| 1 | package Physics::Ellipsometry::VASE::EMA; | ||
| 2 | use strict; | ||
| 3 | use warnings; | ||
| 4 | use PDL; | ||
| 5 | use PDL::NiceSlice; | ||
| 6 | use Exporter 'import'; | ||
| 7 | |||
| 8 | our @EXPORT_OK = qw(ema_linear ema_bruggeman ema_maxwell_garnett ema_looyenga); | ||
| 9 | |||
| 10 | our $VERSION = '1.04'; | ||
| 11 | |||
| 12 | =encoding utf8 | ||
| 13 | |||
| 14 | =head1 NAME | ||
| 15 | |||
| 16 | Physics::Ellipsometry::VASE::EMA - Effective Medium Approximation mixing | ||
| 17 | models for composite thin films | ||
| 18 | |||
| 19 | =head1 SYNOPSIS | ||
| 20 | |||
| 21 | use PDL; | ||
| 22 | use Physics::Ellipsometry::VASE::EMA qw(ema_linear ema_bruggeman | ||
| 23 | ema_maxwell_garnett ema_looyenga); | ||
| 24 | use Physics::Ellipsometry::VASE::Dispersion qw(cauchy_nk); | ||
| 25 | |||
| 26 | my $lambda = sequence(500) * 2 + 300; # 300–1298 nm | ||
| 27 | |||
| 28 | # Dielectric functions of two constituent materials | ||
| 29 | my ($n_a, $k_a) = cauchy_nk($lambda, 1.46, 0.003, 0.0); # SiO2 | ||
| 30 | my ($n_b, $k_b) = (ones($lambda), zeroes($lambda)); # void (air) | ||
| 31 | |||
| 32 | my $eps_a = ($n_a + i() * $k_a) ** 2; | ||
| 33 | my $eps_b = ($n_b + i() * $k_b) ** 2; | ||
| 34 | |||
| 35 | # 30% porosity via Bruggeman mixing | ||
| 36 | my $eps_eff = ema_bruggeman($eps_a, $eps_b, 0.30); | ||
| 37 | my $N_eff = sqrt($eps_eff); | ||
| 38 | my $n_eff = $N_eff->re; | ||
| 39 | my $k_eff = $N_eff->im->abs; | ||
| 40 | |||
| 41 | printf "n(500 nm) = %.3f (bulk SiO2 ≈ 1.46)\n", | ||
| 42 | $n_eff->at(100); | ||
| 43 | |||
| 44 | =head1 DESCRIPTION | ||
| 45 | |||
| 46 | Many thin films encountered in ellipsometry are not homogeneous | ||
| 47 | single-phase materials. Surface roughness, porosity, inter-diffusion | ||
| 48 | layers, nano-composites, and graded interfaces are all cases where | ||
| 49 | the optical response arises from a B<mixture> of two (or more) | ||
| 50 | constituent materials at a scale smaller than the wavelength of light. | ||
| 51 | |||
| 52 | B<Effective Medium Approximation> (EMA) theories provide the dielectric | ||
| 53 | function of such a composite given the dielectric functions of the | ||
| 54 | constituents and their volume fractions. The choice of EMA model | ||
| 55 | depends on the microstructure: | ||
| 56 | |||
| 57 | =over 4 | ||
| 58 | |||
| 59 | =item * | ||
| 60 | |||
| 61 | B<Linear> — simplest volume-weighted average; reasonable when the | ||
| 62 | constituents have similar dielectric functions. | ||
| 63 | |||
| 64 | =item * | ||
| 65 | |||
| 66 | B<Bruggeman> — symmetric in both constituents; no distinction between | ||
| 67 | host and inclusion. Best for random mixtures, porous films, and | ||
| 68 | surface roughness layers. | ||
| 69 | |||
| 70 | =item * | ||
| 71 | |||
| 72 | B<Maxwell-Garnett> — asymmetric; treats one material as isolated | ||
| 73 | spherical inclusions in a continuous host matrix. Best when the | ||
| 74 | volume fraction of the inclusion is small (< 30%). | ||
| 75 | |||
| 76 | =back | ||
| 77 | |||
| 78 | All three functions operate on the B<dielectric function> ε = N² | ||
| 79 | (not on n and k directly), because the mixing rules are derived in | ||
| 80 | terms of ε. | ||
| 81 | |||
| 82 | =head2 Surface roughness modelling | ||
| 83 | |||
| 84 | A common use is modelling surface roughness as a 50/50 Bruggeman mix | ||
| 85 | of the underlying film material and void: | ||
| 86 | |||
| 87 | my $eps_rough = ema_bruggeman($eps_film, $eps_void, 0.50); | ||
| 88 | |||
| 89 | This roughness layer is then included in the TMM stack with a | ||
| 90 | thickness on the order of the RMS roughness. | ||
| 91 | |||
| 92 | =head1 FUNCTIONS | ||
| 93 | |||
| 94 | =head2 ema_linear | ||
| 95 | |||
| 96 | my $eps_eff = ema_linear($eps_a, $eps_b, $vf); | ||
| 97 | |||
| 98 | B<Linear (volume-weighted) mixing>: | ||
| 99 | |||
| 100 | ε_eff = (1 − f) · ε_a + f · ε_b | ||
| 101 | |||
| 102 | where C<$vf> (f) is the volume fraction of material B. This is the | ||
| 103 | simplest mixing rule and does not account for local-field effects. | ||
| 104 | It is exact when the two phases form a laminar (layered) structure | ||
| 105 | with the electric field parallel to the layers. | ||
| 106 | |||
| 107 | # 20% TiO2 in SiO2 matrix (linear blend) | ||
| 108 | my $eps_mix = ema_linear($eps_sio2, $eps_tio2, 0.20); | ||
| 109 | |||
| 110 | =head2 ema_bruggeman | ||
| 111 | |||
| 112 | my $eps_eff = ema_bruggeman($eps_a, $eps_b, $vf); | ||
| 113 | |||
| 114 | B<Bruggeman EMA> solves the self-consistent equation: | ||
| 115 | |||
| 116 | f_a · (ε_a − ε_eff)/(ε_a + 2·ε_eff) | ||
| 117 | + f_b · (ε_b − ε_eff)/(ε_b + 2·ε_eff) = 0 | ||
| 118 | |||
| 119 | For a two-component system this has the closed-form solution: | ||
| 120 | |||
| 121 | ε_eff = ( b + sqrt(b² + 8·ε_a·ε_b) ) / 4 | ||
| 122 | |||
| 123 | where b = (3f − 1)·ε_b + (2 − 3f)·ε_a | ||
| 124 | |||
| 125 | The implementation selects the root with positive real part to ensure | ||
| 126 | a physically meaningful result. | ||
| 127 | |||
| 128 | Bruggeman EMA is B<symmetric>: neither material is privileged as host | ||
| 129 | or inclusion. This makes it the standard choice for: | ||
| 130 | |||
| 131 | =over 4 | ||
| 132 | |||
| 133 | =item * Surface roughness layers (50% film / 50% void) | ||
| 134 | |||
| 135 | =item * Porous films (variable void fraction) | ||
| 136 | |||
| 137 | =item * Interdiffusion or intermixed interfaces | ||
| 138 | |||
| 139 | =item * Cermet and nano-composite films | ||
| 140 | |||
| 141 | =back | ||
| 142 | |||
| 143 | # Surface roughness: 50% Ta2O5 + 50% void | ||
| 144 | my $eps_void = ones($lambda) + i() * zeroes($lambda); # ε = 1 | ||
| 145 | my $eps_rough = ema_bruggeman($eps_ta2o5, $eps_void, 0.50); | ||
| 146 | |||
| 147 | # Include as a thin layer in the TMM stack | ||
| 148 | my ($psi, $delta) = psi_delta($lambda, $theta, | ||
| 149 | [$N_air, sqrt($eps_rough), $N_ta2o5, $N_si], | ||
| 150 | [2.0, 100.0], # 2 nm roughness, 100 nm film | ||
| 151 | ); | ||
| 152 | |||
| 153 | =head2 ema_maxwell_garnett | ||
| 154 | |||
| 155 | my $eps_eff = ema_maxwell_garnett($eps_host, $eps_incl, $vf); | ||
| 156 | |||
| 157 | B<Maxwell-Garnett EMA> treats material B as dilute spherical inclusions | ||
| 158 | embedded in a continuous host matrix of material A: | ||
| 159 | |||
| 160 | ε_eff = ε_a · (ε_b + 2ε_a + 2f·(ε_b − ε_a)) | ||
| 161 | / (ε_b + 2ε_a − f·(ε_b − ε_a)) | ||
| 162 | |||
| 163 | This model is B<asymmetric>: swapping host and inclusion gives a | ||
| 164 | different result. It is most accurate when the inclusion volume | ||
| 165 | fraction is below about 30% and the inclusions are well separated. | ||
| 166 | |||
| 167 | Typical applications: | ||
| 168 | |||
| 169 | =over 4 | ||
| 170 | |||
| 171 | =item * Metal nanoparticles in a dielectric (Au in SiO2) | ||
| 172 | |||
| 173 | =item * Quantum dots in a polymer matrix | ||
| 174 | |||
| 175 | =item * Dilute dopant phases | ||
| 176 | |||
| 177 | =back | ||
| 178 | |||
| 179 | # 5% Au nanoparticles in SiO2 host | ||
| 180 | my $eps_composite = ema_maxwell_garnett($eps_sio2, $eps_au, 0.05); | ||
| 181 | |||
| 182 | =head1 CONVERTING BETWEEN ε AND N | ||
| 183 | |||
| 184 | The functions accept and return the complex dielectric function | ||
| 185 | ε = ε₁ + iε₂. To convert to/from the complex refractive index | ||
| 186 | N = n + ik: | ||
| 187 | |||
| 188 | # N → ε | ||
| 189 | my $eps = ($n + i() * $k) ** 2; | ||
| 190 | |||
| 191 | # ε → N | ||
| 192 | my $N_complex = sqrt($eps); | ||
| 193 | my $n = $N_complex->re; | ||
| 194 | my $k = $N_complex->im->abs; | ||
| 195 | |||
| 196 | =head2 ema_looyenga | ||
| 197 | |||
| 198 | my $eps_eff = ema_looyenga($eps_a, $eps_b, $volume_fraction_b); | ||
| 199 | |||
| 200 | The B<Looyenga mixing rule> (1965) uses cube-root averaging: | ||
| 201 | |||
| 202 | eps_eff^(1/3) = (1-f)*eps_a^(1/3) + f*eps_b^(1/3) | ||
| 203 | |||
| 204 | This formula treats both components symmetrically (unlike Maxwell-Garnett) | ||
| 205 | and is simpler than solving the self-consistent Bruggeman equation. It | ||
| 206 | works well for random mixtures where neither material forms a clear | ||
| 207 | host-inclusion geometry. | ||
| 208 | |||
| 209 | B<When to use Looyenga:> | ||
| 210 | |||
| 211 | =over 4 | ||
| 212 | |||
| 213 | =item - Random mixtures with comparable volume fractions of both phases | ||
| 214 | |||
| 215 | =item - Porous materials (e.g., porous Si, aerogels) | ||
| 216 | |||
| 217 | =item - Quick estimate without iterative solving | ||
| 218 | |||
| 219 | =back | ||
| 220 | |||
| 221 | B<Example:> | ||
| 222 | |||
| 223 | # Porous silicon (50% Si / 50% void) | ||
| 224 | my $eps_Si = pdl(15.0) + i() * pdl(0.2); | ||
| 225 | my $eps_void = pdl(1.0) + i() * pdl(0.0); | ||
| 226 | my $eps_eff = ema_looyenga($eps_void, $eps_Si, 0.5); | ||
| 227 | my $n_eff = sqrt($eps_eff)->re; # ~2.35 | ||
| 228 | |||
| 229 | # Compare with Bruggeman for same system | ||
| 230 | my $eps_brug = ema_bruggeman($eps_void, $eps_Si, 0.5); # ~2.50 | ||
| 231 | |||
| 232 | =head1 SEE ALSO | ||
| 233 | |||
| 234 | L<Physics::Ellipsometry::VASE>, | ||
| 235 | L<Physics::Ellipsometry::VASE::TMM>, | ||
| 236 | L<Physics::Ellipsometry::VASE::Dispersion>, | ||
| 237 | L<Physics::Ellipsometry::VASE::Materials> | ||
| 238 | |||
| 239 | D.E. Aspnes, "Local-field effects and effective-medium theory: A | ||
| 240 | microscopic perspective", I<Am. J. Phys.> B<50>, 704 (1982). | ||
| 241 | |||
| 242 | H. Fujiwara, I<Spectroscopic Ellipsometry: Principles and Applications>, | ||
| 243 | John Wiley E<amp> Sons, 2007, Chapter 5. | ||
| 244 | |||
| 245 | =cut | ||
| 246 | |||
| 247 | # Linear (volume-weighted) mixing: ε_eff = (1-f)ε_a + f·ε_b | ||
| 248 | sub ema_linear { | ||
| 249 | my ($eps_a, $eps_b, $vf) = @_; | ||
| 250 | return (1.0 - $vf) * $eps_a + $vf * $eps_b; | ||
| 251 | } | ||
| 252 | |||
| 253 | # Bruggeman EMA: f_a·(ε_a - ε_eff)/(ε_a + 2ε_eff) + f_b·(ε_b - ε_eff)/(ε_b + 2ε_eff) = 0 | ||
| 254 | # Solved analytically for 2-component mixtures: | ||
| 255 | # ε_eff = (b ± sqrt(b² + 8ε_aε_b)) / 4 | ||
| 256 | # where b = (3f_b - 1)ε_b + (3f_a - 1)ε_a = (3vf - 1)ε_b + (2 - 3vf)ε_a | ||
| 257 | sub ema_bruggeman { | ||
| 258 | my ($eps_a, $eps_b, $vf) = @_; | ||
| 259 | my $fa = 1.0 - $vf; | ||
| 260 | |||
| 261 | # Quadratic solution for 2-component Bruggeman | ||
| 262 | my $b = (3*$vf - 1) * $eps_b + (3*$fa - 1) * $eps_a; | ||
| 263 | my $discriminant = $b**2 + 8 * $eps_a * $eps_b; | ||
| 264 | |||
| 265 | # Take the root with positive real part | ||
| 266 | my $sqrt_disc = sqrt($discriminant + i()*0); | ||
| 267 | my $eps_eff = ($b + $sqrt_disc) / 4.0; | ||
| 268 | |||
| 269 | # Ensure physical result (positive real part of ε) | ||
| 270 | my $eps_eff2 = ($b - $sqrt_disc) / 4.0; | ||
| 271 | my $use_alt = ($eps_eff->re < 0) & ($eps_eff2->re > 0); | ||
| 272 | if (ref $use_alt && $use_alt->any) { | ||
| 273 | my $idx = which($use_alt); | ||
| 274 | $eps_eff->index($idx) .= $eps_eff2->index($idx); | ||
| 275 | } | ||
| 276 | |||
| 277 | return $eps_eff; | ||
| 278 | } | ||
| 279 | |||
| 280 | # Maxwell-Garnett EMA: inclusion (b) in host matrix (a) | ||
| 281 | # ε_eff = ε_a · (ε_b + 2ε_a + 2f(ε_b - ε_a)) / (ε_b + 2ε_a - f(ε_b - ε_a)) | ||
| 282 | sub ema_maxwell_garnett { | ||
| 283 | my ($eps_a, $eps_b, $vf) = @_; | ||
| 284 | |||
| 285 | my $numer = $eps_b + 2*$eps_a + 2*$vf*($eps_b - $eps_a); | ||
| 286 | my $denom = $eps_b + 2*$eps_a - $vf*($eps_b - $eps_a); | ||
| 287 | |||
| 288 | return $eps_a * $numer / $denom; | ||
| 289 | } | ||
| 290 | |||
| 291 | # Looyenga EMA: ε_eff^(1/3) = (1-f)·ε_a^(1/3) + f·ε_b^(1/3) | ||
| 292 | # Valid for random mixtures where neither component is clearly host/inclusion. | ||
| 293 | # More symmetric than Maxwell-Garnett; simpler than Bruggeman. | ||
| 294 | sub ema_looyenga { | ||
| 295 | my ($eps_a, $eps_b, $vf) = @_; | ||
| 296 | my $fa = 1.0 - $vf; | ||
| 297 | |||
| 298 | # Cube root of complex numbers: |z|^(1/3) * exp(i*arg(z)/3) | ||
| 299 | my $eps_a_third = _complex_cbrt($eps_a); | ||
| 300 | my $eps_b_third = _complex_cbrt($eps_b); | ||
| 301 | |||
| 302 | my $mix = $fa * $eps_a_third + $vf * $eps_b_third; | ||
| 303 | |||
| 304 | # Cube the result | ||
| 305 | return $mix**3; | ||
| 306 | } | ||
| 307 | |||
| 308 | # Complex cube root: z^(1/3) | ||
| 309 | sub _complex_cbrt { | ||
| 310 | my ($z) = @_; | ||
| 311 | my $mag = abs($z); | ||
| 312 | my $arg = carg($z); | ||
| 313 | return $mag**(1.0/3.0) * (cos($arg/3.0) + i() * sin($arg/3.0)); | ||
| 314 | } | ||
| 315 | |||
| 316 | 1; |