Boundary Conditions Boundary Conditions Attempt to define and categorise BCs in financial PDEs Mathematical and financial motivations Unifying framework (Fichera function) One-factor and n-factor examples 2 Background ‘Fuzzy’ area in finance Boundary conditions motivated by financial reasoning BCs may (or may not) be mathematically correct A number of popular choices are in use We justify them 3 Challenges Truncating a semi-infinite domain to a finite domain Imposing BCs on near-field and far-field boundaries Boundaries where no BC are needed (allowed) Dirichlet, Neumann, linearity … 4 Techniques Using Fichera function to determine which boundaries need BCs Determine the kinds of BCs to apply Discretising BCs (for use in FDM) Special cases and ‘nasties’ 5 The Fichera Method Allows us to determine where to place BCs Apply to both elliptic and parabolic PDEs We concentrate on elliptic PDE Of direct relevance to computational finance New development, not widely known 6 Elliptic PDE (1/2) Its quadratic form is non-negative (positive semi-definite) This means that the second-order terms can degenerate at certain points Use the Oleinik/Radkevic theory The application of the Fichera function 7 Domain of interest Unit inward normal Region and Boundary 8 Elliptic PDE Xn @2 u Xn @u Lu ´ ai j + bi + cu = f in @ x @ x @ x i j i i ;j = 1 i= 1 where Xn ai j »i »j ¸ 0 in - [ P i ;j = 1 8 » = (»1 ; : : : ; »n ) ² Rn 9 Remarks Called an equation with non-negative characteristic form Distinguish between characteristic and noncharacteristic boundaries Applicable to elliptic, parabolic and 1st-order hyperbolic PDEs Applicable when the quadratic form is positivedefinite as well Subsumes Friedrichs’ theory in hyperbolic 10 case? Boundary Types n P 3 On = x² P ¡ P : Pn i ;j = 1 o ai j Ài Àj > 0 P examine Fichera function à ! Xn Xn @ai k b´ bi ¡ Ài @ x k i= 1 k= 1 3 11 Choices On P P 3 0 : b= 0 1 : b> 0 2 : b< 0 P P P ¡ P ´ P 0[ de¯ne P 1[ P 2[ P 3 12 Example: Hyperbolic PDE (1/2) ¡ @u @x = f in - = (0; L) £ (0; 1) b= ¡ À1 @u a @x + @u b@y = f in (0; 1) £ (0; 1) 13 Example: Hyperbolic PDE (2/2) y 1 x L 14 Example: Hyperbolic PDE a; b> 0 y Ficherab= aÀ1 + bÀ2 x 15 Example: CIR Model Discussed in FDM book, page 281 What happens on r = 0? We discuss the application of the Fichera method Reproduce well-known results by different means 16 CIR PDE @B @t + 1 2 @2 B 2 ¾ r @r 2 B + (a ¡ br) @ @r ¡ rB = 0 Fichera b = (a ¡ br ) + 2¾ § 2 : b< 0 ! ¾> § 0 : b= 0 ! ¾= § 1 : b> 0 ! ¾< p p p 2a 2a 2a (No BC needed) 17 Convertible Bonds Two-factor model (S, r) Use Ito to find the PDE 18 Two-factor PDE (1/2) 19 Two-factor PDE (2/2) V S 20 Asian Options Two-factor model (S, A) Diffusion term missing in the A direction Determine the well-posedness of problem Write PDE in (x,y) form 21 PDE for Asian ¡ @u @t y= 1 2 2 @2 u 2 ¾ x @x 2 + 1 T Rt 0 x @u T @y ¡ ru = 0 x(t)dt Fichera b= (r x ¡ = x(r ¡ u + rx@ @x + 2 ¾ )À1 + x 2 ¾ x)À1 + T À2 x T À2 22 PDE Formulation I (1/2) Lu = f in u = g1 on § 2 u ®u + ¯ @ @l = g2 on § 3 (®1¯ ¸ 0; l is a direction) 23 PDE Formulation I (2/2) y x 24 Special Case Lu = f in u = g3 on § 2 [ § 3 25 Example: Skew PDE Pure diffusion degenerate PDE Used in conjunction with SABR model Critical value of beta (thanks to Alan Lewis) 26 PDE y 1 2¯ 2 @2 u 2 S y @S2 + 1 2 @2 u 2 y @y 2 =0 S 27 Fichera Function µ Ficherab= ¡ § 2i= 1 bi ¡ =¡ ai k 2 2 @ § i = 1§ k= 1 Ài @xk =¡ ai i 2 @ §i= 1 @xi =¡ ¡ ai k 2 @ § k= 1 @xk ¶ Ài Ài ¯S2¯¡ 1y2À1 + yÀ2 ¢ 28 Boundaries ¡ ¯> 1 2 ¢ ¡ 1 : b = 0(§ 0 ) ¡ 2 : b = 0(§ 0 ) 9 ¡3 > > = belong to § 3 ¡4 > > ; 29
© Copyright 2026 Paperzz