「どの向きから見るか」が、数学的な判定を変える。 円周を使う観察から、複素多変数の世界へ。The direction you choose can change what you see. From circles on a graph to functions of several complex variables.
01 / A CIRCLE AROUND A POINT
中心の高さを、円周の平均と比べるCompare the center with a circle average
関数 u が劣調和とは、大まかには「中心の値が、周りの円周上の値の平均を超えない」ことです。領域に閉円板が収まる、どの中心 a・半径 r についても確かめます。一つの円で成立するだけでは足りません。
Subharmonicity means that the value at a center does not exceed the average around it, for every closed disc contained in the domain. One successful circle is not enough.
The definition also requires upper semicontinuity; values of −∞ are allowed. For a smooth real-valued function, subharmonicity is equivalent to Δu = uxx + uyy ≥ 0. Convexity in every real direction is stronger.
例えば u(x+iy)=4x²−2y² は鞍型ですが、Δu=4>0 なので劣調和です。見た目の「くぼみ」と、円周平均の判定を区別してみましょう。The saddle u(x+iy)=4x²−2y² has Δu=4>0, so it is subharmonic despite bending downward in one direction.
多重劣調和は「すべての複素直線で劣調和」Subharmonic on every complex line
複素 n 変数の関数 u では、点 a と方向 v∈ℂⁿ を選び、ζ↦u(a+ζv) を見ます。ζ は複素数なので、これは実1次元の線ではなく実2次元の切り口です。上半連続な u が、どの複素直線上でも劣調和(または恒等的に −∞)なら、多重劣調和(plurisubharmonic、略して psh)と呼びます。
For u of n complex variables, choose a point a and direction v∈ℂⁿ and restrict to ζ↦u(a+ζv). Since ζ is complex, the slice has two real dimensions. An upper semicontinuous u is plurisubharmonic (psh) when every such restriction is subharmonic or identically −∞.
滑らかな場合は、複素 Hessian(Levi 形式)がすべての v に対して非負であることと同値です。z 方向と w 方向を別々に見るだけでは、斜め方向の負の曲率を見逃します。
For smooth u, the complex Hessian, or Levi form, must be nonnegative in every direction. Checking the z and w axes alone can miss negative curvature in a mixed direction.
Lu(a;v) = ∑j,k (∂²u / ∂zj∂z̄k)(a) vjv̄k ≥ 0
TRY A DIRECTION
軸では Yes、斜めでは?The axes pass. What about a diagonal?
u(z,w)=|z|²+|w|²+4 Re(z w̄) v=(cos θ, sin θ)
切り口 u(ζv)=c|ζ|² の実軸断面。紫は |ζ|=1 の円周平均、黄色は中心値0です。Real-axis profile of u(ζv)=c|ζ|². Purple marks its mean on |ζ|=1; yellow marks the center value 0.
Here c=1+2 sin(2θ). At θ=−45°, c=−1 and the circle mean is below the center: one direction disproves psh. This slider explores only a subset of complex directions.
The context cards after each answer link here. A construction's properties do not automatically carry over after changing signs or adding terms.
Hartogs 領域と −logdomains and −log
正で連続な u に対して、底面の各点 z に半径 u(z) の複素円板を置きます。底面が単位円板のとき、この Hartogs 領域 {|w|<u(z)} の擬凸性は −log u の劣調和性と同値。1枚のグラフが、複素2次元の領域を符号化します。
Over the unit disc, positive continuous u defines the Hartogs domain {|w|<u(z)}. Its pseudoconvexity is equivalent to subharmonicity of −log u. A graph of radii encodes a domain in ℂ².
For |x|, second derivatives vanish away from x=0, yet positive curvature is concentrated on the crease. Distributional derivatives capture it as a measure.
The Dirichlet problem prescribes boundary temperatures and asks for the steady interior temperature. Poisson integration extends the data by weighted averaging. Steep boundary features coexist with interior harmonicity.
次数 d≥2 の多項式 P の逃避率は gP(z)=lim d−n log⁺|P∘n(z)|(log⁺t=max{0,log t})。無限遠へ逃げない点の集合で0となり、その境界が Julia 集合です。有限反復の描画は近似であり、拡大して見える細部の精度には限界があります。
For a degree d≥2 polynomial, gP(z)=lim d−n log⁺|P∘n(z)|, with log⁺t=max{0,log t}, measures escape. It vanishes on the filled Julia set; its boundary is the Julia set. Finite iteration limits the precision of the image.
In Σ0.55k(z/1.03)2k, rapidly separated exponents produce a natural boundary. The series is holomorphic inside its circle of convergence but cannot be continued through it. Its real part is still harmonic inside.
Logarithmic potentials model point masses. The potentials −log(1−|z|²/R²) and log(Σ|fⱼ|²) connect to hyperbolic and projective geometry. Their complex Hessians encode geometric coefficients.
The Bergman kernel probes a domain through square-integrable holomorphic functions. Berndtsson's theorem gives plurisubharmonic variation of its logarithm for pseudoconvex families. Nonholomorphic evaluation paths need not preserve that property.