The extent to which the rate constant depends on temperature is determined by the activation energy, Ea. Transient plasmas generate the initial
37 radical pool via electron impact dissociation, and ionization. The initiation reaction
(2-12) is replaced with the electron impact dissociation reactions of (2-17), and (2-
3 + 3 - 18). For the oxygen molecule this occurs via A Σu and B Σu , with thresholds of ~7 eV. Note the electron impact dissociation cross section increases with vibrational level, thus the threshold decreases for vibrationally excited molecules. This gives credence to the possibility of multi step ionization in addition to single electron ionization.
RH + e → RH• + H• + e, (chain initiation) (2-17)
O2 + e → O• + O• + e, (chain initiation) (2-18)
Reactions (2-17) and (2-18) are examples of what is believed to occur when a non-equilibrated plasma in the form of a fast ionization wave or a transient plasma is applied to the mixture. The mean electron energy, ε , scales with the reduced electric field, E/n. Typical values of the reduced electric field across the streamer head is generally measured in hundreds of Td (Townsend, 1 Td = 10-17 V-cm2). This is important as the reaction rate constant for electron impact dissociation reactions scale with the reduced electric field. Considering the simple case of H2 – O2, with a reduced electric field of about 100-300 Td, the reaction rate constant for (2-18) is of the order of 10-9 cm3/s, whereas for (2-13) the reaction rate constant varies from 10-15
- 10-11 cm3 /s. These conditions facilitate electronic excitation and ionization in the mixture. The large difference in reaction rate values implies that initiating the reaction via electron impact is a more efficient and faster process than via traditional thermal (arc) methods. Another consideration is that in the case of electron impact 38 dissociation the reaction rate constant is dependent on the reduced electric field, and is independent of temperature. This process of electron impact dissociation is thought to occur over the streamer paths, with the highest density of hydrocarbon fragments generated at the point of the highest field, near the anode.
It is well known that operating in a pulse mode will cause the paschen curve to shift, allowing for an increase in the breakdown voltage. By reducing the pulse length from 85 ns to 20 ns, we were able to increase the operating pulse voltage from
44 kv to 52 kV. The 20 ns pulse allowed a substantial increase in the applied electric field while holding off spark breakdown. The higher applied voltage of the 20 ns pulse increased the reduced electric field and thus the number of high energy electrons capable of dissociation and ionization; increasing in number of free radicals that initially seed the discharge volume. Peak pressure achieved with the 20 ns pulse, was 20% greater than peak spark pressure, indicating a larger net heat release
(see figure 2.10).
39
Figure 2.10: Pressure v. crank angle, for the spark, 75 ns pulse, and 20 ns pulse, φ =.72.
Transient plasma ignition in this experiment was found to generate faster flame speeds than those produced by spark ignition. Images taken in 200 µs intervals with a high speed camera showed the transient plasma generated flame speeds to be 200- 400 µs faster than that of a spark. Flame initiates near the electrode for both the 75 ns and the 20 ns pulse which is consistent with the area of the volume experiencing the highest applied field and thus a higher radical density.
Chapter 4 describes work to determine radial densities at the time of ignition and any electrode material or electrode temperature effects on ignition delay.
40 It was found that by using transient plasma ignition that the ignition delay period, which is an index of performance in a conventional gasoline engine, can be improved relative to a spark plug. Additionally stable lean combustion was shown to occur at levels not realizable by spark ignition when a 20 ns pulse is used. The 20 ns pulse was found to perform substantially better in the small gap high pressure conditions of the engine. Additionally it is thought that the electron impact is the primary mechanism for molecular dissociation. The results indicate that transient plasma ignition is a viable ignition methodology with current engine design, and suggest that by going to a shorter pulse and overvolting the gap to a greater extent will further extend stable lean combustion while reducing the total amount of heat lost to the walls.
Transient plasma applications to internal combustion engines are at not as well developed as that of the PDE. The technological challenges are greater for an internal combustion engine, specifically insulation, electrode-engine interface, and pulse generator design. The challenge is that the transient plasma operates at 50-60 kV whereas in a spark ignition system the voltage is typically 25 kV - 40kV. This presents some material problems in selecting an insulator able to hold off the required voltage. The feed through for most car engines are 14mm in diameter, however, in modern cars space is a commodity and the sparkplug interface is shrinking, with some designs calling for a 10 mm feed through. The spark ignition system has been around for a century and is extremely well developed. In order for transient plasma ignition to catch on in the automotive industry size, efficiency,
41 reliability, and interoperability must be optimized to create a viable automotive ignition system.
Pulse Detonation Engines
The aerospace community has a vested interest in the development of propulsion technologies based on PDEs, and work is underway to determine whether this is a feasible technology. The PDE is a descendant of the pulse jet engine that powered the German V-1 rockets during World War II. The key difference between the pulse jet and the PDE is that a PDE will detonate its fuel whereas in a pulse jet it is a deflagration process. The PDE provides impulse through pulsed operation, and through fuel detonation, thrust is provided by a supersonic shock wave that is generated by the exhausting gases. Potential advantages include intrinsic efficiencies, design simplicity (less rotating parts), and operation at both subsonic and supersonic speeds.
In theory a PDE can efficiently operate from Mach 0 to more than Mach 4.
Additionally there seems to be a potential niche for PDEs around Mach 4 where they are in theory a very efficient architecture, more so than the Turbojet whose efficiency decreases sharply for speeds greater than Mach 2 (Figure 2.11).10 Observe that this is the only engine platform that can in theory, start from a dead stop and accelerate to speeds of Mach 4+, while maintaining efficiency. Note that for high speed operation its primary competitor is the RAMJET which needs a booster to reach operational speeds.
42
Figure 2.11: Flight Mach number vs. Specific Impulse for different engine designs.
Transient plasma ignition is attractive as an ignition source for PDEs because it produces reductions in ignition delay and has been shown to provide the capability to ignite under leaner conditions. This allows for high repetition rates, high altitude
11, 12 operation, and reduced NOx emissions. The geometry of the discharge area is such that ignition is achieved with a high degree of spatial uniformity over a large volume relative to traditional spark ignition. Also due to the short timescale of the pulse ( < 100 ns) an arc is never formed and the streamers are used for ignition. The
43 electron energy distribution of the streamers is such that a percentage of the electrons have energies on the order of 10 eV or more (mean energy ~5- 10 eV)13, and therefore are effectively able to disassociate hydrocarbon chain molecules, producing active radicals throughout the ignition volume. Another advantage of TPI is that electrons carry most of the energy, and only a small portion of the energy goes little of the energy going into thermal heating of the mixture. These effects allow for large numbers of active species to be generated throughout the volume with the end state being reductions in the ignition delay times and possibly direct initiation of a detonation wave if appropriate conditions are met.
A fundamental component of PDE operation is the creation of the detonation wave. Currently there are three ways this normally occurs, via direct initiation, shock induced initiation, and deflagration to detonation transition (DDT) process.
DDT is the methodology employed in the following discussion of PDE performance with a transient plasma igniter. The DDT process begins with ignition and flame propagation, the onset of flame wrinkling and turbulence will cause a large increase in the burning rate. The increased burning rate creates an increase in the flow velocity ahead of the mixture due to the expanding gases. The unsteady compression waves ahead of the mixture will result in an increase in temperature and will accelerate reaction rates. When the compression waves coalesce a shock front is formed, and the onset of detonation occurs when there is a sudden appearance of explosion centers in the shock front. If detonation successfully occurs it will travel in a relative steady self sustaining wave traveling at the Chapman-Jouguet (CJ) conditions. There has been a lot of work done using this methodology for PDE 44 research, as it requires minimal energy and has been used successfully over a wide range of conditions. It is important to note, that obstacles in the tube are often necessary to successfully create a detonation wave, an example of which is a
Schelkhin spiral. The obstacles increase turbulence which is necessary for the creation of the detonation wave. However the obstacles come at the expense of drag and mass penalties.
Figure 2.12 shows the operating cycle of a simple PDE. (1) A fuel-air mixture is injected into the tube. (2) The mixture expands throughout the tube, and once the mixture has completely filled the tube an ignition event occurs. In (4) and
(5) the flame is initially deflagrating and it will transition to a detonation wave as it progress down and exits the tube. (6) and (7) show a series of rarefaction waves that reduce the pressure inside the tube and combustion products are purged from the tube. In a multi-cycle system this process will repeat.
45
Figure 2.12: PDE cycle of operation.
The duration of the PDE’s operating cycle is generally on the order of 10s of milliseconds. It is important to note that for a PDE to achieve enough thrust to be a viable technology repetition rates of 60 Hz are needed, and 100 Hz or more are desired. What is valuable about transient plasma ignition is that it can directly reduce the time it takes for events 3-5 to occur. It important to note that not only does thrust scale with frequency, it also scales with tube volume.
46 PDE Results
In a PDE thrust scales linearly with repetition rate (and tube volume), thus if ignition delay is reduced by a factor of two, the thrust is potentially doubled. The two major limiting factors for high repetition rate operation in a PDE are ignition delay and the gas exchange time. Through the use of transient plasma as an ignition methodology ignition delay has been substantially reduced (2-9) relative to a traditional spark, making it potentially an enabling technology for multi-cycle PDE operation. The duration of the PDE’s operating cycle is generally on the order of 10s of milliseconds. It is important to note that for a PDE to achieve enough thrust to be a viable technology repetition rates of 60 Hz are needed, and 100 Hz or more are desired. The generation of a large number of radicals over the discharge volume seeds chain branching and propagation reactions such that multipoint ignition rapidly occurs.
Transient plasma ignition is on the cutting edge of ignition methodologies for combustion engines. In particular the PDE community is interested in this technology for several reasons; 1) capability to ignite over a volume, 2) energy is coupled into the mixture more effectively than a traditional spark, 3) extension of the lower flammability limit of mixture, 4) the ability to tailor the ignition volume by adjusting the electrode discharge volume. For these reasons transient plasma has the potential to overcome the traditional capacitive and inductive spark discharge, and laser discharge ignition techniques.
47
Figure 2.13: The left figure shows a valve less PDE setup at the Naval Postgraduate School. This type of architecture requires a booster, and its anticipated applications are missiles or rockets. The right figure shows a factor of 4 reduction in ignition delay for ethylene-air.
In work performed in collaboration with NPS we demonstrated at high flow rates where spark initiated flames are normally extinguished, the transient plasma is able to ignite and effectively create a detonation wave.14 Significant reduction (x4) in ignition delay was shown for C2H4 – air mixtures. Additionally the TPI ignition delay seemed relatively invariant to temperature and was at comparable energy levels with the conventional sparkplug baseline (Figure 2.13). Tests at NPS prior to the introduction of transient plasma as an ignition source were limited to low frequency operation, unless extra oxygen was introduced into the system. While extra oxygen in the lab may be okay, on an airborne platform the extra oxygen adds complexity and danger to the overall system. Additionally this experiment acted as a preliminary study showing that the mass flow rate has little effect on the ignition delay, however, as the mass flow rate increases a decrease in the detonation wave speed is observed. The ignition delay varied from 3.95 ms to 4.25 ms as the mass flow rate varied from .1 to .4 kg/sec. Where the detonation wave speed ranged from 48 1.5 km/sec to just over 1 km/sec, where at 1 km/sec a detonation wave is not achieved. Additionally this experiment tested possible multiple electrode configurations to facilitate pulsing the mixture several times as it transitioned down the tube. It was found that multiple pulses did in fact further reduce ignition delay, however, the drag losses introduced by the secondary electrode made this approach unfeasible. More recently work has been performed at NPS using our igniter and looking for optimal spiral length for DDT reduction.15 This is a preface to work that is currently being done at NPS testing a new staged PDE engine design, where TPI is the ignition source.
Figure 2.14: The left figure shows a valved PDE at Wright Patterson Air Force Base. The valved architecture would be used for an aircraft and would need no booster. The right figure shows a factor of 2 reduction in ignition delay for aviation gasoline.
Tests conducted at Wright Patterson Air Force Base in collaboration with Dr
Fred Schauer’s group were performed over the past two summers. This was the first time TPI was used on a valved PDE, tests were performed initially on 73 inch tubes,
49 with an ID of 2.067.” For this fist test we burned H2-air, and aviation gasoline
(AVGAS)-air mixtures. Shchelkin-like spirals were used for the H2, and for the
AVGAS mixture to promote the DDT process. In hydrocarbon-air mixtures the reduction in ignition delay is of primary importance when at low hypersonic speeds where the ignition delay times can be orders of magnitude larger than the flow residence time. The ignition delay results for AVGAS are depicted in Figure 2.14 where a reduction in the ignition delay by a factor of 2. The lean burn advantages of the transient plasma was also demonstrated. The transient plasma was able to reliably combust AVGAS at equivalence ratios of ~.65, whereas the baseline’s lower limit was φ =.71. The widening of the range of operation into lean side of the curve is advantageous for high altitude engine operation. Also the lean operation allows for a more economical use of the fuel when cruising, and reduced NOx emissions.
The reduced performance of the 66 kV pulse relative to the 59 kV pulse is likely due to small energy losses due to inadequate insulation between the electrode and the engine head. Where at higher pulse voltage small parasitic arcing was occurring from the HV cable/electrode interface to the engine head.
50
Figure 2.15: Ignition delay and DDT time for H2-air mixtures.
For Hydrogen-air mixtures the reduction in ignition delay is important primarily for high hypersonic speeds where the flow residence time is very short.
Hydrogen was also burned during this first experiment, primarily because it is relatively easy to detonate. Figure 2.15 depicts ignition delay and the DDT times under these conditions. Observe that TPI was able to almost halve the ignition delay time for H2. The figure also depicts the DDT times for H2 which seems to be on the order of the baseline with little or no improvement. The DDT times for AVGAS were similar in that TPI seemed to have little effect. Having said that it should be noted that the error spread for the DDT times for H2 and AVGAS were quite large, and efforts are underway to try further process the data to bring in the error. To date there has not been a lot of work done on TPI’s effects on DDT. Looking at the data taken at WP over the past two years, it suggests that it is primarily controlled by fluid dynamics, and may be independent of ignition methodology.
51 A second experiment in collaboration with WP was performed, this time burning AVGAS and C2H4. The main purpose of this experiment was to determine if there was a correlation between discharge volume, and ignition delay time. For this experiment a HV TPI electrode was developed building on the lessons learned from the first experiment. The discharge volume was varied by testing with 8”, 5”, and 3” electrodes. A factor of 2 reduction in ignition delay for AVGAS was confirmed, and an optimal electrode length near 5” for a ~2” ID tube was found. There was potentially significant error introduced in the energy calculations as the engine data was taken at 20 Hz, and there was only V&I curve taken for every complete data set.
Additionally a source of error may be the 3” electrode; which was put together adhoc when the insulator was damaged on the USC electrode during installation on the engine.
The ignition delay and DDT times are seen in Figure 2.16 depicts energy/volume correlation with ignition delay for a C2H4-air mixture. Observe that in general the ignition delay decreases with increasing energy per unit volume. When observing the AVGAS-air data, a similar trend is seen, however, the 3” electrode data for the AVGAS is an outlier. The error comes from two sources, the single V&I curve taken per data set and possibly the jury-rigged electrode used for the 3” data.
So while a definitive trend can not be identified, it does suggest a potential correlation that needs to be further explored. The AVGAS data is placed in the
Appendices because there is some data reduction that still needs to be done. It should be noted that the DDT times (Appendix 1) for the TPI data seems to independent of equivalence ratio, and energy/volume of he discharge. This implies 52 that they are wholly dependent on the fluid dynamics of the engine and not the ignition source.
Figure 2.16: Ignition delay as a function of equivalence ratio for a C2H4-air mixture.
Table 2-1 depicts some of the ignition delay data taken at Stanford
University, the Naval Postgraduate School, and Wright Patterson Air Force Base.
Thrust scales linearly with repetition rate, and it is apparent from the results, tested under a variety of conditions that transient plasma is potentially an enabling technology for high repetition rate operations. Delay to detonation reductions ranging from factors of 2-9 were found, under varying test conditions. Generally hydrocarbon-air mixtures were tested, however, hydrogen-air was also tested. This figure depicts the delivered energy per pulse. While in general TPI deliver mores 53 energy than a conventional sparkplug, it can deliver pulses of comparable energy and still provide significant reductions in ignition delay.16
Table 2.1: Transient plasma ignition of PDE results17
The Naval Postgraduate School and USC at the time of publication are testing transient plasma’s capability to ignite JP-10 in PDEs, which is an excellent example of transient plasma’s advantage over traditional spark ignition. JP-10
18, (C10H16) is a single molecule fuel composed of the exo isomer of tricycle[5.2.1.0].
Due to its highly strained cyclic structure JP-10 is a highly energetic fuel, with a heat value of 39.6 MJ/L (JP-8 for example has a heat value of 34.5 MJ/L); making it attractive as a fuel for both aircraft and missiles. The combustion kinetics of JP-10 are relatively slow, making it very difficult to ignite. The transient plasma was able to ignite the majority of the cycles at a mass flow rate of .1 kg/sec, a gas temperature 54 of 500 K, at 4 atm of pressure, conditions under which the spark was not able to ignite the mixture.
Transient plasma ignition is definitely a viable technology and potentially an enabling technology for PDEs. This is still a relatively new area of research, and there is still a lot that needs to be explained, such as the physical processes behind transient plasma ignition. Transient plasma as an ignition source has a significant impact on the ignition delay times of the system, allowing for higher frequencies of operation. This increase in frequency directly correlates with thrust, and thus solves one of the biggest obstacles in flying PDEs to date at potentially comparable energy cost to the traditional spark.
Transient Plasma Ignition and Soot Reduction
There is a lot of interest in the combustion community to reduce soot formation, particularly ultra-fine particles, which are small enough such that they can transfer from the lungs into the blood and circulate throughout the body. In an attempt to characterize soot formation in mixtures ignited via transient plasmas a combustion particulate experiment was performed with Aamir Abid, in Professor Hai
Wang’s lab with some unexpected results. A line-type pseudospark switched pulse generator charged by a Glassman HV DC supply was used to create a 55ns, 60 kV electrical pulse and ignite the mixture via a transient plasma. Atmospheric ethylene
– air mixtures (C2H4, 19.4% mol fraction, 2.82 L/min; O2, 28.1% mol fraction,4.09
L/min; N2, 52.5% mol fraction, 7.65 L/min) were combusted under a fuel rich
(equivalence ratio = 2.1) condition to ensure soot formation. Immediately after the 55 mixture was ignited valves were opened manually to start the exhaust flow to the
3080 TSI SMPS mobility analyzer. The overpressure created by combusting the ethylene-air mixture was used to drive the process and a sonic nozzle was used to help keep the flow rate constant. The results were compared with a similar mixture ignited via a commercial automobile spark ignition circuit. The experimental setup is seen in Figure 2.17.
Figure 2.17: Soot particulate experimental setup.
It was found that the ethylene mixture ignited via transient plasma produced less substantially less soot than that mixtures ignited via the traditional spark. Figure
2.18 depicts the particle size distribution function found for both the transient plasma and spark ignited mixtures. The transient plasma (listed as corona) produced 56 approximately a factor of 50 less soot than the spark ignited case. It is extremely important to note several sources of error in this experiment. The first is that after the mixture was ignited we would immediately open the valves to start the flow to the mobility analyzer. The problem with this methodology is that the ultra-fine particles immediately start to coalesce into large particles, potentially giving inaccurate readings for the smaller particles. The second source of error results from the constant flow rate that the mobility analyzer requires to obtain its measurements.
The flow was driven by the overpressure created during the combustion process and it was not sufficient to allow the mobility analyzer to scan through an entire range of particles. Data was taken by scanning in one direction for a particular combustion event, and then scanning backwards in a separate shot to create the full particle size distribution.
57 5 108
Spark Ignition 4 108
D 3 108 g lo d / 2 108 dN
1 108
Corona discharge 0 10 100 Mobility Diameter (nm)
Figure 2.18: Particle size distribution functions of soot for an ethylene/oxygen/nitrogen mixture (equivalence ratio = 2.1), comparing corona- discharge and spark-plug ignition.
The problems described above prevent truly clean data from being taken regarding particle size, however, it is sufficient to see that transient plasma ignition does have a significant effect on the raw number of soot particles produced. While further study needs to be performed on transient plasma’s effects on soot production, it is postulated that it produces less soot because it takes less time to combust the mixture, essentially shortening the window in which soot is produced. In order to further explore these results a laser scattering experiment is proposed in Chapter 5.
58
Chapter 2 Endnotes
1 H. Wang, Course Notes AME 599: Combustion Chemistry and Physics, Fall 2006.
2 Cynthia S Trevisan and Jonathan Tennyson , “Calculated rates for the electron impact dissociation of molecular hydrogen, deuterium and tritium,” Plasma Phys. Control. Fusion 44 (2002).
3 S M Starikovskaia, “Plasma Assisted Ignition and Combustion,” J. Phys. D: Appl. Phys. 39 (2006) R265–R299.
4 S. Pancheshnyi, M. Nudnova, A. Starikovskii, Phys. Rev E , 71, (2005).
5 N. L. Aleksandrov, N. B. Anikin, E. M. Bazelyan, D. V. Zatsepin, S. M. Starikovskaia, A. Yu. Starikovskii, 32nd AIAA Plasmadynamics and Lasers Conference and 4th Weakly Ionized Gases Workshop, 2001-2949, (2001).
6 N. L. Aleksandrov, N. B. Anikin, E. M. Bazelyan, D. V. Zatsepin, S. M. Starikovskaia, A. Yu. Starikovskii “ Chemical Reactions and Ignition Initiation in Hydrocarbon-Air Mixtures by High-Voltage Nanosecond Gas Discharge” 32nd AIAA Plasma dynamics and Lasers Conference and 4th Weakly Inized Gases Workshop, June 2001 Anaheim, CA, AIAA paper 2001-2949.
7 L. Rosocha, Y. Kim, G. Anderson, J. Lee, and S. Abbate, “Decomposition of Ethane in Atmospheric-Pressure Dielectric – Barrier Discharges: Experiments,” IEEE Trans. Plasma Sci., vol. 24, bo. 6, 2526-2536, Dec. 2006.
8 J. Liu, F. Wang, L. C. Lee, N. Theiss, P. D. Ronney, and M. A. Gundersen, “Effect of discharge energy and cavity geometry on flame ignition by transient plasma,” 42nd Aerospace Sciences Meeting, 6th Weakly Ionized Gases Workshop, Reno, Nevada 5 - 8 Jan 2004, Paper Number : AIAA-2004-1011.
9 C. Cathey, T. Tang, T. Shiraishi, T. Urushihara, A. Kuthi, and M. A. Gundersen, “Nanosecond Plasma Ignition for Improved Performance of an Internal Combustion Engine,” IEEE Trans on Plasma Sci., submitted.
10 T. R. A. Bussing, T.E. Bratkovich, and J. B. Hinkley Jr, “Practical Implementation of Pulse Detonation Engines, “ AIAA 2748-1997.
11 S. A. Bozhenkov, S. M. Starikovskaya, A. Yu. Starikovskii, Comb. and Flame 133, 133-146 (2003).
12 S.M. Starikovskia, E. N., Kukaev, and A. Yu. Kuksin, Comb. and Flame 139, 177- 187 (2004). 59
13 E M van Veldhuizen andW R Rutgers, “Pulsed positive corona streamer propagation and branching,” J. Phys. D: Appl. Phys. 35 (2002).
14 J. Sinibaldi, J. Rodriguez, B. Chanel, C. Brophy, F. Wang, C. Cathey, and M. A. Gundersen, “Investigation for Transient Plasma for Pulse Detonating Engines,” AIAA 2005-3774.
15 P. Hutcheson, C. Brophy, J. Sinibaldi, C. Cathey, and M. A. Gundersen, “Investigation of Flow Field Properties on Detonation Initiation,” 42nd AIAA/ASME/SAE/ASEE Joint Propulsion Conference 2006, Sacramento, California, 9 -12 July 2006.
16 J. B. Liu, N. Theiss, P. D. Ronney, and M. A. Gundersen, “Minimum ignition energies and burning rates of flames ignited by transient plasma discharges,” 2003 meeting of Western States Section/Combustion Institute, UCLA, Oct 20-21, 2003, Paper 03F-88. 2002.
17 C. Cathey, F. Wang, T. Tang, A. Kuthi, M. A. Gundersen, J. Sinibaldi, C. Brophy, J. Hoke, F. Schauer, J. Corrigan , J. Yu, E. Barbour, and R. Hanson, “Transient Plasma Ignition for Delay Reduction in Pulse Detonation Engines,” 45th AIAA Aerospace Sciences Meeting and Exhibit, Reno, Nevada, 2007, TBP.
18 C. Q. Jiao, C. A. DeJoseph Jr., and A. Carscadden, “Dissociative Ionization of JP- 10 (C10H16) by Electron Impact,” International Journal of Mass Spectrometry, submitted.
60 Chapter 3 Transient Plasma Generation
Introduction
Pressure and gap conditions will generally dictate breakdown conditions.
Combustion experiments possess gaps that are usually less than 5 cm, and are at pressures of 10 atm or less. Streamer velocities are typically on the order of ~108 cm/sec, meaning the time to bridge the gap is less than 100 ns. In order to prevent the streamers from collapsing into an arc, the voltage pulse needs to be less than
100ns, requiring custom built high voltage pulse generators. Creating high voltages with pulse widths of 100 ns or less is quite challenging, and requires careful attention not only in the design of the pulse generator, but also of the diagnostics used.
Switches
The switch is the heart of pulse generator designs, and is particularly important for high voltage, short pulse systems. In general switches can be separated by function (closing v. opening switch), and type (solid, liquid, gas, plasma, etc.).
Switch performance can have large effects on the shape of the voltage pulse. The function of a closing switch is to transition rapidly from an insulator capable of holding off high voltages to a conductor. Then the switch must be able to quickly return to its insulating state. While there are many characteristics used to compare switches, the most critical to high voltage, short pulse generators, are voltage hold- off, current rise time, current density, forward drop, repetition rate, and lifetime. The characteristics of various solid state, gas, and plasma switches are seen in Table 3.1.
61
Max Hold-Off Peak Rise Lifetim Switch Forward di/dt Rep Voltage Current time e Type Drop (V) (A/s) Rate (kV) (kA) (ns) (shots) (Hz) 10 to > Spark Gap 100 1000 20 1012 5 102 -103 106-108 Thyratron 125 20 150 1012 20 4 · 1010 1010 Pseudospar k 100 5 -100 200 1012 30 103 109 Thyristor 1 to 5 1 to 50 2 1010 1000 103 1012 IGBT 7 1 3 5 · 109 350 104 1012 9 6 12 MOSFET 1 0 VDS 5 ·10 <5 10 10
Table 3.1: Summary of various switching characteristics.1
High Power MOSFET
High power metal oxide semiconductor field effect transistors (MOSFETs) are very similar to the MOSFETs introduced in the 70s. The development of the high power MOSFET was in part due to the limitations of bi-polar junction transistors (BJTs) which until relatively recently was standard for power electronics.
A BJT is a current controlled device that requires as much as a fifth of the collector current to remain in the on state, additionally higher reverse base driver currents are required to turn the device off rapidly. Another limitation of BJTs is that current is carried by both holes and electrons, and the long carrier lifetime of the holes reduces the switching speed to a comparable sized power MOSFET (are a majority carrier device). The last major disadvantage of BJTs is that they can be subject to thermal runaway as the forward drop decreases with increasing temperature, which can cause diversion of current to a single device when many are in parallel. This is in contract to a power MOSFET where the forward drop increases with temperature ensuring
62 that the current is evenly distributed among components. A power MOSFET is seen in Figure 3.1. power MOSFETs are generally used for voltages of a kilovolt or less, in applications where fast switching times are needed. For current combustion applications a power MOSFET is not able to hold enough voltage to be practical.
Figure 3.1: Power MOSFET schematic, and device symbol.
IGBT
The insulated gate bipolar transistor (IGBT) essentially combines the advantages of a BJT with those of a MOSFET. IGBTs have the low on-state voltage drops of a BJT while maintaining high blocking voltage capabilities in addition to the fast switching speeds of MOSFETs. IGBTs have a vertical structure as seen in
Figure 3.2, which is quite similar to that of the vertically diffused MOSFET, aside from the p+ layer that acts as the drain of the device. This layer forms a pn junction
(J1) which will inject minority carriers into what would appear to be the drain drift region of a vertical MOSFET. The 4 layers of PNPN, which comprises the PNP
63 transistor and the NPN transistor, form a thyristor structure, which causes the possibility of a latch-up. IGBTs are an excellent solid state switch for low to medium high voltage applications, however, they still have hold off voltage limitations when compared to gas switches. An IGBT is used in the rapid charge outlined later in this chapter.
Figure 3.2: Insulated Gate Bipolar Transistor (IBGT).
Thyristors
The main solid state switch candidates are MOSFETs, IGBTs and thyristors.
What makes thyrsistors attractive is they have voltage hold-off capabilities and are able to conduct large amounts of current. The benefit of using a solid state switch is thought to be long lifetime, assuming it is operated within its specified parameters.
It is important to note that in a solid state switch if breakdown occurs the switch will
64 probably become unusable, whereas in a gas type the switch the insulating medium can repair itself.
A thyristor is a solid state closing switch consisting of four doped regions
(PNPN) such that three interacting pn junctions exist. Thyristors can have relatively high hold-off voltages and conduct large currents when compared to other solid state switches; however, they are much slower than transistors. Although there are several different forms of thyristors the most common is the silicon controlled rectifier
(SCR), which is a gate controlled thyristor.
The SCR pictured in figure 3.3 consists of the four doped regions, P1, N2,
P3, N4, the three junctions, J12, J23, J34, and the anode (A), cathode (K), and the gate
(G).
(a)
65
(b)
Figure 3.3: Silicon Controlled Rectifier (SCR), with doping regions (P1, N2, P3, N4), terminals (anode A, Gate G, cathode K), and junctions (J12, J23, J34) (a), and SCR circuit symbol and circuit characteristics (b).2
In figure 3.3 (b) the circuit characteristics of an SCR are delineated. Where VAK portion of the curve is independent of the gate current (IG), conducting little current until breakdown occurs at VAK = -VBR. However the SCR is typically operated under a forward bias where there is strong dependence on IG. In this regime the SCR conducts little current until breakdown occurs, when VAK > VBF (maximum forward bias blocking voltage). Once VBF is exceeded the device changes from a high impedance blocking mode to a low impedance conducting mode, operating near VAK
= 0. Applying IG will lower VBF, permitting the SCR to enter the conducting regime at voltages less than the applied voltage (VAK).
66 It is important to note that for a PNPN junction to switch from a high impedance to a low impedance mode, it is necessary for the carriers injected from the cathode and anode to cross the adjacent base. The average time it takes for carriers to cross a quasineutral base region is:
W 2 (3-1) 2DB
If W2 and W3 are the widths of the quasineutral regions of the N2 and P3 base regions then the transit times across those regions will be
2 W2 t1 = (3-2) 2DP
2 W3 t2 = (3-3) 2DN
Taking the geometrical mean will give a first order approximation of the triggering time:
W2W3 tON ≈ t1t2 = (3-4) 2 DP DN
What is important about equation (3-4) is that the time is proportional to the widths of the quasineutral regions. Meaning for fast operation these must be very small; however, smaller widths will lower the hold off voltage of the switch, creating a trade off between fast switching and high power handling. Another important characteristic of SCRs is that the current must drop below IH for the device to turn off. Meaning once the SCR turns on, it is locked on and will remain in the conducting mode as along as current is flowing in excess of IH. Additionally di/dt
67 values are limited; if the current rise time is too fast, the injection current may rise to excessive levels near the gate before the injections becomes uniform across the gate.
This can result in localized heating of the device, possibly resulting in thermal failure, or di/dt burnout. Additionally it is possible for premature conduction to occur during high a.c. conduction or if exposed to a high voltage noise spike; this is known as the dv/dt effect.
Gas Switches
Gas switches are one of the most widely used for high voltage, high power applications. In a gas switch the dielectric strength of the gas itself acts as the insulator in the off state. This allows the breakdown voltage to be determined by the gas itself and the gap between electrodes. When the switch is triggered a portion of the gas will become ionized creating plasma (the on state of the switch). Depending on the degree of ionization of the gas the plasma can act almost like a metal. The switch trigger is usually electrical or optical in nature and accomplishes the same basic function: detaching valance electrons in the gas atoms and creates a partially ionized plasma. There are three gas switch types relevant to the research presented here, the thyratron, the pseudospark, and the Gas Spark Gap. The pseudospark switch is used for the experiments outlined in subsequent chapters, and the coverage of it will be more detailed than that of the thyratron and spark gap.
Thyratron A thyratron is a uni-directional closing switch. The switch is typically filled with low pressure Hydrogen or Deuterium, and operates on the left hand side of the
68 Paschen curve. A basic thyratron, consists of the anode, cathode, control grid, gas reservoir, and baffle. A thyratron has a hot cathode that is heated to emit electrons
(thermionic emission) into the neutral plasma formed by the gas. If a sufficiently large electrical pulse is placed across the grid, rapid ionization will occur between the grid and the cathode, then the anode-cathode will break down. The recovery time is highly dependent on when the current drops low enough for deionization of the plasma to occur.
Figure 3.4: Basic thyratron.3
69 Figure 3.4 depicts the configuration of a basic thyratron. The baffle is used to prevent spurious electrons from causing the device to breakdown, and its geometry is such that is prevents a line of signal path between the cathode and the anode. The grid is placed within one mean free electron path of the anode for the tube operating pressure. Whereas the cathode is located multiple mean free paths away from the grid, typically near the Paschen minimum to minimize to reduce the voltage necessary to drive the grid.
The big advantage of the thyratron is its ability to switch large amounts of current relatively fast with little erosion of the cathode, which is a major problem with spark gap switches. A disadvantage of the thyratron is the hot cathode. It adds overhead to pulse generator design, and increases the power consumption of the device.
Pseudospark
A pseudospark switch is a cold (hollow) cathode thyratron for high power applications. The device is encased in a ceramic metal housing, and the copper flanges provide good transport for the power dissipated in the switch. There is also a
Deuterium reservoir in the switch, whose gas pressure is determined by selectively heating the reservoir. Figure 3.5 depicts an Alstom FS2000, and its 5 major components. The anode, hollow cathode, keep alive (glow discharge) input, trigger input, and reservoir heater.
70
Figure 3.5: Alstom Pseudospark switch (left), switch schematic (right).4
There are several key differences in the cathode between a Pseudospark and a thyratron. The first is a Pseudospark uses a cold cathode. The heater for a pseudospark is to control the gas pressure in the device (the Thyratron has a similar device). However, there is no heater for the cathode. Geometrically the cathodes are also different the pseudospark has a hollow cathode see Figure 3.6. Now looking at the two gaps d, and d’, the longer gap will breakdown first. This is because in the small gap the spacing is comparable to the mean free path of an electron. This results in very inefficient impact ionization, also electron loss due to diffusion is more rapid, therefore a higher breakdown voltage is required. Positive ions in the anode cathode region will drift down to the cathode, and build a very thin positive layer at the cathode, creating a virtual anode right at the cathode, and thus a very large electric field. This type of discharge will remain as long as the balance of
71 electron production/loss rate is maintained. If by some external means (say an electrical or optical trigger) extra electrons are produced, a plasma channel will develop and the anode-cathode will breakdown rapidly.
Cathode
d
d’
Anode
Figure 3.6: Hollow cathode structure of a pseudospark switch.
One of the features of this switch is the gas Deuterium reservoir which allows for adjustment of the gas pressure in the switch. The pseudospark operates on the left side of the Paschen curve so by adjusting the heater current you can increase the gas pressure and reduce the breakdown voltage of the switch.
Spark Gaps
Spark gaps are an extremely simple and effective switch that covers an impressive range of conditions. Generally their geometry consists of two electrodes separated by an insulating medium (gas, vacuum, liquid, or solid). Switching occurs by over-volting the gap, or by applying a trigger to a third electrode (usually electric). Common spark gap geometries are seen in figure 3.7, where you have two electrodes separated by a distance, d, and a third triggering electrode. Perhaps the most common is the trigatron configuration where the trigger is a needle isolated 72 from the electrodes. The potential across the switch is just short of the static breakdown voltage. Breakdown of the switch is initiated by a fast rising pulse to the trigger electrode, which is though to produce streamers, and after a delay on the order of 10’s of nanoseconds breakdown of the electrodes occurs.5
Figure 3.7: Field distortion spark gap (left), trigatron spark gap (right).
There has been substantial work done on spark gap development for a variety of applications. The basic components of spark gap design are field geometry, dielectric material, triggering methodology, load characteristics, repetition rate, life time, and applied voltage. It is important to note that like the gas switches, the repetition rate of a spark gap, regardless of the dielectric used, will be determined on how long it takes for a “fresh gap” to re-establish itself. Also the high current densities associated with spark gaps can result in rather rapid electrode erosion, which can limit the lifetime of this switch. 73 Switch Discussion
The previous sections have outlined several different switch architectures. It is useful to compare actual devices in the context of using them for transient plasma generation.
Perkin Elmer makes several Hydrogen thyratrons. One that has a similar hold off voltage to the pseudospark switch is the HY – 3003, which has a maximum hold off voltage of 35 kV. The maximum switched current if 5 kA. This is much lower than the 30 kA that the FS2000 can handle. Another big difference is that the
HY -3003 has a heated cathode, so there will be a continual drop across the cathode, and extra circuitry needed to operate the switch versus the pseudospark. POWEREX makes the FT1500AU-240 thyristor. It is capable of switching 12 kV and 1500
Amps with a rise time of 100 A/us. The solid state switch is not able to hold off as much voltage, switch as much current, or have a faster rise time then the FS2000.
Spark gaps have the ability to obtain faster current rise times, then that of the pseudospark. However, for a given voltage the spark gap is not able to conduct as much current as the pseudospark. One thing to note is that the spark gap is bi- directional whereas the switches mentioned above have a preferred direction of current flow.
The pseudospark is an excellent choice for medium high voltage switching applications. Its high rise time, and ability to conduct large amounts of current with little electrode wear makes it ideal relative to other switching options. When higher voltages ( >100kV ) or very short pulses ( < 50 ns ) are required, sparkgaps start to
74 become the better choice. However, as in any application, what switch you choose will depend on the specific conditions of the experiment.
Magnetic Switching
An alternate to traditional switches for high power applications are magnetic switches, also know as saturable inductors. The use of ferromagnetic materials to produce fast pulses was first proposed by Melville in 1976.6 Essentially by driving a large amount of current through the windings on a magnetic core so that the applied field H will produce sufficient flux density B in excess of the core’s saturation flux.
This will cause the inductance of the windings to change from relatively high value to to a very low value, making the inductor behave like a switch. The change in µ can be in excess of three orders of magnitude, a typical change would be from 1000 to ~ 2 or 3.
The change in flux density when a rectangular voltage pulse, Vp, is applied is given by:
t 1 p V t ∆B = V dt = p p (3-5) ∫ p NA 0 NA
Where A is the core area, N is the number of turns in the winding, and tp is the duration of the applied pulse. If the applied voltage pulse is of sufficient amplitude the core will saturate at time, ts.
∆BAN ()Bsat ± Br AN ts = = (3-6) V p V p
75 Bsat is the saturation flux density, and Br is the remanent (residual) flux density of the core at the beginning of the pulse. Typical saturation values for Bsat for a powder ferrite are 300 to 500 mT, and for iron cores .5 – 2.5 T. It is desirable to set the remanent flux density at the start of the pulse such that ∆B is maximized (position a in figure 3.8), minimizing the amount of core material needed. After each pulse a negative pulse needs to be applied to re-set the core prior to firing the switch again.
Recall that B=µrµ0H, so it is apparent that by moving from a to c results is a large change in the relative permeability, µr. Whereas in going from point b to point c, where the slope is shallow gives a much smaller change in µr. The inductance of a winding around a core is given by:
µ µ N 2 µ µ N 2 L = r 0 = r 0 (3-7) le / Ae C1
Where le is the effective length of the magnetic core, Ae is the effective core area, and
C1 is the core factor. Equation (7) illustrates how the changes in permeability can cause large changes in the inductance of the winding.
76
Figure 3.8: BH loop describing the magnetic switch.7
The magnetic switch does not suffer the erosion of the cathode like gas switches. Additionally they are able to operate at frequencies of 1 kHz or more, and are reliable. The saturation flux density of many materials is temperature dependent.
The main disadvantage is that the core loss induced heating can change the time for the switch to saturate.
Pulse Generator Architectures
Conventional continuous power is relatively low voltage and is delivered to a load slowly and steadily for long periods of time. A typical television may consume
1 million joules of energy over 3 hours, however, the power delivered is likely near
100 watts. In many applications large amounts of energy must be delivered in short 77 amounts of time, necessitating the use of high voltage pulse generators. The combustion applications to date require voltages up to 50-75 ns, 60-90kV pulse into
~300 Ω load. The short pulse width increases the complexity of the pulse generator design & pulse diagnostics, as losses generally increase with increasing frequency.
The pulse generator system in its simplest form consists of four components, a power source, an energy storage device, a pulse forming section, and the load.
Figure 3.9: Pulse generator block diagram.8
The Marx Bank, the inductive adder, the Blumlien, and magnetic compression are pulse generator architectures that are capable of meeting the short pulse, high- voltage requirements need to create transient plasmas.
Marx Bank
Prior to the development of modern pulse generator topologies scientists interested in pulse power were limited to simply discharging capacitors across a load.
Then Erwin Marx developed his “Marx Bank” which is a capacitor based pulse generator with voltage multiplication capabilities. A Marx Bank essentially charges multiple capacitors in parallel and then when triggered, the capacitors are discharged in series through a number of switches, giving a voltage gain that is approximately
78 equal to the number of capacitors stages (N) used (in practice it is generally lower than NV0).
(a)
(b)
Figure 3.10: Idealized three stage Marx bank (a), Marx generator with stray capacitance, and spark gap capacitance (b).
An idealized Marx generator is show in Figure 3.10(a), however, a practical circuit diagram will include stray capacitance (C1, C2, C3, C4, C5, C6), and the capacitance associated with the switch (Cg) as seen in Figure 3.10 (b). When switch
79 S1 closes looking at the stage defined by the nodes, a, b, c, d, the diagram can be reduced to what is shown in Figure 3.11.
Figure 3.11: Reduced stage of a-b-c-d.
The potential at point c is given by:
C g C3 + C4 Vc ≈ Vo −Vo = Vo (3-8) C g + C3 + C4 C g + C3 + C4
And Vcb is given by:
⎡ ⎤ C3 + C4 Vcb = Vc −Vb = Vo ⎢1+ ⎥ (3-9) ⎣⎢ C g + C3 + C4 ⎦⎥
It is apparent from equation (9) that if the capacitance of the switch is zero there will be a doubling of the voltage at this stage. However, this not a practical result and there will always be a reduced voltage in actual cases, 1.5Vo would be a reasonable result. This analysis can be preformed over subsequent stages to determine the final output voltage of the device.
The leading edge of a Marx generator pulse is controlled by the series resistance of the generator and the load capacitor, the falling edge of the pulse can be
80 tailored by the capacitance of the Marx generator and the load resistor. In practice the load capacitance and the capacitance of the Marx generator are typically fixed, leaving adjustment of the pulse to tailoring the series resistance and the load resistance.
Magnetic Compression
Another pulse generator methodology is based on magnetic compression.
Taking a relatively slow pulse and passing it through a series of magnetic compression stages where the pulse is compressed in time while concurrently increasing the pulse amplitude. Using a circuit like that depicted in Figure 3.12 pulse compression is achieved, where saturable inductors are switched successively down the line, which each capacitor in the stage charging and discharging faster than the one in the previous stage. Additionally if the capacitors in each successive stage are smaller than the previous stage, voltage gain can be achieved. The capacitor, C0 is initially charger to V0, and as the switch is closed it is discharge through the inductor, L0, and charges capacitor, C1. As C1 charges a point will be reached where
L2 will saturate, causing C1 to discharge into C2. As the potential rises in C2 a point will be reached where L2 saturates, as long as it has a lower inductance than L1. This will cause C2 to discharge more rapidly into the subsequent stage, and the process will repeat progressively along the line.
81
Figure 3.12: Ideal magnetic pulse compression circuit.
A complete analysis of a magnetic compression circuit is quite difficult due to the nonlinear behavior of its components. However, an approximate analysis was developed by Melville, and later extended by Birx.9, 10
Figure 3.13: Middle stage in a magnetic compression pulse generator.
Looking at one of the middle stages of the compression circuit, as seen in Figure
3.13, the voltage charging the capacitor Cn is given by:
⎛ C ⎞ ⎜ n ⎟ vn (t) = V (0)⎜ ⎟(1− cos(ωt)) (3-10) ⎝ Cn−1 + Cn ⎠
1 where ω = (3-11) L(n−1)sCnCn−1
Cn + Cn−1
82 Letting all the capacitors be equal reduces equations (3-10) and (3-11) to:
V (0) v (t) = (1− cos(ωt)) (3-12) n 2
2 ω = (3-13) L(n−1)sC0
The time, Tn, to charge the capacitor, Cn, is:
L C T = π (n−1)s 0 (3-14) n 2
The compression stages are designed such that the inductor, Ln, will saturate when the capacitor, Cn prior to it reaches peak voltage. The change in flux Bn(t) of the core of the inductor Ln is:
1 Tn V Tn B (t) = v (t)dt = 0 (1− cos(ωt))dt (3-15) n ∫ n ∫ N n An 0 2N n An 0
TnV0 Bn (Tn ) = ∆Bn = = B(sat)n + Brn (3-16) 2N n An
Solving for Tn, which is the time for inductor Ln to saturate gives,
2∆Bn An N n Tn = (3-17) V0
Setting equations (14) and (17) equal to each other gives:
2∆B A N L C n n n = π (n−1)s 0 (3-18) V0 2
2 ⎡2∆Bn An N n ⎤ 2 Which gives, L(n−1)s = ⎢ ⎥ (3-19) ⎣ πV0 ⎦ C0
The saturated inductance of Ln is:
83 2 2 µo µrs N n An µo µrs N n An Lns = = (3-20) l Voln
With µrs is the saturated permeability, and the Voln is the core volume of Ln, the ratio of the saturated inductance for inductors in successive stages is:
L ⎛ C V 2 ⎞ 1 π 2 ns = µ µ ⎜ 0 0 ⎟ (3-21) 0 rs ⎜ ⎟ 2 L(n−1)s ⎝ 2 ⎠Voln 4∆Bn
The term in brackets is the pulse energy, it is apparent that the core volume must scale with pulse energy. The gain, G, of the compressor is the ratio of the charging times of the successive stages, and is given by:
T L G = n = (n−1)s (3-22) Tn+1 Lns
A magnetic compression pulse generator which uses diodes to further sharpen the output pulse, a solid state opening switch (SOS) pulse generator, is seen in Figure 3.14. This pulse generator takes a long relatively low voltage pulse and compresses it though a series of LC resonant stages. 11 What is novel in this architecture is that a diode chain is used to sharpen the final output pulse. Currently this pulse generator is capable of producing a 20 ns pulse of 66 kV. The advantage of this pulse generator is that it can potentially deliver more energy into a smaller gap than the pseudospark pulse generator. An immediate application is internal combustion engines, and smaller gap PDEs. To date our research indicates for small gap systems this type of system can perform better than the pseudospark pulse generator because it can deposit more energy prior to arcing. The caveat to that is due to the solid state pulse generators short pulse width, for a large gap system the
84 output pulse voltage needs to be approximately 1.5 times the pseudospark pulse generator’s output pulse voltage to perform on the same level. This means the insulation requirements can become a limiting factor to making a pulse generator of this architecture a compact ignition system for large gaps.
Resonant Magnetic reactor DOS sharpening Charging
L1 L2 900 - 1.2 K VDC Nanocrystal Nanocrystal C4 41 turns 11 turns 1 nF
Load C1 C2 C3 6 VMI 200 Ohm 1.68 uF 4 nF 4 nF K100UF
Gnd T_1 Gnd T_2 2 Custom Metglas Nanocrystal DSEI1210 IGBT 2:42 3:6 1200V 600A
Figure 3.14: Magnetic pulse compression pulse generator.
Blumlein Pulse Generator
There is always a need for high voltage flat rectangular pulses in many application fields having short duration, fast rise time, and fast energy transfer. A simple and cost effective option is to implement Blumlein line that can meet the high voltage and high current applications, such as lasers, high power microwave generators, and X-ray generation. Essentially, it is two identical transmission lines that are connected in such a way that they are charged in parallel configuration and discharged in series. Using proper termination, it can produce output voltage as high as the charging one. Additionally, the Blumlein line is one of the most popular high power pulsed generator because it is easy to operate and relatively reliable.
85 Higher voltage or current generation can be accomplished by stacking several transmission lines together using proper switching and termination. For high voltage applications, stack the individual lines together in series discharge configuration. For high current applications, stack the transmission lines in parallel discharge configuration.
Blumlein can be constructed either in cylindrical form, parallel plate configuration, or out of capacitors and inductors tailored to replicate a transmission line. Our design uses the coaxial cable geometry due to the easiness in handling and fabrication. The principle of a Blumlein can be summarized as having load impedance, ZLoad, to be twice the value of the characteristic impedance, ZLine, of the transmission lines to be matched. ZLine is the line impedance for each equal section as illustrated in Figure 3.15.
Line Line Impedance Impedance
High Charging RZLine ZLine Voltage
Current VO Limiting Resistor Switch ZLoad = 2ZLine Load Impedance
Figure 3.15: Blumlein Configuration Using Two Transmission Lines
Conceptually, the Blumlein can be viewed as two charged capacitors connected in a parallel configuration via the load impedance. Once the lines are charged to the charging voltage and the switch is closed at time t equals to zero, a negative voltage pulse equals to the switch breakdown is produced at the shorted 86 end. It will propagate along the line towards the load impedance. At time τ, this pulse arrives at the midpoint and the impedance seen by the pulse is the sum of ZLoad and ZLine. Half of the charging voltage is reflected back and a pulse V appears across the load impedance. Additionally, a negative half pulse of V is transmitted to the other section of the line. Both reflected and transmitted pulses are reflected again by the shorted and open ends of the lines respectively, and return to the midpoint in time
2τ. This cancels the initial charging voltage along the entire transmission line. As a result, the output voltage across the load impedance is VO with pulse duration of 2τ.
This is seen below in Figure 3.16.
V Zload
-V
-V/2
-V/2
-V/2
V/2
- V/2 V/2
Figure 3.16: Ideal Blumlein Line Operation
87 The output voltage of a Blumlein line was explained and derived in Pai and
Zhang as:
VO = 2V * ZLoad / (ZLoad + 2ZLine) (3-23)
An ideal flat rectangular pulse will be obtained if the load impedance is matched to the characteristic impedance of the coaxial line (i.e. ZLoad = 2ZLine). From the output voltage equation, this will yield VO = V. If the load is mismatched, pulses are continuously reflected in the lines, resulting in a train of successive pulses of decreasing amplitude across the resistor as shown in Figure 3.17. The maximum voltage possible to get in a Blumlein is twice the charging voltage, across an open load. Furthermore, the output voltage cannot exceed the charging voltage under nominal conditions.
VO
ZLoad = 2ZLine
2τ Time
VO
ZLoad < 2ZLine
2τ Time
VO
ZLoad > 2ZLine
2τ Time
88 Figure 3.17: Ideal (top) and Mismatched (bottom two) Output Pulses of a Blumlein.
The transient plasma is generated by a line type pseudospark switched pulse generator. The pseudospark switched pulse generator is capable of creating a 50-75 ns pulse up to 90 kV.12 The key element in the pulse generator is the pseudospark switch. Which is a gas based cold, hollow cathode switch that is capable of switching 30 kA, with a rise time of 8 kA/ns. The pulse generator is based on the
Blumlein architecture. However, instead of using transmission line, capacitors are used to make the pulse forming network. This results in a minimum pulse that is a critically damped pulse, which will give a voltage amplitude of 64% of the charging voltage across the load, whereas in a traditional Blumlien you are able to get 100% of the charging voltage across the load. The pulse generator can be used in conjunction with a high voltage DC supply or a rapid charger for high repetition rate operation, typically for ignition applications operation is around 100 Hz. Figure 3.18 depicts the pseudospark pulse generator.
Figure 3.18: Pseudospark based pulse generator schematic.
89 Inductive Adder
As solid state switch capabilities have improved they are being incorporated into high voltage modulator designs, one in particular is the inductive adder. This topology is scalable to high voltage, and has demonstrated consistency, high repetition rates, short pulses, and the ability to maintain fast rise and fall times for pulses. These characteristics make it extremely attractive for high voltage kickers for accelerators at SLAC and LNL.13 The inductive adder (Figure 3.19) is a circuit topology that uses multiple relatively low voltage stages that are coupled together via their transformers, where their secondary windings are connected in series. The individual pulse stages are usually driven simultaneously for maximum output pulse voltage. In order to achieve fast pulses (fast rise time) there needs to be as little inductance as possible, which often means single turn windings in both the primary and secondary of the output transformer in each stage.
90
Figure 3.19: Basic 3 stage inductive adder.
On the primary side it is important that the source impedance of the switch and the capacitor bank be quite low (less than one ohm). This will ensure that it will be able to provide the total drive current (consists of the secondary current, additional current loads in the primary, and the magnetization current of the transformer core) without a lot of voltage droop across the pulse.
91
Charging Supplies
The Pulse Forming Networks (PFNs) that are used need to be charged by a
HV supply. If a HV DC supply is used, the pulse generator is charged with a large resistive element in series with the line. The size of the resistor is largely dependent on the switch. If the resistor is too small the switch will not be able to recover and will be locked on by the current coming from the HV DC supply. Additionally when resistive charging is used, the RC time constant will limit the repetition rate to about
60 Hz. This creates a need for a source methodology that is capable of efficiently charging the pulse generator for high repetition rate operation. Lastly it is important to note that an additional loss of wasted energy across the charging resistor is introduced, resulting in a significant drop in “wall plug” efficiency.
A Flyback Resonant Charger (Rapid Charger) can be used to efficiently and rapidly charge the PFN (Figure 3.20). The Rapid Charger has a large on board capacitor bank that is charged via 208 3φ power. The power to be delivered is first stored in the primary winding of the transformer. Using an IGBT as an opening switch causes the transformer to demagnetize through the secondary and the HV diode to the PFN. Using this architecture allows for continuous operation well in excess of 100 Hz. In a 100 pulse burst mode it can reach repetition rates of 2 kHz, and by increasing the size of the capacitor bank this can be scaled upward. In fact the rapid charger’s repetition rate is limited by the Pseudospark switch which is rated at
1 kHz. This device has been field tested at Nissan and on the PDEs at both Wright
Patterson AF Base and the Naval Postgraduate School.
92 35kV Output to PS Generator RG8 Coxial cable 230V, 25A 6 : 300 EDI DIODE RELAY 250, 100W +150 V out Fly-back Tansformer KVP35
NT-15 +15V in 3k, 10W + + GND -15V in 6600uF, 450V
I out 300k, 2W 330k, 2W -150 V out 1000uF, 25V IGBT 0.5 ohm 600V, 30A + V out IGBT 5k 20V + - Gate Drive Board 1400V 800A Fuji 1MB1600NP-140 0.1 AD202 10k 10 uF 600V 1/4W 600V
330 uH JWM 6721
20 A, 3 Phase Circuit Breaker U 3-ph, 20 A 3 Ph, 20 A Rectifier mains filter V 208 V 3 phase FN 355-20/03 W AC plug
Gnd
+15V
AUX. 30 W -15V Supply AUX Power out +5V
GND
Figure 3.20: Rapid Charger Circuit Diagram.
Ignition Electrode Design
Transient plasma electrodes must be able to survive high voltage, and temperature conditions. A typical PDE coaxial transient plasma high voltage electrode (Figure 3.21). The ceramic is MACOR, which is rated at about 1000
Volts/mil, and has a coefficient of thermal expansion relatively close to steel. The outer ground cylinder is steel, and acts as the return path for the discharge current.
The electrode is an 8-32 steel rod, which is threaded for field enhancement, to assist in streamer development. The electrode can be installed or removed from the system without the removal of the PDE tube (improvement over previous TPI electrode designs). Additionally the threaded rod’s length can be adjusted to control the volume of the discharge. In previous designs the electrode had to be installed prior to mounting the tube, and thus any changes to its configuration or need for repair 93 were difficult. Also observe that it is a purely coaxial design, this was done largely to keep the noise down to a manageable level. EMI becomes a very large problem for nearby devices if unshielded connections are used.
Figure 3.21: HV TPI electrode developed to feed through the 14mm interface used in the WP PDE.
94
Chapter 3 Endnotes
1E Schamiloglu, R.J. Barker, M. Gundersen, and A. A. Neuber, “Modern Pulsed Power: Charlie Martin and Beyond,” Proceedings of the IEEE, 92: 1014-1020, July 2004.
2R. Pierret, Semiconductor Device Fundamentals, Addison-Wesley Publishing Company: Massachusetts, 1996, pg 465.
3T. R. Burkes, J. P. Craig, M. O. Hagler, M. Kristiansen, W. M. Portnoy, “A review of high-power switch technology,” IEEE Transactions on Electron Devices, Vol 26, Issue 10, Oct 1979 Page(s):1401 – 1411.
4Alsom FS200 Pseudospark switch data sheet.
5 T. H. Martin, 5thh IEEE Pulsed Power Conference, 1985, 74.
6 W. S. Meville, “The Use of Saturable Inductors as Discharge Devices for Pulse Generators,” Proc. IEE 98 (1951) 185-207.
7 P. Smith, Transient Electronics, John Wiley & Sons: New Jersey, 2002, pg 238.
8 S. T. pai, and Q. Zhang, High Power Pulse Technology, World Scientific: Singapore, 2003, pg 3.
9 W. S. Meville, “The Use of Saturable Inductors as Discharge Devices for Pulse Generators,” Proc. IEE 98 (1951) 185-207.
10 D. L. Birx, E. J. Lauer, L. L. Reginato, J. Schmidt, and M. Smith, “Basic Principles Governing the Design of Magnetic Switches,” Lawrence Livermore Laboratory Report No. UCID – 18832 (1980).
11 T. Tang, A. Kuthi, F. Wang, C. Cathey, and M. A. Gundersen, “Design of 60kV 20ns solid state pulse generator based on magnetic reactor driven diode opening switch,” 27th International Power Modulator Conference 2006, Washington D. C., District of Columbia, May 14-18th, 2006.
12 F. Wang , A. Kuthi, M. A. Gundersen, “Compact High Repetition Rate Pseudospark Pulse Generator,” IEEE Trans. on Plasma Science, 33 , Issue: 4 , Part 1, 1177 – 1181, Aug. 2005.
95
13 E.G. Cook, G. Akana, E.J. Gower, S.A. Hawkins, B.C. Hickman, C.A. Brooksby, R.L. Cassel, J.E. De Lamare, M.N. Nguyen, G.C. Pappas, “Solid-State Modulators for RF and Fast Kickers,” SLAC-PUB-11759.
96 Chapter 4 Transient Plasma Induced OH Production
Introduction
As discussed in Chapters 1 and 2 free radicals are key to initiating and maintaining combustion as they seed many of the chain branching and chain propagation reactions. Even relatively small amounts of free radicals (10-5 – 10-3 of the total number of gas particles in the system) can shift the system equilibrium and initiate combustion.1 Essentially this proposal will determine the spatial distribution and concentrations of OH, and O, created via TPI relative to those created by spark ignition. Of particular interest is the time period between application of the transient plasma and ignition of the mixture. This means that laser absorbance or laser induced fluorescence (LIF) spectroscopic measurements of the species will be needed. LIF type measurements were chosen to facilitate seeing the spatial distribution of the radicals the plasma produces.
The vast majority of the work done looking at active species produced by non-equilibrated plasmas is at much lower pressures (~50 torr), which is especially true when looking to determine an absolute number of generated species. In order to reduce the complexity of the experiment this research only considers the relative concentrations produced by the transient plasma and a traditional spark.
Additionally this work is meant to build upon the past work with PDEs and ICEs, where for a fully realistic experiment parameters such as pressure, mixture temperature, discharge volume, and flow rate need to be matched. Again in order to make this experiment manageable I neglected flow considerations and operating
97 temperature. There are a number of radicals that are interesting to the combustion process, the notable ones being, H, O, OH, and CH3 for a CH4-air reaction. OH is monitored in this experiment because of the relative ease it can be monitored with respect to other possible radicals.
The subsequent sections outline results of OH emission experiment, a planar laser induced fluorescence (PLIF) experiment, and a high speed imaging experiment looking at OH and OH* in a quiescent CH4 – air mixture at STP. The purpose of these experiments are tri-fold; 1) determine spatial distribution of OH in the pre- stages of combustion, 2) determine relative concentrations of OH and OH*, and 3) determine to what extent transient plasma acts as a volumetric ignition source. This experiment shows for the first time graphically how our discharge perturbs the mixture prior to ignition, and to what extent the transient plasma is seeding the discharge volume with radicals.
OH Emission Spectroscopy
Optical emission spectroscopy (OES) is one of the simplest optical diagnostics that can be used to probe elements and select molecules in the visible and ultraviolet. OES takes advantage of the unique band structures of materials and can often be used to identify a particular element or molecule by monitoring emission as an excited state in the atom or molecule relaxes and emits a photon. OES is a powerful tool for monitoring emission from a single source, however, its utility can be diminished when observing emission from non-singular sources where overlapping transitions can cause interference.
98 OH Emission Experiment
It is believed that the non-equilibrium electrons can exist in the head of streamer with mean electron energies in the 5-10 eV range and can reach up to 15 eV.2, 3 If this is true some of the electrons will have enough kinetic energy to
3 effectively dissociate CH4 to H and CH3 radicals, and dissociate O2 to O( P) and
O(1D).
1 3 O2 + e → O( D) + O( P) + e (4-1)
The reaction threshold for equation 4-1 is about 7 eV. The modeling done by
Chungi Jiang (see Figure 2.2) shows that equation (4-1) accounts for 70% of all the
-15 2 processes in a pure O2 mixture at E/n > 10 V/ cm (100 Td). As our E/n value is expected to much higher, we should see a substantial amount of O(1D). It should be
1 noted that at room temperature O( D) reacts with CH4 almost at the collision limit (k
-10 3 3 = 1.4x10 cm /s), as opposed to O( P) that only reacts with CH4 at high temperature
(k = 6x10-19 cm3/s at 293K).
1 -10 3 CH4 + O( D) → CH3• + OH• (Rate coefficient, kf = 1.4×10 cm /s) (4-2)
The CH3 and OH radicals produced via equation 3 are ready to initiate combustion chain reactions depicted below.4
Low-Temperature Mechanism
CH3• + O2 → CH3O2• (chain propagating) (4-3)
CH3O2• + CH4 → CH3O2H + CH3• (4-4)
CH3O2H → CH3O• + OH• (chain branching) (4-5)
CH3O• + O2 → CH2O + HO2• (4-6)
99 CH2O + OH• → H2O + HCO• (4-7)
CH3O2H + OH•→ CH3O2• + H2O (4-8)
HCO• + O2 → CO + HO2• (chain propagating) (4-9)
HO2• + CH4 → H2O2 + CH3• (4-10)
HO2• + CH2O → H2O2 + HCO• (4-11)
OH• → wall (chain terminating) (4-12)
CH3• → wall (4-13)
HO2• → wall (4-14)
1 -11 -11 3 The quenching rates of O( D) by O2 is 4.2x10 and N2 is 2.4x10 cm /s. Observe that they are about an order of magnitude slower than the reaction rate with CH4
-10 1 (1.4×10 ); thus, O( D) has sufficient fraction to react with CH4.
What is interesting about equation (4-1), is that this is a process that does not appear in normal thermal ignition, as the electron energies are not high enough ( <
1eV ) to produce O(1D). It should be noted that positive ions can be formed via electron impact of stable molecules, which then can recombine with electrons to form reactive species. However, the number of electrons with energies of ~13 eV or more is expected to be small, thus the percentage of ionic species formed via an ionization-electron recombination process is also small. A possible conclusion is that the ionic species may not be as important as the species produced in equation (4-
1).
The excited OH*(A) radicals which can be produced by the recombination:
O+H+M → OH(A)*+M (4-15)
100 The emission intensity of OH (A→X) is proportional to the production of O and H.
This experiment was performed in the Pulse Power Laboratory at USC. The relatively simple experimental setup seen in Figure 4.1 using an Acton
Monochromator (resolution .5 nm) containing a fast photomultiplier tuber (rise time
~ 2ns) sensitive from 200-1100 nm as the detector. The cylindrical chamber is 8” long with an ID of 4.” The transient plasma electrode is an 8-32 threaded steel rod, with 3” of exposed length. The window is 1” diameter sapphire. A line type pseudospark switched pulse generator5 charged with a Glassman High Voltage DC supply was used to create a 70ns FWHM, 60 kV pulse responsible for generating the transient plasma that ignites a stoichiometric methane-air mixture initially at ambient pressure and temperature. Spark ignition was accomplished using a conventional car sparkplug (Champion RV17YC6) and a commercially available automobile ignition circuit, which created a discharge of ~ 40 mJ. For spark ignition tests the transient plasma electrode was removed and replaced with the sparkplug, which was put in the center of the plate opposite the main viewing window. Chamber pressure was monitored by an Omega PX-105 piezoresistive pressure transducer with a frequency response of 10 kHz (3 dB at 1.5 kHz).
101
Figure 4.1: Optical emission experimental setup for indirect detection of O(1D).
OH Emission Results and Discussion
The major mechanisms for OH production in both humid air and in CH4-air mixtures are listed below:
TPI initiation: Mechanism in humid air6
1 3 O2 + e → O( D) + O( P) + e (4-16)
e + H2O → H• + OH• + e (4-17)
1 -10 3 -1 7 O( D )+H2O → 2OH• (kf = 3.5 ·10 cm · s ) (4-18)
TPI initiation: Mechanism in CH4-air mixture
1 3 O2 + e → O( D) + O( P) + e (4-19)
1 -10 3 -1 CH4 + O( D) → CH3• + OH• ( kf = 1.4×10 cm · s ) (4-20)
CH4 + e → CH3• + H• +e (4-21)
102 Looking at the OH emission intensity of the discharge in both humid air (figure 4.2) and CH4-air a duel peak structure was found at both 309 and 314 nm.
Figure 4.2: OH emission in humid air.
There are several possibilities as to why there two peaks occur. There is the chance that the observed emission is from several different overlapping transitions, however, as it occurred at both 309 and 311 nm and the resolution of the monochromater was ~ 5 Å, this seems unlikely. Streamers under these conditions will have a velocity of ~1.5-3.3 mm/ns, and the discharge gap is ~ 50 mm, giving a streamer transit time of 15-16 ns. This corresponds to the pulse width of the first peak, implying that the first peak results from the streamers crossing the gap, and
103 the second peak corresponds to the channels starting to conduct large amounts of current.
Figure 4.3 illustrates how the emission intensity scales with applied pulse voltage. Additionally it shoes that the emission signal pulse width remains relatively constant. The quenching rate is on the order of a nanosecond, which means that the emission signal pulse width scales with the applied pulse width.
Figure 4.3: OH Emission (309 nm) signal peak (left) and pulse width (right) as a function of applied pulse voltage.
Figure 4.4 depicts emission intensity after ignition by both a transient plasma and conventional spark ignition. OH emission in flames ignited by the transient plasma was typically a factor of 5-8 greater than in flames ignited by conventional spark discharge.
104 Optical Emission (308 nm) Traces for Spark & Transient Plasma Ignited Stoichiometric CH4-Air Flames 6 s t i 5 n u . b
ar 4
, ) m
n 3
08 TPI Emission 3 (
n Spark Emission
o 2 ssi i
m 1 E cal ti
p 0 O
-1 -20-10 0 102030405060708090100 Time (ms)
Figure 4.4: OH Emission (308 nm) during flame propagation.
Laser Induced Fluorescence (LIF)
Laser induced fluorescence is a useful and sensitive optical diagnostic tool where a laser (pump wavelength) is used to excite an electronic transition in a particular molecule, and after some finite time the spontaneous emission will occur from the excited manifold. The term manifold is used to illustrate that fluorescence may not occur from the pumped state. When fluorescence does occur at the excitation wavelength, it is called resonance fluorescence. In general the fluorescence wavelength is downshifted from the pump wavelength, “Stokes
Shifted.” Collisional and radiative transfer between energy levels is an important consideration for measurements of temperature or concentration. As LIF is an indirect method of measuring the ground state of a molecule (versus emission which
105 measures the excited state) it is important to understand the population dynamics of the system. The cross sections for absorption (NO~ 10-16 cm3) are much larger than
Raman cross section (NO ~10-30 cm3) in which a molecule is perturbed from resonance. This is why the fluorescence is LIF experiments is orders of magnitude higher than Raman Scattering, and why LIF is able to reach such high detection sensitivities (less than ppm). The following derivation follows Eckbreth and
Radziemski et al.8,9
Figure 4.5: Two level energy diagram.
The two level model depicted in Figure 4.5 illustrates the LIF process on its most fundamental level. The major assumption here is that the only two levels that are coupled by the laser radiation undergo any population change. The rate equation or the rate of change in the population of state N2 is given by the following. dN 2 = N ()W + Q − N (W + Q + A ) (4-22) dt 1 12 12 2 21 21 21
106 -3 -1 Where N1 (cm ) is the population of the ground state (level 1), W12 and W21 (s ) are
-1 the laser absorption and stimulated emission rates, Q12 (s ) is the collisional
-1 -1 excitation rate, Q21 (s ) is the collisional quenching rate, and A21 (s ) is the spontaneous emission rate. If we assume that the total number density, NT is constant (N1+N2 =NT) and that the system is in steady state, (d=dN2/dt=dN1/dt=0) , we can obtain the following:
NT ()W12 + Q12 N 2 = (4-22) Q21 + Q12 + A21 +W21 +W12 In most cases the collisional excitation from the ground state to upper states can be neglected when compared to other rates (may not be true for plasmas), so Q12= 0, and we can assume the initial population of N2 is negligible prior to laser excitation.
When there are no coherent laser effects W12 and W21 are approximately proportional to the laser power for a given absorption spectrum, and laser frequency spectrum.
The stimulated emission and absorption rates are related by g2W21 = g1W12, where g2 and g1 are the level degeneracies. When the laser power level is low the upper population (W12 + W21 << Q21 + A21) will be proportional to the laser induced absorption rate:
W12 NT N 2 = (4-23) Q21 + A21
When operating in this low power regime where the excited population is proportional to the laser power, it is termed linear fluorescence measurements. In this regime the excited population is inversely proportional to the collisional quenching and spontaneous emission rates.
107 In the linear regime the number of photons incident, on the detector per unit
-1 time, np (s ) is given by:
N A V εΩ n = 2 21 c c (4-24) p 4π
Where ε is the optical collection efficiency, Ωc is the solid angle subtended by the
3 collection optics, and Vc is the fluorescence probe volume (cm ). Note isotropic fluorescence is assumed. The fluorescence signal is,
V f = n p hcν 21φ pG p (4-25)
Where h is Planck’s constant, c is the velocity of light, ν21 is the fluorescence
-1 transition frequency (cm ), φp is the quantum efficiency of the detector, and Gp
(V/W) is the detector gain. In the linear regime the observed signal depends on the fluorescence yield, φp:
A21 φ p = (4-26) A21 + Q21
The dependence of fluorescence yield on the quenching rate complicates linear fluorescence measurements. This is because in flames and plasmas there are many spatial in-homogeneities and a variety of quenching partners make correction for the quenching rate difficult. One method that can be used to circumvent this problem is saturation of the laser pump transition. At higher laser powers the absorption and stimulated emission rates can become much greater than that of the collisional quenching and spontaneous emission rates (W12 + W21 >> Q21 + A21). This causes the upper population level to approach the saturated limit:
108 g 2 N 2 = NT (4-27) g1 + g 2
Under saturation conditions the upper level population is independent of the laser power, collisional quenching rate, and the spontaneous emission rate. The fluorescence signal for both cases are given below in the form Vf = (excited state number density) (power emitted per excited species)(optical system collection efficiency)(detection electronics gain).
NTW12 Vcε cΩc V f (linear) = hcν 21 A21 φ pG p (4-28) Q21 + A21 4π
NT g 2 Vcε cΩc V f (sat) = hcν 21 A21 φ pG p (4-29) g1 + g 2 4π
The two level system described gives a basic understanding of the LIF process, however, can not be applied in a real situation. This is due to the fact that population changes occur over many levels and are not solely limited to those directly pumped by the laser. For example Figure 4.6 illustrates the situation where there are three levels ( NT=N1+N2+N3 ), where N2/NT is no longer independent of the collisional transfer rates, unlike the two level model described above.
109
Figure 4.6: Three level energy diagram.
Laser excitation dynamics in a molecule are seen in Figure 4.7, illustrating the radiative and collisional interactions a molecule undergoes during laser excitation. A particular vibrational-rotational level is excited via laser to another single vibrational-rotational level in the excited electronic level. It is important to note that whether or not a single transition is excited depends on the rotational line spacing, the laser line width, and laser power. Once the molecule is excited they can decay back to the original state, or to other ground state levels via spontaneous emission.
110
Figure 4.7: Molecular excitation dynamics. The transfer rate is given as Tij = Qij + Aij.
The radical of interest is OH, recall the relation between the fluorescence signal and the number density (or mole fraction) of OH is proportional to laser energy, fluorescence yield, OH number density, and the Boltzmann population fraction of the absorbing transition.10 For effective imaging of OH this requires that the mole fraction in the region of interest affects the signal intensity more than the parameters discussed above. The Boltzmann population fraction of OH up to 2500
Kelvin is seen in Figure 4.8. This shows 2-3% of the OH population in the excited
111 states of interest. It is interesting to note that as temperature increases the population in each state reduces. This is because as temperature increases the number of available states increases, and the OH re-distributes across all the states.
Figure 4.8: Boltzmann Population fractions of F1e state of OH (v = 0). Calculations provided by Dr. Cam Carter.
112 OH PLIF Experimental Procedure
This experiment consists of a Planar Laser Induced Fluorescence (PLIF) measurement to probe OH development and production induced by the transient plasma, a complete description is found in Cathey et. al.11 Figure 4.9 depicts the cylindrical combustion chamber that was designed for this experiment. The main imaging window is 10.16 cm diameter, 2.54 cm thick fused silica, and the laser entrance and exit windows are 2.54 cm diameter sapphire. The chamber is 10.16 cm inner diameter, 20.32 cm long stainless steel. The anode was an 8-32 threaded rod, approximately 7.62 cm long. It is important to note that during the combustion process water vapor is a product and condensation on the windows can be a problem.
As condensation only becomes a problem in post ignition work and this experiment is primarily focused on the radical production prior to ignition, no heaters were used to prevent water condensation on the windows during the imaging process.
Figure 4.9: Combustion chamber for PLIF experiment.
A line type pseudospark switched pulse generator charged with a Glassman
High Voltage DC supply was used to create a 70ns FWHM, 60 kV pulse to create the 113 transient plasma. 12 Chamber pressure was monitored with an Omega PX-105 piezoelectric pressure transducer. Spark ignition was accomplished using a conventional car sparkplug (Champion RV17YC6) and a commercially available automobile ignition circuit, which created a discharge of ~ 40 mJ. For spark ignition tests the transient plasma electrode was removed and replaced with the sparkplug, which was put in the center of the plate opposite the main viewing window.
Figure 4.10: Depicts electronic states and energy levels of OH employed in common LIF measurements. W indicates the absorption process, Q the electronic quenching process, V the vibrational energy transfer process, and the fluorescence process.13
Figure 4.10 depicts the relevant OH band structure when excited by pumping a transition in the A2Σ+ ← X2Π (1,0) band. Both UG-5 and WG-305 Schott-glass filters were used to block the (1,0) fluorescence and 282 nm scattering while allowing passage of the strong (1,1) and (0,0) bands at 314 nm and 309 nm (Figure
114 4.10). Note that the v′ = 0 (prime designating an electronically excited state) vibrational state is populated through vibrational energy transfer (VET), from v′ = 1. This excitation scheme mitigates fluorescence trapping (however it still exists) from the absorption of OH fluorescence by OH molecules that lie along the detection
14 path. The Q1(3) transition at 282.209 nm was chosen for excitation, as its ground state offers a reasonable compromise between low-temperature population and insensitivity to temperature changes at low temperatures.
The time-integrated fluorescence signal (counts), Sf, from collection volume Vc is proportional to the detected number of fluorescence photons:
Ω S = β c V f N Fˆ (4-30) f 4π c B OH
Here, β accounts for net optical transmission and detector gain; Ωc is the detection solid angle; W is the laser excitation rate and its time integral is proportional to laser pulse energy; NOH is the OH number density while fB is the Boltzmann fraction of
OH molecules in the absorbing state. Fˆ , the specific fluorescence, is defined as
N ε A + N ε A ⋅ dt ∫ ()1 1 1 0 0 0 Fˆ ≡ ∆t (4-31) f B N OH which for linear fluorescence conditions becomes
⎛ ε A + ε A V (Q + A )⎞ Fˆ = Wdt ⋅⎜ 1 1 0 0 1 e,0 0 ⎟ (4-32) ∫ ⎜ V + Q + A ⎟ ∆t ⎝ 1 e,1 1 ⎠
Here, Ni, εi, Ai and Qe,i are the respective number density, fluorescence collection efficiency, fluorescence rate, and electronic quenching rate for the A-state v′= i (0 or
1) level (see Fig. 2); and, QV,10 is the rate for VET from v′=0 to 1. Note that in
115 deriving Eq. (3b) it is assumed that upward VET, from v′=0 to 1 or from v′=1 to 2, is negligibly small.
Figure 4.11: Experimental setup for OH-PLIF Experiment.
An injection seeded pulsed Nd:YAG (Spectra-Physics GCR-170) laser was used to pump a dye laser (Lumonics Hyperdye), which was then frequency doubled to generate the beam at 282 nm. Specifically, the output of the frequency doubled Nd:YAG laser (400 mJ-per-6 ns pulse) pumps a Lumonics Hyperdye HD-300 dye laser to produce ~80 mJ/pulse at 564 nm. The 564 nm beam was then passed through a frequency doubler (Inrad Autotracker III) to obtain 282 nm radiation. The 282 nm beam was then separated from the dye beam with an Inrad Prism Harmonic Separator. The laser probe sheet was formed using a combination of a cylindrical (100 mm focal length) and spherical lens (1 m focal length). The sheet height was 116 apertured by the 25 mm high window slit, and the sheet thickness is estimated to be 300 µm. The reflection from the front surface of the cylindrical lens was directed over a reference flame and then to a photodiode to monitor the respective LIF from the nascent OH, using a photomultiplier tube, and the beam energy. Both signals were recorded on a digital oscilloscope. The purpose of monitoring the OH LIF signal was to ensure that the laser remained tuned to the OH Q1(3) transition throughout the experiment. Additionally, a second pick off at the exit window of the chamber was setup to monitor scattering and absorption of the beam due to condensation on the windows during combustion. The fluorescence was collected by a PIMAX UV sensitive (Superblue) intensified CCD camera with a Cerco 45 mm focal length f/1.8 UV lens. The camera gate was normally set to 200 ns, but was also set as low as 50 ns to reduce collection of the discharge emission. Furthermore, the camera’s bracket pulsing option was also used to further reduce the collected emission from the discharge. The experimental setup, as seen in Figure 4.11, was the same for the high speed camera work, excluding the laser. The Princeton Instruments camera was replaced with a Photron FASTCAM-Ultima APXi2 intensified high speed camera. Images were recorded with both CH (centered at 430 nm with a pass band of 10 nm, FWHM) and OH line filters (centered at 307 nm with a pass band of 25 nm, FWHM) at 2000 frames-per-second.
OH Number Density Calibration Procedure
To estimate the OH number densities, a 25.4-mm-square Hencken burner
[Barlow et al.] was used as a calibration source (Figure 4.12). 15 Previously
[Ombrello et al.], absorption measurements had shown that the OH number density
16 -3 in the burnt-gas region of a premixed CH4-air flame was (0.94 ± 0.07) × 10 cm , a 117 value consistent with a burnt-gas temperature of 2170 K (versus the adiabatic equilibrium value of 2194 K).16 The ratio of fluorescence signals along with the number density in the Hencken flame, allowed estimation of the discharge number densities.
Figure 4.12: Hencken Burner H2-air flame (left), CH4-air (right).
Here, it assumed that the fluorescence is in the linear regime. However, based an estimated laser sheet irradiance of ~6.7×1010 W/m2 and the derived excitation rate
~5×109 s-1, some degree of transition saturation exists, particularly for the flame conditions. For the linear fluorescence condition, it is necessary to determine the collision rates Q1, Q0, and V1. This was done using cross sections from Steffens and
Crosley, for low temperatures, and from Paul, for the high temperature (i.e., calibration flame conditions); however, a complete set of cross sections do not exist, and some assumptions and estimations are necessary to derive the ratio of the fluorescence yield (the term in parentheses in the second part of Eq. 2) between the 118 two conditions.17, 18 It is important to emphasize that the two conditions, the burnt gas and the combustible mixture, are very different collisional environments, and a more suitable low temperature calibration condition is desired.
OH PLIF Results and Discussion
It was found that transient plasma did in fact induce OH production throughout the discharge volume, likely along the streamer propagation path (figure
4.13). Contrary to the original supposition that ignition occurs over the entire discharge volume, it looked to occur only near the anode, creating an outward cylindrically propagating flame. Additionally it was found that the OH that was initially induced by the discharge decayed below detectable levels near 100 µs, while ignition did not occur till ~ 1 ms.
Figure 4.13: PLIF images of OH production when combusting quiescent stoichiometeric CH4-air mixtures. White haze on the bottom of the frames is an 119 artifact of the processing, and brightness of the frames were adjusted for view ability.
Due to the non-thermal nature of the plasma the gas temperature changes negligibly during the discharge. Additionally the amount of energy that is being delivered is quite small and is not enough to significantly raise the gas temperature above ambient conditions. By assuming that the gas temperature is ~ 300 K the ratio the Boltzman ratio can be calculated to determine the number density in the mixture.
For a complete description of the process see, Cathey et. al..13 Figure 4.14 is a decay curve of the OH that was induced by the plasma, with peak mean values of ~ 4·1014 molecules – cm-3. It is important to note that while the mean values are on the order of 1014 molecules – cm-3, there are spots within the OH structures where OH levels are higher by an order of magnitude or more (see inset in Figure 4.14).
120 Figure 4.14: OH decay curve. The inset is a PLIF image calibrated to show OH number density at t =1 µs, with value in 1014 molecules – cm-3.
High Speed Camera Imaging
High speed or high frame rate cameras, for example those with frame rates of
1000 framers per second or more can effectively be used to image and characterize events that last on the order of 100s of milliseconds or less. Being able to create a slow motion video of transient events can be extremely informative in allowing you to capture a visual record of the entirety of an event, unlike traditional camera where
121 they can only capture a single shot. While single shot analysis is useful, its utility is diminished when looking at events which do not repeat exactly in the same manner.
High Speed Camera Experimental Procedure
The experimental setup as seen in Figure 4.11 was essentially the same for the high speed camera work, minus the laser. The Princeton Instruments camera was replaced with a Photron FASTCAM-Ultima APXi2 intensified high speed camera.
Images were recorded with both CH (centered at 430 nm with a pass band of 10 nm,
FWHM) and OH line filters (centered at 307 nm with a pass band of 25 nm, FWHM) at 2000 frames per second. Additionally the field of view of the camera was adjusted from the PLIF experiment such that the camera was able to image the entire discharge volume. Like the OES and PLIF experiment the combustion of quiescent, stoichiometric CH4 – air mixtures were observed.
High Speed camera Results and Discussion
The high speed camera images seen in figure 4.15 & 4.16 confirm that ignition via a transient plasma occurs approximately 1 ms after the discharge, and only along the anode, creating an expanding cylindrical flame. Multi-point ignition was only observed along the anode, and no ignition was observed elsewhere in the discharge gap.
122
Figure 4.15: Flame propagation in the visible of transient plasma and spark ignited CH4-air mixtures during a single combustion cycle. The exposure time of each frame is 500 µs.
It is important to note that ignition along the anode and faster flame speed was also seen during the automobile internal combustion engine experiments at
Nissan (figure 4.16). The pressure-gap conditions for the Nissan example are 4864 torr-cm and for the methane experiments 3860 torr-cm. This result is additional support for the idea that the electrode may have an effect on ignition.
123
Figure 4.16: Visible flame propagation for transient plasma and spark ignited iso-octane-air mixtures in a single cylinder test engine.
There are several possible explanations as to why ignition is solely occurring along the anode. The first and perhaps the simplest is that this is the region of highest field and therefore highest radical production. Recall that the coaxial geometry produces a highly non-uniform electric field, using the PLIF chamber test conditions a peak field of ~ 3 MV/m is at the anode whereas a parallel plate geometry would produce only ~ 1 MV/m under the conditions tested. Additionally the anode is a threaded rod; whose points further enhance the field locally above and beyond 3 MV/m. Another potential cause of ignition near the anode is a plasma- induced local increase in the concentration of atomic oxygen in the mixture, where the high currents of the transient plasma remove, at least partially, the electrode’s surface oxide layer. This local generation of oxygen would enhance combustion
124 near the anode. A final postulate is that the plasma discharge likely heats the anode.
This heating may assist the combustion reactions by providing the heat needed to assist local ignition along the anode. Further study is needed in order to determine why ignition initiates along the electrode and if ignition within the chamber volume is possible with increasing reduced electric field.
The difference in flame geometries is important to note between the transient plasma and the spark ignited systems. The conventional spark plug, can be assumed to be a point source when compared to the transient plasma. Additionally it creates essentially an expanding hemispherical flame that grows along the rear wall of the chamber, and then propagates forward towards the camera, always in contact with the wall. Whereas the transient plasma ignited flame begins over a much large volume surrounding the ~75mm long 8-32 threaded rod, and propagating outward towards the chamber wall. When the transient plasma ignited flame reaches the chamber wall ( t ≈ 20 ms ) it has combusted ≈ 40% of the chamber volume, whereas at the same time the spark ignited flame has not left the rear chamber wall. The wall contact is important because it acts as a heat & radical sink to the flame and retards increases in flame temperature and speed.
Looking at the transient plasma ignited flame there appears to be some wrinkling of the flame front. This is likely due to multiple ignition kernels along the anode surface. It is also possible that wrinkling at large radii is caused by concentration gradients across the surface that increase the rate of diffusive mixing across the flame front .19
125
Figure 4.17: Flame propagation of transient plasma and spark ignited CH4-air mixtures during a single combustion cycle. The top two rows are imaged using a CH line filter, and the bottom two rows are imaged using an OH line filter.
Using the high speed camera in conjunction with line filters produced the images Notice that after the flame front reaches the chamber wall at approximately
20 ms, chamber emission continues to increase until about 60ms. This is likely due to the combustion of the mixture in the fore and aft portions of the chamber (see Fig.
126 4.5). Additionally the back of the chamber is a reflective surface which may be falsely increasing the signal level at these long time scales. The OH signal is particularly strong, and is attenuated with respect to the CH images taken. This was done to mitigate saturation of the camera detector.
Figure 4.18: Visible flame propagation for transient plasma and spark ignited CH4-Air, φ =1, mixtures at t = 20 ms, and associated pressure traces.
Figures 4.13 and 4.14 suggest that the flame speed of the transient plasma is faster than that of the spark ignited flame. This is easily seen when a particular time is looked at for both transient plasma and spark ignited flames. Figure 4.15 depicts flame propagation 20 ms post transient plasma/spark and the associated pressure curves. Additionally it is seen that the ignition delay time (time it takes to reach 10% of the peak pressure) for a transient plasma ignited flame is 34% faster than that of the spark ignited flame. Additionally the peak pressure reached were higher, the
127 transient plasma reached 6.7 atm, whereas the spark assisted flame reached only 6.3 atm. Lastly the pressure gradient (or rise time from 10% – 90% ) was 59% larger for the transient plasma case. It is important to note that the transient plasma was delivering ~ 800mJ, whereas the spark was delivering ~ 40 mJ. Additionally the ignition delay and rise time for transient plasma ignited flames are highly dependent on electrode – chamber geometry. An example of the geometric dependence is seen in previous work using CH4-air mixtures where a factor of 3 reduction in ignition delay was achieved using a smaller gap chamber (higher field conditions) with a longer electrode (increased number of ignition kernels).20
The increase in flame speed is potentially from several sources: 1) does enhancement of the flame speed occur from OH produced in the gap prior to ignition
(but seen to decay below ~ 1·1014 molecules –cm-3 near 100 ms), 2) is it solely a property of the flame geometry (cylindrical v. point source). Jeremy Cain and Hai
Wang answered this question by analyzing the high speed camera images to determine the flame speed, and then compared them to established values of stoichiometric CH4 –air mixtures. They found the laminar flame speed to be 39 ±13 cm cm/sec for the transient plasma, well within the established values for cylindrical propagating flames. This result indicates that the OH seen in the first 100 µs of the discharge has little or no effect on flame propagation, as ignition does not occur till 1 ms. Additionally when combined with the observation that ignition always occur near the anode it implies that there may be some plasma induced chemistry near the anode that is responsible for ignition.
128 Conclusions
The transient plasma does indeed produce radicals, potentially over a very large volume. The original supposition was that high energy electrons in the streamer would form radicals that would ignite the fuel-air mixture. While this has not been disproved, the previous experiment has shed considerable doubt on this hypothesis. Under the conditions tested they decayed rapidly, within 100 µs falling below the detectible limit of the experiment, while ignition did not occur for another
900 µs. Also it was shown that the transient plasma induced radicals did not enhance the speed of flame propagation and ignition always occurred solely along the anode, creating an outward propagating cylindrical flame.
When these results are taken together is suggests that the anode may have previously unconsidered or neglected effects on ignition. The rapid decay of radicals with little proximity to the anode suggests that something is occurring providing enough heat release for combustion to occur. The peak field density near the anode may allow enough radicals to be produced during the discharge for ignition to occur, however, the time decay OH in the PLIF images suggest additional potential chemical processes. Such as the plasma removing the oxide layer of the anode which locally enhances the combustion process or local elevation of gas temperature by joule heating of the electrode during the discharge which provides the heat release necessary for ignition. The potential for electrode material effects on ignition could provide the engine designer another tool to control combustion in addition to fuel selection and stoichiometry.
129
Chapter 4 Endnotes
1 S M Starikovskaia, “Plasma Assisted Ignition and Combustion,” J. Phys. D: Appl. Phys. 39 (2006) R265–R299.
2 A. Klimov, V. Brovkin, V. Vinogradov, and D. VanWie, AIAA paper no. 2001- 0491 (2001).
3 Y. P. Raizer, Gas Discharge Physics (Springer-Verlag, Berlin, 1997).
4 I. Glassman, Combustion, Academic Press: New York, pg 90-91. ADJ by Prof Hai Wang.
5 F. Wang, A. Kuthi, and M.A. Gundersen, IEEE Trans. Plasma Sci. 33 (2005) 1177-1181.
6 J. J. Lowke and R. Morrow, IEEE Trans. Plasma Sci. 23, 661, 1995.
7 H. Okabe, Photochemistry of Small Molecules ~Wiley, New York, 1978.
8 Eckbreth, A., Laser Diagnostics for Combustion Temperature and Species, Abacus Press: Cambridge 1998.
9 Radziemski, L. J., Solarz, R. W, and Paisner, J. A., laser Spectroscopy and its Applications, Marcel Dekker: New York, 1987.
10 R. K. Hanson, J. M. Seitzman, and P. H. Paul, “Planar laser-fluorescence imaging of combustion gases,” Applied Physics B: Lasers and Optics, 50, 6 (1990).
11 C. Cathey, J. Cain, H. Wang, M. A. Gundersen, M. Ryan, and C. Carter, “OH Production by Transient Plasma and Mechanism of Flame Ignition and Propagation in Quiescent Methane-Air Mixtures,” Combustion and Flame, submitted.
12 F. Wang, A. Kuthi, and M. A. Gundersen, “Compact High Repetition Rate Pulse Generator,” IEEE Transactions on Plasma Science, Volume: 33 , Issue: 4 , Part 1, Aug. 2005, pg 1177-1181.
13 K. Kohse-Hoinghaus, “Laser Techniques for the Quantitative Detection of Reactive Intermediates in Combustion Systems,” Prog. Energy Combust. Sci. Vol 20, 1994.
130
14 J.M. Seitzman and R.K. Hanson, “Planar Fluorescence Imaging in Gases,” Chapter 6, (pp. 405466) In Experimental Methods for Flows with Combustion, ed. A. Taylor, Academic Press, London, 1993.
15 Barlow, R. S., Karpetis, A. N., Frank, J. H., and Chen, J.-Y., “Scalar Profiles and NO Formation in Laminar Opposed-Flow Partially Premixed Methane/Air Flames,” Combustion and Flame, Vol. 127, No. 3, 2001,pp. 2102–2118.
16 Ombrello, T., Qin, X., Ju, Y., Gutsol, A., Fridman, A., and Carter, C., “Combustion enhancement via stabilized piecewise nonequilibrium gliding arc plasma discharge,” AIAA Journal, vol. 44, 2006, 142-150.
17 Steffens, K. L., and Crosley, D. R., “Vibrational energy transfer in OH A2 between 195 and 295 K,” J. Chem. Phys. Vol. 112, 2000, pp. 9427-9432.
18 Paul, P. H., “Vibrational energy transfer and quenching of A2Σ+ (v′=1) measured at high temperatures in a shock tube,” J. Phys. Chem. Vol. 99, 1995, pp. 8472- 8476.
19 R. Abu-Gharbieh, G. Hamarneh, T. Gustavsson, and C. Kaminsk,i Optics Express 5 Vol. 8 (2001) 278-287.
20 J. Liu, F. Wang, L.C. Lee, P.D. Ronney, and M.A. Gundersen, “Effects of Fuel Type on Flame Ignition by Transient Plasma Discharges,” 42nd Aerospace Sciences Meeting, 6th Weakly Ionized Gases Workshop, Reno, NV, Jan 2004.
131 Chapter 5 Conclusion
Transient Plasma Ignition
Transient plasma ignition has been shown to have substantial advantages over traditional spark ignition for both PDEs and automobile engines. Transient plasma ignition ability to reduce ignition delay and increased capability to ignite over a broader range of flow, temperature, pressure, and fuel types; make it a potentially enabling technology for this engine platform. Additionally, it is an attractive ignition source for automobile manufactures as it can ignite leaner mixtures than traditional spark ignition, which retards NOx production, a pre-curser to photochemical smog. Additionally preliminary data suggests that the rapid burn rate of transient plasma ignited flames retards soot formation.
The PLIF results confirmed, as expected, OH production over the entire discharge gap. Contrary to expectations, was the rapid decay of OH produced (< 100
µs) in the discharge gap (seen in the laser sheet) relative to when ignition occurred
(1ms). The high speed imaging confirmed that ignition occurred only near the anode, creating a cylindrically propagating flame, and that it was only volumetric ignition source in that ignition occurred over the electrode length. Ignition along the anode when combined with the rapid decay of the OH seen in the PLIF images suggests that there may be some plasma induced chemistry near the anode. In order to fully understand the ignition mechanism, be it heating of the electrode, local oxygen enhancement, or simply high field density, further study is required.
132 While the ignition mechanism may not be fully understood, transient plasma has outperformed spark ignition over a variety of conditions. However, in order to make it a viable technology research and development must continue in creating compact sources of pulsed power capable of producing high voltage pulses less than
100 ns long.
Future Work
At the core of my research are applications of pulsed power. The non- equilibrium atmospheric plasmas that pulsed power allows us to create have several extremely interesting applications. Additionally the work I have done so far on
PDEs and internal combustion engines has allowed me to identify areas in pulse generator design that need to be improved in order to make a more reliable and compact product.
Pulse Generator Design
The pseudospark switched (PS) line type pulse generator used for the majority of my research while highly versatile is a large, bulky device, which also requires a charging device. The solid state opening switch (SOS) pulse generator described in Chapter 3 is much smaller than the PS switched pulse generator and its architecture has potential to be easily modified for multi-cylinder (automobile) or multi-tube (PDE) operation.
The last improvement that is needed to the PS Pulse generator is a computer controlled waveform acquisition system. Currently voltage and current waveforms are taken from probes in the box directly to an oscilloscope. This setup is adequate
133 for single shot experiments, however, for multiple shot experiments it is far from ideal. In a typical experiment at the Naval Post Graduates schools the PS pulse generator is fired at 100 Hz for 1 second, generating 100 different voltage and current waveforms respectively. Under the current system we are only collection a single voltage and current waveform which is not necessarily typical of the 100 shots. What would be a huge improvement is a LabView based control that not only saved every wave form, but also is able to perform energy and power calculations.
This would allow a direct correlation between pulse data to the ignition delay and wave speed data that is collected on the PDE. A trigger function could also be introduced, with an optical link to the pulse generator and rapid charger. Thus allowing control and monitoring of the pulse from inside the control room.
These improvements are necessary if this technology is going to be viable as an actual ignition system on an airborne platform. In particular if the group starts testing the PS pulse generator as an ignition source for a SCRAMJET (see section
4.2). SCRAMJET testing is incredibly expensive to run. At Wright Patterson Air
Force Base they only run the experiments at night, and have a about a 3 hour window to test. This is because the electrical load is on the order of MegaWatts, and they need to use the power at off-peak times. Working under such conditions requires that your equipment is extremely reliable as there will be little if any time to repair during the testing cycle.
134 High Speed Thermal Imaging
Recall that transient plasma as an ignition mechanism seems to results in ignition near the anode (in the geometry & conditions tested). Hundreds of amps of conducted during the transient plasma pulse and it plausible that they would create local hotspots along the anode that heat the mixture enough for combustion reactions to progress. In order to validate this hypothesis a high speed infrared (IR) camera could be used to monitor thermal emission from the electrode during and post discharge. The experimental setup would be similar to the high speed imaging experiment in Chapter 4, however, the detector, optics, and baffling would have to be optimized for the IR. The detector could be calibrated by heating the anode to know temperatures, and matching the image intensities to the know anode temperature.
Additionally this experiment would be useful in creating a 2D map of the flame temperature post ignition.
Laser Scattering Soot Experiment
Recall from chapter 2 that transient plasma was found to produce a factor of
50 less soot than spark ignited ethylene – air mixtures. However, due to limitations with the experiment we were not able to fully characterize soot production. Using
Mie scattering experiment or possibly laser induced incandescence a full characterization of the soot number density, and particle size can be made.
Additionally the experiment can be modified to monitor emissions in a PDE.
135 SCRAMJET Ignition and Relight
A SCRAMJET engine has no low speed propulsion capability, and will only begin to function at approximately Mach 4. There is significant interest in both hydrogen and hydrocarbon high speed fuel efficient operation for these engines. A significant problem in scramjet engines is reliable high speed and high altitude ignition and relights.1 In a supersonic engine, like a Ramjet or a PDE the flow is slowed to subsonic speeds in the combustor, this process limits the upper operational speeds. However, in a SCRAMJET the flow is supersonic through the combustor allowing for much higher operational speeds.
A SCRAMJET must be propelled to approximately Mach 4 before it will be able to function. Once ignition of the supersonic flow is achieved the flame is self- sustaining under the proper speed and pressure conditions. However, there is the possibility that flame blow off will occur, and there needs to be a reliable way to relight the engine. As you may have surmised many of the techniques used for testing in the lab are not readily transferable to an actual airborne platform. In ground tests there is a restriction (aerodynamic or physical) induced in the supersonic flow to raise the static pressure, and temperature. This will establish a pre-combustion shock train prior to flame holding and fuel injection region. Another method uses a kinetic accelerant which will cause the fuel-air mixture to release enough heat to backpressure the combustion region. If the heat released is high enough combustion will become self-sustaining, and the ignition aid can be removed.
For a flyable system a secondary supply of gas can act as the aerodynamic restrictor or the use of material like silane (accelerant) are used to facilitate the ignition 136 process. These technologies are effective however; they increase system weight and complexity, and introduce serious safety issues (silane reacts explosively with air).
However, the real problem lies with the reliability of these systems. Ignition is achieved through a balance between blockage introduced for ignition, and that resulting from the combustion released heat. If there is too much backpressure the engine inlet will un-start, if there is too little a self sustaining combustion condition is not achievable. Another important factor is that these systems allow for a limited number of ignition attempts, and probably no re-light attempts.
Subcritical microwave discharges, fast ionization waves, plasma jet injectors, and pulsed DC discharges with magnetic amplification are all implementations of plasma assisted combustion technologies in SCRAMJETS. These technologies have been shown to enhance ignition, extend flameholding operational limits, and increase mixing and combustion efficiencies. The areas where plasma assisted combustion can potentially make its mark, are a) reduction in ignition delay time, b) improved flameholding, c) shortening of the mixing and combustion length, and d) allowing for stable burning and ignition for external burning systems.
Subcritical microwave discharges are an efficient, compact ignition method.2
In this technology a microwaves at power levels lower than that required to initiate a discharge in free space are used. This technology requires the use of a field concentration device or a laser initiation system. This ignition methodology can also occur over a relatively large volume. Once the discharge is ignited it will expand to fill the volume or to the point where the field strength is not longer able to support it.
137 Plasma jet injectors operate by creating radicals that are subsequently injected into the fuel-air flow. In a plasmadynamic injector a large discharge current creates a plasma from a portion of the fuel, which is then accelerated via a self induced magnetic field to high speeds. This injector is able to penetrate deeply into the cross stream flow.3,4
There has also been work done in pulsed DC discharges. In which hot filaments are formed between wall mounted electrodes that interact with the chemical energy that is released during combustion, generating instabilities in the flow. These instabilities result in rapid movement and mixing of the discharge filaments.
Plasma based ignition technologies may be able to deliver enough energy to accelerate fuel air kinetic rates to the point where self sustaining combustion occurs.
The capability exits to build powerful, relatively compact pulsed power devices, however, the question still remains if plasma assisted ignition is viable under high speed, high temperature conditions. Research to date has primarily been with a plasma torch. A plasma torch is a high energy arc which has been coupled to a stream of gas. However, transient plasma generated via electric and microwave discharges are currently in development.
In using the PS pulse generator for this application there are several areas that need to be addressed. In all likelihood it will take the repetition rates in the kilohertz in order to ignite the supersonic flow, so the rapid charger capacitor bank will have to be increased. Also the components in the rapid charger will probably have to be replaced with those with a higher power rating. Another potential issue is the 138 electrode interface and what kind of discharge geometry can be generated, and what is necessary to ignite the flow.
In order to obtain combustion of high speed flows it is important that the streamer velocity is as high as possible. Typical streamer velocities in air range from
107-109 cm/sec. This may be fast enough so that it can have applications for supersonic combustion in aircraft engines, which occurs in hypersonic platforms like a SCRAMJET. In order to obtain high streamer velocities it is important that the rise time of the pulse is as fast as possible. There also seems to be a difference in the breakdown voltage for positive and negative streamers. There has been some limited work showing that the breakdown voltage increases for positive streamers with increasing flow rate, whereas in negative streamers the breakdown voltage remains constant. Currently this is not totally explained, however, it is postulated that the differences arise from the different transport mechanism for each polarity streamer.5
OH Absorption and O Emission Shock Tube Experiment
Temperature is generally an extremely important factor in combustion reactions, especially the speed at which the reactions progress. In general for thermal processes the ignition delay (τ) is exponentially dependent on temperature see equation (5-1), which exhibits the temperature dependence of the reaction rate constants.
τ = A e(B/T), where A and B are constants (5-1)
139 Recall that the rule of thumb that the reaction rates double with every 10 degrees
Celsius.
The third experiment is a shock tube experiment where laser absorbance measurements of OH, and emission of O will be looked at under similar conditions as experiment 1, only now temperature can be a variable, and we can see its affect on radical production and ignition delay. This experiment will be preformed at Stanford in Professor Hanson’s optical diagnostics lab. Additionally the mixture temperature over the time prior to ignition can be measured optically with a resolution of several kHz.
In order to get an accurate representation of the conditions in a PDE the temperature of the gas needs to be about 200 - 400 K, and pressure of about 1 atm.
This experiment will be able to correctly simulate the temperature and flow conditions seen in a PDE. This is a viable method to replicate flow temperature and pressure over a large range, and has been used successfully by the Starikovskii’s group, Moscow Institute of Physics Laboratory, to research non-equilibrium plasma assisted combustion.6 This experiment is an optical study of OH production via transient plasma under realistic temperature and pressure conditions. Since we are interested in the time between ignition and application of the transient plasma an absorbance measurement is needed to accurately monitor the OH progression.
Additionally atomic O2(b→X) emission at 760nm will be recorded as an indirect measurement of O2(b→X). As this experiment will be performed at Stanford in
Professor Ronald Hanson’s laboratory this will also serve as a baseline for integrating our TPI system with their Shock tube which will facilitate future 140 SCRAMJET ignition studies.
A shock tube is essentially a tube with a gas at low pressure (driven gas) separated by a gas at high pressure (driver gas) by a diaphragm. When the diaphragm bursts a compression wave is formed in the driven gas which rapidly transitions to a shock wave. Once the incident shock wave hits the end of the tube it will reflect back into the already heated gas, causing the temperature, pressure, and density of the gas to raise further to the desired test conditions. The current experiment will be conducted in reflected shock waves. The Stanford kinetics shock tub is seen if Figure 5-1.
Figure 5.1: Stanford kinetics shock tube (left), the optical section of the shock tube where the emission, absorption measurements are made (right). This section is where the TPI electrode will interface with their system.
The diagnostics of this experiment are divided into three sections; monitoring shock wave parameters, monitoring the transient plasma, and the optical spectroscopy. This experiment I will be primarily responsible for the incorporating
141 the transient plasma into the shock tube, shock tube operation and the optical measurements will be taken my members of Professor Hanson’s group.
Plasma Assisted Flow Control
Another interesting application that our pulsed power technology can be applied to is flow control for aerodynamic systems. For this application the non- equilibrated plasma is generated by an RF (typically several kHz) dielectric barrier discharge (DBD) is used to prevent gas flows from separating, which results in increased drag mostly due to the pressure differential between the front and rear surfaces of the object.7 There has been work recently at Princeton that points toward an increase in efficiency if the RF discharge is replaced by short (10s of nanoseconds) repetitive pulses. If our system is going to be scaled for high repetition rate operation for the SCRAMJET ignition work, it can then be easily latterly transferred for flow control research.
142
Chapter 5 Endnotes
1 L. Jacobson, C. Carter, R. Baurle, T. Jackson, 41st Aerospace Sciences Meeting, AIAA 2003-0871, Reno, Nevada (2003).
2 D. Van Wie, D.J. Risha, C.F. Suchomel, 42nd Aerospace Sciences Meeting, 6th, Reno, Nevada, AIAA-2004-1357 (2004).
3 L. Timofeev, Milestone 6, Task 2, JHU/APL Subcontract 808713, (2000).
4 A. P. Ershov, N. V. Arvelyan, V. L. Bychkov, V. A. Chernikov, V. M. Shibkov, O. S. Surkont, I. B. Timofeev, AIAA 872 (2003).
5 A. Jaworek, A Krupa, ELMECO’97, “Electromagnetic Devices and Processes in Environment Protection,” Lublin, (1997).
6 S.M. Bozhenkov, E.N.Kykaev, A.Yu.Kuksin, M.M.Nudnova, S.M.Starikovskaia, A.Yu.Starikovskii “Plasma Control Of Ignition Of Hydrogen-Air And Methane- Air Mixtures.” 39th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, 2003. AIAA Paper AIAA2003-5045.
7 S. Macheret, “Nonequilibrium Surface Plasma for Aerodynamic Control,” Gordon Research Conference on Plasma Processing Science, July 2006.
143 Bibliography
Aleksandrov, N. L., Anikin, N. B., Bazelyan, E. M., Zatsepin, D.V., Starikovskaia, S.M., Starikovskii, A.Y., “ Chemical Reactions and Ignition Initiation in Hydrocarbon - Air Mixtures by High-Voltage Nanosecond Gas Discharge” 32nd AIAA Plasma dynamics and Lasers Conference and 4th Weakly Ionized Gases Workshop, June 2001 Anaheim, CA, AIAA paper 2001-2949.
Abu-Gharbieh, R., Hamarneh, G., Gustavsson, T., and Kaminsk, C., Optics Express 5 Vol. 8 (2001) 278-287.
Aleksandrov, N. L., Anikin, N.B., Bazelyan, E. M., Zatsepin, D. V., Starikovskaia, S. M, Starikovskii, A. Y., 32nd AIAA Plasmadynamics and Lasers Conference and 4th Weakly Ionized Gases Workshop, 2001-2949, (2001).
Alstom FS200 Pseudospark switch data sheet.
Barlow, R. S., Karpetis, A. N., Frank, J. H., and Chen, J.-Y., “Scalar Profiles and NO Formation in Laminar Opposed-Flow Partially Premixed Methane/Air Flames,” Combustion and Flame, Vol. 127, No. 3, 2001,pp. 2102–2118.
Birx, D. L., Lauer, E. J., Reginato, L. L., Schmidt, J., and Smith, M., “Basic Principles Governing the Design of Magnetic Switches,” Lawrence Livermore Laboratory Report No. UCID – 18832 (1980).
Bozhenkov, S. A., Starikovskaya, S. M., Starikovskii, A. Yu., Comb. and Flame 133, 133-146 (2003).
Bozhenkov, S. M., Kykaev, E. N., Kuksin, A. Yu., Nudnova, M. N., Starikovskaia, S. M., and Starikovskii, A. Yu., “Plasma Control Of Ignition Of Hydrogen-Air And Methane-Air Mixtures.” 39th AIAA/ASME/SAE/ASEE Joint Propulsion Conference and Exhibit, 2003. AIAA Paper AIAA2003-5045.
Burkes, T.R., Craig, J. P., Hagler, M. O., Kristiansen, M., Portnoy, W. M., “A review of high-power switch technology,” IEEE Transactions on Electron Devices, Vol 26, Issue 10, Oct 1979 Page(s):1401 – 1411.
Bussing, T. R. A., Bratkovich, T. E., and Hinkley Jr, J. B., “Practical Implementation of Pulse Detonation Engines, “ AIAA 2748-1997.
Cathey, C. ,Cain, J., Wang, H., Gundersen, M. A., Ryan, M., and Carter, C., “OH Production by Transient Plasma and Mechanism of Flame Ignition and Propagation in Quiescent Methane-Air Mixtures,” Combustion and Flame, submitted.
144 Cathey, C., Tang, T., Shiraishi, T., Urushihara, T., Kuthi, A., and Gundersen, M. A., “Nanosecond Plasma Ignition for Improved Performance of an Internal Combustion Engine,” IEEE Trans on Plasma Sci., submitted.
Cathey, C., Wang, F., Tang, T., Kuthi, A., Gundersen, M.A., Sinibaldi, J., Brophy, C., Hoke, J., Schauer, F., Corrigan, J., Yu, J., Barbour, E., and Hanson, R., “Transient Plasma Ignition for Delay Reduction in Pulse Detonation Engines,” 45th AIAA Aerospace Sciences Meeting and Exhibit, Reno, Nevada, 2007, TBP.
Cook, E. G., Akana, G., Gower, E. J., Hawkins, S. A., Hickman, B. C., Brooksby, C. A., Cassel, R. L., De Lamare, J. E., Nguyen, M. N., Pappas, G. C., “Solid-State Modulators for RF and Fast Kickers,” SLAC-PUB-11759.
Eckbreth, A., Laser Diagnostics for Combustion Temperature and Species, Abacus Press: Cambridge 1998.
Ershov, A. P., Arvelyan, N. V., Bychkov, V. L., Chernikov, V. A., Shibkov, V. M., Surkont, O. S., and Timofeev, I. B, AIAA 872 (2003).
Fridman, A., and Kennedy L. A., Plasma Physics and Engineering, Taylor and Francis, New York, 2004.
Glassman, I., Combustion, Academic Press: New York, pg 90-91. ADJ by Prof Hai Wang.
Hanson, R. K., Seitzman, J. M., and Paul, P. H., “Planar laser-fluorescence imaging of combustion gases,” Applied Physics B: Lasers and Optics, 50, 6 (1990).
Hippler, R., Pfau S., Schoenbach (Eds) K. H. , Low Temperature Plasma Physics, Wiley-VCH, New York, 2001.
Hutcheson, P., Brophy, C., Sinibaldi, J., Cathey, C., and Gundersen, M.A., “Investigation of Flow Field Properties on Detonation Initiation,” 42nd AIAA/ASME/SAE/ASEE Joint Propulsion Conference 2006, Sacramento, California, 9 -12 July 2006.
Jacobson, L., Carter, C., Baurle, R., Jackson, T.,41st Aerospace Sciences Meeting, AIAA 2003-0871, Reno, Nevada (2003).
Jaworek, A., and Krupa,A., ELMECO’97, “Electromagnetic Devices and Processes in Environment Protection,” Lublin, (1997).
Jiao, C. Q., DeJoseph Jr., C. A., and Carscadden, A., “Dissociative Ionization of JP- 10 (C10H16) by Electron Impact,” International Journal of Mass Spectrometry, submitted. 145
Klimov, A., Brovkin, V. , Vinogradov, V., and VanWie, D., AIAA paper no. 2001- 0491 (2001).
Kohse-Hoinghaus, K., “Laser Techniques for the Quantitative Detection of Reactive Intermediates in Combustion Systems,” Prog. Energy Combust. Sci. Vol 20, 1994.
Liu, J. B., Theiss, N., Ronney, P. D., and Gundersen, M. A., “Minimum ignition energies and burning rates of flames ignited by transient plasma discharges,” 2003 meeting of Western States Section/Combustion Institute, UCLA, Oct 20-21, 2003, Paper 03F-88. 2002.
Liu, J., Wang, F., Lee, L. C., Ronney, P. D., and Gundersen, M. A., “Effects of Fuel Type on Flame Ignition by Transient Plasma Discharges,” 42nd Aerospace Sciences Meeting, 6th Weakly Ionized Gases Workshop, Reno, NV, Jan 2004.
Liu, J., Wang, F., Lee, L.C., Theiss, N., Ronney, P.D., and Gundersen, M.A. , “Effect of discharge energy and cavity geometry on flame ignition by transient plasma,” 42nd Aerospace Sciences Meeting, 6th Weakly Ionized Gases Workshop, Reno, Nevada 5 - 8 Jan 2004, Paper Number : AIAA-2004-1011.
Loeb, L .B., Basic Processes of Gaseous Electronics, University of California Press: Berkeley (1960).
Lowke, J. J., and Morrow, R., IEEE Trans. Plasma Sci. 23, 661, 1995.
Macheret, S., “Nonequilibrium Surface Plasma for Aerodynamic Control,” Gordon Research Conference on Plasma Processing Science, July 2006.
Martin, T. H., 5thh IEEE Pulsed Power Conference, 1985, 74.
Meek, J. M., Craggs, J. D., Electrical Breakdown of Gases, Wiley: New York (1978).
Meville, W. S., “The Use of Saturable Inductors as Discharge Devices for Pulse Generators,” Proc. IEE 98 (1951) 185-207.
Okabe, H., Photochemistry of Small Molecules ~Wiley, New York, 1978.
Ombrello, T., Qin, X., Ju, Y., Gutsol, A., Fridman, A., and Carter, C., “Combustion enhancement via stabilized piecewise nonequilibrium gliding arc plasma discharge,” AIAA Journal, vol. 44, 2006, 142-150.
Pai, S. T., and Zhang, Q., High Power Pulse Technology, World Scientific: Singapore, 2003, pg 3. 146
Pancheshnyi, S., Nudnova, M., Starikovskii, A., Phys. Rev E , 71, (2005).
Paul, P. H., “Vibrational energy transfer and quenching of A2S+ (v¢=1) measured at high temperatures in a shock tube,” J. Phys. Chem. Vol. 99, 1995, pp. 8472-8476. Pierret, R., Semiconductor Device Fundamentals, Addison-Wesley Publishing Company: Massachusetts, 1996, pg 465.
Radziemski, L. J., Solarz, R. W, and Paisner, J. A., laser Spectroscopy and its Applications, Marcel Dekker: New York, 1987.
Raether, H., Electron Avalanches and Breakdown in Gases, Butterworth & Co: London (1964).
Raizer, Y., Gas Discharge Physics, Springer: New York, 1991.
Rosocha, L., Kim, Y., Anderson, G., Lee, J., and Abbate, S., “Decomposition of Ethane in Atmospheric-Pressure Dielectric – Barrier Discharges: Experiments,” IEEE Trans. Plasma Sci., vol. 24, bo. 6, 2526-2536, Dec. 2006.
Schamiloglu, E., Barker, R. J., Gundersen, M., and Neuber, A. A., “Modern Pulsed Power: Charlie Martin and Beyond,” Proceedings of the IEEE, 92: 1014-1020, July 2004.
Seitzman, J. M., and Hanson, R. K., “Planar Fluorescence Imaging in Gases,” Chapter 6, (pp. 405466) In Experimental Methods for Flows with Combustion, ed. A. Taylor, Academic Press, London, 1993.
Sinibaldi, J., Rodriguez, J. , Chanel, B., Brophy, C., Wang, F., Cathey, C. , and Gundersen, M. A., “Investigation for Transient Plasma for Pulse Detonating Engines,” AIAA 2005-3774.
Smith, P., Transient Electronics, John Wiley & Sons: New Jersey, 2002, pg 238. Starikovskaia, S. M., “Plasma Assisted Ignition and Combustion,” J. Phys. D: Appl. Phys. 39 (2006) R265–R299.
Starikovskia, S. M., Kukaev, E. N., and Kuksin, A. Yu., Comb. and Flame 139, 177- 187 (2004).
Steffens, K. L., and Crosley, D. R., “Vibrational energy transfer in OH A2 between 195 and 295 K,” J. Chem. Phys. Vol. 112, 2000, pp. 9427-9432.
147 Tang, T., Kuthi, A., Wang, F., Cathey, C., and Gundersen, M. A., “Design of 60kV 20ns solid state pulse generator based on magnetic reactor driven diode opening switch,” 27th International Power Modulator Conference 2006, Washington D. C., District of Columbia, May 14-18th, 2006.
Timofeev, L., Milestone 6, Task 2, JHU/APL Subcontract 808713, (2000). Trevisan, C. and Tennyson, J., “Calculated rates for the electron impact dissociation of molecular hydrogen, deuterium and tritium,” Plasma Phys. Control. Fusion 44 (2002).
Van Veldhuizen, E. M., and Rutgers W. R., “Pulsed positive corona streamer propagation and branching,” J. Phys. D: Appl. Phys. 35 (2002).
Van Wie, D., Risha, D. J., Suchomel, C. F., 42nd Aerospace Sciences Meeting, 6th, Reno, Nevada, AIAA-2004-1357 (2004).
Wang, F., Kuthi, A., and Gundersen, M. A., “Compact High Repetition Rate Pulse Generator,” IEEE Transactions on Plasma Science, Volume: 33 , Issue: 4 , Part 1, Aug. 2005, pg 1177-1181.
Wang, F., Kuthi, A., and Gundersen, M.A., IEEE Trans. Plasma Sci. 33 (2005) 1177-1181.
Wang, H., Course Notes AME 599: Combustion Chemistry and Physics, Fall 2006.
148