Chapter Two. Kinetic theory of gases The kinetic theory of gases is of great importance for understanding the physical vapor deposition. Although it is a classic model, its predictions are remarkably accurate and it affords a physical picture of the behavior of a gas. The fundamental assumptions are: The gas contains an enormous number of classical particles. The particles are infinitely hard spheres. All collisions are elastic. There are no forces on the particles except during collisions with each other or with the container. 2.1 The Boltzmann distribution The distribution in energy of the particles (f) is given by the classical Boltzmann distribution function: f ( E ) exp E kT II-1 where k is Boltzmann’s constant, T is the absolute temperature, and E is the energy. It is suitable for identical, distinguishable (thus classical) particles that do not obey the Pauli exclusion principle. The number of particles having a kinetic energy between E and E+dE is therefore given by dN=Nf(E)dE/kT, where N the total number of particles and 1/kT is the necessary normalized factor. dN/N is the ratio of the particle number in this region to the total number. When dE is larger, the value of dN/N could be larger also, thus dN/NdE. Actually the possibility for energy ranged from E1 to E2 is given by dN 1 E2 1 E2 E f E dE exp dE N kT E1 kT E1 kT II-2 Obviously for the whole range (E from 0 to ), this ratio must be 1: 1 f E dE 1 II-3 kT 0 The Boltzmann factor gives the fraction of the total number of particles in the system whose energy is greater than or equal to some specified energy Ea. It is calculated from E dE E II-4 Ea exp kT kT exp kTa . In our theory the energy of a gas particle is entirely translational kinetic energy, given by 2 2 2 mv 2 m(v x v y v z ) E 2 2 II-5 It is frequently useful to work with the subgroup of particles that have essentially the same velocity vector. Consider those particles occupying a certain differential volume of `velocity space’. A probability distribution function may be constructed from the Boltzmann distribution. It is based the probability of having the total energy associated with the velocity v (vx , v y , vz ) : F (v x , v y , vz ) exp m(v v ) / 2kT dv dv dv exp m(v x2 v 2y vz2 ) / 2kT 2 x v 2y 2 z x y . II-6 z This is the probability of occupancy of a differential volume of velocity space (dvxdvydvz), per unit volume of that space, by a gas particle. The probability is F (v x , v y , v z ) dv x dv y dv z . The dominator, a normalizing factor, is (2kT/m)3/2. Example 2.1 Find the height distribution of air particles. Suppose the number of air particles per unit volume at z=0 is n00, then the number in dV=dxdydz at a height z can be expressed as dN=n00exp(-mgz/kT) dxdydz. Thus the number of particles per unit volume at z is n0= dN /dV=n00exp(-mgz/kT) 2.2 Characteristic particle speeds Speed is the magnitude of the velocity vector. To gain the probability distribution function for speed F(v), we use a spherical coordinate. The differential volume in velocity space, instead of an infinitesimal cube, 2 2 2 becomes that of a thin spherical shell given by 4v2dv; vx v y vz becomes simply v2. The new expression is F (v ) exp mv 2 / 2kT exp mv 2 / 2kT v 2 sin ddv . II-7 The probability itself is 4 v2F(v)dv. The dominator is, again, (2 kT/m)3/2. There are several characteristic speeds: The most probable speed (vp) is the speed for which d[4v2F(v)]/dv is zero (ignoring the v=0 case): 1/ 2 2kT vp m . II-8 3/ 2 2 2 The average speed is equal to m / 2kT 0 v exp mv / 2kT 4v dv 1/ 2 8kT vav m . II-9 The root-mean-square speed (vrms) is found from the square root of m / 2kT 3 / 2 0 v 2 exp mv2 / 2kT 4v 2dv : 1/ 2 vrms 3kT m . II-10 The average kinetic energy (Eav) of the particle is 2 mvrms 3kT Eav . 2 2 II-11 It is always true for the Boltzmann distribution that vp<vav<vrms. Example 2.2 Calculate the rms speed and the average kinetic energy of a nitrogen molecule at room temperature. 1/ 2 3 (1.38 10 23 J / K ) 298K vrms 515m / s 27 28 1 . 66 10 kg 28 1.66 1027 kg (515m / s )2 Eav 6 10 21 J 0.04eV . 2 2.3 The ideal gas law The ideal gas law follows directly from the kinetic theory of gases. We can derive it by compute the pressure as the rate of transfer of momentum, per unit area, to the wall of the container. The pressure due to particles with velocity component between vy and vy+dvy is equal to the number of particles striking the wall per unit area, times momentum transferred per collision, per unit time: dP(v y ) nFy dv y v yt 2mv y t II-12 The total pressure then is P 2mn 0 1/ 2 m 2kT mv 2y 2 v y dv y nkT exp 2 kT II-13 It relates the thermodynamic state variables P and T (through n). The more popular form of the ideal gas law is M PV RT C or PV T II-14 where V is the volume of gas, T is absolute temperature (unit: K), C is a constant, M the mass of gas (unit: gram), the mass of gas molecular, and R is the Molar gas constant: R=8.31441 Jmol-1K-1. For mixture of gases The partial pressure of a mixture of gases may be calculated by separating the preceding integrals into individual integrals for each gaseous specie. The total pressure is the sum of the partial pressures. P Pn n II-15 Thus M M M M PV ( Pn )V 1 2 3 ... n RT 2 3 n n 1 Example 2.3 The height dependence of the pressure. From the ideal gas law (Eq. II-13), P=n0kT II-16 According to the example 2.1, we have P=n00kTexp(-mgz/kT) =P0exp(-mgz/kT) =P0exp(-gz/RT), where P0= n00kT is the pressure at the sea level (z=0), is the mass of air. Mean free path The mean free path () is the distance a particle travels, on average, before experiencing a collision. To estimate , we can calculate the survival probability Q(x) that a given particle travels a distance x, from the point of previous collision to the next collision. 1-Q(x) is the probability to have a collision after traveling a distance x. Q decreases with x according to the following expression for the probability (d[1Q(x)]/dx) of particles experiencing a collision while traveling between x and x+dx: dQ QPc dx II-17 where Pc is the average number of collisions per unit distance traveled, a constant. Thus we have Q(x)=exp(-Pcx). II-18 The mean free path is found by calculate the average distance traveled: 1 x ( dQ / dx )dx xPc exp( Pc x )dx . II-19 0 0 Pc Thus the mean free path is the reciprocal of the probability of a collision per unit distance traveled. The collision section We use the assumption that the particles are hard sphere of diameter d. Therefore a particle’s collision section is d2. On average, a particle sweeps out a volume of space at the rate given by d2vav and thus encounters other particles at a rate given by d2vavn. Here we assumed the target particles are stationary. But this is not true and we should use the average relative speed of particles, e.g. Hence the correct collision rate r is r = 2 vav d2n 2 vav . II-20 On the other hand, the collision rate is also the average speed multiplied by the collision probability per unit distance traveled: r=vavPc II-21 Thus Pc= 2 d 2 n II-22 1 2d 2n II-23 and The mean free path is typically determined experimentally from viscosity measurements. Use eq.II-23, the effective particle diameter can be found. It is in the range 2-5 Å for most common gases. d Example 2.3 Estimate the mean free path in N2 gas at one atmosphere and room temperature. The density under these conditions, from ideal gas law, is n 760torr 25 3 2 . 46 10 m 1.38 10 23 J / K 298K 1 2 o 2 3.75 2.46 1025 m 3 o 650 Pressure Suppose the gas container with a volume V contains N molecular of mass m. Thus the number per unit volume is n0=N/V. To simplify the discussion, we separate the gas into several groups, in which the gas particles have the same speed with the same direction. n0 n0i II-24 i Consider the pressure on a small area dA perpendicular to the x-axis. A particle with a speed vi collision with dA. Since the collision is fully elastic, vix changes to –vix while viy and viz keep unchanged. The variation of the momentum is –2mvix, which is the impulse within a period dt, the number of particles can reach dA is n0ivixdtdA. The total impulse of gas particles with speed vi on dA: 2n0imvix2dtdA. Thus the total impulse of all gas particles on dA is dK 2n 0i v ix 0 mvix2 dAdt . II-25 The condition for vix0 is necessary as the particle with vix<0 will not reach dA. In average, the particle with vix0 should be a half of the total particles, therefore we have dK n v ix 0 2 mv 0i ix dAdt II-26 It shows the interaction of gas to dA in time dt. dK PdAdt 2 n mv 0i ixdAdt vix 0 II-27 The pressure P is then: P 1 n0imvix2 dAdt mn0i vix2 dAdt vix 0 vix 0 II-28 There are various systems for the pressure units, including bar, torr, Pa, psi (lb/ft2), kg/cm2, etc. 2.4 Adsorption and condensation The most important factors in physical vapor deposition are vapors and gases. Indeed, the key of PVD is to produce a vapor that will condense on a solid surface (substrate). For vapor creation, if the condensed phase is a liquid, we speak of evaporation; is a solid, of sublimation. The two together are referred to as vaporization and the reverse, condensation. Adsorption of gases Gas particles adsorb because there are stable positions for them on surface. The adsorption energy, EdA, is the depth of the potential well with respect to infinite separation of the molecule from the surface. If Ed is less than ~0.4 eV, the particle (whether a diatomic molecule or something else) is said to be in a state of physisorption, which is an energy range typical of van der Waals bonding. The noble gases physisorb on all surfaces; a molecule probably also physisorb if it is able to adsorb without dissociating. If Ed is large than ~1eV, the particle is said to be chemisorbed, which is a state of relatively strong chemical bonding. Single oxygen atoms chemisorb on virtually all solid surfaces. Mean residence time The mean residence time indicates how long will an individual particle stay adsorbed. It can be estimated as the inverse of its desorption rate: 1 exp Ed / kT , II-29 where is a frequency of vibration perpendicular to the surface, almost always assumed to be ~1013 Hz. Example 2.4 A particle physisorbed with a desorption energy of 0.4 eV has a mean residence time at a room temperature: 1 13 0.58s 10 Hz exp 0.4eV /{( 8.62 10 5 eV / K ) 298K } A desorption energy of 1.2 eV gives a room temperature of 1.9107 s; at 500 K this is reduced to 0.12 s. Condensation of vapors A film will condense on a substrate when there exists a supersaturated vapor above the substrate. For film deposition of pure elements by condensation from a vapor, there is a critical incident flux at the substrate below which no film accumulates. The value of this critical flux increases with the substrate temperature. The following principles are illustrated: For a given substrate temperature, there is critical incident flux above which a film will form, but below which no deposition is obtained. The greater the substrate temperature, the greater the critical incident flux. For compounds whose vapors consists of particles having the stoichiometric composition, stoichiometric films may be obtained by direct vaporization of these compounds. An important thing is to obtain a stoichiometric vapor, which is technique dependent.