Amoebas, ArchTrop and all that jazz Roman Kogan March 24, 2014 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 1 / 18 Overview of the talk Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Overview of the talk Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Overview of the talk Motivation and defintions Newton Polyhedron Amoeba ArchTrop Geometric results Complexity results Gnarly details Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Overview of the talk Motivation and defintions Newton Polyhedron Amoeba ArchTrop Geometric results Complexity results Gnarly details Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Overview of the talk Motivation and defintions Newton Polyhedron Amoeba ArchTrop Geometric results Complexity results Gnarly details Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Overview of the talk Motivation and defintions Newton Polyhedron Amoeba ArchTrop Geometric results Complexity results Gnarly details Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Overview of the talk Motivation and defintions Newton Polyhedron Amoeba ArchTrop Geometric results Complexity results Gnarly details Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Overview of the talk Motivation and defintions Newton Polyhedron Amoeba ArchTrop Geometric results Complexity results Gnarly details Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Overview of the talk Motivation and defintions Newton Polyhedron Amoeba ArchTrop Geometric results Complexity results Gnarly details Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 2 / 18 Motivation Motivation: say something about zeros without evaluating FTA: # zeros ≤ degree Descartes: # real roots ≤ # coefficient sign changes Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 3 / 18 Motivation Motivation: say something about zeros without evaluating FTA: # zeros ≤ degree Descartes: # real roots ≤ # coefficient sign changes Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 3 / 18 Motivation Motivation: say something about zeros without evaluating FTA: # zeros ≤ degree Descartes: # real roots ≤ # coefficient sign changes Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 3 / 18 Newton Polyhedron Newton polygon = generalization of the degree Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 4 / 18 Newton Polyhedron Newton polygon = generalization of the degree ai ∈ Zn Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 4 / 18 Newton Polyhedron Newton polygon = generalization of the degree ai ∈ Zn τ X p(x) = ci xai i=1 = τ X ai a ci x1 1 . . . xnin i=1 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 4 / 18 Newton Polyhedron Newton polygon = generalization of the degree ai ∈ Zn τ X p(x) = ci xai i=1 = τ X ai a ci x1 1 . . . xnin i=1 Newt(p) = Conv{a1 , a2 , . . . , aτ } Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 4 / 18 Newton Polyhedron Newton polygon example p(x) = x2 y + y 2 + y 4 + x4 y 4 + x3 y 3 ; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 5 / 18 Newton Polyhedron Newton polygon example p(x) = x2 y + y 2 + y 4 + x4 y 4 + x3 y 3 ; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 5 / 18 Newton Polyhedron Newton polygon example p(x) = x2 y + y 2 + y 4 + x4 y 4 + x3 y 3 ; Newt(p) = Conv{(2, 1), (0, 2), (0, 4), (4, 4), (3, 3)} Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 5 / 18 Newton Polyhedron Newton polygon example p(x) = x2 y + y 2 + y 4 + x4 y 4 + x3 y 3 ; Newt(p) = Conv{(2, 1), (0, 2), (0, 4), (4, 4), (3, 3)} Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 5 / 18 Amoeba Q: What does the Newton polyhedron tell about the geometry of roots? Amoeba: zero set on log paper Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 6 / 18 Amoeba Q: What does the Newton polyhedron tell about the geometry of roots? Amoeba: zero set on log paper Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 6 / 18 Amoeba Q: What does the Newton polyhedron tell about the geometry of roots? Amoeba: zero set on log paper Z(p) := {(z1 , . . . , zn ) ∈ (C − 0)n : p(z1 , . . . , zn ) = 0} Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 6 / 18 Amoeba Q: What does the Newton polyhedron tell about the geometry of roots? Amoeba: zero set on log paper Z(p) := {(z1 , . . . , zn ) ∈ (C − 0)n : p(z1 , . . . , zn ) = 0} Amoeba(p) := {(log |z1 |, . . . , log |zn |) : (z1 , . . . , zn ) ∈ Z(p)} Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 6 / 18 Amoeba Amoeba(p) := {(log |z1 |, . . . , log |zn |) : (z1 , . . . , zn ) ∈ Z(p)} A simple example: p(x) = x + y + 1; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 7 / 18 Amoeba Amoeba(p) := {(log |z1 |, . . . , log |zn |) : (z1 , . . . , zn ) ∈ Z(p)} A simple example: p(x) = x + y + 1; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 7 / 18 Amoeba Amoeba(p) := {(log |z1 |, . . . , log |zn |) : (z1 , . . . , zn ) ∈ Z(p)} A simple example: p(x) = x + y + 1; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 7 / 18 Amoeba Amoeba(p) := {(log |z1 |, . . . , log |zn |) : (z1 , . . . , zn ) ∈ Z(p)} A simple example: p(x) = x + y + 1; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 7 / 18 Amoeba Amoeba: more complicated example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 ; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 8 / 18 Amoeba Amoeba: more complicated example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 ; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 8 / 18 Amoeba Amoeba: more complicated example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 ; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 8 / 18 Amoeba Amoeba: more complicated example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 ; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 8 / 18 Amoeba Amoeba: more complicated example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 ; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 8 / 18 ArchTrop The missing link: Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 9 / 18 ArchTrop The missing link: ai ∈ Zn Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 9 / 18 ArchTrop The missing link: ai ∈ Zn τ X p(x) = ci xai i=1 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 9 / 18 ArchTrop The missing link: ai ∈ Zn τ X p(x) = ci xai i=1 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 9 / 18 ArchTrop The missing link: ai ∈ Zn τ X p(x) = ci xai i=1 ArchTrop := {(log |x1 |, . . . , log |xn |) ∈ Rn : max |ci xai | is attained at least twice} i Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 9 / 18 ArchTrop The missing link: ai ∈ Zn τ X p(x) = ci xai i=1 ArchTrop := {(log |x1 |, . . . , log |xn |) ∈ Rn : max |ci xai | is attained at least twice} i log x ∈ ArchTrop ⇔ norms of at least two monomials of p(x) are equal and larger than norms of other monomials at x Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 9 / 18 ArchTrop: simple example ArchTrop := {log x ∈ Rn : max |xai | is attained at least twice} i Let p(x) = x + y + 1. Then ArchTrop = union of solutions to ( ( ( log |x| = log |y| log |x| = log |1| log |y| = log |y| log |x| ≥ log |1| log |1| ≥ log |y| log |1| ≥ log |x| Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 10 / 18 ArchTrop: simple example ArchTrop := {log x ∈ Rn : max |xai | is attained at least twice} i Let p(x) = x + y + 1. Then ArchTrop = union of solutions to ( ( ( log |x| = log |y| log |x| = log |1| log |y| = log |y| log |x| ≥ log |1| log |1| ≥ log |y| log |1| ≥ log |x| Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 10 / 18 ArchTrop: simple example ArchTrop := {log x ∈ Rn : max |xai | is attained at least twice} i Let p(x) = x + y + 1. Then ArchTrop = union of solutions to ( ( ( log |x| = log |y| log |x| = log |1| log |y| = log |y| log |x| ≥ log |1| log |1| ≥ log |y| log |1| ≥ log |x| Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 10 / 18 ArchTrop: simple example ArchTrop := {log x ∈ Rn : max |xai | is attained at least twice} i Let p(x) = x + y + 1. Then ArchTrop = union of solutions to ( ( ( log |x| = log |y| log |x| = log |1| log |y| = log |y| log |x| ≥ log |1| log |1| ≥ log |y| log |1| ≥ log |x| Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 10 / 18 ArchTrop Example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 11 / 18 ArchTrop Example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 11 / 18 ArchTrop Example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 11 / 18 ArchTrop Example p(x) = x2 y + y 2 + 3x2 y 3 + y 4 + x4 y 4 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 11 / 18 ArchNewt and ArchTrop The missing link: duality Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 12 / 18 ArchNewt and ArchTrop The missing link: duality ai ∈ Zn Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 12 / 18 ArchNewt and ArchTrop The missing link: duality ai ∈ Zn τ X p(x) = ci xai i=1 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 12 / 18 ArchNewt and ArchTrop The missing link: duality ai ∈ Zn τ X p(x) = ci xai i=1 ArchNewt(p) := Conv{(ai , − log |ci |) : i = 1..τ } ⊂ Rn+1 Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 12 / 18 ArchNewt and ArchTrop The missing link: duality ai ∈ Zn τ X p(x) = ci xai i=1 ArchNewt(p) := Conv{(ai , − log |ci |) : i = 1..τ } ⊂ Rn+1 ArchTrop := {v ∈ Rn : (v, −1) is maximal on a +dim lower face of ArchNewt} Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 12 / 18 ArchNewt and ArchTrop The missing link: duality ai ∈ Zn τ X p(x) = ci xai i=1 ArchNewt(p) := Conv{(ai , − log |ci |) : i = 1..τ } ⊂ Rn+1 ArchTrop := {v ∈ Rn : (v, −1) is maximal on a +dim lower face of ArchNewt} (v is an outer normal to a 1+-dimensional cell ) Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 12 / 18 ArchNewt and ArchTrop The missing link: duality ai ∈ Zn τ X p(x) = ci xai i=1 ArchNewt(p) := Conv{(ai , − log |ci |) : i = 1..τ } ⊂ Rn+1 ArchTrop := {v ∈ Rn : (v, −1) is maximal on a +dim lower face of ArchNewt} (v is an outer normal to a 1+-dimensional cell ) Easy to compute! Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 12 / 18 Hausdorff distance Q: How close is Amoba to ArchTrop? Hausdorff distance - measure of closeness: ∆X, Y = min{sup inf d(x, y), sup inf d(x, y)} x∈X y∈Y Roman Kogan y∈Y x∈X Amoebas, ArchTrop and all that jazz March 24, 2014 13 / 18 Hausdorff distance Q: How close is Amoba to ArchTrop? Hausdorff distance - measure of closeness: ∆X, Y = min{sup inf d(x, y), sup inf d(x, y)} x∈X y∈Y Roman Kogan y∈Y x∈X Amoebas, ArchTrop and all that jazz March 24, 2014 13 / 18 Hausdorff distance Q: How close is Amoba to ArchTrop? Hausdorff distance - measure of closeness: ∆X, Y = min{sup inf d(x, y), sup inf d(x, y)} x∈X y∈Y y∈Y x∈X X Y Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 13 / 18 Hausdorff distance Q: How close is Amoba to ArchTrop? Hausdorff distance - measure of closeness: ∆X, Y = min{sup inf d(x, y), sup inf d(x, y)} x∈X y∈Y y∈Y x∈X X Y E.g. above: supx∈X inf y∈Y d(x, y) is small: X close to Y ; supy∈Y inf x∈X d(x, y) is big: Y not close to X. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 13 / 18 Main Theorem Theorem For all p with τ terms, ∆(Amoeba(p), ArchTrop(p) ≤ (2τ − 3) log(τ − 1)) Theorem Furthermore: if vertices of support of p = vertices of Newt(p), then ArchTrop(p) ⊂ Amoeba(p) and they are homotopy equivalent; supx∈Amoeba inf y∈ArchTrop d(x, y) ≤ log(τ − 1), so there is hope for improvement! For p(x1 , . . . , xk ) = (x1 + 1)τ −k + x2 + . . . + xk , ∆(Amoeba(p), ArchTrop(p) ≥ log(τ − k)). These bounds in number of of terms only are new. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 14 / 18 Main Theorem Theorem For all p with τ terms, ∆(Amoeba(p), ArchTrop(p) ≤ (2τ − 3) log(τ − 1)) Theorem Furthermore: if vertices of support of p = vertices of Newt(p), then ArchTrop(p) ⊂ Amoeba(p) and they are homotopy equivalent; supx∈Amoeba inf y∈ArchTrop d(x, y) ≤ log(τ − 1), so there is hope for improvement! For p(x1 , . . . , xk ) = (x1 + 1)τ −k + x2 + . . . + xk , ∆(Amoeba(p), ArchTrop(p) ≥ log(τ − k)). These bounds in number of of terms only are new. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 14 / 18 Main Theorem Theorem For all p with τ terms, ∆(Amoeba(p), ArchTrop(p) ≤ (2τ − 3) log(τ − 1)) Theorem Furthermore: if vertices of support of p = vertices of Newt(p), then ArchTrop(p) ⊂ Amoeba(p) and they are homotopy equivalent; supx∈Amoeba inf y∈ArchTrop d(x, y) ≤ log(τ − 1), so there is hope for improvement! For p(x1 , . . . , xk ) = (x1 + 1)τ −k + x2 + . . . + xk , ∆(Amoeba(p), ArchTrop(p) ≥ log(τ − k)). These bounds in number of of terms only are new. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 14 / 18 Main Theorem Theorem For all p with τ terms, ∆(Amoeba(p), ArchTrop(p) ≤ (2τ − 3) log(τ − 1)) Theorem Furthermore: if vertices of support of p = vertices of Newt(p), then ArchTrop(p) ⊂ Amoeba(p) and they are homotopy equivalent; supx∈Amoeba inf y∈ArchTrop d(x, y) ≤ log(τ − 1), so there is hope for improvement! For p(x1 , . . . , xk ) = (x1 + 1)τ −k + x2 + . . . + xk , ∆(Amoeba(p), ArchTrop(p) ≥ log(τ − k)). These bounds in number of of terms only are new. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 14 / 18 Main Theorem Theorem For all p with τ terms, ∆(Amoeba(p), ArchTrop(p) ≤ (2τ − 3) log(τ − 1)) Theorem Furthermore: if vertices of support of p = vertices of Newt(p), then ArchTrop(p) ⊂ Amoeba(p) and they are homotopy equivalent; supx∈Amoeba inf y∈ArchTrop d(x, y) ≤ log(τ − 1), so there is hope for improvement! For p(x1 , . . . , xk ) = (x1 + 1)τ −k + x2 + . . . + xk , ∆(Amoeba(p), ArchTrop(p) ≥ log(τ − k)). These bounds in number of of terms only are new. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 14 / 18 Main Theorem Theorem For all p with τ terms, ∆(Amoeba(p), ArchTrop(p) ≤ (2τ − 3) log(τ − 1)) Theorem Furthermore: if vertices of support of p = vertices of Newt(p), then ArchTrop(p) ⊂ Amoeba(p) and they are homotopy equivalent; supx∈Amoeba inf y∈ArchTrop d(x, y) ≤ log(τ − 1), so there is hope for improvement! For p(x1 , . . . , xk ) = (x1 + 1)τ −k + x2 + . . . + xk , ∆(Amoeba(p), ArchTrop(p) ≥ log(τ − k)). These bounds in number of of terms only are new. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 14 / 18 Complexity results Complexity is in terms of problem bitsize. Deciding whether 0 ∈ Amoeba(p), p ∈ Z[x] is NP-hard [Plaisted, 84]; For fixed n, if f has rational logs, computing Archtrop(p) is in P. For fixed n, if f has rational coefficients, deciding q ∈ Archtrop(p) is in P. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 15 / 18 Complexity results Complexity is in terms of problem bitsize. Deciding whether 0 ∈ Amoeba(p), p ∈ Z[x] is NP-hard [Plaisted, 84]; For fixed n, if f has rational logs, computing Archtrop(p) is in P. For fixed n, if f has rational coefficients, deciding q ∈ Archtrop(p) is in P. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 15 / 18 Complexity results Complexity is in terms of problem bitsize. Deciding whether 0 ∈ Amoeba(p), p ∈ Z[x] is NP-hard [Plaisted, 84]; For fixed n, if f has rational logs, computing Archtrop(p) is in P. For fixed n, if f has rational coefficients, deciding q ∈ Archtrop(p) is in P. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 15 / 18 Complexity results Complexity is in terms of problem bitsize. Deciding whether 0 ∈ Amoeba(p), p ∈ Z[x] is NP-hard [Plaisted, 84]; For fixed n, if f has rational logs, computing Archtrop(p) is in P. For fixed n, if f has rational coefficients, deciding q ∈ Archtrop(p) is in P. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 15 / 18 Gnarly details Proof outline: Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 16 / 18 Gnarly details Proof outline: One-dimensional case: Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 16 / 18 Gnarly details Proof outline: One-dimensional case: Use basic properties to reduce to a special case; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 16 / 18 Gnarly details Proof outline: One-dimensional case: Use basic properties to reduce to a special case; Use elementary and classical results to show that sup inf x∈Archtrop y∈Amoeba Roman Kogan d(x, y) ≤ 1 + log(τ − 1) Amoebas, ArchTrop and all that jazz March 24, 2014 16 / 18 Gnarly details Proof outline: One-dimensional case: Use basic properties to reduce to a special case; Use elementary and classical results to show that sup inf x∈Archtrop y∈Amoeba d(x, y) ≤ 1 + log(τ − 1) In many dimensions, specialize to one to generalize. Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 16 / 18 Gnarly details Proof outline: One-dimensional case: Use basic properties to reduce to a special case; Use elementary and classical results to show that sup inf x∈Archtrop y∈Amoeba d(x, y) ≤ 1 + log(τ − 1) In many dimensions, specialize to one to generalize.Use inequalities and convexity to find q ∈ Archtrop near a p ∈ Amoeba; Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 16 / 18 Gnarly details Proof outline: One-dimensional case: Use basic properties to reduce to a special case; Use elementary and classical results to show that sup inf x∈Archtrop y∈Amoeba d(x, y) ≤ 1 + log(τ − 1) In many dimensions, specialize to one to generalize.Use inequalities and convexity to find q ∈ Archtrop near a p ∈ Amoeba; Use results of Passare and Rullgard to show that Amoeba is homotopic to the spine Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 16 / 18 Gnarly details Showing supx∈Archtrop inf y∈Amoeba d(x, y) ≤ 1 + log(τ − 1) : 1 if g(x) = αxβ (γ1 x1 , γ2 x2 , . . . , γn xn ), then Amoeba(g) = Amoeba(f ) − log |γ| Archtrop(g) = Archtrop(f ) − log |γ| 2 Obtain a constant bound for p(x) of the form p(x) = c0 + . . . + cd xd , |ci | ≤ 1, |c0 | = |cd | = 1; 3 Use a result of Montel to show the O(log |τ − 1|) bound for p(x) = c0 + . . . + cp xp + γ1 xn1 + . . . + γq xnq ; |ci |, ||γi | ≤ 1 4 and |c0 | = |cd | = 1; use observation 1 to generalize Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 17 / 18 Gnarly details Showing supx∈Archtrop inf y∈Amoeba d(x, y) ≤ 1 + log(τ − 1) : 1 if g(x) = αxβ (γ1 x1 , γ2 x2 , . . . , γn xn ), then Amoeba(g) = Amoeba(f ) − log |γ| Archtrop(g) = Archtrop(f ) − log |γ| 2 Obtain a constant bound for p(x) of the form p(x) = c0 + . . . + cd xd , |ci | ≤ 1, |c0 | = |cd | = 1; 3 Use a result of Montel to show the O(log |τ − 1|) bound for p(x) = c0 + . . . + cp xp + γ1 xn1 + . . . + γq xnq ; |ci |, ||γi | ≤ 1 4 and |c0 | = |cd | = 1; use observation 1 to generalize Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 17 / 18 Gnarly details Showing supx∈Archtrop inf y∈Amoeba d(x, y) ≤ 1 + log(τ − 1) : 1 if g(x) = αxβ (γ1 x1 , γ2 x2 , . . . , γn xn ), then Amoeba(g) = Amoeba(f ) − log |γ| Archtrop(g) = Archtrop(f ) − log |γ| 2 Obtain a constant bound for p(x) of the form p(x) = c0 + . . . + cd xd , |ci | ≤ 1, |c0 | = |cd | = 1; 3 Use a result of Montel to show the O(log |τ − 1|) bound for p(x) = c0 + . . . + cp xp + γ1 xn1 + . . . + γq xnq ; |ci |, ||γi | ≤ 1 4 and |c0 | = |cd | = 1; use observation 1 to generalize Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 17 / 18 Gnarly details Showing supx∈Archtrop inf y∈Amoeba d(x, y) ≤ 1 + log(τ − 1) : 1 if g(x) = αxβ (γ1 x1 , γ2 x2 , . . . , γn xn ), then Amoeba(g) = Amoeba(f ) − log |γ| Archtrop(g) = Archtrop(f ) − log |γ| 2 Obtain a constant bound for p(x) of the form p(x) = c0 + . . . + cd xd , |ci | ≤ 1, |c0 | = |cd | = 1; 3 Use a result of Montel to show the O(log |τ − 1|) bound for p(x) = c0 + . . . + cp xp + γ1 xn1 + . . . + γq xnq ; |ci |, ||γi | ≤ 1 4 and |c0 | = |cd | = 1; use observation 1 to generalize Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 17 / 18 Gnarly details Showing supx∈Archtrop inf y∈Amoeba d(x, y) ≤ 1 + log(τ − 1) : 1 if g(x) = αxβ (γ1 x1 , γ2 x2 , . . . , γn xn ), then Amoeba(g) = Amoeba(f ) − log |γ| Archtrop(g) = Archtrop(f ) − log |γ| 2 Obtain a constant bound for p(x) of the form p(x) = c0 + . . . + cd xd , |ci | ≤ 1, |c0 | = |cd | = 1; 3 Use a result of Montel to show the O(log |τ − 1|) bound for p(x) = c0 + . . . + cp xp + γ1 xn1 + . . . + γq xnq ; |ci |, ||γi | ≤ 1 4 and |c0 | = |cd | = 1; use observation 1 to generalize Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 17 / 18 Applications and further thoughts This can be used to isolate roots; Connection to triangulations and Voronoi can be explored; Details in a paper [AKNR] Metric Estimates and Membership Complexity for Archimedean Amoebae and Tropical Hypersurfaces Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 18 / 18 Applications and further thoughts This can be used to isolate roots; Connection to triangulations and Voronoi can be explored; Details in a paper [AKNR] Metric Estimates and Membership Complexity for Archimedean Amoebae and Tropical Hypersurfaces Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 18 / 18 Applications and further thoughts This can be used to isolate roots; Connection to triangulations and Voronoi can be explored; Details in a paper [AKNR] Metric Estimates and Membership Complexity for Archimedean Amoebae and Tropical Hypersurfaces Roman Kogan Amoebas, ArchTrop and all that jazz March 24, 2014 18 / 18